Identification of a transcription factor that confers resistance to oxidative stress

Initially, this work aimed at uncovering previously uncharacterised TFs that are important for carbon catabolite repression (CCR), through screening an A. fumigatus transcription factor deletion library for growth defects in the presence of allyl alcohol (AA), a compound indicative for potential defects in CCR[13,14]. A strain that was deleted for gene Afu1g09190 (AFUB_008610) was highly sensitive to AA (Fig 1A and 1B) but presented no growth defects in non-stress conditions (S1A and S1B Fig) nor in the presence of 2-deoxyglucose (2DG), an additional indicator for defects in CCR[15] (S1C Fig). AA is converted by alcohol dehydrogenase (ADH) to acrolein, a highly toxic compound shown to also induce oxidative stress within the cell[16,17]. To investigate any ADH regulatory or activity defects, enzyme activity was measured in the presence of ethanol and ethanol and glucose, with glucose causing the CCR-mediated transcriptional repression of ADH-encoding genes[18]. ADH was not significantly different between the parental-type (referred to as wild-type, WT) CEA17 strain and the ΔAfu1g09190 strain in all conditions (S1D Fig), suggesting that the observed sensitivity of strain ΔAfu1g09190 was not due to regulatory defects in ADH activity and in CCR.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The transcription factor encoded by Afu1g09190 is important for resistance against oxidative stress.

a-d, Strains were grown for 5 days from 10⁵ spores at 37°C on glucose minimal media supplemented with increasing concentrations of the oxidative stress-inducing compounds (a) ally alcohol, (b) acrolein, (c) menadione and (d) t-butyl hydroperoxide (t-butyl). Graphs indicate the % of growth in the presence of the respective drug with respect to the control condition (without drug). Graphs are the quantitation of radial growth of the pictures shown in the same panel, with standard deviations representing three biological replicates (*P-value < 0.01; **P-value < 0.001; ***P-value < 0.0001 in a two-way ANOVA test).

https://doi.org/10.1371/journal.ppat.1008645.g001

We therefore decided to determine whether strain ΔAfu1g09190 was also sensitive to other oxidative stress-inducing compounds. Growth was characterised in the presence of acrolein, menadione and t-butyl hydroperoxide (t-butyl). Growth of strain ΔAfu1g09190 was significantly inhibited in the presence of all the aforementioned compounds, when compared to the WT and complemented strains (Fig 1C–1F). The aforementioned results therefore suggest that the A. fumigatus TF Afu1g09190 is important for mediating resistance against different types of oxidative stresses.

Gliotoxin (GT) biosynthesis-related genes are under the transcriptional control of Afu1g09190

To gain understanding of the cellular processes governed by Afu1g09190, RNA-sequencing was carried out for the wild-type (WT) and ΔAfu1g09190 strains, when grown for 24 h in GMM (glucose minimal medium, control) and after the addition of 10 mM AA for 30 min. The Afu1g09190 TF was confirmed to localise to the nucleus in the presence of AA (S1E Fig) by microscopy analysis of a GFP-tagged Afu1g09190 strain (S1F and S1G Fig).

In the first series of analyses, the transcriptional response to AA in the WT strain was determined. A total of 4,563 (2,365 up-regulated, 2,198 down-regulated, log2FC ≥ 1.0 and ≤ -1.0, P-value < 0.05) genes were significantly differentially regulated in the WT strain between the control and AA condition (S1 File). Functional categorization (FunCat) (https://elbe.hki-jena.de/fungifun/fungifun.php) analyses (P-value < 0.05) of all significantly up-regulated genes showed enrichment for cellular processes such as oxidative stress response, oxygen and radical detoxification, protein folding and stabilization, detoxification, stress response and glutathione conjugation reaction (Fig 2A, S1 File). Glutathione (GSH) is a cellular antioxidant that provides protection against intra- and extracellular oxidative stresses[19]. To determine whether the observed sensitivity of the ΔAfu1g09190 strain to AA was due to a defect in GSH metabolism, intracellular GSH levels were quantified in both the WT and mutant strains when exposed to AA for different amounts of time. No difference in intracellular GSH concentrations was observed between the WT, ΔAfu1g09190 and Afu1g09190 complemented strains in the presence of AA (S1H Fig). FunCat analysis (p-value < 0.05) of all down-regulated genes showed enrichment mainly for DNA metabolism (transcription, repair) and growth (budding, filamentation, cell cycle) (Fig 2A, S1 File). These results indicate that the addition of AA caused a fungal transcriptional response that aims at dealing with increased oxidative stress whilst down-regulating growth-related processes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The Afu1g09190-encoded TF regulates the transcription of gliotoxin (GT) biosynthetic genes and of gtmA through directly binding to the promoter regions in the presence of allyl alcohol (AA).

a, AA induces transcriptional up-regulation of genes required for the oxidative stress response whilst down-regulating genes encoding proteins involved in growth in the A. fumigatus wild-type (WT) strain. Functional categorisation (FunCat) description of the 5 most significant enriched categories (p-value < 0.0005), that contain up- (UP) or down (DOWN)-regulated genes with a -1 > Log2 fold-change > 1 (p-value < 0.05), as identified by RNA-sequencing. The WT strain was exposed to 10 mM AA for 30 min after 24 h growth in glucose minimal medium (GMM, control) and gene expression was compared between both conditions. b, Fumagillin and GT biosynthetic genes are severely repressed in the ΔAfu1g09190 strain. Heatmap of the Log2FC values, as determined by RNA-sequencing, of genes that were significantly (p-value < 0.05) differently (-1 > Log2FC < 1) expressed between the WT and ΔrglT in the presence of 10 mM AA for 30 min. Genes were classified into virulence-relevant categories, such as oxidative stress-related proteins, transcription factors, secondary metabolite biosynthesis and transporters. Log2FC gene expression values are also shown for the WT strain, when comparing gene expression in the presence of AA to the control condition. Hierarchical clustering was performed in MeV (http://mev.tm4.org/), using Pearson correlation with complete linkage clustering. c, Validation of the RNA-sequencing data by RT-qPCR. RNA was extracted from strains grown for 24 h in GMM (control) before 10 mM AA was added for 30 min and RT-qPCR was performed. Gene expression values were normalized by gene Afu1g05340, whose expression remained constant across all conditions, as confirmed by RNA-seq. Standard deviations represent three biological replicates (**P-value < 0.001; ***P-value < 0.0001 in a one-way ANOVA test). d, Chromatin immunoprecipitation coupled to DNA sequencing (ChIP-seq) of the WT and Afu1g09190::HA strains, when grown for 24 h in GMM (control) and then exposed to 10 mM AA for 30 min, identified 538 RglT::HA binding peaks that could be classified into three main groups: i) unique (150) or enriched binding (122) in the control condition, ii) consistent binding (160) in both the control and AA conditions and iii) unique (67) or enriched (39) binding in the AA condition. e, Afu1g09190 binds to the promoter regions of gliZ, gliM, gliA, gliF and gliT and was subsequently named RglT (regulator of GT). Heat map depicting ChIP-seq signal for all genes of the GT biosynthetic gene cluster in the control and AA conditions. Also shown is the differential binding analysis fold change (FC) between both conditions. f, RNA-seq results are in agreement with the ChIP-seq results, confirming that GT biosynthetic genes are under the direct transcriptional control of RglT. Heat maps depicting RNA-seq FPKM values for the WT and ΔrglT strains as well as Log2FC (fold change) values between the WT to the ΔrglT strain of all gli genes in the control and AA conditions.

https://doi.org/10.1371/journal.ppat.1008645.g002

Subsequently, gene expression comparison between the WT and the ΔAfu1g09190 strains was carried out in the control and AA conditions. In the control condition, a total of 338 genes (166 genes up-regulated, 172 genes down-regulated) were significantly differentially regulated, whereas exposure to AA resulted in 115 genes (17 genes up-regulated, 98 genes down-regulated) as significantly differentially regulated between both strains (S1 File). We then further focused on genes that were significantly differentially regulated in the mutant strain, when compared to the WT strain, in the presence of AA and manually classified them into functional categories that have been reported to be important for A. fumigatus virulence, including cell signalling components, oxidative stress-related proteins, secondary metabolite biosynthesis, transcription factors and transporters (S2 File and Fig 2B). Of particular interest was the significant down-regulation of several genes, including the genes required for self-protection from each SM, of the fumagillin and gliotoxin (GT) biosynthetic gene clusters in the ΔAfu1g09190 strain in the presence of AA (S2 File). Subsequently, we determined the MIC (minimal inhibitory concentration) of fumagillin and GT on the WT and ΔAfu1g09190 strains. There was no difference in the MIC of fumagillin for the WT and ΔAfu1g09190 strains (100 μg/ml), whereas the ΔAfu1g09190 strain (35 μg/ml) was significantly (p-value < 0.05 in a 1-way ANOVA test) more sensitive to GT when compared to the WT (70 μg/ml) strain (Table 1). We therefore decided to focus on characterising the role of Afu1g09190 in GT biosynthetic regulation. We identified 8 (gliP 3.5 log2-fold reduction; gliF -5.4 log2-fold reduction; gliT -5.5 log2-fold reduction, gliG -6.6 log2-fold reduction, gliH 6.7 log2-fold reduction, gliA 7.0 log2-fold reduction, gliN 7.2 log2-fold reduction and gliM 8.5 log2-fold reduction) out of the 13 GT-related biosynthesis genes that were severely reduced in expression in the ΔAfu1g09190 strain in the presence of AA (S2 File, Fig 2B). Furthermore, the expression of gtmA (Afu2g11120, 5.6-fold decrease), that codes for a bis-thiomethyltransferase involved in the down-regulation of GT biosynthesis through catalysing the formation of bisdethiobis(methylthio)gliotoxin (BmGT) from dithiogliotoxin[20], was also significantly reduced in the mutant strain (S2 File, Fig 2B). The RNA-seq data was validated by RT-qPCR of three randomly chosen genes (Fig 2C). These results suggest that the transcription factor encoded by Afu1g09190 is important for the expression of GT biosynthesis-related genes in the presence of AA.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Minimal inhibitory concentrations (MIC) of fumagillin and gliotoxin on the growth of the wild-type (WT) and ΔAfu1g09190 strains.

Standard deviations are shown for 3 independent replicates for the two strains in the presence of each mycotoxin.

https://doi.org/10.1371/journal.ppat.1008645.t001

Afu1g09190 binds to GT biosynthetic gene promoter regions

To discover whether GT biosynthetic gene promoter regions are directly contacted by Afu1g09190, genome-wide binding sites were determined by ChIP-seq (chromatin immunoprecipitation coupled to DNA sequencing) of the WT (non-tagged control) and the Afu1g09190:3xHA (S1F and S1G Fig) strains in the same conditions as described for the RNA-seq analysis. Distinct binding peaks were observed throughout the genome for the Afu1g09190:3xHA strain, but not for the untagged strain, in both the control and AA conditions (S2A Fig). A total of 538 sites were bound by Afu1g09190 genome-wide under the two conditions. Among these, 160 sites were constitutively bound by Afu1g09190::3xHA (i.e. common targets) with the remaining 272 and 106 binding events being unique to or enriched in the control and AA conditions, respectively (Fig 2D; S3 File). Systematic assignment of intragenic and intergenic binding sites to nearest genes revealed that the Afu1g09190 intragenic binding sites were primarily located at the 5’ UTR (untranslated) region of the genes (219 sites) as opposed to very few binding sites at 3’ UTR (35 sites) regions and within exons (18). On the other hand, 130 intergenic binding sites were within promoter regions [-500 bp to TSS (transcription start site)] and the remaining 139 intergenic sites were further upstream than 500 bp from the TSS. Fifteen Afu1g09190 binding sites were not associated with any gene as they were at convergent gene intergenic regions. Altogether, 481 unique genes were annotated with Afu1g09190 peaks.

Next, GO (gene ontology) analysis was performed on the target genes. Gene targets that were unique to or enriched in the control condition encode proteins with functions in pathogenesis, various metabolic pathways and asexual development (S2B Fig and S3 File). On the other hand, genes with unique or increased Afu1g09190:3xHA binding in the presence of AA are significantly enriched with functions in biosynthetic pathways for GT, mycotoxin, cysteine, isoleucine and beta-glucan; as well as in oxidative stress response and oxidant detoxification (S2B Fig). These results are consistent with the above findings that Afu1g09190 is important for oxidative stress resistance and regulation of genes required for GT biosynthesis.

Indeed, differential binding analysis (DiffBind) revealed that Afu1g09190 bound to the promoter regions of the transcription factor-encoding gene gliZ (2.14 log2-fold stronger binding in the AA condition when compared to the control condition), the O-methyltransferase-encoding gene gliM (3.11 log2-fold stronger binding in the AA condition when compared to the control condition), the efflux transporter-encoding gene gliA (3.62 log2-fold stronger binding in the AA condition when compared to the control condition), the bidirectional promoter region of gliT and gliF (1.87 log2-fold stronger binding in the presence of AA than when compared to the control condition) (Fig 2E, S3 File). Afu1g09190 also bound to the promoter region of gtmA (1.75 log2-fold stronger binding in the AA condition when compared to the control condition) in the presence of AA (S3 File). This increase in gene promoter binding correlated with a strong induction in gene expression in the WT strain in the presence of AA as shown by RNA-seq; but this induction is abolished in the ΔAfu1g09190 strain (Fig 2F). We therefore named this transcription factor RglT (Regulator of gliotoxin). These results suggest that RglT-mediated regulation of the expression of the GT biosynthetic genes gliP, gliG, gliH and gliN is indirect, whereas RglT modulates the expression of the GT biosynthetic genes gliF, gliT, gliM and gliA as well as of gtmA in the presence of AA through directly binding to the respective gene promoter regions.

RglT is crucial for GT biosynthesis

Due to the observed transcriptional down-regulation of several gli genes, we wondered whether the ΔrglT strain was able to produce GT, through performing HPLC on culture supernatants of the WT, ΔrglT and ΔrglT::rglT strains, when grown in GT-producing conditions. Strains were grown in Czapek-Dox broth for 72 h as this medium has previously been shown to result in detectable gliotoxin levels in culture supernatants[12]. The ΔgliT strain, which does not produce GT[12], was used as a negative control. GT eluted at a retention time of 8.45 min and chromatograms showed a peak corresponding to GT in all strains except for the ΔgliT strain (Fig 3A). In contrast, the UV-vis absorption spectra for GT in the supernatants of the ΔrglT and ΔgliT strains were different from the standard, WT and rglT complemented strains (Fig 3B), indicating that the ΔrglT strain did not produce GT but rather a metabolite that had a similar retention time as GT. To determine the nature of this metabolite, high-resolution LC-micrOTOF-MS analysis was carried out on the same samples and the presence of a molecular ion with m/z 357.0945 ([M+H]⁺) was detected in all samples, including the ΔgliT strain (S3A Fig). The accurate mass values of this metabolite correspond to a predicted molecular formula of C₁₅H₂₁N₂O₄S₂, indicating the biosynthesis of bisdethiobis(methylthio)-gliotoxin (BmGT) by ΔrglT[21]. BmGT was also detected in the WT, rglT complemented and ΔgliT strains (S3A Fig). LC-micrOTOF-MS detected GT in the supernatants of the WT and rglT complemented strains but not in the supernatants of the ΔrglT and ΔgliT strains (S3B Fig, Fig 3C and 3D). These results indicate that the transcription factor RglT is essential for GT biosynthesis in A. fumigatus.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. RglT is crucial for gliotoxin (GT) biosynthesis.

a, HPLC (high performance liquid chromatography) analysis for GT of supernatants of different strains, that were grown for 72 h in GT-inducing conditions, showed similar chromatographs for all strains except for the ΔgliT strain (negative control). Standard GT (positive control) eluted after 8.45 min at 254 nm. All experiments were performed in biological triplicates. b, UV-vis spectra of the ΔrglT and ΔgliT strains differ from the spectra of the wild-type (WT) and ΔrglT::rglT⁺ strains, indicating the absence of GT. c, LC-MS (Liquid Chromatography-Mass Spectrometry) of biological triplicates, confirms the absence of GT from the supernatant of the ΔrglT strain. Shown are the extended chromatograms of organic extracts from: (1, brown) standard gliotoxin; (2, orange) WT; (3, dark green) ΔrglT::rglT; (4, pale green) ΔrglT; (5, blue) ΔgliT. d, The ΔrglT strain does not produce GT. Diagram summarizing in which strains GT or BmGT [bisdethiobis(methylthio)gliotoxin] was detected. e, RglT is essential for gliZ, gliT, gliF and gtmA expression in GT-inducing conditions. RT-qPCR was performed after strains were grown for 72 h in GT-inducing conditions. Gene expression was normalized by the β-tubulin-encoding gene. Standard deviations represent three biological replicates (**P-value < 0.001, ***P-value < 0.0001 in a one-way ANOVA test). f, RglT binds to the promoter regions of gliZ, gliT and gliF in GT-inducing conditions. ChIP (chromatin-immunoprecipitation)-qPCR was performed for the WT and RglT::HA strains when grown for 48 h in GT-inducing conditions. Binding motifs were identified by ChIP-sequencing and binding was calculated using the % input method, normalizing the RglT::HA-specific binding by the total cellular DNA (input). Standard deviations represent three biological replicates (**P-value < 0.001, ***P-value < 0.0001 in a paired, one-tailed student t-test). g-i, Endogenous GT resistance is completely abolished in the ΔrglT strain. Strains were grown for 5 days at 37°C on glucose minimal medium supplemented with or without GT (g, h). Graphs (i) show the percentage of growth in the presence of GT when compared to the control, no GT condition and standard deviations represent biological triplicates (***P-value < 0.0001 in a two-way ANOVA test).

https://doi.org/10.1371/journal.ppat.1008645.g003

RglT directly regulates GT biosynthetic genes in GT-inducing conditions

To determine whether GT biosynthesis-related gene expression and RglT binding dynamics follow a similar pattern in GT-inducing conditions as observed for AA oxidative stress-inducing conditions, qRT-PCR and ChIP-qPCR analyses were carried out in these conditions. The expression of gliZ, gliT, gliF and gtmA, was severely reduced in the ΔrglT strain when grown in GT-inducing conditions (Fig 3E). We included the GT biosynthesis cluster-resident transcription factor GliZ[22], which is essential for GT biosynthesis, in our analysis in order to describe a potential transcriptional dependency between gliZ and rglT in GT-producing conditions, as binding of RglT to the gliZ promoter region was already observed in the presence of AA. Subsequently, we carried out ChIP-qPCR to determine whether RglT binds to the promoter regions of gliZ, gliT, gliF and gtmA in GT-inducing conditions. Using the ChIP-seq data (S4 File), we looked for motifs that were unique to or enriched in the AA condition and found 4 potential bindings sites in the gliZ (sense strand) promoter region, 6 potential binding sites in the gliT (sense strand) promoter region, whereas the gliF (antisense strand) and gtmA (sense strand) promoter regions contained 4 and 2 potential binding motifs, respectively (S2C Fig). ChIP-qPCR confirmed that RglT bound to the promoter region of gliZ, gliT and gliF (Fig 3F), suggesting that RglT regulates gliZ, gliT and gliF expression directly in conditions that result in GT production (Fig 3F).

MEME (Multiple EM for Motif Elicitation)-ChIP analysis was also carried out on the 500 bp region surrounding the peaks identified in our ChIP-seq data in order to determine conserved RglT DNA binding motifs that were specifically enriched in the presence of AA (S4 File and S2D Fig). Two related but different DNA motifs (5’-TCGG-3’ and 5’-CGGNCGG-3’) are specifically enriched among unique and stronger RglT:3xHA binding sites in the presence of AA, as opposed to a GC rich motifs among the peaks with similar binding under both conditions (S2D Fig). In addition, there is also a difference in dyad composition enriched between Control-enriched and AA-enriched RglT:3xHA binding classes (S2E and S2F Fig).

RglT is crucial for protection against exogenous GT

GT MIC assays showed that the ΔrglT strain was sensitive to exogenously added GT, probably due to gliT expression being severely reduced in this strain in the presence GT-producing conditions. gliT encodes a GT oxidoreductase that is crucial for GT disulphide bridge formation and cleavage as well as for protection against exogenous GT[12]. To further explore the role of RglT and GliT in mediating GT self-protection, the respective deletion strains were grown on solid minimal medium supplemented with GT. The ΔrglT strain lost all ability to protect itself against exogenously added GT (Fig 3G and 3I). The ΔgliT strain was also sensitive to GT but not to the same levels as the ΔrglT strain (Fig 3H and 3I). These results suggest that RglT is essential for mediating protection against exogenous GT, probably through an unidentified mechanism that likely involves regulation of gliT and gtmA expression.

The ΔrglT strain is hypovirulent in a chemotherapeutic murine model of IPA

Due to the importance of GT for fungal survival, colonization and immunomodulation of the mammalian host[10,11], the virulence of the ΔrglT strain was assessed both in vitro and in vivo. Macrophages contribute to innate immunity, fungal clearance and the generation of a pro-inflammatory response during A. fumigatus infection[23]. The capacity of bone marrow-derived macrophages (BMDMs) to phagocytize and kill wild-type (WT), ΔrglT and ΔrglT::rglT conidia was determined and the ΔrglT strain was found to be significantly more susceptible (P-value < 0.05) to macrophage-mediated phagocytosis (Fig 4A) and killing (Fig 4B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RglT is essential for A. fumigatus virulence.

a, b Bone marrow-derived murine C57BL/6 (wild-type) and CGD (chronic granulomatous disease) macrophages phagocytize (a) and kill (b) a significant higher number of ΔrglT conidia in vitro than when compared to the wild-type (WT) and ΔrglT::rglT strains. Standard deviations represent three biological replicates (*P-value < 0.05 in a two-tailed, unpaired student t-test). c, Survival curve of chemotherapeutic BALB/c mice, which were immunosuppressed with a combination of cyclophosphamide and hydrocortisone, infected intra-nasally with the WT, ΔrglT and ΔrglT::rglT strains show that the ΔrglT strain is hypovirulent (***p-value < 0.0001 in the Mantel-Cox and Gehan-Brestow-Wilcoxon tests, comparing the ΔrglT to the WT and ΔrglT::rglT strains). d, Fungal burden of the naïve control (CTRL = phosphate buffered saline), WT, ΔrglT and ΔrglT::rglT strains show a significant (***p-value < 0.0001 in a two-tailed, unpaired student t-test) reduction of the ΔrglT growth in chemotherapeutic murine lung tissue at 3 days post-infection. e, Histopathology of murine lungs at three days post-infection with the WT, ΔrglT and ΔrglT::rglT strains indicates the absence of ΔrglT proliferation in the lung tissue. 5 μm lung tissue sections were stained with haematoxylin and eosin (HE) or with Grocott’s Methenamine Silver (GMS) before slides were viewed and pictures taken. f, Tumor necrosis factor alpha (TNF-α) and g, interleukin (IL)-12p40 concentrations in macrophages infected with heat killed (HK) conidia or live-resting (LR) conidia.

https://doi.org/10.1371/journal.ppat.1008645.g004

Next, virulence of all three strains was assessed in a clinically-relevant chemotherapeutic BALB/c murine model of IPA. The WT and ΔrglT::rglT strains were both equally virulent (p-value > 0.05) and killed all mice after 11 and 8 days post-infection (p.i.), respectively; whereas the ΔrglT strain was hypovirulent [p-value > 0.05 when compared to the phosphate buffer saline (PBS) negative control] with 80% of mice infected with this strain being alive after 15 days p.i. (Fig 4C). Fungal burden in murine lungs showed a significant reduction of ΔrglT strain presence/growth in comparison with the WT and ΔrglT::rglT⁺ strains that was not significantly different from the negative PBS control (Fig 4D), suggesting that the ΔrglT strain was cleared from the lung tissue. In agreement, histopathology of murine lungs of animals infected with PBS or the ΔrglT strain showed the absence of inflammation and fungal hyphae (Fig 4E). In contrast, murine lungs of mice infected with the WT and the ΔrglT::rglT⁺ strains showed multiple foci of invasive hyphal growth, which penetrated the pulmonary epithelium in the major airways, while pockets of branched invading hypha originated from the alveoli (Fig 4E). These results suggest that the absence of GT in the ΔrglT strain is not the major cause for the observed reduction in virulence, since GT-deficient strains, such as the ΔgliP strain, which is deleted for the GT nonribosomal peptide synthase, is virulent in a chemotherapeutic murine model of IPA[24].

To exclude changes in cell wall and growth defects of the ΔrglT strain as a possible cause for the strain-specific avirulence, macrophages were inoculated with live resting (LR) and swollen, heat-killed (HK) conidia and assayed for cytokine production. No difference in the production of TNF-α and IL-12p40 was observed between the WT, deletion and complemented strains in the presence of HK conidia; whereas LR conidia of the ΔrglT strain elicited a significantly greater production of both cytokines than when compared to the WT and ΔrglT complemented strains (Fig 4F and 4G). These results suggest that the increased macrophage pro-inflammatory response to the ΔrglT strain is not due to changes in strain-specific cell wall structure. In addition, growth of the ΔrglT strain on agar plates supplemented with ground porcine lung fragments and whole murine lung explants did not significantly differ from growth of the WT and ΔrglT::rglT⁺ strains (S3C and S3D Fig), suggesting that the ΔrglT strain grows normally on mammalian lung tissue in vitro.

The RNA-seq data showed that RglT regulates many genes involved in oxidation-reduction processes (S2 File), which are important for resistance against oxidative stress that is exerted by the mammalian immune system in order to fight fungal infections [25]. To determine whether oxidative stress resistance genes are under the transcriptional control of RglT, we assessed the expression of the TF yap1 (Afu6g09930, major factor in the control of the fungal response to oxidative stress [26]), the mycelial catalase cat1 (Afu3g02270), the superoxide dismutases sod1 (Afu5g0924) and sod2 (Afu4g11580), the cytosolic peroxiredoxin prx1 (Afu4g08580), the mitochondrial peroxiredoxin prxB (Afu5g15070) and the thioredoxin reductase trr1 (Afu4g12990) by qRT-PCR in the WT and ΔrglT strains when grown in the presence of 10 and 30 mM AA for 1 h (short incubation) and 3 h (prolonged incubation) (S4A Fig). In these conditions, RglT positively regulates the expression of yap1, prxB and trr1; and negatively regulates the expression of cat1 and sod1, whereas no induction occurred for sod2 and prx1 (S4A Fig). All these genes were not significantly differently expressed after 30 min in the presence of 10 mM AA between the WT and ΔrglT strains (S4B Fig), suggesting that prolonged incubation times and higher concentrations of oxidative stress-inducing compounds result in a different transcriptional profile. It remains to be determined whether this regulation is direct or indirect, as the ChIP-seq data showed direct regulation of only trr1 after 30 min incubation with 10 mM AA (S4B Fig). In addition, we grew the WT, ΔgliP and ΔgliP::gliP strains in the presence of oxidative stress-inducing compounds in order to further determine whether the virulence defect of the ΔrglT strain is due to defects in oxidative stress resistance. A relationship between gliotoxin and oxidative stress has previously been noted [11,27,28], although the mechanism has not been described to date. Growth of the ΔgliP strain was not significantly different from the WT and ΔgliP complemented strains in these conditions (S5 Fig).

To definitely determine whether the ΔrglT strain has defects in oxidative stress resistance, phagocytosis and killing by macrophages derived from a murine model of CGD (chronic granulomatous disease) with NADPH deficiency was performed. CGD macrophages are still able to phagocytize (Fig 4A) and kill (Fig 4B) a significantly higher number of ΔrglT conidia, suggesting that the virulence defect of the ΔrglT strain is not due to defects in resistance to oxidative stress exerted by macrophages. Together, these results suggest that the inability of the ΔrglT strain to infect chemotherapeutic mice are due to unidentified mechanisms that are essential for in vivo survival and that have not been described in vitro as of date.

The homologue of RglT has a conserved function for oxidative stress resistance and protection against exogenous GT in the non-GT-producing fungus A. nidulans

Secondary metabolite production is predicted to be essential for fungal survival in the highly competitive soil habitat of filamentous fungi[1], and GT is secreted by only a few Aspergillus species[29]. We therefore asked whether non-GT-producing Aspergillus species, such as A. nidulans, were also able to protect themselves against exogenously added GT and if so, whether this resistance against GT is mediated by RglT. Examination of the taxonomic distribution of RglT, GliT, and GT biosynthetic gene cluster (BGC) homologs across Eurotiomycetes and Sordariomycetes (Fig 5A; taxon names have been removed for clarity, but the full phylogeny is shown in S3E Fig) inferred their presence in 327, 416, and 111 taxa, respectively (S5 File). Patterns of presence or absence of the two genes harbored strong phylogenetic signal (p < 0.001; S5 File). Furthermore, both RglT and GliT were more commonly found in Eurotiomycetes than in Sordariomycetes. Specifically, we identified 175 / 216 (81.02%) taxa with both RglT and GliT across Eurotiomycetes, 61 of which also had a GT BGC homolog. Among Sordariomycetes, we identified 124 / 242 (51.24%) taxa with both RglT and GliT, 23 of which also had a homologous GT BGC. Lastly, phylogenetically informed model testing of the distributions of RglT and GliT showed that the pattern of occurrence of RglT is statistically dependent on the distribution of GliT (p = 0.002; S5 File) but not vice versa. This finding is consistent with an evolutionary scenario in which the GliT-based resistance mechanism is ancestral and the evolutionary recruitment of RglT regulation of GliT occurred subsequently.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. The homologue of A. fumigatus RglT is important for gliotoxin (GT) self-protection, through regulating the expression of the gliT homologue in the non-GT-producing fungus A. nidulans.

a, The phylogenetic distribution of RglT, GliT and GT biosynthetic gene cluster homologs across 458 fungal genomes. A 3-gene phylogeny of genomes from Eurotiomycetes (shown by the red branches) and Sordariomycetes (shown by the blue branches). For every tip in the phylogeny, the presence of RglT, GliT and gliotoxin BGC homologs is depicted using orange, purple, and green bars, respectively; absences are depicted in white. The grey bar plots depict how many of the 13 Gli genes are present in the gliotoxin BGC homolog. The tip corresponding to the A. fumigatus Af293 genome is indicated by a red dot, and the tip corresponding to the A. nidulans A4 genome is indicated by a maroon dot. b, Clustal W alignment between AfRglT and AnRglT (An1368) shows 54% identity at the protein level (e-value 8e-137). c-f, AnRglT (AN1368) is important for resistance against oxidative stress and GT self-protection. Strains were grown for 5 days from 10⁵ spores at 37°C on glucose minimal media supplemented with increasing concentrations of (c) gliotoxin, (d) allyl alcohol, (e) t-butyl hydro-peroxide (t-butyl) and (f) menadione. Legend for all graphs is displayed. Graphs indicate the % of growth in the presence of the respective drug with respect to the control condition (without drug). Standard deviations represent three biological replicates (*P-value < 0.01; **P-value < 0.001; ***P-value < 0.0001 in a two-way ANOVA test). g-i, The expression of (g) AngliT (AN3969) and (h) AfgliT, as determined by qRT-PCR, is transcriptionally dependent on RglT when GT is added exogenously for 3 h at a final concentration of 5 μg/ml. As a control, (i) AfgliT gene expression was confirmed to be independent of GliZ. Gene expression was normalised by tubulin gene expression in both fungi. Standard deviations represent three biological replicates (***P-value < 0.0001 in a two-way ANOVA test).

https://doi.org/10.1371/journal.ppat.1008645.g005

According to FungiDB (https://fungidb.org/fungidb/), the orthologue of A. fumigatus RglT is encoded by gene AN1368 in A. nidulans, which has 54% identity (e-value 8e-137) at the protein level with RglT from A. fumigatus (Fig 5B). Similarly, FungiDB predicts A. nidulans gene AN3963 to encode the putative orthologue of A. fumigatus gliT. AN3963 has 50% identity (e-value 3e-103) at the protein level with GliT of A. fumigatus and is predicted to code for a putative GT reductase[12]. Indeed, A. nidulans grew in the presence of exogenously added GT (Fig 5C), whereas deletion of AN1368 resulted in a strain with abolished growth in the presence of GT (Fig 5C). In addition, the ΔAn1368 strain also presented reduced or abolished growth in the presence of the oxidative stress-inducing compounds AA (Fig 5D), t-butyl (Fig 5E), but not menadione (Fig 5F). These results suggest a conserved function of RglT for protection against oxidative stress in both A. fumigatus and A. nidulans.

To determine whether the A. nidulans GliT homologue-encoding gene is also transcriptionally dependent on AnRglT, qRT-PCR was performed when GT was added exogenously to the cultures. The expression of gliT is dependent on RglT in both A. fumigatus (Fig 5G) and A. nidulans (Fig 5H) when GT was added exogenously to the medium. In agreement with Schrettl et al.[12], transcriptional expression of A. fumigatus gliT is independent of GliZ in the presence of gliotoxin (Fig 5I). These results confirm that RglT regulates the expression of GT self-protection and resistance genes when GT is added exogenously to cultures of both A. fumigatus and A. nidulans, and suggest a conserved function of RglT for resistance against exogenous GT, even in GT non-producing Aspergillus species.

Ethics statement

Animal experiments were performed in strict accordance with the Brazilian Federal Law 11,794 establishing procedures for the scientific use of animals in strict accordance with the principles outlined by the Brazilian College of Animal Experimentation (CONCEA), and the State Law establishing the Animal Protection Code of the State of São Paulo. All protocols adopted in this study were approved by the local ethics committee for animal experiments from the Federal University of São Paulo (Permit Number: 1468200318; Studies on the interaction of Aspergillus fumigatus with animals). All efforts were made to minimize suffering. Animals were clinically monitored at least twice daily and humanely sacrificed if moribund (defined by lethargy, dyspnea, hypothermia and weight loss). All stressed animals were sacrificed by cervical dislocation after intraperitoneal (i.p.) injection of ketamine and xylazine.

Strains and media

All strains used in this study are listed in S1 Table. Strains were grown at 37°C, except for growth on allyl alcohol-containing solid medium, which was carried out at 30°C. Complete and glucose minimal media (GMM) were prepared exactly as described previously[14]. Gliotoxin production was induced by growing the strains in Czapek-Dox (http://himedialabs.com/TD/M076.pdf) broth for 72 h. Porcine lungs were cut into small pieces and freeze-dried before they were ground under liquid nitrogen and 1 g was dissolved in a 100 ml solution of 1% w/v agarose and autoclaved. Lungs were extracted from healthy C57BL/6 mice and one lung was transferred to a 1% w/v agarose plate. All experiments were carried out in biological triplicates.

Alcohol dehydrogenase (ADH) activity

ADH activity in total cellular protein extracts of biological triplicates of strains grown in the defined conditions was measured exactly as described previously[39].

Microscopy

Microscopic analyses of GFP fluorescence was carried out as described previously[14], with the following modifications: strains were grown for 16 h at 30°C before allyl alcohol was added to a final concentration of 10 mM for 15 min. Experiments were carried out in biological triplicates and nuclei of 50 hyphae were observed for GFP fluorescence for each replicate. Afu1g09190::GFP nuclear localization was calculated as a percentage in reference to the total amount of nuclei.

Intracellular glutathione concentrations

Mycelia were ground to a fine powder in liquid nitrogen and 200 mg was re-suspended in 250 μl assay buffer (5% sulfosalicylic acid dissolved in 0.125M sodium phosphate monobasic and 6.3 mM EDTA, pH 7.5) and centrifuged (13,000 rpm, 10 min, 4°C). Total GSH concentrations were measured as described previously[40]. Briefly, GSH concentrations were determined with DTNB (Ellman’s reagent) in the presence of glutathione reductase and NADPH. Sample glutathione concentrations were calculated from a GSH standard curve. In parallel, total cellular protein was extracted for each sample, by re-suspending 200 mg powder in 1 ml extraction buffer (50 mM Tris-HCl pH 7.0, 50 mM NaF, 1 mM DTT). Samples were centrifuged for 10 min at 13,000 rpm at 4°C before protein concentrations were determined by Bradford assay (Bio-Rad), according to manufacturer’s instructions. GSH concentrations were normalised by total intracellular protein concentrations and expressed as nmol/mg intracellular protein.

RNA extractions, cDNA synthesis and RNA-sequencing

All experiments were carried out in biological triplicates and strains were grown from 10⁶ conidia/ml in glucose minimal medium (GMM). For the allyl alcohol (AA), strains were grown for 24 h in GMM before allyl alcohol was added to a final concentration of 10 mM for 30 min. Gliotoxin (GT) production was induced by growing strains for 72 h in Czapek-Dox medium. Alternatively, GT was added for 3 h to the culture medium to a final concentration of 5 μg/ml after strains were grown for 21 h in GMM. As GT was dissolved in DMSO, control cultures received the same volume of DMSO for 3 h. RNA was extracted as described previously[14] and cDNA was synthesized from 500 ng RNA using the ImProm-II Reverse Transcriptase (Promega), according to manufacturer’s instructions. qRT-PCR was carried out on 1 μl template cDNA as described previously[14], with the exception that the total reaction volume was 10 μl (5.5 μl SYBR Green, 0.055 μl of a 100 pmol/μl primer and 3.39 μl ddH₂O) using the 7500 Fast Real-Time PCR Thermocycler and the 7500 Fast System v1.4.0 (AB Applied Biosystems).

For RNA-sequencing, RNA quality was verified using the Agilent Bioanalyser 2100 (Agilent technologies) using a minimum RIN (RNA Integrity Number) value of 7.0. RNA-sequencing was carried out using Illumina sequencing, as described previously[41]. The TruSeqStranded mRNA LT Set B kit (Illumina) was used for library preparation. Libraries were sequenced (2x100bp) on CTBE NGS sequencing facility HiSeq 2500 instrument, generating approx. 13.1x10⁶ fragments per sample. Short reads were submitted to the NCBI’s Sequence Read Archive under accession number SRP154617. Data processing (quality check, cleaning, removal of rRNA, genome mapping) was carried out as described previously[41], with the following modifications. Raw read counts for A. fumigatus genes were imported into DESeq2 using Bioconductor package tximport (version 1.12.3). DESeq2 (version 1.24.0) was used for differential gene expression quantification. Default Benjamini & Hochberg method was used for multiple hypothesis correction of DESeq2 differentially expressed genes.

Determination of minimal inhibitory concentrations (MIC)

Strains were grown in 96-well plates at a concentration of 10⁴ spores/well in 200 μl glucose minimal medium supplemented with increasing concentrations of fumagillin or gliotoxin, according to the protocol elaborated by the Clinical and Laboratory Standards Institute (CLSI, 2017). Three independent experiments were carried out for each mycotoxin.

Chromatin immunoprecipitation coupled to DNA sequencing (ChIP-seq)

Strains were grown in the indicated conditions and cross-linking was performed as described previously[42]. Press-dried mycelia was frozen in liquid nitrogen and stored at -80˚C until chromatin preparation, which was performed as described previously[43]. For immunoprecipitation, 2 ug of anti-HA antibody (F7, Santa Cruz) was used. Library preparation was carried out according to[44] with the exception of the first end-repair step, which was replaced by the NEBNext End-Repair module (Cat. no. E6050). Libraries were checked and concentration determined using DNA High Sensitivity Bioanalyzer assay, mixed in equal molar ratio as described[44] and subjected to sequencing using Illumina HiSeq2500.

ChIP-seq analysis

Raw sequencing reads were mapped to the A. fumigatus strain Af293 reference genome (AspGD version s03-m05-r06) using Bowtie2[45] (v2.3.5). For ChIP-seq signal data visualization purpose, aligned reads were extended to 200bp using ‘macs2 pileup’ command and scaled to 1 million mapped reads using ‘macs2 bdgopt’ command from MACS2[46] (v2.1.1) tool. UCSC Kent utils’ programs[47] ‘bedSort’ and ‘bedGraphToBigWig’ were used to generate BigWig files, which were visualized in Integrative Genomics Viewer[48] (v2.5.3). Peaks were called using MACS2 with parameters ‘—nomodel—extsize 200’. MACS2 -log10(p-value) cutoff of 25 was used to select significant RglT peaks in the both AA and control ChIP-seq datasets.

Differential binding analysis was performed using DiffBind[49,50] R package (v2.12.0 with DESeq2 v 1.24.0). DiffBind combines the input peak lists into a consensus region set and performs differential binding affinity analysis over these consensus regions. Consensus regions which had peak in only one of the replicates in Control and AA condition were removed. Additionally, if a consensus region had no peak in any of the samples with -log10(p-value) > = 25, it was removed. After this quality filtering, total 538 genome-wide RglT binding sites were retained in Control and AA condition. Regions with RglT binding signal difference larger than 2-fold and FDR < = 0.05 were considered as significantly differentially bound regions between AA vs control comparison of RglT binding. Based on the DiffBind and peak occupancy analysis, RglT peaks were grouped into 5 categories such as AA specific (peak unique under AA condition), AA enriched (RglT peaks with stronger signal in AA condition compared to control), common (both conditions has RglT signal with no significant difference), control enriched (stronger signal in control compared to AA) and control specific (unique to control condition). RglT binding sites were assigned to nearest genes in A. fumigatus reference annotation using inhouse script.

Motif enrichment was performed on three sets of sequences, common peaks, AA enriched (AA specific + AA enriched) peaks and control enriched (control specific + control enriched) peaks. 500bp sequence around peak summit position was extracted for all significant RglT peaks and analyzed using MEME-ChIP[51] for motif enrichment. Selected motifs of interest were mapped back to the genome using FIMO[52] program from MEME-suite. Heatmap figures were made using the online program FungiExpresZ (https://cparsania.shinyapps.io/FungiExpresZ/). The ChIPseq data are available from NCBI SRA (sequence read archive) database under accession number PRJNA574873.

ChIP-qPCR

ChIP-qPCR was carried out as described previously[14], with the exception that HA-tagged protein immunoprecipitation of 1 mg total cellular proteins was carried out as described in[53]. DNA was purified using the Qiaquick gel extraction kit (Qiagen), according to manufacturer’s instructions. qPCR was run as described above.

High-performance liquid chromatography (HPLC) and liquid chromatography-mass spectrometry (LC-MS)

All solvents used during this study were of analytical grade (Synth) or HPLC grade (Merck or J.T. Baker). The analyses of extracts were performed in a high-performance liquid chromatography system (HPLC-Shimadzu, Kyoto, Japan), which consisted of a LC-6AD solvent pump, a SCL 10AVB system controller, a CTO-10ASVP column oven, a Rheodyne model 7725 injector, a SPD-M10AVP diode array detector (DAD), and Class VP software for data acquisition. The LC-MS analysis was performed using a UPLC (Shimadzu) coupled to Sigma-Aldrich-ascentis (2.7μm, 100mm x 4.6mm) column attached to a guard column (Agilent, 4.6 mm, 12.5 mm, 5 μm) and with the micrOTOF II mass spectrometer (Bruker Daltonics), using a gradient solvent system with acetonitrile:water with 0.1% v/v acetic acid.

All experiments were performed in biological triplicates. Extraction of 20 ml culture supernatants was performed with 3 x 50 ml chloroform and organic extracts were washed with saturated NaCl, dried with anhydrous Na₂SO₄ and filtered. Samples were concentrated on a rotary evaporator (Rotavapor R-300, Buchi) under reduced pressure to yield the corresponding extracts (S1 Table). Extracts (samples) were dissolved in 350 μL MeOH, while the standard gliotoxin was prepared by dissolving 200 μg gliotoxin in 350 μL MeOH and doing a 1:6 dilution (50 μL in 300 μL MeOH). Each sample was analyzed using reverse-phase HPLC (C-18 ascentis column), attached to a guard column. The gradient analysis was performed using acetonitrile:water with 0.1% v/v acetic acid. The analysis time was 35 minutes, starting from 10% to 100% acetonitrile in 30 min; then, 100% 30–32.5 min, returning to 10% 32.5–34 min, finishing in 10% 35 min; using a flow of 1ml/min. Chromatograms were analyzed at 254 nm. For LC-MS analysis, each sample was analyzed using reverse-phase HPLC (C-18 ascentis column, 2.7μm, 10cm x 4.6mm, Sigma-Aldrich). The gradient analysis was performed using acetonitrile:water with 0.1% v/v acetic acid. The analysis time was 20 minutes, starting from 30% to 70% acetonitrile in 15 min; then, 70% to 100% 15–18 min, returning to 30% 18–19 min, finishing in 30% 20 min; using a flow of 1ml/min.

Western blotting

Total cellular protein extractions were carried out as described previously[53], and quantified using Bradford reagent (Bio-Rad), according to manufacturer’s instructions. A total of 50 μg of protein per sample was run on a 4% stacking and 12% resolving, self-made gel before being transferred to a membrane as described previously[14]. Blocking, primary (1:2000 dilution of the mouse monoclonal GFP antibody, #sc-9996, Santa Cruz Biotechnology) and secondary antibody (1:2000 dilution,[14]) incubations, and membrane revelation was carried out as described previously[14].

Bone marrow-derived macrophage (BMDM) preparation and corresponding assays

BMDMs were prepared according to [54], with modifications. Femurs and tibia from 8- to 12-week old C57BL/6 mice and gp91phox knockout (CGD—chronic granulomatous disease) mice were flushed with RPMI 1640 medium before flushed cells were cultured for 7 days in RPMI 1640 medium supplemented with 10% fetal cow serum (FCS) and 20 ng/ml M-CSF (Invitrogen). Non-adherent cells were removed, and adherent macrophages were collected and washed twice with cold PBS. Macrophage cell concentrations were determined using a Neubauer chamber. The phagocytic index was determined as described previously [55]. For fungal viability determination, 96-well plates were inoculated with 200 μl of RPMI-FCS medium containing macrophages and 1 x 10⁵ A. fumigatus conidia in a 1:2 macrophage:conidia ratio. Plates were incubated at 37°C, 5% CO₂ for 4 h before they were centrifuged at 400 × g for 10 min at 4°C. Supernatants were discarded and macrophage lysis was induced through the addition of ddH₂O. Serial dilutions of samples were carried out and plated on Sabouraud agar plates. Fungal colonies were counted after 18 h incubation at 37°C. Cytokine production of macrophages incubated with conidia (1:5 macrophage/conidia ratio) for 4 h at 37°C in 5% CO₂ was measured as described previously [56]. Heat-killed swollen conidia were prepared by inoculating 5 x 10⁶ conidia/ml in RPMI medium for 5 h at 37°C, before samples were centrifuged and conidia were heat-killed for 15 min at 121°C. All experiments were carried out in biological duplicates with 4 replicates each.

Infection of chemotherapeutic mice with A. fumigatus, fungal burden and histopathology

Immunosuppression of female BALB/c mice, A. fumigatus conidia suspension preparations, murine infections with A. fumigatus via intranasal instillation, fungal burden and histopathology of murine lungs were carried out exactly as described previous [56]. Briefly, mice were immunosuppressed intraperitoneally with cyclophosphamide on days -4, -1, and 2 prior to and post infection, and subcutaneously with hydrocortisonacetate on day -3 prior to infection. A. fumigatus conidia suspensions were prepared freshly a day prior to infection, washed three times with PBS, counted and diluted to a concentration of 5.0 x 10⁶ conidia/ ml. The viability of the administered conidia was determined by growing them in serial dilutions on YAG medium, at 37°C. Mice (n = 10/A. fumigatus strain) were infected by intranasal instillation of 1.0 x 10⁵ conidia diluted in 20 μl of PBS. Mice (n = 5) which received 20 μl of PBS served as the negative control. Survival curves were carried out in two independent experiments.

To determine fungal burden in the murine lungs, mice were infected as described above, but with a higher inoculum of 1×10⁶ conidia in 20 μl PBS. After 72 h, both lungs were extracted from the animals, frozen in liquid nitrogen, before DNA was extracted via the phenol-chloroform method. qPCR on the A. fumigatus 18S rRNA region and on the mouse GAPDH intronic region was carried out on at least 500 μg of total DNA from each sample.

For histopathology, murine lungs were fixed in 10% formaldehyde solution, diaphanized and embedded in paraffin before being sliced into 5 μm thick slices. Samples were stained with Hematoxylin and Eosin (HE) or GMS.

Phylogenetic tree construction and alignment

To determine the taxonomic distribution of the RglT, GliT, and the gliotoxin (GT) biosynthetic gene cluster (BGC), we identified homologs of all three genetic elements across genomes in the taxonomic classes Eurotiomycetes and Sordariomycetes that we retrieved from NCBI’s GenBank; only genomes with gene annotations were retrieved. From the 467 genomes that we downloaded, we removed 9 because they lacked RNA polymerase I β-subunit (RPA2), the largest RNA polymerase II subunit (RNAPol2), and the α-subunit of the translation elongation factor (EF-1), which we used for our fungal phylogenetic analyses (see below). This filtering resulted in the genomes of 242 Sordariomycetes and 216 Eurotiomycetes (458 total genomes).

To infer the presence or absence of RglT, GliT, and the GT BGC, we blasted the A. fumigatus Af293 RglT and GliT genes as well as each gene in the gliotoxin BGC against each of the 458 genomes using NCBI’s Blast+ v2.3.0[57]. For all genes, we used an E-value of 1e-4 and a query coverage of 50% (GliT and gliotoxin BGC genes) or 40% (RglT). To determine the presence or absence of a homologous gliotoxin BGC, we determined the physical proximity in the genome of all homologous sequences to genes in the BGC. We considered the gliotoxin BGC to be present in a genome if 7 / 13 Gli genes were physically clustered and the major synthase, GliP, was present among the clustered genes. Physical clustering was defined using a gene distance boundary of three.

To infer the evolutionary history of the 458 taxa in Sordariomycetes and Eurotiomycetes, we employed a maximum likelihood framework to reconstruct their evolutionary history using a concatenated matrix of RPA2, RNAPol2, and EF-1. Each gene was aligned using Mafft, v.7.402 [58] as previously described [59] and trimmed using trimAl, v.1.2rev59 [60], with the ‘gappyout’ parameter. The trimmed alignments were concatenated into a single matrix that contained 458 taxa and 3,377 amino acid sites. The phylogeny of the 458 taxa was inferred using IQ-Tree v.1.6.11 [61] with 20 independent tree searches and 10 candidate trees during each independent tree search. The best-fitting model of sequence substitution, LG+F+I+G4 [62,63], was determined automatically using Bayesian Information Criterion [64]. Bipartition support was determined using 5,000 ultrafast bootstrap approximations [65]. Poorly supported bipartitions–defined as less than 80% ultrafast bootstrap approximation support–were collapsed using TreeCollapseCL v.4 (http://emmahodcroft.com/TreeCollapseCL.html). The resulting phylogeny correctly recovered broad taxonomic relationships with the exception of Pencillicium antarcticum IBT 31811, which was placed among Sordariomycetes (S2F Fig). We ruled out that P. antarcticum IBT 31811 is a contaminant because genome-scale approaches have placed this strain among other Penicillium species [59]; the very long branch of this species suggests that this placement is likely a long branch attraction artifact.

To determine if RglT, GliT, and the gliotoxin were randomly distributed across the phylogeny or whether their distribution was correlated with the phylogeny, we measured the phylogenetic signal associated with the presence / absence of each genetic element using the D statistic [66] as implemented in the Caper, v.1.0.1 [67], package in R, v.3.5.2 (https://www.r-project.org/). To assess if there is a dependence relationship between RglT and GliT when accounting for phylogeny, we conducted model testing of four hypotheses using phytools, v.0.6–99 [68]. Specifically, we tested if (i) the pattern of occurence of RglT and GliT is independent of one another (our null hypothesis), (ii) the pattern of occurrence of GliT is dependent on RglT, (iii) the pattern of occurrence of RglT is dependent on GliT, and (iv) the patterns of occurrence of RglT and GliT are interdependent. We evaluated model fit using weighted Akaike information criterion and compared the best fitting model to the null hypothesis using a chi-squared test.