A Network of HMG-box Transcription Factors Regulates Sexual Cycle in the Fungus Podospora anserina

High-mobility group (HMG) B proteins are eukaryotic DNA-binding proteins characterized by the HMG-box functional motif. These transcription factors play a pivotal role in global genomic functions and in the control of genes involved in specific developmental or metabolic pathways. The filamentous ascomycete Podospora anserina contains 12 HMG-box genes. Of these, four have been previously characterized; three are mating-type genes that control fertilization and development of the fruit-body, whereas the last one encodes a factor involved in mitochondrial DNA stability. Systematic deletion analysis of the eight remaining uncharacterized HMG-box genes indicated that none were essential for viability, but that seven were involved in the sexual cycle. Two HMG-box genes display striking features. PaHMG5, an ortholog of SpSte11 from Schizosaccharomyces pombe, is a pivotal activator of mating-type genes in P. anserina, whereas PaHMG9 is a repressor of several phenomena specific to the stationary phase, most notably hyphal anastomoses. Transcriptional analyses of HMG-box genes in HMG-box deletion strains indicated that PaHMG5 is at the hub of a network of several HMG-box factors that regulate mating-type genes and mating-type target genes. Genetic analyses revealed that this network also controls fertility genes that are not regulated by mating-type transcription factors. This study points to the critical role of HMG-box members in sexual reproduction in fungi, as 11 out of 12 members were involved in the sexual cycle in P. anserina. PaHMG5 and SpSte11 are conserved transcriptional regulators of mating-type genes, although P. anserina and S. pombe diverged 550 million years ago. Two HMG-box genes, SOX9 and its upstream regulator SRY, also play an important role in sex determination in mammals. The P. anserina and S. pombe mating-type genes and their upstream regulatory factor form a module of HMG-box genes analogous to the SRY/SOX9 module, revealing a commonality of sex regulation in animals and fungi.


Introduction
High-mobility-group box (HMGB) proteins [1] include chromatin architectural proteins as well as specific transcription factors that are involved in highly diverse functions ranging from sex determination [2] to extracellular immune signaling [3,4]. All of these functions rely on the HMG-box, an HMGB conserved motif containing approximately 80 amino acids arranged in a distinctive L-shaped three-a-helical fold [5,6]. This HMG-box motif sharply bends DNA [7] and facilitates the assembly of transcriptional complexes that involve other proteins by distorting chromatin [8]. Based on phylogenetic analyses of the HMG-box, the HMGB superfamily can be divided into two families, HMGB-UBF_HMG and SOX/TCF/MATA_HMG [9]. These families were named after the best known representative in each group, namely hUBF [10], the SOX (Sry-type HMG-box) genes [11], TCF-1 [12] and MATa-1 [13]. The HMGB-UBF_HMG family is considered to be sequence non-specific, as generic UBF proteins can bind both ribosomal DNA regulatory sequences and sequences across the entire ribosomal DNA repeat [14]. Members of this family are present in plants, fungi and animals. On the other hand, the SOX/TCF/MATA_HMG family contains proteins that bind specific DNA sequences, with a common T/A rich core (reviewed in [2]). The SOX/TCF/MATA_HMG family is subdivided into the SOX-TCF_HMG and the MATA_HMG subfamilies. The MATA_HMG subfamily includes exclusively fungal proteins, most of which are involved in sexual processes, while SOX-TCF_HMG genes appear to be restricted to animals [9].
Members of the HMGB-UBF_HMG family function in many processes such as transcription, genomic stability and the three R's (replication, recombination and DNA repair) (reviewed in [8]). An exhaustive analysis of HMGB-UBF_HMG_box encoding genes was performed in Saccharomyces cerevisiae in a previous study, and it showed HMGB-UBF_HMG-box genes to have highly diversified functions. In brief, the six HMGB-UBF_HMG_box genes from S. cerevisiae are involved in ribosomal DNA transcription [15],

Identification of HMG-box genes in P. anserina
The Podospora protein database (Materials and Methods) was searched for HMGB proteins using the previously defined HMGbox consensus sequence [27] as query in a Blastp search [48]. P. anserina genome annotation in Fungal Transcription Factor [49] and Superfamily databases [50] was examined to identify possible missing HMGB genes in our analyses. A total of 12 HMG-box genes were identified. Of these, eight had not been previously characterized and were here named PaHMG2 to PaHMG9 (Table 1). CD-Search [51] identified five MATA_HMG-box proteins and five HMGB-UBF_HMG-box proteins (Table 1), but failed to categorize two HMG-box proteins. These two proteins, PaHMG7 and PaHMG9, contained an atypical charged residue instead of a conserved aliphatic or aromatic amino acid at position 9 ( Figure 1). However, they contained the specific aromatic amino acids that anchor the recognition helix of the HMG-box domain to the hydrophobic core at position 8 and 11, confirming that they belong to the HMGB superfamily. Additional domains are shown in Table 1. Except for mtHMG1, which contained a mitochondrial targeting signal, the remaining 11 HMG-box proteins from P. anserina were predicted to localize to the nucleus.
A search for orthologs of P. anserina HMG-box genes with FUNGIpath [52] in selected fungal species indicated that three had an ortholog in S. cerevisiae, six had an ortholog in S. pombe, and all had an ortholog in Neurospora crassa (Table 1). PaHMG6, PaHMG7 and PaHMG8 had orthologs in Basidiomycota but the functions of these orthologs have not been investigated yet. Prf1, Rop1 and HMG3 from U. maydis [37,38], Mat2 from C. neoformans [42] and Pcc1 from C. cinerea [40] encoded HMG-box regulators that are related to the sexual cycle in Basidiomycota. FUNGIpath did not detect any P. anserina orthologs for these HMG-box genes.

Author Summary
Podospora anserina, a coprophilous fungus, is used extensively as a model organism to address questions of sexual development and mating-type functions. Its mating-type locus contains three HMGB genes that encode transcription factors involved in fertilization and fruit-body development. We present the functional characterization of the remaining HMGB genes, which revealed that 11 of 12 HMGB genes were involved in sexual development. An analysis of the relationships between these genes uncovered a regulatory network governing the expression of mating-type genes. PaHMG5 is a key transcription factor that operates upstream of mating-type genes in this network. A homolog of PaHMG5 performs a similar function in the fission yeast Schizosaccharomyces pombe, which diverged from P. anserina 550 million years ago. The conservation of a regulatory circuit over such a prolonged timeframe is a striking exception to the general observation that sex developmental pathways are highly variable, even across closely related lineages. A module consisting of two HMGB transcription factors (Sry and Sox9) is a key regulator of sex determination in mammals. We propose that the module containing PaHMG5 and mating-type HMGB genes is the fungal counterpart of the mammalian module, revealing a commonality of sex regulation in animals and fungi.
Most HMG-box genes localized to chromosome 1, which contains the mating-type locus [53]. Only three out of 12 HMGB genes mapped outside of chromosome 1, suggesting that the distribution of HMG-box genes may be biased to chromosome 1. However, statistical tests indicated this bias to be inconclusive (Material and Methods).

Evolutionary analysis of fungal HMG-box proteins
The HMG-box of the 12 P. anserina HMGB proteins was extracted and grouped in a phylogram with the HMG-box motif of selected plant, animal and fungal species, including S. cerevisiae, S. pombe, N. crassa, Aspergillus nidulans, Cochliobolus heterostrophus and U. maydis ( Figure 2). A total of 154 HMG-box domains were clustered in four groups, which overall corresponded to the previously defined HMGB groups [9]: MATa_HMG (groupe A, Figure 2) [27], MATA_HMG (group B, Figure 2), SOX-TCF_HMG (group C, Figure 2) and HMGB-UBF_HMG (group D, Figure 2). Group A ( Figure 2) included exclusively the mating-type transcription factors with an a1 domain, which was proposed to correspond to a new class of HMG-box [27,54]. Group A formed a clade related to the MATA_HMG subfamily and contained FMR1 (Podan1). The other HMG-box mating-type transcription factors FPR1 (Podan2a) and SMR2 (Podan3) were placed in group B (Figure 2), which contained mostly MATA_HMG-box proteins. PaHMG5 (Podan2c) clustered within clade G, with SpSte11 (Schpo2c) of S. pombe [30] and NCU09387/FMF-1 (Neucr2c) of N. crassa [55,56]. The orthology of these three proteins, suggested by their phylogenic position, was independently confirmed with FUNGIpath. Moreover, the relationship of NCU09387 (Neucr2c) with SpSte11 (Schpo2c) was previously reported [57]. PaHMG8 (Podan2b) was the only MATA_HMG-box protein encoded by a gene outside chromosome 1. Its ortholog in S. cerevisiae is ROX1p (Sacce2), a repressor of hypoxic genes, and one of the few MATA_HMG-box proteins not related to mating processes [29]. Earlier evolutionary analysis of the oxygen-responding system in Kluyveromyces lactis and S. cerevisiae suggested that ROX1 was recruited specifically to control this system in S. cerevisiae [58]. Rfg1p (Canal2), the ortholog of ROX1p (Sacce2) in Candida albicans, controls filamentous growth and virulence [59]. Interestingly, the clade defined by PaHMG8 (Podan2b), ROX1p (Sacce2) and Rfg1p (Canal2) included a MAT1-2-1 mating-type protein (Ajeca2, see '&' in clade B, Figure 2) and was placed close to the MAT1-2-1 (Podan2a, Aspni2a, Neucr2a and Sorma2) and MAT1-1-3 mating-type proteins (clade E, Figure 2). The placement of the PaHMG8 (Podan2b)/ROX1p (Sacce2) clade in the MATA_HMG subfamily and our functional analyses of PaHMG8 (Podan2b) (see below) supports the idea that the  Figure 1. Alignment of HMG domains from the 12 HMG-box proteins of P. anserina. The alignment was performed using ClustalW2 [105] and colored according to the Clustal X color scheme provided by Jalview [106]. This color scheme is displayed in Table S1. doi:10.1371/journal.pgen.1003642.g001 LG+G [107]. ancestral ROX1 gene was primarily involved in sexual development before being recruited for various other functions in S. cerevisiae and C. albicans. Surprisingly, group B contained an HMGB-UBF_HMG clade (clade H, Figure 2), which included PaHMG3 (Podan5e) and several fungal orthologs, notably the histone demethylase SpLsd1 (Schpo7a) of S. pombe [21]. All proteins from clade H have SWIRM and amino oxydase domains, which are characteristic of histone demethylases, thus supporting their orthologous relationship. Five P. anserina HMG-box proteins were classified into group D ( Figure 2), which comprised members of the HMGB-UBF_HMG family, except the clade grouping PaHMG7 (Podan6b), NCU06874 (Neucr6b) from N. crassa and NP_595970 (Schpo7b) from S. pombe. These three proteins are characterized by a conserved atypical residue in the HMG-box domain [see the above section for its description in PaHMG7 (Podan6b)] and a SprT-like domain (Interpro accession number IPR006640). The four HMGB-UBF_HMG-box proteins, PaHMG2 (Podan5c), PaHMG4 (Podan5b), mtHMG1 (Podan5d), and PaHMG6 (Podan5a), branched into different clades of group D. PaHMG2 (Podan5c) belonged to a branch that included proteins with a Sterile Alpha Domain (SAM), which is involved in protein-protein and protein-RNA interactions [60]. In some cases, the orthologous relationships, as defined by FUNGIpath, were not in agreement with the HMG-box based phylogeny. The first instance was the mitochondrial protein mtHMG1 (Podan5d) and its ortholog NCU02695 (Neucr5d) in N. crassa, which failed to group into the same clade within the HMGB-UBF_HMG family although FUNGIpath provided a high confidence score for the orthology of mtHMG1 (Podan5d) and NCU02695 (Neucr5d) HMG-box domain. Moreover, mtHMG1 (Podan5d) and NCU02695 (Neucr5d) have a mitochondrial targeting signal and are characterized by a DUF1898 domain, supporting an orthologous relationship. In another instance, Nhp6Ap (Sacce5a) and Nhp6Bp (Sacce5b), two functionally redundant putative inparalogs from S. cerevisiae collectively referred to as Nhp6p [61,62], were placed in distant clades with PaHMG6 (Podan5a) and PaHMG4 (Podan5b), respectively, and may be outparalogs instead. In contrast, FUNGIpath identified Nhp6p (Sacce5a, b) as co-orthologs of PaHMG6 (Podan5a). Similarly, Hmo1p (Sacce5d) from S. cerevisiae belonged to the same clade as PaHMG4 (Podan5b), while FUNGIpath identified Hmo1p (Sacce5d) as an ortholog of PaHMG6 (Podan5a). However, the FUNGIpath confidence score for orthology of PaHMG6 (Podan5a), Nhp6p (Sacce5a, b), and Hmo1p (Sacce5d) was low. Further analyses will be necessary to resolve these phylogenetic ambiguities.
PaHMG9 (Podan6a), along with its N. crassa ortholog NCU07568 (Neucr6a), were placed between group C and D ( Figure 2). Accordingly, CD-Search failed to place PaHMG9 (Podan6a) and NCU07568 (Neucr6a) into either the MA-TA_HMG or HMGB-UBF groups (Table 1). Interestingly, the PaHMG9 HMG-box (Podan6a) is characterized by an atypical residue (see above section) that is conserved in NCU07568 (Neucr6a). Our phylogenic analysis also placed a MATA_HMGbox protein from N. crassa (NCU02326, Neucr2d) between group C and D ( Figure 2). Previous work reported that NCU02326 (Neucr2d) is related to SpeSte11 (Schpo2c) and is the closest homolog to NCU09387 (Neucr2c) [57]. In agreement with this report, FUNGIpath identified NCU02326 (Neucr2d) as an ortholog of SpSte11 (Schpo2c, clade G in Figure 2) and an inparalog of NCU09387 (Neucr2c). However, unlike NCU09387 (Neucr2c), NCU02326 (Neucr2d) was not placed within clade G with all other SpSte11 orthologs. Such inconsistency is not unprecedented, as there are several evidences in N. crassa that effective defense against duplicated sequences prevents the maintenance of closely related paralogs in this species [63]. To independently assess which of the two N. crassa inparalogs is conserved in P. anserina, we compared the environment of these genes in the two species. The shared synteny observed upstream and downstream of PaHMG5 (Podan2c) and NCU09387 (Neucr2c) ( Figure S1A) confirms that these two genes are orthologs. The absence of conserved organization between the NCU02326 (Neucr2d) locus and its putative counterpart in P. anserina indicates that this gene is absent in P. anserina ( Figure S1B).

Phenotypic analysis of HMG-box deletion strains during vegetative growth
The eight additional HMG-box genes identified in this study were inactivated by targeted gene deletion to assess their role in the life cycle of P. anserina. Deletions were verified by Southern blot analysis (Materials and Methods). The eight mutants were examined for growth and for macroscopic mycelium alterations on minimal medium in mat + and mat2 context. The DPahmg2, DPahmg6 and DPahmg9 strains displayed reduced growth ( Table 2).
The eight mutants were also tested for cold-and thermo-sensitive growth at 18uC and 36uC and for sensitivity to caffeine (phosphodiesterase inhibitor), fludioxonil, which acts on osmoregulation, and sodium dodecylsulfate, which acts on the cell membrane, without revealing any additive phenotype.
Phenotypic analysis of HMG-box deletion strains during sexual development P. anserina is a heterothallic (self-sterile) fungus that has two mating types, mat+ and mat2. Haploid strains of each mating type initiate the sexual cycle by differentiating male gametes (spermatia) and female organs (protoperithecia). A phenotypic analysis of sexual development separately evaluates the male and female fertility of each mutant. The female organ is a multicellular structure comprising protective maternal hyphae and the ascogonium, which contains the female gametic nuclei. The spermatia are independent cells that can be collected and used to fertilize any strain of opposite mating type. A pheromone/receptor signaling system allows the ascogonium to recognize and fuse with sexually compatible spermatia. Therefore, fertilization can be controlled to initiate the development of the fruit-body. The male gametic nucleus is delivered into the ascogonium, which differentiates into the hymenium. Karyogamy takes place in the hymenium and is immediately followed by meiosis. Subsequently, the haploid nuclei are packed into ascospores. Ultimately, the mature ascospores are forcibly discharged from fruit-bodies (perithecia).
To determine whether HMG-box gene deletions impair sexual reproduction, each mutant strain was tested for male and female fertility in both mating types (mat+ and mat2). The test for male fertility consisted in fertilizing wild-type female organs with spermatia from each HMG-box deletion mutant ( Figure 4A). This analysis indicated that all mutants, except DPahmg5, produced functional spermatia. Moreover, fertilized female organs differentiated into mature fruit-bodies producing ascospores, indicating that HMG-box deletions in the male nuclei did not affect any developmental steps in the fertilized female organs. For each mutant, reciprocal crosses with mat2 and mat+ wild-type strains behaved similarly, indicating that the observed phenotype was not dependent on mating type. We further quantified spermatia produced by each deletion mutant and measured their activity in fertilization assays (Materials and Methods). Two mutants, DPahmg4 and Dkef1 produced five and seven times more spermatia, respectively, than the wild-type strain. Spermatia from these two mutants and from other mutants displayed the same fertilization ability as wild-type spermatia. The overproduction of spermatia observed in Dkef1 was in agreement with microscopic observations, illustrating that even young hyphae produce spermatia. The DPahmg5 mat+ and DPahmg5 mat2 strains were found to produce as many spermatia as wild-type strains. Thus, male sterility in these DPahmg5 strains can be attributed either to an inability of spermatia to fertilize wild-type protoperithecia or to an arrest of perithecial development shortly after fertilization.
To test the female fertility, we examined the formation of fruitbodies by mycelia from each mutant fertilized with wild-type spermatia ( Figure 4A). No fruit-bodies were produced on mutant mycelium in crosses involving DPahmg5, DPahmg6, DPahmg8, Dkef1 and Dmthmg1 strains, demonstrating that these mutants were sterile as female partners. In contrast, DPahmg2, DPahmg3, DPahmg4 and DPahmg7 strains produced fruit-bodies that developed normally and ejected as many ascospores as fruit-bodies from a wild-type cross. For each mutant, reciprocal crosses with mat2 and mat+ wild-type strains behaved similarly, indicating that the observed phenotype was not dependent on mating type.
We also examined the distribution of perithecia on the mycelium. These assays were based on homozygous crosses and were performed by inoculating a mixture of fragmented mat+ and mat2 mycelia in the center of a plate containing minimum medium under constant light illumination ( Figure 4B). In this test, a wild-type strain differentiates fruit-bodies mostly in a ring-like area, which was 1 cm wide and located approximately 1 cm away from the inoculation point [70]. Only DPahmg2 and DPahmg7 differentiated fruit-bodies with the same pattern as the wild-type strain. The alterations were striking for DPahmg4, since fruit-bodies were distributed on the entire surface of the culture and the ringlike area was no longer visible. The pattern was also different for DPahmg3, which developed a wider ring-like fruit-body area. Furthermore, we observed that this phenotype was exacerbated by  the deletion of PaHMG7. This difference was confirmed by quantitative measurement of the ring-like diameter in the DPahmg3 DPahmg7 double mutant ( Table 2). The increase in ring diameter (54.9 mm instead of 53.7 mm) observed in the double mutant suggested that the PaHMG3 and PaHMG7 genes together control the distribution of fruit-bodies. Fruit-bodies were never observed for DPahmg5, DPahmg8 and Dkef1. The DPahmg6 mutant did not produce fruit-bodies as fast as the wild-type strains, but extended incubation revealed that it was weakly female fertile. Fruit-bodies were 50 times less abundant than in a wild-type cross and they displayed an altered shape, with a smaller body and a larger neck than the wild-type fruit-bodies (data not shown). They began to eject ascospores 8 days after fertilization, whereas completion of the sexual process took 4 days in a wild-type cross. Genetic analysis of their progeny confirmed that they were produced by DPahmg6 parents and not by contaminating mycelium. In the wild type, anastomoses were never observed in apical hyphae (1), while they were occasionally observed in the subapical area (2). In Dkef1 strain, anastomoses were profuse in apical (1) and in subapical (2) areas. Most notably, anastomosis occurred between the apex of hyphae at the leading edge and neighbouring apical hyphae (see arrows in Dkef1 in 1). (C) Schematic of appressorium-like development in P. anserina. a: appressorium-like structure including the palm-like structure and the needle-like hyphae; c: cellophane layer on which mycelium is growing. (D) Spermatia and their spermatogonia were present in the subapical area of the Dkef1 strain but were absent from the same area in the wild-type strain. Scale bar = 10 mm in all panels. doi:10.1371/journal.pgen.1003642.g003 In conclusion, all the eight HMG-box deletion strains analyzed in this study are altered phenotypically. Complementation assays confirmed that the deletion was responsible for the observed phenotype (Materials and Methods). None of the eight HMG-box genes analyzed in this study were essential for viability, although the deletion of three genes (PaHMG2, PaHMG6 and KEF1) resulted in growth alterations. Most HMG-box gene deletions affected sexual reproduction (Table 2). PaHMG2 is the only P. anserina HMG-box gene whose deletion exclusively affected vegetative growth. Deletions of PaHMG3, PaHMG4 and PaHMG7 affected the distribution of fruit-bodies. More importantly, these deletions failed to affect vegetative growth and fruit-body development. Deletion of PaHMG6, PaHMG8 and KEF1 impaired female fertility, but mutant strains remained male fertile. Strikingly, the DPahmg5 mutant exhibits both male and female sterility, an uncommon phenotype in P. anserina that was only reported for mating-type mutants [26].

Heterokaryotic complementation of female sterility caused by HMG-box gene deletions
Microscopic observation of DPahmg5, DPahmg6, DPahmg8 and Dkef1 cultures revealed the presence of protoperithecia, indicating that female sterility was not due to the absence of female reproductive structures. Female sterility of DPahmg5, DPahmg6, DPahmg8 and Dkef1 can be attributed to a defect either in the development of the fruit-body envelope or in the formation of the hymenium, or both. To determine which tissue requires the HMG-box proteins, we performed trikaryon mosaics. In this experiment, mat+ and mat2 HMG-box mutant strains were cocultured with a strain containing a deletion of the mat+ idiomorph Analysis of male and female fertility of HMG-box gene deletion mutants in crosses with wild-type tester strains. When cultures were confluent, sterile water was poured and dispersed over the surface of the mycelium. As each strain produced spermatia (male cells) and protoperithecia (female organs) regardless of its mating type, reciprocal fertilization of the mutant and wild-type strains took place and indicated whether the mutant was fertile as a male (donor) or as a female (receptor). For DPahmg5 some perithecia differentiated at the contact zone where mutant and wild-type mycelia fuse. In the resulting heterokaryotic mycelium, wild-type nuclei complement the male and/or female sterility defect of the mutant. Those perithecia were fertile and expelled numerous asci allowing genetic analysis. (B) Analysis of perithecium distribution in homozygous crosses of HMG-box gene deletion mutants. Fragmented mycelium from mat+ and mat2 strains with the same HMG-box deletion were deposited at the center of a Petri dish and incubated until perithecia formed. Typically, the wild-type strains differentiated perithecia within a ring-like area. doi:10.1371/journal.pgen.1003642.g004 (Dmat) [71]. Except for the loss of the mat+ mating-type gene, the Dmat strain had a wild-type genotype. Because the Dmat strain lacked the mating-type gene needed for fertilization, it could not participate in sexual reproduction and, thus, could not rescue hymenium defects. However, the Dmat strain maintained wild-type vegetative characteristics and could act as a helper strain by complementing maternal defects in a contest of sexually compatible mutant strains. Notably, the Dmat strain can provide maternal haploid tissues to form the perithecial envelope [72].
Fertile fruit-bodies were observed in the three trikaryotic cultures of DPahmg6, DPahmg8 and Dkef1 (data not shown), indicating that Dmat cells surrounding the hymenia of these HMG-box mutants promote the development of functional perithecia. These results demonstrated that PaHMG6, PaHMG8 and KEF1 are required for making the perithecial envelope and that they are dispensable for the hymenium. We also examined the behaviour of the Dmthmg1 strain in a trikaryon mosaic test. As previously described [44], most Dmthmg1 cultures died quickly and did not differentiate protoperithecia. A trikaryotic culture involving sexually compatible Dmthmg1 strains and Dmat strain produced perithecia, but no ascospores were formed (data not shown). These results indicate that the Dmat tissue complemented the vegetative requirement for mtHMG1 during protoperithecium formation, while the absence of mtHMG1 in the hymenium led to an arrest of its development. The protein mtHMG1 is therefore required for both protoperithecial and hymenium development.
Trikaryotic cultures of DPahmg5 did not produce any perithecia. It is noteworthy that mating-type gene mutations do not affect PaHMG5 transcription [73], thus excluding the possibility that the Dmat strain is unable to complement the DPahmg5 defects because it is itself affected for PaHMG5 expression. Failure to restore fertility of DPahmg5 in the trikaryotic test could reflect the need for this gene in both male and female fertility, which can be difficult to be restored simultaneously. Therefore, we designed an assay to separately evaluate the restoration of female and male fertility of DPahmg5. To evaluate whether the female sterility of DPahmg5 could be rescued by a Dmat PaHMG5 + strain, co-cultures of mat+ DPahmg5 and Dmat PaHMG5 + strains were used as female partners in crosses fertilized by mat2 wild-type spermatia. Conversely, to evaluate whether Dmat PaHMG5 + could restore the male competency of DPahmg5, cultures of the mat2 wild-type strain were fertilized with spermatia issued from a co-culture of mat+ DPahmg5 and Dmat PaHMG5 + strains (Materials and Methods). Two fruit-bodies were observed on 10 Petri dishes for the female sterility restoration assay, while three fruit-bodies were observed on 10 mat2 plates for the male restoration assay. Several thousand perithecia were formed with wild-type strains in similar experiments, indicating that male and female sterility of DPahmg5 strain was inefficiently complemented by the Dmat PaHMG5 + strain. Taken together, these data are consistent with a requirement of PaHMG5 for fertilization or an early stage of hymenium development. The arrest in the development of the hymenium at an early stage precludes any conclusion on the role of PaHMG5 in the development of the maternal perithecial envelope.

Defects caused by overexpression of PaHMG5
In the complementation assay of the DPahmg5 mutant, introduction of the PaHMG5 wild-type allele restored male fertility but transformants remained female sterile (Materials and Methods). Moreover, the transformants displaying the highest efficiency as male partners were vegetatively altered, displaying a flat vegetative mycelium without aerial hyphae. These data suggested that deregulation of PaHMG5 may be detrimental to the fungus. To examine the consequences of unregulated PaHMG5 expression, a plasmid was constructed to express a fusion of PaHMG5 with the Pagpd (glyceraldehyde-3 phosphate-dehydrogenase) promoter and initiation codon [74] (Materials and Methods). When this fusion construct was introduced into protoplasts from the wild-type strain, most of the recovered transformants displayed a flat mycelium. In addition, they exhibited female sterility in crosses with the wild-type strain. Genetic analysis (Materials and Methods) demonstrated that the vegetative defect (flat mycelium) and female sterility consistently co-segregated with the Pagpd::PaHMG5 transgene. These observations strongly suggested that female fertility relies on tight regulation of PaHMG5 expression.
Transcriptional expression of mating-related HMG-box genes and downstream target genes PaHMG5, PaHMG6, PaHMG8, KEF1 and mtHMG1 [44] are involved in the development of male and female organs, raising the question of their genetic interactions. To assess the relationships between these genes, quantitative real-time RT-PCR (RT-qPCR) was used to examine their expression patterns in DPahmg5 (Table S3), DPahmg6 (Table S4), DPahmg8 (Table S5), Dkef1 (Table  S6) and Dmthmg1 (Table S7) strains in a mat+ and mat2 context. Genes were defined as up-regulated in the mutant strain if the fold change (FC) was .1, with a p-value of ,0.05 (see Materials and Methods for FC computation). On the other hand, genes were defined as downregulated in the mutant strain if 0,FC,1 with a p-value of ,0.05. FCs with a 95% confidence interval including the value of 1 were not considered significant [75]. The results are summarized in Figure 5. Most of the deletions had similar effects in mat+ and mat2 strains. The few exceptions were statistically non-significant results in mat+ or mat2 strains, e.g., mtHMG1 had a significant FC (1.35) in mat+ Dkef1 relative to wild-type and a nonsignificant FC (1.1) in mat2 Dkef1 (Table S6). A map of the genetic interactions among PaHMG5, PaHMG6, PaHMG8, KEF1 and mtHMG1 was constructed based on the assumption that deletion of a regulatory gene can affect downstream target genes, while nonaffected genes are either upstream regulatory genes or genes acting in an independent pathway ( Figure 6). Interestingly, PaHMG6, PaHMG8, KEF1 and mtHMG1 appeared to interact and converge on a key regulator, PaHMG5. We detected an effect of mtHMG1 on the transcription of KEF1 and downstream HMG-box genes, suggesting that this protein is localized to the nucleus although it has been shown to be targeted to the mitochondria [44]. Reexamination of its sequence using PSORTII [76] revealed the presence of several monopartite and bipartite nuclear localization signals.
We further examined the transcription levels of mating-type genes and of a selection of specific target genes [73], including the pheromone and pheromone receptor genes, in the different mutant strains (Table S3 to S7 and Figure 5). We found that the transcription of the mating-type genes FMR1 and FPR1 was reduced in DPahmg5 and DPahmg8 strains. As expected from the cascade shown in Figure 6, FMR1 was downregulated in mat2 DPahmg6; in contrast, we failed to identify a significant reduction in FPR1 transcription in mat+ DPahmg6. However, all tested FPR1 target genes were either up-or downregulated in this latter strain, suggesting that PaHMG6 by-passes FPR1 to control its target genes. Overall, transcription of the mating-type target genes increased in Dkef1 strains, in agreement with the repressor effect of KEF1 on mating-type genes. Interestingly, transcription of pheromone receptor genes decreased in female sterile HMG-box mutants, even in the Dkef1 strain. This is in agreement with their essential role in female fertility [73].
PaHMG5 appears to be the major regulator of FPR1 and FMR1 mating-type genes in the network presented in Figure 6. This raises the possibility that constitutive expression of FPR1 and FMR1 could compensate for the absence of PaHMG5 and, thus, rescue sterility of the DPahmg5 mutant. Transgenic versions of FMR1 and FPR1 driven by the Angpd constitutive promoter were previously reported to complement loss-of-function of the corresponding gene [77]. Functional Angpd-FMR1 or Angpd-FPR1 transgenes were introduced by genetic crosses in mat2 and mat+ DPahmg5 mutant strains, respectively. We observed that DPahmg5 strains carrying these transgenes remained male and female sterile, indicating that PaHMG5 function is not limited to the transcriptional activation of mating-type genes. It may regulate mating-type target genes as a cofactor of mating-type transcription factors. Alternatively, it may regulate fertility genes that are different from mating-type target genes.

Promoter region analysis of HMG-box genes and target genes of HMG-box transcription factors reveals HMG-box binding sites
The regulatory network shown in Figure 6 consists of a cascade of HMG-box genes, suggesting that each gene may contain a binding site for the upstream regulating HMG-box factor. A search for a conserved binding site using MEME [78] identified a consensus motif (A/G)ACAAAGAA in KEF1, mtHMG1, PaHMG5, PaHMG8, and the FMR1 and FPR1 mating-type genes ( Figure 7A). This consensus motif is very similar to the common core DNA motif A(A/T)CAA(A/T)G that is recognized by HMG-box transcription factors [79] (reviewed in [2]). The remaining P. anserina HMG-box genes either contained a sequence that displayed some differences to the A(A/T)CAA(A/T)G core sequence (PaHMG2, PaHMG3, PaHMG4 and PaHMG6) ( Figure 7A), or they did not contain any related sequence (PaHMG7 and SMR2). Further analyses of mating-type target genes using MEME revealed that the mat+ pheromone receptor gene (PRE1), alternative oxydase gene (AOX), phospho-enol pyruvate kinase gene (PEPCK), Pa_1_24410 and Pa_6_7350 also contained the core HMG-box binding site ACAAAGA ( Figure 7A). Interestingly, the two pheromone genes (MFM and MFP) displayed the same conserved core sequence, ATCAAAG. The mat2 pheromone receptor (PRE2), Pa_4_80, Pa_4_3858 and Pa_5_9770 did not contain the core HMG-box binding site, suggesting that these genes are secondary targets of HMG-box genes. A total of eight genes contained the (A/ G)ACAAAGAA consensus site. The comparison with the distribution of this site in the P. anserina genome indicated that the consensus site is significantly enriched in the selected set of genes examined here (pvalue = 0.016) (Materials and Methods).

Electrophoretic mobility shift assays
We further tested the binding of the heterologously expressed His-tagged version of the entire PaHMG5 protein (HMG5His) in  Table S3 to S7 were converted to a heat map using Matrix2png [108].  . Genetic network of HMG-box genes that regulate mating in P. anserina. Arrows with heads and blunt ends indicate activation and repression, respectively. The numbers next to the arrows indicate the average FC in gene expression between the wild-type and mutant strains.The relationship between mtHMG1 and PaHMG5 may be mediated by KEF1. Alternatively, mtHMG1 may by-pass KEF1 to repress PaHMG5 directly or indirectly. The consistency in the FCs suggest that PaHMG6 by-passes this cascade to activate PaHMG5 and PaHMG8 either directly or indirectly. The numbers next to the arrows connecting the mating-type genes (FMR1 and FPR1) and the downstream target genes are FCs that were obtained from [73]. doi:10.1371/journal.pgen.1003642.g006 electrophoretic mobility shift assays (EMSAs) with primers containing the putative binding sites identified above. We first examined the binding of HMG5His to its putative target genes (FMR1, FPR1 and PaHMG8, Figure 6), including PaHMG5. These genes contained the consensus site (A/G)ACAAAGAA. Under the binding conditions used, HMG5His strongly bound to primers containing these sequences and showed decreased binding upon addition of unlabelled competitor ( Figure 8A). Moreover, a primer with a scrambled FMR1 sequence did not display any binding of HMG5His, confirming the specificity of the sequence recognition. These experiments support the idea that PaHMG5 directly regulates these target genes as well as its own transcription, as previously demonstrated for SpSte11 [80]. We further examined the affinity of HMG5His for the putative binding sites of the mating-type target genes ( Figure S3), which displayed binding sites different from the consensus. Binding of HMG5His to sites that were different from the consensus (A/G)ACAAAGAA was much weaker than to the consensus binding site (see MFM, MFP AOX and PEPCK in Figure S3). The binding to MFM, MFP and PRE1 sites was more carefully investigated (Figure 8, B to D). Reciprocal competition of MFM, MFP and PRE1 sites with the FMR1 consensus site confirmed that HMG5His had a greater affinity for the consensus sequence than for the sites of these mating-type target genes. These results indicate that PaHMG5 recognized the sites in the 59UTR of mating-type target genes, but may require the mating-type transcription factors to increase the efficiency of binding. We also evaluated the affinity of HMG5His for the putative binding sites of the HMG-box genes that are not regulated by PaHMG5 in our proposed network ( Figure 6). HMG5His bound strongly to KEF1 and mtHMG1 sites, which are identical to the consensus sequence, while binding to PaHMG2, PaHMG3, PaHMG4 and PaHMG6 sites was much less efficient ( Figure S4). Strikingly, the less efficient binding occurred with sites that were different from the consensus by modification of the central part of the site (see PaHMG2, PahMG3 and PaHMG6 in Figure S4). Reciprocal competition confirmed that HMG5His has much less affinity for the PaHMG2, PaHMG3, PaHMG4 and PaHMG6 sites than for the consensus site ( Figure S5). We assume that the binding of HMG5His to these sites corresponded to the recognition of an HMG-box binding site by an HMG-box protein, but this did not demonstrate in vivo regulation by PaHMG5.
The Weblogo ( Figure 7B) showed a consensus binding site obtained from the entire set of analyzed sequences. This consensus corresponded to the sequence that was recognized with the best efficiency by HMG5His in EMSA. Comparison of this consensus with other HMG-box binding sites ( Figure 7C) reveals a striking similarity with the TR-box, which is bound by SpSte11 of S. pombe [30,81]. The P. anserina consensus site matched the TR-box in nine consecutive basepairs. The similarity of the HMG5His consensus binding site with the PRE-boxes from U. maydis and C. neoformans was reduced. In particular, the PRE-box of U. maydis had an A instead of a G in position 7 in the P. anserina consensus. This transition correlates with a weak binding of HMG5His (see PaHMG2 and PaHMG6 in Figure S4). Taken together, these data indicate that PaHMG5 binds most efficiently to sites that are almost identical to those that are recognized by SpSte11, its ortholog in S. pombe.
An analysis of the distribution of the (A/G)ACAAAGAA binding site in the P. anserina genome indicated that this motif is preferentially localized on segments of 1000 bp upstream of the predicted translational start sites (p-value of ,0.0001) (Materials and Methods). In contrast, a scrambled binding site did not display significant enrichment in these 1000 bp regions corresponding to promoter regions (p-value = 0.08). A total of 502 genes contain the (A/G)ACAAAGAA motif, suggesting that up to 5% of P. anserina genes may be directly controled by MATA_HMG-box transcription factors.

Significance and conservation of the HMG-box gene functions in fungi
The data presented here provide new insights into the role and the relationships of HMG-box genes in the fungus P. anserina (Pezizomycotina). We revealed a network of HMG-box genes upstream of the FMR1 and FPR1 mating-type genes, which are themselves HMG-box genes. PaHMG5 plays a central role in this network by regulating mating-type gene transcription. We identified SpSte11 as the ortholog of PaHMG5 in S. pombe by different phylogenetic analyses, and we demonstrated that both recognize almost identical binding sites. KEF1 is another important member of the network, acting upstream of PaHMG5 as a repressor. KEF1 also appears as a repressor of hyphal anastomoses, which are dramatically deregulated in a Dkef1 strain. Moreover, the control of mating-type target genes by FMR1 and FPR1 was described in a previous report [73], which, together with the results presented here, provides the first exhaustive view of the regulatory circuits upstream and downstream of mating-type genes in a filamentous Ascomycete.
Strikingly, ten of the 12 HMG-box genes identified in the genome of P. anserina control fertility and sexual development. Although two genes (PaHMG2 and PaHMG7) were not directly involved in sexual reproduction, the deletion of PaHMG7 displayed a synergistic effect with the deletion of PaHMG3 on the perithecium distribution in co-cultures of mat2 and mat+ strains. This observation demonstrates that 11 of the HMG-box genes have a direct or indirect function during sexual reproduction. The genome-wide systematic deletion analysis of F. graminearum transcription factors [43] allowed us to compare the resulting phenotypes with those of P. anserina (Table S8). A total of six out of the 11 HMG-box genes deleted from F. graminearum were involved in perithecium development. One HMG-box gene (FGSG_06760) was not deleted, and the perithecium distribution on the mycelium was not tested in F. graminearum as described here. Taken together, these data implicate the HMGB superfamily in sexual reproduction in Pezizomycotina, although further analyses in other species should validate this finding.
Comparative analyses of sex regulatory pathways indicate that transcription factors have often evolved to accommodate unique rather than conserved functions, even across closely related lineages [82,83] (reviewed in [84,85]). For example, the pheromone-response pathway is controlled by an HMG-box protein in C. neoformans [41] and a transcription factor distantly related to the homeodomain family in S. cerevisiae [86], providing evidence for a dramatic change in a key regulator in these species, despite the presence of strong effector conservation [41]. Our study revealed that PaHMG5 and SpSte11 [30] are striking exceptions to the unusual plasticity of pathways regulating sex. These two orthologous HMG-box proteins positively regulate mating-type gene transcription in P. anserina and S. pombe. Moreover, the transcription factor analysis published by Son et al [43] suggests that the SpSte11 ortholog in F. graminearum (FGSG_01366, GzHMG010) also shares this conserved function. The phenotype of F. graminearum deleted for GzHMG010 recapitulates the phenotype of mating-type gene deletions (Table S8, see GzHMG010, MAT1-1-1 and MAT1-2-1) and mating-type locus deletion [87]. This feature is expected for genes operating in the same pathway, but molecular analyses will be necessary to determine whether the control of matingtype gene expression by SpSte11 orthologs extends to other fungi. N. crassa contains two co-orthologs of PaHMG5, indicating that control of mating-type gene expression in N. crassa may be more complex than in P. anserina. Based on phylogenetic analyses and analyses of synteny, we identified NCU09387 as the ortholog of PaHMG5, while NCU02326, the inparalog of NCU09387, has no counterpart in P. anserina. Mutations of NCU09387 resulted in strains that could mate, but fruit-body development arrested before ascospore formation [55,56]. This phenotype is clearly different from the mating-type deletion, which resulted in a strain that is unable to mate [88]. This observation raises the interesting possibility that NCU02326 may control the expression of mating-type genes during fertilization, while NCU09387 is involved in the regulation of mating-type genes after fertilization.
The HMG-box gene network regulates vegetative and mating-type fertility genes Sexual reproduction relies on an interplay between vegetative tissues (the mycelium and the maternal hyphae) and sexual tissue (the hymenium). Vegetative tissues critically contribute to optimal conditions for ascogonium formation and to fertility by providing nutrients to growing fruit-bodies [68]. Inside the fruit-body, the hymenium goes through karyogamy and meiosis, and provides signals to the vegetative tissues to sustain nutrient mobilization [26]. This duality raises the question as to whether the alteration  (HMG5His) with probes corresponding to FMR1, FMR1 scrambled sequence (FMR1Sc), FPR1, PaHMG8 and PaHMG5 oligonucleotides. The probes are indicated below each panel. The interaction of HMG5His with probe was analyzed without competitor and in the presence of increasing amounts (given as fold molar excess below the triangles) of competitor (indicated above the triangles). (B) Interaction of the HMG5His protein with the MFM probe and reciprocal competition between FMR1 and MFM oligonucleotides. The probes are indicated below each panel. The competitors are indicated above the triangles. A control of the interaction of HMG5His with the FMR1 probe was performed as in A and included in the assay. HMG5His has a greater affinity for FMR1 than for MFM sequence, as indicated by the efficient exclusion of MFM probe by FMR1 competitor and the very inefficient exclusion of FMR1 probe by MFM competitor. (C) Interaction of the HMG5His protein with the MFP probe and reciprocal competition between FMR1 and MFP oligonucleotides. Legend as in B. HMG5His has a greater affinity for FMR1 than for MFP sequence, as indicated by the efficient exclusion of MFP probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by MFP competitor. (D) Interaction of the HMG5His protein with the PRE1 probe and reciprocal competition between FMR1 and PRE1 oligonucleotides. Legend as in B. HMG5His has a greater affinity for FMR1 than for PRE1 oligonucleotides, as indicated by the efficient exclusion of PRE1 probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by PRE1 competitor. doi:10.1371/journal.pgen.1003642.g008 that entailed sterility in the mutant strains affects vegetative tissues, or sexual tissues via mating-type genes. The strains deleted for PaHMG6, PaHMG8 and KEF1 showed female sterility which was restored by complementation with the Dmat strain in trikaryotic mosaic tests. Although PaHMG6, PaHMG8 and KEF1 control transcription of mating-type genes, the trikaryotic test revealed that they also control fertility genes that are independent of mating-type genes but, nevertheless, critical for sexual reproduction.
The regulatory circuit shown in Figure 6 and the complete sterility resulting from the PaHMG5 deletion points to this gene being a major regulator of sexual reproduction in mat+ and mat2 strains. We demonstrated that this gene controls mating-type genes and the pheromone/receptor systems. Pheromone and pheromone receptor genes were identified as the most critical matingtype targets for male and female fertility, respectively [73,89]. The deletion of PaHMG5 had a moderate effect on the transcription of mating-type genes (FPR1, FC = 0.25; FMR1, FC = 0.1; Table S3), but strongly reduced the transcription of pheromone genes (MFP, FC = 0.0004; MFM, FC = 0.004; Table S3). This effect on pheromone genes can explain the male sterility phenotype of DPahmg5 strains. Moreover, some genes necessary for the biogenesis of pheromones are regulated by mating-type genes [73], and their transcription may also be decreased in DPahmg5 strains, thereby enhancing male sterility. By contrast, transcription of the PRE2 receptor gene was reduced 4-fold, a value that is unlikely to result in complete female sterility, which was characteristic of the mat+ DPahmg5 strain. This observation suggests that the deletion of PaHMG5 affects the transcription of another target gene that is critical for female fertility. Further experiments will be necessary to identify this target gene and to determine if it is a target of mating-type genes or a vegetative critical fertility gene.

Relationship between the stationary phase and the HMGbox gene network
Several lines of evidence indicate that the competence of P. anserina for sexual reproduction is acquired during the stationary phase. Development of reproductive structures takes place during this phase [90], and the expression of mating-type genes increases up to 1000-fold upon entry into the stationary phase [73]. Our study provides the first evidence of a link between the stationary phase and the HMG-box genes that control the sexual cycle. Deletion of KEF1 resulted in the formation of anastomoses, appressorium-like structures, and spermatia in young hyphae, all hallmarks of the stationary phase. We propose that this gene is a critical repressor of the switch to the stationary phase, which underlies the morphological transitions of this stage. However, this gene only contributed moderately to the increase in mating-type gene expression, as its deletion resulted in a 3-fold increase of FMR1 and FPR1 transcription. Other pathways are likely to connect entry into the stationary phase and sexual competence. Two MAP kinase pathways, PaMpk2 and PaMpk1, have essential roles in establishing the stationary phase in P. anserina [67,90,91]. No experimental evidence is yet available in P. anserina to support the link between these MAP kinase pathways and HMG-box genes. Exhaustive analyses of MAP kinase pathways in S. cerevisiae and S. pombe revealed numerous connections with HMG-box proteins. For instance, SpSte11 is a direct target of the Spk1 MAP kinase pathway in S. pombe [92], suggesting that PaHMG5 may be a phosphorylation target of the Spk1 ortholog in P. anserina (PaMpk2). Another interesting connection was uncovered in S. cerevisiae between Nhp6p and the Mpk1p (Slk2p) MAP kinase pathway [61], which is orthologous to the PaMpk1 MAP kinase cascade in P. anserina. These connections provide future direction to find the molecular pathways linking the stationary phase and the sexual cycle.

A module of two HMG-box genes involved in sexual development is present in animals and fungi
Sex determination is highly variable, in contrast to other developmental systems that are well conserved through evolution. Even within a single kingdom, studies on common laboratory model organisms reveal that the genetic mechanisms of sex determination bear little, if any, resemblance. Strikingly, PaHMG5 and SpSte11 have conserved their function as matingtype regulators in P. anserina and S. pombe [30], although these two organisms have diverged 550 million years ago [93]. Moreover, the mating-type genes in both fungi are themselves HMG-boxgenes, thereby defining an HMG-box module that is conserved in both fungi. An HMG-box module involved in sexual reproduction is also present in the basidiomycete fungus U. maydis. The Rop1 protein directly regulates the transcription of prf1, another HMGbox gene [37]. The Prf1 protein in turn induces the expression of mating-type genes [38,39]. HMG-box genes also play a critical role in sex determination in vertebrates. Sry, the mammalian Ychromosomal testis-determining HMG-box gene is an activator of Sox9. Sox9 is also conserved among non-mammalian vertebrate species and has an ancestral and pivotal role in sex determination [2]. The conservation of a similar regulatory HMG-box module in vertebrates and Dikarya reveals a commonality of sex regulation in animals and fungi. Sry and Sox9 are not orthologs of the fungal HMGB module genes. The two modules are thus analogs, not homologs. However, Martin et al [27] noted that the SexM protein of Phycomyces blakesleeanus was classified within the SOX-TCF_HMG subfamily. This placement is confirmed in the phylogram presented here (Figure 2: SexM, Phybl8). The discovery of a SOX-TCF_HMG-box regulator of mating-type genes in P. blakesleeanus would indicate an ancestral origin for an HMG box module involved in sex determination in Opisthokonta. Further investigations on HMG-box transcription factors and sex regultation in fungi should provide relevant information about the conservation and evolution of such modules.

Strains and media
The genetic and biological features of P. anserina were first described by Rizet and Engelmann [94] and current culture techniques can be found at http://podospora.igmors.u-psud.fr/ methods.php. The strains used in this study were all derived from the S strain [95], which was used to determine the P. anserina genome sequence [53].
Statistical test for HMG-box gene distribution in the P. anserina genome A total of nine HMG-box genes mapped to chromosome I and three HMG-box genes mapped to other chromosomes. Among the nine genes that mapped to chromosome I, three mapped to the mating-type locus and corresponded to mat+ and mat2 idiomorphs (reviewed in [26]). The idiomorph considered for statistical analysis was the mat+ idiomorph, which contained one HMGbox gene (FPR1). The total number of HMG-box genes on chromosome I was therefore, seven. The sizes of chromosome I to VII were 8,813,526 bp, 5,165,605 bp, 4,712,833 bp, 3,808,397 bp 4,734,309 bp, 4,264,133 bp and 4,087,213 bp, respectively. The total size of the P. anserina genome was 35,686,016 bp. The expected number of HMG-box genes on chromosome I was between two and three. The p-value was calculated on a contingency table http://www.graphpad.com/ quickcalcs/contingency1.cfm using Fisher's exact test (p-value = 0.07 for two genes; p-value = 0.18 for three genes). The expected number of HMG-box genes on chromosome I was not significantly different from seven genes.

Phylogenetic analyses
Sequence acquisition, identification of consensus amino acids and phylogenetic analysis were performed as described previously [27]. The HMG-box domains used to build the phylogenetic tree ( Figure 2) and their alignment are in dataset S1 and dataset S2, respectively.

Gene deletion and complementation
To delete the chromosomal copy of the eight HMG-box genes, eight plasmids containing deletion cassettes conferring resistance to hygromycin B [Hyg R ] were constructed according to the N. crassa strategy for high-throughput generation of gene deletion [97] with modifications aimed at minimizing errors in the 59 and 39 flanking regions [73] (see Table S9 for primer sequences). The deletion cassette was released from the vector by AscI digestion prior to transformation of DPaKu70 protoplasts [98]. This transformation assay consistently yielded a high percentage of transformants with the correct deletion (.90%). One or two transformants obtained from each assay were subjected to Southern blot analysis to confirm the deletion ( Figure S2), and one transformant with the expected hybridization pattern was selected for further analysis. The eight primary transformants containing corresponding targeted deletions were genetically purified by crossing with a wild-type strain of opposite mating type. This eliminated untransformed nuclei and segregated out the DPaKu70 mutation through its phleomycin resistant [Phleo R ] phenotype. Screening for [Hyg R , Phleo S ] strains allowed the identification of mat+ and mat2 strains containing the HMG-box deletion but lacking DPaKu70. These strains constituted the stock of the deletion mutant for subsequent studies.
To ensure that the phenotype(s) observed for the HMG-box gene deletion mutants was actually due to inactivation of the relevant gene, the wild-type allele was reintroduced by transformation into the corresponding mutant. Wild-type alleles were obtained by amplifying fragments encompassing the corresponding gene (see Table S10 for primer sequences) and these were used directly for co-transformation of the mutant strain with the pPable vector [77], which conferred resistance to phleomycin. A significant number of co-transformants displaying a restored wild-type phenotype were recovered in each assay, demonstrating that phenotypes were not due to additional mutations (Table S11). It should be noted that introduction of the PaHMG5 wild-type allele into the DPaHMG5 mutant only rescued the male defect without restoring female fertility, indicating partial complementation.

Measure of spermatium production and activity
Mutants were grown for 7 and 14 days on Petri dishes containing minimal agar medium. Spermatia were recovered by washing the surface of the dish with 1.5 ml of water and were counted with a haemacytometer chamber. A diluted spermatium suspension was used to fertilize a wild-type strain of opposite mating type and perithecia were counted after incubation for 4 days. Typically, 30% to 50% of spermatia from the wild-type strain were fertilizing, giving rise to perithecia. Duplicates were carried out for all strains and the whole experiment was performed twice to confirm the data.

Microscopy
Microscopic observations were made on 4-day-old mycelia growing on cellophane placed on solid M0 medium (minimal medium lacking dextrin as carbon source) in a Petri dish. Small pieces (1 cm 2 ) of cellophane containing mycelium were cut with a scalpel and mounted upside down in water. Pictures were taken with a Leica DMIRE 2 microscope coupled to a 10 MHz Cool SNAPHQ charge-coupled device camera (Roper Instruments). Since penetration of the cellophane through apressorium-like structures occurs perpendicularly to the surface, pictures were obtained at one micrometer increments to capture this process. Stacks of pictures were analyzed with ImageJ (http://rsb.info.nih. gov/ij) and deconvolution was performed with CombineZP (Alan Bradlay; alan@micropics.org.uk). Calculations of the mean and the standard deviation of the distance between emergence of appressorium-like structures and the leading edge of the thallus were made using 20 and 10 individual measurements in wild-type and DPahmg9 strains, respectively.

Construction of the DPahmg3 DPahmg7 double mutant
One heterokaryotic DPahmg3 mat+/DPahmg7 mat2 culture was self-crossed, and mat+ and mat2 homokaryotic double mutants were isolated from the progeny. Single and double mutants displayed a [Hyg R ] phenotype and were, thus, undistinguishable. Therefore, a search was performed for asci showing first division segregation of the [Hyg R ] phenotype (i.e., asci with two [Hyg R ] and two [Hyg S ] ascospores). Sensitive ascospore-derived cultures from these asci carried wild-type alleles of both genes; hence, resistant ones harbored mutations in both.

Deregulation of the PaHMG5 gene
The plasmid pBHGSTE11 contains a fusion of the P. anserina gpd promoter and initiation codon [74] with the coding phase of PaHMG5. The Pagpd promoter was a 0.38 kbp fragment obtained from the pPable plasmid [77], digested with NcoI, treated with Klenow and digested with XbaI. The PaHMG5 sequence was amplified with Pfu (Promega) from GA0AB103CF05 [53] using 5PSTE11 and 3HindSTE11 primers (see Table S10 for primer sequences) and digested with HindIII. The Pagpd promoter and the PaHMG5 fragment were then ligated into the HindIII and XbaI sites of plasmid pBCHygro [99] to yield the pBHGSTE11 plasmid. Sequencing of the entire fusion confirmed that the first Pagpd codon was in frame with PaHMG5 and that no mutation altered PaHMG5.The pBHGSTE11 plasmid was introduced into mat+ wild-type protoplasts and 10 [Hyg R ] transformants were phenotypically analyzed in a cross with a mat2 wild-type strain. A total of nine transformants showed a flat and female sterile mycelium, but they were fertile as male partners. To determine more precisely the phenotypic effects resulting from the integration of the Pagpd::PaHMG5 fusion, the progeny from three representative transformants were subjected to genetic analysis. Segregation of the Pagpd::PaHMG5 fusion was scored through the [Hyg R ] phenotype. Most unpigmented ascospores did not germinate. Their genotype could nevertheless be deduced from tetrad analysis. For two transformants, the presence of the Pagpd::-PaHMG5 fusion was responsible for an ascospore pigmentation and germination defect (although most unpigmented ascospores did not germinate their genotype could nevertheless be deduced from tetrad analysis). However, pigmented ascospores giving rise to a [Hyg R ] mycelium were recovered in the same progeny. These displayed a similar phenotypic vegetative alteration as observed in the primary transformants (flat mycelium and female sterility). In a transformant corresponding to a different integration site (different second division segregation % of [Hyg S ]/[Hyg R ]), the presence of the Pagpd::PaHMG5 fusion did not affect ascospore pigmentation; instead it conferred the vegetative mycelium defect (flat mycelium and female sterility). These data and conclusions were subsequently confirmed by analyzing second generation progeny which were obtained by crossing the purified [Hyg R ] Pagpd::PaHMG5 bearing transformants with the wild-type strain.

RT-qPCR experiments
Vegetative cultures for RNA preparation were performed on Petri dishes containing minimal medium and covered with a cellophane sheet (Bio Rad Hercules, USA). These cultures were inoculated with nine implants from mat+, mat2 or HMG-box mutants of either mating type. Dishes were placed at 27uC under constant light (0 h) and were removed from the incubation room at 96 h, at which time P. anserina was competent for fertilization [73]. Mycelia were harvested and RNAs were extracted as described previously [73]. Purified RNAs were submitted to an additional DNase digestion in solution and cleaned up once more on RNeasy Plant Mini Kit (Qiagen, Hilden, Germany). Total RNAs were reverse transcribed with SuperScript III (Lifes Technologies) according to manufacturer's instructions. Each time the expression of an intronless gene was quantified, a non-reverse transcribed (NRT) control was performed for each biological replicate. For genes with introns, all primers were designed against two consecutive exons (see Table S12 for primer sequences) and an NRT control was systematically performed on a pool of biological replicates. Each RT-qPCR experiment contained at least five biological replicates and each point was performed in technical duplicate. Normalization genes for DPahmg5, DPahmg8 and Dmthmg1 were selected from a pool of ten housekeeping genes using geNorm [100] as described previously [73]. geNorm failed to select normalization genes for DPahmg6 and DPahmg9 strains, probably because metabolism was altered in these strains. A single stable reference gene was identified in these strains using NormFinder [101]. The normalization genes are listed in Table  S13. RT-qPCR normalization was performed according to the relative quantification method with kinetic PCR efficiency correction. Standard error and 95% confidence interval calculations, and other statistical analyses were performed using REST 2009 software (Qiagen, Hilden, Germany) [102]. The FC in the expression of a gene of interest was computed as the normalized relative quantity of cDNA in sample relative to that in the control: FC = relative quantity of cDNA for the gene of interest x (geometric mean of relative quantity of cDNA for the normalization genes) 21 The relative quantity (RQ) of a cDNA was: RQ = efficiency of amplification (arithmetic mean for WT strain replicates -arithmetic mean for mutant strain replicates) The efficiency of amplification was above 1.8 for all analyzed genes. Genes were defined as downregulated in the mutant strain if 0,FC,1 with a p-value of ,0.05. On the other hand, genes were defined as up-regulated in the mutant strain if FC.1, with a p-value of ,0.05. FCs with a 95% confidence interval including the value of 1 were not considered significant [75].

Consensus motif search and analysis of HMG-box binding motif distribution
Motif searches were conducted using MEME http://meme. nbcr.net/meme/cgi-bin/meme.cgi [78] on segments of 1000 bp upstream of the predicted translational start sites. A group of core sequences including KEF1, mtHMG1, PaHMG5, PaHMG8, and the FMR1 and FPR1 mating-type genes was first analyzed using MEME to identify the consensus HMG-box binding site. Subsequently, each candidate sequence was included with the core sequences for MEME analysis. Segments of 1500 bp upstream of the predicted translational start sites were analyzed when genes yielded no hit using MEME. The number of occurrences of the (A/G)ACAAAGAA binding site was counted for the P. anserina genome (n = 1149 in 35,686,016 bp). Assuming a random distribution throughout the genome, the (A/G)ACAAA-GAA sequence is expected to occur 0.55 time in the promoter regions of the genes selected for MEME analysis (17,000 bp). The observed and expected numbers of (A/G)ACAAAGAA sequence were compared on a contingency table (http://www.graphpad. com/quickcalcs/contingency1.cfm) and a p-value was computed using Fisher's exact test. The (A/G)ACAAAGAA sequence is significantly enriched in the selected set of gene analyzed with MEME (p-value = 0.016). We further analyzed the distribution of the (A/G)ACAAAGAA sequence in the promoter region of all P.

Protein expression, purification and electrophoretic mobility shift assays
The full length PaHMG5 cDNA was amplified by PCR with LA Taq (TaKaRa, Shiga, Japan) from reverse transcribed total RNAs with primers Nde13 and HisBam13 (Table S10) according to the manufacturer's instructions. HisBam13 was designed to introduce an His 6 tag downstream of the 39 coding sequence. The PCR products were cloned into the pET28 vector (Novagen) between NdeI and BamHI restriction sites and inserts were sequenced to identify cDNA without mutations. E. coli BL21 (DE3) transformed with the recombinant PET28 vector was grown in 26YT medium (MP Biomedicals) supplemented with kanamycin at 50 mg/ml. Approximately 800 ml of culture medium was incubated in a shaker at 200 rpm at 37uC until OD600,0.6-0.8. Protein expression was then induced with 0.5 mM isopropyl b-Dthiogalactopyranoside (Sigma) and the cell culture was further incubated at 15uC overnight. Cells were harvested by centrifugation, resuspended in 40 ml of 20 mM Tris pH9.0, 500 mM NaCl, 5 mM b-mercaptoethanol and protease inhibitor cocktail (Roche), and stored at 220uC. Cell lysis was achieved by sonication, and the cell extract was centrifuged at 20000 g for 30 min at 4uC. The His-tagged protein from the soluble fraction was purified on a nickel-nitrilotriacetic acid column (Qiagen Inc.) and eluted with an isocratic imidazole gradient, followed by a cation exchange step on a HiTrap Heparin column (GE Healthcare) equilibrated in 20 mM Tris (pH9.0), 300 mM NaCl, 5 mM b-mercaptoethanol, and 5% glycerol. The protein was eluted using a linear salt gradient. The production of recombinant protein was confirmed by SDS-PAGE. Lysis, soluble and some purified fractions were tested by western blot analysis. Proteins were resolved by SDS-PAGE and transferred onto nitrocellulose membrane (Protran, Whatman). The membrane was blocked by incubation for 1 hour at room temperature with 5% nonfat milk in TBS-T (Tris buffered saline-Tween 20, pH 7.5), incubated with rocking for 1 hour with anti-6His IgG conjugated with peroxidase (1/2,000) (Roche), and developed using the BM Blue POD substrate reagents from Roche.
Complementary primers (Table S14) were annealed to yield double stranded 33 bp oligonucleotides with single base 59 overhangs consisting of a guanine to promote efficient labeling by T4 polynucleotide kinase [103]. The double stranded oligonucleotides were 59-end labeled by T4 polynucleotide kinase (Thermo Scientific) and [c-32 P]ATP (222TBq/mmole), according to the manufacturer's instructions. The probe was purified on MicroSpin G-25 columns (GE Healthcare) and further processed as described in [104].
EMSAs were performed with 340 mg of purified HMG5His in 20 mM Tris (pH9.0), 500 mM NaCl, 5 mM b-mercaptoethanol, 5 mM MgCl 2 and 5% glycerol supplemented with 1.25 mg of poly(dI-dC) in a total volume of 11 ml. Samples were incubated for 20 min on ice, labeled probe was added, and incubation was continued for 1 h on ice. For competition experiments, unlabeled double stranded oligonucleotides were added after incubating 20 min with poly(dI-dC) and incubated further for 15 min before adding the labeled probe.
Protein-DNA complexes were separated in polyacrylamide gel (6%) in 0.256 Tris-borate-EDTA buffer at 200 V per gel for ,80 min. Radioactive probes were visualized using a Typhoon laser scanner (GE Healthcare).

Supporting Information
Dataset S1 HMG-box domains used for Figure 2.  Figure S1 Comparative organization around SpeSte11 orthologs in P. anserina and N. crassa. Orthologs were determined using FUNGIpath [52]. The gene names are enclosed in arrowed boxes indicating gene orientation. Orthologous genes are enclosed in boxes of identical color and connected with double arrows. Genes enclosed in white boxes do not have any ortholog in P. anserina or N. crassa. Gene sizes are not to scale. A: synteny in P. anserina and N. crassa for genes upstream and downstream of PaHMG5 and NCU09387. The conserved synteny indicates that PaHMG5 is the ortholog of NCU09387. B: organization of genes upstream and downstream of NCU02326 in N. crassa and search for a synteny in P. anserina. The absence of a conserved synteny confirms that the ortholog of NCU02326 is absent in P. anserina. (TIFF) Figure S2 Genomic Southern blots of HMG-box mutant strains probed with the hygromycin sequence.   Figure S4 Electrophoretic mobility shift assays with PaHMG5 and HMG-box gene oligonucleotides. Interaction of His tagged PaHMG5 (HMG5His) with probes corresponding to HMG-box genes: KEF1, mtHMG1, PaHMG2, PaHMG3, PaHMG4 and PaHMG6. Legend as in Figure S3 (A). (TIFF) Figure S5 Electrophoretic mobility shift reciprocal competition assays with PaHMG5 and HMG-box gene oligonucleotides. Legend as in Figure 8 (B). (A) Interaction of the HMG5His protein with the PaHMG2 probe and reciprocal competition between FMR1 and PaHMG2 oligonucleotides. HMG5His has a greater affinity for FMR1 than for PaHMG2 oligonucleotides, as indicated by the efficient exclusion of PaHMG2 probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by PaHMG2 competitor. (B) Interaction of the HMG5His protein with the PaHMG3 probe and reciprocal competition between FMR1 and PaHMG3 oligonucleotides. HMG5His has a greater affinity for FMR1 than for PaHMG3 oligonucleotides, as indicated by the efficient exclusion of PaHMG3 probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by PaHMG3 competitor. (C) Interaction of the HMG5His protein with the PaHMG4 probe and reciprocal competition between FMR1 and PaHMG4 oligonucleotides. HMG5His has a greater affinity for FMR1 than for PaHMG4 oligonucleotides, as indicated by the efficient exclusion of PaHMG4 probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by PaHMG4 competitor.
(D) Interaction of the HMG5His protein with the PaHMG6 probe and reciprocal competition between FMR1 and PaHMG6 oligonucleotides. HMG5His has a greater affinity for FMR1 than for PaHMG6 oligonucleotides, as indicated by the efficient exclusion of PaHMG6 probe by FMR1 competitor and the inefficient exclusion of FMR1 probe by PaHMG6 competitor. (TIFF)