Dothideomycete–Plant Interactions Illuminated by Genome Sequencing and EST Analysis of the Wheat Pathogen Stagonospora nodorum

Abstract

Stagonospora nodorum is a major necrotrophic fungal pathogen of wheat (Triticum aestivum) and a member of the Dothideomycetes, a large fungal taxon that includes many important plant pathogens affecting all major crop plant families. Here, we report the acquisition and initial analysis of a draft genome sequence for this fungus. The assembly comprises 37,164,227 bp of nuclear DNA contained in 107 scaffolds. The circular mitochondrial genome comprises 49,761 bp encoding 46 genes, including four that are intron encoded. The nuclear genome assembly contains 26 classes of repetitive DNA, comprising 4.5% of the genome. Some of the repeats show evidence of repeat-induced point mutations consistent with a frequent sexual cycle. ESTs and gene prediction models support a minimum of 10,762 nuclear genes. Extensive orthology was found between the polyketide synthase family in S. nodorum and Cochliobolus heterostrophus, suggesting an ancient origin and conserved functions for these genes. A striking feature of the gene catalog was the large number of genes predicted to encode secreted proteins; the majority has no meaningful similarity to any other known genes. It is likely that genes for host-specific toxins, in addition to ToxA, will be found among this group. ESTs obtained from axenic mycelium grown on oleate (chosen to mimic early infection) and late-stage lesions sporulating on wheat leaves were obtained. Statistical analysis shows that transcripts encoding proteins involved in protein synthesis and in the production of extracellular proteases, cellulases, and xylanases predominate in the infection library. This suggests that the fungus is dependant on the degradation of wheat macromolecular constituents to provide the carbon skeletons and energy for the synthesis of proteins and other components destined for the developing pycnidiospores.

Dothideomycetes is a large class of filamentous fungi containing many necrotrophic fungal pathogens. This work describes the genome sequence and extensive EST data of the wheat pathogen Stagonospora nodorum. Gene prediction analyses suggest a minimum of 10,762 nuclear genes contained in 38 Mbp and reveal gene families and enzymatic functions likely to be important in the host–parasite interaction.

INTRODUCTION

Stagonospora nodorum (syn. Phaeosphaeria) is a major pathogen of wheat (Triticum aestivum) in most wheat-growing areas of the world. It is the major cause of losses due to foliar pathogens in Western Australia and north central and northeastern North America (Solomon et al., 2006a), and until the 1970s, it was the major foliar necrotrophic pathogen in Europe (Bearchell et al., 2005). Infection begins when spores (ascospores or asexual pycnidiospores) land on leaf tissue (Bathgate and Loughman, 2001; Solomon et al., 2006b). The spores rapidly germinate to produce hyphae that invade the leaf, using hyphopodia to gain entry to epidermal cells or by growing directly through stomata (Solomon et al., 2006b). The hyphae rapidly colonize the leaves and begin to produce pycnidia in 7 to 10 d. The infection has been divided into three metabolic phases (Solomon et al., 2003a). The first phase is penetration of the host epidermis and is fuelled predominantly by lipid stores in the spores (Solomon et al., 2004a); the second is proliferation throughout the interior of the leaf, involves toxin release (Liu et al., 2006), and uses host-derived simple carbohydrate sources (Solomon et al., 2004a); the final phase produces the new spores, but so far the metabolic requirements for pycnidiation are unclear.

The infection epidemiology follows a polycyclic pattern with repeated cycles of both asexual (Bathgate and Loughman, 2001) and sexual (B.A. McDonald, unpublished data) infection throughout the growing season. New rounds of infection are initiated by rain-splash of pycnidiospores and wind dispersal of ascospores. Eventually, wheat heads become infected, causing the glume blotch symptom. In Mediterranean climates, the fungus over-summers in senescent straw and stubble. Seed transmission can be important but is readily controlled by fungicides. Infected stubble harbors the pseudothecia that produce airborne (Bathgate and Loughman, 2001), heterothallic (Bennett et al., 2003) ascospores to reinitiate the infection in the following growing season. Epidemiological and population genetic evidence suggest that the fungus undergoes meiosis in the field regularly and that gene flow is global (Stukenbrock et al., 2006).

S. nodorum is a member of the Dothideomycetes class of filamentous fungi. This is a newly recognized major class that broadly replaced the long-recognized loculoascomycetes (Winka and Eriksson, 1997) and includes the causal organisms of many economically important plant diseases; notable examples are black leg (Leptosphaeria maculans), southern maize (Zea mays) leaf blight (Cochliobolus heterostrophus), barley (Hordeum vulgare) net blotch (Pyrenophora teres), apple scab (Venturia inaequalis), black sigatoka (Mycosphaerella fijiensis) of banana (Musa spp), wheat leaf blotch (M. graminicola), tomato (Solanum lycopersicum) leaf mold (Passalora fulva), and Ascochyta blight of many legume species (Ascochyta spp). The taxon also includes large numbers of saprobic species occurring on substrates ranging from dung to plant debris, a few species associated with animals, and several lichenized species (Del Prado et al., 2006).

Many plant pathogens in this group produce host-specific toxins (Wolpert et al., 2002). These molecules interact with specific host factors to produce disease symptoms only in selected genotypes of specific plant hosts. Well-known examples are found in Alternaria alternata, Cochliobolus heterostrophus, and Pyrenophora tritici-repentis, while species in the genera Mycosphaerella, Corynespora, and Stemphylium are also thought to produce host-specific toxins (Agrios, 2005). Proteinaceous host-specific toxins have recently been shown to be important virulence determinants in S. nodorum (Liu et al., 2004a, 2004b), including one whose gene is thought to have been interspecifically transferred to the wheat tan spot pathogen, P. tritici-repentis (Friesen et al., 2006). Only one example of a host-specific toxin has been identified outside of the Dothideomycetes (Wolpert et al., 2002). Non-host-specific toxins produced by species in this group include cercosporin (Cercospora) and solanopyrone (Ascochyta) but are also produced by many other fungal taxa (Agrios, 2005).

S. nodorum is an experimentally tractable organism, which is easily handled in defined media, was one of the first fungal pathogens to be genetically manipulated (Cooley et al., 1988), and has been a model for fungicide development (Dancer et al., 1999). Molecular analysis of pathogenicity determinants is aided by facile tools for gene ablation and rapid in vitro phenotypic screens, and thus far, a small number of genes required for pathogenicity have been identified (Cooley et al., 1988; Bailey et al., 1996; Bindschedler et al., 2003; Solomon et al., 2003b, 2004a, 2004b, 2005, 2006a). It has thus emerged as a model for dothideomycete pathology.

Whole-genome sequences have been described for a handful of fungal saprobes and pathogens (Galagan et al., 2003, 2005; Jones et al., 2004; Dean et al., 2005; Kamper et al., 2006). Here, we present an initial analysis of the genome sequence of the dothideomycete S. nodorum. Gene expression studies using EST libraries from axenic mycelium and heavily infected wheat leaves complement the genome sequence and provide a broad-based analysis of the genomic basis of infection by S. nodorum.

RESULTS

Acquisition and Analysis of the Genome Sequence

The genome sequence was obtained using a whole-genome shotgun approach. Approximately 10× sequence coverage was obtained as paired-end reads from plasmids of 4 and 10 kb plus 40-kb fosmids. Reads were assembled using Arachne (Jaffe et al., 2003), forming 496 contigs totaling 37,071 kb. The contig N50 was 179 kb, meaning an average base in the assembly lies within a contig of at least 179 kb. Greater than 98% of the bases in the assembly have a consensus quality score of ≥40, corresponding to the standard error rate of fully finished sequence. The contigs are connected in 109 scaffolds spanning 37,202 kb with a scaffold N50 of 1.05 Mb. More than 50% of the genome is contained in the 13 largest scaffolds. Within the scaffolds, only ∼140 kb is estimated to lie within sequence gaps. Thus, in terms of representation, contiguity, and sequence accuracy, the draft assembly is of high quality. The mitochondrial genome comprises a circle of 49,761 bp (GenBank accession number EU053989). It replaces two of the auto-assembled scaffolds, 52 and 65. The nuclear genome is therefore assembled in 107 scaffolds of total length 37,164,227 bp.

General features of the assembly are described in Table 1  

. As detailed below, the genomic sequencing was complemented by sequencing EST libraries. One library was constructed from axenically grown mycelium, and after removing vector, poly(A) sequences, poor-quality sequences, and sequences <100 bp, there were 7750 remaining ESTs. Of these, 97.6% mapped to the genome assembly via Sim4, and 1.4% mapped to unassembled reads. This indicates that the assembly has achieved good coverage of the genome. In all, seven EST contigs and 13 singletons clustered to the unassembled reads.

Analysis of Repetitive Elements

Repetitive elements were identified de novo by identifying sequence elements that existed in 10 or more copies, were greater than 200 bp, and exhibited >65% sequence identity. The analysis revealed the presence of 25 repeat classes. Table 2  

lists the general features of the repeats (see Supplemental Data Set 1 online). Only three, Molly, Pixie, and Elsa, had been detected earlier (GenBank accession numbers AJ277502, AJ277503, and AJ277966, respectively). Molly, Pixie, X15, and R37 show sequence characteristics of inverted terminal repeat–containing transposons, while Elsa, R9, and X11 appear to be retrotransposons. Repetitive elements were found individually throughout the genome but were often found in clusters spanning several kilobases.

Telomere-associated repeats were identified by searching for examples of the canonical telomere repeat TTAGGG at the termini of auto-assembled scaffolds. Physically linked repetitive sequences were then analyzed for association with the TTAGGG sequence repeats. Between 19 and 38 copies of telomere-associated repeats were found in the assembly.

Repeat-induced point (RIP) mutation is a fungal-specific genome-cleansing process that detects repeated DNA at meiosis and introduces C-to-T mutations into the copies (Cambareri et al., 1989). Using the parameters defined for Magnaporthe grisea (Dean et al., 2005), we identified RIP-like characteristics in several of the repeat classes (Table 2; see Supplemental Data Set 1 online). The transposons Molly and Elsa were the most clearly affected classes. None of the telomere-associated repeats displayed RIP characteristics.

Mitochondrial Genome

Gene Content

The initial gene prediction process identified 16,957 gene models of which 16,586 were located on the 107 nuclear scaffolds. A revised gene prediction procedure, using new ESTs (see below) and 795 fully supported and manually annotated gene models, identified 10,762 version 2 nuclear gene models. Of these, 617 genes corresponded to a merging of version 1 genes; in 50 cases, three prior gene models were merged, and in one case, four genes were merged. ESTs that aligned to unassembled reads identified one supported gene (SNOG_20000.2). A total of 5354 version 1 gene models were not supported but did not conflict with second-round predictions. We therefore conclude that S. nodorum contains a minimum of 10,762 nuclear genes of which all but 125 are supported by two gene prediction procedures and 2696 are supported by direct experimental evidence via EST alignment. These genes, with an identification format SNOG_xxxxx.2, were compared with the GenBank nonredundant protein database at the National Center for Biotechnology Information (NCBI). Informative (not hypothetical, predicted, putative, or unknown) BLASTP (Altschul et al., 1990) hits with e-values <1 × 10⁻⁶ were found for 7116 genes (see Supplemental Data Set 2 online). It is estimated that at least 46.6% of the nuclear genome is transcribed and 38.8% is translated.

The 5354 gene models without support from the reannotation have unaltered accession numbers as SNOG_xxxxx.1 and are retained for further possible validation and analysis. As 952 of these unsupported gene models have BLASTP hits with e-values <1 × 10⁻⁶, we predict that some will be validated as new evidence emerges.

Gene Expression during Infection

Two EST libraries were constructed and analyzed as part of this project. An in vitro library was constructed from axenic fungal mycelium transferred to media with oleate as the sole carbon source; this is referred to as the oleate library. An in planta library was made from bulked sporulating disease lesions on wheat 9, 10, and 11 d after infection (DAI). For both libraries, 5000 random clones were sequenced in both directions. The oleate library was entirely fungal and hence particularly suited for primary genome annotation purposes. The lipid growth media was chosen to replicate the early stages of infection in which lipolysis, the glyoxalate cycle, and gluconeogenesis are thought to be critical metabolic requirements (Solomon et al., 2004a). The in planta library was designed to reveal both plant and fungal genes expressed at a late stage of infection. After trimming and removal of plant ESTs and alignment to the genome assembly, the in planta library formed 1448 and the oleate library formed 1231 unigenes (see Supplemental Data Set 2 online). Only 427 of the unigenes were found in both libraries. Although this represents 19% of the unigenes, it encompasses 46% of the ESTs, showing the genes expressed uniquely in one library are of relatively low expression.

The unigenes obtained from the in planta and oleate libraries were classified according to gene ontology (GO) categories. GO matches were obtained for 851 (59%) of the unigenes found in the in planta and 736 (59%) of those found in the oleate library. Relative numbers of ESTs and genes in different GO classes were compared (Table 3  

). GO categories were ranked by statistical discrimination between the two libraries, and the top 10 are shown for each of biological process, cellular component, and molecular function.

Coexpression of fungal gene clusters responsible for the synthesis of secondary metabolites (Bok and Keller, 2004) and pathogenicity (Kamper et al., 2006) has been observed. We compared the number of ESTs found in the in planta and oleate libraries to search for clusters of coexpressed genes. Using a window of 10 contiguous genes, we scanned for regions with biased ratios of ESTs from either library. One putatively in planta–induced region stood out. These six genes, SNOG_16151.2 to SNOG_16157.2, were exclusively in the in planta library with 4, 20, 0, 4 10, and 1 EST clones, respectively. The genes have best hits to a major facilitator superfamily transporter, a phytanoyl-CoA dioxygenase, a CocE/NonD hydrolase, and a salicylate hydroxylase and are next to a transcription factor. Such a cluster may be involved in the degradation of phytols, phenylpropanoids, and catechols either for nutritional purposes, providing trichloroacetic acid intermediates via the β-ketoadipate pathway or to detoxify wheat defense compound(s). It is intriguing that overexpression of the thiolase in the β-ketoadipate pathway in L. maculans markedly reduced pathogenicity (Elliott and Howlett, 2006). Experiments to test these ideas in S. nodorum are in progress.

DISCUSSION

The estimated genome size of S. nodorum is 37.2 Mbp, which is significantly larger than the 28 to 32 Mbp previously estimated by pulsed-field gel analysis (Cooley and Caten, 1991). Electrophoretic karyotypes have proven to be unreliable estimators of total genome size in many fungal species (Orbach et al., 1988; Orbach, 1989; Talbot et al., 1993). In the case of S. nodorum, this discrepancy may be a consequence of comigration of chromosomal bands on pulsed-field gels leading to an underestimation of genome size (Cooley and Caten, 1991). The electrophoretic karyotype was found to be highly variable between strains, even when isolated from a single ascus, consistent with the generalization that many fungal species have plastic genome structures (Zolan, 1995). The revised genome size estimate is comparable to other sequenced filamentous fungi, such as the rice pathogen M. grisea, which is currently estimated to be 41.6 Mbp, and the nonpathogen Neurospora crassa now estimated at 39.2 Mbp.

Phylogenetic Analysis

Repeated Elements in the Nuclear Genome

Prior to this study, only one unpublished study had been made of repetitive elements in the S. nodorum genome (Rawson, 2000). The de novo analysis of repeats is likely to become a feature of future genome sequencing projects as organisms with little prior molecular work are selected. The total amount of repetitive DNA in the S. nodorum nuclear genome is estimated at 4.5%. This compares with 9.5% in M. grisea. Some of these repeats could be associated with telomeres. Between 19 and 38 copies of telomere-associated repeats were found in the assembly. These numbers accord well with the 14 to 19 chromosomes visualized by pulsed-field electrophoresis (Cooley and Caten, 1991).

Studies of S. nodorum life cycle have indicated that it undergoes regular sexual crossing, particularly in areas with Mediterranean-style climates with the associated need to over-summer as ascospores (Bathgate and Loughman, 2001; Stukenbrock et al., 2006). The relatively low content of repetitive DNA is consistent with this observation. The prevalence of clear cases of RIP further confirms that meiosis is a frequent event. RIP has been found to varying degrees in other fungal genomes. It is notable that the closely related L. maculans has previously been shown experimentally to exhibit RIP (Idnurm and Howlett, 2003).

The Mitochondrial Genome

A distinctive feature of fungal mitochondrial genomes is the clustering of tRNA genes (Ghikas et al., 2006), and it is thought that both the tRNA gene content and their placement will be conserved in fungi (Table 4  

). The 27 tRNA genes clustered into five groups, with the two larger tRNA gene clusters flanking rnl, a pattern common to other fungal mitochondrial DNAs (mtDNAs) (Tambor et al., 2006). The tRNA gene cluster 5′ to rnl had GDS¹WIS²P as a consensus, with its closest sequenced relative M. graminicola (GenBank accession number EU090238), while the 3′-downstream consensus was EAFLQHM, having many tRNA genes in the same order found in other Ascomycetes. Variation in the order of tRNAs, such as inversions in Epidermophyton floccosum (Ile-Trp) or in Podospora anserina and S. nodorum (Met-His), suggest that rearrangements are common in fungal mtDNAs (Table 4).

Kouvelis et al. (2004) argued that gene pairs nad2-nad3, nad1-nad4, atp6-atp8, and cytb-cox1 would remain joined in ascomycetes with some possible exceptions, as already detected in M. graminicola that present only two of these gene pairs coupled. In S. nodorum, the atp8-9 genes were not present, cytb-cox1 were uncoupled, and the nad1 and nad4 genes were on sections of the mtDNA in an inverted orientation. Two orfs (orf1 and orf2) had homologous sequences in the in planta EST library. While the M. graminicola mtDNA genome had no introns, four intron-encoded genes were found in S. nodorum, with high homology to LAGLIDADG-type endonucleases or GIY-YIG–type nucleases (see Supplemental Table 1 online). Intron-encoded proteins have also been reported in P. anserina (Cummings et al., 1990) and Penicillium marneffei (Woo et al., 2003).

Functional Analysis of Proteins

Our primary goal in obtaining the genome sequence was to reveal genes likely to be involved in pathogenicity. While some genes seem to be specifically associated with pathogenicity in a single organism, other genes and gene families have been generally associated with pathogenicity, albeit they are also found in nonpathogens (see http://www.phi-base.org/about.php; Baldwin et al., 2006). Some of the generic functions are listed (Table 5  

; see Supplemental Data Set 2 and Supplemental Table 2 online). EST support was found for 59 of the genes, with no statistically significant difference in the numbers of ESTs in the in planta and oleate libraries.

Nonribosomal peptide synthetases (NRPSs) are modular enzymes that synthesize a diverse set of secondary metabolites, including the dothideomycete host-specific toxins AM-toxin, HC-toxin, and victorin from Alternaria and Cochliobolus species (Wolpert et al., 2002). NRPS genes from C. heterostrophus have been analyzed in detail (Lee et al., 2005). Comparison of both protein sequences and domain structure of the C. heterostrophus NRPS genes (NPS1-11) with the eight identified in S. nodorum was undertaken (Table 6  

). Putative orthology was determined by identifying reciprocal best hits among the gene sets. Three pairs were identified. SNOG_02134.2 was linked to NPS2, and both appear to be orthologous to the Aspergillus nidulans gene SidC (Eisendle et al., 2003), which is responsible for the synthesis of ferricrocin, an intracellular iron storage and transport compound involved in protection against iron toxicity (Eisendle et al., 2006). It is likely that SNOG_02134.2 and NPS2 play similar roles. SNOG_14638.2 was linked to NPS6, a ubiquitous gene with a related role in siderophore-mediated iron uptake and oxidative stress protection (Oide et al., 2006). Third, SNOG_14834.2 appears to be directly related to NPS4, Psy1 from Alternaria brassicae (Guillemette et al., 2004), and NPS2 from Alternaria brassicicola (Kim et al., 2007). Ab NPS2 mutants are reduced in virulence, and the gene is predicted to encode a component of the conidial wall. It is likely that SNOG_14834.2 will have a similar generic role. Reciprocal best-hit relationships were observed between a further two SNOG NRPS genes and other fungal genes; SNOG_09081.2 was related to PesA (Bailey et al., 1996) and SNOG_14908.2 to the Salps2 gene from Hypocrea lixii (Vizcaino et al., 2006). These genes plus SNOG_14834.2, SNOG_09488.2, and SNOG_01105.2 are all closely related to NPS4, suggesting that this five-gene subfamily is expanded in S. nodorum. Interestingly, a further seven NRPS genes from C. heterostrophus (NPS3, 5, 8, 10, 11, and 12) have no obvious ortholog in S. nodorum, suggesting expansion of this subfamily in the maize pathogen (Yoder and Turgeon, 2001). Intensive studies in C. heterostrophus showed that individual gene knockouts produced altered phenotypes only for NPS6 (Lee et al., 2005), indicating redundancy of function. The smaller complement of NRPS genes in M. grisea (eight) and S. nodorum indicate that these organisms may be more fruitful models to study NRPS function. Furthermore, although toxins have been implicated in the virulence of both pathogens, nonribosomal synthetases cannot be expected to be a major source. No ESTs were found corresponding to these genes. This may be due to their large size or to a generally low level of expression.

Polyketide synthases (PKSs) are a second modular gene family strongly associated with pathogenicity (Kroken et al., 2003; Gaffoor et al., 2005), being responsible for the production of T-toxin from C. heterostrophus and PM-toxin from Mycosphaerella zeae-maydis (Yun et al., 1998; Baker et al., 2006). S. nodorum is predicted to contain 19 PKS genes compared with the 24, 14, and 24 in the pathogens M. grisea, Fusarium graminearum, and C. heterostrophus, respectively, and 28 in the saprobe A. nidulans but significantly more than the seven in N. crassa (Table 5). Orthology relationships between the well-studied F. graminearum and C. heterostrophus gene sets and S. nodorum are shown in Table 7  

. Reciprocal best BLAST hits were observed with eight genes from C. heterostrophus and five from F. graminearum. This close relationship, particularly with C. heterostrophus, reflects the close phylogenetic relationship and suggests an ancient origin for the majority of the PKS paralogs (Kroken et al., 2003). SNOG_11981.2 is supported by five ESTs from the in planta library and none from the oleate library, consistent with upregulation during infection and sporulation (no other PKSs have EST support). This gene is orthologous to nonreducing clade 2 PKS genes that are associated with 1,8-dihydroxynaphthalene melanin biosynthesis in Bipolaris oryzae and many other pathogens (Kroken et al., 2003; Moriwaki et al., 2004; Amnuaykanjanasin et al., 2005). We showed previously that S. nodorum produces melanin from dihydroxyphenlyalanine (Solomon et al., 2004b), for which a PKS would not be necessary. This finding suggests that either S. nodorum produces 1,8-dihydroxynaphthalene melanin in addition to dihydroxyphenlyalanine melanin or that SNOG_11981.2 plays another role.

A single PKS-NRPS hybrid protein (SNOG_00308.2.) was identified in the genome of S. nodorum as was found for F. graminearum (Gaffoor et al., 2005) and C. heterostrophus (Lee et al., 2005). C. heterostrophus NPS7 and SNOG_00308.2 do not appear to be closely related, and it is likely that they evolved independently. By contrast, the hybrid protein of F. graminearum (Gz FUS1/FG12100) is closely related to SNOG_00308.2. Overall, they are 54.5% similar, and apart from an acyl binding domain found only in SNOG_00308.2, the domain structures are identical. Gz FUS1 produces fusarin C, a mycotoxin (Gaffoor et al., 2005). The best hit (with 59.4% similarity) to SNOG_00308.2 in M. grisea is Ace1, a hybrid PKS/NRPS that confers avirulence to M. grisea during rice (Oryza sativa) infection (Bohnert et al., 2004). It will be intriguing to determine if the product of SNOG_00308.2 is required for pathogenicity.

G-alpha proteins have been extensively studied in fungi (Lafon et al., 2006), and many are required for pathogenicity. Distinct roles for three g-alpha genes have been revealed in M. grisea, A. nidulans, and N. crassa. It was therefore a surprise to find a fourth g-alpha gene in the S. nodorum genome. This gene (SNOG_06158.2) was shown to be expressed in vitro and in planta. It was investigated by gene disruption, and its loss resulted in no discernable phenotype (data not shown). A fourth g-alpha protein has been recently identified in both A. oryzae (Lafon et al., 2006) and Ustilago maydis (Kamper et al., 2006), but neither shows significant sequence similarity to that of S. nodorum. Indeed, SNOG_06158.2 is most similar to Gba3 among the U. maydis g-alphas and GpaB among the A. oryzae genes.

G-alpha proteins transduce extracellular signals leading to infection-specific development (Solomon et al., 2004b). Pth11 and ACI1 are two M. grisea genes encoding transmembrane receptors that defined a new protein domain, CFEM (Kulkarni et al., 2003), with roles in g-protein signaling. M. grisea has nine CFEM domain proteins, and F. graminearum has eight. Using a combination of domain searches and BLAST searches seeded with the M. grisea and F. graminearum genes, we identified six related genes (Table 8  

). Four of these have at least a weak match to the CFEM domain, but only three are predicted to be transmembrane proteins. SNOG_09610.2 appears to be the ortholog of Pth11 and the F. graminearum gene fg05821, suggesting these genes are ancient and conserved. Pth11 is required for appressorial development and perception of suitable surfaces (DeZwaan et al., 1999). Neither S. nodorum nor F. graminearum form classical appressoria (Solomon et al., 2006b), suggesting that the genes have different functions in these species. SNOG_05942.2 is also closely related to Pth11. Knockout strains for this gene were obtained but had no obvious phenotype (data not shown). SNOG_03589.2 is also related to Pth11 and appears to be a transmembrane protein but lacks the CFEM domain. Both SNOG_08876.2 and SNOG_15007.2 possess CFEM domains and a signal peptide but do not appear to be integral membrane proteins. Knockout strains for SNOG_08876.2 were obtained but revealed no obvious phenotype (data not shown). SNOG_15007.2 is similar to glycosylphosphatidylinositol-anchored CFEM-containing proteins from three other fungal species. It appears to be heavily expressed both in the oleate and in planta EST libraries with 17 and 27 ESTs, respectively. A clear role for this gene is as yet unknown. CFEM domain proteins are thought to be involved in surface signal perception, and three candidates for this role have been identified. The paucity of such genes in S. nodorum and lack of phenotype associated with deletion of two of them suggests that this function may be covered by a different class of transmembrane receptors.

Hydrophobins are small, secreted proteins with eight Cys residues in a conserved pattern that coat the fungal mycelium and spore (Wessels, 1994). Class 2 hydrophobins are restricted to ascomycetes, whereas class 1 hydrophobins are also found in other fungal divisions (Linder et al., 2005). Two class 2 and no class 1 hydrophobin genes were found in the S. nodorum genome (see Supplemental Figure 1 online). This is an unusual example of an ascomycete genome with only class 2 hydrophobin genes. It is interesting to note that although a range of Neurospora species all contained a gene orthologous to the class 1 hydrophobin EAS gene from N. crassa, they were expressed significantly only in species that produced aerial conidia (Winefield et al., 2007). S. nodorum produces pycnidiospores in a gelatinous cirrhus adapted for rain dispersal (Solomon et al., 2006b). It may be that the absence of aerial mitosporulation negates the need for class 1 hydrophobins.

Two rhodopsin-like genes were found in the genome, one of which is similar to the bacteriorhodopsin-like gene found in L. maculans that is a proton pump (Sumii et al., 2005). The likely physiological roles of these genes are currently under investigation.

One newly recognized aspect of S. nodorum is the production of secreted proteinaceous toxins. SNOG_16571.2 encodes a host-specific protein toxin called ToxA that determines the interaction with the dominant wheat disease susceptibility gene Tsn1 (Friesen et al., 2006; Liu et al., 2006). Population genetic evidence suggests that this gene was interspecifically transferred to the wheat tan spot pathogen, P. tritici-repentis, prior to 1941, thereby converting a minor into a major pathogen. An RGD motif that is involved in import into susceptible plant cells is required for activity (Meinhardt et al., 2002; Manning et al., 2004; Manning and Ciuffetti, 2005). Biochemical and genetic evidence suggests that S. nodorum isolates contain several other proteinaceous toxins (Liu et al., 2004a, 2004b). ToxA is a 13.2-kD protein, and current research indicates that the other toxins are also small proteins. Among the predicted extracellular proteins, 840 are <30 kD and 26 have an RGD motif (see Supplemental Data Set 2 online). Analysis of these toxin candidates is underway.

Gene Expression during Infection

When analyzed by biological process and molecular function, the in planta EST library was dominated by genes involved in the biosynthesis of proteins and in the degradation and use of extracellular proteins and carbohydrates, specifically cellulose and arabino-xylan (Table 3; see Supplemental Table 5 online). The secretion of proteases (Bindschedler et al., 2003), xylanases, and cellulases suggests that the fungus catabolizes plant proteins and carbohydrates for energy in planta at this late stage of infection. Studies in a range of pathogens suggest that early infection is characterized by the catabolism of internal lipid stores and that mid stages are characterized by the use of external sugars and amino acids (Solomon et al., 2003a). These studies of late infection suggest that polymerized substrates are used after the more easily available substrates are exhausted and provide a novel perspective to in planta nutrition. The abundance of transcripts linked to protein synthesis indicates that this process is very active during this stage of infection and may be related to the massive secretion of degradative enzymes and to the synthesis of proteins needed for sporulation. A similar picture was observed when the genes were analyzed by cellular component (Table 3) with gene products destined for the ribosome dominating. Genes whose products are targeted to the nucleolus and cytoskeleton illustrate the importance of nuclear division and cytokinesis at this stage of infection as many new cell types are elaborated in the developing pycnidium. Thirty-six gene products targeted to the extracellular region include a protease, an α-glucuronidase, and various glucanases consistent with the picture of polymerized substrate breakdown.

Late-stage infection has rarely been examined in plant–fungal interactions, and it is therefore noteworthy that upregulation of transcripts involved in protein synthesis was also observed in late-stage infections of wheat by M. graminicola (Keon et al., 2005). The high rate of protein synthesis may be required for pycnidial biogenesis and may also represent an attractive target for fungicidal intervention.

The oleate EST library was dominated by genes involved in lipid and malate metabolism, the trichloroacetic acid cycle, and electron transport. To determine whether the ratio of ESTs in the in planta and oleate libraries was indicative of early- and late-stage infection, transcript levels from eight candidate genes were determined by a quantitative PCR. The genes were chosen from those found exclusively in the in planta library. Transcripts were quantified in cDNA pools made from RNA isolated from in vitro and in planta growth of SN15 at early and late infection time points. The in vitro cultures sampled at 4 DAI did not contain any pycnidia, whereas cultures sampled at 18 DAI were heavily sporulating with many pycnidia. To isolate RNA from both early and late infections of wheat, an SN15 infection latent period assay was used to provide in planta infection transcripts (Solomon et al., 2003b). Lesions were excised after 8 and 12 DAI, representing early and late infection stages. The three genes most upregulated during growth in planta (SNOG_00557.2, SNOG_03877.2, and SNOG_16499.2) all had >20 in planta ESTs, indicating that the EST frequencies were useful predictors of gene expression at these levels. SNOG_00557.2 was the gene with the highest peak expression level in planta and the largest difference between in vitro and in planta conditions. As the gene is predicted to be an arabinofuranosidase, this reaffirmed the important role of carbohydrolases during colonization of the host tissue. Overall, the four tissue types represented clear states of nonsporulation and sporulation during both in vitro and in planta growth. All the genes were found to be expressed in the 12-DAI infected sample, and this was the highest level observed in all but one case (see Supplemental Figure 2 online). On the other hand, moderate to high levels of gene expression were found in the in vitro samples from six of the eight genes. Attempts to use axenic samples to mimic infection has a long history (Coleman et al., 1997; Talbot et al., 1997; Solomon et al., 2003a; Trail et al., 2003; Thomma et al., 2006). Our data indicate that oleate feeding is not a significantly closer model for infection than starvation has proved to be.

Genome Architecture

Colinearity of gene order is a notable feature of closely related plant and animal genomes but is represented in fungi by a complex pattern of dispersed colinearity, such as observed between Aspergillus spp (Galagan et al., 2005). This low degree of organizational conservation can be attributed to the 200 million year separation of these species and the rarity of meiotic events that would tend to maintain chromosomal integrity. S. nodorum is the first dothideomycete to have its sequence publicly released; thus, the closest species for which whole-genome comparisons could be made are the Aspergilli and N. crassa, which are separated from S. nodorum by ∼400 million years of evolution. Thus far, we have been unable to detect significant regions of colinearity with other sequenced genomes, but it is possible that more sensitive methods and the release of more closely related genome sequences will succeed in revealing patterns of chromosomal level evolution.

There is tantalizing evidence of residual colinearity of the genes present in a contiguous 38-kb region of L. maculans DNA (Idnurm et al., 2003). Six of the nine L. maculans genes have orthologs within a 200-kb region of S. nodorum DNA. Gene orientation is retained, but they are interspersed with ∼60 other predicted genes (Figure 4B).

Fungi have a truly ancient history, many times longer than that of animals and flowering plants (Padovan et al., 2005). The extent of genetic diversity is correspondingly large. Phylogenetic analysis of fungi has been hampered by a paucity of reliable morphological indicators. As a consequence, phylogenetic reconstruction of the fungi has been particularly unstable until the widespread introduction of multigene-based DNA sequence comparisons. This study confirms the overall monophyletic characters of the Dothideomycetes and the Pleosporales taxa. These groups contain many thousands of species and are notable for their content of major plant pathogens infecting many important plant families. The origins of this class, likely >400 million years ago, is considerably older than their plant hosts. It is particularly interesting to note that all the fungal plant pathogens that are well established as host-specific toxin producers are in this class (Stagonospora, Cochliobolus, Pyrenophora, M. zeae-maydis, and Alternaria). Taken as a whole, the pathogens in this class are mainly described as necrotrophic, while a few are debatably described as biotrophic (Venturia and C. fulvum) or hemibiotrophic (L. maculans, M. graminicola, and M. fijiensis). None of the species in this class are obligate pathogens, and none possess classical haustoria (Oliver and Ipcho, 2004); furthermore, many other species are nonpathogenic. The genome sequence of S. nodorum represents an important point of comparison from which to derive hypotheses about the genetic basis of pathogenicity in these organisms. Genome sequences of several of these species are in progress (Goodwin, 2004), including A. brassicicola, M. fijiensis, M. graminicola, L. maculans, and P. tritici-repentis, or have been completed but not released (C. heterostrophus; Catlett et al., 2003). Three of these species, S. nodorum, M. graminicola, and P. tritici-repentis, are wheat pathogens. The possibility of multiple pairwise comparisons of gene content between phylogenetically and ecologically close species promises to be a powerful method to derive workably small lists of candidate effector genes controlling pathogenicity, host specificity, and life cycle characters and thus provide ideas for the generation of novel crop protection strategies. The genome sequence provides the tools for global transcriptome analysis, thereby identifying the genes expressed during different phases of infection. This study has also highlighted secreted proteins, which appear to be markedly more numerous than in nonpathogens but which have predominately mysterious roles. Finally, the genome sequence is an essential prerequisite for the critical analysis of hypotheses of interspecific gene transfer. This has already been identified in pathogens in general (Richards et al., 2006) and S. nodorum in particular (Friesen et al., 2006; Stukenbrock and McDonald, 2007) and may emerge as a major feature in the evolution of these organisms.

METHODS

Fungal Strains

Stagonospora nodorum strain SN15 (Solomon et al., 2003b) was used for both genome and EST libraries and has been deposited at the Fungal Genetics Stock Center. The genome sequence was obtained as described (http://www.broad.mit.edu/annotation/genome/stagonospora_nodorum/Assembly.html). The sequences are available for download from GenBank under accession number AAGI00000000.

Phylogenetic Analysis

A combined matrix of 41 taxa was generated from DNA sequences obtained from two ribosomal (nuclear large and small subunit [nuc-LSU and nuc-SSU]) and three protein genes (elongation factor 1 α [EF-1α] and the largest and second largest subunits of RNA polymerase II [RPB1 and RPB2]). Data were obtained from the Assembling the Fungal Tree of Life (AFTOL; www.aftol.org) and GenBank sequence databases, which aligned in ClustalX (Thompson et al., 1997) and manually improved where necessary. After introns and ambiguously aligned characters were excluded, 6694 bp were used in the final analyses. In some cases, genes were missing (see Supplemental Table 3 and Supplemental Data Set 3 online). The resulting data were combined and delimited into 11 partitions, including nuc-SSU, nuc-LSU, and the first, second, and third codon positions of EF-1α, RPB1, and RPB2, with unique models applied to each partition. Metropolis coupled Markov chain Monte Carlo analyses were conducted using MrBayes 3.1.2 (Huelsenbeck and Ronquist, 2001) with a six-parameter model of evolution (generalized time reversible) (Rodriguez et al., 1990) and gamma distribution approximated with four categories and a proportion of invariable sites. Trees were sampled every 100th generation for 5,000,000 generations. Three runs were completed to ensure that stationarity was reached, and 5000 trees were discarded as “burn in” for each. Posterior probabilities were determined by calculating a 50% majority-rule consensus tree of 45,000 trees from a single run. Maximum likelihood bootstrap proportions were calculated by doing 1000 replicates in RAxML-VI-HPC (Stamatakis, 2006) with the GTRCAT model approximation and 25 rate categories with the same data partitions as for the Bayesian runs.

Repetitive Elements

Repetitive elements were identified de novo using RepeatScout v1.0.0 (Price et al., 2005) with a minimum threshold of 10 matches and a minimum repeat length of 200 bp. Newly generated repeats were aligned to the genome assembly via BLASTN v2.0 (Altschul et al., 1990). Hits were discarded if sequence identity fell below 35% or alignment length was <200 bp. For each repeat, the number of filtered hits was counted, and repeats were discarded if the number of hits to the assembly was <10. Prototype repeat regions were defined by identification of sequence similarities among themselves via BLASTN and deletion of redundant repeat sequence. Regions of the genome assembly that matched to a repeat were aligned using MUSCLE v3.6 (Edgar, 2004). To identify repeat type and function, repeats were compared with the nonredundant sequence database hosted by NCBI via BLASTN/BLASTX, and subrepeat regions within repeats were also analyzed. Tandem repeats were identified using Tandem Repeat Finder (Benson, 1999), direct repeats were identified via MegaBLAST (Zhang et al., 2000), and inverted repeats were identified using both MegaBLAST and eINVERTED (Rice et al., 2000).

Putative telomeric regions were predicted by considering scaffold ends containing successive repetitive elements without interspersed predicted protein coding genes. The occurrence of repeat classes within these regions was counted and compared with occurrences throughout the genome. Where >85% of the repeats were found to be at nongenic scaffold end regions, these classes were classified as telomeric repeats. The scaffold ends were predicted to be physical telomeres if they contained three or more telomeric repeats.

To analyze and compare the prevalence of RIP mutation, we developed a program to compare RIP mutations between multiple sequences (J.K. Hane and R.P. Oliver, unpublished data). RIPCAL compares the aligned sequences against a designated model sequence (in this case the trimmed de novo repeats) and was configured to calculate RIP (Dean et al., 2005).

Gene Content

An automated genome annotation was initially created using the Calhoun annotation system. A combination of gene prediction programs, FGENESH, FGENESH+, and GENEID, and 317 manually curated transcripts were used. GENEWISE was used with non-species specific parameters to predict genes from proteins identified by BLASTX. To refine the initial genome annotation, a further 10,752 EST reads were obtained from an EST library from oleate-grown mycelium and 10,751 from S. nodorum–infected wheat (see below for details). These EST sequences were screened against a library of wheat ESTs, trimmed for vector sequence manually, for poly(A) tail sequence using TrimEST (Rice et al., 2000), screened for unusable sequences of poor quality, filtered for remaining sequence of ≥50 bp, and aligned to the genome assembly using Sim4 (Florea et al., 1998). ESTs with multiple genomic locations were assigned their optimum location based on percent identity, total alignment length, and best location of the EST mated pair. Gene models were manually annotated according to optimum EST alignments using Apollo (Lewis et al., 2002).

Genes that were fully supported by EST data were used to train the gene prediction program UNVEIL (Majoros et al., 2003). Second-round gene annotations were created by combining gene predictions with EST data, with EST data replacing predicted gene models, and UNVEIL predictions preferred over first-round predictions. Genes with coding regions <50 amino acids were discarded. The numerical identifiers assigned during the first-round predictions have been retained. New gene models were assigned loci from 20,000 onwards. Updated loci (UNVEIL and EST supported) have been given the numerical suffix 2 (e.g., SNOG_16571.2). The 5354 unsupported genes have retained the numerical identifiers and suffix 1. Where genic regions lack EST support, only the coding sequence is reported.

ESTs not aligned to the assembly by Sim4 were compared with the unassembled reads via BLASTN. EST with matches to unassembled reads with an e-value of <1E-10 were clustered into contigs using cap3 (Huang and Madan, 1999). The assembled ESTs were tested for single open reading frames using getORF (Rice et al., 2000), and possible genes were determined by BLASTP (Altschul et al., 1990) to protein databases at NCBI. One new gene (STAG_20208.1) was identified by this method. We have chosen not to alter genes where gene models conflicted with homologs except where colinearity evidence confirms orthology (Figure 4). Updated files of gene and protein sequences are available from R. Oliver (roliver@murdoch.edu.au).

Proteins were assigned putative functional classes by searching for relevant PFAM (Bateman et al., 1999) domains (Bateman et al., 2004) with HMMER v2.0 (Eddy, 1998) (see Supplemental Data Set 2 and Supplemental Table 2 online). Putative gene ontology was assigned via BLASTP with BLAST2GO (Conesa et al., 2005). Due to the relatively poor identification of certain PKS and NRPS domains using resources like PFAM (Bateman et al., 2004) and CD-SEARCH (Marchler-Bauer and Bryant, 2004; Marchler-Bauer et al., 2005), PKS and NRPS domains and their modular organization were further elucidated with online tools available at NRPS-PKS (Yadav et al., 2003; Ansari et al., 2004). Subcellular localization and secretion was predicted via WoLF PSORT (Horton et al., 2007).

EST Library Construction

For the construction of the in planta cDNA library, wheat (Triticum aestivum) cv Amery was infected with S. nodorum SN15 as a whole plant spray and processed as a latent period assay (Solomon et al., 2006b). Necrotic tissue was excised from the leaves at 10, 11, and 12 DAI for RNA isolation. RNA was isolated by the Trizol method (Sigma-Aldrich).

Material for the oleate library was generated as follows: minimal medium (100 mL) + sucrose (0.5% [w/v]) was inoculated with S. nodorum SN15 pycnidiospores (2.75 × 10⁷) and incubated at 22°C with shaking (130 rpm) for 4 d. Mycelia was harvested, washed, and added to 100 mL of minimal media, including 0.05% (v/v) Tween 80 and 0.2% (w/v) oleate as the carbon source. The culture was incubated as before for 30 h before being harvested, washed, snap frozen in liquid nitrogen, and freeze-dried in a Maxi Dry Lyo (Heto Holten). mRNA was extracted using the Messagemaker mRNA purification-cloning kit (Gibco/Invitrogen) according to manufacturer's instructions. A total of 2.3 μg of mRNA was used as template for reverse transcription to cDNA. The in planta cDNA library was constructed using the SMART cDNA library construction kit (Clontech), and the oleate library was constructed in pSPORT1 (Gibco/Invitrogen). All manipulations were performed according to the manufacturer's instructions. Phage DNA was packaged using the GigapackIII Gold phage packaging system (Stratagene) according to the manufacturer's instructions. Phages were grown, amplified, and mass-excised to bacterial clones according to the provided protocols. The final libraries were both estimated to contain ∼500,000 clones.

Quantitative PCR

Total RNA (1 μg) was reverse transcribed to cDNA using iScript reverse transcriptase premix (Bio-Rad) according to the manufacturer's instructions. RNA from three biological replicates was pooled for a single cDNA synthesis. cDNA reactions were used as PCR template at 1:50 dilution for in vitro–grown SN15 samples and at 1:5 dilution for in planta–grown SN15 samples. Quantitative PCR reactions consisted of 10 μL of iQ SYBR green supermix (Bio-Rad), forward and reverse primers (each at 1.2 μM), and 5 μL of template DNA in a 20 μL reaction. Reactions were incubated in a Rotor-Gene 3000 thermocycler (Corbett Research). Cycling conditions were 3 min/95°C and then 35 cycles of 10 s/95°C, 20 s/57°C, and 72°C/20 s. Amplicon fluorescence from template of unknown concentration was compared with that from genomic DNA standards of 25, 2.5, 0.25, and 0.025 ng/reaction. All reactions were performed with two technical replicates. Data were analyzed using the Rotorgene software version 6.0 (Corbett Research). Primer sequences are listed in Supplemental Table 4 online.

Accession Number

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession number AAGI00000000.

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. Aligned Kyte and Doolittle Hydrophobicity Plots.

• Supplemental Figure 2. Quantitative PCR Analysis of Gene Expression from in Planta–Associated Genes.

• Supplemental Table 1. Mitochondrial Genes.

• Supplemental Table 2. Summary of PFAM Domains Used for Identification.

• Supplemental Table 3. List of Species Used in This Study.

• Supplemental Table 4. PCR Primers for Abundant in Planta–Associated Genes.

• Supplemental Table 5. EST Data for Abundant in Planta–Associated Genes.

• Supplemental Data Set 1. Repetitive Elements: Telomeres, Subrepeats, and Similarity Scores.

• Supplemental Data Set 2. Gene Summary.

• Supplemental Data Set 3. Alignments for Figure 1 as a Noninterleaved Nexus Formatted Text File.

ACKNOWLEDGMENTS

Funding was provided by the Australian Grains Research and Development Corporation (UMU14; we particularly thank James Fortune, John Sandow, and John Lovett for their support of this project), by the Swiss Federal Institute of Technology Zurich, by the Swiss National Science Foundation (Grant 3100A0-104145), by MIT, and by the National Science Foundation (DEB-0228725; Assembling the Fungal Tree of Life). We thank Soeren Rasmussen for help with bioinformatics, Robert Lee for the oleate EST library, and Tim Friesen (USDA) for comments on the manuscript. We dedicate this article to the recently retired Chris Caten who has inspired our work on S. nodorum.

REFERENCES

Author notes