Genome-Wide Gene Expression Profiling of Fertilization Competent Mycelium in Opposite Mating Types in the Heterothallic Fungus Podospora anserina

Background Mating-type loci in yeasts and ascomycotan filamentous fungi (Pezizomycotina) encode master transcriptional factors that play a critical role in sexual development. Genome-wide analyses of mating-type-specification circuits and mating-type target genes are available in Saccharomyces cerevisiae and Schizosaccharomyces pombe; however, no such analyses have been performed in heterothallic (self-incompatible) Pezizomycotina. The heterothallic fungus Podospora anserina serves as a model for understanding the basic features of mating-type control. Its mat+ and mat− mating types are determined by dissimilar allelic sequences. The mat− sequence contains three genes, designated FMR1, SMR1 and SMR2, while the mat+ sequence contains one gene, FPR1. FMR1 and FPR1 are the major regulators of fertilization, and this study presents a genome-wide view of their target genes and analyzes their target gene regulation. Methodology/Principal Findings The transcriptomic profiles of the mat+ and mat− strains revealed 157 differentially transcribed genes, and transcriptomic analysis of fmr1− and fpr1− mutant strains was used to determine the regulatory actions exerted by FMR1 and FPR1 on these differentially transcribed genes. All possible combinations of transcription repression and/or activation by FMR1 and/or FPR1 were observed. Furthermore, 10 additional mating-type target genes were identified that were up- or down-regulated to the same level in mat+ and mat− strains. Of the 167 genes identified, 32 genes were selected for deletion, which resulted in the identification of two genes essential for the sexual cycle. Interspecies comparisons of mating-type target genes revealed significant numbers of orthologous pairs, although transcriptional profiles were not conserved between species. Conclusions/Significance This study represents the first comprehensive genome-wide analysis of mating-type direct and indirect target genes in a heterothallic filamentous fungus. Mating-type transcription factors have many more target genes than are found in yeasts and exert a much greater diversity of regulatory actions on target genes, most of which are not directly related to mating.


Introduction
Sexual reproduction is driven by specific gene expression programs required for identifying a sexually compatible partner, mating, meiosis and generation of progeny. Regulation of the genes involved in the identification of the mating partner and mating is well understood at a genome-wide level in the budding yeast Saccharomyces cerevisiae [1] and the fission yeast Schizosaccharomyces pombe [2], and these analyses have identified a limited number of cell specific-genes, most of which have a known function. Filamentous fungi provide an opportunity to study mating-type evolution and function in multicellular organisms, which is not possible in unicellular yeasts; however, no complete description of the regulation exerted by mating-type regulatory genes is available in the filamentous Ascomycetes. Furthermore, our understanding of the functions controlled by the mating-type transcription factors is very incomplete as most of their target genes remain unknown.
Various other idiomorphic genes have been described in Pezizomycotina (see [33] for a review), notably MAT1-1-2 and MAT1-1-3. MAT1-1-2 proteins contain a PPF domain, which is characterized by the conservation of three invariant residues, two prolines and one phenylalanine [34], while MAT1-1-3 transcription factors are characterized by an HMG-box and form a subgroup within MATA_HMG [21]. Their roles have been investigated in P. anserina [29,32] and N. crassa [4,35,36], and they are required for post-fertilization development of the fruiting body. Despite numerous genetic analyses of mating-type gene functions, only a few mating-type target genes, essential for mating, have been identified. These genes are involved in the conserved pheromone/receptor system initially found in budding yeast.
The control of cell-type specificity exerted by mating-type genes has been examined in detail through genetic approaches in the heterothallic filamentous ascomycete P. anserina (see Figure 1 for the correspondence of P. anserina and standard nomenclature for MAT genes). These studies revealed that the main regulators of fertilization, FMR1 and FPR1, have activator and repressor activities on the functions required for fertilization ( Figure 1) ( [13,28] and reviewed in [3]). Two other genes, SMR1 (MAT1-1-2) and SMR2 (MAT1-1-3), are present in the mat2 idiomorph. These two genes are not involved in the activation of the mat2 functions required for fertilization, although the MATA_HMG transcription factor, SMR2, plays a minor role in the repression of the mat+ fertilization functions in mat2 strains. Subsequent studies have identified only two target genes of FMR1 and FPR1. These target genes encode the MFM and MFP pheromone precursors and are transcriptionally activated in a mating type-specific manner ( Figure 1) [13]. Coppin et al. [13] have suggested that genes encoding pheromone precursor processing proteins are subjected to repression in a mating-type specific manner (Figure 1), but no such genes have been identified. The dual regulatory action of matingtype genes initially reported in P. anserina was also reported in G. moniliformis for target genes not directly involved in mating [10].
In this study, we present a transcriptomic genome-wide identification of the genes that are differentially transcribed between the mat+ and mat2 strains in a vegetative stage competent for fertilization. A total of 157 genes were identified and repression or activation of these genes was determined by analyzing the transcriptomic profile of fpr1 2 and fmr1 2 mutants. The data provided, for the first time, a complete description of the regulation exerted by mating-type genes on their target genes in a heterothallic Pezizomycotina. Many of the target genes are not involved in mating, in agreement with previous observations in G. moniliformis [10] and S. macrospora [7,37], and deletion of 32 selected genes identified only two genes essential for mating. In addition, the search for common mating-type target genes in P. anserina and G. moniliformis or S. macrospora revealed statistically significant numbers of orthologous pairs; however, these conserved target genes have different transcriptional profiles.

Results
Time-course RT-qPCR analysis of mating-type gene transcript level during vegetative growth FMR1 and FPR1 play essential roles in mating-type determination and fertilization (see Introduction). An RT-qPCR experiment was performed to determine the transcription pattern of these two genes and to identify the optimal time to search for differentially expressed genes in mat2 and mat+ mycelia. The transcript levels of mating-type genes were investigated on mat+ and mat2 mycelia harvested after incubation in Petri dishes for 24  The quantification cycle (Cq) values for FMR1 and FPR1 cDNAs were very high at 24 h, indicating that their expression was very low at the beginning of vegetative growth (Table S1). The transcript levels of these two genes increased markedly at 48 h and reached a plateau at 72 h (Figure 2A and B). The nine colonies grown in a Petri dish made contact at 48 h ( Figure 2C); thus, maximal FMR1 and FPR1 transcript levels occurred 24 h after confluence. A decrease in FMR1 and FPR1 transcript levels was observed 48 h after fertilization.
Transcript levels of the SMR1 and SMR2 genes were undetectable throughout vegetative development until 120 h, at which time very low levels of transcripts were detectable in the mat2 mycelium (Table S1). The transcript levels of these two genes increased strongly after fertilization ( Figure 2A) in agreement with their essential roles during perithecium development [32,38]. In line with our usual experimental conditions, fertilization was performed typically with 96 h-old mycelia, corresponding to the middle of the plateau phase of maximum transcript accumulation for FMR1 and FPR1. This time-point was thus used in subsequent experiments to determine the genes that were differentially transcribed in mat2 and mat+ cultures.
Transcriptomic profiling of mat2 and mat+ strains A transcriptomic comparison of different mating-type strains provided an in-depth analysis of the genes that are differentially transcribed in the mat2 and mat+ strains. RNA was extracted from cultures grown under constant light for 96 h and prepared for hybridization (see Materials and Methods) to whole genome geneexpression microarrays developed for P. anserina [39].
Genes were defined as up-regulated in the mat+ strain if their transcript levels showed a $2-fold change (FC$2) with a p-value of ,0.005 in the mat+ strain compared to the mat2 strain (mat+ vs mat2 comparison). Reciprocally, genes were defined as upregulated in the mat2 strain when they showed a #2-fold change (FC#22) with a p-value of ,0.005 in the mat+ vs mat2 comparison. A total of 157 genes were differentially transcribed in the mat+ vs mat2 comparison (Table S2); 88 genes accumulated more transcripts in mat+ than in mat2 strains (Table S3), and 69 genes accumulated more transcripts in mat2 than in mat+ strains (Table S4).
Next, we searched for genes expressed in a mating-type specific manner. Assuming that genes not expressed in one mating type should display a signal-to-standard-deviation ratio (SSR)#3 [39], 13 genes were identified that were transcribed exclusively in the mat+ strain and six genes were identified that were transcribed exclusively in the mat2 strain ( Table 1). Two of these genes have an ortholog with a known function in S. cerevisiae: STE6, the ortholog of Pa_5_11640 (ABC transporter), is involved in the transmembrane export of the lipophilic pheromones [40], and FBP1, the ortholog of Pa_4_9360 (fructose-1,6-bisphosphatase), which has no reported role in yeast mating. These 19 genes with mating type-specific expression did not include genes encoding the pheromone precursors and the pheromone receptors. This is in agreement with previous genetic analyses, which suggested that, although mat+ and mat2 pheromone genes are transcribed at low levels in mat2 and mat+ vegetative nuclei, respectively, tight repression of the expression of these genes in inappropriate cells takes place at the post-transcriptional level [13]. Interestingly, a low level of transcription was reported for an N. crassa pheromone receptor gene in inappropriate mating-type cells and a role in pheromone gene transcription was evidenced [41].
The consistency of microarray data with RT-qPCR analyses was tested for 14 genes selected to cover a wide range of FC values. Eleven genes displayed a similar FC direction in both experiments (Table S5). As commonly observed in the validation of array data, the FC values were much greater in the RT-qPCR than in the microarray experiments for the genes that were induced strongly [42,43]. The strongest differences were observed for the MFM and MFP transcripts, which yielded dramatically low Cq values in the mat2 and mat+ strains, respectively (Table S6), suggesting that these two genes may be among those with the highest transcription level at this stage of development. Similarly, the pheromone precursor gene of N. crassa, mfa-1, was identified as the most abundant clone in starved mycelial cDNA libraries [18]. Of the 14 genes selected for microarray validation, three genes had an FC value close to one with non-significant p-values; microarray experiments gave typically more reliable results than RT-qPCR for genes that displayed low induction levels.
Transcriptomic profiling of fmr1 2 and fpr1 2 mutant strains Two types of genetic differences can account for differential transcription in mat+ and mat2 strains. The first type includes the two mating-type genes that are expressed at 96 h, namely FPR1 in the mat+ strain and FMR1 in the mat2 strain. Another possible type of genetic difference is the single nucleotide polymorphism (SNP) between mat+ and mat2 strains. In these isogenic strains, the polymorphism is restricted to the environment of the mating-type locus and may be associated with differential gene transcription in mat+ and mat2 strains. The effect of FMR1 and FPR1 on the transcription of the 157 genes, differentially transcribed in mat+ and mat2 strains, was determined by transcriptomic profiling of the fmr1 2 and fpr1 2 strains [44]. The mat2 vs fmr1 2 and mat+ vs fpr1 2 comparisons (Table S2) enabled the identification of the target genes (direct or indirect) of FMR1 and FPR1, respectively. Genes that were differentially transcribed due to mating-type linked SNPs were differentially transcribed in the fpr1 2 vs fmr1 2 comparison (Table S2). Taken together, these comparisons allowed us to determine the contribution of FMR1, FPR1 and SNPs towards the transcription of each gene differentially expressed in mat+ and mat2 strains.
The 157 genes differentially transcribed in the mat+ vs mat2 strains were divided into eight classes based on their transcription patterns (Table S7), and the regulatory effect of FMR1 or FPR1 (activation or repression) on each target gene was inferred from the FC values. It should be noted that the technology used in this study did not enable us to determine whether FMR1 and FPR1 were direct or indirect regulators of their target genes. The number of genes in each class and the control exerted by the mating-type genes are summarized in Figure 3.
Class 1 contained 13 genes differentially transcribed in mat+ and mat2 strains as a consequence of the mating-type linked SNPs and independent of any control by mating-type proteins. This class included the FPR1 (Pa_1_20590) and FMR1 genes. These two Figure 1. Regulation of fertilization by mating-type genes in P. anserina. Mating-type protein names are enclosed in colored circles: magenta, MATa-HMG protein; cyan, MATA-HMG proteins. Grey squares represent target genes present in mat+ and mat2 strains. The standard nomenclature is indicated below the P. anserina-specific gene names. MFM and MFP encode the pheromone precursors [13]. Arrows with heads and blunt ends indicate target-gene activation and repression, respectively. Data were compiled from [13,28]. doi:10.1371/journal.pone.0021476.g001 genes were not differentially transcribed in the mat+ vs fpr1 2 and mat2 vs fmr1 2 comparisons, indicating that their transcription was not under the positive or negative control of the proteins that they encode. Classes 2 to 7 displayed all possible combinations of control by FMR1, FPR1 and mating-type linked SNPs. Strikingly, class 2 contained genes, which were differentially transcribed in mat+ and mat2 strains due to the effects of the three factors: FPR1 (in mat+ strain), FMR1 (in mat2 strain), and mating-type linked SNPs (in either mat+ or mat2 strains). Class 8 contained 41 genes differentially transcribed in mat+ and mat2 strains, but their control by FMR1, FPR1 or mating-type linked SNPs could not be clearly established because their FC scores or p-values were outside the selected significant values. A total of nine genes had FC scores close to the cut-off values (22,FC,21.8 or 1.8,FC,2) in the mat+ vs fpr1 2 comparison, as exemplified by Pa_4_5280 (FC = 21.97, P = 0.00048). The highly significant p-values associated with these FCs indicated that these genes were controlled by FPR1. Similarly, five genes appeared to be under the control of FMR1, based on their FCs and significant p-values in the mat2 vs fmr1 2 comparison (e.g. Pa_1_3100, FC = 1.97, P = 0.00056). The analysis of FC values facilitated the identification of five additional genes from class 8, which were controlled by both mating-type genes. For each of these genes, FCs of wild-type-to-mutant comparisons were below the cut-off value, but their combination resulted in an FC value that corresponded to that observed in the mat+ vs mat2 comparison. This was exemplified by Pa_7_9690 (protein-S-isoprenylcysteine O-methyltransferase), activated by FPR1 (FC = 1.69 in the mat+ vs fpr1 2 comparison) and repressed by FMR1 (FC = 21.75 in the mat2 vs fmr1 2 comparison). Activation by FPR1 and repression by FMR1 were combined in Figure 2. Time-course RT-qPCR analysis of mating-type gene transcript levels during vegetative growth and fruiting-body development. A: Relative quantification of mat2 mating-type gene transcription using the Relative Expression Software Tool (REST) [69]. RNA was extracted from mat2 colonies harvested during vegetative growth (24 h, 48 h, 72 h, 96 h and 120 h) and from mat2 colonies grown for 48 h after fertilization with mat+ microconidia (crossing 48 h). The reference time for FMR1 fold-change determination was 24 h. The reference time for SMR1 and SMR2 fold-change determination was 120 h. Quantification cycle (Cq) values used for relative quantification are in Table S1. B: Relative quantification of mat+ mating-type gene transcription with REST [69]. RNA was extracted from mat+ colonies harvested during vegetative growth ( the mat+ vs mat2 comparison, yielding a 2.95-fold change. Following the same rationale, two genes (Pa_4_4560 and Pa_6_6730) were repressed by FPR1 and activated by FMR1, and two genes (Pa_4_3860 and Pa_2_11120) were activated by FPR1 and induced by SNPs linked to the mat+ mating type. None of the remaining genes in class 8 displayed differential transcription in the mat+ vs mat2 comparison, and FCs close to a value of one with significant p-values in the other comparisons. This precluded the existence of an undetermined regulatory factor that controls class 8 genes in addition to mating-type genes and SNPs. An additional ninth class (Table S7 and Figure 3) contained mating-type target genes that were not differentially transcribed in the mat+ vs mat2 comparison (e.g. Pa_5_4645, FC = 1.33) because they were activated or repressed to a similar level in mat+ and mat2 strains (e.g. Pa_5_4645, FC = 3.46 in the mat+ vs fpr1 2 comparison and FC = 3.4 in the mat2 vs fmr1 2 comparison). This class contained 10 genes with diverse functions ( Table 2). Among the 167 genes examined in total, 132 genes were ultimately controlled by FMR1 or FPR1 or by both transcription factors, as shown in Figure 3B. The remaining 35 genes (13 genes from class 1 and 22 genes from class 8) were not included in this clustering. Class 1 consisted of genes exclusively under the control of matingtype linked SNPs, while the remaining genes from class 8 had FC values below the cut-off or not significant p-values and could not be included, with confidence, as genes controlled by FMR1, FPR1 or mating-type linked SNPs.
The transcriptomic profiling of wild-type vs mutant strains revealed many more genes than the 157 that were differentially transcribed in the mat+ vs mat2 comparison ( Figure 4 and Table  S2). The mutant strains may thus have accumulated adventitious mutations in addition to the fpr1 2 and fmr1 2 mutations. Ten genes were selected from the fpr1 2 and fmr1 2 strains to identify such putative mutations by sequencing (see Materials and Methods). No mutations were found in these ten genes, suggesting that their differential transcription might be the indirect result of adventitious mutations that have yet to be identified. The complete sequencing of the genome for the fpr1 2 and fmr1 2 strains is underway to identify these putative mutations.

Functional annotation of mating-type regulated genes
The published functional annotation of the 157 genes, differentially transcribed in the mat+ and mat2 strains (class 1 to 8), and the 10 genes that were up-or down-regulated to similar levels in both strains (class 9) was completed for the proteins with unknown functions. Each of these protein sequences was searched against the non-redundant protein database at NCBI using BLASTP [45] and their putative functions were inferred from conserved domains (Table S8 and Table S9). The Functional Catalogue (FunCat) created by MIPS [46] was used to group the genes, differentially transcribed in mat+ and mat2 strains, according to their cellular or molecular functions (Table 3). Chisquared tests indicated that the FunCat main functional categories were not differently distributed in the mat+ and mat2 strains. Approximately 50% of these proteins were of unknown function (FunCat class number 99). Among these proteins, eight did not produce a hit on any other protein and were specific to P. anserina (Table S8 and Table S9). Three genes (nad1, 2, 3) belonged to the mitochondrial genome and encoded the three sub-units of NADH dehydrogenase [47]. They were placed in class 1 (genes under the control of mating-type linked SNPs, Figure 3A and Table S7).
The search for genes involved in the sexual process identified genes encoding the pheromones, their receptors and some genes required for pheromone biogenesis. Five of these genes were upregulated in the mat+ strain: the genes encoding the mat+ lipophilic pheromone (MFP, Pa_2_2310) and the mat2 pheromone receptor (PRE2, Pa_4_1380), and the orthologs of RAM1 (farnesyl transferase b, Pa_4_7760), STE24 (protein-S-isoprenylcysteine O-methyltransferase, Pa_7_9690) and STE6 (ABC transporter, Pa_5_11640), which are required for the biogenesis of the lipophilic pheromone in S. cerevisae [48]. The first three genes clustered with the genes activated by FPR1 (class 4, Table S7), while the two last genes were activated by FPR1 and repressed by FMR1 (class 8, Table S7). The p-value for Pa_5_11640 (ABC transporter, STE6) was .0.005 but below the 0.05 upper limit of significance (class 8, Table S7). The prenylation of the S. cerevisiae lipophilic pheromone requires also the a subunit of farnesyl transferase, which is encoded by RAM2 [49]. The ortholog of RAM2 (Pa_2_9270) was not among the 157 genes that were differentially transcribed in mat+ and mat2 strains because its upregulation in the mat+ strain was below the cut-off value (FC = 1.59). However, the p-value for this gene was highly significant (P = 0.00001) ( Table S2), suggesting that this gene may be considered as up-regulated in the mat+ strain. Orthologs of the other genes required for lipophilic pheromone biogenesis were either not differentially regulated in mat+ vs mat2 strains (RCE1, AFC1/STE24, STE14 and STE23) or were not found in the P. anserina genome (AXL1). Two genes involved in the sexual process were up-regulated in the mat2 strain and were activated by FMR1: the genes encoding the mat2 hydrophilic pheromone (MFM, Pa_1_8290) and the mat+ pheromone receptor (PRE1, Pa_7_9070) (class 5, Table S7). Orthologs of the KEX2 and KEX1 genes, which are required for hydrophilic pheromone biogenesis in   S. cerevisiae [48] were not differentially transcribed in the mat+ and mat2 strains. Pa_6_7350, the only protease-encoding gene to be up-regulated in the mat2 strain, had no ortholog in S. cerevisae and no ortholog was found in P. anserina for the aminopeptidase STE13.

Phenotypic analyses of target-gene deletions
The genes that displayed an FC value higher than 5 in the mat+ vs mat2 comparison (Table S3) were selected for deletion and phenotype analysis, while two genes, Pa_2_2310 (MFP) and Pa_3_1710 (AOX), were deleted in previous studies [13,50]. Fourteen additional genes from this list were selected based on (i) putative roles in reproduction (e.g. Pa_4_7760, farnesyltransferase b, RAM1), (ii) conservation in a wide range of fungi (e.g.  Pa_2_7180), or (iii) functions of special interest (e.g., the C6 transcription factors, Pa_2_6830 and Pa_6_3770). Nine additional candidates were selected among the genes that were up-regulated in the mat2 strain (Table S4) based on similar criteria. The phenotype of the Pa_1_8290 (MFM) deletion was investigated in a previous study [13]. A total of 32 deletion strains were constructed and examined for male and female fertility and for possible defects in perithecium development ( Table 4). The two mutant strains deleted for the mat+ lipophilic pheromone receptor (PRE1, Pa_7_9070) and the mat2 hydrophilic pheromone receptor (PRE2, Pa_4_1380) showed defects in sexual reproduction. The loss of the receptor for the mat+ lipophilic pheromone resulted in female sterility in mat2 strains, while DPRE1 mat+ strains displayed wild-type fertility. Identical phenotypes were reported in N. crassa strains that were deleted for the lipophilic pheromone receptor, pre-1 [41]. The PRE2 deletion resulted in reciprocal phenotypes, namely female sterility in mat+ strains, while mat2 strains were not affected. The deletion of Pa_4_7760 (farnesyl transferase b, RAM1) was lethal, in contrast to the disruption of its ortholog in S. cerevisiae [49]. Farnesylation of Ras proteins is required for essential Ras function, suggesting that RAM1 is an essential gene. Viable RAM1 loss-of-function mutations in yeast have been attributed to an alternative prenylation process involving Cdc43p [49]. Similar to P. anserina, the deletion of RAM1 is lethal in Cryptococcus neoformans [51], indicating that the farnesylation of Ras proteins cannot be compensated for by alternative pathways in these two fungi, although both contain orthologs of Cdc43p (Pa_5_4440 in P. anserina, CNAG_02756 in C. neoformans). The 29 remaining deletion strains were not impaired for mating and showed no defects in perithecium development. The mutant strains were also tested for their sensitivity to several inhibitors (see Materials and Methods) without revealing any differences between the mutant and the wild-type strains.

Interspecies comparison of mating-type target gene transcriptional profiles
A cDNA macroarray analysis was conducted in G. moniliformis to compare a strain in which the MAT1-2-1 gene was deleted with the MAT1-2 wild-type strain [10]. The mutant and wild-type strains used in these experiments were the counterparts of those used in the mat+ vs fpr1 2 comparison, thus providing the opportunity to search for conserved mating-type target genes in these two heterothallic species. A total of 29 orthologous pairs were found in G. moniliformis MAT1-2-1 and P. anserina FPR1 target genes (P,0.0001) ( Table 5 and Table S10), none of which encoded a protein with a known function directly related to mating. The P. anserina ortholog of clone #263, PaMpk3 (Pa_1_23930), is involved in osmotic stress resistance, but has no function in sexual development (Hervé Lalucque, Fabienne Malagnac and Philippe Silar, personal communication). The cDNA macroarrays did not provide an exhaustive representation of the genome, which possibly explains the low number of orthologous pairs identified as well as the absence of the expected pheromone and receptor pheromone gene pairs. Of the 29 genes regulated by MAT1-2-1 and FPR1 in G. moniliformis and P. anserina, none was repressed by both transcription factors, only 12 were activated by both transcription factors and a total of 17 genes were activated by MAT1-2-1 and repressed by FPR1 or vice versa, suggesting low conservation of the transcriptional profiles in P. anserina and G. moniliformis. Further correlation analysis is precluded because specific FC values were not available for G. moniliformis target genes.
S. macrospora is a homothallic species that contains a MATa_HMG gene (SmtA-1) and a MATA_HMG gene (Smta-1) homologous to FMR1 and FPR1, respectively [52]. The SmtA-1 and Smta-1 target genes were identified in S. macrospora, using N. crassa microarrays [7,37] and a search of the common target genes of FPR1 and Smta-1 revealed 16 orthologous pairs (P,0.0001) ( Table 5 and Table S11). A total of 57 orthologous pairs were also found when searching for common target genes of FMR1 and SmtA-1 (P,0.0001) ( Table 5 and Table S12). Only one gene (NCU03727, ham-2), the ortholog of Pa_ 2_9440, was reported to be involved in sexual reproduction in N. crassa [53]. The ham-2 mutants are unable to differentiate female organs but male fertility is not affected. The transcriptional profiles of each group of orthologous genes were compared as described in [54]. The calculation of correlation based on the FC values of orthologous pairs reveal no shared transcriptional patterns for FPR1/Smta-1, nor for FMR1/SmtA-1 target genes (Table 5).

Mating-type transcription factors FMR1 and FPR1 control a large number of target genes not directly involved in mating
The fertilization-competent mycelium analysis described in this study provides the first global view of differential gene expression in strains of opposite mating type in Pezizomycotina. These analyses revealed that the sexual differentiation correlates with the differential transcription of many genes which are not directly involved in sexual reproduction. Genome-wide transcriptomic analyses were conducted in S. cerevisiae and S. pombe to determine differentially transcribed genes in cells of different mating types. Six a-specific genes and five a-specific genes were identified in S. cerevisiae [1] ( Figure 5) and with the possible exception of one a-specific gene, all these genes are directly involved in some aspect of mating. In S. pombe, Mata and Bähler identified 12 M-specific genes and 4 P-specific genes [2] ( Figure 5). Three of these cell type-specific genes were not essential for mating. In P. anserina, 157 genes were identified that were differentially transcribed in mat+ vs mat2 strains and 10 genes were identified that were up-or down-regulated to a similar level in both strains. Thus, in P. anserina, the number of mating-type controlled genes (167) greatly outnumbers that in yeasts. In contrast with the cell-type specific genes in fission and budding yeasts, many P. anserina mating-type target genes appeared to be involved in a variety of biological processes, including general metabolism and energy production. A total of 32 genes were selected for deletion, including a few genes expressed in a mating-type specific manner. Only the genes encoding the pheromone receptors PRE1 (Pa_7_9070) and PRE2 (Pa_4_1380) were found to play a role in sexual reproduction, and one gene was found to be essential for viability (Pa_4_7760, farnesyl transferase b, RAM1). The deletion of the 29 other genes did not reveal any obvious defect in growth and/or mating. These 29 genes included three transcription factors (Pa_2_6830, Pa_4_9520 and Pa_6_3770), which were apparently unrelated to mating. Furthermore, DPa_2_6830 was genetically associated with either DPa_4_9520 or DPa_6_3770, and both double mutants behaved as the wild-type strain (data not shown). Although systematic deletions were not performed, these results suggested that the 167 mating-type target genes identified included very few candidates essential for fertilization or subsequent developmental steps. Currently, only pheromones [13] and pheromone receptor genes have been identified as mating-type target genes essential for mating in P. anserina.
Two hypotheses can account for the surprisingly high number of mating type-controlled genes not directly involved in mating.  The first is based on the observation that P. anserina is a pseudohomothallic (or secondary homothallic) species: it is encountered in nature as a self-fertile mycelium containing mat+ and mat2 nuclei. The optimal balance of the two types of nuclei for mycelium self-fertility may require that each type of nucleus controls complementary metabolic pathways. Although no differences in growth rate have been reported between homokaryotic and self-fertile mycelia in P. anserina in laboratory conditions, self-fertile mycelia may be favored in nature when less easily digestible nutrients are available. The second, and not exclusive, hypothesis is based on the idea that perithecium development relies on nutrients supplied by the mycelium [33,55,56]. The mating-type genes, which control perithecium development as well as mating, may be involved in the mobilization of nutrients necessary for the developing perithecia.
In particular, the differences observed in the transcription profiles of mat+ and mat2 strains concern the metabolism of lipids whose role during sexual reproduction has been emphasized (reviewed in [57], [58]); for example, three genes encoding esterase-lipase are up-regulated in the mat+ strain. Furthermore, six additional mat+ up-regulated genes involved in lipid metabolism were identified, not including those necessary for processing of the lipophilic pheromone MFP. The lipid reserves may be metabolized during vegetative growth to release fatty acids and other compounds which could be used as nutrients to provide carbon skeletons and metabolic energy during sexual reproduction [56]. Although there are many obvious transcriptional differences between mat+ and mat2 strains, no conspicuous differences have ever been noticed during sexual reproductions, besides cell-type specificity. We cannot exclude that some differences do exist between mat+ and mat2 female strains during ascogonium formation and initial development after fertilization, although subsequent developmental stages in the maturing fruiting bodies converge in mat+ and mat2 female strains due to the complementary information brought by the male nucleus.

Mating-type transcription factors FMR1 and FPR1 have a dual activating and repressing regulatory activity
Our analysis in P. anserina revealed a very complex regulatory pattern, which uses all eight possible combinations of regulation (activation and repression) by FMR1 and/or FPR1 ( Figure 5). This pattern indicated that each mating-type protein has activator and repressor activity, and that either one or both mating-type proteins may control each target gene. The complex regulation exerted by mating-type genes has been anticipated by previous genetic analyses, which predicted that FPR1 acts as an activator for mat+ function and as a repressor for mat2 functions. FMR1 was predicted to have a symmetric but inverse activity to FPR1 (see Introduction and Figure 1). Subsequent genetic studies indicated that the repressor activity of each transcription factor may operate at the level of post-transcriptional pheromone maturation ( [13] and reviewed in [3]). In agreement with this prediction, our study indicated that the protein-S-isoprenylcysteine O-methyltransferase (Pa_7_9690), which is required for the maturation of the lipophilic pheromone in yeast [48], is activated by FPR1 and repressed by FMR1 (see Results and Table S7, class 8). Pa_5_11640 (ABC transporter, STE6) could be a good candidate for a similar type of regulation, but this requires confirmation because the p-values were above the threshold. Unfortunately, this study did not succeed in identifying any gene that could be activated by FMR1 and repressed by FPR1 and was required for hydrophilic pheromone maturation.
The dual action (repression and activation) described in this study for the P. anserina mating-type genes is also conserved in budding yeast ( Figure 5) (reviewed in [59]) and in G. moniliformis [10]. The fact that the Pc and Mc proteins of S. pombe have been reported to use only gene inductions [2] may be due to experimental bias. Mata and Bä hler have compared the transcriptome of M and P vegetative cells which overexpress Ste11p relative to corresponding M and P control cells. This enabled the authors to identify activated target genes, but the experimental design could not reveal target genes that were tightly repressed in the control cells since these genes could not be repressed further in cells overexpressing Ste11p. It is therefore possible that M-specific genes are activated in M cells and repressed in P cells in S. pombe, whereas P-specific genes are subject to a symmetrical but reversed action in M cells ( Figure 5). The dual action of mating-type genes (activation and repression) may be a conserved feature in Ascomycota, although both types of control are operated by two mating-type proteins in yeast and only one in P. anserina and G. moniliformis. Further analyses in S. pombe and dissection of mating-type regulatory circuits in other heterothallic Ascomycota will help to test this hypothesis.
Interspecies comparisons of mating-type regulated genes revealed conserved genes but no shared transcriptional pattern The search for common target genes of the mating-type transcription factors revealed a highly significant number of orthologous pairs in the P. anserina/G. moniliformis and P. anserina/S. macrospora comparisons. This result was unexpected for SMTA-1 and FMR1, because SMTA-1 is not essential for fertilization and fruiting-body development in S. macrospora [7], in contrast to FMR1 in P. anserina [24]. However, SMTA-1 is a positive regulator of pheromone gene transcription, although it is not essential for their expression [7]. This observation suggests that SMTA-1 may be involved in the control of the same genes as essential mating-type transcription factors. Orthologous pairs of genes, which are essential for mating, have typically similar expression patterns across species. For instance, the transcription of genes encoding lipophilic pheromones is activated by MATA_HMG transcription factor FPR1 in P. anserina and mat a-1 in N. crassa [17]. Similarly, the genes encoding hydrophilic pheromones are activated by MATa_HMG transcription factor FMR1 in P. anserina and mat A-1 in N. crassa [17]. In contrast, correlation analysis of FC values in orthologous pairs revealed that transcriptional profiles were not maintained among the numerous conserved mating-type target genes that were not involved in mating. One possible explanation to this paradoxical observation is that, for various combinations of genes, different transcriptional profiles might result in the same metabolic effect. This effect may be the mobilization of nutrients during the vegetative phase to fulfil the formation of the fruiting bodies, as proposed above. Another possible explanation is that there is a selection for the conservation of a binding site for a mating-type transcription factor at a conserved chromosomal locus, rather than selection for the function of the target genes. The inter-species variations of transcription patterns would thus reflect random variations in the expression of target genes. Conservation of these binding sites around diverse genes may be required for spatial and temporal sequestration of transcription factors, thereby contributing to the control of the expression of these regulatory proteins.

Conclusions
Our study demonstrated that mating-type genes control many more target genes than those that are directly required for fertilization. Dissection of the regulatory circuits revealed all eight combinations that are possible with two transcription regulators exerting dual control (repression or induction). It must be noted that these experiments were conducted in homokaryotic strains. The control in a heterokaryotic mat+/mat2 self-fertile strain cannot be inferred from our results, as competition of FMR1 and FPR1 for binding sites may impact the transcription of target genes under dual control. Moreover, variation in the nuclear balance during the development of the heterokaryotic mycelium [60] may result in an additional layer of complexity to the temporal expression of target genes. The role of the mating-type target genes in a biological process related to sexual reproduction is still elusive and its elucidation will require both biochemical analyses of the metabolic pathways involving these genes and ChIP experiments to determine if they are primary or secondary target genes of FMR1 and FPR1. Mating-type regulatory proteins also play essential roles in the development of the ascogenous hyphae in the fruiting body, in association with the other matingtype proteins (reviewed in [33]). Further investigations will be necessary to determine how the switch from vegetative to reproductive tissues modifies the set of target genes that are controlled by FMR1 and FPR1.

Strains
The genetic and biological features of P. anserina were first described by Rizet and Engelmann [61] 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 [62], which was used to determine the P. anserina genome sequence [63]. The fmr1 2 and fpr1 2 mutant strains used for transcriptomic profiling have been described previously [44]. Successive nonsense mutations were introduced at the beginning of FMR1 and FPR1 coding sequences, resulting in loss-of-function mutations. Deletion of genes identified in transcriptomic profiling was performed by targeted gene replacement in a strain defective for PaKU70, a gene necessary for the non-homologous end joining recombination [44].

Preparation of biological samples for RNA extraction
Vegetative cultures. Petri dishes containing minimal medium were covered with a cellophane sheet (cat#1650193, Bio Rad, Hercules, USA) and were inoculated with nine implants from P. anserina mat+ or mat2 cultures. The dishes were placed at 27uC under constant light (0 h) and were removed from the incubation room at different culture time points (24 h, 48 h, 72 h, 96 h and 120 h). With a growth rate of 7 mm/day, the mycelium reached confluence at 48 h. Mycelia were harvested by scraping with a microscope coverglass and excess water was removed with Precision Wipes (Kimberly-Clark, Reigate, United Kingdom). The mycelium was then stored in aliquots of no more than 100 mg in liquid nitrogen until RNA extraction. Quantitative PCR experiments were performed with five mat+ plates and five mat2 plates for each time point. Transcriptomic profiling at 96 h was performed with four biological replicates for each time point, with each plate being considered as a biological replicate.
Fertilized cultures. Plates were prepared as indicated above, except that the medium was covered with cheesecloth (Sefar Nitex 03-48/31, Sefar, Heiden, Swiss) instead of cellophane, which was completely degraded by P. anserina after 120 h. Crossing was performed at 96 h by spermatization: each mat+ plate was washed with 2 ml of water to harvest the mat+ spermatia, and the suspension was collected and poured onto a mat2 plate to fertilize the mat2 female organs. After applying the mat+ spermatia suspension onto the mat2 plate, the suspension, which contained the mat2 spermatia, was collected and poured back onto the mat+ plate to fertilize the mat+ female organs. Mycelium and perithecia were harvested by scraping the cheesecloth with a microscope coverglass. Quantitative PCR experiments were performed with five mat+ plates and five mat2 plates. Each plate was considered to be a biological replicate.

RNA extraction
A maximum of 100 mg of mycelium was recovered from liquid nitrogen and ground in a Mikro-Dismembrator (Sartorius, Goettingen, Germany) for 1 min at 2600 rpm in vessels containing 9 mm chromium steel grinding balls frozen in liquid nitrogen. The resulting powder was suspended in RLT buffer (700 ml) (Qiagen, Hilden, Germany) and the suspension was passed through a Qiashreddrer column (Qiagen, Hilden, Germany). Total RNA was purified on RNeasy Plant Mini Kit columns (ref 74904, Qiagen, Hilden, Germany) following the manufacturer's protocol, which included DNase treatment. For microarray experiments, washing with RPE buffer (500 ml and 350 ml) was performed a further two times to improve RNA quality. The quantity and quality of the total RNA was determined using a NanoDrop ND-1000 spectrophotometer (Nanodrop Technologies, Wilmington, USA) and by electrophoresis on 1% agarose gel. For microarray experiments, a further control was performed on the Bioanalyzer 2100 system (Agilent, Santa Clara, USA) as described previously [64].

RT-qPCR analyses
Total RNA was reverse-transcribed with Superscript III (Invitrogen, Carlsbad, USA) and oligo(dT) 20 as recommended by the manufacturer. RT-qPCR amplification mixtures (10 ml) contained 16 FastStart Universal SYBR Green Master (Rox) (Roche Diagnostics, Mannheim, Germany), 2 ml of a 1/10 dilution of the reverse-transcription reaction and 500 nM of forward and reverse primers. Reactions were run on a CFX96 device (Bio Rad, Hercules, USA). The cycling conditions comprised 10 min polymerase activation at 95uC and 45 cycles at 95uC for 10 sec and 60uC for 30 sec with plate read, followed by a melting curve. Each plate contained cDNA samples representing all of the conditions examined for one gene, a standard curve of three serial dilutions points (1/10 dilutions) and a no-template control. All measures were performed with five biological replicates for each condition. For microarray validation, the biological replicates used for RT-qPCR were different from those used in microarray experiments. All of the PCR efficiencies were over 85%. At least one primer in each primer pair was positioned on two consecutive exons (Table S13), ensuring specific detection of cDNA with minimal interference by genomic DNA. MFM and MFP genes are intron-less genes and were detected with primers that do not discriminate cDNA from genomic DNA: therefore, the detection of their cDNA was not below the level of genomic DNA contamination. For these two genes, a non-reversetranscribed control was performed on each biological replicate under the same conditions as the reverse-transcribed samples to determine the level of genomic DNA contamination. Measurements were considered as significant when the genomic contamination was five cycles above the cDNA signal.
For each RT-qPCR analysis (time course and microarray validation), eight housekeeping genes (Table S13) were screened with geNorm [65] to identify the best reference genes for each type of experiment. The minimum number of reference genes was selected that had a pairwise variation (V) below 0.15. If two genes were sufficient to reach a value of 0.15 for V, three reference genes were used according to MIQE guidelines [66], provided that the addition of the third reference gene did not increase V above 0.15. Pa_7_8490 (ubiquitin), Pa_4_8980 (TATA-box binding protein), Pa_3_5110 (glyceraldehyde phosphate dehydrogenase) [67] and Pa_5_5390 (histone H2A) were used for normalization of the timecourse RT-qPCR (V = 0.15). Pa_7_8490 (ubiquitin), Pa_3_5110 (glyceraldehyde phosphate dehydrogenase) [67] and Pa_1_16650 (ribosomal protein AS1) [68] were used as reference genes for normalization of the microarray RT-qPCR validation (V = 0.031).

Labelling of cDNA, hybridization and microarray analyses
The construction and optimization of gene expression microarrays for P. anserina has been described previously [39]. Briefly, the microarrays consisted of a 4644 K platform (AMADID 018343, Agilent, Santa Clara, USA) containing 10,556 probes on each array with each probe present in four replicates.
For the transcriptome microarray experiments, target preparation, hybridization and washing were performed according to the two-color microarray-based gene expression analysis instructions (version 5.0, February 2007) as described by the manufacturer (Agilent, Santa Clara, USA). One-microgram aliquots of total RNA were amplified and Cy-labelled with Agilent's Low RNA input fluorescent linear amplification (LRILAK) PLUS kit and the Two-Color RNA Spike-in Kit (Agilent, Santa Clara, USA). The labelling efficiency and the product integrity were checked as described previously [64].
Four biological replicates labelled with Cy-3, for each of the different experimental conditions were compared with a common reference labelled with Cy-5, in indirect comparisons. The common reference was obtained by mixing RNA extracted from the different conditions as indicated in [39]. A total of 825 ng of each of the Cy3-and Cy5-labeled targets was mixed and incubated on microarray slides for 17 h at 65uC in a rotating oven (10 rpm) using an Agilent in situ hybridization kit (Agilent, Santa Clara, USA). The slides were washed and dried by centrifugation at 800 rpm for 1 min.
Microarrays were scanned using the Agilent DNA microarray Scanner (Agilent, Santa Clara, USA) at a resolution of 5 microns using the extended dynamic range (XDR) feature. Spot and background intensities were extracted with the Feature Extraction (FE, v9.5.3) software (Agilent, Santa Clara, USA) using the GE2-v4_95_Feb07 default protocol. Preliminary array quality was assessed through the use of Agilent control features as well as spike-in controls (Agilent 2-Color Spike-in Kit for RNA experiment). Subsequent flagging was done according to the GenePix Pro software (Molecular Devices Sunnyvale, USA) nomenclature, which included four flag levels (good [100], bad [2100], not found [250], moderate [0]). FE-software normalized data (Lowess normalized, local background subtracted) were processed with MAnGO [70]. A moderated t test with adjustment of p-values [71] was computed to measure the significance of each difference of expression. Genes were considered as differentially transcribed if their transcript levels showed a $2-fold change (FC$2) or #2-fold change (FC#22) with a p-value of ,0.005.

Microarray data accession number
All microarray data is MIAME compliant. The raw data has been deposited in the MIAME compliant Gene Expression Omnibus database [72] and is accessible through the GEO Series accession number GSE27297.
Selection and sequencing of genes from fpr1 2 and fmr1 2 strains Ten genes were selected in fpr1 2 and fmr1 2 strains to identify putative adventitious mutations that might have occurred during the introduction of site-directed mutations in FPR1 and FMR1 [44]. As the two mutant strains were generated independently, it is highly unlikely that they would harbor identical mutations. Therfore, it was assumed that an adventitious mutation in a given gene should affect its expression in either one of the two mutant strains, while its expression should be similar in the three other strains considered, namely the other mutant strain, the mat+ and the mat2 strains. Pa_1_23560, Pa_3_11210, Pa_5_4750, Pa_5_10440 and Pa_7_130 displayed differential transcription with highly significant p-value in the fpr1 2 vs fmr1 2 and fpr1 2 vs mat+ comparisons but not in the mat+ vs mat2 and fmr1 2 vs mat2 comparisons (Table S2); thus these genes were selected as the best putative targets for adventitious mutations in the fpr1 2 strain. Based on a similar rationale, Pa_1_970, Pa_1_16110, Pa_2__4340, Pa_4_4670 and Pa_6_6260 were selected as possible targets for adventitious mutations in the fmr1 2 strain. The selected genes were amplified from the appropriate DNA with Dream Taq (Fermentas, St Leon-Rot, Germany) in accordance with the manufacturer's instructions using primers located 1000 bp upstream and 1000 bp downstream of the translation initiation codon. Each amplification product was sequenced with upstream and downstream primers and primers targeted to regions around the initiation codon. Sequences were assembled with CAP3 [73], which is available on http://pbil.univ-lyon1.fr/cap3.php, and were compared with wild-type sequences using BLAST [45].

Gene annotation and the search for orthology
Protein sequences were searched against the NCBI nonredundant protein sequence library using BLASTP [45]. FUNGIpath [74] was used to search for orthologous genes on http:// embg.igmors.u-psud.fr/fungipath/.

Gene deletion
The construction of deletion cassettes was based either on the AscI/MluI method or on the N. crassa strategy for high-throughput generation of gene deletions [75].
The AscI/MluI method is based on cloning an MluI fragment containing the resistance selection gene in an AscI linearized plasmid containing the 39 and 59 flanking regions of the gene that was selected for deletion. This plasmid was obtained by the amplification of a vector containing the selected gene with primers targeted to the 59 and 39 borders of the coding sequence to be deleted and with their 39 ends oriented towards the 59 and 39 untranslated regions. These diverging primers were designed with an AscI site at their 59 end in order to amplify the vector in a linearized form, ending with an AscI site and lacking the coding sequence of the selected gene. The selection gene in an MluI fragment was obtained by amplification of the hph gene from the pBCHygM vector with the primer pair, 5_CPC1_Mlu (59gcgacgcgtccgagatgcgccgcgtgc-39) and 3_Ttrpc_Mlu (59-gtcacgcgtagaggatcctctagcta-39), followed by digestion with MluI. The pBCHygM vector is a derivative of pBC-hygro [76], which has lost the MluI site in the trpC terminator.
The N. crassa strategy has been previously described [75] and included two modifications aimed at minimizing errors in the 59 and 39 flanking regions and the hph cassette obtained from pMOcosX [77]. First, the high-fidelity Pfu polymerase (Promega, Madison, USA) was used instead of the low-fidelity LA Taq polymerase (TaKaRa BIO Inc, Shiga, Japan) and the cycle conditions were adapted to minimize amplification errors. If amplifications with Pfu on 20 cycles failed, 10 additional cycles were performed, and if 30 cycles were not successful, the LA Taq polymerase was used according to the manufacturer's instructions. The second modification involved the addition of an AscI site at each end of the deletion cassette. These sites were used to linearize and separate the deletion cassette from the vector prepared by amplification in Escherichia coli prior to the transformation of P. anserina.
Linearized DNA containing the 39 and 59 untranslated regions of the gene to be deleted and the hph selection gene [77], was used for the transformation of the P. anserina DPaKU70 strain as described in http://podospora.igmors.u-psud.fr/methods.php# transfo. In each mutant strain, accurate gene replacement was checked by Southern blot analysis and one transformant containing the correct deletion was crossed with a wild-type strain to generate both mat+ and mat2 mutant strains lacking the DPaKU70 deletion.

Phenotype analyses
For each deletion, both mat+ and mat2 mutant strains were tested for fertility. The number of male cells active in the fertilization of a compatible wild-type tester strain (used as a female) was determined as previously described [13]. Crosses homozygous for each deletion were performed to search for possible defects in the development of the fertilized female organs. The capacity of mat+/mat2 heterokaryotic deletion strains to grow and reproduce on different carbon sources was tested as described in [78]. Growth and fertility of mat+, mat2 and mat+/mat2 deletion strains were tested on minimal media containing inhibitors that act on different cell compartments: nikomycine (0.1 mg/ml), which acts on the cell wall by inhibiting chitin synthase; fludioxonil (0.01 mg/ml), which acts on osmoregulation; sodium dodecylsulfate (0.03%) which acts on the cell membrane; and caffeine which induces pleiotropic effects. The concentration of each inhibitor caused an approximately 50% decrease in the wild-type growth rate.

Calculation of interspecies correlation
FUNGIpath [74] was used to identify systemically orthologous groups of genes for pairwise comparisons between P. anserina FPR1 and G. moniliformis MAT1-2-1 target genes (Table S10), P. anserina FPR1 and S. macrospora Smta-1 target genes (Table S11), and P. anserina FMR1 and S. macrospora SmtA-1 target genes (Table S12). The cut-off for selecting differentially transcribed genes in P. anserina comparisons was lowered to $1.5-fold change (FC$1.5) and a #1.5-fold change (FC#21.5) with a p-value of ,0.005. Pvalues for the overlap of gene sets between P. anserina and G. moniliformis or S. macrospora (Table 5) were determined from a 262 contingency table using a Chi-square test with Yates correction run on http://www.graphpad.com/quickcalcs/contingency1.cfm (GraphPad Software Inc, La Jolla, USA). The contingency table compared the number of orthologs (group 1, outcome 1) identified in the total number of target genes (group 1, outcome 2) of a given mating-type transcription factor (G. moniliformis MAT1-2-1, Smta-1 or SmtA-1) with the number of targets genes (group 2, outcome 1) identified for the corresponding transcription factor in the 10,556 genes of P. anserina (group 2, outcome 2). Calculation of interspecies correlation (Table 5) were performed as described in [54]

Table S2
Whole genome transcriptomic comparisons of wildtype and mutant strains. Fold change (FC), adjusted p-value (pvalue) and the binary logarithm of the arithmetic mean of compared intensities (A) are indicated for the 10,556 P. anserina genes and for each comparison (mat+ vs mat2, fpr1 2 vs fmr1 2 , mat+ vs fpr1 2 and mat2 vs fmr1 2 ). Genes with FC values above cut-off and significant p-values are marked by ''1'' in the ''selection'' column for each comparison. FC cut-offs: FC#22 or FC$2. pvalue cut-off: P,0.005. (XLS)

Table S7
Partitioning of the mat+ vs mat2 differentially transcribed genes according to their FC value in wild-type to mutant comparisons. Genes with positive FCs (FC$2, P,0.005) in the mat+ vs fpr1 2 and mat2 vs fmr1 2 comparisons were considered as being activated (A) by FPR1 and FMR1, respectively. Genes with negative FCs (FC#22, P,0.005) in the mat+ vs fpr1 2 and mat2 vs fmr1 2 comparisons were considered as being repressed (R) by FPR1 and FMR1, respectively. (XLS)