Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Defining the Predicted Protein Secretome of the Fungal Wheat Leaf Pathogen Mycosphaerella graminicola

  • Alexandre Morais do Amaral ,

    alexandre.amaral@embrapa.br

    Affiliations Embrapa LabEx Programme, Rothamsted Research, Harpenden, Herts, United Kingdom, Department of Plant Biology and Crop Science, Rothamsted Research, Harpenden, Herts, United Kingdom

  • John Antoniw,

    Affiliation Department of Plant Biology and Crop Science, Rothamsted Research, Harpenden, Herts, United Kingdom

  • Jason J. Rudd,

    Affiliation Department of Plant Biology and Crop Science, Rothamsted Research, Harpenden, Herts, United Kingdom

  • Kim E. Hammond-Kosack

    Affiliation Department of Plant Biology and Crop Science, Rothamsted Research, Harpenden, Herts, United Kingdom

Defining the Predicted Protein Secretome of the Fungal Wheat Leaf Pathogen Mycosphaerella graminicola

  • Alexandre Morais do Amaral, 
  • John Antoniw, 
  • Jason J. Rudd, 
  • Kim E. Hammond-Kosack
PLOS
x

Abstract

The Dothideomycete fungus Mycosphaerella graminicola is the causal agent of Septoria tritici blotch, a devastating disease of wheat leaves that causes dramatic decreases in yield. Infection involves an initial extended period of symptomless intercellular colonisation prior to the development of visible necrotic disease lesions. Previous functional genomics and gene expression profiling studies have implicated the production of secreted virulence effector proteins as key facilitators of the initial symptomless growth phase. In order to identify additional candidate virulence effectors, we re-analysed and catalogued the predicted protein secretome of M. graminicola isolate IPO323, which is currently regarded as the reference strain for this species. We combined several bioinformatic approaches in order to increase the probability of identifying truly secreted proteins with either a predicted enzymatic function or an as yet unknown function. An initial secretome of 970 proteins was predicted, whilst further stringent selection criteria predicted 492 proteins. Of these, 321 possess some functional annotation, the composition of which may reflect the strictly intercellular growth habit of this pathogen, leaving 171 with no functional annotation. This analysis identified a protein family encoding secreted peroxidases/chloroperoxidases (PF01328) which is expanded within all members of the family Mycosphaerellaceae. Further analyses were done on the non-annotated proteins for size and cysteine content (effector protein hallmarks), and then by studying the distribution of homologues in 17 other sequenced Dothideomycete fungi within an overall total of 91 predicted proteomes from fungal, oomycete and nematode species. This detailed M. graminicola secretome analysis provides the basis for further functional and comparative genomics studies.

Introduction

Plant pathogenic fungi and oomycetes secrete an arsenal of proteins and metabolites during infection of their hosts. Protein secretion in eukaryotes can occur via a classical or non-classical route [1].Whilst it is apparent that certain fungi secrete proteins via a non-classical route [2], the mechanism is to date unknown and no clear molecular signatures for this exist. Classical secretion on the other hand is mediated through the presence of an N-terminal signal peptide which establishes the basis of transit though the endoplasmic reticulum and golgi systems on route to secretion from the cell. Some of the proteins secreted by plant pathogenic fungi have assignable functions and may include a diversity of hydrolytic enzymes which attack the plant cell wall or degrade other complex carbon or nitrogen sources, for example secreted proteases and glucanases. However many secreted proteins lack any obvious functional motifs and are frequently too small to encode catalytic activities. These small secreted proteins are often referred to as effectors (or candidate effectors) and are believed to function to counteract or suppress host defences and/or mask detection by the plant immune systems. In many cases plants have evolved the capability to recognise either directly or indirectly these effectors through disease resistance (R) proteins and/or guardee proteins giving rise to the widely accepted “gene-for-gene” model of effector triggered immunity [3]. This frequently activates a particular defence response involving highly localised cell death, termed hypersensitive cell death, which is particularly effective against biotrophic pathogens. For fungi which ultimately have a necrotrophic lifestyle, some of the small secreted effector proteins have been shown to target plant susceptibility (S) proteins encoded by homologues of resistance genes. In these cases, the resulting S protein – effector interaction triggers widespread HR for the benefit of the pathogen. This remarkable “hijack” of plant disease resistance mechanism has been termed an “inverse gene-for-gene” interaction [4], [5]. Both examples highlight that pathogens require a very specific and selective effector repertoire which enable them to infect their often restricted range of host plants and/or particular plant tissues, to then cause disease and finally complete their lifecycle through asexual/sexual sporulation.

Mycosphaerella graminicola (anamorph Septoria tritici, recently renamed Zymoseptoria tritici (Desm.) Quaedvlieg & Crous, comb. nov. [6]), is the causal agent of Septoria tritici blotch (STB) disease [7]. STB currently ranks as one of the most economically important diseases of wheat in the UK and Western Europe, and a threat to yields worldwide [8]. M. graminicola exhibits both host and tissue specificity, infecting only the leaves of wheat (Triticum spp) plants. Following leaf penetration through stomata, the hyphae then grow intercellularly throughout the leaf mesophyll cell layer for at least seven days post inoculation in the absence of any visible symptoms of disease [9]. Typically 8–10 days after inoculation disease lesions begin to form on susceptible plants [9]. This transition is associated with induction of host defence responses sharing characteristics with a hypersensitive response and involving differential regulation of plant defence signalling pathways [10], [11]. These events culminate in loss of control of cell permeability resulting in leakage of nutrients from dying plant cells into the intercellular (apoplastic) spaces. This coincides with an exponential increase in fungal growth rate. Asexual sporulation structures (pycnidia) then form in the sub-stomatal cavities of necrotic leaf tissues. The asexual pycnidiospores extrude through stomatal openings and are dispersed via rain splash throughout the crop canopy, giving rise to polycyclic infections [9]. This strictly intercellular (or apoplastic) growth lifestyle of M. graminicola is a characteristic shared with other Mycosphaerellaceae plant pathogens, which distinguishes them from various other globally important fungal plant pathogens including Fusarium spp and Magnaporthe spp which penetrate host cells during at least one stage of their infection cycle [8].

The sequenced genome of M. graminicola isolate IPO323 was recently published [12]. The current gene model prediction (11,035) was greatly aided by the production of over 50,000 expressed sequence tags deriving from various libraries. The genome spans 21 chromosomes, although it has been established that the eight smallest of these can be lost without affecting the ability of the fungus to cause disease [12], [13], and these are now referred to as dispensable chromosomes (the dispensome). First analysis of the constituents of this genome indicated that M. graminicola may have less potential for degrading plant cell walls due to possessing a limited number of genes encoding appropriate hydrolytic enzymes. Conversely several protease encoding gene families were expanded suggesting that host cell protein degradation might be an important source of fungal nutrition during infection [12]. However it was unclear how many of these putative proteins were likely to reside in the overall protein secretome, as to date no proteome based analyses have been published for this organism.

It is assumed that many fungal effectors are most likely to be soluble, extracellular secreted proteins that do not become cross-linked into the fungal cell wall [14]. It was also unclear, until recently, to what extent M. graminicola might rely on the deployment of effectors to either facilitate the initial symptomless growth phase (evasion or suppression of plant defence), and/or to trigger host cell death underlying the appearance of disease lesions [15]. However we recently identified a key role for at least one fungal predicted secreted protein effector in facilitating the symptomless phase of leaf infection through its activity in suppressing chitin-mediated plant defences. This effector, referred to as Mg3LysM, contains three predicted LysM domains which function to bind chitin fragments and prevent the elicitation of plant defences [16]. Mg3LysM and other MgLysM effectors were identified on the basis of their homology to CfECP6, the first LysM effector identified in plant pathogenic fungi, via a purely biochemical route, from the tomato leaf mold fungus Cladosporium fulvum [17], [18]. The functional conservation of the unique LysM effector activities in several plant pathogenic fungi irrespective of host range highlights the power of comparative genomics for putative fungal effector discovery. Despite the first evidence for secreted effector protein function during symptomless colonisation, it remains wholly unclear how many putative secreted protein effectors M.graminicola possesses, and how many might function in suppressing early defences or in subsequently triggering defences to support the activation of host cell death signalling and to facilitate the necrotrophic growth phase and asexual sporulation.

The computationally predicted classical secretomes of phytopathogenic fungi and oomycetes present powerful tools to compare and contrast between species with different host and tissue specificities as well as nutritional preferences. For filamentous fungi, predicted and well annotated classical secretomes have been deciphered by purely bioinformatics approaches for the basidiomycetes Ustilago maydis [19] and Puccinia graminis [20] as well as the ascomycete Fusarium graminearum [21]. The three species, like M. graminicola infect one or more cereal host plant species but have differing tissues specificities and nutritional lifestyles. To further analyse the predicted classical secretome of M. graminicola we have performed a rigorous bioinformatics analysis. The results of these analyses are reported here, together with a 91 member interspecies comparison, involving 17 additional Dothideomycete fungi, most of which are plant pathogens including 5 other Mycosphaerellaceae species, in addition to other fungi with contrasting lifestyles, oomycete species and plant pathogenic nematodes. This comprehensive analysis provides a basis for further candidate effector protein discovery via follow-up genomics based approaches.

Results

The total predicted and refined secretome of M. graminicola isolate IPO323

In the current study we analysed the combined filtered and frozen gene call (11,035 unique proteins) of version 2 the M. graminicola genome in two phases. In the first stage (Figure 1A), designed to predict all possible secreted proteins (the “total” secretome), SignalP and TargetP were used to identify secreted proteins with signal peptides (1,369 proteins). Fifty-eight of these proteins were subsequently predicted to contain GPI anchors. After removal of the signal peptide sequence from each sequence, any mature proteins that contained a transmembrane domain (TM) were excluded. We then used the ProtComp software to exclude proteins that were probably not located in the extracellular space. This predicted the total secretome for M. graminicola and contained 970 proteins (including those with GPI anchors). This represents 8.8% of the total current predicted M. graminicola protein models. For completeness, and to assist with follow up comparative analyses, the results for the predicted total secretome with the larger size of 970 genes arising from stage 1 of the analysis are presented in Tab 1 in File S1.

thumbnail
Figure 1. The two stage automated secretome prediction pipeline of M. graminicola.

(A) Stage 1 used to predict the total secretome and (B) stage 2 used to predict the refined secretome. In stage 1, 37 of the 949 sequences are also predicted to be GPI-anchor proteins.

https://doi.org/10.1371/journal.pone.0049904.g001

Stage 2 (Figure 1B), designed to identify a subset of proteins with an even higher probability of being secreted (the “refined” secretome), applied more stringent conditions to further analyse this set of proteins. A second software package (WoLF PSORT) that also predicts the eventual location of proteins was used to define proteins that have a high probability of being secreted into the extracellular spaces (extracellular score >17). This resulted in a reduced set of 556 secreted proteins (Tab 2 in File S1). After this stage, the proteins that lacked a methionine in the first position (9 proteins) were removed (Tab 3 in File S1) and the 55 proteins with a TM domain predicted within the signal peptide sequence were also removed. This resulted in the final prediction for the refined secretome for M. graminicola (Tab 4 in File S1) and contained 492 proteins (i.e. 4.5% of the total current predicted M. graminicola protein models). Overall, the predicted mature protein length for the refined secretome ranged in size from 33 to 1,369 amino acids. We next downloaded from the JGI genome portal the annotation and functional classification (where available) for the 492 secreted proteins present in the MG 2 gene call (http://genome.jgi-psf.org/Mycgr3/Mycgr3.home.html). This information is also presented in Tab 4 in File S1. Out of these 492 proteins, 321 (65%) possessed information on protein function (Tab 5 in File S1, columns PFAM, KOG, CDD, jgi-domains, jgi-go_info, jgi-kog_info and definition) whilst 171 (35%) (Figure 1B, Table 1) were described as hypothetical or conserved hypothetical (Tab 6 in File S1).

thumbnail
Table 1. Distribution of secreted proteins throughout the 21 chromosomes of Mycosphaerella graminicola.

https://doi.org/10.1371/journal.pone.0049904.t001

EST expression support for the predicted refined secretome

Prior to sequencing the genome of the reference isolate IPO323, 27,000 ESTs were produced from 10 libraries using this isolate [22]. In addition, approximately 4,000 ESTs were produced from a UK field isolate [23]. All these ESTs are displayed and retrievable from the JGI genome website (http://genome.jgi-psf.org/Mycgr3/Mycgr3.download.ftp.html). We checked each of the 492 genes for any level of EST support displayed on the JGI genome browser. This led to a final total of 262 which have some level of EST support (Tab 4 in File S1). The remainder currently have none. Therefore to date approximately 53% of the refined secretome has EST support. However, this value is likely to be an underestimate of the proportion that encodes transcribed genes. For example, there are currently no ESTs aligned to the M. graminicola gene MgNLP, shown by qPCR to vary in expression during fungal growth in liquid culture and across phases of plant infection [24]. In summary therefore, this indicates that at least a minimum of 53% of the predicted refined secretome are actively transcribed genes in at least one biological situation. More specifically we identified 75 predicted proteins which had EST support from both in vitro culture and plant infection based libraries (see later). Whereas 127 had EST support solely from in vitro libraries and 60 had support only from plant infection libraries.

The distribution of the refined secretome throughout the genome sequence

To reiterate, the sequenced reference isolate of M. graminicola (IPO323) possesses 21 chromosomes with the smallest eight chromosomes being dispensable for plant infection [13]. Whilst 24 of the total set of 970 genes predicted to code for secreted proteins (Fig. 2A) reside on these dispensable chromosomes, none of the refined 492 predicted proteins reside on them.

thumbnail
Figure 2. The genes predicted to code for the secretome of M. graminicola displayed over the 21 chromosomes.

(A) The total 970 genes arising from stage 1 of the analysis. (B) The refined 492 genes arising from stage 2 of the analysis. (C) The 85 genes found so far to be M. graminicola specific.

https://doi.org/10.1371/journal.pone.0049904.g002

The location of the genes coding for the refined secretome proteins was explored over the 13 core chromosomes (Fig. 2B). The predicted secreted protein coding gene density (ie, number of secreted proteins per Mb) was comparable for chromosomes 1 through 11, with an average of 13.7 genes per Mb (Table 1). By contrast, on the two shortest core chromosomes, 12 and 13, the density was slightly higher at 20.6 and 19.5 genes per Mb, respectively (Table 1).

Identification of predicted secreted proteins with enzymatic functions

Among all 321 predicted secreted proteins with known or presumed functions, at least 64 are suggested to be involved in the degradation of polysaccharides (Table 2 and Tab 7 in File S1). Thirty-nine proteins have functions related to protein degradation and 29 are implicated in the modification of lipids (Tables 3 and 4). The genome sequencing of isolate IPO323 identified 184 glycoside hydrolases in total. Of the 184 predicted glycoside hydrolases, 54 are present in the refined secretome (Table 2) with 22 of these proteins predicted to be directly involved in modifying the plant cell wall (cellulose, hemicellulose and pectin). These secreted plant cell wall degrading enzymes (PCWDEs) comprise 28 glycoside hydrolases, 6 esterases and 1 pectate lyase (Table 5 and Tab 8 in File S1). A direct comparison with the predicted refined secretome of the wheat ear attacking fungus Fusarium graminearum [21] highlights a drastically reduced overall complement of secreted PCWDE's in M. graminicola (Figure 3) as initially alluded to by previous analysis of the total genome content [12]. All five predicted cellulases identified in the genome of M. graminicola [12] were found to be present in the refined secretome. In addition, the refined secretome possessed only two members of the Glycoside hydrolase family 61 (GH61), with protein Ids 33254 and 103512. Although still poorly studied, GH61 functions have been associated with improving cellulose breakdown when acting alongside cellulases [25][27], and members of this protein family were also found in reduced numbers in the genome of M. graminicola as compared to other plant pathogens [12]. For comparison, in F. graminearum at least 11 members of GH61were found in the predicted refined secretome [21]. M. graminicola is also predicted to secrete at least four cutinases (protein Ids 43394, 68483, 77282 and 99331) (Table 4). Finally, we identified 28 predicted secreted proteins with functions relating to the modification of the fungal cell wall (Tab 9 in File S1), including 2 members of the glycoside hydrolase family 18 and one member of glycoside hydrolase family 75 where the substrate is predicted to be chitin.

thumbnail
Figure 3. Comparison of the distribution of functional categories (main PCWDEs) identified in the predicted secretomes of M. graminicola and F. graminearum.

The various enzyme classes are given along with the actual number of enzymes in each class (in brackets).

https://doi.org/10.1371/journal.pone.0049904.g003

thumbnail
Table 2. The sub-set of M. graminicola genes that code for secreted proteins involved in the degradation of polysaccharides.

https://doi.org/10.1371/journal.pone.0049904.t002

thumbnail
Table 3. The sub-set of M. graminicola genes that code for secreted proteins involved in the degradation of proteins.

https://doi.org/10.1371/journal.pone.0049904.t003

thumbnail
Table 4. The sub-set of M. graminicola genes that code for secreted proteins involved in the lipids degradation.

https://doi.org/10.1371/journal.pone.0049904.t004

thumbnail
Table 5. The sub-set of total putative secreted proteins related to plant polysaccharides degradation found in M. graminicola.

https://doi.org/10.1371/journal.pone.0049904.t005

Hydrophobin-like proteins

Fungal secreted hydrophobins facilitate attachment to hydrophobic surfaces. Typically, they have eight cysteines in the mature protein sequence [28]. In the original annotation of the M. graminicola genome eight hydrophobin-like proteins were predicted [12]. Of these 4 are found in both the total and the refined secretome (48129, 96543, 108349 and 88691), 2 were only identified in the total secretome (95491and 96536) whilst the other 2 proteins were not predicted to be secreted (40724 and 117719). The sizes of the 4 hydrophobin-like sequences predicted in the refined secretome are very different (96, 143, 463 and 816 aa's) and two contain a different number of cysteine residues (8, 8, 15 and 46, respectively). This finding suggests that the current gene models should be reassessed for at least two of the predicted secreted hydrophobins.

The relative abundance of individual PFAM domains in the refined secretome

The refined M. graminicola secretome of 492 proteins contains 235 with at least one PFAM domain identified (see Tab 5 in File S1). Their relative abundance was determined and the most frequent PFAM domains identified (Table 6). The most frequently observed PFAM in the M. graminicola refined secretome was PF01328 corresponding to peroxidase_2. This was detected in 11 predicted secreted proteins. Other PFAMs present in high copy numbers included PF00135 corresponding to carboxylesterase present in 10 proteins, PF00732 corresponding to the glucose-methanol-choline oxidoreductase family (FAD ADP-binding domain) present in 9 proteins, PF00450 corresponding to Peptidase_S10 is present in 9 proteins and PF05199 corresponding to GMC_oxred_C (steroid-binding domain) is present in 8 proteins. Two M. graminicola proteins in the refined secretome possessed a very large number of PFAM domains. These were protein Id 95631 (PF02993, PF03154, PF03276, PF04652, PF05109, PF05518, PF06070, PF07174, PF08639, PF09726, PF09770, PF10667). However there is no clear function associated with these and manual inspection suggested the protein to be very rich in serine, threonine and proline residues. Protein Id 109621 possessed 5 predicted PFAM motifs (PF04625, PF06676, PF07174, PF10287 and PF10290) and was also rich in these particular amino acids but with no clear function predicted.

thumbnail
Table 6. Most frequent PFAM domains found throughout the secretome of Mycosphaerella graminicola (Mg), and corresponding frequency in Fusarium graminearum (Fg).

https://doi.org/10.1371/journal.pone.0049904.t006

Analysis of secreted proteins with no predicted enzymatic functions

A total of 171 proteins present within the refined M. graminicola secretome are currently described as hypothetical or conserved hypothetical proteins (Tab 6 in File S1). We therefore explored these sequences for the presence of a number of interesting features.

a. Tandem repeat containing proteins.

We have previously described a small family of predicted secreted proteins possessing internal tandem coding repeats, referred to as the MgTRPs (M. graminicola Tandem Repeat Proteins), some of which showed increased expression during plant infection [29]. The original prediction for secretion of these twenty-three proteins arose from SignalP scores and WoLF PSORT predicting extracellular as the most likely protein location. The current analysis predicted 21 of these proteins in the total secretome but did not predict secretion for MgTRP4 or MgTRP17 (Tab 1 in File S1). In contrast the refined analysis predicted secretion for only MgTRP1, 2, 14, 15, 16 and 17 (Tab 4 in File S1).

b. Putative effector and cysteine-rich secreted proteins.

M. graminicola isolate IPO323 possesses sequence homologues of a number of functionally characterised effector proteins identified in the tomato leaf mould pathogen Cladosporium fulvum. These include two in planta expressed homologues of CfECP6, referred to as Mg3LysM and Mg1LysM, which have demonstrated effector functions in M. graminicola with both similar and unique properties [16]. M. graminicola also has three putative homologues of the C. fulvum effector ECP2 which are referred to as MgECP2, MgECP2-1 and MgECP2-2 [30]. Finally M. graminicola possesses one functional copy of the secreted necrosis and ethylene inducing proteins (NEPs) referred to as MgNLP. This protein does not play a significant role in virulence but its transcript is strongly up-regulated during leaf infection [24]. Our analysis of both the total and refined predicted secretome of M. graminicola identified all of these predicted proteins in both categories (Tabs 1 and 4 in File S1) adding further weight to our predictive approach. By analysing proteins containing 6 or more cysteines in the refined secretome, we identified 94 proteins (Tab 10 in File S1).

In fungi and oomycete plant pathogens a number of small (<200 amino acids) cysteine-rich apoplastic effectors have been described which lack homology to proteins in other species. We therefore examined the refined M. graminicola secretome for the presence of putative effectors of this type. In the refined secretome an overall total of 70 proteins have cysteine representing over 5% of the mature protein length. Intriguingly for all but 9 of these predicted proteins the mature protein length was less than 200 amino acids with a mean overall length of 143 amino acids. Eight of the 9 cys-rich proteins larger than 200 amino acids have some form of functional annotation. In contrast, of the 61 cys-rich proteins below 200 amino acids in length only 11 have any functional annotation. Therefore a total of 50 proteins have >5% cysteine and no functional annotation (Table 7) with a smaller sub-set of 22 proteins being comprised of more than 10% cysteine (Tabs 11 and 12 in File S1). As anticipated there were no putative cys-rich proteins detected on any of the dispensable chromosomes 14–21 and the largest number on any one chromosome was detected on chromosome 13 (seven proteins – Table 7).

thumbnail
Table 7. The predicted <200 amino acid cysteine-rich proteins (>5% cys) from M. graminicola isolate IPO323.

https://doi.org/10.1371/journal.pone.0049904.t007

Analysis of sequence motifs associated with fungal and oomycete effectors

All 492 refined secretome sequences were inspected for the presence of the degenerative RxLR-dEER [31] and Y/F/WxC motifs [32] located in close proximity to the predicted signal peptide sequence. No exact RxLR-dEER matches were found within the refined M. graminicola secretome. By contrast, the degenerative Y/F/WxC motif was present in close proximity to the predicted signal peptide cleavage site in 16 proteins (Tab 4 in File S1). All three motifs were identified YxC (5), FxC (5) and WxC (6) and 10 of these proteins had a mature length of <150 amino acids. Only three proteins were annotated, protein Id 37166 a xyloglucan xyloglucosyl transferase (WxC), protein Id 76589 a Glycosyl hydrolase family 45 member with the PFAM domain PF02015 (WxC) and protein Id 109621 with 5 PF domains (PF04625, PF06676, PF07174, PF10287, PF10290) (FxC). In addition, the Y/F/WxC motifs were found in the correct location in 10 hypothetical proteins with no annotation (Tab 6 in File S1).

Upon inspection of the refined secretome dataset for the presence of bipartite nuclear localisation signals (NLS), no proteins containing this motif were identified (Tab 4 in File S1). By contrast 130 proteins were predicted to contain at least one nuclear export sequence (NES) (Tab 4 in File S1).

Interspecies comparative analyses

Initially, these comparative analyses on the predicted refined secretome were done between the cereal ear and stem base infecting ascomycete species F. graminearum and the solely wheat leaf infecting M. graminicola. This revealed there were dramatic differences in both the number of predicted PCWDEs as well as the repertoire of other PFAM domain present (Table 6). As stated previously, the most frequently observed PFAM in the M. graminicola refined secretome was PF01328 which corresponds to peroxidase_2, detected in 11 copies. Whereas in the refined F. graminearum (Fg) secretome only 2 copies of PF01328 were identified. This represents a major quantitative difference between the two species. Other PFAMs clearly identified more frequently in the Mg refined secretome include PF09286 (Pro-kuma_activ protease), PF08760 (DUF1793), PF07971 (Glyco_hydro_92 glycosylhydrolase), PF05577 (Peptidase_S28) and PF01593 (Amino_oxidase). Conversely a number of PFAMs frequently identified in the Fg refined secretome were less frequent or absent from the Mg refined secretome (Table 8). Notably several PFAMs encoding different glycosyl hydrolases were reduced in number including PF04616, PF03443, PF00722, PF00704 and PF00295. Mg also has less predicted secreted proteins possessing the PFAMs; PF01083 (Cutinase), PF00544 (Pec_lyase_C pectate lyase), PF00150 (Cellulase) and PF01522 (Polysacc_deac_1) which may be implicated in the degradation of plant cell walls (as alluded to in the previous section). In addition, several other PFAMs frequently observed in the Fg refined secretome were entirely absent from the Mg secretome, including PF00657 (Lipase_GDSL), PF08031 (BBE Berberine bridge-like enzymes) and PF09044 (Kp4 killer toxin). Therefore there are clear qualitative differences overall in the protein activities predicted to reside in the refined secretomes of the two tissue specific fungal pathogens of wheat.

thumbnail
Table 8. Most frequent PFAM domains found throughout the secretome of Fusarium graminearum (Fg), and corresponding frequency in Mycosphaerella graminicola (Mg).

https://doi.org/10.1371/journal.pone.0049904.t008

The 492 members of the refined M. graminicola secretome were then subjected to an extensive interspecies comparison based upon 126 genome datasets representing 88 fungal species (including two genome databases for M. graminicola itself), 1 plant parasitic oomycete (Phytophthora infestans) and 2 plant parasitic nematodes (Meloidogyne incognita and M. hapla). The fungi analysed included predominantly ascomycete and basidiomycete species, one mucoromycotina and spanned a range of lifestyles including micro and macro saprophytes as well as plant and animal pathogens (Table S1). They include 17 additional Dothideomycete species, 14 of which are recognised as plant pathogens. Within this class were 5 additional members of the Mycosphaerellaceae (Tabs 13 and 14 in File S1). This analysis revealed a remarkable predicted expansion of homologues of the eleven predicted secreted peroxidases possessing PFAM domain PF01328 in the Mycosphaerellaceae and to a lesser extent to other members of the plant pathogenic Dothideomycetes (Figure 4). BLASTP analysis with one member of this putative protein family (protein Id 90087) identified a total of 19 unique proteins in the genome of M. graminicola (Tab 18 in File S1 and Figure 4A). Similarly high numbers were found in the genomes of the Mycosphaerellaceae species, including Cercospora zeae-maydis (18), Dothistroma septosporum (20), M. fijiensis (17) and Septoria musiva (15), whilst slightly fewer were observed in Septoria populicola (7). This latter number was similar to other non- Mycosphaerellaceae plant pathogenic Dothideomycetes including Stagonospora nodorum (12) which possessed the most homologues outside the Mycosphaerellaceae species, and Pyrenophora tritici-repentis and P. teres (9 each). The Dothideomycetes with the fewest homologs were Rhytidhysteron rufulum which colonises wood or other dead plant tissues (1) and the saprophyte Hysterium pulicare (3). Outside the Dothideomycetes most homologues were found in the plant pathogen Colletotrichum higginsianum (8) whilst Fusarium graminearum was predicted to have four homologues. The overall pattern described for protein 90087 was also observed when total BLASTP hits from all 11 members of the predicted M. graminicola secreted peroxidases were plotted (Figure 4B). Homologues of these genes were notably absent from all analyses members of the Saccharomycotina including the animal pathogens (Candida species) and the free living yeasts (Saccharomyces and Schizosacharomyces species).

thumbnail
Figure 4. Interspecies genome BLASTP counts for homologues of the M. graminicola secreted protein family with peroxidase/chloroperoxidase activity (PFAM; PF01328).

(A) The number of unique protein homologues of protein 90087, a representative member of the secreted peroxidase family. (B) Total number of BLASTP hits in each species towards the 11 predicted secreted peroxidases from M. graminicola. Black bars highlight Mycosphaerellaceae species; red bars highlight other Dothideomycetes. Analysis performed at a BLASTP cut-off of e-5.

https://doi.org/10.1371/journal.pone.0049904.g004

a. The distribution of M. graminicola functionally uncharacterised cysteine rich proteins.

The global comparative analysis of all 492 proteins from the refined secretome against these 126 genome datasets also identified 85 proteins that may be unique to M. graminicola (no homologues identified in any species at an e value cutoff of <10−5) (Tabs 15–18 in File S1). None of these 85 proteins were located on the dispensable chromosomes (Figure 2C). Of the 85 potentially M. graminicola specific proteins, 38 (45%) were cysteine-rich (>5% cys) (Tables 1 and 7). Some predicted secreted proteins also had homology to one another (i.e. they could be considered twins) or to genes present in the genome sequence currently not predicted to encode secreted proteins (Table 9). Amongst, these predicted 85 unique M. graminicola proteins, 10 possessed the Y/F/WxC motifs [32] located in close proximity to the predicted signal peptide sequence (Tab 17 in File S1). Of these, 5 proteins had a cysteine content >10%.

thumbnail
Table 9. Features of the predicted secreted proteins unique to M. graminicola based on interspecies BLASTP analysis (e-5).

https://doi.org/10.1371/journal.pone.0049904.t009

In total, 12 functionally unclassified cys-rich proteins possessed homologues amongst the other species analysed. The distribution of these predicted proteins varied across the genomes analysed. For example 11 of these proteins had a homologue in the genome sequence of at least one other Dothideomycete species (Tab 15 in File S1). In almost every case at least one copy was found in the related maize leaf infecting species Cercospora zeae-maydis. Homologues of one cys-rich protein in particular (97526) were found only in this related species, appearing as a small gene family comprising six members. Whilst, homologues of protein Id 102996, which is both cysteine rich and possess a YSC motif nearby the signal peptide, was found only in Cercospora zeae-maydis as a single copy gene. The distribution of homologues of the other conserved proteins varied considerably. Whilst many were detected only in related Mycosphaerellaceae pathogens (eg proteins 95714, 96536, 97077, 97526, 88664, 90533 and 108482) others had more widespread distribution. Homologues of protein 41491 were identified in fungal pathogens and saprophytes but curiously not from any other Mycosphaerellaceae species or any other Dothideomycete fungus. Homologues of protein 96910 were only found in plant pathogenic fungi whilst homologues of protein 110386 were most widespread overall indicated by their presence in 71 of the queried genome databases representing over 50 different fungal species (Tab 15 in File S1).

b. Copy number and distribution of the functionally non-annotated sequences.

The 171 predicted secreted proteins with no annotation were inspected for copy number in M. graminicola itself, and in 90 other species of predominantly fungi, but also including the oomycete P. infestans and the nematodes M. incognita and M. hapla (Tab 15 in File S1). To do this each protein was subjected to BLASTP analysis against each of the named genomes using a cut off value of e-5 and the unique protein hits then identified. Sixty-three predicted proteins were found to be M. graminicola species specific, of which 48 were single copy, 11 were present in two copies and 4 as three copies. For the sequences 91702, 90420, 41491 and 107811 although no homologous sequences were found in other Dothideomycete species, similar sequences were found in one or more species outside this class (Tab 19 in File S1). This rare distribution suggests these novel sequences may have arisen from elsewhere in the fungal kingdom.

The distribution of some of the M. graminicola sequences within the 17 Dothideomycete species was particularly striking (Table 10). For example, protein 91252 was found as 11 copies in M. graminicola, but was present between 0 and 4 copies in the taxonomically related species. The unrelated proteins 68660, 89162 and 89375 were found as 7, 6 or 4 copies in M. graminicola, but were present between 0 and 14 copies, 1 and 8 copies and 0 to 8 copies, respectively, in the other species. The related proteins 32157 and 33493 were found in 5 copies in M. graminicola but between 4 and 13 copies in the other Dothideomycete. A number of the other sequences were particularly prevalent in a single species. For example, the related proteins 107050 and 73873 were most prevalent in Botryosphaeria dothidea, 6 or 7 copies versus 4 in M. graminicola. Whilst, 77324 present only as a single copy in M. graminicola, had 12 copies in Botryosphaeria dothidea and 97526 again a single copy sequencing in M. graminicola had 6 copies in Cercospora zeae –maydis and none in the other Dothideomycete species. The species distribution of a few of the single copy M. graminicola proteins was also particularly striking. For example, 110386 was present between 12 and 17 copies in the two Cochliobolus species whereas the other Dothideomycete species had only a few copies (<6 throughout). In contrast, it was noticeable that for a few proteins, the copy number was slightly higher in M. graminicola compared to any of the other Dothideomycete species. Finally, 15 proteins had a very narrow taxon distribution being found only in M. graminicola and one other Dothideomycete species.

thumbnail
Table 10. Distribution and copy number of M. graminicola homologues in other Dothideomycetes and other fungal species.

https://doi.org/10.1371/journal.pone.0049904.t010

c. Global analysis of all refined secreted proteins shared with other organisms.

Of the 407 M. graminicola predicted proteins which had homologues in at least one other species analysed, the largest number of homologues were found in the maize leaf infecting Mycosphaerellaceae species Cercospora zeae-maydis with 352 of the 407 having at least one homologue at e-5. Tabs 15 and 18 in File S1 highlight that there was an overall trend for the largest number of homologues to reside in the predicted proteomes of other Mycosphaerellaceae species. These were then followed by other Dothideomycete cereal infecting pathogens and then other plant pathogenic fungi (Table 11). The organisms that had fewest homologues to the 407 M. graminicola proteins were the plant parasitic nematodes (only 55 and 56 of the 407 possessed homologues respectively) followed by the ascomycete budding yeasts (typically having <112 with homology to the 407), the plant parasitic oomycete P. infestans (125) then the cereal infecting basidiomycete Puccinia species (<134) and finally lower fungi from the phylum Chytrids (<144).

thumbnail
Table 11. Conservation of the M. graminicola (Mg) genes, predicted to encode secreted proteins, amongst the 124 fungal, oomycete and Chytrid genomes assessed and then presented according to overall species distribution or lifestyle.

https://doi.org/10.1371/journal.pone.0049904.t011

This analysis revealed nine proteins from the Mycosphaerella graminicola secretome only found in fungal pathogens of wheat plants or other cereals (Table 12), including an expansion with six copies of the 97526, a protein with unknown function, in the genome of Cercospora zeae-maydis, a fungus of the class Dothiodiomycete that causes the gray leaf spot, a foliar disease of corn, and an expansion of the hydrophobin-like protein 92962, with five other homologues in the genome of M. graminicola (108349, 109435, 95631, 94883 and 96944), all secreted. In addition, the gene 92805, with PFAM domain related to fibronectin attachment, shows a unique homologue in Fusarium verticillioides, a very common fungal species (class Ascomycete) that can infect all organs of maize plants. However, only the 102996, a protein without function assigned, found exclusively in C. zeae-maydis shows EST support so far (Tab 20 in File S1).

thumbnail
Table 12. Number of putative secreted proteins from Mycosphaerella graminicola secretome only found in fungal pathogens of wheat plants or other cereals.

https://doi.org/10.1371/journal.pone.0049904.t012

For completeness, we included the BLASTP analyses results with e-values less than e-40 and e-100 in the Tabs 21 and 22 in File S1 and the analysis applied to secreted proteins found in other plant pathogenic fungi in the Tab 23 in File S1 and Table S2 for comparison.

EST analysis of the predicted M. graminicola unique genes

To provide some further annotation for the 85 M. graminicola unique genes, and also to explore for any possible biological roles these may genes play, their presence/absence was explored in the previously published ESTs libraries prepared from the 9 in vitro conditions and 4 in planta conditions [22], [23]. Interestingly, 9 sequences were only present in one or more of the in planta derived libraries, 6 were present in at least one in vitro and one in planta library, whilst a further 24 were only present in one or more of the in vitro libraries. The three M. graminicola unique genes with the highest level of EST support were 80321 and 99917, present in most of the in vitro and in planta libraries and 105608 present in the most of the in vitro libraries. For completeness, the EST support present in the 13 libraries for each of the 85 M. graminicola unique genes as well as the rest of the refined secretome gene set is provided (Tab 20 in File S1). In total, 39 of the 85 M. graminicola unique genes had EST support.

Discussion

Septoria tritici blotch disease of wheat caused by Mycosphaerella graminicola represents a significant economic threat to global wheat production in the context of future food security concerns. It is also emerging as another model pathosystem to investigate mechanisms of fungal pathogenesis in plants and to a lesser extent host resistance responses [15]. Re-sequenced genomes of specific isolates and new molecular tools are fast appearing to investigate important questions relating to host specificity and mechanisms of infection [8]. This latter point is clearly of interest as this would appear to differ in one clear respect to that used by the more established models, in particular Magnaporthe oryzae and Fusarium graminearum, in that the entire cereal infection process occurs without host cell penetration. This suggests that the M. graminicola fungus must deploy an intricate means of communication with plant cells from the extracellular (apoplastic or intercellular) environment to facilitate infection. It is likely that this fungus uses various secreted protein effectors to assist the leaf infection process. Moreover this mode of plant infection would appear to be widespread amongst the Mycosphaerellaceae pathogens in the fungal class Dothideomycetes. These pathogens collectively are responsible for widespread and devastating diseases of major cereal and non-cereal crop plants as well as several cultivated tree species [33].

The current study has made use of the finished genome of what is regarded as the reference isolate of M. graminicola, IPO323. This isolate possesses 21 chromosomes, currently the largest number identified for any individual strain. However, the smallest eight chromosomes are dispensable for plant infection [12], [13]. We aimed to provide here both a “total” and “refined” predicted protein secretome for this fungus. The reason for this is the current lack of experimental proteome data for this fungus in which to verify our predictions. Therefore we chose to predict those that have some likelihood of secretion (total secretome) along with those with even greater likelihood (refined secretome) on the basis of the available bio-informatic resources. It was recently determined that M. graminicola has functional homologs of ECP6, a chitin binding secreted protein effector from the Mycosphaerellaceae species Cladosporium fulvum, which infects the leaves of the dicotyledonous tomato plant [16][18]. The analysis presented here identified both M. graminicola LysM effectors within the refined predicted secretome adding further weight to our approach. Moreover these data highlight the power of effector discovery by comparative genomics. However many secreted protein effectors identified in C. fulvum or various other plant pathogenic fungi and oomycetes are unique to these species, which complicates the identification of other key effector proteins. For this reason we performed a genome wide in silico analysis of the predicted M. graminicola proteome and compared this to fungi and oomycetes for which predicted proteome data are available to explore further this secretome. The set of 492 secreted proteins represents 4.4% of the M. graminicola genome, which is in accordance with many other predicted secretomes [34].

Global analyses highlighted several interesting features of the refined M. graminicola secretome. Firstly no members were predicted to reside on the eight smallest dispensable chromosomes perhaps supporting their redundant (or at least currently cryptic) roles in plant infection. Secondly we found no evidence for specific micro-regions or clusters of secreted proteins in contrast to what has been discovered for effector proteins in the genome of the basidiomycete plant pathogen Ustilago maydis [35]. Thirdly there was no particular association of the distribution of the 492 predicted secreted proteins with regions of the genome containing repetitive DNA or transposable elements (data not shown). This distinguishes M. graminicola from other Dothideomycete plant pathogens including Leptosphaeria maculans which possesses a variety of effector proteins in such regions [36]. Fourthly the global interspecies analysis identified that most homologues of the 492 proteins were found in the genome of the related Mycosphaerellaceae species Cercospora zeae-maydis which like M. graminicola is a leaf-specific cereal pathogen, but in this case of maize. The next most similar set of species were all plant pathogenic species of the Mycosphaerellaceae including Dothistroma septosporum a foliar pathogen of many pine species; Mycosphaerella fijiensis a foliar pathogen of banana causing Black Sigatoka disease and Septoria musiva and S. populicola which cause leaf spots and cankers on poplars, the latter most commonly affecting poplars of the section Tacamahaca and theirs hybrids. The ‘most closely related’ list then generally moves into other members of the Dothideomycetes with the exception of the presence of Colletotrichum graminicola, a plant pathogenic member of the Sordariomycetes. Interestingly this fungus is considered an archetypal hemibiotroph [37] and also displays specificity to leaves of graminaceous plants thereby suggesting some similarities in mode of infection and host specificity to M. graminicola. At the other end of the spectrum the low similarity to the ascomycete budding yeasts (members of the Saccharomycotina) is notable, perhaps highlighting specialisations associated with filamentous growth and/or plant pathogenesis.

The composition of the 492 proteins in the refined secretome is also noteworthy. In total, 65% had some annotation leaving 35% lacking annotation. Amongst the annotated genes were a number of secreted glycoside hydrolases implicated in plant cell wall degradation. However this number was comparatively small when compared for example with the number reported in the predicted refined secretome of the cereal ear, stem base and root infecting fungus Fusarium graminearum [21] (and this study) (Figure 3). Whilst there is evidence for initial intercellular colonisation of wheat ears by F. graminearum prior to cell wall attack [38], this difference in numbers most likely reflects the reduced capacity of the strictly apoplastic colonising non cell penetrating hyphae of M. graminicola to breach the cell walls of living plant cells, as had been previously suggested from analysis of the total genome content [12]. The analysis of PFAM domain abundance between the two secretomes of these cereal attacking fungi was also particularly striking and revealed many differences (Tables 6 and 8). The most abundant PFAM in the M. graminicola secretome was PF01328 corresponding to peroxidase_2 or “chloroperoxidase”. This was present in 11 members of the refined secretome. In contrast this was found in only two members of the F. graminearum refined secretome. More remarkably the global interspecies BLASTP analysis clearly highlighted that almost all of the analysed Mycosphaerellaceae plant pathogens contained similarly high numbers of these predicted proteins, above and beyond the numbers found in other Dothideomycete plant pathogens, and non-Dothideomycete fungi (plant pathogens, animal pathogens or saprophytes). Analysis of the physical distribution of these genes in the M. graminicola genome did not suggest clustering (Figure S1). These predicted secreted proteins function in various processes ranging from halogenation of natural products (eg synthesis of the antibiotic chloramphenicol) but also perhaps significantly all use H2O2 as a substrate [39]. Hydrogen peroxide and related reactive oxygen species (ROS) are well studied components of inducible plant defence responses [40]. With respect to M. graminicola infection of wheat leaves roles for H2O2 in inhibiting (or slowing) initial colonisation by the fungus have been suggested [41]. H2O2 is also produced in very large amounts during disease lesion formation and asexual sporulation [10], [42], [43]. It is therefore tempting to speculate that this enlarged family of putative secreted chloroperoxidases might in part allow the fungus to overcome initial plant defences and/or allow hyphae to tolerate environments with high H2O2 and oxidative stress typical of photosynthetically active leaves. The high numbers of predicted protein homologues in most of the plant pathogenic Mycosphaerellaceae species we studied, suggests that this may be an adaptation important for plant infection by many members of this family of fungi. This attribute might distinguish them from other plant pathogenic fungi with different tissue specificities and/or modes of infection.

Based on the interspecies BLASTP analysis a total of 85 predicted secreted proteins appeared to be unique to M. graminicola and most had no annotation. A significant proportion of these (37) were deemed to be small (<200 aa's) and cysteine-rich, which are features particularly well described for apoplastic effector proteins identified from a number of fungi and oomycetes [44], [45] . Effector proteins are considered to be virulence factors which assist colonisation by the pathogen through interfering with the activation of plant defence responses or counteracting components of these defence responses. They can also act as avirulence factors if the plant has evolved resistance proteins which may serve to “guard” or monitor changes which occur on the virulence targets of these effector proteins [45]. In total 55% of these small predicted secreted proteins (<200 aa's) so far have EST support. In addition, homologues of the bona fide effectors ECP2 and ECP6 from the exclusively tomato leaf apoplast infecting species Cladosporium fulvum have been identified. Very interestingly 10 of the predicted M. graminicola secreted protein possess a Y/F/WxC motifs [32] located in close proximity to the predicted signal peptide sequence. Of these 5 were also very cysteine rich. The function(s) of these motifs, originally identified in abundance in the genome of the haustorium forming and barley leaf infecting ascomycete species Blumeria graminis f. sp hordei, is not yet known.

This study established a baseline for further analyses which ideally should focus on the changes in gene expression throughout infection, on direct proteomics approaches to verify their predicted secretion and on refined evolutionary analyses. The large number of non-annotated sequences that still remain despite this study poses a further challenge, and therefore exploring their temporal gene expression patterns may provide the first clues to function. Additionally the sequencing of other Dothideomycetes species with different pathogenic and saprophytic lifestyles, other isolates of M. graminicola and the subsequent comparative analyses should reveal the repertoire of species -specific secreted proteins found in most M. graminicola isolates and those also found in very closely related species. An example of the latter would be the grass infecting species currently termed S1 which also has some ability to infect wheat leaves [46]. These types of studies on highly related species and/or other M. graminicola isolates will also reveal the more flexible parts of the M. graminicola secretome and which parts of the repertoire of predicted effector proteins exhibit the greatest sequence differences between isolates and/or species.

Materials and Methods

All protein sequence information can be retrieved from JGI (http://genome.jgi-psf.org/pages/search-for-genes.jsf?organism=Mycgr3) using the unique numerical identifier.

Bioinformatic analyses of the secretome

Stage 1.

Version 2 of the M. graminicola genome was downloaded from the JGI genome portal (http://genome.jgi-psf.org/Mycgr3/Mycgr3.download.html). The prediction of the refined M. graminicola secretome was based on the procedure described by Muller and colleagues [47] for U. maydis. We developed an automated secretome prediction pipeline based on this procedure using bash shell, awk and python scripts on a PC running Red Hat Enterprise Linux 5.2. Initially all proteins with a Target P Loc = S (TargetP v1.1; http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?targetp) or a Signal P D-score = Y (SignalP v3.0; http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?signalp) were combined [48], [49]. These were then scanned for transmembrane spanning regions using TMHMM (TMHMM v2.0; http://www.cbs.dtu.dk/cgi-bin/nph-sw_request?tmhmm) and all proteins with 0 TMs or 1 TM, if located in the predicted N-terminal signal peptide, were kept. GPI-anchor proteins were predicted by big-PI (http://mendel.imp.ac.at/gpi/cgi-bin/gpi_pred_fungi.cgi) [50]. ProtComp was also used to predict localization of the remaining proteins using the LocDB and PotLocDB databases (ProtComp v8.0; http://www.softberry.com).

Stage 2.

WoLF PSORT analysis was done using “runWolfPsortSummary fungi” in the WoLF PSORT v0.2 package, which estimates localisation sites with a sensitivity and specificity of approximately 70% [51]. All proteins predicted with an extracellular score >17 were kept in the final secretome dataset. The selection of this ‘cut-off’ point was tested using a range of experimentally verified secreted fungal proteins from other phytopathogens (Table S2 and Tab 23 in File S1). An extracellular score >17 had previously been used to refine the prediction of the secretome for Fusarium graminearum [21] and the results obtained agreed well (68%) with the available proteomics datasets obtained for the in planta and in vitro secretome of this fungus [52]. PFAM analysis was done using the PFAM database (ftp://ftp.ncbi.nih.gov/pub/mmdb/cdd/) and the rpsblast program in the NCBI blast+ software package (ftp://ftp.ncbi.nlm.nih.gov/blast/executables/blast/). The number of cysteine residues within the mature peptide and the search for degenerative YxC motifs were computed using custom python scripts. The number of internal sequence repeats was found using RADAR (http://www.ebi.ac.uk/Tools/Radar/) [53]. The detection of RNA transcripts for the genes of interest was explored by BLASTN analysis (e-100) of the 13 designated EST libraries available from the JGI website (http://genome.jgi-psf.org/Mycgr3/Mycgr3.download.ftp.html).

Analysis of chromosome location alongside other key features of the M. graminicola genome

To inspect the position of individual genes on the 21 M. graminicola chromosomes (Figure 2), the MgraMap tool was downloaded from www.OmniMapFree.org which displays a map of the complete M. graminicola genome (JGI version 2.0) [12]. The MgraMap was used according to methods described elsewhere [54].

Comparative analysis of the refined M. graminicola secretome

For the detailed follow up analyses, only proteins in the refined secretome from Stage 2 were used. The M. graminicola secretome was compared with 90 proteomes from other fungal, oomycete and plant pathogenic nematode species genomes, varying in host range, tissue specificity and lifestyle as well as several exclusively saprophytic species (Table S1). The fungal and oomycete genomes and their predicted gene repertoires were downloaded from either the BROAD or JGI websites or from species specific websites maintained by various research communities. For the comparative analyses, the conservation, absence or expansion of the genes coding for the M. graminicola secreted proteins was explored by BLASTP analysis, determined at two levels of confidence, p<e−5 and p<e−40.

Genes coding for proteins with a known function

The total putative secreted proteins related to plant polysaccharides degradation found in M. graminicola was predicted using annotation based on the protein family classification from the CAZy – Carbohydrate Active Enzyme website (http://www.cazy.org/). All proteins from the four classes of proteins from the CAZy catalogue (Glycoside Hydrolases, Glycosyl Transferases, Polysaccharide Lyases and Carbohydrate Esterases) were identified initially throughout the genome by a keyword search on the genome browser (http://genome.jgi-psf.org/pages/search-for-genes.jsf?organism=Mycgr3) and then selected for secreted proteins based on a comparison against the predicted secretome. Finally, a manual annotation based on BLASTP analysis against the non-redundant protein database and CAZy classification was done for each protein.

Data obtained from the JGI and displayed in several of the Supplementary files

For completeness and to assist in inter- sequence comparisons, in each row of Tabs 1 through 6 in File S1, the following information is given in the following columns: (A) protein ID, (B) gene name ID, (C) effector homologue, (D) chromosome location, (E) chromosome position, (F) coding strand, (G) predicted in frozen gene model, (H) predicted in filtered gene model and (I) number of introns predicted.

Supporting Information

Figure S1.

The distribution of genes encoding secreted peroxidases/chloroperidases (PFAM01328) in the M. graminicola genome.

https://doi.org/10.1371/journal.pone.0049904.s001

(PPTX)

Table S1.

The list of species used for the various comparative analyses.

https://doi.org/10.1371/journal.pone.0049904.s002

(XLSX)

Table S2.

Experimentally verified secreted proteins related to pathogenicity and virulence, present in other plant pathogenic fungi.

https://doi.org/10.1371/journal.pone.0049904.s003

(DOC)

File S1.

List of all proteins used for analysis in this study.

https://doi.org/10.1371/journal.pone.0049904.s004

(XLSX)

Acknowledgments

We would like to thank Martin Urban, John Lucas and Steph Heard at Rothamsted Research, for comments on the manuscript.

Author Contributions

Conceived and designed the experiments: AMDA JJR JA KEHK. Performed the experiments: AMDA JJR JA KEHK. Analyzed the data: AMDA JJR JA KEH. Contributed reagents/materials/analysis tools: AMDA JJR JA KEHK. Wrote the paper: AMDA JJR JA KEHK.

References

  1. 1. Prudovsky I, Tarantini F, Landriscina M, Neivandt D, Soldi R, et al. (2008) Secretion without Golgi. J Cell Biochem 103: 1327–1343.
  2. 2. Ridout CJ, Skamnioti P, Porritt O, Sacristan S, Jones JD, et al. (2006) Multiple avirulence paralogues in cereal powdery mildew fungi may contribute to parasite fitness and defeat of plant resistance. Plant Cell 18: 2402–2414.
  3. 3. Jones JD, Dangl JL (2006) The plant immune system. Nature 444: 323–329.
  4. 4. Friesen TL, Faris JD, Solomon PS, Oliver RP (2008) Host-specific toxins: effectors of necrotrophic pathogenicity. Cell Microbiol 10: 1421–1428.
  5. 5. Faris JD, Zhang Z, Lu H, Lu S, Reddy L, et al. (2010) A unique wheat disease resistance-like gene governs effector-triggered susceptibility to necrotrophic pathogens. Proc Natl Acad Sci U S A 107: 13544–13549.
  6. 6. Quaedvlieg W, Kema GH, Groenewald JZ, Verkley GJ, Seifbarghi S, et al. (2011) Zymoseptoria gen. nov.: a new genus to accommodate Septoria-like species occurring on graminicolous hosts. Persoonia 26: 57–69.
  7. 7. Eyal Z (1999) The Septoria tritici and Stagonospora nodorum blotch diseases of wheat. Eur J Plant Pathol 105: 629–641.
  8. 8. Dean R, Van Kan JA, Pretorius ZA, Hammond-Kosack KE, Di Pietro A, et al. (2012) The top 10 fungal pathogens in molecular plant pathology. Mol Plant Pathol 13: 414–430.
  9. 9. Kema GHJ, Yu DZ, Rijkenberg FHJ, Shaw MW, Baayen RP (1996) Histology of the pathogenesis of Mycosphaerella graminicola in wheat. Phytopathology 86: 777–786.
  10. 10. Keon J, Antoniw J, Carzaniga R, Deller S, Ward JL, et al. (2007) Transcriptional adaptation of Mycosphaerella graminicola to programmed cell death (PCD) of its susceptible wheat host. Mol Plant Microbe Interact 20: 178–193.
  11. 11. Rudd JJ, Keon J, Hammond-Kosack KE (2008) The wheat mitogen-activated protein kinases TaMPK3 and TaMPK6 are differentially regulated at multiple levels during compatible disease interactions with Mycosphaerella graminicola. Plant Physiol 147: 802–815.
  12. 12. Goodwin SB, M'Barek SB, Dhillon B, Wittenberg AH, Crane CF, et al. (2011) Finished genome of the fungal wheat pathogen Mycosphaerella graminicola reveals dispensome structure, chromosome plasticity, and stealth pathogenesis. PLOS Genetics 7: e1002070.
  13. 13. Wittenberg AH, van der Lee TA, Ben M'barek S, Ware SB, Goodwin SB, et al. (2009) Meiosis drives extraordinary genome plasticity in the haploid fungal plant pathogen Mycosphaerella graminicola. PLOS ONE 4: e5863.
  14. 14. Kleemann J, Takahara H, Stuber K, O'Connell R (2008) Identification of soluble secreted proteins from appressoria of Colletotrichum higginsianum by analysis of expressed sequence tags. Microbiol 154: 1204–1217.
  15. 15. Deller S, Hammond-Kosack KE, Rudd JJ (2011) The complex interactions between host immunity and non-biotrophic fungal pathogens of wheat leaves. J Plant Physiol 168: 63–71.
  16. 16. Marshall R, Kombrink A, Motteram J, Loza-Reyes E, Lucas J, et al. (2011) Analysis of two in planta expressed LysM effector homologs from the fungus Mycosphaerella graminicola reveals novel functional properties and varying contributions to virulence on wheat. Plant Physiol 156: 756–769.
  17. 17. Bolton MD, van Esse HP, Vossen JH, de Jonge R, Stergiopoulos I, et al. (2008) The novel Cladosporium fulvum lysin motif effector Ecp6 is a virulence factor with orthologues in other fungal species. Mol Microbiol 69: 119–136.
  18. 18. de Jonge R, van Esse HP, Kombrink A, Shinya T, Desaki Y, et al. (2010) Conserved fungal LysM effector Ecp6 prevents chitin-triggered immunity in plants. Science 329: 953–955.
  19. 19. Mueller O, Kahmann R, Aguilar G, Trejo-Aguilar B, Wu A, et al. (2008) The secretome of the maize pathogen Ustilago maydis. Fungal Genet Biol 45(Suppl 1): S63–70.
  20. 20. Joly DL, Feau N, Tanguay P, Hamelin RC (2010) Comparative analysis of secreted protein evolution using expressed sequence tags from four poplar leaf rusts (Melampsora spp.). BMC Genomics 11: 422.
  21. 21. Brown NA, Antoniw J, Hammond-Kosack KE (2012) The predicted secretome of the plant pathogenic fungus Fusarium graminearum: a refined comparative analysis. PLOS ONE 7: e33731.
  22. 22. Kema GHJ, van der Lee TAJ, Mendes O, Verstappen ECP, Lankhorst RK, et al. (2008) Large-scale gene discovery in the Septoria tritici blotch fungus Mycosphaerella graminicola with a focus on in planta expression. Mol Plant Microbe Interact 21: 1249–1260.
  23. 23. Keon J, Antoniw J, Rudd J, Skinner W, Hargreaves J, et al. (2005) Analysis of expressed sequence tags from the wheat leaf blotch pathogen Mycosphaerella graminicola (anamorph Septoria tritici). Fungal Genet Biol 42: 376–389.
  24. 24. Motteram J, Kufner I, Deller S, Brunner F, Hammond-Kosack KE, et al. (2009) Molecular characterization and functional analysis of MgNLP, the sole NPP1 domain-containing protein, from the fungal wheat leaf pathogen Mycosphaerella graminicola. Mol Plant Microbe Interact 22: 790–799.
  25. 25. Harris PV, Welner D, McFarland KC, Re E, Navarro Poulsen JC, et al. (2010) Stimulation of lignocellulosic biomass hydrolysis by proteins of glycoside hydrolase family 61: structure and function of a large, enigmatic family. Biochemistry 49: 3305–3316.
  26. 26. Battaglia E, Benoit I, van den Brink J, Wiebenga A, Coutinho PM, et al. (2011) Carbohydrate-active enzymes from the zygomycete fungus Rhizopus oryzae: a highly specialized approach to carbohydrate degradation depicted at genome level. BMC Genomics 12.
  27. 27. Couturier M, Navarro D, Olive C, Chevret D, Haon M, et al. (2012) Post-genomic analyses of fungal lignocellulosic biomass degradation reveal the unexpected potential of the plant pathogen Ustilago maydis. BMC Genomics 13: 57.
  28. 28. Bayry J, Aimanianda V, Guijarro JI, Sunde M, Latge JP (2012) Hydrophobins–unique fungal proteins. PLOS Pathog 8: e1002700.
  29. 29. Rudd JJ, Antoniw J, Marshall R, Motteram J, Fraaije B, et al. (2010) Identification and characterisation of Mycosphaerella graminicola secreted or surface-associated proteins with variable intragenic coding repeats. Fungal Genet Biol 47: 19–32.
  30. 30. Stergiopoulos I, van den Burg HA, Okmen B, Beenen HG, van Liere S, et al. (2010) Tomato Cf resistance proteins mediate recognition of cognate homologous effectors from fungi pathogenic on dicots and monocots. Proc Natl Acad Sci U S A 107: 7610–7615.
  31. 31. Tyler BM (2009) Entering and breaking: virulence effector proteins of oomycete plant pathogens. Cell Microbiol 11: 13–20.
  32. 32. Godfrey D, Bohlenius H, Pedersen C, Zhang Z, Emmersen J, et al. (2010) Powdery mildew fungal effector candidates share N-terminal Y/F/WxC-motif. BMC Genomics 11: 317.
  33. 33. Goodwin SB (2004) Minimum phylogenetic coverage: an additional criterion to guide the selection of microbial pathogens for initial genomic sequencing efforts. Phytopathology 94: 800–804.
  34. 34. Lum G, Min XJ (2011) FunSecKB: the Fungal Secretome KnowledgeBase. Database (Oxford) 2011: bar001.
  35. 35. Kämper J, Kahmann R, Bolker M, Ma LJ, Brefort T, et al. (2006) Insights from the genome of the biotrophic fungal plant pathogen Ustilago maydis. Nature 444: 97–101.
  36. 36. Gout L, Fudal I, Kuhn ML, Blaise F, Eckert M, et al. (2006) Lost in the middle of nowhere: the AvrLm1 avirulence gene of the Dothideomycete Leptosphaeria maculans. Mol Microbiol 60: 67–80.
  37. 37. Perfect SE, Hughes HB, O'Connell RJ, Green JR (1999) Colletotrichum: A model genus for studies on pathology and fungal-plant interactions. Fungal Genet Biol 27: 186–198.
  38. 38. Brown NA, Urban M, Van De Meene AML, Hammond-Kosack KE (2010) The infection biology of Fusarium graminearum: Defining the pathways of spikelet to spikelet colonisation in wheat ears. Fungal Biol 114: 555–571.
  39. 39. Neidleman S, Geigert J (1992) Biohalogenation: Principles, basic roles, and applications. Chichester: Ellis Horwood. 203 p.
  40. 40. Nimchuk Z, Eulgem T, Holt BF 3rd, Dangl JL (2003) Recognition and response in the plant immune system. Annu Rev Genet 37: 579–609.
  41. 41. Shetty NP, Mehrabi R, Lutken H, Haldrup A, Kema GHJ, et al. (2007) Role of hydrogen peroxide during the interaction between the hemibiotrophic fungal pathogen Septoria tritici and wheat. New Phytologist 174: 637–647.
  42. 42. Shetty NP, Kristensen BK, Newman MA, Moller K, Gregersen PL, et al. (2003) Association of hydrogen peroxide with restriction of Septoria tritici in resistant wheat. Physiol Mol Plant Pathol 62: 333–346.
  43. 43. Keon J, Rudd JJ, Antoniw J, Skinner W, Hargreaves J, et al. (2005) Metabolic and stress adaptation by Mycosphaerella graminicola during sporulation in its host revealed through microarray transcription profiling. Mol Plant Pathol 6: 527–540.
  44. 44. Birch PRJ, Boevink PC, Gilroy EM, Hein I, Pritchard L, et al. (2008) Oomycete RXLR effectors: delivery, functional redundancy and durable disease resistance. Curr Opinion Plant Biol 11: 373–379.
  45. 45. de Wit PJGM, Mehrabi R, van den Burg HA, Stergiopoulos I (2009) Fungal effector proteins: past, present and future. Mol Plant Pathol 10: 735–747.
  46. 46. Stukenbrock EH, Jorgensen FG, Zala M, Hansen TT, McDonald BA, et al. (2010) Whole-genome and chromosome evolution associated with host adaptation and speciation of the wheat pathogen Mycosphaerella graminicola. PLOS Genetics 6: e1001189.
  47. 47. Muller O, Schreier PH, Uhrig JF (2008) Identification and characterization of secreted and pathogenesis-related proteins in Ustilago maydis. Mol Genet Genomics 279: 27–39.
  48. 48. Emanuelsson O, Nielsen H, Brunak S, von Heijne G (2000) Predicting subcellular localization of proteins based on their N-terminal amino acid sequence. J Mol Biol 300: 1005–1016.
  49. 49. Emanuelsson O, Brunak S, von Heijne G, Nielsen H (2007) Locating proteins in the cell using TargetP, SignalP and related tools. Nat Protoc 2: 953–971.
  50. 50. Eisenhaber B, Schneider G, Wildpaner M, Eisenhaber F (2004) A sensitive predictor for potential GPI lipid modification sites in fungal protein sequences and its application to genome-wide studies for Aspergillus nidulans, Candida albicans, Neurospora crassa, Saccharomyces cerevisiae and Schizosaccharomyces pombe. J Mol Biol 337: 243–253.
  51. 51. Horton P, Park KJ, Obayashi T, Fujita N, Harada H, et al. (2007) WoLF PSORT: protein localization predictor. Nucleic Acids Res 35: W585–587.
  52. 52. Paper JM, Scott-Craig JS, Adhikari ND, Cuomo CA, Walton JD (2007) Comparative proteomics of extracellular proteins in vitro and in planta from the pathogenic fungus Fusarium graminearum. Proteomics 7: 3171–3183.
  53. 53. Heger A, Holm L (2000) Rapid automatic detection and alignment of repeats in protein sequences. Proteins 41: 224–237.
  54. 54. Antoniw J, Beacham AM, Baldwin TK, Urban M, Rudd JJ, et al. (2011) OmniMapFree: A unified tool to visualise and explore sequenced genomes. BMC Bioinformatics 12: 447.