The genomes of A. benhamiae and T. verrucosum were sequenced by a whole-genome shotgun hybrid approach. The assembly of A. benhamiae spans 22.3 Mb [DDBJ/EMBL/GenBank:ABSU00000000] and that of T. verrucosum comprises 22.6 Mb [DDBJ/EMBL/GenBank:ACYE00000000] (Table 1; Additional file 2; both genomes are also deposited in the Broad Institute database 18 ). Thus, these genomes are smaller than those of phylogenetically related ascomycetes, such as aspergilli (28 Mb and 37.3 Mb in case of Aspergillus clavatus and Aspergillus niger, respectively), Coccidioides species (27 to 29 Mb), and Histoplasma capsulatum (30 to 39 Mb).

The genomes of A. benhamiae and T. verrucosum contain 7,980 and 8,024 predicted protein-encoding genes, respectively (Table 1). Introns were found in 5,809 of the A. benhamiae and 5,744 of the T. verrucosum genes. Both genomes comprise a mosaic of long G + C rich, gene-containing portions separated by A + T rich 'islands' with a GC content below 40%, ranging from a few kilobases to more than 25 kb. As expected from previous reports based on nuclear ribosomal internal transcribed spacer regions 1 and 2 15 19 20 21 , the comparison of the two genome sequences revealed a strong similarity. Using the software Mummer 22 , approximately 21.8 Mb of the genomes (98.0% of the available A. benhamiae and 96.7% of the T. verrucosum genomic sequences) can be aligned to each other, indicating that the vast majority of genes lie in collinear regions and are shared between the two organisms. The average identity of the alignable portion of the genomes is 94.8%. The alignment of the two genomes points to only five major genomic rearrangements, one inversion and four balanced translocations between chromosomes (Figure S1 in Additional file 2). The presence of only a few rearrangements between the two genomes suggests very recent speciation. These findings are reflected by the phylogenetic tree constructed by use of the available genome sequences (Figure 2; Figure S2 in Additional file 3).

However, we also identified notable dissimilarities between the genomes of A. benhamiae and T. verrucosum. After having detected the orthologous pairs with best bidirectional hits, we came up with lists of proteins that presumably were unique for either species. Since the best bidirectional hits were identified using protein Blast, we next applied BlastN to correct for possible gene prediction errors. We used a filter threshold for significant hits of 80% identity between sequences over less than 50% of the query length. There were 238 A. benhamiae sequences that gave no hits or non-significant hits in T. verrucosum, and 219 T. verrucosum genes were not found in A. benhamiae. Of these, 83 and 78 genes (A. benhamiae and T. verrucosum, respectively) have assigned names and/or functional domains. A list of the predictions is provided in Additional file 4. Given the overall strong genome sequence similarity, a future functional investigation of these distinctions appears to be of interest, in particular with respect to the tremendous differences between the two species in terms of in vitro growth ability and animal host preference (see also the 'Other interesting genes' section).

We analyzed the A. benhamiae fast-evolving genes in comparison to T. verrucosum. Using the dN/dS ratio as a measure for selective pressure, we obtained a list of positively selected genes (dN/dS >1) (Additional file 5). In total we found 132 positively selected genes with assigned functions, enabling assumptions about their roles in the cell and, hence, the reasons for their accelerated evolution. Of particular interest are the two most abundant groups of these genes, those encoding transcription factors (18 genes) and MFS transporters (5 genes). The latter are known to be usual constituents of secondary metabolite (SM) gene clusters.

Both dermatophyte genomes encode the basic metabolic machinery for glycolysis, tricarboxylic acid cycle, glyoxylate cycle, pentose phosphate shunt, and synthesis of all 20 standard amino acids and the five nucleic acid bases. Moreover, dermatophytes appear to be capable of producing a wide range of SMs, which is reflected by the presence of polyketide synthase (PKS)- and non-ribosomal peptide synthetase (NRPS)-encoding genes (see the 'Genetic basis for secondary metabolism gene clusters' section). The outstanding ability of dermatophytes to specifically infect superficial host structures may be supported by the possession of a broad repertoire of genes encoding hydrolytic enzymes, the expression of many of which was also proven experimentally (see the next paragraph and the 'Identification of secreted fungal proteins during keratin degradation by secretome analysis' section). In addition, the ability of dermatophytes to assimilate lipids, major constituents of the skin, is putatively reflected by the presence of 16 lipase genes in either genome. A putative link between the possession of lipases and fungus-induced skin disease has previously been revealed for basidiomycetes of the genus Malassezia 23 .

Of particular note is the apparent relative paucity of tRNA genes in both dermatophytes in comparison with other closely related ascomycetes. The genomes of A. benhamiae and T. verrucosum contain 80 and 77 tRNA genes, respectively, whereas the number of tRNA genes varies between approximately 100 to 130 in Coccidioides species and 150 to 370 in aspergilli. However, some strains of H. capsulatum, representing a comparatively closely related pathogen, also possess only 83 to 89 tRNA genes, suggesting that the low number of tRNA genes is not specific to dermatophytes.

Dermatophytes are keratinophilic fungi, sharing the ability to utilize compact hard keratin as a sole source of carbon and nitrogen. In line with this knowledge, the two sequenced genomes reflect a remarkable metabolic capability for protein degradation. They contain 235 predicted protease-encoding genes, 87 of the deduced proteins possessing a secretion signal (Table S3 in Additional file 6). We did not detect any protease in A. benhamiae or T. verrucosum unique to either species, a finding that may reflect similar life styles and/or host adaptation mechanisms, especially with respect to their common keratinophilic growth. In general, deviations in the number of proteases per genome are rather large in the fungal kingdom, ranging from approximately 90 in Ustilago maydis to approximately 350 in Gibberella zeae (according to the MEROPS database 24 ). Dermatophytes belong to the most protease-rich species.

The protein sequence of each protease is highly conserved across dermatophyte species 25 . Collections of predicted secreted proteases of A. benhamiae and T. verrucosum as well as Coccidioides spp. (Onygenales) were compared to those of A. fumigatus as a member of the Eurotiales, for which many secreted proteases have previously been characterized. Most A. fumigatus proteases in A1 (pepsins), M28 (leucine aminopeptidases), S9 (dipeptidylpeptidases), S10 (carboxypeptidases) and S53 (tripeptidylpeptidases) families have an orthologue in dermatophytes and Coccidioides spp. (Table S4 in Additional file 7). The major striking differences found between the secreted protease batteries of A. fumigatus and Onygenales are the following: subtilisin (S8), deuterolysin (M35), and fungalysin (M36), which belong to endoprotease gene families, have expanded in Onygenales (Table S4 in Additional file 7); the same is true for exopeptidases of the M14 family (metallocarboxypeptidases) and the M28 family (aminopeptidases) - a major carboxypeptidase (McpA) homologous to the human pancreatic carboxypeptidase A was previously characterized in dermatophytes 26 , and of particular note, Aspergillus spp. have no McpA orthologue; and genes encoding acidic glutamic proteases (G1 family) were not detected in either dermatophytes or Coccidioides spp.

Major differences between dermatophytes and Coccidioides spp. proteases were found in M35, M36 and S8 proteases families (see the phylogenetic trees in Additional file 8). Proteases of these three families of dermatophytes and Coccidioides spp. form distinct clades in phylogenetic trees (Additional file 8). Members of the S8 and M36 families have undergone additional amplifications in the dermatophyte lineage, and expansion of the M35 family appears to be different in Coccidioides spp. and dermatophytes. In the latter, a clade was apparently lost. In addition, three genes encoding proteases of the S41 family were found in the dermatophyte genomes while only one gene encoding a protease of this family was identified in Coccidioides spp.

Recent comparative genomic analyses of Coccidioides species with other members of the Onygenales showed gene family sizes are associated with a host/substrate shift from plants to animals in these microorganisms 27 . Experimentally, the expression of genes encoding fungalysins and subtilisins was recently monitored in A. benhamiae by cDNA microarray analysis during growth on keratin, and also during cutaneous infection of guinea pigs 8 . Interestingly, the prominent keratin induced A. benhamiae subtilisin-encoding genes, such as SUB3 and SUB4, were not observed in this former analysis to be strongly activated in vivo, in contrast to others that conversely were not found to be induced during in vitro growth on keratin. A role for Sub3 was recently observed in adhesion of the dermatophyte Microsporum canis to feline epidermis, but not for the invasion thereof 28 . These findings suggest additional functions of secreted proteases during host adaptation other than keratin degradation. Since the formerly used cDNA microarray does not comprise the full genome of A. benhamiae, the future identification of in vivo specific dermatophyte proteases on the basis of the presented genome appears to be of major interest.

A potential role of secreted proteases, in particular serine proteases, in pathogenesis has been widely reported in many prokaryotes and fungi 2 29 30 31 , including functions as allergens 32 . In order to apply insights from the present genome sequences to determine putative virulence gene function, we set out to reveal the basic panel of factors that are secreted during growth of A. benhamiae on keratin. To achieve this, secretome analysis was performed, an approach that, to our knowledge, has not been applied to A. benhamiae before. Experimental analysis (after 2 days of growth) led to the identification of 203 single electrophoretic species (Figure 3b). From these entities, 53 different proteins were detected (Table S5 in Additional file 6). By far the largest group of identified proteins is formed by putative proteases (approximately 75% relative spot volume). In addition, we found other, different hydrolases and proteins involved in carbohydrate metabolism (Table S5 in Additional file 6). Three of the subtilisin-like serine proteases (Sub3, Sub4, and Sub7), three fungalysine-type metalloproteases (Mep1, Mep3, and Mep4), the leucine aminopeptidases Lap1 and Lap2, as well as the dipeptidyl-peptidases DppIV and DppV were detected in the secretome, consistent with gene expression analysis in A. benhamiae during keratin degradation 8 . Supporting our results, the pattern of proteins secreted by the two related dermatophyte species Trichophyton rubrum and T. violaceum during growth on soy protein was previously described in 4 . In that study, a gel-based approach led to the identification of 19 proteins secreted by at least one of these species. Remarkably, 15 of the corresponding homologs were also found to be secreted in the present study by A. benhamiae on keratin medium, including major keratinases of the subtilisin family of secreted proteases (also see Table S6 in Additional file 6). Individual differences between the present and formerly observed secretion patterns might be due to the different dermatophyte species analyzed and/or to the different protein substrates and cultivation parameters used. In conclusion, the set of dermatophyte secreted proteases in a protein medium is similar to that of A. fumigatus, which includes endoproteases such as the major subtilisin Alp1 and the fungalysin Mep and exoproteases such as Lap1, Lap2, DppIV and DppV.

Endo- and exoproteases secreted by microorganisms cooperate very efficiently in protein digestion to produce oligopeptides and free amino acids that can be incoporated via transporters. During the process of protein digestion the main function of endoproteases is to produce a large number of free end peptides on which exoproteases may act. At neutral and alkaline pH, synergistic action of Lap and DppIV was shown in Aspergillus spp. 13 24 . Laps degrade peptides from the amino terminus until reaching an X-Pro sequence, which acts as a stop. In a complementary manner, the X-Pro sequences can be removed by DppIV, thus allowing Laps access to the next residue. Dermatophyte and Aspergillus spp. Lap1, Lap2, DppIV and DppV have shown comparable substrate specificity 33 . Therefore, our proteomics approach allows us to hypothesize common basic mechanisms in dermatophytes during extracellular protein digestion. However, the presence of large protease gene families in dermatophytes reflects selection during evolution and the ability of these fungi to adapt to different environmental conditions during infection and saprophytic growth.

Growth of A. benhamiae on keratin might mimic selected in vivo growth substrates, yet may not reflect the entire process of infection. In order to gain more insights into basic host adaptation mechanisms, we studied the global transcriptional response of A. benhamiae during infection of human keratinocytes. After 12 h of co-cultivation, germinating A. benhamiae microconidia were observed to be localized and concentrated on the host cells, suggesting that the fungus actively adheres to the keratinocytes (Figure 1c,d). To perform 454 RNA sequencing, the fungal cells were harvested after incubation for 96 h with and without keratinocytes. About 50 A. benhamiae genes showed differential expression with a fold change >5 (P-value < 0.05; Table S7a in Additional file 6); 45 genes encoding putatively secreted proteins (Table S5 in Additional file 6) and 13 genes coding for proteins involved in the biosynthesis of SMs are expressed either only with or without keratinocytes, or under both conditions. Of the 235 predicted protease-encoding genes, 158 are expressed under both conditions. Sixteen potentially secreted proteins, including three proteases, are differentially expressed (Table S7b in Additional file 6). In particular, the expression profile of the genes encoding carboxypeptidase S1 and dipeptidyl-peptidase DppV implies their potential involvement in the infection process. The transcript levels of two NRPS genes were reduced during co-cultivation with keratinocytes, a finding that is noticeable but cannot be explained at this stage.

To confirm the RNAseq results, we selected several genes that were predicted to be differentially expressed and tested them by Northern blotting. We used two housekeeping genes, actin (ARB_04092) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH, ARB_00831), as controls as they are not expected to be differentially regulated between the control and co-incubation conditions. All tested genes were regulated as expected from the RNAseq data (Figure S4 in Additional file 9). The expression level alterations of metabolic enzymes (ARB_07891, ARB_04156, ARB_01650 and ARB_04856) and membrane transporters (ARB_01027) reflect the adaptation of the fungus to the different nutrition provided by keratinocytes and their remnants, whereas the strong up-regulation of the hydrophobin ARB_06975 indicates altered binding properties and adhesivity during growth on epithelial cells and during infection. In conclusion, this independent experimental method shows that the accuracy of the RNAseq data was exemplary.

The A. benhamiae and T. verrucosum genomes encode a relatively high number (26 and 25, respectively) of SM biosynthesis gene clusters (Table 2), a finding that contrasts with observations made in other fungi and bacteria highly adapted to humans. For comparison, Candida albicans or Staphylococcus aureus hardly produce SMs and Histoplasma species have no more than seven SM gene clusters per genome; more closely related to dermatophytes is Coccidioides immitis, which has 16 SM gene clusters, the main difference being in the number of NRPSs (5 versus 15 in A. benhamiae). Nine PKS, 15 NRPS and 3 PKS/NRPS hybrid genes were identified in the A. benhamiae genome, all of which except for one NRPS gene (ARB_02149) are conserved in both species (Table 2). Addressing the question of whether the absence of the latter gene in T. verrucosum is associated with phenotypic and/or host-specific differences between the two species will be of future interest. To see whether only the NRPS or the entire associated gene cluster is absent from T. verrucosum, we examined the conservation of the other constituents of the ARB_02149 gene cluster and observed that the 'missing' NRPS belongs to an otherwise very well conserved and collinear region that spans more than 75 kb (the whole T. verrucosum supercontig 79). However, one other gene besides ARB_02149 is missing in T. verrucosum, the MFS transporter ARB_02151 (Figure 4). Interestingly, the 'missing' genes are separated by a perfectly conserved ABC multidrug transporter (ARB_02150 = TRV_01489). The Arthroderma ARB_02149 gene cluster has several traits typical of functional SM gene clusters, such as the presence of genes for the MFS transporter, feruloyl esterase and C6 transcription factor. This makes us suppose that the NRPS was lost in Trichophyton rather than acquired by Arthroderma. However, it remains unclear if the MFS transporter was deleted simultaneously, and why the deletion did not capture the 'middle' ARB_02150 gene.

All nine PKS genes detected in A. benhamiae have unequivocal counterparts in the T. verrucosum genome (Table 2). An interesting feature of the dermatophyte PKS set is the unusual proportions of reducing and non-reducing PKSs. Whereas in all other closely related ascomycetes (such as aspergilli) most of the PKSs are non-reducing, in dermatophytes most are reducing PKSs. A comparison with the closest sequenced relative, C. immitis (Table 2; see more details below), also revealed substantial differences in the composition of the PKS set: the ratio of reducing to non-reducing in dermatophytes is 2:1, whereas in C. immitis it is 2:3. This observation suggests dermatophytes have an uncommon SM profile, which deserves future investigation. Particular attention should be paid to the fact that these fungi are characterized by intense pigmentation, a phenotype that may be related to their pathogenicity. For the related species T. rubrum, the polyketide-derived mycotoxin xanthomegnin has been suggested to be responsible for the characteristic red colony reverse pigment. Most interestingly, xanthomegnin production has even been detected in epidermal material infected by T. rubrum, in contrast to non-infected controls 34 . A putative link between SM production and host adaptation of A. benhamiae might also be reflected by our observation that several genes associated with the synthesis of such molecules were found to be differentially regulated during infection of human keratinocytes (see the 'Differential gene expression in A. benhamiae during infection of keratinocytes' section).

To get an impression of possible expansions of families and evolutionary relationships, we compared the sets of SM producers in dermatophytes with that of C. immitis (Table 2; Figure S5.1 and S5.2 in Additional file 10). As mentioned above, the total number of SM gene clusters is higher in dermatophytes, mainly due to the more abundant NRPSs. However, we observe differences also in the PKS set as well as in the number of PKS/NRPS hybrids: C. immitis possesses only one hybrid, whereas each dermatophyte has three. The higher number of non-reducing PKSs in C. immitis is mainly due to the expansion of one clade; most likely we are seeing the results of duplication of some ancestor genes with a domain architecture of a beta-ketoacyl synthase domain, an acetyltransferase domain, an acyl carrier protein domain, and a methyltransferase domain (KS-AT-ACP-ME). Four of six C. immitis non-reducing PKSs belong to this clade. Of the other two, one has a clear ortholog in dermatophytes, and the other has an unusual structure (AT-KS-ACP-thioesterase domain (TE)) without an orthologous dermatophyte gene. In comparison to C. immitis, dermatophytes possess two additional non-reducing clades, which means that, in spite of the lower number of non-reducing PKSs, they have more various potential capacities. The reducing C. immitis PKSs also cannot boast great variety: two of four C. immitis genes are most likely the result of a duplication (they form a separate clade and do not have dermatophyte orthologs), one PKS has orthologs in dermatophytes, and one is only a probable homolog (see below). On the other hand, in dermatophytes we see an expansion of the group with a fumonisin synthase-like structure (KS-AT-ME-enoyl reductase domain (ER)-ketoacyl reductase domain (KR)-ACP): three orthologous pairs formed by out-paralogs in each species have only one close homolog in C. immitis. Since the C. immitis gene lacks one of the domains (methyltransferase), we cannot consider it as a fumonisin-like ortholog. Besides the 6-methyl-salicylic acid synthase, completely lacking in C. immitis, another not completely reducing PKS (KS-AT-ME-KR-ACP), as well as two PKS/NRPS hybrids, do not have homologs in C. immitis. Taken together, these data agree with our hypothesis that highly adapted parasites such as Coccidioides do not require a large arsenal of SMs.

Sexual reproduction is known for A. benhamiae but not for T. verrucosum 35 36 . The A. benhamiae and T. verrucosum genomes revealed the whole sets of genes for mating and meiosis in both species, suggesting that the lack of a known sexual cycle in T. verrucosum is not due to major deletions of genes essential for sexual reproduction and meiosis (Table S8 in Additional file 6). Both sequenced strains showed a single mating type encoding an HMG box transcription factor. To identify the complementary mating type, we sequenced the corresponding region of an A. benhamiae mating partner strain (strain CBS 809.72; Figure 5). The newly identified region encodes an alpha-box type transcription factor, indicating that A. benhamiae exhibits two mating types, as described for other closely related fungal pathogens such as H. capsulatum and C. immitis 37 . A. benhamiae mating type + strains as well as mating type - strains are often routinely isolated 36 . There is no apparent disequilibrium between mating type + and mating type - strain frequencies.

We did not identify a striking defect in the T. verrucosum mating type locus, which appears to be intact. Several strains of T. verrucosum were found to be of the same mating type as the sequenced strains, suggesting a strong disequilibrium towards mating type +.

In Aspergillus (Eurotiales), Coccidioides and Histoplasma (Onygenales) the mating type (MAT) loci are flanked by APN2 and the SLA2 genes encoding a DNA lyase and a cytoskeleton protein, respectively 37 . The MAT idiomorphs and flanking regions described here for A. benhamiae and T. verrucosum are essentially identical to those of other closely related dermatophytes 38 .

Of particular interest are the genes of A. benhamiae that have no obvious counterpart in T. verrucosum (Additional file 4) and whose predicted functions suggest their potential involvement in basic biological phenotypes and/or pathogenicity. Two such genes, ARB_04713 and ARB_02149, encoding a phosphopantetheine-binding domain and an NRPS, respectively, were found in the transcriptome analysis, although not expressed differentially. The expression pattern of the A. benhamiae-specific NRPS ARB_02149 further suggests that its as yet unidentified product is produced during infection by the fungal cells.

Another gene of particular interest encodes hydrophobin. In A. fumigatus, surface hydrophobin was shown to prevent immune recognition 39 . The A. benhamiae hydrophobin gene (ARB_06975) shows 99% similarity with the respective T. verrucosum gene (TRV_00350) and displays moderate overexpression (1.6×) under co-cultivation conditions (Table S7b in Additional file 6). The analysis of a potential role of dermatophyte hydrophobins in immune response functions and/or adhesion to host surfaces will be part of future research.

A clinical isolate of A. benhamiae strain 2354 was used (isolate LAU2354) 15 . T. verrucosum strain 44 17 was kindly provided by Yvonne Gräser (Charité, Berlin, Germany). Strains were cultivated at 28°C on Sabouraud 2% (w/v) glucose agar (SG, Merck, Darmstadt, Germany) for 12 days; liquid cultures were shaken at 180 rpm at 30°C for 5 to 7 days. Hyphae and conidia were separated by filtration using a 40 μm cell strainer (BD Bioscience, Heidelberg, Germany). Conidia were counted with a cell counter (Beckman, Coulter, Krefeld, Germany) or manually using a Thoma chamber. For crossing experiments of A. benhamiae LAU2354 with the opposite mating type CBS 809.72, MAT medium 40 (1/10 SG, 0.1% (w/v) MgSO₄ and 0.1% (w/v) KH₂PO₄) was used.

For DNA preparation, mycelia were separated from the medium by filtration through Miracloth (Calbiochem, Darmstadt, Germany) and ground in a mortar under liquid nitrogen. After evaporation, the powder was suspended in a solution containing 150 mM EDTA, 50 mM Tris-HCl, pH 8.0, 1% (w/v) SDS, 20 mM NaCl and 100 μg/ml proteinase K (Merck). After incubation for 1 h at 55°C, the solution was gently mixed with 1/4 volume of 4 M NaCl and kept on ice for 30 minutes. After centrifugation for 10 minutes at 6,000 rpm and 4°C, polyethylene glycol 6000 (Serva, Heidelberg, Germany) was added to the supernatant to a final concentration of 10% (w/v). The DNA was precipitated for 1 h on ice and centrifuged for 10 minutes at 10,000 rpm at 4°C. The pellet was dissolved in a solution containing 25 mM Tris-HCl, pH 8.0, 5 mM EDTA, 10 mM NaCl and 1% (v/v) Triton X100. For density centrifugation, 1g CsCl and 12 μl bisbenzimide (10 mg/ml) for each milliliter of solution were added 41 42 . Ultracentrifugation was performed in a vertical rotor at 44,000 rpm for 24 h at 25°C. DNA was separated into two bands of different density according to the AT-content of the DNA. The upper band contained a DNA fraction highly enriched for mitochondrial DNA. For T. verrucosum, two rounds of density gradient centrifugation were necessary. In the first round, ethidium bromide was used instead of bisbenzimide. For RNA preparation, SG medium was inoculated with conidia to a final concentration of 3 × 10⁴ conidia/ml and shaken at 180 rpm for three days at 30°C. Total RNA was isolated using a commercial kit as described by the manufacturer (Qiagen, Hilden, Germany). After RNA extraction, a cDNA library was constructed from this material according to the manufacturer's protocols (MINT cDNA synthesis kit, Evrogen, Moscow, Russia).

Nuclear DNA of A. benhamiae and T. verrucosum was sheared, size fractionated (3 to 4 kb), end-repaired, and cloned into the SmaI site of pUC18. For both fungal species, two fosmid libraries each were prepared in pCC1FOS (Epicentre Biotechnologies, Madison, WI, USA) as described by the manufacturer, one for the high-GC chromosomal DNA fraction and one for the AT-rich mitochondrial DNA fraction. For T. verrucosum, 40,000 fosmids from GC-rich and 80,000 fosmids from AT-rich DNA were obtained. For A. benhamiae, the corresponding yields were around 50,000 (GC-rich) and 20,000 (AT-rich), respectively. End sequences of plasmid and fosmid clones were obtained using dye terminator chemistry and a 3730×l sequencer (Applied Biosystems, Foster City, CA, USA). Moreover, a fosmid library was generated with a GC-rich DNA fraction of the A. benhamiae strain CBS 809.72 encoding the opposite mating type locus. We tested 1,000 fosmids by colony filter hybridization and in PCR experiments. Fosmids were identified by hybridization experiments with a digoxygenin-labeled part of the apn2 gene (ARB_07318) and in PCR experiments using apn2 amplifying primers (5'-CTTCTAGTGACTCGCCACAGG-3' forward and 5'-GAGTTGGAGGTTGAGATGCTGAC-3' reverse). Three clones were positively identified by both methods. To test whether the fosmids contained the full length MAT region, the clones were tested in PCR experiments amplifying parts of other flanking genes, such as the sla2 gene (ARB_07317) and the rps4 gene (ARB_07322). For sla2, PCR primer pair 5'-CTTGTTCAGGAGAGCTATGG-3' and 5'-CAGCTTCTCGAGCTCCTCCC-3' was used; for rps4, PCR primer pair 5'-CAGCGCCTGGTCAAGGTCGACG-3' and 5'-GGTCACGCTCCTCAGCAATGG-3' was used. DNA of a positive fosmid was shotgun sequenced using dye terminator chemistry (ABI).

In addition, genomic 454 libraries were generated according to the manufacturer's protocol and sequenced using a GS FLX (Roche, Mannheim, Germany). The nucleotide sequences were assembled species-specific using the newbler software. Clone gaps were filled using a primer walking strategy with custom primers. Isolation, quantification and quality control of total RNA was performed as described 43 . A cDNA-library was constructed according to the manufacturer's protocols (Evrogen) and 1,411 ABI dye terminator sequences were obtained mainly from the 5' end. The sequences were matched to the assembled genomic sequences to determine exon/intron structures and to obtain an intron signature for the species.

The same DNA as for the preparation of the plasmid/fosmid libraries was used for the preparation of genomic libraries for the 454/FLX system (Roche) according to the manufacturer's protocols. Three runs each were performed on the 454/FLX sequencing machine. All 454/FLX sequence data were assembled species-specific using the newbler software. The Sanger based sequencing reads were assembled onto this 'backbone'. Clone gaps were filled using a primer walking strategy with custom primers.

Both genomes are deposited in NCBI with accession codes [DDBJ/EMBL/GenBank:ABSU00000000] for A. benhamiae and [DDBJ/EMBL/GenBank:ACYE00000000] for T. verrucosum.

Gene models of both fungi were generated by using in silico predictions and sequence data from an EST library constructed from cultured A. benhamiae cells. We matched 1,411 ABI dye terminator sequences obtained from the cDNA library sequencing to the assembled genomic sequences to determine exon/intron structures and to obtain an intron signature for the species. The alignments of the cDNA sequences to the genomic backbone yielded evidence for 861 introns and at least 653 protein-coding open reading frames (coding sequences), which were validated by BLAST. These data were used to train the gene prediction program geneid 44 . To validate the accuracy of the gene prediction, 47 gene structures in one genomic region were annotated manually and compared to the automated predictions, indicating a specificity of 82% at a sensitivity of 97%. For the annotation and comparative analyses of the genomes a web based genome browser was set up using the GenColors database/software system 45 .

Blast analysis of all coding sequences of one genome against the other yielded best bidirectional hits. We used a filter threshold for significant hits of 30% identity between amino acid sequences over at least 50% of the protein.

A BlastN analysis of the genomic sequences was performed for all protein coding genes of T. verrucosum against all A. benhamiae contigs. A filter threshold for significant hits was 80% identity between sequences over at least 60% of the query length; 239 T. verrucosum sequences gave no hits or non-significant hits.

The human keratinocyte line HaCaT was obtained from Prof. Fusenig (Deutsches Krebsforschungszentrum, Heidelberg, Germany). The cells were cultivated in DMEM supplemented with 10% (v/v) fetal calf serum, gentamycin (28 μg/ml) and 1% (w/v) ultraglutamine at 37°C in a humidified atmosphere and 5% (v/v) CO₂ for 2 days. Medium and supplements were purchased from Lonza (Basel, Belgium). Human keratinocytes were infected by A. benhamiae conidia with a multiplicity of infection (MOI) of 6. Infected human cells were cultivated in fetal calf serum-free DMEM supplemented with both gentamycin and ultraglutamine for 96 h at 28°C. As a control, A. benhamiae conidia were grown in the absence of keratinocytes under the same conditions. After infection, the human keratinocytes were lysed by addition of 0.03% (v/v) Triton X for 2 minutes and A. benhamiae was harvested. Fungal cells were collected by centrifugation for 3 minutes at 3,500 g. A. benhamiae cells were washed twice in Dulbecco's phosphate buffered saline (Lonza) and stored in aliquots at -80°C. For RNA sequencing, total RNA was isolated using RiboPure™-Yeast Kit (Ambion Europe, Huntingdon, UK) according to the manufacturer's instructions from keratinocytes co-incubated with conidia and conidia only grown in cell culture medium for 96 h.

RNA was reverse transcribed using a SMART technique (Evrogen). The single-stranded DNA was then amplified using SMART primers for 20 cycles to produce double-stranded DNA in sufficient quantity for GS-FLX sequencing (Roche). We generated A. benhamiae transcriptome data by sequencing parts of individual cDNAs after fragmentation by nebulization using 454/FLX sequencing technology. For postprocessing of these sequences, SMART adapters were identified and clipped using a combination of perl scripts plus cross_match. After further cleaning with seqclean (removal of polyA tails and low complexity regions), 682,580 ESTs (98.8 Mb) remained for mapping.

Mapping of the ESTs to the repeat-masked A. benhamiae genome as a backbone was done in two major steps. First, we used Blat 46 to assign each EST to its most probable position in the genome allowing a maximum intron length of 10 kb. A valid hit required a minimum length of 30 bp and a minimum identity of 90% to the backbone sequence. In the second step, each EST was realigned to its most probable position utilizing a slightly modified version of Exalin 47 that implements the Smith-Waterman algorithm and information theory for better alignments and intron prediction. Using this approach, we were able to align 571,963 ESTs to the genome. Finally, EST positions were translated to positions of known gene models if possible. In this way, we determined for each gene a set of ESTs and thereby its raw expression level. The data were normalized to the total number of mapping ESTs. Table S9 in Additional file 6 shows the total numbers of generated reads, the reads mapped to a genome, and the reads in gene models for each technical replicate of infection and control samples.

The raw counts for the transcripts were analyzed using the R Statistical Computing Environment and the Bioconductor packages DESeq 48 and edgeR 49 . Both packages provide statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting P-values were adjusted using the Benjamini and Hochberg's approach for controlling the false discovery rate 50 . Genes with an adjusted P-value <0.05 found by both packages were assigned as differentially expressed.

The RNAseq data are submitted to the Sequence read archive of NCBI and are available with the accession numbers [NCBI:SRR070551] and [NCBI:SRR070552] (sample runs) and [NCBI:SRR070553] and [NCBI:SRR070554] (control runs).

Total RNA from mycelial samples was isolated using RiboPure™-Yeast Kit (Ambion) according to the manufacturer's instructions. Total RNA was denatured (15 minutes, 60°C; 5% (v/v) formaldehyde, 50% (v/v) formamide, 40 mM MOPS, pH 7) and separated by agarose gel electrophoresis (1.2% agarose, 40 mM MOPS, 10 mM sodium acetate, 2 mM EDTA, 2% (v/v) formaldehyde, pH 7). Blotting, hybridization and chemoluminescent signal detection were performed according to the manufacturer's instructions (DIG Application Manual for Filter Hybridization, Roche). Gel load and blot signal strength were quantified and normalized using Bio-Rad (Munich, Germany) Quantity One (v4.6.7) software.

For cultivation of A. benhamiae, medium was prepared as follows: 10 g/l keratin (MP Biomedicals Europe, Illkirch, France) was autoclaved in water and subsequently 20 mM potassium phosphate pH 5.5, 0.4 mM MgSO₄, 77 mM NaCl, 5 mM glucose and 0.5% (v/v) SL-8 trace elements 51 were added. Microconidia obtained from A. benhamiae cultivated for 7 days on MAT agar at 30°C were used to inoculate shaking flasks at a final spore concentration of 10⁶ per milliliter. After cultivation for 2 days at 200 rpm and 30°C, cultures were filtered through Miracloth (Calbiochem, Darmstadt, Germany) and the supernatant was centrifuged at 4,000 g for 20 minutes at 4°C. Secreted proteins were precipitated with 10% (w/v) trichloroacetic acid/6.5 mM DTT overnight at 4°C. The precipitate was pelleted at 4,000 g for 20 minutes at 4°C and resuspended twice in ice-cold acetone/water (9:1)/6.5 mM DTT followed by subsequent centrifugation steps. The air-dried pellet was dissolved in lysis buffer 3, as described 52 . Immobiline DryStrips of 11 cm covering a pH range from 3 to 10 (GE Healthcare Life Sciences) were rehydrated overnight according to the manufacturer's instructions. Isoelectric focusing was carried out in an Ettan IPGphor II using a 0 to 1 kV gradient for 11 h, 1 to 8 kV for 3 h and finally 8 kV for 24 kVh. Afterwards, strips were incubated for 15 minutes in equilibration buffer (6 M urea, 2% (w/v) SDS, 75 mM Tris^.Cl pH 8.8, 30% (v/v) glycerol) with 65 mM DTT, followed by an alkylation step of the proteins with 135 mM iodoacetamide in equilibration buffer under the same conditions. Separation of proteins by the second dimension was carried out using pre-cast Criterion gels (12.5% (w/v), Tris-HCl; Bio-Rad) according to the manufacturer's instructions. Proteins were visualized by Colloidal Coomassie Brilliant Blue G-250 staining 53 .

Protein spots were excised from the gels and digested with sequencing-grade Trypsin (Promega, Mannheim, Germany) as described elsewhere 54 . Eluted peptides were mixed with an equal amount of a saturated alpha-cyano-hydroxycinnamic acid (Bruker Daltonics, Bremen, Germany) solution in aqueous 30% (v/v) acetonitrile and spotted on an MTP anchor-chip 800/384 (Bruker Daltonics). Mass spectrometry spectra were acquired with an Ultraflex I TOF/TOF (Bruker Daltonics) mass spectrometer using Peptide Mass Standard II (Bruker Daltonics) as calibrant. The five most intense mass spectrometry signals were selected for tandem mass spectrometry analysis. MASCOT (version 2.1.02; Matrix Science, London, UK) searching against protein predictions from the A. benhamiae genome and the NCBI database (taxa fungi) was used for protein identification with the following the parameters: fixed modification of cysteine to S-carbamidomethyl derivatives, variable methionine oxidation, no missed cleavage and a peptide mass tolerance of 200 ppm.

The PKS and NRPS domain architecture was predicted using the InterProScan 55 and NRPS-PKS 56 tools.

For genome-based phylogeny, 23 proteins from 28 fully sequenced fungal genomes were used for the reconstruction of the phylogenetic relationships of A. benhamiae and T. verrucosum (Additional file 3). The 23 ortholog groups were selected based on the KOG (clusters of orthologous groups for eukaryotes) assignments, as described by Xu et al. 23 . Only KOGs without paralogs, that is, proteins represented by a single protein in a species, were taken into consideration. Five proteins from the publication of Xu et al. 23 were not confirmed as fulfilling this requirement. Thus, they were not included. The genome set selected for the survey was non-redundant, that is, we did not consider four closely related Candida species as well as six Saccharomyces species, but only representatives of each clade, that is, C. albicans and S. cerevisiae, respectively. By contrast, we included all available Pezizomycetes, since A. benhamiae and T. verrucosum presumably belong to this phylum. A representative of Zygomycota (Rhizopus oryzae) was used as an outgroup. The considered genomes were as follws. Eurotiomycetes: Arthroderma benhamiae, Trichophyton verrucosum, Aspergillus clavatus, Aspergillus flavus, Aspergillus fumigatus, Aspergillus nidulans, Aspergillus oryzae, Aspergillus terreus, Botrytis cinerea, Coccidioides immitis, Histoplasma capsulatum, Paracoccidioides brasiliensis, Sclerotinia sclerotiorum, Stagonospora nodorum, Uncinocarpus reesii. Sordariomycetes: Chaetomium globosum, Fusarium graminearum, Magnaporthe grisea, Neurospora crassa. Saccharomycotina: Candida albicans, Lodderomyces elongisporus, Saccharomyces cerevisiae. Taphrinomycotina: Schizosaccharomyces japonicus. Basidiomycota: Coprinus cinereus, Cryptococcus neoformans, Puccinia graminis, Ustilago maydis. Zygomycota: Rhizopus oryzae.

The protein sets for each KOG protein shared among the 28 genomes were collected. Each set was then aligned by ClustalX, and the conserved blocks were extracted using the Gblocks tool 57 with allowance of smaller final blocks (five amino acids) and gap positions within the final blocks using otherwise default parameters. The extracted blocks were concatenated for each species. The phylogenetic analysis was performed using PHYML 58 for the construction of the maximal likelihood tree, and PHYLIP for the construction the neighbor joining tree, with the Jones-Taylor-Thornton model of the amino acid substitution in both cases. The neighbor joining and maximal likelihood trees had identical architecture.

The phylogenetic trees for proteases and enzymes involved in SM production were obtained using PHYLIP for the construction of the neighbor joining tree, with the Jones-Taylor-Thornton model of the amino acid substitution.