A. fumigatus strains used in this study are listed in Table S1. Strains were maintained as glycerol stocks and activated on solid glucose minimal medium (GMM) at 37°C with appropriate supplements (Shimizu and Keller, 2001). For pyrG auxotrophs, the growth medium was supplemented with 5 mM uridine and uracil. Conidia were harvested in 0.01% Tween 80 and enumerated using a hemacytometer. For RNA-seq analysis strains Af293, TWY32.1, and TWY24.121 were inoculated into 50 mL of liquid GMM at 5 x 10⁶ conidia/mL in duplicate and grown at 25°C and 250 rpm for 96 h in ambient light conditions. The mycelium was harvested and lyophilized before RNA extraction. For biomass, siderophore and iron-dependent gene expression analysis, strains CEA17, TWY37.2, TJW109.3, TWY25.5, and TWY28.3 were grown in 50 mL Aspergillus minimal media (AMM) according to (Pontecorvo et al., 1953) containing 20 mM glutamine and iron concentrations as indicated in the text according to Schrettl et al. (2008). Strains were grown in triplicates (duplicates for gene expression analysis) for 24 h at 37°C, 250 rpm and ambient light conditions with an initial spore concentration of 5 × 10⁶ conidia/mL. For analysis of secondary metabolites and cluster gene expression, strains ATCC46645, Δ hapX, and Δ sreA were inoculated into 50 mL of GMM containing different iron concentrations as indicated in the text at 5 × 10⁶ conidia/mL. Strains were grown in triplicates (duplicates for gene expression analysis) for 5 days (3 days for gene expression analysis) at 25°C and 250 rpm at ambient light conditions. For growth assays strains TJW55.2, TWY32.1, TWY38.6, TWY24.121, TWY35.1, and TW36.1 indicated amount of conidia were inoculated in 2 μL on solidified (Noble Agar, Difco™, BD, USA) GMM containing indicated iron concentrations and FEC, respectively, and incubated for 3 days at 37°C in the dark.

For DNA isolation, A. fumigatus strains were grown for 24 h at 37°C in steady state liquid GMM, supplemented with 1 mM FeSO₄. DNA isolation was performed as described by Green and Sambrook (2012). The A. fumigatus mutant strains were constructed using a double-joint fusion PCR approach (Szewczyk et al., 2006) and all primers used in the study are listed in Table S2. Briefly, approximately 1 kb fragments flanking the targeted deletion region were amplified by PCR from A. fumigatus strain Af293 or CEA17 genomic DNA and the A. parasiticus pyrG marker gene was amplified from the plasmid pJW24 (Calvo et al., 2004). The three fragments were subjected to fusion PCR to generate deletion cassettes. The A. parasiticus pyrG marker gene was amplified from the plasmid pJW24 using the primer pair pyrG_prom_F/pyrG_term_R (Table S2). The primers “gene”-5R and “gene”-3F contain complement sequences to the primers pyrG_prom_F and pyrG_term_R at their 5'-region, respectively (Table S2). The fusion construct was created by PCR containing 5′ and 3′ gene flanks and the pyrG gene fragment functioning as templates and primers simultaneously. The final PCR fusion product was amplified using primer pairs “gene”-5F/“gene”-3R and the previously PCR-generated fusion construct as template. Transformation was performed as described by Palmer et al. (2008). For selection of Δ sidA transformants, 1 mM FeSO₄ was added to the selection media. For multiplex diagnostic PCR, primer pair “gene”-F/“gene”-R were used to identify transformants that lost the respective gene locus and primer pair gpd_int-F/gpd_int-R as internal control (Table S2). Single integration of the transformation construct was confirmed by Southern analysis as described by Green and Sambrook (2012) using P³²-labeled probes created by amplification of the respective deletion construct using primer pair “gene”-5F/“gene”-3R (Table S2 and Figure S3).

Mycelia were harvested by filtration through Miracloth (Calbiochem). Total RNA was extracted with TRIzol reagent (Invitrogen) from freeze-dried mycelia, following the manufacturer's protocol. Northern analysis was performed as described by Green and Sambrook (2012). Probes for northern analysis were constructed at regions internal to the gene of interest using primers listed in Table S2 (“gene”-F/“gene”-R) and labeled with dCTP αP³².

To identify transcriptionally active genes, extracted RNA samples were subject to DNase treatment using the RNeasy kit (Qiagen), and sequencing libraries were generated using the ScriptSeq kit v2 (Epicenter) following manufacturer's directions. All libraries were sequenced (7 sample per lane) using the Illumina HiSeq2000 instrument (www.illumina.com) on a 2 × 100 bp paired-end run. Transcripts were assembled and expression levels were estimated using normalized reads per kilobase per million mapped reads (RPKM) (Mortazavi et al., 2008) as calculated by the TopHat, Bowtie, and Cufflinks packages (Trapnell et al., 2009, 2010, 2012; Langmead et al., 2010) and CLC Genomics Workbench (CLC Bio). After trimming for quality (>Q30), reads were mapped at 90% length, 90% identity and reads that mapped to more than one location in the genome were excluded. A minimum of 4 reads in each condition were required to analyze gene expression, over 88% of genes had over 4 reads in every condition. Differentially expressed genes were determined in duplicate data by identifying genes that were at least two-fold difference using a T-test cutoff of p < 0.05 (Bullard et al., 2010) and by using SAM with a false discovery rate of 0 as implemented in MeV (Saeed et al., 2006). Replicates of each condition were highly reproducible (R² > 0.99) (Dataset S1). For gene enrichment analysis, genes differentially regulated were analyzed for enrichment using the Munich Information Center for Protein Sequences (MIPS) Functional Catalog (FunCat) (Ruepp et al., 2004) at http://www.helmholtz-muenchen.de/en/ibis/resourcesservices/index.html (Dataset S2).

Analysis of siderophores was carried out by reversed phase high performance liquid chromatography (HPLC) as described previously (Oberegger et al., 2001). Moreover, to quantify extracellular or intracellular siderophores, culture supernatants or cellular extracts were saturated with FeSO₄ and siderophores were extracted with 0.2 volumes of phenol. The phenol phase was separated and subsequent to addition of five volumes of diethylether and 1 volume of water, the siderophore concentration of the aqueous phase was measured photometrically using a molar extinction factor of 2996 M⁻¹ cm⁻¹ at 440 nm.

For analysis of pseurotin A, fumitremorgin C and fumagillin supernatant of fungal cultures was lyophilized and re-dissolved in 5 mL double destilled (dd) H₂O 1% (v/v) formic acid (FA). From the water crude, 800 μL were partitioned with 800 μL ethyl acetate for three times. The combined organic layers of each sample were evaporated in vacuo, and re-dissolved in 500 μL of 20% (v/v) acetonitrile (ACN) 1% FA (v/v) and 50 μL were examined by HPLC photo diode array (PDA) analysis. The samples were separated on a ZORBAX Eclipse XDB-C18 column (Agilent, 4.6 mm by 150 mm with a 5 μm particle size) by using a binary gradient of 1% (v/v) FA as solvent A and 1% FA in ACN as solvent B using a Flexar Binary Liquid Chromatography (LC) Pump (PerkinElmer) coupled to a Flexar LC Autosampler (Perkin Elmer) and a Flexar PDA Plus Detector (PerkinElmer). The binary gradient started with an isocratic step at 80% A for 2 min followed by a linear gradient to 40% A in 10 min and an additional linear gradient to 100% B in 0.5 min. After each run the column was washed for 5 min using 100% B and was equilibrated for 4 min using 80% A. The flow rate was set to 1.5 mL/min. Identification and relative quantification of secondary metabolites was performed using Chromera Manager (PerkinElmer) by comparison to standards for fumagillin, pseurotin A and fumitremorgin C (Cayman Chemicals, MI, USA) (Dataset S3). Areas under the curve were normalized to dry weights of each sample. For terezine D analysis liquid fungal cultures including fungal tissue and media were frozen using a dry ice acetone bath, and lyophilized. The lyophilized residues were extracted with 75 mL of 10% methanol in ethyl acetate for 3.5 h with vigorously stirring. Extracts were filtered over cotton, evaporated to dryness, and stored in 8 mL vials. Preparation for HPLC-mass spectrometry (MS) Analysis: Crude extracts were suspended in 0.1 mL of methanol and centrifuged to remove insoluble materials, and the supernatant was subjected to HPLC-mass spectrometry analysis using an Agilent 1100 series HPLC connected to a Quattro II low resolution mass-spectrometer (Micromass/Waters) operated in electrospray positive ionization (ESI⁺). Data acquisition and processing for the HPLC-MS was controlled by Waters MassLynx software. An Agilent Zorbax Eclipse XDB-C18 column (4.6 × 250 mm, 5 μm particle diameter) was used with 0.1% acetic acid in 50/50 methanol/acetonitrile (organic phase) and 0.1 % acetic acid in water (aqueous phase) as solvents at a flow rate of 1.0 mL/min. A solvent gradient scheme was used, starting at 5% organic for 3 min, followed by a linear increase to 100% organic over 35 min, holding at 100% organic for 15 min, then decreasing back to 5% organic for 1 min and holding at 5% organic for the final 6 min, a total of 1 h. Relative abundance of terezine D was determined by integrating their extracted ion chromatograms using MassLynx software (Dataset S3).

Statistical analysis was performed by using GraphPad Prism software using analysis of variance (ANOVA) followed by Tukey's test to show significant differences.

Since HAS was shown to chelate ferric iron, similar to the intra- and extra-cellular siderophores produced by A. fumigatus, we investigated whether HAS could restore growth defects in an A. fumigatus sidA deletion mutant deficient of siderophore production. Deletion mutants of sidA in A. fumigatus were previously shown to exhibit strong growth defects under iron deficient conditions and to be avirulent in animal models of invasive aspergillosis (Schrettl et al., 2004a, 2007; Hissen et al., 2005; Knox et al., 2014). Single (ΔsidA) and double (OE::hasA/ΔsidA) mutants of sidA with or without an overexpression hasA allele were created as described in Material and Methods (Figure S1). HasA is the has cluster transcription factor positively activating other members of the cluster and HAS biosynthesis (Yin et al., 2013).

Confirming previous results (Schrettl et al., 2004a), sidA deletion mutants showed significant growth defects under iron depleted conditions compared to the wild type (WT), the hasA deletion mutant (ΔhasA), the overexpression hasA mutant (OE::hasA), as well as a mutant overexpressing hasA in a hasD deletion background (OE::hasA/ΔhasD) where hasD encodes the non-ribosomal peptide synthetase required for HAS biosynthesis (Figure 1). We found that OE::hasA in the ΔsidA background did not rescue growth (Figure 1), thereby ruling out the hypothesis that HAS could function as an extra-cellular siderophore assisting in iron uptake. Growth of ΔsidA and OE::hasA/ΔsidA could be restored to wild-type like levels when 2 μM of the Schizosaccharomyces pombe siderophore ferrichrome (Schrettl et al., 2004b) was supplemented to iron deplete conditions. Under extreme iron excess (10 mM), growth rates of the WT and the OE::hasA strains were decreased to similar levels (Figure 3), indicating that HAS, similar to FC and HFC, does not participate in iron detoxification in contrast to the vacuolar iron transporter CccA (Gsaller et al., 2012) (Figure 1).

In order to confirm the ability of HAS and its precursor astechrome to bind iron as reported previously (Arai et al., 1981; Yin et al., 2013), we cultivated the WT, OE::hasA, OE::hasA/ΔhasD, and OE::hasA/ΔhasG strains under increasing iron conditions. HasG is a FAD binding protein required to convert astechrome to HAS and both compounds chelate iron (Arai et al., 1981; Yin et al., 2013). OE::hasA and OE::hasA/ΔhasG exhibited a reddish mycelial coloration particularly during iron excess in contrast to the wild type and OE::hasA/ΔhasD (Figure 2).

Since HAS could be excluded to function as an extracellular siderophore, we investigated if it could function in iron detoxification under iron-replete and iron excess conditions. Therefore, we assessed biomass of the WT, ΔhasA, OE::hasA, and OE::hasA/ΔhasD strains. During iron-replete (30 μM), iron deficient (1 μM) and no iron (0 μM) conditions OE::hasA showed decreased biomass production compared to the other strains tested (Figure 3). In case of OE::hasA this growth defect could be rescued by addition of 1 mM iron to the media (Figure 3). Taken together these data suggest that HAS causes growth deficiency due to its iron-chelating ability.

In line with the observed biomass reduction, production of intra- and extra-cellular siderophores was increased under iron-replete conditions (30 μM) in strains overproducing HAS (OE::hasA) compared to the WT or OE::hasA/ΔhasD (Figure 4). At iron-replete conditions (30 μM iron), OE::hasA produces more extracellular and intracellular siderophores than WT or the other has mutants thus supporting the hypothesis that HAS causes the fungus to experience iron-deficiency in this condition. However, under iron starvation (0 μM) no increase in siderophore production could be observed in OE::hasA compared to the WT, most likely due to the fact that siderophore production is already fully induced under those conditions in all strains. In iron excess conditions, siderophore production is similar across all strains (Figure 4).

In A. fumigatus, iron starvation induces the expression of genes involved in siderophore biosynthesis (e.g., sidA) and reductive iron uptake (e.g., ftrA). Moreover, iron starvation represses genes involved in iron dependent pathways such as cycA (Schrettl et al., 2010a). During iron-replete conditions (30 μM), OE::hasA and OE::hasA/ΔhasG increase expression of sidA and ftrA but repress cycA expression compared to the WT, ΔhasA and OE::hasA/ΔhasD, respectively (Figure 5A). These data indicate that HAS or astechrome accumulation causes iron starvation, which is in agreement with the reduced biomass formation and increased siderophore production under this condition (Figures 3, 4). Since expression of genes involved in iron acquisition and iron-utilization are repressed by SreA and HapX, respectively, in A. fumigatus (Schrettl et al., 2008, 2010a), we tested expression of sreA in iron deplete conditions and conditions exposed to iron sufficiency for a short time. As previously described (Schrettl et al., 2008), shifting to iron-replete conditions caused induction of sreA expression compared to iron-deplete conditions in the WT. This induction is less pronounced in OE::hasA and OE::hasA/ΔhasG compared to the other strains tested (Figure 5B). These data are in line with the growth and siderophore production data indicating that HAS overexpression decreases iron bioavailability.

As our cumulative data demonstrate a role of HAS in iron homeostasis of A. fumigatus we were curious if the has gene cluster itself is subject to HapX- or SreA-mediated iron-dependent expression. Therefore, we grew the WT, ΔhapX, and ΔsreA under iron deplete (0 μM), replete (30 μM) and excess (1 mM) conditions and assessed HAS production and gene expression by northern blot analysis. Expression of select has genes was repressed by iron depleted conditions in a HapX-independent manner (Figure 6A). Gene expression of hasB and concomitant terezine D production, the first stable intermediate of the HAS pathway, was induced by medium-high iron concentrations and SreA-deficiency, which increases the cellular iron accumulation (Figures 6A,B).

Since we observed an iron-dependent expression of the has cluster and HAS production, we investigated expression of select genes from the gene clusters responsible for fumitremorgin C, pseurotin A and fumagillin production, as these secondary metabolites are known to be produced under the same wild-type conditions and therefore convenient to measure (Wiemann et al., 2013). Similar to the observed regulation for the has cluster, select genes from the pso and fma supercluster (Wiemann et al., 2013) were co-activated followed by increased product formation with increasing iron concentration in the WT (Figures 7, 8).

Both SreA and HapX loss perturbed metabolite production, reflected in transcriptional regulation in the sreA but not hapX mutant. In contrast to SreA negative regulation of HAS, fumagillin and pseurotin A production and gene expression was decreased in the ΔsreA mutant with increasing iron concentrations (Figures 7, 8). Production of fumagillin and pseurotin A, as well as fumitremorgin, was greatly reduced in ΔsreA in high-iron conditions (Figure 9). All four secondary metabolites were differentially produced in the ΔhapX mutant depending on iron concentration (Figure 8) with production of fumagillin and pseurotin significantly increased in ΔhapX compared to the WT under no-iron and iron-replete conditions (Figure 9).

A curious observation was made for fumitremorgin C where its production followed much of the same trend as fumagillin and pseurotin A in both ΔsreA ΔhapX mutants (Figures 8, 9) but, unlike fumagillin and pseurotin, stayed unaffected by increasing iron concentration in WT (Figure 8). Interestingly—and in contrast to fumagillin and pseurotin A cluster genes - expression of ftmH was highest under iron-deficient conditions, despite the lack of product formation under those conditions (Figure 7). Notably, fumitremorgin C production is significantly increased in the ΔhapX mutant compared to the WT under no-iron but not iron excess conditions, indicating that HapX functions as a repressor under these conditions (Figure 9).

The above data strongly suggested HAS to be a player in an iron acquisition and consumption circuitry involving secondary metabolite production. In order to more thoroughly interrogate these hypotheses, we performed RNA-seq analysis comparing overexpression hasA (e.g., OE::hasA = high HAS) to WT (low HAS) or to OE::hasA/ΔhasD (no HAS). Compared to the WT, 583 and 435 genes were up- and down-regulated in the OE::hasA strain, respectively. When comparing the OE::hasA strain to the strain OE::hasA/ΔhasD, 767 were up- and 394 genes were down-regulated in the strain able to produce HAS (Figure 10; Dataset S1). A common set of 404 and 291 genes to up- and down-regulated, respectively, was identified between the two comparisons (Figure 10; Dataset S1). Enrichment analysis of these common gene sets of the OE::hasA strain compared to both the WT and the OE::hasA/ΔhasD strains, shows that “secondary metabolism” is a category to be both positively and negatively affected (Figures S1, S2; Dataset S2). Analyzing genes from this category more closely, based on characterized and predicted (Khaldi et al., 2010; Medema et al., 2011; Inglis et al., 2013) gene clusters, provided an accurate view on the genes affected. As expected, genes belonging to the HAS cluster were found to be up-regulated (Table 1).

Additionally, genes from several predicted but uncharacterized clusters also are positively affected: five and eight genes from two uncharacterized NRPS-like gene clusters on chromosome III (AFUA_3g02670; 02680, 02685, 02690, and 02760) and V (AFUA_5g09960; 10040, 10090, 10160, 10180, 10200, 10210, and 10220), respectively, as well as one gene belonging to an uncharacterized PKS cluster on chromosome I (AFUA_1g17690), four genes belonging to an uncharacterized PKS gene cluster on chromosome III (AFUA_3g01400; 01460; 01500, and 01560) and three genes belonging to an uncharacterized PKS cluster on chromosome IV (AFUA_4g14530, 14570, and 14580) (Table 1). Interestingly, the uncharacterized PKS-encoding gene on chromosome I (AFUA_1g17740) as well as one of the uncharacterized PKS-encoding gene on chromosome III (AfuA_3g14700) and additional five genes from the same cluster (AFUA_3g14710, 14740, 14750; 14760, and 14770) were only found to be up-regulated comparing the OE::hasA strain to OE::hasA/ΔhasD, in which neither HAS nor any intermediate is produced (Table 1). Similarly, three (fgaPT1/easL, fgaDH/easD, and fgaP450-2/easM) of the 11 characterized genes from the fumigaclavine gene cluster (Unsold and Li, 2006; Wallwey et al., 2010; Robinson and Panaccione, 2012; Wallwey et al., 2012) as well as one uncharacterized NRPS-encoding gene [AFUA_3g13730; nrps6/pesG; (Cramer et al., 2006b; Stack et al., 2007)] were only found to be up-regulated when comparing OE::hasA to OE::hasA/ΔhasD (Table 1).

Of the secondary metabolite gene clusters found to be down-regulated in the common gene set, all of the 13 gliotoxin cluster genes (gliZ, gliI, gliJ gliP, gliC, gliM, gliG, gliK, gliA, gliN, gliF, gliT, and gliH) (Gardiner et al., 2005; Balibar and Walsh, 2006; Schrettl et al., 2010b; Davis et al., 2011; Forseth et al., 2011; Scharf et al., 2012a,b) (Figure 10), as well as the adjacently located NRPS-encoding gene nrps9/pesJ [AFUA_6g09610; (Cramer et al., 2006b; Stack et al., 2007)], seven (ftmA, ftmB, ftmC, ftmD, ftmE, ftmG, and ftmH) of the eight fumitremorgin cluster genes (Maiya et al., 2006; Grundmann et al., 2008; Kato et al., 2009; Steffan et al., 2009; Wiemann et al., 2013), and three (pyr6, pyr7, and pyr8) of the nine pyripyropene cluster genes (Itoh et al., 2012) were identified (Table 1). Additionally, one gene (AFUA_1g10370) adjacent to the NRPS-encoding gene nrps1/pes1/pesB [AFUA_1g10380; (Cramer et al., 2006b; Reeves et al., 2006; O'Hanlon et al., 2012)] and the uncharacterized PKS-encoding gene (AFUA_3g02570) were also negatively affected (Table 1).

Overall, as noted above, the effect of hasA overexpression was dampened in the absence of hasD, with the fold differences in the secondary metabolite clusters often being smaller and the number of differentially expressed genes fewer. These results suggest that for the majority of gene clusters HAS itself, and not just HasA, plays a role in regulation of secondary metabolite biosynthetic genes.

Another category that we found to be enriched specifically among the genes positively affected comparing OE::hasA to the WT and OE::hasA to OE::hasA/ΔhasD belong to the functional group “heavy metal ion transport” (Figure S1; Table S2). This included the metalloreductase-encoding gene fre2/freB, the putative ferric chelate reductase-encoding gene fre7, the putative zinc transporter fetD/fet4, and two putative siderophore transporters (AFUA_3g13670 and AFUA_7g04730) (Table 2) all previously identified to be induced by iron or zinc starvation (Schrettl et al., 2008; Yasmin et al., 2009; Blatzer et al., 2011a,b). Two of the genes in this category were also induced by exposure to neutrophils (Sugui et al., 2008), one of them encoding a putative high affinity copper transporter (AFUA_2g03730) and the other one encoding a putative metalloreductase (AFUA_6g02820) (Table 2). These results support the hypothesis that HAS is a component of an iron-centered regulon, that when over-produced induced iron starvation responsive genes.

On the other hand, several genes repressed by iron sufficiency in an SreA-dependent manner (Schrettl et al., 2008), were only identified to be up-regulated when comparing the OE::hasA strain to OE::hasA/ΔhasD (Table 3). Specifically, the gene cluster located on chromosome III (AfuA_3g03390-03440) involved in siderophore production is up-regulated in this comparison. Genes identified in this cluster encode for the hydroxyornithine transacylase SidF, the fusarinine C NRPS SidD, the ABC transporter SitT and the fusarinine C acyltransferase SidG (Table 3).

Several of the genes negatively affected in both comparisons (OE::hasA to WT and OE::hasA to OE::hasA/ΔhasD) belong to Fe-dependent enzymes involved in primary metabolism (Figure S2; Dataset S2), supporting the hypothesis that overproduction of HAS leads these strains to experience iron-deficiency. Among those genes were eight putative oxidoreductases (AFUA_1g17430; AFUA_2g03820; AFUA_2g09355; AFUA_3g00150; AFUA_3g01070; AFUA_4g13780; AFUA_5g09720; AFUA_6g03290) (Table 4). In line with this finding we observed an up-regulation of the cytochrome C-encoding gene (cycA) in northern blot analysis (Figure 5B). Similar to what was observed for the genes belonging to secondary metabolism, the global effects of hasA overexpression were also lessened in the absence of hasD. In this case, only 101 and 194 genes were up- or down-regulated in the OE::hasA/ΔhasD strain compared to WT, with less than a third of those having differences greater than four-fold, supporting our the hypothesis that HAS itself causes the major effects observed (Dataset S1).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.