Investigation of the Fusarium virguliforme Transcriptomes Induced during Infection of Soybean Roots Suggests that Enzymes with Hydrolytic Activities Could Play a Major Role in Root Necrosis

Sudden death syndrome (SDS) is caused by the fungal pathogen, Fusarium virguliforme, and is a major threat to soybean production in North America. There are two major components of this disease: (i) root necrosis and (ii) foliar SDS. Root symptoms consist of root necrosis with vascular discoloration. Foliar SDS is characterized by interveinal chlorosis and leaf necrosis, and in severe cases by flower and pod abscission. A major toxin involved in initiating foliar SDS has been identified. Nothing is known about how root necrosis develops. In order to unravel the mechanisms used by the pathogen to cause root necrosis, the transcriptome of the pathogen in infected soybean root tissues of a susceptible cultivar, ‘Essex’, was investigated. The transcriptomes of the germinating conidia and mycelia were also examined. Of the 14,845 predicted F. virguliforme genes, we observed that 12,017 (81%) were expressed in germinating conidia and 12,208 (82%) in mycelia and 10,626 (72%) in infected soybean roots. Of the 10,626 genes induced in infected roots, 224 were transcribed only following infection. Expression of several infection-induced genes encoding enzymes with oxidation-reduction properties suggests that degradation of antimicrobial compounds such as the phytoalexin, glyceollin, could be important in early stages of the root tissue infection. Enzymes with hydrolytic and catalytic activities could play an important role in establishing the necrotrophic phase. The expression of a large number of genes encoding enzymes with catalytic and hydrolytic activities during the late infection stages suggests that cell wall degradation could be involved in root necrosis and the establishment of the necrotrophic phase in this pathogen.

Introduction may play a major role in establishing the necrotrophic phase of the pathogenic fungus. The candidate virulence genes identified in this study lay the foundation for identification of F. virguliforme virulence mechanisms and the molecular basis of the root necrosis caused by this soybean pathogen.
RNA extraction from germinating conidia, mycelia and root samples F. virguliforme infected root samples were collected at 3, 5, 10 and 24 days post inoculation (dpi) from three independent experiments. The overview of the experimental strategy is shown in Fig 1. Necrotic symptoms were visible at 10 dpi with gradual spreading of root rot symptoms until 24 dpi. RNA samples of infected roots from three experiments, collected 3 and 5 dpi, were bulked and named as "early infection" whereas the pooled RNA sample isolated from the infected roots, collected 10 and 24 dpi, were termed as "late infection". Early time points (3 and 5 dpi) were selected to study the putative biotrophic phase with very little root necrosis; while late time points (10 and 24 dpi) were selected for investigating the necrotrophic phase (S1 Fig).
For semi-quantitative RT-PCR and quantitative (qRT-PCR) analyses, root samples were harvested at 1, 3, 5 and 10 dpi either following F. virguliforme infection or treatment with water. In RT-PCR, we included 1 dpi in RT-PCR to determine the early changes in gene expression while 10 dpi was considered to represent the late time point ignoring 24 dpi treatment. In each experiment, roots of five plants were pooled for each treatment and frozen in liquid nitrogen and stored at -80˚C until further use.
Germinating conidia were grown for 12 h in liquid modified Septoria medium (MSM) [13]. Mycelia were harvested from spores grown in MSM liquid media for two weeks. Total RNAs were isolated from the germinating conidia, mycelia and infected root tissues using TRIzol Reagent (Invitrogen, Carlsbad, CA, U.S.A.). The quality of RNA samples was determined by running RNAs on a denaturing agarose gel. cDNA library preparation for sequencing transcripts 10 ug total RNAs from early and late time points of F. virguliforme infected soybean root tissues, germinating conidia, and mycelia were used to purify poly (A) + RNAs using oligo (dT) attached to magnetic beads (Promega, Madison, WI). Poly (A) + RNAs were fragmented into short sequences in the presence of divalent cations at 94˚C for 5 min. RNA samples were reverse transcribed using a cDNA synthesis kit from Illumina (Illumina, Inc. San Diego, CA, U.S.A.).
Sequencing of the F. virguliforme transcripts isolated from infected soybean roots, germinating conidia and mycelia cDNAs of an individual RNA sample were sequenced in a single lane of the Illumina NGS platform GAII (Illumina, Inc. San Diego, CA, U.S.A.) at the DNA Facility, Iowa State University. The raw sequence reads were generated using the Solexa GA pipeline 1.6 (deposited in GEO; accession GSE86201). The draft F. virguliforme genome sequence (available at http:// fvgbrowse.agron.iastate.edu/gb2/gbrowse/fvirguli/ [26]) has been shown to contain 14,845 predicted genes [26]. Reads with quality scores more than Phred 33 were used to map transcripts to the cDNA sequences of predicted F. virguliforme genes using the Bowtie program [27] with the default parameters. The generated SAM (Sequence Alignment/Map) output for each condition was used to extract mapped reads (S1 Table) for corresponding genes using a custom script [28]. The reads per kb per million reads (RPKM) for each gene was calculated according to the formula R = 10 9 C/NL [29], where C is the number of mappable reads aligned onto the exonic sequence of a gene, N is the total number of mappable reads in the sample, and L is the length of the gene. CIMminer was used to generate color-coded Clustered Image Maps (CIMs) ("heat maps") representing "high-dimensional" data sets such as gene expression profiles (http://discover.nci.nih.gov/cimminer/). A heat map was generated in one matrix CIM for the normalized values for each gene after RPKM analysis [30]. TargetP1.1 (http:// www.cbs.dtu.dk/services/TargetP/) set for non-plant as organism with no cutoffs, winnertakes-all was used to identify the candidate secreted proteins.

Identification of the candidate F. virguliforme virulence genes
We looked for infection-induced F. virguliforme genes that showed high identity to functionally characterized virulence genes by running the NCBI BlastX program against the nonredundant protein sequences (nr) database (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Semi-quantitative RT-PCR and quantitative real-time PCR (qRT-PCR) analyses
Semi-quantitative PCR of several induced F. virguliforme genes was conducted to validate the expression profiles deciphered from deep sequencing of single RNA samples pooled from three biological experiments. Genes were selected randomly from individual functional categories. Total RNA samples were isolated from the F. virguliforme-infected or water treated etiolated root samples of soybean cv. Essex at 1, 3, 5 and 10 days post inoculation or water treatment by using TRIzol (Invitrogen, Carlsbad, CA, U.S.A.). E. coli DNase I treatment was performed in order to remove any contaminating genomic DNA from the total RNA samples (Invitrogen, Carlsbad, CA, U.S.A.). cDNA was prepared from individual RNA samples using random primers (Invitrogen, Carlsbad, CA, U.S.A.). cDNA samples were used to conduct semi-quantitative RT-PCR at 94˚C for 2 min, and then 35 cycles of 94˚C for 30 s, 50˚C or 55˚C for 30 s and 72˚C for 30 s then 72˚C for 10 min. Primers for RT-PCR are listed in S2 Table. The amplified products were resolved on a 2% (w/v) agarose gel through electrophoresis at 8 V/cm. Normalization of the gene expression for F. virguliforme genes in qRT-PCR was carried out using the FvTox1 (g6924) transcript levels because the transcript levels of

Results
The goal of this study was to identify candidate pathogenicity genes for understanding the mechanisms used by F. virguliforme to cause root necrosis in soybean. Seven day-old etiolated seedlings were inoculated with the F. virguliforme Mont-1 isolate showed symptoms on the 10th day after inoculation (S1 Fig). Infected seedlings showed very mild symptoms until five day post inoculation (dpi). Severe root necrosis was observed by 10 dpi. Therefore, we classified the infected seedlings to two groups: (i) infected roots harvested 3-d and 5-d post inoculation as early infection phase (early infection) and (ii) 10-d and 24-d post inoculation as the late infection phase (late infection) (Fig 1; S1 Fig). We consider that during the early infection phase the pathogen is most likely in the biotrophic phase and turns to the necrotrophic phase at least by 10-d post inoculation. In order to identify most of the F. virguliforme genes induced during the infection process, deep sequencing of individual RNA samples was conducted.
Since only a small proportion of the RNA transcripts of infected soybean root tissues is encoded by F. virguliforme genes, deep sequencing of an individual RNA sample collected from three biological replications was conducted in a single lane to detect transcripts of most transcribed genes including those that are expressed at low levels. Since the deep-transcript sequencing was conducted only once, we could not conduct any statistical analysis. We considered only those genes as expressed genes that have shown to contain at least three sequence reads. We validated the transcriptomic data by conducting semi-quantitative RT-PCR and qRT-PCR analyses of 34 and eight F. virguliforme genes, respectively. Furthermore, only F. virguliforme genes showing at least 10-fold or more changes in transcript levels between infected roots and average expression levels of germinating conidia and mycelia were considered as differentially expressed genes (DEGs) for GO term enrichment analyses.

F. virguliforme gene expression patterns among germinating conidia, mycelia and infected soybean roots
The numbers of F. virguliforme and soybean sequence reads were calculated for early and late infection stages. F. virguligforme transcripts comprised less than 4% of the total transcripts in infected soybean roots (Fig 2A). An increased proportion of F. virguliforme transcripts was detected in late infection as compared to the early infection stage (Fig 2A). The sequence reads of four tissue samples: (i) germinating spores, (ii) mycelia; (iii) early infection, and (iv) late infection are presented in Fig 2B. Of the 14,845 predicted genes, 13,224 F. virguliforme genes were expressed among the three-tissue samples. Of the 13,224 F. virguliforme expressed genes, . Venn diagram showing unique and common F. virguliforme genes among germinating spore, mycelia and infected soybean roots. S, genes expressed only in germinating conidia; M, genes expressed only in mycelia; I, genes expressed in infected roots; SM, genes expressed in both spores and mycelia; SI, genes expressed in both spores and infected roots; MI, genes expressed in mycelia and infected roots; SMI, genes expressed in spores, mycelia and infected roots. The genes with a minimum of three reads were considered in developing the Venn diagram. The 984 genes having sequence reads below three were not considered for the Venn diagram.
doi:10.1371/journal.pone.0169963.g002 9,815 genes were expressed in all samples. Eighty-one percent of F. virguliforme (12,017 genes) were expressed in germinating spores (Table 1). Among these genes, 620 genes were specifically expressed in germinating conidia (Table 1; Fig 2C). In mycelia, 82% of the predicted F. virguliforme genes (12,208 genes) were expressed, of which 548 ( Fig 2C) are unique to mycelia (Table 1; Fig 2C). Of the 14,845 predicted F. virguliforme genes, 10,626 (72%) were expressed in infected tissues, of which 224 were expressed only in infected roots. The number of expressed genes was increased during infection from 64% of the predicted F. virguliforme genes during early infection to 70% in the late infection stage (Table 1). Among the 14,845 predicted F. virguliforme genes, 2,578 were expressed only in germinating spores and mycelia and are most likely involved in fungal growth and development ( Fig 2C).

Identification of differentially expressed F. virguliforme genes during infection
To quantify the expression patterns of F. virguliforme genes, a digital measure of relative gene expression was applied [31]. Typically, large genes have more reads than small genes even if they have the same transcript levels. To avoid this gene size-associated read number bias, the RPKM value for each gene was calculated as described in materials and methods. Fold changes in gene expression during early and late infection stages was calculated by comparing RPKM values of individual genes in infected roots with the average RPKM values of the corresponding genes in spores and mycelia. Of the 10,626 genes expressed in infected roots, 1,886 genes showed at least a 2-fold increase in expression levels when compared to germinating conidia and mycelia (Table 2). Of these 1,886 genes, 80 showed over 50-fold, and 33 showed over 100-fold induction in infected roots as compared to their corresponding average transcript levels in germinating conidia and mycelia ( Table 2). In infected soybean roots, expression of 4,204 genes, transcribed in germinating spores and mycelia, were suppressed during infection ( Table 2).

Semi quantitative RT-PCR and qRT-PCR validation
To validate the expression data gathered by deep-sequencing of transcripts, 34 genes with !10-fold average induction in infected roots as compared to the corresponding averages in germinating spores and mycelia were selected to conduct semi-quantitative RT-PCR (S2 Table). Expression patterns revealed by RT-PCR analyses were consistent with the digital expression data of 30 of the 34 selected genes (Fig 3). Four failed to show expression in RT-PCR, which could be due to primer-or PCR-condition-related issues. For 11 F. virguliforme genes, we observed amplification of transcripts collected from germinating spores in RT-PCR. The RPKM values of these 11 genes were very low in germinating spores (Fig 3).
To validate the semi-quantitative expression data gathered by RT-PCR, qRT-PCR was conducted on a few randomly selected expressed genes (Fig 4), RT-PCR data of which are shown in Fig 3. Overall, expression pattern of F. virguliforme genes determined by RT-PCR was in agreement with the expression pattern of the selected genes in qRT-PCR analysis. Our RT-PCR and qPCR data validated the transcriptomic data gathered from single RNA samples, pooled from three biological replications (Figs 3 and 4).

Blast2GO Functional categorization of induced F. virguliforme genes
Gene ontology associates molecular functions, cellular components and biological processes to a gene if it shows high sequence identity to previously characterized genes. We used Blast2GO to classify the F. virguliforme infection-induced genes based on their biological processes and molecular function [32]. F. virguliforme genes with !10 and !50 fold increase in expression in infected roots as compared to germinating conidia and mycelia were classified for biological processes and molecular functions to understand the overall basic mechanisms of virulence for this fungal pathogen. Considering the availability of transcript sequences for only a single pooled sample of RNAs, genes with lower than 10 fold changes in expression were not considered in this study. Among the 369 genes with !10-fold up-regulation, 273 genes (79%) could be classified by Blast2GO analysis into 17 biological process groups. The most prominent categories of genes are involved in: (i) carbohydrate metabolic processes, (ii) transport, (iii) transcription, (iv) regulation of biological processes, and (v) lipid metabolic processes (Fig 5A and  5C).
GO molecular function term enrichment analysis was also conducted for the 369 F. virguliforme genes with !10-fold up-regulation. This led to the classification of 273 genes into 11 categories. The largest GO category includes genes encoding enzymes with oxidoreductase and hydrolase activities (Fig 6A and 6C). On the other hand, GO annotation for molecular functions of F. virguliforme genes with !50-fold induction in infected tissues revealed five classes of genes, with most genes encoding enzymes with oxidoreductase, peptidases and hydrolase activities, and transporters (Fig 6B and 6C).

Genes with unknown function induced during infection
Ninety-six F. virguliforme genes with !10-fold increased expression could not be assigned to a functional category by Blast2GO analysis (S3 Table). Of these 96 genes, 64 were annotated by conducting BLASTX search. One of the genes (g9838), showed high identity to a stress responsive a/b barrel domain containing protein identified in Aspergillus niger (XP_003188860). Three polyketide synthetic genes including: snoal-like polyketide cyclase protein (g13126), snoal-like polyketide cyclase family protein (g14688) and lovastatin-like diketide synthase gene  (g14617) were identified. Polyketides such as the T-toxin produced by Cochliobolus heterostrophus [33,34] have been shown to be virulence factors and host specific toxins. We also identified four genes (g12373, g12372, g14510, and g5504) encoding CFEM domain containing proteins. CFEM domains are unique to fungi and are found in large numbers in pathogenic ascomycetes relative to non-pathogenic ascomycetes [35]. We failed to find homologous functionally characterized genes for 32 F. virguliforme genes that showed !10-fold induction during infection. These genes could encode species-specific factors, some of which could be involved in virulence.

Identification of candidate secretory proteins
The secretory peptides including virulence factors and digestive enzymes are used by fungi to attack plants. Therefore, secreted proteins are considered as candidate virulence factors involved in suppression of defense mechanisms (e.g. effector proteins), degradation of host proteins (e.g. peptidases), degradation of antifungal compounds (e.g. glyceollin detoxification enzymes) or degradation of host tissues (e.g. hydrolytic enzymes). Seventy-three of the 369 F. virguliforme genes with !10-fold induction in infected roots were shown to encode putative secretory proteins (S4 Table). Blast2Go analysis of these proteins for cellular components classified these genes mainly into extracellular and membrane-associated proteins. We investigated the early and late infection-specific genes encoding candidate secretory peptides/proteins (Table 3). Secretory genes induced during early infection stage (S1 Fig) encode proteins that are unlikely have hydrolytic activities; whereas, in the late infection stage many induced genes encode enzymes that have hydrolytic activities ( Table 3).

Identification of candidate virulence genes
Virulence genes are involved in disease development, but are not essential for completing the pathogen's life cycle [36]. In order to identify virulence genes, the sequences of the 369 genes with !10-fold induction during root infection were investigated to determine if any of them show high identity to any genes with known virulence functions. Twenty-seven candidate virulence genes were identified from this search (S5 Table).
Genes expressed during infection only may also be candidate virulence genes. Two-hundred and twenty-four F. virguliforme genes were expressed only in the infected soybean roots. Some of these genes showed very low RPKM values. However, 83 of the 224 genes showed high transcript levels (more than 20 reads in infected tissues) and could be considered candidate virulence genes. Blast2GO analysis for possible molecular functions of these genes showed that a majority of the 83 induced genes encode the pectate lyases and glucoside hydrolyases involved in cell wall degradation (S6 Table).

Discussion
To better understand how F. virguliforme causes root necrosis in soybean, we conducted a comparative transcriptomic analysis to identify genes that are induced during infection of soybean roots. F. virguliforme is considered to be a hemi-biotrophic fungal pathogen with an initial biotrophic phase [8][9][10]37]. To provide molecular evidence supporting this hypothesis, we developed a model system with 7-day old etiolated seedlings (Fig 1; S1 Fig; [11]). This system provides a uniform infection of root tissue by the pathogen. We observed that over 10,000 (72%) F. virguliforme genes are induced during infection as opposed to only few hundreds in our initial study of the infected roots of light-grown soybean seedlings (B.B. Sahu and M.K. Bhattacharyya, unpublished). Less than 4% of the transcripts from infected roots of etiolated  Fold changes in expression levels during early and late stages are calculated against the mean RPKM values of individual genes in germinating conidia and mycelia. 2 Minimum of two-fold differences in expression levels between early and late infection stages are presented. Rest of the genes encoding putative secretory peptides or proteins can be found in S4 Table. 3 RC values are confidence scores for secretory protein prediction. RC value 1 indicates highest confidence; while 5 the lowest. soybean seedlings are F. virguliforme-specific (Fig 2). Our deep sequencing approach identified 10,626 (72%) of the 14,845 predicted F. virguliforme genes that were expressed during infection. Of these genes, 224 were expressed only in infected roots based on their absence in germinating conidia and mycelia. In addition to classifying the F. virguliforme genes based on their identities with annotated genes, we identified genes encoding putative secretory peptides/protein that may play a role in virulence. We identified 27 candidate virulence genes based on their homologies to other functionally characterized virulence genes (S5 Table).

Genes induced during early infection stage
Over all, relatively few of the infection-inducible genes showed higher expression levels in the early infection stage as compared to that in the late infection stage. These genes may be important in establishment of the infection process. Of these genes, g9652 (Table 3) showed high identity to a triacylglycerol lipase gene, a virulence factor gene in Mycobacteria tuberculosis [38,39]. In pathogenic fungi, it has been suggested that lipases in general are involved in the penetration of the waxy cuticle [40]. It has been shown that a lipase (FGL1) determines the virulence of the wheat pathogen, F. gramminerum [41]. F. virguliforme g13527 protein (Table 3) shows high identity to an extracellular matrix protein with GPI anchored domain and a transmembrane domain found in epidermal growth factors. Extracellular matrix proteins may be involved in cell adhesion and are secreted by fungi. GPIs are membrane or cell wall proteins. GPI7 in F. graminearium governs the virulence function [42].
Phytopathogenic fungi relay signals upon sensing the cuticle. CHIP2 and CHIP3 encoding hard surface inducible proteins are induced in Colletotrichum gloeosporioide following contact with the cuticle [43]. Expression of the F. virguliforme homologue (g12834) of CHIP2 during early stage infection may suggest its role in surface sensing and signaling for tissue entry (S3 Table).
Fungal pathogens use secreted extracellular enzymes to degrade structural barriers to penetrate host cells [44,45]. Upon successful entry into the host plant, the fungi must resist plant produced antimicrobial compounds [46]. Fungi employ enzymes with oxidative-reductive properties to degrade antimicrobial compounds such as phytoalexins. In response to the pathogenic fungal attack, plants secrete polyamines [47] and phytoalexins [48] to defend against invading pathogens. Consistent with this, F. virguliforme genes encoding enzymes involved in detoxification of plant defense molecules show high expression during the early infection stages with declining transcript levels as infection progresses. Glyceollin is a soybean phytoalexin produced in response to microbial invaders [49]. The expression of the Fusarium oxysporum f. sp. pisi pisatin demethylase has been documented to detoxify pisatin, a pea phytoalexin, for establishing compatible interaction [50]. The F. virguliforme g13127 gene showing high identity to pisatin demethylase was induced to a high level during infection (S5 Table). Most likely g13127 encodes glyceollin demethylase and is involved in glyceollin metabolism. Along with phytoalexins, plants produce aromatic defense compounds that are toxic to fungi. The F. virguliforme g8644 gene encodes dienelactone hydrolase (S5 Table). The enzyme is involved in the β-ketoadipate pathway that catabolizes aromatic compounds into acetyl co-A and succinyl co-A [51][52][53].

Expression of toxin genes for foliar SDS development
As the fungus switches to the necrotrophic phase it produces toxins and necrosis inducing factors and starts killing plant cells for rapid nutrient uptake and growth. Trichothecene is a mycotoxin that severely impacts protein synthesis by inhibiting either the initiation or the elongation process of translation by interfering with peptidyl transferase activity [54]. Hydroxylation at carbon-2 (C-2) is the first committed intermediate of the trichodiene pathway and is mediated by the cytochrome P450 monooxygenase [55]. Expression of multiple cytochrome p450 related genes (g5125, g9837, g9839, g9830) increased in this pathogen during the different stages of infection (S1 and S5 Tables). In addition to the role in the trichodiene pathway, cytochrome P450 enzymes are considered to be involved in fungal adaptations [56,57]. The related cytochrome P450 genes are arranged in a group in the F. virguliforme genome. This grouped arrangement suggests that these genes may be coordinately expressed and operate as a cassette.
To date at least 12 potential toxins have been identified in F. virguliforme including proteinacious FvTox1, non-ribosomally produced peptides, polyketides and effector proteins [11,12,14]. We observed increased expression of necrosis inducing peptide 2 (FvNIP2, g12151) during infection of F. virguliforme, thus identifying it as a potential F. virguliforme toxin (S1 and S5 Tables; [14]). We did not find increased expression of FvTox1 (g6924) in infected root. FvTox1 has been shown to be important for foliar SDS symptom development [11,17]. This gene is however constitutively expressed in the fugal mycelia and germinating spores (S1 Table).

Induction of genes for root necrosis during late infection phase
The necrotrophic phase in infected soybean roots presumably begins with the accumulation of hydrolytic and glycosyl hydrolases to cause root necrosis. In our model system, roots of etiolated seedlings showed very little visible signs of root necrosis at early stage of infection; however, in the late infection stage root rotting due to necrosis was observed (S1 Fig). Among the genes encoding candidate virulence peptides/proteins with secretory signals (S3 Table), only a few are strongly induced in the early infection stage; whereas, a large number of genes encoding secretory proteins are induced during late infection stage ( Table 3). The pattern of gene expression completely changed as F. virguliforme entered the necrotrophic phase. A large number of pectate lyases with secretary signals involved in degradation of the cell wall component pectin are induced (Table 3). Among the infection-induced genes, a majority showed increasing levels of expression during the late stage of infection. The genes with enhanced expression during late stage infection include functional categories such as localization, cell hydrolysis, oxidation-reduction, membrane transport, many degradative enzymes such as pectate lyases, glucoside hydrolases, and elastinolytic metalloproteinase (Figs 5A, 6A and 6B; S4 and S6 Tables).
Investigation of the infection-induced (i) genes encoding putative secretory peptides/proteins (Table 3; S4 Table), (ii) genes specifically induced during infection (S6 Table) or (iii) infection-induced genes with homology to functionally characterized virulence genes (S5 Table), strongly suggest that during the late infection stage a large number of genes encoding cell wall degrading enzymes are induced. Expression of genes encoding degradative enzymes is common in other necrotrophic fungi. Necrotrophic fungi have also been shown to use a wide array of cell wall degrading enzymes to breakdown host tissues for penetration. Taken together, this suggests that degradative enzymes may play a major role in changing the fungus from its biotrophic phase to the necrotrophic phase. In the case of the pea pathogen Nectria haematococca, pathogenesis associated protein PEP2 has been shown to play a role in infection [58]. PEP2 is possibly involved in degrading the extracellular matrix proteins, and cleavage of the cell surface receptors. The F. virguliforme gene g8640, with high identity to PEP2 and induced during infection of soybean roots, may be involved in degradation of soybean root tissues (Fig 6; S1 and S5 Tables).
Plants cell walls are composed of complex carbohydrate and degradation of these complex walls requires specialized carbohydrate degrading enzymes. Hemicellulose constitutes a large fraction of the cell wall and contains xyloglucan, glucuronoarabinoxylan, mannan, galactan, arabinan, mixed-linked glucan, and glucuronoarabinoxylan [59]. During infection, necrotrophic pathogens degrade these building blocks by increasing the secretion of cell wall degrading enzymes such as pectate lyase and polygalacturonases [60][61][62].
Many cell wall degrading enzymes were found to be induced during infection with a dramatic increase in expression during the late infection phase, presumably to degrade the plant cells when the fungus becomes necrotrophic. In F. virguliforme, two pectate lyase genes (g7676 and g11622), one exopolygalacturonase gene (g4533), one endo-beta-xylanase (g11621), and one cellobiohydrolase ii (g14515) gene were induced during late infection stage (S1, S4 and S6 Tables). Endo-β-1,4-xylanase has been shown to be important for infection by Fusarium oxysporum f. sp. lycopersici [63].
F. virguliforme gene g14032 encoding a putative elastinolytic metalloproteinase Mep is highly induced during late infection (S5 Table). Mep proteins belong to a family of proteins likely involved in degradation of plant tissues [64]. Increased accumulation of g14032 transcripts during late stage infection suggest that it may be involved in transitioning the pathogen from its biotrophic to the necrotrophic phase.
As the carbohydrates of the cell walls are degraded, the fungi encounter cell wall associated peptides that could be toxic. Following infection of soybean, F. virguliforme, genes g8474, and g10809 encoding carboxypeptidases, and g12211 and g12135 encoding leucyl aminopeptidases were induced (S1 and S5 Tables). Carboxypeptidases are specific proteases which hydrolyze the peptide bond of an amino acid at the C-terminus [65], whereas the leucyl aminopeptidases catalyze the hydrolysis of the leucine residues at the N-terminus. These infection-induced F. virguliforme genes may be involved in digesting the cell-wall associated peptides.
After successful entry the fungal pathogens travel through the intra and intercellular passages and come in contact with the structural proteins and enzymes important for host defense. Enzymes such as subtilases and alkaline proteinases have been shown to degrade defensive enzymes [66]. Increased expression of F. virguliforme gene g14667 encoding subtilase in the infected tissues especially during late infection suggests its possible involvement in nullifying host defense-related enzymes (S5 Table).

Conclusions
In this comparative transcriptomic study, we were able to identify putative virulence factors by the following approaches: (i) investigation of the functions of infection-induced genes based on sequence homology using Blast2GO analyses; (ii) identification of secretory proteins and their GO annotation for possible functions; (iii) search for candidate virulence genes through sequence homology search with functionally characterized virulence genes; and (iv) studying the genes that are only induced in infected roots. Expression of several infection-induced genes encoding enzymes with oxidation-reduction properties for degradation of antimicrobial compounds such as the phytoalexin glyceollin could be an important virulence mechanism in this pathogen during early biotrophic phase. Induced expression hydrolytic and cell wall degrading enzyme genes (pectate lyase, glycoside hydrolase, polygalacturonases) in F. virguliforme during soybean root infection parallels the similar observations made recently in the transition of multiple Zymoseptoria tritici genotypes [67]. Expression of a large number of genes encoding enzymes with catalytic and hydrolytic activities during late infection stage suggests that cell wall degradation is involved in establishing the necrotrophic phase in this pathogen. This study suggests that enzymes with hydrolytic and catalytic activities play an important role in the transitioning the pathogen from biotrophic to necrotrophic phase.  Table) was used for normalization of the FvTox1 expression levels. Data are means and standard deviations (SD) of two independent biological replications with three technical replications (n = 6). (TIF) S1