A Proteomic Approach to Investigating Gene Cluster Expression and Secondary Metabolite Functionality in Aspergillus fumigatus

A combined proteomics and metabolomics approach was utilised to advance the identification and characterisation of secondary metabolites in Aspergillus fumigatus. Here, implementation of a shotgun proteomic strategy led to the identification of non-redundant mycelial proteins (n = 414) from A. fumigatus including proteins typically under-represented in 2-D proteome maps: proteins with multiple transmembrane regions, hydrophobic proteins and proteins with extremes of molecular mass and pI. Indirect identification of secondary metabolite cluster expression was also achieved, with proteins (n = 18) from LaeA-regulated clusters detected, including GliT encoded within the gliotoxin biosynthetic cluster. Biochemical analysis then revealed that gliotoxin significantly attenuates H2O2-induced oxidative stress in A. fumigatus (p>0.0001), confirming observations from proteomics data. A complementary 2-D/LC-MS/MS approach further elucidated significantly increased abundance (p<0.05) of proliferating cell nuclear antigen (PCNA), NADH-quinone oxidoreductase and the gliotoxin oxidoreductase GliT, along with significantly attenuated abundance (p<0.05) of a heat shock protein, an oxidative stress protein and an autolysis-associated chitinase, when gliotoxin and H2O2 were present, compared to H2O2 alone. Moreover, gliotoxin exposure significantly reduced the abundance of selected proteins (p<0.05) involved in de novo purine biosynthesis. Significantly elevated abundance (p<0.05) of a key enzyme, xanthine-guanine phosphoribosyl transferase Xpt1, utilised in purine salvage, was observed in the presence of H2O2 and gliotoxin. This work provides new insights into the A. fumigatus proteome and experimental strategies, plus mechanistic data pertaining to gliotoxin functionality in the organism.


Introduction
Following the publication of A. fumigatus Af293 [1] genomic sequence and the sequencing of a second A. fumigatus strain, A1163 [2], extensive efforts have been undertaken to characterise the proteome of this opportunistic human pathogen [3][4][5][6][7][8][9][10]. Traditional proteomic strategies have utilised 2-D separation with subsequent protein identification by MS. Shotgun MS-based proteomics has developed more recently and provides a complementary method to 2-D for proteome profiling [9,10], since 2-D can occasionally be limiting for the identification of particular subsets of proteins, especially hydrophobic proteins, membrane proteins, and proteins with large molecular mass or extreme pI [11].
MS-based or shotgun proteomics can adopt multiple approaches including, (i) direct LC-MS/MS, (ii) indirect LC-MS/MS and (iii) 2-D-LC-MS/MS (multidimensional protein identification technology, MudPIT) [12,13]. Direct LC-MS/MS involves the on-line separation of complex peptide mixtures using reversed phase nano-LC columns with extended acetonitrile gradients to effect peptide separation [14]. Indirect LC-MS/MS is where complex peptide or protein mixtures are pre-fractionated off-line (e.g. by SDS-PAGE) before LC-MS/MS analysis [15]. Subproteome strategies have also been implemented to investigate glutathione binding [4] and mitochondrial proteins [6]. Indeed, the recent emergence of MS-based proteomics studies of A. fumigatus has been undertaken whereby 530 plasma membrane associated proteins were identified by utilising a combination of SDS-PAGE fractionation of total protein followed by peptide separation and identification by 2-D-LC-MS/MS [16]. This study would have been difficult to perform using 2-D due to the incompatibility of hydrophobic proteins, and proteins with transmembrane (TM) regions, with detergents used in isoelectric focusing, the first separation stage of 2-D [17]. Quantitative MSbased proteomics, both label-free and using isobaric tagging for relative and absolute quantitation (iTRAQ), have been used to comparatively profile the stages of A. fumigatus germination [9,10]. Activity-based MS proteomics has also recently been developed to investigate A. fumigatus following incubation with human sera [18]. The application of MS-based proteomics to dissect the proteome of A. fumigatus has the potential to provide a global overview of the pathways and biological processes active under a set of conditions. In addition, (i) bioinformatic analysis can expand the characterisation of large datasets generated by MSbased proteomics, and (ii) shotgun proteomics offers the possibility of identifying the presence of either hypothetical proteins or proteins of unknown function, whose existence may either be unclear, or only previously demonstrated at the transcript level. Furthermore, (iii) shotgun MS-based proteomics has the potential to be used for the non-targeted identification of secondary metabolite (SM) cluster expression, which, coupled with subsequent metabolomics, could result in the identification of novel cluster products [19].
Proteomic approaches may also have an application in characterizing the effect of exogenous SMs on A. fumigatus [20]. Indeed, despite the many advantages of shotgun proteomics, 2-D has been successfully deployed to inform on proteomic alterations in A. fumigatus under various conditions [20][21][22][23][24][25][26][27]. Thus, a complementary strategy of shotgun and 2-D proteomics offers much in terms of the ability to reveal the nature of the proteome in pathogenic microorganisms, provide further insight into SM biosynthesis-and explore how apparently synergistic stressors may interact in unexpected ways. Interestingly, both gliotoxin and H 2 O 2 , separately, have been shown to result in numerous, growth inhibitory-associated, alterations to the proteome of A. fumigatus [20,23,28]. Indeed, exposure of mammalian cells to gliotoxin has been shown to increase the production of ROS, while H 2 O 2 induces oxidative stress [29]. Paradoxically however, it has been revealed [30] that gliotoxin actually relieved H 2 O 2 -induced growth inhibition of A. fumigatus in a concentration-dependent manner, although the basis for this phenomenon was not investigated.
The aim of the work presented here was to investigate the potential for shotgun MS to dissect the mycelial proteome, particularly with respect to identifying SM cluster expression, allied to cognate metabolite biosynthesis. Moreover, dissection of the molecular basis of SM (gliotoxin)-mediated relief of H 2 O 2induced oxidative stress in A. fumigatus was explored by 2-D and LC-MS/MS analysis. Overall, these combinatorial approaches reveal new insights into the expression, functionality and dynamic nature of the A. fumigatus proteome during normal growth and consequent to attenuated oxidative stress conditions.

Mycelial proteomics
For shotgun proteomics, mycelia from A. fumigatus ATCC26933 shaking cultures were harvested after 48 h, 200 rpm, 37uC in Aspergillus minimal media (AMM) and snapfrozen in liquid nitrogen. A. fumigatus ATCC26933 mycelia were also harvested from shaking cultures grown for 72 h in Czapek-Dox media, 37uC, 200 rpm. Protein was extracted and subjected to trypsin digestion as described [31]. Briefly, frozen mycelia (1 g) were ground in liquid nitrogen and resuspended in 6 ml of 25 mM Tris-HCl, 6 M Guanidine-HCl, 10 mM DTT pH 8.6. Extracts were sonicated five times at 10% power, cycle 6 for 10 sec intervals, followed by centrifugation at 10000 g for 10 min at 4uC. DTT (1 M; 10 ml per ml lysate) was added to the supernatants and incubated at 56uC for 30 min. Iodoacetamide (1 M; 55 ml per ml lysate) was added and incubated in the dark for 20 min. Whole cell lysates were dialysed twice against 100 mM ammonium bicarbonate. Aliquots of denatured protein solutions (100 ml) were digested with trypsin (5 ml; 0.4 mg/ml in 10% (v/v) acetonitrile, 10 mM ammonium bicarbonate), overnight at 37uC. Tryptic peptide mixtures were spin-filtered (Agilent Technologies, 0.22 mm cellulose acetate), separated on extended liquid chromatography gradients on a nanoflow Agilent 1200 LC system and subjected to tandem mass spectrometry using an Agilent 6340 Ion Trap LC-MS System (Agilent Technologies, Santa Clara, CA). Database searches for identification of proteins were carried out using Spectrum Mill MS Proteomics Workbench (Revision B.04.00.127). Validation criteria were set to (i) maximum of two missed cleavages by trypsin, (ii) fixed modification: carbamidomethylation of cysteines, (iii) variable modifications: oxidation of methionine, (iv) mass tolerance of precursor ions 62.5 Da and product ions 60.7 Da were employed and searches were carried out against a protein database of Aspergillus fumigatus strains Af293 (reference strain) and A1163, acquired from [32]. Protein grouping was carried out based on the presence of $1 shared peptide. Protein identifications were validated based on fixed thresholds (minimum protein score set to 20), with single peptide identifications requiring a Spectrum Mill score $17.0 and SPI . 70%. In order to determine the relative hydrophobicity of the identified proteins, the grand average of hydropathy (GRAVY) index was calculated using GRAVY calculator (www.gravycalculator.de). Using Phobius (http://phobius.cbr.su.se), the number of putative transmembrane regions present in each identified protein was determined. Identified proteins were grouped into functional categories based on the FunCat (Functional Catalogue), GO (Gene Ontology) and KEGG (Kyoto Encyclopedia of Genes and Genomes) annotations, using the FungiFun application (https://www.omnifung.hki-jena.de/FungiFun/) [33].

Detection of Secreted Secondary Metabolites in A. fumigatus
Culture supernatants were harvested from A. fumigatus ATCC26933 grown for 48 h in AMM or 72 h in Czapek-Dox, as described above. Organic extractions were carried out using a 1:1 mixture of chloroform to culture supernatant [34]. Chloroform extracts were dried to completion using a rotary evaporator and resuspended in methanol for LC-MS analysis using an Agilent 6340 Ion Trap LC-MS System (Agilent Technologies, Santa Clara, CA). Settings were adjusted to include the detection of singly charged molecules and molecules were separated on an acetonitrile gradient over a 15 min runtime.

2-D and LC-MS/MS
The mechanisms involved in gliotoxin-mediated relief of H 2 O 2induced stress in A. fumigatus ATCC26933 were investigated by comparative 2-D and LC-MS/MS. A. fumigatus ATCC26933 was grown in Sabouraud dextrose media at 200 rpm, 37uC for 24 h prior to addition of one of the following four treatments: (i) Solvent control (500 ml methanol added per 50 ml culture), (ii) Gliotoxin alone (gliotoxin, dissolved in methanol, added to a final concentration of 10 mg/ml), (iii) H 2 O 2 alone (H 2 O 2 added to a final concentration of 2 mM; 500 ml methanol added per 50 ml culture), (iv) Gliotoxin and H 2 O 2 combined (gliotoxin added to a final 10 mg/ml and H 2 O 2 added to final 2 mM). Mycelia (n = 5 biological replicates/treatment) were harvested after 4 h and ground in liquid nitrogen. Crushed mycelia were resuspended in 10% (w/v) TCA and sonicated three times at 10% power, cycle 6, 10 sec. Samples were incubated on ice for 20 min and centrifuged at 10000 g, 4uC for 10 min. Pellets were washed with ice-cold acetone to remove excess TCA and resuspended in IEF buffer [4]. Protein was separated on pH 4-7 IEF strips, followed by resolution by SDS-PAGE [35,36]. Colloidal Coomassie staining was carried out on gels, with subsequent protein spot analysis using Progenesis SameSpot software (Nonlinear Dynamics, Newcastle upon Tyne, United Kingdom). Gels (n = 20; 5 replicates/ treatment) from all four treatments were aligned and subsets of treatments were compared. Spots demonstrating significant changes in abundance (p,0.05, fold change $1.5) were excised from gels and trypsin digested according to the method described by Shevchenko (2007) [37]. Digested peptides were analysed by LC-MS/MS as described previously, with SpectrumMill used for database searching.

Fluorescent Detection of ROS in A. fumigatus
A. fumigatus ATCC26933 conidia (4610 6 per well, in 6 well plates) were added to Sabouraud liquid media (4 ml/well) and incubated for 24 h at 37uC, static. Each well contained a glass microscope slide. After removal of the top layer of mycelia, H 2 O 2 (2 mM final) only, or with gliotoxin (10 mg/ml final) was added. H 2 O 2 alone acts as a positive control as it is a known ROS inducing agent. Gliotoxin (10 mg/ml final) and an equivalent volume of MeOH were also added individually to separate wells as negative control. Plates were re-incubated 37uC, static, for 30 min, culture supernatants removed and mycelial mats washed once with 4 ml PBS for 5 min. After PBS removal, Sabouraud medium (3 ml) containing 29,79-dichlorodihydrofluorescein diacetate (2.5 mg/ml; H 2 DCFDA; Life Technologies) was added to each well followed by incubation 37uC for 40 min, static. After washing twice with PBS (2620 min each), mycelia were visualised using a fluorescent microscope (GFP filter: Ex/Em: 492-495/517-527 nm). Fluorescence was quantified by measuring Integrated Density Value (IDV) of selected areas from each image (n = 5/ treatment) representing subsequent data as means 6 standard error. Differences were evaluated using ANOVA, and statistical significance was accepted at p,0.05.

Results
Shotgun mass spectrometry identifies 414 proteins from A. fumigatus mycelia grown in AMM Utilising a direct shotgun proteomics approach, a total of 1826 unique peptides were identified in A. fumigatus mycelial lysates, corresponding to 414 distinct A. fumigatus proteins from 405 protein groups (Table S1). A false discovery rate (FDR) of 1.97% was determined for the distinct peptides identified in this study, using the validation criteria outlined (Table S1). These proteins (n = 414) spanned a theoretical pI range of 3.9 to 11.8 and a M r range of 7.8 to 444 kDa ( Figure 1; Table S1). All peptides identified contributed to a sequence coverage range of between 1 and 64% of the respective proteins, with Spectrum Mill scores ranging between 17 and 1327.
The GRAVY index for identified proteins ranged from 21.632 to 0.483, with positive scores indicating hydrophobicity ( Figure 1). A number of hydrophobic proteins were identified (n = 33; 7.97% of total identified proteins), based on positive GRAVY scores. Additionally, GRAVY scores were computed for the entire predicted proteome of A. fumigatus and it was observed that 15.6% of the total proteome possess positive GRAVY scores ( Figure 1). The majority of proteins identified by shotgun mass spectrometry (70.05%) were slightly hydrophilic, with GRAVY scores ranging from 20.5 to 0. This is in accordance with the total predicted proteome of A. fumigatus, where 55.1% of all predicted proteins fall within this range. Proteins with transmembrane helices (n = 44; 10.62% of total identified proteins) were detected. Several proteins were detected with 10 or more putative TM regions, including a plasma membrane H + -ATPase (AFUA_3G07640), an amino acid permease (Gap1) (AFUA_7G04290) and two ABC transporters (AFUA_1G14330 and AFUA_5G06070). One protein, a small oligopeptide transporter (OPT family) (AFUA_2G15240) was detected with 14 putative transmembrane regions and a GRAVY score of 0.276. Annotations were available for 89.37%, 86.47% and 49.27% of identified proteins using the FunCat, GO and KEGG schemes, respectively ( Figure 1). Based on the FunCat classification, functional categories that were significantly over-represented were protein synthesis (n = 86, p = 4.68610 223 ), energy (n = 91, p = 4.22610 217 ), protein with binding function or cofactor requirement (n = 254, p = 4.43610 214 ) and transcription (n = 32, p = 2.79610 29 ). Proteins (n = 23; 5.5%) were identified which have no functional classifications using the aforementioned methods.

Identification of A. fumigatus secondary metabolite cluster expression at protein level
Proteins identified by shotgun mass spectrometry were mapped based on their relative loci on each of the eight A. fumigatus chromosomes, using their gene locus identifiers ( Figure S1). A number of proteins (n = 15) that comprise a secondary metabolite supercluster, involved in the production of pseurotin A, fumitremorgins and fumagillin [38][39][40][41][42], were identified (Table 1). In addition, proteins were identified from the gliotoxin biosynthetic cluster on chromosome 6 [43], including GliT, the gliotoxin oxidoreductase responsible for self-protection against gliotoxin [28], and two clusters responsible for the production of unknown metabolites on chromosomes 3 and 4 respectively ( Table 1). A phosphoglycerate kinase PgkA protein (AFUA_1G10350) was also identified, which is predicted to be part of the Afpes1 NRPS cluster on chromosome 1 [1]. The identification of these proteins is indicative of the respective cluster activity under the growth conditions used. To confirm whether this detection of secondary metabolism-associated proteins correlated with the production of the respective molecules, LC-MS analysis was carried out on culture supernatants. This analysis revealed the presence of fumagillin, tryprostatin B and pseurotin A, along with an array of other, as yet unidentified, molecules ( Figure 1F). These secondary metabolites are all products of the 'supercluster' on Chromosone 8 [44][45][46], demonstrating correlation between the proteomic and metabolomic profiles. Interestingly, expression of the clusters identified here is partially or completely regulated by the transcription regulator LaeA [38]. Two additional proteins, encoded by AFUA_3G03280 and AFUA_3G03330, were also detected, from a cluster with predicted involvement in the production of a siderophore and a toxin [38] (Table 1). Following on from this observation of SM cluster protein detection, mycelia from cultures grown in Czapek-Dox media for 72 h were also analysed. Shotgun proteomic analysis revealed the presence of four proteins from the gliotoxin biosynthetic cluster under the conditions used. The identification of peptides from GliN, GliF, GliH and GliT (Table S2), correlated with the presence of gliotoxin in the culture supernatants of these cultures ( Figure 1G). Proteins (n = 13) were found to be significantly differentially abundant (p,0.05) when A. fumigatus was co-exposed to gliotoxin/H 2 O 2 compared to H 2 O 2 alone ( Figure 2; Table 2). These comprised six proteins with an abundance increase, and seven proteins with a decrease, of at least 1.5 fold. Furthermore, comparative analyses were carried out on all comparitor sets, to identify alterations in protein abundance between individual treatments (i.e., gliotoxin alone, H 2 O 2 alone, co-addition and solvent control). Redundancy was noted, with some proteins included in multiple comparison sets, resulting in the net differential expression of 27 unique proteins ( Figure 2). These proteins were excised and subjected to in-gel trypsin digestion, followed by LC-MS/MS analysis for protein identification.   (Tables S3 and S4,  Table 2). Protein abundance was assessed for all conditions relative to the solvent control (Tables S3 and S4). The abundance of proteins after gliotoxin/H 2 O 2 co-exposure was also assessed relative to the individual treatments of gliotoxin alone or H 2 O 2 alone (Table 2). Proteins (n = 13) were significantly altered in abundance following co-addition relative to H 2 O 2 alone (p,0.05) (Table 2, Figure 2). Proteins exhibiting increased abundance in the co-addition included those with oxidation-reduction activity. GliT, the gliotoxin oxidoreductase [28,47], was more abundant in the co-addition condition relative to H 2 O 2 alone (3.5 fold) but was unaltered by H 2 O 2 addition alone ( Figure 2, Table 2 and Table  S4). An increase in abundance of the Ran-specific GTPase and the proliferating cell nuclear antigen (PCNA), involved in cell-cycle regulation and DNA-repair [48,49], respectively, was also observed in the presence of gliotoxin/H 2 O 2 together. The HAD family hydrolase, also exhibited increased abundance when both gliotoxin and H 2 O 2 were present, relative to any of the control conditions. Proteins involved in amino acid and nucleic acid metabolism [50,51], glutamine amidotransferase: cyclase and methylenetetrahydrofolate reductase, also showed differential abundance. The class V chitinase, associated with cell autolysis [52], was significantly less abundant (p = 5.2610 25 ) upon coaddition relative to H 2 O 2 alone. A decrease in abundance of proteins associated with response to stress was observed following gliotoxin/H 2 O 2 co-exposure, relative to H 2 O 2 alone. Hsp90 and the oxidative stress protein Svf1 were of lower abundance in the co-addition (2.7 and 1.6 fold, respectively), reflective of the relief of H 2 O 2 -induced stress (Table 2, Figure 2). Additionally two unknown function proteins were detected, which underwent a 4.9 fold increase (AFUA_2G11120) and a 2.9 fold decrease (AFUA_6G03460) in abundance in the co-addition, relative to H 2 O 2 alone. This latter observation underpins the necessity to undertake both shotgun and 2-D based approaches to identify novel proteins.

Gliotoxin inhibits H 2 O 2 -induced ROS formation in A. fumigatus
Previous work has demonstrated that GliT is essential to protect A. fumigatus against exogenous gliotoxin [28] but contraintuitively, that gliotoxin reverses H 2 O 2 -mediated growth inhibition of A. fumigatus ATCC26933 and DgliK 26933 [30], yet, no mechanistic explanation was forthcoming. Results from the comparative proteomics analysis indicated that gliotoxin may effect an alleviation of H 2 O 2 induced-oxidative stress. Data in Figure 3 show ROS formation consequent to H 2 O 2 exposure in A. fumigatus. However, co-addition of gliotoxin results in a significant reduction (p.0.0001) in the production of reactive metabolites, as judged by decreased 29,79-dichlorofluorescein fluorescence. This suggests that gliotoxin acts as an anti-oxidant and functions to impede H 2 O 2 -mediated growth inhibition.

Discussion
A comprehensive mycelial proteome reference map, produced by Vödisch et al. [53], identified proteins with a GRAVY score up to 0.158 and fourteen proteins with 1-2 putative TM regions, which computes to 4.2% of the identified proteins possessing TM regions. Here, by comparison, 44 proteins possessing predicted TM domains were identified from A. fumigatus mycelia, corresponding to 10.62% of the proteins identified using shotgun mass spectrometry alone (Table S1). This represents a substantial increase (2.5 fold) in the identification of proteins with TM regions, compared to previous 2-D based studies with similar targets [6]. Using the shotgun proteomics approach, 33 hydrophobic proteins, corresponding to 7.97% of total identified proteins, were detected, including a protein transport protein SEC61 alpha subunit (AFUA_5G08130) with a GRAVY score of 0.4828, compared to 3.4% hydrophobic protein content by previous 2-D coupled MS analysis [53]. The standard molecular mass resolution of A. fumigatus mycelial proteins, using 2-D, ranges from 10 to 142 kDa [3,4,6]. The constraint of high molecular mass did not apply to the shotgun proteomic approach used in this study, with the identification of 12 proteins possessing a molecular mass greater than 142 kDa. The largest protein detected was encoded by AFUA_5G02570, with predicted histone acetyltransferase activity and a theoretical mass of 444 kDa. PesO [54], a hybrid polyketide synthase/ non-ribosomal peptide synthetase (PKS/ NRPS) (AFUA_8G00540), with a theoretical molecular mass of 434 kDa was also identified. Here, 73 unique peptides were identified from this protein, contributing to sequence coverage of 28%. PesO is involved in the production of pseurotin A [40] and its identification provides evidence of expression of this secondary metabolite cluster. A 267 kDa polyketide synthase (AFUA_8G00370) was also identified by 10 unique peptides, contributing to 10% sequence coverage. These findings represent some of the largest A. fumigatus proteins to be identified by mass spectrometry to date. Cagas et al. [9] utilised iTRAQ in order to profile the early development proteome of A. fumigatus. This gelfree method of large scale proteomic identification extended the Table 2. Proteins (n = 13) exhibiting significant differential abundance1 in A. fumigatus ATCC26933 following the co-addition of gliotoxin and H2O2, relative to H2O2 alone. molecular mass limits of detection to 9 to 255 kDa, thereby confirming the value of alternative methods for proteomic investigation. Supplementary information (Text S1) provides further discussion of shotgun MS data. Verification of proteins encoded by genes within secondary metabolite clusters was achieved in this study using shotgun mass spectrometry. Proteins identified from AMM cultures are putatively encoded by six clusters involved in the production of up to nine secondary metabolites (SM) [38]. Products of these SM clusters include up to 3 unknown metabolites, fumigaclavine C [55], fumitremorgins [39,44,45], pseurotin A [40], fumagillin [41,46], gliotoxin [43] and a putative siderophore [38]. Perrin et al. [38] annotated a 'supercluster' on Chromosome 8 (AFUA_8G00100-AFUA_8G00720) that is involved in the production of fumitremorgins, pseurotin A and fumagillin [39,40,46]. Fifteen proteins identified by shotgun mass spectrometry are annotated members of this 'supercluster', with one identified protein involved in the production of fumitremorgins and tryprostatins, four proteins involved in the pseurotin A biosynthetic portion of the cluster and the remaining nine proteins associated with fumagillin biosynthesis. Metabolomic investigation confirmed the presence of fumagillin, tryprostatin B and pseurotin A in culture supernatants, confirming the activity of these clusters, as indicated by shotgun proteomics. Phytanoyl-CoA dioxygenase family protein (FtmF) (AFUA_8G00230) was identified by 3 unique peptides and a sequence coverage of 14%. FtmF, a nonheme Fe(II) and a-ketoglutarate-dependent dioxygenase, catalyses the conversion of fumitremorgin B to verruculogen via endoper-oxide bond formation [56]. This enzyme is also capable of converting fumitremorgin B to 12a,13a-dihydroxyfumitremorgin C and 13-oxo-verruculogen, by deprenylation and oxidation mechanisms respectively [57]. Verruculogen, like fumitremorgin B, is a tremorgenic mycotoxin and has been shown to produce deleterious effects on respiratory epithelial cells [58]. A second protein (AFUA_8G00280) was also identified from the border region of this cluster, with putative oxidoreductase activity, and along with the detection of FtmF, suggests the production of fumitremorgins or tryprostatins by A. fumigatus under the conditions of culture. Pseurotin A production is also encoded by the 'supercluster' on Chromosome 8 [40], and four enzymes, that form part of the pseurotin biosynthetic cluster, were detected here; an alpha/beta hydrolase (AFUA_8G00530), a hybrid PKS-NRPS enzyme PesO (AFUA_8G00540), a methyltransferase SirN-like (AFUA_8G00550) and a putative glutathione S-transferase (AFUA_8G00580) [4]. This cluster has demonstrated increased expression at both the transcript and protein level under hypoxic conditions [21]. Furthermore, up-regulation of the methyltransferase and PesO transcripts were also shown in the mouse lung during infection by A. fumigatus [21]. Identification of nine proteins from the portion of the supercluster associated with fumagillin biosynthesis, represents significant coverage of this fifteen member cluster by shotgun mass spectrometry. Fumagillin is a meroterpenoid, with demonstrated anti-angiogenic activity through interaction with methionine aminopeptidase II (MetAP2) [59]. Fumagillin has also been associated with disruption of NADPH oxidase function could represent a putative virulence factor [60]. Indeed, cognate transcript expression of six of the proteins identified from this cluster was up-regulated in A. fumigatus Af293 during the initiation of murine infection [61]. Identification of a number of proteins from both the pseurotin A and fumagillin clusters is in-line with the recent identification of a transcription factor, FapR, which co-regulates expression of genes in these two clusters [46]. Further investigation of a second minimal media culture condition (Czapek-Dox, 72 h) revealed a similar observation. Proteins (n = 4) from the gliotoxin biosynthetic cluster were identified (Table S2), in line with the detection of gliotoxin in culture supernatants ( Figure 1G). Enlisting a shotgun proteomic approach provides a non-targeted method to detect the expression of proteins involved in secondary metabolism, in any given growth condition, and could prove useful as a tool for the identification of novel metabolites.
Proteomics also revealed changes in protein abundance associated with SM (gliotoxin)-mediated relief of H 2 O 2 -induced stress. Proteins (n = 13) were differentially abundant following exposure to a combination of H 2 O 2 (2 mM) and gliotoxin (10 mg/ ml), relative to H 2 O 2 alone (2 mM), which facilitates dissection of the mechanisms involved in gliotoxin-mediated relief of H 2 O 2induced stress ( Table 2). Increased abundance of two proteins, in response to gliotoxin and H 2 O 2 in combination, relative to H 2 O 2 alone, with predicted or demonstrated oxidoreductase activity included the gliotoxin oxidoreductase GliT and the NADHquinone oxidoreductase (23 kDa subunit), with 3.5 and 1.9 fold increase in abundance, respectively. In addition to a key role in the gliotoxin biosynthetic process, GliT also mediates self-protection against the harmful effects of gliotoxin [28,47]. Increased expression of GliT was detected following exposure to gliotoxin alone (5.1 fold), as previously noted [20,28] and combined with H 2 O 2 (4.8 fold) relative to the solvent control ( Figure 2). There was no significant alteration to abundance of GliT in response to H 2 O 2 alone (p = 0.297) and this demonstrates that GliT abundance is not regulated by H 2 O 2 and increased levels in the co-addition condition is solely a result of gliotoxin presence. Choi Purine-related enzymes and pathways undergoing an increase in expression, relative to the solvent control, are indicated in red and decreased expression is indicated in green. Metabolites are indicated in black. Ade1, bifunctional purine biosynthetic protein; Xpt1, xanthine-guanine phosphoribosyltransferase; Apt1, adenine phosphoribosyltransferase. Enzymes of the histidine and de novo purine biosythesis converging pathways, glutamine amidotransferase: cyclase and Ade1, are down-regulated in response to gliotoxin. Expression of enzymes involved in the purine salvage pathways, Xpt1 and Apt1, is up-regulated in the presence of H 2 O 2 and gliotoxin, repsectively, relative to a solvent control [20,23]. et al. [62] noted that gliotoxin catalysed H 2 O 2 reduction, mediated by the mammalian thioredoxin redox system and proposed that gliotoxin replaces 2-cys peroxiredoxin as an electron acceptor, in the reduction of H 2 O 2 to H 2 O in mammalian cells.
The proliferating cell nuclear antigen (PCNA) exhibited a 2.4 fold increase in abundance following incubation with a combination of gliotoxin and H 2 O 2 , relative to H 2 O 2 alone. Moreover, abundance of this protein was also further increased in the coaddition condition, relative to gliotoxin alone (7.4 fold) and the solvent control (5.9 fold; Table 2). Thus, PCNA abundance is induced by H 2 O 2 (approximately 3.5 fold) and not by exposure to gliotoxin alone. Clearly, a combination of H 2 O 2 with gliotoxin leads to further induction of this protein. PCNA is involved in the process of DNA-repair following H 2 O 2 -mediated damage [48] and acts as an anchor to the DNA template for binding partners [63]. The increase in PCNA abundance observed in response to H 2 O 2 , alone or coupled with gliotoxin, may therefore be indicative of H 2 O 2 -induced DNA damage in these conditions. Furthermore, additional induction of PCNA abundance in the co-addition condition relative to H 2 O 2 , alone, may account for the recovery of growth, due to enhanced DNA repair capacity.
Proteins involved in the response to cellular stress underwent decreased abundance in the co-addition condition, relative to H 2 O 2 alone. A decrease in abundance of Hsp90 (2.7 fold) and the oxidative stress protein Svf1 (1.6 fold) was noted. Hsp90 displayed increased abundance in the presence of H 2 O 2 alone, relative to the solvent control (2.0 fold), with this response reversing upon coincubation with gliotoxin. In accordance with these observations at the proteomic level, the transcript of Hsp90 was also reported to be up-regulated in A. fumigatus in response to exogenous H 2 O 2 [64]. Hsp90 is a stress-induced protein involved in the refolding of denatured proteins and signal transduction [64,65]. The decrease in abundance of Hsp90 is indicative of the decrease in oxidative stress, correlating with the relief of growth inhibition observed. The decreased abundance of Svf1, with a nuclear localisation and a role in the response to oxidative stress [7,66], is also diagnostic for the attenuation of oxidative stress in the co-addition condition relative to H 2 O 2 alone. Indeed, a significant reduction (p,0.0001) in ROS levels was detected following co-application of gliotoxin and H 2 O 2 , relative to H 2 O 2 alone (Figure 3), thus providing biochemical verification of the proteomics data. A decrease in abundance (1.8 fold) of the class V chitinase in the presence of a combination of gliotoxin and H 2 O 2 was observed, relative H 2 O 2 alone. This protein belongs to subgroup A of fungal/bacterial chitinases which are associated with fungal growth and autolysis [67,68]. The orthologous A. nidulans protein, ChiB, has demonstrated involvement in the autolysis of fungal mycelia in response to stress [52]. A higher abundance of this protein, in the presence of H 2 O 2 alone, may indicate the occurrence of mycelial autolysis, which could have been stimulated by the presence of oxidative stress. This autolysis could also account for the growthinhibited phenotype observed in the presence of H 2 O 2 alone [30].
Proteins involved in amino acid and nucleotide metabolism also decreased in abundance in the presence of gliotoxin and H 2 O 2 combined, relative to H 2 O 2 alone. Glutamine amidotransferase: cyclase and methylenetetrahydrofolate reductase underwent a 1.7 and 1.6 fold decrease in abundance, respectively, relative to H 2 O 2 alone. Additionally, abundance of both of these proteins was decreased 1.9 fold in the co-addition condition, relative to the solvent control. Considering these observations, H 2 O 2 does not appear to be involved in the controlling the levels of these proteins. Instead, gliotoxin, either independently or in combination with H 2 O 2 , is responsible for triggering the decrease in abundance of these proteins. Glutamine amidotransferase: cyclase catalyses two Table 3. Summary of abundance changes of proteins involved in purine biosynthesis pathways.

Protein Name
Gliotoxin steps in the biosynthesis of histidine, producing both a histidine precursor and 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR), an intermediate of the purine biosynthetic process, thus linking these pathways [50]. Interestingly, the bifunctional purine biosynthetic protein, Ade1, was reduced in abundance in the presence of gliotoxin relative to the solvent control (1.8 fold) ( Figure 2). Conversely, xanthine-guanine phosphoribosyl transferase Xpt1, was more abundant in the presence of gliotoxin and H 2 O 2 combined, relative to the solvent control (1.8 fold) and gliotoxin alone (1.6 fold), indicating that levels of this protein are influenced by H 2 O 2 . Indeed, Lessing et al. [23] observed an increase in abundance of Xpt1 following exposure to H 2 O 2 for 45 min. Xpt1 is involved in the purine nucleotide salvage pathway, whereby XMP and GMP are formed from precursors, xanthine and guanine, respectively [69] (Figure 4; Table 3). Additionally, the increased abundance of another component of the purine salvage pathway, adenine phosphoribosyltransferase, has been noted in response to exogenous gliotoxin [20]. These observations reveal a diminution of de novo purine biosynthesis in the presence of gliotoxin and that the alternative salvage pathway is utilised in its place ( Figure 4; Table 3). Together, these observations underline the influence of gliotoxin and H 2 O 2 , either alone or in combination, on nucleotide biosynthesis in A. fumigatus. Furthermore, while no definitive functions have been demonstrated for the unknown function proteins encoded by AFUA_6G03460 and AFUA_2G11120, computational analysis has assigned the function of D-alanine-D-alanine ligase and methyltransferase to these proteins, respectively [32]. An orthologue of this methyltransferase (MT-II) was found to be upregulated in A. niger in response to reductive stress from DTT [70], which may resemble the stress induced by gliotoxin and presents an interesting target for future investigations.

Conclusions
In summary, shotgun proteomics has revealed expression of multiple proteins involved in secondary metabolite biosynthesis coincident with detection of the cognate metabolites, and provides strong evidence for the activation of multiple clusters under the control of the transcriptional regulator LaeA, in the conditions tested. Our findings also demonstrate how proteomics can inform how the SM, gliotoxin, effects attenuation of H 2 O 2 -mediated oxidative stress. Figure S1 Distribution of proteins identified using shotgun mass spectrometry (n = 414) based on gene locus (blue lines). Identification of proteins (n = 15) from a supercluster on chromosome 8, involved in the production of fumitremorgin B, pseurotin A and fumagillin (red circle). (DOC)