Deciphering the Cryptic Genome: Genome-wide Analyses of the Rice Pathogen Fusarium fujikuroi Reveal Complex Regulation of Secondary Metabolism and Novel Metabolites

The fungus Fusarium fujikuroi causes “bakanae” disease of rice due to its ability to produce gibberellins (GAs), but it is also known for producing harmful mycotoxins. However, the genetic capacity for the whole arsenal of natural compounds and their role in the fungus' interaction with rice remained unknown. Here, we present a high-quality genome sequence of F. fujikuroi that was assembled into 12 scaffolds corresponding to the 12 chromosomes described for the fungus. We used the genome sequence along with ChIP-seq, transcriptome, proteome, and HPLC-FTMS-based metabolome analyses to identify the potential secondary metabolite biosynthetic gene clusters and to examine their regulation in response to nitrogen availability and plant signals. The results indicate that expression of most but not all gene clusters correlate with proteome and ChIP-seq data. Comparison of the F. fujikuroi genome to those of six other fusaria revealed that only a small number of gene clusters are conserved among these species, thus providing new insights into the divergence of secondary metabolism in the genus Fusarium. Noteworthy, GA biosynthetic genes are present in some related species, but GA biosynthesis is limited to F. fujikuroi, suggesting that this provides a selective advantage during infection of the preferred host plant rice. Among the genome sequences analyzed, one cluster that includes a polyketide synthase gene (PKS19) and another that includes a non-ribosomal peptide synthetase gene (NRPS31) are unique to F. fujikuroi. The metabolites derived from these clusters were identified by HPLC-FTMS-based analyses of engineered F. fujikuroi strains overexpressing cluster genes. In planta expression studies suggest a specific role for the PKS19-derived product during rice infection. Thus, our results indicate that combined comparative genomics and genome-wide experimental analyses identified novel genes and secondary metabolites that contribute to the evolutionary success of F. fujikuroi as a rice pathogen.


Introduction
The genus Fusarium is one of the most important groups of phytopathogenic fungi. They infect a broad spectrum of crops worldwide and are responsible for huge economic losses due to yield reductions and mycotoxin contamination. The Gibberella fujikuroi species complex (GFC) constitutes a monophyletic but diverse subgroup of over 50 Fusarium species with similar morphological features. The complex is divided into the African, American and Asian clades, according to DNA-based phylogenetic analyses [1][2][3] (Figure 1A).
The species Fusarium fujikuroi Nirenberg (teleomorph Gibberella fujikuroi (Sawada) Wollenweber) was first described more than 100 years ago as the causative agent of the ''bakanae'' (''foolish seedling'') disease of rice in Japan [2][3][4]. The most characteristic symptom of this disease is excessively elongated seedlings with chlorotic stems and leaves ( Figure 1B). In addition, affected plants are infertile and therefore do not produce edible grains. The disease symptoms result from the ability of F. fujikuroi to produce and secrete gibberellic acids (GAs), a family of plant hormones [5,6]. Today, the fungus is used worldwide for the commercial production of GAs, which are applied extensively in horticulture to regulate plant growth and development [7]. Although GAs control growth in higher plants, they are considered as secondary metabolites (SMs) in Fusarium because they are not essential for fungal growth and development but instead are thought to contribute to the virulence of the pathogen.
Many fusaria, including multiple species in the GFC, are noted for their production of other SMs, particularly pigments and mycotoxins. In fungi, genes responsible for the synthesis of a SM are typically located adjacent to one another in a gene cluster. Such clusters typically include a gene encoding a polyketide synthase (PKS), non-ribosomal peptide synthetase (NRPS) or terpene cyclase (TC) that is responsible for conversion of primary metabolite(s) to a molecule that serves as a precursor for synthesis of a biologically active SM or family of structurally related SMs (e.g. GAs). SM biosynthetic gene clusters can also include genes that encode: 1) tailoring enzymes that catalyze modification of the precursor molecule or subsequent intermediates in a SM biosynthetic pathway; 2) proteins that transport SMs or intermediates across cellular membranes; and 3) pathway-specific transcription factors that typically induce expression of all the genes in a cluster.
The best studied SMs in F. fujikuroi are the diterpenoid GAs. Two major milestones in research on GA biosynthesis in this fungus were the identification of the seven-gene GA biosynthetic gene cluster in F. fujikuroi [8,9] and the discovery that these genes are regulated by the global nitrogen regulator AreA which had not  60 mM glutamine (fusarins), 6 mM NaNO 3 (fusarubins) and 6 mM glutamine (bikaverin); b) Perithecia resulting from a sexual cross of two isolates of F. fujikuroi with opposite mating types. Fusarubins account for the dark color of perithecia [15]; c) F. fujikuroi grown on complete medium and regeneration medium; d) fluorescent microscopy image of the DsRed-labeled F. fujikuroi wild type penetrating rice root cells during infection; e) Microscopic image of microconidial chain on KCl agar medium; f) characteristic symptoms of bakanae disease due to wild-type-infected rice seedlings (left) compared to the GA-deficient mutant (right). doi:10.1371/journal.ppat.1003475.g001

Author Summary
Fungi produce numerous ''secondary metabolites'' (SMs) that are not essential for life but can provide an advantage under natural conditions, e.g. in fungal-host interactions. Here, we conducted the most comprehensive analysis to date of secondary metabolism in fungi using Fusarium fujikuroi. This fungus causes ''bakanae'' disease of rice and is best known for its ability to produce gibberellins (GAs). We show that GA production is limited to F. fujikuroi and provides a selective advantage during infection of the preferred host plant rice. Generation and analysis of a high-quality de novo F. fujikuroi genome sequence combined with comparisons to six other Fusarium genomes revealed the presence of 45 mostly unknown SM gene clusters. We provide a broad spectrum of experimental data including epigenetic, transcriptional, proteomic and chemical product analyses under different nitrogen and pH conditions. Two of the SM clusters (PKS19 and NRPS31) are not present in any other sequenced fungal genome. In planta expression studies revealed that the otherwise silent PKS19 cluster is induced on rice, but not on maize, suggesting a specific role for the PKS19derived product during rice infection. Together, our results demonstrate the tremendous potential of a single fungal species to produce a diversity of SMs that likely contributes to adaptation to environmental changes.
The availability of genome sequences has significantly impacted examination of secondary metabolism in fungi [19][20][21][22][23][24][25][26][27][28][29][30]. To date, publicly available genome sequences of five Fusarium species (F. graminearum, F. oxysporum, F. pseudograminearum, F. solani and F. verticillioides) have aided in silico examination of secondary metabolism in Fusarium. The sequences have facilitated the establishment of a preliminary catalogue of PKS and NRPS genes in Fusarium [31,32], examinations of their phylogenetic relationships [33,34], and bioinformatic identification of novel SM biosynthetic gene clusters [22]. However, none of the functionally characterized gene clusters have led to the identification of any new secondary metabolite that has not been found to be produced by Fusarium spp. before.
Most species in the GFC produce multiple SMs [2], but only a fraction of them have been linked to specific biosynthetic genes. Prior to this study, a complete genome sequence has been available for only one member of the GFC, the African clade species F. verticillioides, which causes ear and stalk rot of maize [22]. The genomes of the American clade species F. circinatum, the cause of pitch canker of pine [35], and the Asian clade species F. mangiferae, that causes mango malformation [36,37] (recently sequenced by S. Freeman and coworkers), are not yet publicly available but were used in the current study for comparison of secondary metabolism both in silico and in laboratory experiments.
The objective of the current study is the comprehensive analysis of secondary metabolism in F. fujikuroi, the cause of ''bakanae'' disease of rice. We present a draft genome sequence and de novo assembly of exceptional quality for F. fujikuroi, a member of the Asian clade of the GFC. We assembled 12 scaffolds that correspond to the 12 previously identified GFC chromosomes [38,39]. In addition to the well-known GA gene cluster, we describe novel genes coding for key SM biosynthetic enzymes, such as PKSs, NRPSs, TCs, dimethylallyl tryptophane synthases (DMATSs), and cytochrome P450 monooxygenases (P450s) and thereby decoded the complete potential of this important species to produce SMs. Our analyses revealed differences and similarities between species of different GFC clades and the more distantly related F. oxysporum ( Figure 1A). By applying a combination of microarrays, ChIP-seq, proteomics, and HPLC-FTMS analyses, we demonstrate that nitrogen availability has an enormous impact on secondary metabolism by affecting gene expression, histone modification patterns, protein composition, and SM product levels. Two of the gene clusters (PKS19 and NRPS31) are not present in any other sequenced fungal genome and thus unique to F. fujikuroi. The forced expression of these unique clusters by genetic engineering led to structural characterization of corresponding metabolites by HPLC-FTMS. In planta expression of the PKS19 gene cluster suggest a specific role for the derived chemical product during rice infection thereby adding a second novel metabolite, in addition to GAs, that may contribute to the ability of F. fujikuroi to uniquely infect rice.

Results/Discussion
The F. fujikuroi genome: General features Whole genome shotgun sequencing of F. fujikuroi (strain IMI58289) by 454 pyrosequencing yielded 0.94 Gb of raw sequence data that was assembled into only 12 scaffolds (N50 of 4.2 Mb; 73 contigs spanning 43.9 Mb with an average read coverage of 19). A total of 14,813 gene models were predicted using a combination of gene prediction tools. Table 1 summarizes physical genome features which are similar to those of closely related species. To assess the completeness of the F. fujikuroi genome draft, we did BLAST searches with two separate highly conserved core gene sets from 39 and 21 higher eukaryotes species, respectively [40,41]. None of the expected single-copy core orthologs were missing from the F. fujikuroi gene model set indicating that the core gene space has been completely covered.
In order to predict protein functions and reconstruction of evolutionary genesis, a Similarity Matrix of Proteins (SIMAP) [42] was generated for the F. fujikuroi gene model set and then queried against the Swiss-Prot (UniProt Consortium, 2011) database. This analysis revealed 390 F. fujikuroi proteins with higher than 80% identity to proteins in the database, while 4,639 F. fujikuroi proteins had little similarity (,10%), indicating novel, species-specific proteins. In a bidirectional best hits (BBH) analysis of the protein set from F. fujikuroi, 71, 77 and 90% of the proteins were .50% identical to protein sets from the closely related species F. verticillioides, F. circinatum and F. mangiferae, respectively. In contrast, only 63% of the F. fujikuroi protein set had .50% identity to a F. graminearum protein set.
In F. fujikuroi, the annotated ORFs account for 49.2% of the genome with an average coding length of 1,457 nt and 2.8 exons per gene; the average exon length is 518 nt. The overall GC content is 47.4%, while the average GC content of ORFs is 51.5%. All of these key genome features are similar to those reported for F. verticillioides (Table 1).
Previous electrophoretic karyotype analysis of F. fujikuroi IMI58289 by contour-clamped homogeneous electric field (CHEF) gel electrophoresis led to the assignment of eleven chromosomes to seven bands. By Southern blot hybridization of the CHEF gel, eight genes were localized to either one or two of the separated chromosome bands [39]. Due to the high quality assembly of the F. fujikuroi genome into 12 scaffolds corresponding to the 12 chromosomes, we were able to assign these eight genes to specific chromosomes and to link the hybridization signals with the number of the chromosomes ( Figure S1).
Despite the shared synteny between the F. fujikuroi and F. verticillioides genomes [22], there are several significant differences worth noting. First, chromosome XII is present in the F. fujikuroi but absent in the genome sequence of F. verticillioides ( Figure 2). However, the lack of chromosome XII in F. verticillioides may be strain-specific as it has been detected in other strains of F. verticillioides [38]. In F. fujikuroi, chromosome XII spans 693 kb and includes 173 predicted genes. A majority (139) of the corresponding predicted proteins lack sequence similarity to annotated Swiss-Prot proteins. Functional enrichment analysis using the MIPS 'FunCat' program [43] revealed that the remaining 35 predicted proteins are enriched (FDR,0.01) in the FunCat category 'guidance of longitudinal cell extension and cell migration' [43]. A second significant difference between the two genomes is that F. fujikuroi chromosome IV lacks 285 and 820 kb from the left and right arms, respectively, relative to F. verticillioides. In the latter species, the 285 and 820-kb regions collectively include 408 predicted genes that share no significant similarity to genes located elsewhere in the F. fujikuroi genome and are enriched in the FunCat categories 'secondary metabolism', 'detoxification' and 'metabolism of melanin' (FDR,0.01). There are also three smaller genomic regions in F. verticillioides that are absent in F. fujikuroi: a 70 kb segment on chromosome VII with 29 genes, a 22 kb segment on chromosome III with eight genes, and a 12 kb segment on chromosome V with six genes. The majority of these genes code for proteins of unknown function.
To determine whether the presence of chromosome XII and the significantly shorter chromosome IV are strain-specific features of F. fujikuroi, we analyzed nine additional F. fujikuroi isolates from different geographic regions by PCR. For analysis of chromosome XII, we employed three primer pairs that amplify fragments from the arms and near the center of chromosome XII. The results from this PCR analysis indicate that chromosome XII is most likely absent in two F. fujikuroi strains (m570 and C1995) from Japan, while only some regions of the chromosome are present in three strains (E289, E292 and E325) from Italy ( Figure S2). The fact that all of the F. fujikuroi strains examined were isolated from infected rice plants and cause ''bakanae'' disease suggests that chromosome XII is not essential for pathogenicity on rice. Whether chromosome XII contributes to niche specificity, as has been observed for the supernumerary chromosomes of F. oxysporum and Nectria haematococca (F. solani MP VI) [22,44,45] is not yet known. Variability in presence and absence of chromosome XII in different F. fujikuroi isolates suggests that it might be an ''accessory'' supernumerary chromosome consistent with previous analyses of GFC species [38]. For analysis of chromosome IV, we designed two heterologous primer pairs based on F. verticillioides sequence present in the 285 and 820-kb regions of chromosome IV that were absent in F. fujikuroi IMI58289. All nine isolates of F. fujikuroi examined have the characteristic shorter chromosome IV observed in strain IMI58289 (Figure 2 and S2). Taken together, these results indicate that within F. fujikuroi, the shortened chromosome IV is species-specific, whereas the presence or absence of chromosome XII is strain-specific.
Overall, the F. fujikuroi genome is low in AT-rich regions, as less than 5% of the genome consists of regions with more than 55% of A+T ( Figure 3A). The main AT-rich region on each chromosome is the centromere (Table S1). After Neurospora crassa [46], F. fujikuroi is the second filamentous fungus with an essentially complete assembly of centromeric sequences. None of the other currently available Fusarium genomes contains more than short pericentric regions flanking the actual, yet unknown centromere sequences [46]. The positions of most centromeres are conserved in F. fujikuroi and F. verticillioides, and in many cases there is synteny of genes close to centromeres, even in the more distantly related F. graminearum, in which the genome has been condensed into four chromosomes (Table S2). Overall, centromeres of F. fujikuroi are significantly shorter than those of N. crassa, ,50-80 kb compared to 150-280 kb, but similar in size to those of fission yeast [46,47]. Centromeric DNA of F. fujikuroi resembles that of N. crassa, as it is rich in transposon relics but lacks highly repetitive alpha-satellite DNA that is typical of centromeric DNA of mammals and plants [46,[47][48][49]. Comparisons with other Fusarium species failed to reveal common centromeric or pericentric DNA sequences, suggesting that these regions can serve as lineage or speciesspecific markers. In F. fujikuroi, none of the terminal contig sequences are capped by telomere repeats, indicating that sequence data for all chromosome ends remain incomplete as for the other sequenced fungi.
In silico analysis of prominent gene families and transposable elements revealed significant differences between related species A genome-wide comparison of prominent gene families in Fusarium genomes suggest that some families are expanded while others are underrepresented in F. fujikuroi relative to other species.
The number of predicted transcription factors (TF) genes in F. fujikuroi is significantly higher (950 genes) in comparison to F. verticillioides (640), F. circinatum (841), and F. oxysporum (876), but almost identical to F. mangiferae (945), the closest relative of F. fujikuroi among the species examined (Table S3). The expansion of the total number of TFs in both F. fujikuroi and F. mangiferae is reflected in the Interpro domain group 'fungal-specific TF/ Zn(2)C6 fungal type DNA binding domain' (IPR007219; IPR001138) (Table S3). This TF gene family is expanded to 235 in F. fujikuroi and 208 in F. mangiferae compared to 90 in F. verticillioides and 144 in F. graminearum. F. fujikuroi has 53 TFs that do not have a closely related homologue (less than 60% identity) in  (Table S3).
To determine the complete set of secreted proteins, including those secreted by both classical and non-classical pathways, we applied a combination of five bioinformatic approaches on the predicted protein sets for F. fujikuroi and four other fusaria. The number of proteins (1,336) in the predicted F. fujikuroi secretome is similar to those for the closely related GFC species F. circinatum, F. mangiferae and F. verticillioides, while F. oxysporum has 15% more proteins (1,541) in its secretome (Tables 2 and S3). Similarly, the number of predicted transporters, including ABC and MFS transporters, and all classes of substrate-specific permeases, are very similar in F. fujikuroi and other fusaria as is the number of histone-modifying enzymes (Tables 2 and S3).
We also determined the coverage of transposable elements (TEs) by scanning for known transposons that have been reported to RepBase [50]. Additionally, we searched for novel TEs in a de novo approach which revealed two LTR-retrotransposons encoding a conserved reverse transcriptase or integrase and one DNAtransposon containing a predicted transposase. These three transposon families are not contained in RepBase (BLAST e value,10610 210 , bitscore .1,000) but were present in genome sequences of the closely related F. oxysporum and F. mangiferae. Overall, TEs constitute 2.2% of the F. fujikuroi genome, which is verticillioides calculated using MUMer [121] highlights overall collinearity. Orthologous DNA is represented by red dots, inverted segments are shown as blue dots. Inset magnifies F. fujikuroi chromosome XII, which has no homologue in the F. verticillioides scaffold set. The missing subtelomeric regions of chromosome IV in F. fujikuroi are highlighted by vertical purple lines. Dots that are located above or below the line indicating collinearity represent largely repetitive DNA. doi:10.1371/journal.ppat.1003475.g002 Figure 3. Characterization of F. fujikuroi chromosomes and variation in acetylation and methylation statues of histone H3. A: Information for chromosomes I, V and VIII is shown as examples of the 12 F. fujikuroi chromosomes (seeFigure S3 for additional chromosomes). For each chromosome, the position of the centromere is shown at the top; below this in descending order are: GC content, location of SM biosynthetic gene clusters, acetylation and methylation states of histone H3, and changes in gene expression. Variation in histone H3 modification statues indicates chromosomal regions in which genes are expressed (H3K9ac and H3K4me2) or silent (H3K9me3). ''D expression up'' indicates a more than twofold increase in gene expression during growth of F. fujikuroi in nitrogen-rich medium, whereas ''D expression down'' indicates an at least twofold decrease in gene expression. SM biosynthetic gene cluster locations are indicated by arrows labeled with the PKS, NRPS or TC (DTC means diterpene cyclase; STC means sesquiterpene cyclase) gene in each cluster (see Table 4 and Table S4). For the same analyses of other F. fujikuroi chromosomes, higher than in F. verticillioides (0.5%) but lower than in F. oxysporum (4.8%) (Table S3).
To estimate the genetic potential of F. fujikuroi to produce SMs, we identified genes predicted to encode five key classes of SMassociated enzymes: PKSs, NRPSs, TCs, DMATSs and P450s. The genes were identified by the presence of characteristic domains in predicted proteins and by BLAST analyses (Tables 2  and S4). We also examined flanking genes to identify putative gene clusters, which could include genes encoding TFs, transporters and modifying enzymes (e.g. dehydrogenases and acyl transferases) in addition to genes encoding the SM-associated enzymes noted above. These genes and potential gene clusters were then compared to homologous sequences in the genomes of F. verticillioides, F. circinatum, F. mangiferae and F. oxysporum as well as the more distantly related species F. graminearum and F. solani [22,[31][32][33][34]45]. This analysis revealed that the F. fujikuroi genome comprises 17 genes that encode putative type I PKSs with canonical ketosynthase (KS) and malonyl-CoA:acyl carrier protein (ACP) transacylase (MAT) domains. Based on domain content, 14 of the predicted PKSs are reducing-type PKSs (R-PKS) in that see Figure S3. B: Immunocytological analysis of histone acetylation and methylation in F. fujikuroi. Detection of specific histone markers was performed with H3K9me3 and H3K9ac-specific antibodies. DNA was counterstained with DAPI. H3K9me3 is significantly enriched in heterochromatin that forms several chromocenters, while H3K9ac is evenly distributed in the nuclei (scale bar = 5 mm). doi:10.1371/journal.ppat.1003475.g003  they have the keto-reductase (KR), dehydratase (DH) and enoyl reductase (ER) domains that catalyze complete reduction of bcarbonyl during polyketide synthesis. Four of the R-PKSs include a NRPS module and are also referred to as PKS/NRPS hybrids. The remaining three PKSs are nonreducing-type PKSs (NR-PKSs) because they lack the KR, DH and ER domains. The F. fujikuroi genome also includes one type III PKS, an enzyme class typical for plants but that has recently been found in some bacteria and fungi [22,51]. The analysis also revealed the presence of 15 NRPS, 2 DMATS, 10 TC (2 diterpene and 8 sesquiterpene cyclases) genes. In total, the F. fujikuroi genome has the potential to encode 45 enzymes that could give rise to 45 structurally distinct families of SMs. SMURF analysis and the manual examination of genes flanking the 45 core SM genes indicate that most are part of a SM gene cluster. Finally, the analysis revealed that the F. fujikuroi genome encodes 143 putative P450s of which 28 are located in putative SM gene clusters ( Table 2).

Genome-wide distribution of histone marks
Recent studies in several fungi showed that chromatin modifications differ in regions with active (euchromatin) and silent (heterochromatin) gene transcription [52][53][54]. There are several examples demonstrating that SM gene clusters in fungi can be regulated by chromatin-modifying enzymes, a form of gene regulation that represents a general level for coordinated control of larger chromosomal segments [53][54][55][56][57]. Modifications of histone proteins can thus serve as markers for changes in chromatin structure associated with gene expression and silencing. For example, gene expression has been associated with acetylation of histone H3 lysine 9 (H3K9ac) and dimethylation of histone H3 lysine 4 (H3K4me2), whereas gene silencing has been associated with trimethylation of histone H3 lysine 9 (H3K9me3) [52][53][54]. In the current study, the presence of acetylated and methylated forms of histone H3 were determined across all F. fujikuroi chromosomes by ChIP-seq analysis with H3K9ac, H3K4me2 and H3K9me3specific antibodies. As in N. crassa, H3K9me3 is mainly present near F. fujikuroi centromeres ( Figures 3A and S3) as pericentric regions have higher levels of H3K9me3 compared to the putative centromeric core regions. Some of this decrease may be due to replacement of canonical H3 with the centromere-specific H3, CenH3, in the centromere cores [46,47]. This pattern sets F. fujikuroi centromeres apart from dispersed large heterochromatic regions that show more uniform enrichment of H3K9me3 (Figures 3 and S3).
Immunofluorescence microscopy with H3K9ac-and H3K9me3-specific dyes revealed that nuclei of F. fujikuroi are highly enriched for H3K9ac (e.g. extensive euchromatin) and with no specific subnuclear distribution ( Figure 3B). In contrast, H3K9me3 was concentrated in discrete areas of nuclei, predominantly at the periphery. Counterstaining with the DNA-binding fluorescent dye DAPI suggested that these are constitute heterochromatic regions, as previously observed in N. crassa [58]. H3K9me1 and H3K9me2 appear to be absent, at least under the conditions used in this study (data not shown).
All putative SM biosynthetic genes identified above were mapped to chromosome I to XI (none were present on chromosome XII) and most were located within subtelomeric regions, loosely defined as within several hundred kb of the putative end of a chromosome ( Figures 3A and S3). These chromosomal regions are often subject to regulation by posttranslational modification of histones, including acetylation or methylation of the N-terminal tail of histone H3 [52][53][54]59].
To assess whether the location of many SM gene clusters within subtelomeric regions corresponds to a punctate enrichment of H3K9me3 and are thus heterochromatic, we performed ChIP-seq analysis applying H3K9me3-specific antibody under nitrogenlimiting conditions that are favorable for the production of GAs and several other SM ( Figure 3A). As mentioned above, H3K9me3 occurred largely in the pericentric and centromeric regions but each chromosome had several additional large regions with high levels of this mark for constitutive heterochromatin. All such regions were associated with AT-and transposon-rich DNA that had few, if any, annotated genes. None of the 45 putative SM enzyme-encoding genes and potential clusters was found in regions enriched for H3K9me3 suggesting that this mark is not important for the regulation of SM in F. fujikuroi. In contrast, H3K9me3 has been proposed to be critical for regulation of SM clusters in Aspergillus nidulans [53,60]. When the two marks associated with active gene expression, H3K9ac and H3K4me2, were observed, no H3K9me3 marks were simultaneously present. H3K9ac and H3K4me2 occurred primarily at the centers or arms of most chromosomes under nitrogen-limiting conditions. Chromosomes X, XI and XII are unusual in that they nearly completely lack enrichment for H3K9ac and H3K4me2, which we attribute to the absence of expressed genes on these chromosomes, at least under the conditions tested. To visualize the global alterations in gene expression in response to a change in nitrogen availability, we mapped the chromosomal positions of genes that are up-or downregulated two-fold or more, as determined by microarray analysis, in F. fujikuroi grown in high vs. low glutamine medium ( Figure 3A). Overall, subtelomeric regions exhibited decreased expression in high nitrogen, while genes in regions associated with H3K4me2 and H3K9ac are generally induced in high nitrogen conditions. Comparative analysis of SM gene clusters: GA gene cluster structure and distribution SMs produced by fungi often play a role in triggering plant cell death and disease [61,62]. Therefore, identification of the whole set of potential SM gene clusters can lead to the development of tools to investigate the role of toxins in disease development. The availability of genome sequences of F. fujikuroi, F. circinatum [35], F. mangiferae (S. Freeman and co-workers, this work), and eleven isolates of F. oxysporum (Broad Institute) provides an opportunity for a more comprehensive analysis of Fusarium SM biosynthetic genes than has been possible previously. Due to the association of GA production with ''bakanae'' disease of rice, GA biosynthetic genes are among the most extensively studied SM genes in Fusarium fujikuroi. Southern and PCR-based surveys indicate that the GA cluster, or parts of it, occurs in some species of the GFC ( Figure 4) as well as some closely related species, e.g. F. foetens, F. napiforme, F. miscanthi, and F. nisikadoi [8,63]. However, production of GAs has only been detected in isolates of F. sacchari, F. konzum and F. proliferatum [64,65]. In addition, the cluster is absent in the more distantly related species F. graminearum and F. solani [66]. Here, analysis of genome sequences revealed that homologues of the entire F. fujikuroi GA cluster are present in F. circinatum, F. mangiferae, and five isolates of F. oxysporum ( Figure 4). All intact clusters share the same gene order and orientation as the previously described clusters in F. fujikuroi and F. proliferatum [8,63,64]. In addition, all of the genes appear to encode functional proteins, with two exceptions: the coding regions of P450-2 and P450-3 in F. oxysporum isolates PHW815 and FOSC 3-a, respectively, are interrupted by premature stop codons ( Figure 4). The seven other F. oxysporum genome sequences have partial GA clusters consisting of one to three intact and partial (i.e. pseudo) GA genes ( Figure 4). While some GA cluster flanking regions share considerable synteny, there are marked differences in content and arrangement of flanking genes among other species (Figure 4). Although some of the flanking genes in F. circinatum are homologues of those in F. fujikuroi, the F. circinatum flanking region has a 38-kb insert and the GA cluster itself is inverted relative to F. fujikuroi ( Figure 4).
Despite some differences in gene content and arrangement in the cluster-flanking regions, the presence of four conserved genes (FFUJ_14327, 14328, 14338, and 14340) in all species/isolates with an intact cluster suggests that significant synteny is conserved in the GFC and F. oxysporum. Furthermore, an ancient GA cluster was present in the ancestral Fusarium genome before divergence of GFC and F. oxysporum. This conclusion is supported by phylogenetic analysis that resolved Fusarium GA genes into a single clade that is distinct from GA genes in other fungi. This result indicates that the GA cluster in all Fusarium species examined likely evolved by vertical inheritance from a common ancestor ( Figure S4). In addition, previously reported PCR and Southern data as well as genome sequence data analyzed here indicate that partial GA clusters are derived via similar patterns of gene loss in three phylogenetically distinct lineages of Fusarium: GFC, F. oxysporum and F. minscanthi/F. nisikadoi [8,63]. In all three lineages, partial GA clusters always lack P450-2, GGS2, CPK/KS and P450-3 ( Figure 4). The similar patterns of gene loss in different Fusarium lineages indicate that degeneration of the cluster is not random. It is not clear what selective pressure(s) would drive non-random degeneration of the GA cluster, however, one possibility is that populations of each Fusarium lineage were introduced into an environment(s) or specified to a new host where GA production was disadvantageous.
Interestingly, functional GA clusters are also present in Sphaceloma (Dothideomycetes) and Phaeospheria (Dothideomycetes) Figure 4. Comparison of GA biosynthetic gene clusters in Fusarium genome sequences. Arrows that are the same color represent genes, or gene sets, that have closely related homologues in two or more species/isolates. Light blue arrows represent GA biosynthetic genes. Black arrows represent genes that do not have closely related homologues in the GA cluster region of other species/isolates. Y indicates a pseudogene. For those that are available, gene/protein designations are indicated next to or below species names. FOXB and FOXG designations are for F. oxysporum isolates Fo5176 and Fol 4287, respectively. doi:10.1371/journal.ppat.1003475.g004 [67,68], both being only distantly related to Fusarium (Sordariomycetes). However, the evolutionary mechanisms by which these fungi acquired GA biosynthetic gene clusters are not yet clear. There is an increasing number of indications that the presence of homologous SM biosynthetic gene clusters in distantly related fungi can result from horizontal gene transfer (HGT) [69]. One example of this is evidence for HGT of the bikaverin biosynthetic gene cluster from Fusarium to the distantly related fungus Botrytis cinerea [70,71].

Gibberellins and their role in plant infection
Phylogenetic studies showed a broad distribution of the GA gene cluster among the genus Fusarium. However, nothing is known about the ability of F. circinatum, F. mangiferae, and F. oxysporum to produce GAs. Therefore, we studied the production of GAs by the different Fusarium strains under four culture conditions that varied in nitrogen availability and pH: 1) nitrogen deficient and acidic (6 mM glutamine); 2) nitrogen deficient and alkaline (6 mM nitrate); 3) nitrogen sufficient and acidic (60 mM glutamine); and 4) nitrogen sufficient and alkaline (60 mM nitrate). We also examined five recently sequenced strains of F. oxysporum which have an intact GA biosynthetic gene cluster in contrast to F. oxysporum 4287 ( Figure 4). Despite the presence of the cluster, GA production was detected only in F. fujikuroi. Lack of GA production in other species could be caused by a number of factors, including mutations that leave ORFs intact but render enzymes nonfunctional, reduced transcription of GA genes, improper GA transcript processing and/or altered translation (Tables 3; S5A and S5B). Although GA production was not detected in F. circinatum, production of ent-kaurene, the first committed intermediate in the GA pathway, was detected ( Table 3). The presence of this metabolite is consistent with the detection of transcripts for CPS/ KS, which encodes ent-copalyl diphosphate/ent-kaurene synthase in F. circinatum ( Figure S5).
To determine whether fusaria with a remnant of the GA gene cluster have retained the regulatory mechanisms required for GA production, we transformed F. oxysporum 4287 with a cosmid clone carrying a wild-type copy of the F. fujikuroi GA gene cluster. As in previous experiments with F. verticillioides [66], transformants of F. oxysporum 4287 with the F. fujikuroi GA cluster produced GAs at levels similar to those produced by F. fujikuroi IMI58289 (Table  S5B).
To explore if plant signals can induce GA gene transcription, we examined the expression of GA genes of Fusarium with an intact GA gene cluster (e.g. F. mangiferae, F. circinatum and some F. oxysporum) during infection of maize by qPCR. No GA gene expression was observed in these fusaria, except for low CPS/KS expression levels in F. mangiferae (data not shown). We also analyzed the expression of CPS/KS by F. fujikuroi during growth on the preferred host plant rice as compared to the non-preferred host maize. As expected, significantly higher expression for CPS/KS was observed in rice than in maize ( Figure 5). These differences in GA gene expression suggest a dependency on specific rice signals as expected for the bakanae fungus.
Although ''bakanae'' disease was described more than 100 years ago, the role of GAs in pathogenesis of F. fujikuroi on rice is not well understood. To determine whether GA production is essential for pathogenesis, we compared the ability to infect and invade rice roots between the GA-producing wild type strain (F. fujikuroi IMI58289) and the nonproducing mutant SG139 that is missing the entire GA gene cluster. Microscopic analysis of infected rice roots revealed that the two strains can equally penetrate the rice root epidermis. Both strains also show the same apoplastic growth behavior within the parenchyma cells of the epidermis and the cortex ( Figure 6B). However, the total number of successfully  invaded symplasts per rice root differed significantly ( Figure 6A,C). While we found 103 events of invasive fungal growth of the wild type inside the symplasts of rice root cells, only seven comparable events by the GA-deficient strain in seven independently analyzed roots were observed. Based on these results, we conclude that GAs contribute to the ability of F. fujikuroi to grow invasively in symplasts of parenchyma cells of rice epidermis and cortex rather than play a role in initial root colonization ( Figure 6A,B,C).
In conclusion, our studies on GA production and GA gene expression clearly show that despite the presence of intact gene clusters in multiple Fusarium species/isolates, the ability to produce GAs and express GA biosynthetic genes is limited to the species F. fujikuroi and may provide a selective advantage during infection of the preferred host plant, rice. In addition, presence of a functional GA gene cluster is not essential for F. fujikuroi to colonize rice root cells, but appears to contribute to further invasion in rice tissue.

Comparative analysis of PKS and NRPS biosynthetic gene clusters
To study the relatedness of F. fujikuroi PKS genes to those from other available Fusarium genomes, we performed a phylogenetic analysis of the predicted KS and MAT domains present in the predicted PKS proteins. The analysis also included KS and MAT domains from other fungal PKSs for which the SM products are known ( Figure S6). Fusarium genomes contain 13-17 PKS genes, but only three are present in all genomes examined: PKS3/FSR1/ PGL1 (FFUJ_06826) which is responsible for production of the fusarubin-derived perithecial pigments [15,31,72,73], and PKS2 (FFUJ_00118) and PKS7 (FFUJ_06260) for which the SM products are not known.
PKS10/FUS1 (FFUJ_10058), is required for fusarin production in F. graminearum, F. verticillioides, F. fujikuroi and the distantly related entomopathogenic fungus Metarhizium anisopliae [16,18,31,75,77,78] (Niehaus et al., unpublished). FUS1 is part of a nine-gene fusarin biosynthetic gene (FUS) cluster that is located in distinct genomic locations in all analyzed genomes ( Figure S10). F. mangiferae lacks this cluster but has a partial, nonfunctional FUS1 that is present in a similar syntenic region as the fusarin cluster in F. verticillioides rather than the syntenic region of the cluster in the more closely related F. fujikuroi. Comparison of gene organization of the FUS cluster among genomes indicates that the cluster has undergone at least one major rearrangement during its evolutionary history. F. fujikuroi, F. verticillioides and F. graminearum have one arrangement of cluster genes, while F. circinatum, F. solani and M. anisopliae have a different arrangement of cluster genes ( Figure S9). Phylogenetic analyses of Fusarium PKSs, including FUS1, ( Figure S6) suggest that the second arrangement, present in both M. anisopliae and some fusaria, is ancestral and that the rearrangement occurred in Fusarium after it diverged from Metarhizium. Possible reasons for the different cluster arrangement in the monophyletic species F. fujikuroi and F. circinatum could either be that the cluster was obtained by a progenitor of either species by HGT or the cluster rearranged in the same manner on multiple independent occasions. Although the SM product of the FUS1 homologue in Trichoderma reesei is unknown, the presence of two additional FUS gene homologues (FUS2 and FUS3) in this fungus ( Figure S10) suggests that it is structurally related to fusarin.
Similar to the fusarin cluster, the gene cluster responsible for fumonisin production is present at distinct genomic positions in F. verticillioides, F. fujikuroi, F. oxysporum (O-1819) and A. niger [79][80][81][82] (this work) ( Figure 7). In contrast to the fusarin cluster, the synteny of the fumonisin cluster is perfectly conserved in the three fusaria, with the exception that FUM20 is absent and FUM17 is present as a pseudogene in F. fujikuroi. In A. niger the FUM gene cluster consists of eleven genes arranged markedly different than in Fusarium ( Figure 7). The presence of the FUM cluster in Aspergillus has been attributed to horizontal gene transfer from a distantly related Sordariomycete [83]. Two F. fujikuroi PKS genes of unknown function are located 15 kb apart: PKS17 (FFUJ_12066) and PKS18 (FFUJ_12074) ( Table S4). Homologues of PKS17 and PKS18, as well as their flanking genes, are present in F. mangiferae, and the arrangement of the homologous genes is highly conserved in the two species ( Figure S11). PKS17, a reducing PKS, and PKS18, a nonreducing PKS, share homology to the two PKSs, AfoG and AfoE respectively, involved in asperfuranone biosynthesis in Aspergillus nidulans [84] ( Figure S6). In the proposed biosynthetic pathway for asperfuranone, the C-8 polyketide product of AfoG binds to the starter unit: ACP transacylase (SAT) domain of AfoE which then completes the synthesis of the asperfuranone precursor [84]. The lack of a SAT domain in PKS18 suggests that either the protein is non-functional or that the malonly:ACP transacylase (MAT) domain serves to bind both starter unit, potentially synthesized by PKS17, and extender units during polyketide synthesis.
We also identified a PKS gene (FFUJ_12239 = PKS19) in F. fujikuroi that lacks a closely related homologue in the other Fusarium genomes ( Figure S6; Figure 8; Table S4). PKS19 is part of a putative six-gene SM cluster (PKS19 cluster) that is embedded within an AT-rich region of chromosome VIII (Table S4; Figure 3A; Figure 8). Like PKS19, the five other genes in the putative PKS19 cluster do not have closely related homologues in other Fusarium genome sequences examined. However, genes flanking the AT-rich region have closely related homologues, arranged in a highly syntentic manner, in the other fusaria Figure 6. Infection of roots of rice plants by GA-producing wild-type (IMI58289) and GA-deficient (SG139) strains of F. fujikuroi. Rice plants were co-cultivated with either strain of F. fujikuroi for 7 days at 28uC and 80% humidity. Images of the corresponding whole plants following the incubation period are shown in Figure 1Bf. Both the wild-type strain and GA-deficient mutant were engineered to express the dsRed fluorescent protein. Stars indicate the position of cells that have been penetrated by and are filled with hyphae of F. fujikuroi. A: Microscopic overviews (10-fold magnification) of a root infected with the mutant SG139 (above) or wild-type. Note the absence of invaded cells in the root infected with the GAdeficient mutant. Images are an overlay of images from brightfield and Texas Red filter, which captures fluorescence emitted by the DsRed protein. B/ C: 40-fold (B) and 63-fold (C) magnified images of roots infected with GA-deficient mutant (left) or wild-type strain (right). In most cases, hyphae of the mutant strain were observed between cells, in 'intercellular' spaces, whereas wild-type hyphae were often observed inside the cells (indicated by stars; see also Figure 1Bd) as well as in intercellular spaces. White lines in the images are scale bars corresponding to 200 mm (B) or 50 mm (C). D: Quantification of penetration events per rice root infected with wild-type or SG139, respectively. In addition, the total number of shots per root taken for analysis as well as penetration events of the mutant compared to the wild-type are shown. doi:10.1371/journal.ppat.1003475.g006 ( Figure 8). Together, the presence of the cluster in F. fujikuroi, its absence in closely related species, and the high level of synteny of the cluster flanking genes suggest that the PKS19 cluster was introduced into the F. fujikuroi genome relatively recently, possibly by HGT.
Although there are no closely related homologues of F. fujikuroi PKS19 in the F. circinatum, F. mangiferae, F. verticillioides or F. oxysporum genome sequences, the genomes have PKS8 homologues, which have 55-58% nucleotide identity to PKS19 ( Figure S6). In addition, the two genes adjacent to PKS8 are predicted to encode Horizontal arrows that are the same color represent genes, or gene sets, that have closely related homologues in two or more fungi. Blue horizontal arrows represent FUM genes, and the numbers in them correspond to FUM gene designations rather than designations from genome databases. Y indicates a pseudogene. Gene content and order is well conserved in the Fusarium species but less conserved in A. niger. doi:10.1371/journal.ppat.1003475.g007 Figure 8. Comparison of the PKS19 (left) and PKS8 (right) clusters in genome sequences of Fusarium. Horizontal arrows that are the same color represent genes, or gene sets, that have closely related homologues in two or more fungi. The putative PKS19 cluster is embedded within an AT-rich region, consists of six genes (FFUJ_12235-12250; horizontal green arrows), and among the species examined is unique to F. fujikuroi. However, synteny of genes corresponding to PKS19 flanking genes in F. fujikuroi is highly conserved in F. mangiferae, F. verticillioides, F. circinatum and F. oxysporum. The vertical red arrows indicate the genomic location corresponding to the location of the PKS19 cluster in F. fujikuroi. The putative PKS8 cluster consists of three genes (green horizontal arrows). Only a remnant of PKS8 is present in F. fujikuroi, whereas the intact PKS8 cluster is present in the other species examined. The synteny of the PKS8 cluster-flanking genes is partially conserved among the fusaria examined. For those that are available, gene designations are indicated below species names. doi:10.1371/journal.ppat.1003475.g008 the same types of proteins as the two genes adjacent to PKS19 (DltD-domain protein and a ToxD-like protein; Figure 8), even though the nucleotide identities for the genes between species are low, 52-60%. These observations suggest that PKS8 and PKS19 and the two adjacent genes constitute homologous, but relatively distantly related, biosynthetic gene clusters. Given the similar gene content of the two clusters, their SM product(s) could be similar in structure and perhaps even biological activity. The distribution of the PKS8 and PKS19 clusters is mutually exclusive among the species examined. However, F. fujikuroi has a remnant of PKS8 in a genomic region that is syntenic to the regions flanking the PKS8 cluster in the other fusaria ( Figure 8). This suggests that an ancestral F. fujikuroi genome had an intact PKS8 cluster that subsequently degenerated. It is not clear whether intact PKS8 and PKS19 clusters existed simultaneously in the ancestral genome.
Homologues of the PKS12 gene (FFUJ_10347) have a pattern of distribution similar to that of PKS8 and PKS19. Phylogenetic analysis resolved two types of PKS12 homologues, PKS12 and PKS12a ( Figure S6), but the nucleotide identity between the types is low (58-66%). The presence of a methyltransferase-encoding gene upstream of PKS12 and PKS12a suggests a two-gene SM biosynthetic cluster ( Figure S12). Distribution of the PKS12 and PKS12a clusters is mutually exclusive in the fusaria examined. However, F. oxysporum strain Fol4287 has both an intact PKS12a cluster and remnants of the PKS12 cluster ( Figure S12) in a region that is syntenic to the flanking regions of the PKS12 cluster in other fusaria. This indicates that a PKS12 cluster was present in an ancestral F. oxysporum genome and subsequently degenerated similarly to the PKS8 cluster. Complexity of PKS12/PKS12acluster evolution is further evident by the presence of two paralogues of the PKS12a cluster in F. oxysporum strain Fol4287 ( Figure S12). Apart from these paralogues, the similarities in distribution of the PKS12/PKS12a and PKS8/PKS19 clusters among GFC and F. oxysporum raise the question if distribution of distantly related homologous PKS clusters is mutually exclusive, and this is exclusivity related to similarities in structures and or biological activities of the cluster SM products. Elucidation of the SM products of the PKS12/PKS12a and PKS8/PKS19 clusters and analysis of the distribution of these clusters in additional fusaria should aid in answering this question.
In order to assign some Fusarium NRPSs to a putative function, a BLAST analysis of the predicted NRPSs identified in the Fusarium genomes was conducted. A homologue of NRPS1 (Table S4), responsible for production of the siderophore malonichrome associated with iron uptake and present in F. roseum, F. graminearum, F. culmorum and F. pseudograminearum [85][86][87] is absent in F. fujikuroi and F. mangiferae. The homologous gene in F. verticillioides is located on the portion of chromosome IV that is missing in both F. fujikuroi ( Figure 2) and F. mangiferae. The ability of F. fujikuroi to thrive despite missing an NRPS1 homologue likely reflects the underlying redundancy in iron uptake systems recently described [88].
F. fujikuroi NRPS22 (FFUJ_09296) provides an example of some challenges associated with predicting NRPS function. F. fujikuroi, F. mangiferae, F. circinatum and F. oxysporum possess an NRPS (NRPS22) homologue with two adenylation (A)-domains (Table S4; Figure  S13). The first is most similar to the first A-domain in the enniatin NRPS, EsyN, previously described in F. lateritium, F. sambucinium and F. scirpi [89,90] while the second NRPS22 A-domain is more similar to the second A-domain in the beauvericin NRPS, BeaS, from Beauveria bassiana [91]. Recent functional characterization of NRPS22 in F. oxysporum revealed that it is required for beauvericin production [92] and which immediate suggests that the F. fujikuroi, F. mangiferae and F. circinatum NRPS22 homologues also confer the ability to produce beauvericin rather than enniatin. It is interesting to note that the first A-domain in BeaS and EsyN is responsible for activation of the same amino acid and that this domain in the F. fujikuroi, F. mangiferae, F. circinatum and F. oxysporum homologues is more similar to EsyN than to BeaS ( Figure S13).
Like PKS19, NRPS31 (FFUJ_00003) was present in the F. fujikuroi genome sequence but not in the sequences of the other Fusarium species examined ( Figure 9). However, there is syntenic conservation of the regions flanking the NRPS31 cluster in the genome sequence of all the other Fusarium species except for F. solani (Figure 9). Functional characterization of this gene cluster will be shown below.
The portion of chromosome IV in F. verticillioides that contains NRPS1 and that is missing in F. fujikuroi, also contains a DMATS (FVEG_12218 = DMATS4 in Table S4). As noted above for NRPS1, the lack of this DMATS in F. fujikuroi implies that it is either redundant or that it is not required for growth.
In summary, the presence of highly similar PKS and NRPS gene clusters or remnants of PKS and NRPS-encoding genes in Fusarium species is an indication for their presence in ancestral Fusarium, and the absence of these clusters in other fusaria indicates a subsequent complete or partial loss during the evolution of the genus. On the other hand, the presence of highly similar gene clusters in distantly related fungi, probably distributed by HGT, can facilitate identification of their products.

Does the presence of PKS and NRPS gene clusters correlate with product formation?
In addition to GAs, culture filtrates of the five Fusarium strains (F. fujikuroi IMI58289, F. circinatum Fsp34, F. mangiferae MRC7560, F. verticillioides M-3125 and F. oxysporum 4287) were also analyzed for the PKS-derived metabolites bikaverin, various fusarins, Omethylfusarubin, fumonisin B 1 and fusaric acid, as well as for the NRPS-derived metabolite beauvericin (Table 3; Table S6). The optimal conditions for the production of the different metabolites varied considerably. For example, fusarins were only produced under high nitrogen conditions (60 mM glutamine; 120 mM NaNO 3 ), while bikaverin and fusarubins were only or mainly formed under acidic low nitrogen (6 mM glutamine) or alkaline low nitrogen (6 mM NaNO 3 ) conditions, respectively (Table S6). Fumonisin B 1 was produced under acidic low nitrogen (6 mM glutamine) by F. fujikuroi IMI58289 and F. verticillioidesM-3125, both of which harbor the entire cluster. Surprisingly, F. fujikuroi produced only very low levels of fumonisin B 1 compared to F. verticillioides despite the high (.90%) nucleotide sequence identity of FUM cluster genes in the two species (Table S6).
To determine whether the F. fujikuroi fumonisin cluster is induced in planta and to compare any expression observed with the expression of the F. verticillioides fumonisin cluster in planta, we examined the expression of the PKS-encoding gene FUM1 on rice and maize in both fungi by qPCR. Although significant differences in the level of transcription were observed after growth on the two plant species, the pattern of expression were similar. Specifically, FUM1 was more highly expressed on maize than on rice in both strains and more highly expressed by F. verticillioides than F. fujikuroi ( Figure S14). The higher and specific expression on maize by F. verticillioides could reflect the requirement of fumonisin for late stages of infection [93,94]. Interesting, although F. fujikuroi is not a major pathogen of maize, we detected higher expression of FUM1 in planta than in vitro, perhaps reflecting a similar Fusarium response to maize signals ( Figure S14).
Overall, our data suggest that the presence of orthologous SM gene clusters in closely related Fusarium species does not always correlate with production of the corresponding SM(s), at least under laboratory conditions. For example, although F. circinatum Fsp34 has the FUS cluster, it did not produce fusarins, and while F. oxysporum 4287 has the FSR cluster, it did not produce Omethylfusarubin in our hands (Table S6) (Figures S10 and S7). Genes in some clusters are expressed only or at higher levels in planta than in culture, and in some cases, levels of expression in planta were dependent on the host plant, as shown for fumonisin and GA biosynthetic genes. Similar observations have been made in other fungi, e.g. in A. nidulans where non-standard growth conditions activated otherwise silent gene clusters [95,96].
Nitrogen availability affects the transcriptome, proteome and histone modification patterns Since biosynthesis of multiple F. fujikuroi SMs is regulated by nitrogen availability and pH, we analyzed the transcriptome of the fungus grown under the same conditions used to assess SM production, namely acid low nitrogen, acid high nitrogen, alkaline low nitrogen, and alkaline high nitrogen using Roche-NimbleGen DNA microarrays. We used high quality 126135 K Nimblegen microarrays that were manufactured based on the present genome annotation of F. fujikuroi IMI58289 and performed a genome-wide search for all nitrogen and/or pH-affected genes, but specifically focused on expression patterns of all annotated SM gene clusters (Table S4).
Based on the selection criteria 2-fold change in expression at the 95% confidence interval, 3,117 and 3,242 genes were up-regulated under acidic and alkaline low-nitrogen conditions, respectively. Up-regulation of 2,494 of these genes occurred under both lownitrogen conditions; whereas up-regulation of 560 of the genes occurred only in the acidic condition, and up-regulation of 717 occurred only in the alkaline condition. 63 genes were upregulated in the acidic condition but down-regulated in the alkaline low nitrogen condition; and 31 genes were up-regulated in the alkaline condition but down-regulated in the acidic condition ( Figure S15, Table S7). 'FunCat' analysis indicated an overrepresentation of genes likely involved in transport, carbon metabolism, and detoxification among the low nitrogen-induced genes (Table S7). This included genes known to be nitrogen-repressed, e.g. the nitrite reductase NIIA (FFUJ_06099), the ammonium permeases MEPA, MEPB and MEPC (FFUJ_01144; FFUJ_11805; FFUJ_13836), and a nitrate transporter gene CRNA (FFUJ_00934) [11,97,98].
Under high-nitrogen conditions, there was up-regulation of many genes involved in primary metabolism, e.g. amino acid metabolism, DNA processing, transcription, transport, protein synthesis, and protein folding and modification (Table S7). In total, 3,860 and 4,192 genes were up-regulated under acidic and alkaline high-nitrogen conditions, respectively. Of those genes, 3,021 were up-regulated in both acidic and alkaline high-nitrogen conditions, while 808 were up-regulated only in the acidic and 1,108 were up-regulated only in the alkaline high nitrogen condition ( Figure S15, Table S7).
We were able to identify at least one condition that induced relatively high levels of expression of genes within 30 of the 45 F. fujikuroi SM clusters: 13 clusters with a PKS gene, the two with a diterpene cyclase (DTC) gene, two with a sesquiterpene cyclase (STC) gene, 11 with an NRPS gene, one with a DMATS gene, and the one with the type III-PKS gene. These data are based on differences in expression of putative SM cluster genes in response to low and high nitrogen (Table S7). Among the SM gene clusters for which the corresponding SM product(s) has not yet been identified, genes in the PKS16, STC4, NRPS20, and NRPS21 clusters were expressed at higher levels under both acidic and alkaline low-nitrogen conditions (Tables 4 and S7). Genes in the  NRPS3 cluster were expressed under the acidic low-nitrogen (glutamine) condition, whereas genes in the NRPS13 were expressed under the alkaline low-nitrogen (nitrate) condition.
To determine whether transcription and protein levels are correlated in F. fujikuroi, we quantified whole cell protein extracts from the fungus grown under acidic low-and high-nitrogen conditions. To do this, we employed metabolic labeling and quantitative proteomics. The fungus was grown for three days under acidic low (6 mM glutamine) or high nitrogen (60 mM glutamine) conditions. High-nitrogen cultures were grown with 14 N glutamine, whereas low-nitrogen cultures were labeled with 15 N glutamine. 14 N glutamine and 15 N glutamine-grown mycelia were harvested, mixed on equal protein basis, and then fractionated by SDS-PAGE. Protein bands were excised, in-gel digested with trypsin, and peptides were then subjected to liquid chromatography coupled with high resolution mass spectrometry (LC-MS/MS). In total, two independent biological replicates (A and B) were analyzed resulting in the identification of 2,808 distinct proteins (18.9% of all predicted F. fujikuroi ORFs). 2,060 proteins were present in both biological replicates, and 1,644 of these could be quantified in both replicates. Of the 1,644 proteins quantified, 418 were down regulated (0.5-fold and lower), 618 exhibited weak or no change in abundance, and 347 were upregulated (2-fold and higher). 261 proteins have a too high divergence between replicate A and B, resulting in missing mean ratios.
'FunCat' analysis indicated that proteins likely to be involved in 'nitrogen, sulphur and selenium metabolism' and 'secondary metabolism' were significantly overrepresented in the highnitrogen condition (FDR,0.05), whereas proteins likely to be involved in 'virulence, disease and defense' and 'secondary metabolism' were significantly overrepresented in the lownitrogen condition, considering differential expressed genes according to transcriptomics (2-fold higher or lower, FDR,0.05) and proteomics data (4-fold higher or lower). It is notable that the overrepresentation of 'secondary metabolism' proteins under both low and high-nitrogen conditions differs when allowing for transcriptomics data alone, where 'secondary metabolism' genes were not overrepresented in either condition. The detection of a moderate correlation between fold changes of transcript and protein ratios (Pearson = 0.45, Spearman = 0.36, p,0.01) indicate that distinctive transcriptional and post-translational control mechanisms exist in F. fujikuroi (Table S7). In A. fumigatus, a similar comparison of microarray and RNA-sequencing data with 2D-gel protein quantification data was recently presented [99]. These data underscore the power of combining transcriptome and proteomic analyses and indicate that LC-MS/ MS protein quantification provides a greater set of quantifiable proteins than conventional 2D-gel approaches. For full proteome coverage, extensive subcellular and protein fractionation techniques will be required as described for the S. cerevisiae proteome project [100].
Notably, we found proteins from ten of the putative SM clusters that were significantly enriched under acidic conditions with low or high nitrogen ( Figure 10). There were six STC, three NRPS and two PKS clusters for which expression of genes and proteins was not detected under any of the four growth conditions examined suggesting that these clusters are not or not solely regulated by nitrogen availability or pH (Table 4; Figure 10). No proteins encoded by SM cluster genes were identified that are exclusively expressed under alkaline conditions (PKS2, STC1, NRPS2, NRPS6, NRPS10 and NRPS11) ( Table 4) as the proteomic approach was performed only under acidic low and high nitrogen conditions. As expected from our previous studies [10], GA cluster genes were expressed under low nitrogen conditions (Table 4 and S9), and this expression paralleled abundance of GA proteins ( Figure 10). The fumonisin cluster was expressed under both acidic and alkaline low-nitrogen conditions, and accordingly five fumonisin proteins were identified under low-nitrogen conditions (Tables 4 and S7; Figure 10). Also, the recently reported expression of the fusarin C (PKS10/FUS) cluster under high nitrogen conditions [78] (E.-M. Niehaus et al., unpublished) was confirmed here by genome-wide microarray analysis and enrichment of the corresponding proteins (Table 4; Figure 10). Furthermore, gene expression of the NRPS31 cluster, which was observed only in the F. fujikuroi genome, was correlated with the presence of the corresponding proteins under acidic high-nitrogen condition. For the remaining SM clusters (PKS1, PKS12, PKS14, PKS17/18, PKS19, PKS20, NPS25, DMATS1 and PKS type III) only slight tendencies for regulation by nitrogen under one of the four growth conditions were observed (Table 4). Likewise, there was no significant enrichment of the corresponding proteins under any of the conditions examined ( Figure 10).
Summarizing, we performed microarray analyses under our four standard conditions: low-nitrogen and high-nitrogen at both acidic and alkaline pH. We were able to identify at least one condition that induced expression of 30 out of the 45 SM gene clusters. Comparison of our microarray and proteomic analyses revealed a correlation between gene expression and presence of the corresponding proteins for nine of the SM gene clusters that were expressed under acidic low-or high-nitrogen conditions. Both expression of NRPS31 cluster genes and abundancy of the respective proteins were elevated under the acidic high-nitrogen condition. The tendencies of expression of SM cluster genes to correlate with the level of the corresponding proteins indicates that genome-wide expression analyses are a powerful approach for identifying conditions that induce production of unknown SMs.
Does gene expression correlate with histone modifications?
As mentioned above, we determined whether there is a correlation between histone modifications and gene expression across the genome. This analysis also allowed us to examine whether such a correlation exists for genes throughout the F. fujikuroi genome as well as genes located in the 45 SM clusters (Figure 3; Figure S3). For this analysis, we performed ChIP-seq with two of the standard conditions: low-and high-nitrogen acidic conditions (6 mM and 60 mM glutamine, respectively) with antibodies specific to two activating (H3K9ac and H3K4me2) and one silencing (H3K9me3) modifications.
First, we examined whether gene expression was correlated with the presence of the histone activation marks (H3K9ac and H3K4me2), 1) expressed under acidic low-nitrogen conditions and the presence of activation mark H3K4me2, but repressed under acidic high-nitrogen and absence of activation mark H3K4me2, 2) repressed under acidic low-nitrogen and the absence of activation mark H3K9ac, but expressed under acidic high-nitrogen condition and presence of activation mark H3K9ac (Table S7), 3) repressed under acidic low-nitrogen and the absence of activation mark H3K4me2, but expressed under acidic highnitrogen and presence of activation mark H3K4me2. There was a relatively low positive correlation between gene expression and the presence of the activating histone marks H3K9ac and H3K4me2 for all genes in the genome. Overall, the correlation with H3K9ac and expressed genes was better than that of H3K4me2 with expressed genes, which had been observed also by genome-wide analyses in Trichoderma reesei [101]. However, when comparing the presence of histone marks and transcriptome data specifically for SM gene clusters, the correlation was much better. Thus, the presence of the histone mark H3K9ac was correlated with gene expression across the GA gene cluster at the low-nitrogen condition, while H3K9ac was almost completely absent under the repressing high-nitrogen conditions (60 mM glutamine) (Table S7; Figure 11). The same histone pattern was observed for both the bikaverin ( Figure 12) and fumonisin ( Figure S16) genes clusters: enrichment of H3K9ac under inducing conditions (low nitrogen) and low levels of H3K9ac under repressing conditions (high nitrogen).
Taken together, these findings strengthen the hypothesis that histone acetylation is associated with gene transcription also in F. fujikuroi. A similar correlation between acetylation of histones associated with SM gene clusters and expression was also observed for other fungi, e.g. A. parasiticus and A. nidulans [51,57,102,103]. However, not all gene clusters that exhibit histone acetylation are transcribed, and not all transcribed clusters exhibit histone acetylation. For example, the fusaric acid (PKS6) cluster exhibited little enrichment for H3K9ac under the inducing condition ( Figure  S16A).
In contrast to H3K9ac, H3K4me2 enrichment was observed only for the GA gene cluster. Here, two of the seven GA biosynthetic genes, P450-4 and P450-2, were enriched for H3K4me2 under inducing low-nitrogen conditions ( Figure 11).
These results imply that the state of H3K9 acetylation, and also H3K4 methylation, alone is not sufficient for transcriptional activation but that other factors, including basal or specific transcription factors, additional histone modifications, or plant signals can also regulate expression of cluster genes. Furthermore, up-regulation of a gene cluster does not always result in high levels of production of the corresponding SM(s). For example, lownitrogen conditions induce high levels of expression of fumonisin biosynthetic genes in both F. fujikuroi and F. verticillioides. Although this gene expression is accompanied by relatively high levels of fumonisin production in F. verticillioides it is accompanied by only low levels of production in F. fujikuroi (Table S6) [104].
In conclusion, analysis of the histone marks H3K4me2 and H3K9ac that are typically associated with sites of active transcription, combined with genome-wide expression analysis can serve as a powerful tool to identify new SM gene clusters and culture conditions that induce their expression.

Discovery of two novel gene clusters in F. fujikuroi
Genome sequence comparisons revealed that the F. fujikuroi PKS19 and NRPS31 clusters are not present in the other Fusarium species examined (Figures 8 and 9). Thus, we conducted additional analyses to further define these clusters and to obtain evidence for the corresponding SM metabolites. BLAST analysis revealed that NRPS31 is closely related to the NRPS APS1 (66-71% identical) in F. semitectum (Table S4). In F. semitectum, the APS1 gene is part of a 12-gene cluster responsible for synthesis of apicidin, a histone deacetylase inhibitor with antiparasitic activity [105,106]. In F. fujikuroi, the organization of genes adjacent to NRPS31 (Table S4) is largely syntenic to the organization of genes in the APS cluster in F. semitectum and will be referred to hereafter as APS1 to APS9, APS11 and APS12, respectively ( Figure 13A). Exceptions to this are absence of an APS10 homologue and inversion of the order and orientation of APS2 and APS3 in F. fujikuroi ( Figure 13A). To our knowledge, apicidin production has not been reported in F. fujikuroi. In addition, chemical analyses performed during the course of the current study did not provide evidence for production of apicidin by F. fujikuroi.
Microarray analysis revealed expression of the F. fujikuroi APS cluster homologue under the acidic high-nitrogen condition ( Figure 13B; Table 4; Table S7). In accordance with the microarray data, APS cluster genes were slightly enriched for H3K9ac in the high-nitrogen condition, while almost no H3K9ac was present in the acidic low-nitrogen condition. Only the APS5 homologue was enriched for H3K4me2 in the high-nitrogen condition, and no H3K4me2 enrichment occurred in the low-nitrogen condition ( Figure 13B). The APS cluster did not exhibit enrichment of H3K9me3, which can be indicative of gene silencing, under the low-nitrogen (non-inducing) condition. However, there was enrichment of the AT-rich region immediately downstream of APS1 in low nitrogen. Nevertheless, microarray data and H3K9ac enrichment of the cluster region provide evidence that the acidic high-nitrogen condition can activate expression of the APS cluster homologue in F. fujikuroi.

Functional analysis of F. fujikuroi novel gene clusters
In order to determine the metabolites synthesized by the two novel clusters, we employed our understanding of SM synthesis and a variety of targeted gene expression and natural product characterization approaches. First, a more detailed analysis of sequences of the F. fujikuroi APS homologues indicated that the APS8 gene might be defective because it encodes a protein (Aps8) that is 125 amino acids shorter at the N-terminus compared to the Aps8 protein of F. semitectum. To determine whether the potentially defective APS8 gene was responsible for lack of apicidin production in F. fujikuroi, we generated a strain of F. fujikuroi that overexpressed the F. semitectum APS8 homologue. However, overexpression of the F. semitectum APS8 did not induce apicidin production when the strain was grown in the acidic high-nitrogen condition ( Figure 13C). In F. semitectum, overexpression of the APS2 gene, which encodes a positive-acting Zn(2)C6 transcription factor, enhanced apicidin production [106]. Therefore, we generated an additional strain of F. fujikuroi that overexpressed both F. fujikuroi APS2 and F. semitectum APS8. HPLC-FTMS analysis of this mutant revealed that it was significantly enhanced, compared to the wild type, in production of a metabolite with the molecular formula C 35 H 42 O 7 N 5 , which is similar to C 34 H 48 O 6 N 5 , the molecular formula of apicidin ( Figure 13C). The possibility that the C 35 H 42 O 7 N 5 metabolite (compound 2 in Figure 13C and 13D) is structurally similar to apicidin is also supported by the similar UV spectra of the two metabolites ( Figure 13D). Hydrolysis of compound 2 was carried out to determine its amino acid composition. Apicidin consists of the four amino acids (S)-Nmethoxy-tryptophan, (R)-pipecolic acid, (S)-2-amino-8-oxodecanoic acid and (S)-isoleucine. In contrast, compound 2 is composed  of three amino acids, (S)-N-methoxy-tryptophan, (R)-pipecolic acid, phenylalanine and an fourth amino acid-like material with a molecular formula of C 8 H 15 NO 4 . Thus, two of the four amino acids that make up compound 2 differ from those that make up apicidin. These differences in amino acid composition could result from differences in substrate specificities of the F. fujikuroi and F. semitectum APS1 homologues that catalyze condensation of different amino acids during biosynthesis. The absence of an APS10 homologue in F. fujikuroi could also contribute to the structural difference if it is responsible for modification of an amino acid.
As noted above, the minimal PKS19 cluster consists of genes encoding a PKS, a DltD-domain protein, and a ToxD-like protein. However, based on proximity, the cluster may include three additional genes; one that encodes a putative membrane protein (FFUJ_12242), a second that encodes a putative Zn(2)C6 transcription factor (FFUJ_12243), and the third that encodes a putative P450 monooxygenase (FFUJ_12244) (Figures 8, 14). PKS19 is a reducing PKS, a subclass of PKSs that typically catalyze synthesis of polyketides with a fully reduced carbon skeleton. Thus, it is unlikely that the SM product of the PKS19 cluster is a polycyclic aromatic compound.
Microarray analysis indicated that two and three of the six PKS19 cluster genes were expressed in alkaline and acidic highnitrogen conditions, respectively, albeit at relatively low levels ( Table 4; Table S7). The PKS19 cluster exhibited minimal H3K9ac under high nitrogen and only slightly more under lownitrogen conditions, which is consistent with the overall low expression levels observed for the PKS19 cluster by microarray analysis ( Figure 14A). Only the transcription factor gene FFUJ_12243 was enriched for H3K4me2. The AT-rich regions flanking the ends of the cluster and between FFUJ_12243 and FFUJ_12244 were enriched in H3K9me3, which is associated with heterochromatic/gene silencing.
To enhance expression of the PKS19 cluster genes and potentially induce production of the corresponding SM product(s), we generated strains of F. fujikuroi that overexpressed PKS19 (OE::PKS19) and/or the putative Zn(2)C6 transcription factor gene (FFUJ_12243) Overexpression of PKS19 had no effect on expression of the cluster nor did it noticeably alter the metabolic profile of F. fujikuroi. In contrast, overexpression of the putative transcription factor gene (FFUJ_12243) enhanced expression of PKS19 (FFUJ_12239), FFUJ_12240, FFUJ_12242, and FFUJ_12243, but not FFUJ_12241 or FFUJ_12244 ( Figure 14C). Analysis of culture extracts of the double overexpression strain (OE::PKS19/OE::FFUJ_12243) led to identification of four metabolites ( Figure 14, compounds 1, 2, 3 and 4) (Figure 14B, D). The similar UV spectra of compounds 1-4 observed combined with their similar molecular formula suggest that the chemical structures of the compounds are similar. For example, C 12 H 18 O 5 and C 12 H 16 O 4 differ by 2 hydrogen atoms and 1 oxygen atom which could be due to the loss of H 2 O from the former. Although overexpression of PKS19 alone did not result in production of compounds 1-4, simultaneous overexpression of both PKS19 and FFUJ_12243 resulted in production of ,10-fold more of the compounds than overexpression of the transcription factor-encoding gene (FFUJ_12243) alone. Elucidation of the chemical structures of compounds 1-4 is in progress. This is the first example in fungi where simultaneous overexpression of transcription factor-and PKS-encoding genes of the same SM cluster were performed and resulted in the production of a metabolite(s). The significantly enhanced production of compounds 1-4 via this simultaneous overexpression indicates that this strategy might be an effective tool to induce production of other fungal SMs from otherwise silent SM gene clusters (e.g. the PKS17/PKS18 cluster).
To determine whether the NRPS31 and PKS19 clusters are expressed in planta, we assessed levels of APS1 and PKS19 transcripts in F. fujikuroi-infected maize and rice roots by qPCR ( Figure 15). Surprisingly, expression of both genes was clearly plant-specific. The transcript levels of APS1 were generally significantly higher on maize than on rice while transcript levels for PKS19 were always higher on rice and almost not detectable on maize ( Figure 15). The high levels for PKS19 transcripts detected were in sharp contrast to the failure to detect transcription after growth in liquid culture. These differences in expression in vitro and in planta on one hand, and in the two plants on the other hand, suggest that the SM products of the PKS19 gene cluster may play a specific role in the rice-F. fujikuroi interaction. Thus, together with GAs the new PKS19-derived products may contribute to the evolutionary success of F. fujikuroi as a rice pathogen.
In conclusion, our study provides a high-quality genome sequence of F. fujikuroi that was assembled into only 12 scaffolds corresponding to the 12 chromosomes of the fungus. The availability of the genome sequences of six other Fusarium species (among them are two other newly sequenced Fusarium species) facilitated the most comprehensive genome-wide analysis to date of SM biosynthetic gene clusters in fungi. This, in combination with a broad spectrum of experimental approaches on the level of chromatin, transcription, proteome and chemical product analyses under different nitrogen and pH conditions provided new insights into the complex network of gene regulation. We show that most of the gene clusters are differentially regulated by nitrogen availability and pH and that there is a correlation between activating histone marks (H3K9ac and H3K4me2), high levels of mRNA and protein production for some (e.g. the GA cluster), but not all gene clusters.
As a highlight, we present the identification of two new gene clusters (PKS19, NRPS31) that are thus far unique for F. fujikuroi. Transcriptional engineering of these gene clusters by overexpression of the pathway-specific transcription factors and/or the key enzymes enabled the identification of the respective products for both novel gene clusters by HPLC-FTMS analyses. Furthermore, in planta expression studies revealed expression of the otherwise silent PKS19 gene cluster specifically in rice suggesting a specific role for the PKS19-derived product during rice infection beside GAs. Furthermore, we set out to find bona fide biological functions for gibberellins during rice root infection, which has been unclear so far even though production of the compounds by the fungus and gene expression in and flanking the cluster. Histone marks are described in the legend to Figure 3. Expression data were derived from microarray experiments in low and high nitrogen and are plotted as the changes in log 2 expression values in high-nitrogen medium compared to low-nitrogen medium. H3K9ac and gene expression are overall correlated, as both are increased under high nitrogen conditions. In some genes increased H3K4me2 was observed, also suggesting transcription. C: Chemical analysis of the product of the unique PKS19 gene cluster. was discovered a long time ago. We also discovered that the GA gene cluster is present in some related Fusarium species but that GA biosynthesis is restricted to F. fujikuroi.
The results from our study highlight the importance of genome sequencing in combination with multiple comprehensive (''omics'') analyses to gain insights into the potential for SM production, the various levels of regulation that govern their biosynthesis, and the impact on plant-fungus interaction.

Cultivation methods
For RNA isolation, microarray analyses, SM production and chromatin immunoprecipitation experiments (ChIP), strains were first cultivated for 3 days in 300-ml Erlenmeyer flasks with 100 ml Darken medium [107] on a rotary shaker at 180 rpm at 28uC. 500 ml of this culture were then used to inoculate 100 ml of ICI (Imperial Chemical Industries, UK) media [108] containing either 6 mM glutamine, 60 mM glutamine, 6 mM sodium nitrate or 120 mM sodium nitrate. Growth proceeded for 3 days on a rotary shaker at 28uC in the dark. The mycelia were harvested, washed with deionized water, and flash frozen with liquid nitrogen prior to lyophilization. For SM production, growth proceeded for 6 days on a rotary shaker at 28uC in the dark.

Nucleic acid isolation
Fungal DNA or RNA was prepared by first grinding lyophilized mycelium or maize or rice plant roots into a fine powder with a mortar and pestle followed by suspension in corresponding extraction buffers. Genomic DNA was isolated as described [109]. For isolation of highly pure DNA for genome sequencing, DNA was re-precipitated with 1/10 volume of 3 M sodium acetate (NaOAc) (pH 5.2) plus 3 volumes of cold 96% ethanol in glass tubes. The precipitating DNA was then wound around a glass pipette, washed in 70% ethanol and dissolved in TE buffer. Plasmid DNA from E. coli was extracted using the Gene-JETTM Plasmid Miniprep Kit (Fermentas GmbH, St Leon-Rot, Germany) and sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit and the ABI PRISM 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) according the manufacturer's instructions. Total RNA was isolated from mycelium from liquid cultures or from infected maize or rice roots using the RNAgents total RNA isolation kit (Promega GmbH, Mannheim, Germany).

Plasmid construction
The pOE-HPH and pOE-NAT vectors were generated by first PCR amplification of the gpd-promoter region from the pIG1783 vector [110] using the primer pair gpd-yeast-for and gpd-yeast-rev. Fragments of DNA including the hygromycin resistance cassette were generated by PCR from plasmid pCSN44 [111] using the primer pair hph-OE-Prom and hph-OE-Term while DNA including the nourseothricin resistance cassette was generated by PCR from plasmid pNR1 [111,112] using the primer pair nat-OE-Prom and nat-OE-Term. For pOE-HPH, the gpd and the hygromycin resistance cassette fragments were cloned into the EcoRI and XhoI restricted pRS426 [113]. A similar approach was used for the creation of the pOE-NAT vector, using the nourseothricin resistance cassette instead of the hygromycin resistance cassette. Fragments of DNA containing genes of interest were then cloned into the pOE vectors using the unique HindIII restriction site introduced in the primer sequences (underlined). Specifically, APS8, amplified from genomic DNA of Fusarium semitectum using the primer pair aps8-OE-for and aps8-OE-rev, was cloned downstream of the gpd promoter in the restricted pOE-NAT vector via yeast recombination cloning [114,115]. The resulting pOE::APS8 vector was transformed into wild type F. fujikuroi. A fragment of DNA containing APS2 from wild type F. fujikuroi was generated with the primer pair aps2-OE-for and aps2-OE-rev and cloned into the pOE-HPH vector via yeast recombination cloning. The resulting pOE::APS2 vector was then transformed into the OE:APS8 strain to create the double APS8/ APS2 overexpression strain. A similar strategy was used to create the PKS19 cluster transcription factor (FFUJ_12243) and the PKS19 (FFUJ_12239) over expression vectors. The amplicon containing FFUJ_12243 was generated from wild type F. fujikuroi DNA with the primer pair PKS19TF-OE-for and PKS19TF-OErev and the amplicon containing FFUJ_12239 was generated with the primer pair PKS19-OE-for and PKS19-OE-rev.

Northern analyses
Northern blot hybridizations were accomplished essentially as described by [116]. Total RNA (15 mg) were separated by electrophoresis on a 1% (w/v) agarose gel containing formaldehyde, transferred to Hybond-N+ membranes and then hybridized to radioactively labeled DNA. The agarose gel was illuminated with UV light (256 nm), against the background of a silica thinlayer chromatography plate (Schleicher und Schuell, Dassel, Germany) to visualize 18S and 28S rRNA bands. Labeled bands were visualized by exposure to photography film.

PCR and quantitative Real Time PCR
PCR mixtures contained 25 ng of template DNA, 50 ng of each primer (Table S8), 0.2 mM deoxynucleoside triphosphates, and 1 U of Biotherm Taq polymerase (Genecraft, Lüdinghausen, Germany). Reverse transcription-PCR (RT-PCR) was performed using the Superscript II (Invitrogen, Groningen, The Netherlands) and 1.5 mg of total RNA as the template, according to the manufacturer's instructions. Quantitative PCR (qPCR) was performed using iTaq Universal SYBR Green Supermix (BioRad) and Superscript II cDNA as template, in a Biorad thermocycler iTaq. In all cases the qPCR efficiency was between 90-110% and the annealing temperature was 56-60uC. Every sample was run twice and the results were calculated according to the delta-delta-Ct [117]. As reference genes, related actin gene (primers FRACRTPCRFW and FRACRTPCRRV), GDP-mannose transporter (primers FGMTRTPCRFW and FGMTRTPCRRV) and ubiquitin gene (primers FUBRTPCRFW and FUBRTPCRRV) were used. The following primers were used for amplification of the indicated genes: FFUM1RTPCRFW2 and FFUM1RTPCRRV2 for FUM1, FCPSKSRTPCRFW2 and FCPSCSRTPCRRV2 for CPS/KS, FAPS1RTPCRFW and FAPS1RTPCRRV for APS1, and FPKS19RTPCRFW and PKS19RTPCRRV for PKS19.
Strains IMI58289-DsRed and SG139-DsRed were generated by transformation of the corresponding protoplasts with pNDN-ODT coding for DsRed and a nourseothricin resistance cassette [118]. Integration of the plasmid was verified by diagnostic PCR using primers POliC_seq_F1 and DsRed_seq_R1.

Virulence assay
Infection of rice plants was performed according to [14]. Germinated rice seeds were co-cultivated with F. fujikuroi IMI58289-DsRed, SG139-DsRed or F. verticillioides M-3125 respectively. The cultures were incubated for 10 d at 28uC with a humidity of 80% in 12 h light and 12 h dark conditions. For microscopic analyses the rice roots were cleaned with sterile water and cut from the plant. Quantification of penetration events in seven infected roots for each strain was performed by visual inspection using fluorescence microscopy. For gene expression assays, spores (10 4 /ml) of F. fujikuroi or F. verticillioides were used instead of agar plugs.
Infection of maize plants was performed according to [119]. Germinated maize seeds were cultivated with the same three fungal strains but in cycles of 30uC with light for 14 hours, and 20uC in darkness for 10 hours. Roots were collected from three plants at two day intervals, cleaned with sterile water, flash frozen and lyophilized.

Genome sequencing and mapping
High throughput next generation sequencing using the Roche 454 GS-FLX system and the Titanium series chemistry was performed by Eurofins MWG Operon, Germany. The ultra high throughput sequencing process was carried out by shot gun library sequencing with 350-450 bp read length, reaching in average of 20-fold coverage (altogether approximately 900 million sequenced bases). The genome was assembled with the help of Long-Tag paired end sequences (3 kb and 8 kb libraries), into 18 scaffolds (size : .8.3 kb; N50: 4.2 Mb) containing 73 contigs with an average read coverage of 18.96 and a genome size of 43.9 Mb. The 13 largest scaffolds were mapped to 11 orthologous F. verticillioides chromosomes [22] using Mauve [120] and MUMmer [121]. Two gaps between four scaffolds were closed by PCR sequencing and reassessment of low coverage reads. The final 12 largest scaffolds match well to the predicted 12 chromosomes for F. fujikuroi [39].
Gene models. Gene models for F. fujikuroi were generated by three different prediction programs: 1) Fgenesh [122] with different matrices (trained on Aspergillus nidulans, Neurospora crassa and a mixed matrix based on different species); 2) GeneMark-ES [123] and 3) Augustus [124] with Fusarium ESTs as training sets. Annotation was aided by Blastx hits between the F. fujikuroi genome and protein sequences from F. verticillioides, F. oxysporum and F. graminearum, respectively. In addition, Swiss-Prot was also blasted against the scaffolds to uncover gene annotation gaps. The different gene structures and evidences were displayed in GBrowse [125], allowing manual validation of all coding sequences (CDSs). The best fitting model per locus was selected manually and gene structures were adjusted by manually splitting them or redefining exon-intron boundaries based on EST data where necessary. The final call set comprises 14,813 protein coding genes. In addition, 232 tRNA-encoding genes are predicted for F. fujikuroi using tRNAscan-SE [126]. The predicted protein set was searched for highly conserved single (low) copy genes to assess the completeness of the sequence dataset. Ortholog genes to all 246 single copy genes were identified by blastp comparisons (eVal: 10 23 ) against the single-copy families from all 21 species available from the FunyBASE [127]. Additionally, all of the 248 core-genes commonly present in higher eukaryotes (CEGs) could be identified by Blastp comparisons (eVal: 10 23 ) [40].

Identification of repetitive DNA elements
Interspersed repeat elements were identified using Re-peatScout with default parameters [128]. Repeats with length less than 50 bp were removed. Low complexity and simple sequence repeats were determined by NSEG [129] and Tandem-Repeat-Finder [130], which are part of the RepeatScout procedure. We considered the RepBase database to scan for known repetitive and transposable elements [50]. New putative transposable elements were identified and characterized in selecting interspersed repeats that contain predicted ORFs with transposase, reverse transcriptases or integrase InterPro domains. Additionally we used LTR_FINDER [131] for the identification of long terminal repeats. For repeat mapping and masking RepeatMasker was used [132].

Annotation of predicted open reading frames and comparative analysis
The 14,813 protein coding genes of F. fujikuroi were analyzed and functionally annotated using the Pedant system [133], accessible at http://pedant.helmholtz-muenchen.de/genomes. jsp?category=fungal. The genome and annotation was submitted to the European Nucleotide Archive (ENA, http://www.ebi.ac. uk/ena/data/view/HF679023-HF679034).

Genome comparison
The genomes of F. verticillioides and F. oxysporum were retrieved from http://www.broadinstitute.org/annotation/ genome/fusarium_group/MultiHome.html [22]. The genome of F. graminearum and annotation version 3.2 was retrieved from http : // mips . helmholtz -muenchen . de / genre / proj / FGDB / [134]. Brenda Wingfield, University of Pretoria, South Africa provided access to genome sequence of F. circinatum [35]. Stanley Freeman, The Hebrew University of Jerusalem, Israel provided access to genome sequence of F. mangiferae. All genomes were analyzed using the Pedant system [133] to allow comparative feature analysis which includes computation of the Similarity Matrix of Proteins (SIMAP) [42]. The SIMAP database provides a comprehensive calculation of protein sequence similarities/identities, sequence-based features and protein function predictions. Amino acid identities of homologous stretches are multiplied by the length of the homologous region and divided by the length of the whole protein. Proteins with a bidirectional best hit between two species were used to determine collinear regions using the tool Orthocluster [134]. A collinear block was defined as three consecutive, orthologous genes, allowing one missing or additional gene in between. Whole genome nucleotide alignments were performed using MUMmer [121] with a cluster length of exact matches of at least 100 nt and at most 500 nt mismatches between two exact matches.

In silico identification of secondary metabolite clusters
To scan the genome for SM clusters the InterPro scan results of the Pedant analysis as well as SMURF analysis [83] were used. Proteins containing characteristic domains of signature enzymes were taken as a seed for a putative local gene cluster. Each cluster seed was then extended by determining if neighboring genes encoding typical SM enzymes like P450 monooxygenases, methyltransferases, monooxygenases as well as transporter proteins or transcription factors. Each cluster of genes was manually validated by adjusting to previously published data and comparative analysis to homologous clusters in other Fusarium.

Prediction of secreted proteins
We computed secreted proteins in a pipeline approach using a combination of five methods. Therefore we selected a set of putative secreted proteins using SecretomeP [130] and a cutoff score of 0.6. Out of this set we excluded proteins with mitochondrial target using TargetP [135]. We also filtered on proteins with a TargetP RC-score less than four. A prediction of extracellular target compartments has been done with Wolfpsort [136]. To exclude extracellular, membrane bound proteins we utilized TMHMM [137] for transmembrane domain prediction. The SignalP program [138] was used to differentiate between classically and non-classically secreted proteins. Here we assume that proteins having an s-score greater than 0.5 contain a signal peptide and therefore are secreted classically.

Phylogeny
Calculations of phylogenetic relationships between eight Fusarium species are based on the nucleotide sequences of 28 genes. All genes are involved in energetic processes according to FunCat database [43] and show at least 90% identity on protein level to their respective orthologs of the other species. Alignments of the orthologs have been calculated with Mafft [139] using default parameters. Afterwards all alignments were concatenated and columns containing gaps were removed. For phylogenetic tree calculation we used PhyML, an implementation of the maximum likelihood algorithm [140]. We chose the HKY85 substitution model and determined tree support in performing the bootstrap test with 1000 replicates. The gene codes of the 28 genes used for the calculation are: . For phylogenetic analysis of SM genes, predicted amino acid sequences were aligned with the ClustalW+ alignment tool using the BLOSUM scoring matrix. Aligned sequences were then subjected to maximum parsimony analysis as implemented in the program PAUP version 4.0b10-Unix. GA biosynthetic gene sequences were subjected to maximum likelihood analysis as implemented in MEGA5 [141]. Statistical support for branches within phylogenetic trees was determined by bootstrap analysis with 1000 (maximum parsimony) or 500 (maximum likelihood) pseudoreplications.

Microarray design
Based on the de novo sequence of F. fujikuroi and the gene annotation, in consultation with Roche-NimbleGen Systems, Inc. (Madison, WI), we designed a 12-plex microarray with 16522 probe sets (array design 110407_UG_Ffujikuroi_expr_HX12). Each probe set included nine perfectly matching oligonucleotide probes. The microarray data are available from the NCBI Gene Expression Omnibus (GEO) under the series accession numbers GSM1070182 -GSM1070196, listed at http://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSE43745. Hybridization of microarrays was done at Arrows Biomedical (Münster, Germany) essentially according to manufactures protocols.

Analysis of expression data
Microarray tif images were processed to numeric raw data using NimbleScan v2.6 software. The preprocessing steps including background correction, normalization and summarization of probe intensities, were performed employing the oligo R package [142] and the RMA algorithm [143]. We used the limma R package [144] to identify differentially expressed genes by fitting linear models for each gene and computing moderated t-statistics using the empirical Bayes method. The resulting p-values were corrected for multiple testing by using the Benjamini-Hochberg procedure [145]. We regard genes having a log2 fold-change greater than one and corrected p-value (FDR) less than 0.05 between two conditions as differentially expressed. The MIPS Functional Catalogue (FunCat, http://mips.helmholtz-muenchen. de/proj/funcatDB/) was used to assign specific genes to specific biological processes and the MIPS single enrichment tool was used to identify categories whose members were over-represented in the expression analysis compared to the genes present on the microarray [43].

ChIP sequencing
For chromatin immunoprecipitation (ChIP) experiments, the mycelium of wild type was pre-cultivated for 72 hours in Darken medium. 500 ml of the pre-culture was then used for the inoculation of 100 ml ICI medium with either 6 mM glutamine (for nitrogen starvation conditions) or 60 mM glutamine (for nitrogen excessive condition). After three days the mycelium was crosslinked with 1% formaldehyde and incubated for 30 minutes. Then 5 ml glycine (5 M) was added to the culture for quenching and the mycelium was harvested by filtration. About 150 mg of fresh mycelium was used for subsequent ChIP [47,146]. The antibodies used were H3K4me2 (Millipore, 07-030), H3K9me3 (Active Motif, 39161) and H3K9Ac (Active Motif, 39137). Sequencing libraries were prepared as described [147]. Singleend 36-50 nt sequencing was carried out on an Illumina GAII genome analyzer. Reads with mapped to the F. fujikuroi genome with BWA using default conditions [148] and visualized in a gbrowse2 genome browser [149]. Quantification of ChIP-seq reads present in each gene was performed with EpiChIP [47,146,150].
The ChIP seq data are available from the NCBI Gene Expression Omnibus (GEO) under the series accession numbers GSM 1122108 -GSM 1122111, listed at http://www.ncbi.nlm. nih.gov/geo/query/acc.cgi?acc=GSE46033.

Chemical analysis
Chemicals and solvents were purchased from Sigma-Aldrich (Deisenhofen, Germany) or Grüssing GmbH (Filsum, Germany). Gibberellins were obtained from Serva (Heidelberg, Germany), fusaric acid and beauvericin were obtained from Sigma (Deisenhofen, Germany) and apicidin was obtained from AppliChem (Darmstadt, Germany). Fumonisin B 1 , fusarin C, O-methylfusarubin and bikaverin were isolated from fungal cultures as previously described [14,15,151,152]. All standards for HPLC analysis were dissolved in methanol/water (1/9, v/v) except for beauvericin which was dissolved in methanol/water 7/3, v/v. The concentration for each chemical ranged between 10 and 100 mg/ml. The volatile ent-kaurene released by agar plate cultures were collected by use of a closed loop stripping apparatus (CLSA) and analyzed by GC-MS as previously described [88].

Sample preparation for SM analysis
The fungal strains were grown in a submerged culture as described. The mycelica were removed from the culture by filtration through Miracloth (Calbiochem, Merck KGaA, Darmstadt, Germany) under vacuum. The culture filtrates were filtrated through syringe filters (RC Membrane, 0.45 mm, 4 mm Syringe Filters non-sterile, PP Housing, Luer/Slip, Phenomenex, Aschaffenburg, Germany) and directly used for analysis. The cultivation of each fungus under each culture condition was performed in duplicate. For beauvericin, 0.1 g of the mycelium were extracted with 2 ml of ethyl acetate, methanol and methylene chloride (3/2/1). After shaking for 2 h, 500 mL of the extract were evaporated at 25uC under a stream of nitrogen and suspended in the same volume methanol/water (7/3, v/v). The sample was sonicated (Bandelin Sonorex RK 100 h, Bandelin Electronicm Berlin, Germany) for 30 min and insoluble materials were removed by centrifugation for 3 min at 1400 g prior to HPLC-FTMS analysis.

HPLC-FTMS
The HPLC-FTMS system used an HPLC system (Accela LC with Accela Pump 60057-60010 and Accela Autosampler 60057-60020, Thermo Scientific, Dreieich, Germany) coupled to a Fourier transform mass spectrometer with a heated ESI source (LTQ Orbitrap XL, Thermo Scientific, Dreieich, Germany). Ionization was carried out in the positive ionization mode. The parameters were as follows: Capillary temperature 275uC, vaporizer temperature 350uC, sheath gas flow 40 units, auxiliary gas flow 20 units, source voltage 3.5 kV, tube lens 119 V. The software Xcalibur 2.07 SP1 (Thermo Scientific, Dreieich, Germany) was used for data acquisition and analysis. The HPLC column used was a 150 mm62.00 mm i.d., 5 mm, Gemini C18 with a 4 mm62 mm Gemini NX C18 guard column (Phenomenex, Aschaffenburg, Germany). The gradient was from 10% A to 100% A in 60 minutes followed by column flushing for 5 minutes at 100% A and equilibration at the starting conditions of 10% A for 10 minutes. The flow rate was 250 ml/min and the injection volume set to 10 ml. For the comparative analysis of different Fusarium species, the metabolites were identified by the exact mass of the most intense ion, their isotope pattern and the retention time compared to standard substances (Table S9).
For apicidin analysis, solvent A was methanol (v/v) and solvent B was water (v/v). The gradient was from 10% A to 100% A in 30 minutes followed by column flushing for 5 minutes at 100% A and equilibration of the starting conditions of 10% A for 10 minutes. The flow rate was 1000 ml/min, the injection volume was 30 ml. The retention time of the apicidin-standard was 29.9 min, the retention time of the apicidin-derivative formed by F. fujikuroi was 29.8 min.
For analysis of putative PKS19 metabolites, the gradient was from 10% A to 100% A in 60 minutes followed by column flushing for 5 minutes at 100% A and equilibration of the starting conditions of 10% A for 10 minutes. The flow rate was 1000 ml/ min, the injection volume was 10 ml. The retention times of the newly identified metabolites were as follows: 1st peak 23.4 min, 2nd peak 29.1 min, 3rd and 4th peak 32.1 and 33.1 min.

Isolation of putative NRPS31 metabolites
The mutant strain OE::APS2/OE::APS8 was grown in 1 l ICI medium with 60 mM glutamine. The mycelium was removed from the culture by filtration through Miracloth (Calbiochem, Merck KGaA, Darmstadt) under vacuum. The culture filtrate was fractionated on a Strata C18-E (55 mm, 70 Å ) 10 g/60 ml SPE column (Phenomenex, Aschaffenburg, Germany) essentially as described by [18]. Briefly, the column was first activated by flushing with 50 ml methanol and 50 ml water under vacuum. The aqueous culture filtrate was applied under vacuum and washed with 100 ml of water. Metabolites were eluted from the column in five fractions of 20%, 40%, 60%, 80% and 100% methanol/water (v/v). Only the 80% fraction contained a novel peak of interest. The solvent was removed on a rotary evaporator (Rotavapor-R, Büchi Labortechnik GmbH, Essen, Germany) and the residue dissolved in about 3 ml of 50% 5% tetrahydrofuran in methanol (v/v) (solvent A) and 50% water (v/v) (solvent B).

Preparative HPLC-UV
Purification of the new metabolite for further analysis was carried out on a preparative HPLC-UV system (Varian Polaris pumps with Rheodyne manual injection port and Varian ProStar UV-detector, Varian, Europe) using Galaxie 1.9.302.530 (Varian, Europe) software for data acquisition. The column used was a 250610.0 mm Varian Microsorb 100-5 C18 column with a 10.0 610.0 mm Gemini C6-Phenyl guard column. Solvent A was 5% tetrahydrofuran in methanol (v/v), solvent B was water. The flow rate was 3.5 ml/min. The first preparative run was from 50% A to 80% A in 30 minutes followed by equilibration at 50% A for 5 minutes. The UV-detector was set to 254 nm. The collected fractions were tested by HPLC-MS loop analysis (see below) to identify the fractions containing the new metabolite. Due to coelution of the target peak with fusarins by comparison to standard materials, 3 fractions were collected. The fraction from 18 to about 21 minutes containing only fusarins, the fraction from 21 to 22 minutes containing low amount of the target peak and a high amount of fusarins while the fraction from 22 to 26 minutes contained a high amount of the target peak and a low amount of fusarins. Fractions 2 and 3 were evaporated separately to dryness at the rotary evaporator and dissolved in 70% A. A second preparative HPLC-UV run was carried out on each fraction on the same system with the following change. The samples were loaded and peaks were eluted isocratically at 70% A. The two collected fractions eluting between 8 to 10 minutes contained high amounts of the new metabolite and only minor impurities with fusarins and were combined. The verification of the different fractions was carried out on an API 3200 LC-MS/MS system (Applied Biosystems, Darmstadt, Germany) with a 1200 series HPLC system (Agilent Technologies, Böblingen, Germany). The software Analyst 1.4.2 (Applied Biosystems, Darmstadt, Germany) was used for data acquisition. The HPLC run was at 50% methanol/water (v/v) with a flow rate of 200 ml/min for 2 minutes. 10 mL of the sample were injected and without separation on a column directly infused into the mass spectrometer and analysed after electrospray ionization in the positive ionization mode in a Q1 scan. The ion spray voltage was set to 5500 V, the declustering potential was 50 V and the entrance potential was 3.5 V. Zero grade air was used as nebulizer gas (35 psi), and, heated at 350uC, as turbo gas for solvent drying (45 psi). Nitrogen served as curtain gas (30 psi). The scan range was set to m/z 400-700 and the presence of the ions 646 (new metabolite) and 432 (fusarins) was monitored. The quadrupol was set to unit resolution. The fractions containing the metabolite were combined, dried under a stream of nitrogen and the residue was dissolved in 600 ml methanol and stored at 220uC for about 15 minutes until a white precipitate of the metabolite was observed. The precipitate was collected by centrifugation at 20uC, 1100 g for 2 minutes.

Acid hydrolysis of apicidin and unknown compound
Apicidin and the unknown compound were transferred into separate vials, evaporated to dryness, dissolved in 500 ml 6 M hydrochloric acid and heated on 110uC for 8 h. The acid was then removed at 110uC under a stream of nitrogen and the residue dissolved in 100 ml water and used for HPLC-FTMS analysis. The peptide hydrolyzate was applied to the HPLC-FTMS system with the same column and solvents as described above with the following changes. The gradient was 5% A for 10 minutes followed by column flushing for 5 minutes at 100% A and equilibration at the starting conditions of 5% A for 7 minutes. The flow rate was 250 ml/min and the injection volume set to 10 ml. The composite amino acids for apicidin and the unknown compound were identified based on a comparison of their exact masses to standard amino acids (see Table S10). It should be noted that N-methoxy-tryptophan is degraded by acid hydrolysis.

Proteomics
Protein analyses by LC-MS/MS were conducted as described [153] using samples from 2 independent cell harvests. 15N labeled nitrogen-deficient (6 mM Gln) and 14 N labeled nitrogen-sufficient (60 mM Gln) cells were mixed on an equal protein basis. Proteins were separated by SDS-PAGE, protein bands were excised and digested tryptically.
Chromatographic separation of peptides was performed using an Ultimate 3000 Nanoflow HPLC system (Dionex). The mobile phases consisted of 5% (v/v) acetonitrile/0.1% (v/v) formic acid in ultrapure water (A) and 80% acetonitrile/0.1% formic acid in ultrapure water (B). The sample (1 ml) was loaded on a trapping column (C18 PepMap 100, 300 mM65 mm, 5 mm particle size, 100 Å pore size; Thermo Scientific) and desalted for 4 min using eluent A at a flow rate of 25 ml/min. Then the trap column was switched online with the separation column (Acclaim PepMap100 C18, 75 mm615 cm, 3 mM particle size, 100 Å pore size, Thermo Scientific) and peptides were eluted at a flow rate of 300 nl/min and employing the following gradient profile: 0-50% B over 45 min, 50% B for 5 min, 50-100% B over 1 min, 100% B for 5 min. Afterwards the column was re-equilibrated with 100% A for 10 min. The LC system was coupled via a nanospray source to an LTQ Orbitrap XL mass spectrometer (Thermo Finnigan). MS full scans (m/z 375-1800) were acquired in positive ion mode by FT-MS in the Orbitrap at a resolution of 60,000 (FWHM) with internal lock mass calibration on m/z 445.12003. The 12 most intense ions were fragmented in the linear ion trap by CID (35% normalized collision energy). Automatic gain control (AGC) was enabled with target values of 56105 and 56103 for MS full scans and MS/MS, respectively. One microscan was acquired per MS/ MS spectrum and maximum ion trap fill time was 50 ms [154]. Dynamic exclusion was enabled with an exclusion duration of 60 s, repeat count of 1, list size of 500 and exclusion mass width of +/25 ppm. Unassigned charge states and charged state 1 were rejected. Acquired Thermo Xcalibur raw files were converted to the open mzML format [155] via msconvert (part of Proteowizard, [156].

Identification of proteins
Proteomics data analyses were performed individually for each biological replicate using Proteomatic [157] as a data processing pipeline tool providing access to the later mentioned tools and pymzML [158] for direct access to mzML files. OMSSA (version 2.1.9, [159]) and X! Tandem (version 2012.10.01.1, [160]) programs were used for protein identification. The parent mass error was set to 20 ppm, the fragment mass error was set to 0.5 Da. Oxidation of M was used a variable modification and a maximum number of 2 missed cleavages was allowed. For OMSSA, a linear precursor charge dependency was set. For X! Tandem, noise suppression was enabled. Separate searches were conducted for 14 N and 15 N evaluation, respectively (precursor and product ion search types set to 14 N or 15 N).
For the statistical assessment of the quality of PSMs, a target/ decoy approach was used. Decoy sequences were built from the F. fujikuroi protein database by randomly shuffling tryptic peptides. PSM hits from OMSSA and X! Tandem were filtered (1) using a hit distinctiveness filter as described [161], (2) removing all hits from peptides that occur as a target and as a decoy sequence and (3) protein groups are built for peptides that occur in more than one protein. These steps result in a maximum of one peptide hit per spectrum. Quality (version 2.02) [162] was used for statistical validation of protein identifications with only allowing target hits with a posterior error probability (PEP) less than 0.01. As an additional filtering step, all PSMs with a precursor mass deviation of more than 5 ppm were discarded.

Quantification of proteins
Significantly identified peptides were used for protein quantification with qTrace separately for both replicates as described [153], except that the 15N labeling approach was used. After protein information was added to the qTrace output, the following filtering steps were conducted as described [153] (1) require MS2 identifications and (2) pick most abundant band. The resulting relative peptide amounts were normalized to the median of all peptide amounts from each replicate. After protein ratios ( 15 N/ 14 N) were computed by summing up all amounts of proteotypic peptides, results were filtered for at least two peptide/band/charge (PBC) combinations. Hits that did not meet the PBC criteria were required to have a scan count of at least 20. Mean ratios are calculated as the arithmetic mean of the ratios of replicate A and B, SD is the standard deviation of the ratios. If one of the ratios is ''0'' or ''inf'', the resulting ratio mean is the real ratio (not ''0'' or ''inf''), no SD is given. If both ratio are ''inf'' (or ''0''), the resulting ratio mean is ''inf'' (or ''0''), no SD is given. If the ratios are in opposite directions (''.1'' and '',1''), no ratio mean and SD are calculated. Heatmaps were created with a custom script in Proteomatic. All Protein ratios were log2transformed. Correlation between protein ratios and transcriptomics fold changes have been calculated using Hmisc R package (Harrell EF et al., Hmisc: Harrell Miscellaneous, http://cran.rproject.org/web/packages/Hmisc/index.html, 2012). Ratios of proteins that occur in only one experimental condition were set to minimum or maximum ratio to avoid infinity values.

Immunocytology
For immunocytological analysis, 10-14 day old hyphae of F. fujikuroi IMI58289 were scraped from CM-plates, fixed for 15 min in freshly prepared 4% formaldehyde solution in phosphatebuffered saline (PBS pH 7.3), transferred on glass slides, covered with coverslips and squashed. After freezing in liquid nitrogen the cover slips were removed and slides were transferred immediately into PBS. After pre-incubation for 30 min in 3% BSA in PBS at 37uC the slides were incubated with the primary antibodies in a humidified chamber. The primary antibodies were used in the following dilutions: 1:50 anti-H3K9ac (ab10812, Abcam, Cambidge, UK) and 1:100 anti-H3K9me3 (MAb-146-050, Diagenode, Liège, Belgium). After 1 h incubation at 37uC and washing for 15 min in PBS, the slides were incubated either in Alexa-488 anti mouse IgG or Alexa-555 conjugated anti rabbit IgG secondary antibodies for 1 h at 37uC (1:100; Molecular Probes, Invitrogen, Karlsruhe, Germany). All preparations were washed in PBS for 15 min, stained with DAPI in mounting solution (antifade) and examined with Zeiss Axioscop fluorescence microscope. Variation in histone H3 modification status serves as marker for chromosomal regions in which genes are expressed (H3K9ac and H3K4me2) or silent (H3K9me3). ''D expression up'' indicates a more than twofold increase in gene expression during growth of F. fujikuroi in a nitrogen-rich medium, whereas ''D expression down'' indicates an at least twofold decrease in gene expression. SM biosynthetic gene cluster locations are indicated by arrows labeled with the PKS, NRPS or TC (DTC means diterpene cyclase; STC means sesquiterpene cyclase) gene in each cluster (see Figure 3, Table 4 and  Table S4). F. fujikuroi genes (FFUJ locus ID) to the left and right of the predicted centromere on each chromosome was identified and compared to the location of predicted orthologs in F. verticillioides (FVEG locus ID). ''+'' and ''2''. (TIF) Figure S4 Maximum likelihood trees of selected GA biosynthetic genes. A: tree generated from Cps/Ks and related diterpene synthases from multiple genera of fungi; B: tree generated from concatenated sequences of GA enzymes/genes common to Fusarium species, Claviceps purpurea and Sphaceloma manihoticola. C: trees of four GA biosynthetic enzymes/genes (P450-1, P450-2, P450-4 and Cps/Ks) that are common to Fusarium, Claviceps and Sphaceloma; All trees were inferred from alignments generated by Muscle of deduced amino acid sequences, and all gaps in the alignment were removed prior to maximum likelihood analysis. Numbers near branches are bootstrap values based on 500 pseudoreplicates. Only bootstrap values .70 are shown. Muscle and maximum likelihood analyses were conducted using MEGA version 5 [163]. Sequences for F. proliferatum, and non-Fusarium genera other than Claviceps were obtained from the NCBI database via BLASTp analysis with F. fujikuroi homologues: NCBI accessions for terpene synthases for which accession or strain numbers are not shown in Figure S4A Figure S6 Phylogenetic tree of PKSs encoded by Fusarium genome sequences. The tree was generated by maximum parsimony analysis of alignments of deduced amino acid sequences of the KS and AT/MAT domains. The tree also includes PKSs from other fungi and for which the corresponding secondary metabolites is known. The names of metabolites are indicated to the right. R-PKS and NR-PKS indicate reducing and nonreducing PKS respectively. ''MSAS'' indicates 6-methylsalicylic acid synthase. The protein/gene designations FCIRG, FFUJ, FM, FOXG, FVEG and JGI correspond to PKSs deduced from genome sequences of F. circinatum, F. mangiferae, F. oxysporum f. sp. lycopercisi strain Fol 4287, F. verticillioides, and F. solani f. sp. pisi (Nectria haematococca) respectively. The designations FOXB, FOSC3A and PHW815 are for PKSs from F. oxysporum strains Fo5176, FOSC 3-a and PHW815 respectively. The tree is rooted with the KS and AT domains of the Gallus gallus fatty acid synthase. (TIF) Figure S7 Comparison of the fusarubin biosynthetic gene (FSR/PGL) cluster in genomes sequences of Fusarium. Horizontal arrows that are the same color represent genes, or gene sets, that have closely related homologues in two or more species. Blue arrows represent FSR/PGL genes, and the numbers within these arrows correspond to FSR/PGL gene numbers rather than gene/protein model designations from genome databases. For those that are available, gene designations are indicated below species names. In F. solani, all FSR/PGL genes are not located within a contiguous cluster of genes. (TIF) Figure S8 Comparison of the putative fusaric acid biosynthetic gene (FUB) cluster in genomes sequences of Fusarium. Horizontal arrows that are the same color represent genes, or gene sets, that have closely related homologues in two or more species. Blue arrows represent FUB genes, and the numbers within these arrows correspond to FUB gene numbers rather than gene/protein model designations from genome sequence databases. For those that are available, gene designations are indicated below species names. In F. oxysporum, the FUB cluster is interrupted by two genes. The asterisks mark genes without annotation. (TIF) Figure S9 Comparison of the bikaverin biosynthetic (BIK) cluster in genomes sequences of Fusarium. BIK genes are represented by blue horizontal arrows, and the numbers within these arrows correspond to BIK gene numbers rather than gene/protein model designations from genome sequence databases. For those that are available, gene designations are indicated below species names. The asterisks marks FFUJ_14916. (TIF) Figure S10 Comparison of the fusarin biosynthetic gene (FUS) cluster in genome sequences of Fusarium and the related fungus Metarhizium anisopliae as well as remnants of the FUS cluster in F. mangiferae and Trichoderma reesei. FUS genes are represented by blue horizontal arrows, and the numbers in the arrows correspond to FUS gene numbers. For those that are available, gene designations are indicated below species names. Y indicates a pseudogene. (TIF) Figure S11 Comparison of the putative PKS17 -PKS18 gene cluster in the genome sequences of F. fujikuroi and F. mangiferae. The cluster genes are represented by green horizontal arrows, and cluster flanking genes are represented by blue or yellow arrows; yellow arrows represent genes that were not detected in the flanking regions of F. fujikuroi or F. mangiferae. Synteny of some cluster flanking regions is partially conserved in F. verticillioides, F. circinatum, and F. oxysporum, which lack the cluster. For those that are available, gene designations are indicated below species names. (TIF) Figure S12 Comparison of the putative, two-gene PKS12 clusters (green arrows) in genome sequences of Fusarium. The second gene in the cluster is predicted to encode a methyltransferase. Blue arrows represent cluster flanking genes, which exhibit partial synteny conservation among the species examined. Based on phylogenetic analysis ( Figure 5), PKS12 homologues have been arbitrarily designated as PKS12 and PKS12a until additional information becomes available. PKS12 is absent in the F. solani genome sequence, and only a remnant of it (indicated by Y) is present in the genome sequences of F. oxysporum. However, F. solani has a PKS12a homologue, and F. oxysporum strain Fol 4287 has two PKS12a paralogues (FOXG_14850 and FOXG_15586), which are part of a larger region of duplicated DNA that includes two putative methyltransferase genes, one on either side of each PKS12a paralogue. Other F. oxysporum genome sequences examined have only one PKS12a homologue. (TIF) Figure S13 Phylogenetic tree of Fusarium NRPS22 and Beauveria bassiana BeaS and BslS as well as F. scirpi EsyN. The tree was generated by maximum parsimony analysis of alignements of deduced amino acid sequences of the A domains. The protein/gene designations FCIRG, FFUJ, FOXG and FVEG correspond to NRPSs deduced from genome sequences of F. circinatum, F. fujikuroi, F. oxysporum f. sp. lycopercisi strain Fol 4287, and F. verticillioides respectively. The tree is rooted with the A1 domain of the F. graminearum NRPS19. (TIF) Figure S14 Relative expression of the fumonisin biosynthetic gene FUM1 from F. fujikuroi and F. verticillioides in rice and maize roots. Rice (A) and maize (B) roots were infected with Fusarium fujikuroi and F. verticillioides spores and every 2 days RNA was isolated from three or five plants and used in real time PCR analysis. The expression levels were obtained using the delta-delta Ct and were normalized against three reference genes encoding a related actin, the GDP-mannose transporter and ubiquitin. The expression levels of the F. verticillioides FUM1 at 2 days in rice was arbitrarily set as 1, and all other expression levels were reported relative to it. (TIF) Figure S15 Venn diagram representing the distribution of nitrogen-and pH-regulated genes. Number of differentially regulated genes in wild-type F. fujikuroi cultivated under conditions that vary in nitrogen content and pH. The data were obtained by microarray analyses. The conditions were as follows: 6 mM gln (glutamine) constitutes, acidic low nitrogen; 60 mM gln, acid high nitrogen; 6 mM NaNO 3 alkaline low nitrogen; and 120 mM NaNO 3 alkaline high nitrogen. (TIF) Table S1 Putative F. fujikuroi centromeric regions. Coordinates on chromosomes (Chr) I to XII include 0.5-1 kb of euchromatic regions on either side of pericentric and centromeric DNA and no attempt has been made to separate pericentric and centromeric regions. As none of the 14 scaffolds contain telomere repeats, all chromosome ends are unfinished. ''Incompl R'' indicates that the right arm of the chromosome does not have heterochromatin as indicated by presence of H3K9me3. (DOCX) Table S2 Nearest genes to F. fujikuroi centromeres and synteny with F. verticillioides. Nearest F. fujikuroi genes (FFUJ locus ID) to the left and right of the predicted centromere on each chromosome was identified and compared to the location of predicted orthologs in F. verticillioides (FVEG locus ID). ''+'' and ''2'' indicate strands, ''t'' indicates last gene on a specific F. verticillioides contig, ''i'', indicates internal, non-terminal gene. (DOCX)  Gibberellin production by Fusarium species of the GFC and F. oxysporum 4287 (outgroup). F. oxysporum 4287 does not contain a GA gene cluster, but produces GAs after transforming it with the entire GA gene cluster from F. fujikuroi (cos1). B: Gibberellin production by F. oxysporum isolates with an entire GA gene. (DOCX) Table S7 Expression pattern of all genes under four standard growth conditions. acidic and alkaline low-nitrogen (6 mM glutamine or 6 mM sodium nitrate) and acidic and alkaline high-nitrogen (60 mM glutamine or 120 mM sodium nitrate), as well as proteomic data and the presence of activation histone marks under two different growth conditions (acidic low and highnitrogen). (XLSX)  Overview of the metabolites that were analyzed by HPLC-FTMS. The analytes were identified by their retention time and isotope pattern compared to standard substances. For the estimation of the produced amount, the m/z of their most intense ions were observed. (DOCX)