Mevalonate governs interdependency of ergosterol and siderophore biosyntheses in the fungal pathogen Aspergillus fumigatus

Abstract

Aspergillus fumigatus is the most common airborne fungal pathogen for humans. In this mold, iron starvation induces production of the siderophore triacetylfusarinine C (TAFC). Here we demonstrate a link between TAFC and ergosterol biosynthetic pathways, which are both critical for virulence and treatment of fungal infections. Consistent with mevalonate being a limiting prerequisite for TAFC biosynthesis, we observed increased expression of 3-hydroxy-3-methyl-glutaryl (HMG)-CoA reductase (Hmg1) under iron starvation, reduced TAFC biosynthesis following lovastatin-mediated Hmg1 inhibition, and increased TAFC biosynthesis following Hmg1 overexpression. We identified enzymes, the acyl-CoA ligase SidI and the enoyl-CoA hydratase SidH, linking biosynthesis of mevalonate and TAFC, deficiency of which under iron starvation impaired TAFC biosynthesis, growth, oxidative stress resistance, and murine virulence. Moreover, inactivation of these enzymes alleviated TAFC-derived biosynthetic demand for mevalonate, as evidenced by increased resistance to lovastatin. Concordant with bilateral demand for mevalonate, iron starvation decreased the ergosterol content and composition, a phenotype that is mitigated in TAFC-lacking mutants.

Host-imposed iron limitation necessitates elaborate iron-acquisition mechanisms in pathogenic microbes, whereby fungal agents of human disease favor high-affinity reductive iron uptake and chaperone-assisted iron assimilation (1). The latter mechanism is favored by inhaled spores of the saprobic mold Aspergillus fumigatus, the most common airborne fungal pathogen of humans and an agent of life-threatening invasive disease in immunocompromised patients (2). In the absence of direct utilization mechanisms for human iron sources such as heme, ferritin, or transferrin, A. fumigatus employs both reductive iron assimilation (RIA) and siderophore (low-molecular-mass ferric iron chelators)-mediated iron uptake during murine infection (3). Siderophore biosynthesis is a fungus-specific, virulence-essential biosynthetic process and a significant prospect for rationalized therapeutic intervention.

A. fumigatus produces three major hydroxamate-type siderophores for iron uptake: extracellular triacetylfusarinine C (TAFC), which removes iron from transferrin (4), and hyphal ferricrocin (FC) and conidial hydroxyferricrocin (HFC), which facilitate intracellular sequestration and distribution of iron (5, 6). Genetic abrogation of the entire siderophore system renders A. fumigatus avirulent in a murine model of invasive pulmonary aspergillosis; inactivation of either extra- or intracellular siderophore biosynthesis results in partial attenuation of virulence (3, 5).

The cyclic tripeptide TAFC consists of three N²-acetyl-N⁵-anhydromevalonyl-N⁵-hydroxyornithines, linked by ester bonds and is derived from the N²-acetyl–lacking precursor fusarinine C. The presence of the anhydromevalonyl moiety in fusarinine (N⁵-anhydromevalonyl-N⁵-hydroxyornithine) was observed first by Diekmann and Zähner (7) as a degradation product of fusarinine. Subsequent studies (8) demonstrated that anhydromevalonic acid is synthesized from mevalonic acid and that the CoA derivative of anhydromevalonic acid, along with N⁵-hydroxyornithine, forms fusarinine (9). Thus far we have populated the proposed biosynthetic scheme (Fig. 1) with five genes encoding respective A. fumigatus siderophore biosynthetic enzymes (3, 5). However, candidate genes for enzymes converting mevalonate to anhydromevalonyl-CoA have not yet been identified in any fungal species. Moreover, the regulatory link between siderophore and ergosterol biosynthesis remains elusive.

Fig. 1.

Ergosterol (green) and siderophore (purple) biosynthetic pathways in A. fumigatus. Biosynthesis of both TAFC and FC starts with N⁵-hydroxylation of ornithine. Subsequently, the hydroxamate group is formed by the transfer of an acyl group from acyl-CoA derivatives to N⁵-hydroxyornithine. Here the pathways for biosynthesis of TAFC and FC split because of the choice of the acyl group with acetyl for FC and anhydromevalonyl for TAFC. Assembly of the cyclic siderophores fusarinine C and FC is catalyzed by different nonribosomal peptide synthetases (NRPS). TAFC and hydroxyferricrocin are formed by N²-acetylation of fusarinine C and hydroxylation of FC, respectively. Anhydromevalonyl-CoA originates from mevalonate via CoA-esterification and dehydration. Previously identified A. fumigatus genes encoding respective siderophore biosynthetic enzyme activities are indicated in black (3, 5). The functionally characterized sidI and sidH genes (this study) are shown in orange. TAFC, FC, mevalonate, and ergosterol are in red. (Adapted from ref. 5.)

In A. fumigatus, iron acquisition, including biosynthesis and uptake of siderophores, is repressed by iron via the GATA-type transcription factor siderophore repressor A (SreA), and genome-wide microarray analysis identified 49 SreA-controlled genes (10). Notably, SreA must be distinguished from the basic helix–loop–helix type transcription factor SrbA, a recently identified regulator of ergosterol biosynthesis in A. fumigatus (11). In the current study, we identified the enzymes connecting ergosterol and siderophore biosynthetic pathways within the SreA regulon and elucidated the regulatory link between these pathways. This link might be crucial for optimization of treatment of infections caused by siderophore-producing fungi.

Results

Iron Starvation Up-Regulates Expression of the Gene Encoding 3-Hydroxy-3-Methyl-Glutaryl–CoA Reductase, hmg1.

Mevalonate is a key metabolite in the biosynthesis of sterols such as fungal ergosterol and a variety of nonsterol isoprenoid derivatives (Fig. 1). Mevalonate is synthesized from 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) catalyzed by HMG-CoA reductase, the rate-limiting enzyme of sterol biosynthesis. Subsequently, mevalonate is phosphorylated by mevalonate kinase.

Recently, genome-wide transcription profiling has indicated that expression of the gene encoding HMG-CoA reductase, hmg1, is down-regulated in a shift from iron-depleted to iron-replete conditions in A. fumigatus (10), whereas the expression of the gene encoding mevalonate kinase, erg12, is up-regulated. Northern blot analysis confirmed that iron starvation increases hmg1 transcript levels but decreases erg12 transcript levels (SI Appendix, Fig. S1A). Consistent with the down-regulation of erg12, the total cellular ergosterol content of A. fumigatus mycelia decreased to 38% during iron-depleted as compared with iron-replete growth (Fig. 2). Therefore, the observed up-regulation of hmg1 expression during iron starvation does not affect ergosterol biosynthesis. We speculated that this expression pattern reflects an increased demand for mevalonate as a precursor for TAFC biosynthesis, which is induced by iron starvation. Consistent with this hypothesis, genome-wide transcription profiling data do not indicate that iron availability has any effect on the expression of genes encoding HMG-CoA reductase in fungal species that do not produce siderophores (e.g., Saccharomyces cerevisiae, Cryptococcus neoformans, and Candida albicans) or that produce non–mevalonate-derived siderophores (e.g., Ustilago maydis) (12–15).

Fig. 2.

A. fumigatus total ergosterol content is affected by iron availability, TAFC biosynthesis, and hmg1 overexpression. (A) The ergosterol content of the WT strain was measured after growth for 48 h in minimal medium without iron (Fe^–) and with iron (Fe⁺). (B) The ergosterol contents of ΔsidI, ΔsidH, and hmg1^OE mutants were normalized to that of the WT strain grown under the conditions shown in A. Error bars indicate SD for three independent experiments.

Lovastatin Inhibits TAFC Production.

Inhibition of HMG-CoA reductase by treatment with 1 μM lovastatin (16) reduced the TAFC production to 63% of that in untreated WT A. fumigatus without affecting the biomass, suggesting that HMG-CoA reductase plays an essential role in TAFC biosynthesis (Table 1). Treatment with 8 μM of lovastatin reduced the TAFC production to 14% and the biomass to 16% of the untreated control, underscoring that HMG-CoA reductase activity is essential for both growth and TAFC biosynthesis.

Table 1.

Comparison of biomass and TAFC production of WT and hmg1^OE in response to lovastatin treatment

		Lovastatin (μM)
Strain	TAFC/Biomass	0	0.5	1	2	4	8
WT	TAFC	1.00	0.97	0.63	0.50	0.25	0.14
	Biomass	1.00	1.00	1.00	0.94	0.54	0.16
hmg1OE	TAFC	2.06	1.75	1.19	1.15	0.93	0.88
	Biomass	1.06	1.16	1.00	1.04	0.88	0.91

Cultures were grown for 48 h in minimal medium containing xylose as carbon source and lovastatin. The biomass and TAFC production of the WT without lovastatin were 0.34 g and 511 μmol/g dry weight, respectively. The values represent the means of three experiments, normalized to the untreated WT. SD in each case was <10%.

Overexpression of hmg1 Increases TAFC Production.

To assess directly the involvement of Hmg1 in TAFC production, we tried to delete the hmg1 gene. However, these efforts remained futile, most likely because of the essentiality of this gene. Overexpression of hmg1 by placing the gene under the control of the xylose-inducible xylP promoter (SI Appendix, Figs. S1B and S2C) (17, 18) resulted in more than twofold increased TAFC production compared with WT in the absence of lovastatin (Table 1). Consistent with Hmg1 being the target of lovastatin toxicity, hmg1 overexpression increased lovastatin resistance and TAFC biosynthesis under lovastatin treatment. Thus, 8 μM of lovastatin decreased the biomass of the WT and hmg1^OE strains to 16% and 91% of the untreated WT, respectively. Overexpression of hmg1 also increased the ergosterol content by about 14% in both iron-depleted and iron-replete conditions (Fig. 2). Together, these results indicate that Hmg1 represents a bottleneck in TAFC biosynthesis.

The Acyl-CoA Ligase SidI and the Enoyl-CoA Hadratase SidH Are Conserved in Siderophore-Producing Fungi.

In iron-replete conditions siderophore biosynthesis in A. fumigatus is repressed by the GATA transcription factor SreA (10). Mining of the recently identified SreA regulon identified two genes, sidI (AFUA_1G17190) and sidH (AFUA_3G03410), encoding candidate enzymes for activation and dehydration, respectively, of mevalonate for TAFC biosynthesis (Fig. 1). sidI and sidH are flanked by siderophore biosynthetic genes (5), and Northern analysis confirmed up-regulation of sidI and sidH during iron starvation (10) (SI Appendix, Fig. S3).

SidI and SidH belong to the acyl-CoA ligase and enoyl-CoA hydratase protein families, respectively, with members in all organisms. The best-characterized members are involved in β-oxidation. Phylogenetic analysis (SI Appendix, Fig. S4) revealed that all closely related SidI homologs are found in Pezizomycotina species known or believed to produce mevalonate-derived siderophores (1). In contrast, non–siderophore-producing Pezizomycotina such as S. cerevisiae and C. albicans do not possess SidI orthologs. Similarly, Schizosaccharomyces pombe that produces the siderophore ferrichrome (19), which lacks a mevalonate-derived moiety, does not possess an SidI ortholog. This phylogenetic analysis underscores the possible involvement of SidI in siderophore production and demonstrates its evolutionary conservation. As is evident from the phylogenetic analysis (SI Appendix, Fig. S5), only siderophore-producing fungi possess close SidH homologs.

Deletion of SidI or SidH Blocks TAFC Biosynthesis, Decreases Resistance to Oxidative Stress, and Attenuates Virulence in a Murine Model of Invasive Pulmonary Aspergillosis.

To understand the role of SidI and SidH, we generated respective gene deletion mutants in A. fumigatus ATCC 46645 (WT), termed “ΔsidI” and” ΔsidH,” respectively (SI Appendix, Figs. S2A and S6A). Consistent with the expression of sidI and sidH being restricted to conditions of iron limitation, both ΔsidI and ΔsidH displayed no difference from WT with respect to growth rate on iron-replete solid medium (Fig. 3). In contrast, ΔsidI and ΔsidH showed a mild growth reduction during iron depletion and were unable to grow in the presence of the ferrous iron-specific chelator bathophenanthroline disulfonate (BPS), which inhibits RIA and renders siderophore-mediated iron uptake the only active high-affinity acquisition system (3). In agreement, deficiency in sidI or sidH impaired production of extracellular TAFC (Fig. 4). Interestingly, deficiency in sidI, but not sidH, substantially decreased the FC content (Fig. 4). Complementation of ΔsidI and ΔsidH with the respective WT copy of the deleted gene (SI Appendix, Figs. S2B and S6B) restored the WT phenotype (Fig. 3), demonstrating that all phenotypic consequences are direct results of the gene deletions. These results demonstrate that SidI and SidH are essential for TAFC synthesis and that RIA partially compensates for the lack of siderophore-mediated iron uptake.

Fig. 3.

Both ΔsidI and ΔsidH are highly susceptible to iron starvation and oxidative stress. The fungal WT strain, mutant strains, and complemented strains were grown for 48 h on minimal medium agar under iron-depleted (Fe^–) or iron-replete (Fe⁺) conditions without or with H₂O₂ (2 mM) or BPS (250 μM). (Left) Radial growth of the WT strain. (Right) Growth of the mutant and complemented strains normalized to WT.

Fig. 4.

SidI and SidH are required for the synthesis of TAFC. HPLC analysis of supernatant and cell extract from cultures of WT, ΔsidH, and ΔsidI strains grown in iron-depleted conditions (Fe^–) for 24 h. Units are given in milli absorption units (mAU).

In A. fumigatus, iron starvation decreases the cellular iron content and down-regulates iron-dependent pathways, including heme and iron sulfur cluster biosynthesis (10). Deficiency in sidI or sidH also decreased resistance to hydrogen peroxide-caused oxidative stress (Fig. 3), an effect that might be explained by the decreased iron uptake and consequently impaired iron-dependent antioxidative defense (10). Remarkably, ΔsidI displayed significantly higher susceptibility to hydrogen peroxide and iron starvation than did ΔsidH (Fig. 3). Because TAFC biosynthesis is blocked at different steps in ΔsidI and ΔsidH, the phenotypic differences might be caused by accumulation of different intermediates with different consequences on fungal metabolism (Discussion).

Comparative analysis of WT, ΔsidH, and ΔsidI virulence was conducted by measuring survival in a neutropenic model of invasive pulmonary aspergillosis as previously described (5). At an infectious dose of 5 × 10⁵ spores, the mortality associated with ΔsidH and ΔsidI infections was lower (71% and 80%, respectively) than with WT infections (96%); however, significant attenuation was not reproducibly measurable at the level of murine survival. Further titration of the infectious dose to 9 × 10⁴ spores increased the attenuation of virulence in both the ΔsidH (n = 14) and ΔsidI mutants (n = 15) relative to WT (n = 8) (P = 0.0003 and P = 0.0008 respectively) (Fig. 5). Virulence also was attenuated significantly in the ΔsidH and ΔsidI mutants (P = 0.0032 and P = 0.0008, respectively) relative to the corresponding reconstituted strains sidH^C (n = 7) and sidI^C (n = 8).

Fig. 5.

In neutropenic mice virulence in ΔsidH and ΔsidI siderophore mutants is attenuated relative to isogenic WT (ATCC46645) and reconstituted strains. Inocula of 9 × 10⁴ conidiospores were administered intranasally under anesthesia in 40 μL of saline. Mice were weighed twice daily after infection, and the experimental end point was determined by a weight loss of 20% relative to the day of infection. sidH^C and sidI^C indicate reconstituted strains. Murine survival was compared between groups using Kaplan–Meier plots and log rank analysis.

Genetic Inactivation of TAFC Biosynthesis Decreases Lovastatin Susceptibility.

Inactivation of TAFC biosynthesis by the deletion of sidI or sidH increased resistance to lovastatin (8 μM) in iron-depleted conditions (Fig. 6). Accordingly, the ergosterol content of ΔsidI and ΔsidH increased about 13% during iron depletion (Fig. 2). These results suggest that inactivation of SidI or SidH increases the mevalonate pool for essential isoprenoid biosynthesis, including ergosterol, by blocking TAFC-mediated mevalonate consumption. In contrast, TAFC biosynthesis is not essential, because A. fumigatus possesses the alternative RIA system. Consistent with TAFC biosynthesis and the expression of involved enzymes being repressed by iron, ΔsidI and ΔsidH displayed WT-like susceptibility to lovastatin and ergosterol content in iron-replete conditions.

Fig. 6.

Deficiency in SidI or SidH increases resistance to lovastatin. Approximately 50 conidia were spotted on minimal medium agar without or with iron containing 0–8 μM lovastatin. Radial growth was measured after 48 h of incubation. The values represent the means of three experiments normalized to the untreated WT. The SD was <5%.

Iron Depletion Reduces Cellular Ergosterol Content in WT but Less in TAFC Nonsynthesizing Strains.

Collectively, the data shown above suggest a model in which bilateral demand for mevalonate occurs under iron-depleted conditions, imposed by the requirement for siderophore biosynthesis. Therefore, we examined in more detail cellular sterol content in WT and siderophore biosynthetic mutants under iron-replete and iron-starvation conditions (Fig. 7). Comparative GC-MS analysis of the neutral lipid fraction obtained from a WT strain, the siderophore-deficient l-ornithine-N⁵-monooxygenase mutant, ΔsidA (3), the TAFC-nonsynthesizing mutants, ΔsidF, ΔsidH, ΔsidI, and ΔsidD (5), and a ferricrocin-nonsynthesizing mutant ΔsidC (5) revealed iron-dependent alterations in sterol composition (Fig. 7). In all strains the major sterol was ergosterol; 15 additional sterol constituents were identified by comparison with known sterol spectra (SI Appendix, Table S1).

Fig. 7.

Ergosterol biosynthesis in A. fumigatus. Accumulated intermediates are represented in bar plots as the percentage of total sterols. (A) Ergosterol. Asterisks indicate significantly different sterol content (P < 0.05 by nonparametric Mann–Whitney test). (B) Accumulation of (i) C-4 methyl sterols and (ii) Δ7 sterols according to current understanding of the ergosterol biosynthetic pathway. Fe represents iron-replete (+) or iron-depleted (−) conditions.

In A. fumigatus ergosterol biosynthesis involves several iron-dependent enzymes, which operate at different levels of the pathway. To permit an appraisal of sterol intermediates within the context of potential iron-dependent rate limitations, we partitioned the analyzed sterols into two groups, according to the positioning of heme-dependent enzymes [lanosterol 14-α-demethylase (Erg11) and sterol C-22 desaturase (Erg5)] or oxo-diiron enzymes [methyl sterol oxidase (Erg25) and sterol C-5 desaturase (Erg3)] in the sterol biosynthetic pathway (20). In our classification the category of C-4 methyl sterols, derived from lanosterol via eburicol and sequentially demethylated at C-14 and C-4 by Erg11, Erg24, Erg25, Erg26, and Erg27, includes five di-tritetramethylcholesta-8,24-dien-3β-ol compounds: 4,4,14-trimethylcholesta-8,24-dien-3β-ol, 4α,24-dimethylcholesta-8,24(28)-dien-3β-ol, 4,4,14,24-tetramethylcholesta-8,24(28)-dien-3β-ol, 4α,24-dimethylcholesta-7,24(28)-dien-3β-ol, and 4,4,24-trimethylcholesta-8,24(28)-dien-3β-ol. The category of Δ7 sterols, which are desaturated at carbon 5 by the Erg3 enzyme to produce 24-methylcholesta-5,7,24(28)-trien-3β-ol sterol, includes three 24-methylcholesta-Δ7–3β-ol isomers: 24-methylcholesta-7,22-dien-3β-ol, 24-methylcholesta-7,24(28)-dien-3β-ol, and 4α,24-dimethylcholesta-7,24(28)-dien-3β-ol.

Iron limitation negatively impacted the ergosterol content of the cell in all strains tested (Fig. 7A). The observed reduction in ergosterol content was greater for the WT and ΔsidC mutants than for TAFC-nonsynthesizing mutants ΔsidA, ΔsidD, ΔsidF, ΔsidH, and ΔsidI. C-4 methyl and Δ7 sterols accumulated (Fig. 7B) under iron starvation in the WT and ΔsidC strains but drastically less in the TAFC-nonsynthesizing mutants ΔsidA, ΔsidD, ΔsidF, ΔsidH, and ΔsidI.

Iron Starvation Increases A. fumigatus Susceptibility to Voriconazole and Amphotericin B.

We next tested the effect of iron availability or the inactivation of siderophore biosynthesis on susceptibility to antifungal drugs that target ergosterol biosynthesis (Table 2). BPS-mediated iron chelation heightened the susceptibility of the WT strain and the FC-lacking ΔsidC mutant to both voriconazole and amphotericin B. The susceptibility of ΔsidI and ΔsidH cannot be tested under this condition, because their growth is inhibited by BPS (Fig. 3). The WT strain displayed the same susceptibility to voriconazole and amphotericin B in media containing 10 μM iron and in low-iron conditions. Inactivation of the entire siderophore system (ΔsidA), however, increased the susceptibility to both drugs, in particular during low-iron conditions.

Table 2.

MIC of voriconazole and amphotericin B for WT and siderophore-deficient mutants independence of iron availability

	MIC voriconazole (μg/mL)			MIC amphotericin B (μg/mL)
Strain	Fe–BPS*	Fe–	Fe+	Fe–BPS	Fe–	Fe+
ΔsidA	No growth	0.064	0.125	No growth	0.19	0.50
ΔsidI	No growth	0.125	0.125	No growth	1.00	1.00
ΔsidH	No growth	0.125	0.125	No growth	1.00	1.00
ΔsidC	0.032	0.125	0.125	0.38	1.00	1.00
WT	0.064	0.125	0.125	0.50	1.00	1.00

*Fe^–, Fe⁺, and Fe^–BPS are as described in Fig. 3.

Discussion

Differences in iron handling between mammals and siderophore-producing fungi such as Aspergilli might serve to improve the efficacy of antifungal therapies. Several lines of evidence indicate that A. fumigatus faces iron limitation during mammalian infection: (i) Defects impairing adaptation to iron limitation, such as inactivation of the siderophore system or the iron regulator HapX, attenuate virulence in murine models of pulmonary aspergillosis (3, 5, 21); (ii) RIA and the siderophore system are transcriptionally up-regulated during initiation of murine infection (22); and (iii) TAFC-chelated ⁶⁸Ga, the uptake of which is induced by iron starvation, allows in vivo PET imaging of invasive pulmonary aspergillosis in rats (23). Consistent with these findings, human protection against A. fumigatus includes growth inhibition by polymorphonuclear leukocytes via lactoferrin-mediated iron depletion and possibly siderocalin-mediated scavenging of siderophores (24, 25). Moreover, increased bone marrow iron stores represent an independent risk factor for invasive aspergillosis (26). Inversely, iron chelators were shown to improve treatment with azoles and amphotericin B in vitro and in vivo (27, 28).

Siderophore production in A. fumigatus follows an array of different enzymatic steps (Fig. 1), five of which have been identified recently at the gene level (5). In this study, we identified two additional TAFC biosynthetic genes, sidI and sidH, encoding proteins with similarity to acyl-CoA ligases and enoyl-CoA hydratases, respectively. Involvement of such enzymes in the biosynthesis of fusarinine-type siderophores was proposed initially by Anke and Diekmann (9) working with Fusarium cubense. Previously, mevalonate was thought to be converted to anhydromevalonate before ligation of CoA (9). However, later studies on the substrate stereochemistry of enoyl-CoA hydratases concluded that the CoA derivative (having acidic hydrogens at C-2 favorable for syn elimination), and not mevalonate itself, should be the actual substrate (29). Therefore, it is most likely that the acyl-CoA ligase SidI converts mevalonate to mevalonyl-CoA and that the enoyl-CoA hydratase SidH converts mevalonyl-CoA to anhydromevalonyl-CoA, which is transferred to N⁵-hydroxyornithine by the transacylase SidF (Fig. 1). All known fungal siderophores are derived from ornithine as a precursor (1). For fusarinine-type and coprogen-type siderophores, which are produced by a wide array of fungal species, mevalonate is a second precursor; here we characterize the genes encoding the enzymatic steps for formation of the mevalonate-derived anhydromevalonyl-CoA.

Elimination of sidI or sidH results in the loss of TAFC production, leaving the fungus relying solely on RIA. Apart from impairing TAFC biosynthesis, deficiency in sidH and in particular in sidI caused increased susceptibility to iron starvation and to oxidative stress, likely reflecting iron starvation because iron is, as heme, a cofactor of oxidative stress-detoxifying enzymes such as catalases or peroxidases. The phenotypic differences between mutant strains deficient in sidH and sidI might be caused by the accumulation of different pathway intermediates (mevalonate and mevalonyl-CoA, respectively) with different consequences. In further agreement, deficiency in sidI, but not in sidH, sidD, or sidF (5), decreased the intracellular FC content, possibly by disturbing the acetyl-CoA homeostasis required for FC biosynthesis. The hypersusceptibility to hydrogen peroxide of ΔsidI compared with ΔsidH might be related to the concomitant decrease in FC, which plays a role in resistance to oxidative stress (5).

The Ustilago maydis siderophore ferrichrome A requires another isoprenoid pathway intermediate (30): HMG-CoA is converted to methylglutaconyl-CoA by the enoyl-CoA hydratase Fer4 and then is used by the acylase Fer5 to acylate N⁵-hydroxyornithine that is generated from ornithine by the monoxygenase Sid1. Phylogenetic analysis indicates that A. fumigatus SidH and U. maydis Fer4 are not orthologs (SI Appendix, Fig. S5), suggesting recruitment of different enoyl-CoA hydratases in the evolution of siderophore biosynthesis in these two fungi.

Mevalonate is an important intermediate of isoprenoid biosynthetic pathways such as that of fungal ergosterol and essential nonsterol isopreonoids (31). Production of mevalonate is tightly controlled to satisfy cellular requirements for all these products but to avoid accumulation of these products to toxic levels (32, 33). Like S. cerevisiae, A. fumigatus possesses two putative isoforms of HMG-CoA reductase encoded by hmg1 (AFUA_2G03700) and hmg2 (AFUA_1G11230) (34). In S. cerevisiae, Hmg1 and Hmg2 contribute 83% and 17% of the total activity, respectively (35). In A. fumigatus, the individual contribution of the two putative isoforms is unknown. In Northern blot analyses, we detected only hmg1, but not hmg2, expression (SI Appendix, Fig. S1A). For isoprenoid biosynthesis, mevalonate is phosphorylated by the mevalonate-kinase Erg12 (Fig. 1).

Several lines of evidence indicate a tight interconnection of A. fumigatus TAFC and isoprenoid metabolism by the common intermediate mevalonate: (i) Iron starvation down-regulates the cellular ergosterol content but up-regulates TAFC production; (ii) in accordance with the ergosterol content, iron starvation down-regulates the erg12 transcript level; (iii) in accordance with TAFC production, iron starvation up-regulates the hmg1 transcript level; (iv) inhibition of HMG-CoA reductase by lovastatin treatment decreases TAFC production; (v) overexpression of hmg1 increases TAFC production, lovastatin resistance, and ergosterol content; (vi) deletion of sidI or sidH (blocking TAFC biosynthesis-mediated mevalonate consumption) increases lovastatin resistance and ergosterol content; and (vii) iron starvation and in particular genetic inactivation of siderophore biosynthesis increases the susceptibility to the ergosterol-targeting drugs voriconazole and amphotericin B.

Down-regulation of the cellular ergosterol content by iron starvation, as shown here in A. fumigatus (Fig. 7A) and previously for the siderophore-lacking yeasts C. albicans and S. cerevisiae (20, 36), is likely a consequence of the iron dependence of several enzymes of the late ergosterol biosynthetic pathway, e.g., Erg11/Cyp51, Erg3, and Erg5 (20, 37). In A. fumigatus, iron starvation results additionally in accumulation of C-4 methyl and Δ7 sterols, especially in TAFC-nonsynthesizing strains (Fig. 7B). Accumulation of Δ7 sterols also is caused by inactivation of the iron-dependent Erg3 in A. fumigatus (38), indicating that iron starvation indeed limits Erg3 activity. The increased levels of C-4 sterol intermediates during iron starvation might be attributable to decreased carbon-4 demethylation, most likely because of the reduction of Erg25 activity (Fig. 7B). Inactivation of TAFC biosynthesis increased the ergosterol content and reduced levels of C-4 methyl and Δ7 sterol intermediates during iron starvation, indicating that the sterol composition reflects the impact of iron and mevalonate availability, both of which are affected by TAFC production. Consistent with an increased requirement for mevalonate during iron-depleted conditions for TAFC synthesis, lovastatin susceptibility was higher in iron-depleted than in iron-replete conditions [minimal inhibitory concentration (MIC) of 8 μM vs. 10 μM)] and increased the resistance of mutants deficient in SidI or SidH (Fig. 6).

The polyene amphotericin B, azoles, and echinocandins are the drugs of choice for the treatment of aspergillosis. However, associated toxicities and development of resistances demand new antimycotic agents (39, 40). Statins are fungicidal for A. fumigatus (41) but, although they act synergistically with azoles against S. cerevisiae, C. albicans, C. neoformans, and clinically important Zygomycetes spp (42–44), they seem not to do so for A. fumigatus (45, 46). So far, the inhibitory effect of statins has been attributed solely to the inhibition of isoprenoid biosynthesis, which mainly affects sterol biosynthesis but also affects the synthesis of dolichol, heme-A, isopentenyl tRNA, carotenoids, and ubiquinone as well as protein prenylation (31). We show here that statins also inhibit biosynthesis of mevalonate-derived siderophores, which are essential virulence determinants of animal- and plant-pathogenic fungi (5, 47). Our data indicate that antifungal drug design and testing should consider the metabolic changes caused by iron starvation. Given the importance of the siderophore and ergosterol biosynthetic pathways for fungal virulence and antifungal treatment, the molecular insights provided by our analyses may open new possibilities in the fight against fungal infections.

Material and Methods

Fungal Strains and Growth Conditions.

A. fumigatus strains used in this study are listed in SI Appendix, Table S2. The strains were cultured routinely at 37 °C on minimal medium described previously (48) containing 1% glucose as the carbon source and 20 mM glutamine as the nitrogen source. Iron-replete medium contained 10 μM FeSO₄. For high-iron conditions, FeSO₄ was added to a final concentration of 1.5 mM FeSO₄; addition of iron was omitted for iron-depleted medium. Xylose instead of glucose was used as the carbon source for induction of expression of hmg1 under the control of the xylP promoter (17, 18). For radial growth assays, 50 conidia of respective strains were spotted onto described solid minimal medium. Lovastatin (Alexis) was converted to its active form by hydrolyzing in ethanolic NaOH as described previously (49). The plates were incubated at 37 °C, and radial growth was measured after 48 h. For propagation of plasmids, Escherichia coli strain DH5α was grown in LB medium supplemented with ampicillin (100 μg/mL).

Phylogenetic Analysis of SidI and SidH.

The amino acid sequences of SidI and SidH homologs were retrieved by BLAST search. The alignment was created with the multiple alignment algorithm of CLUSTAL W version 1.83. For phylogenetic analysis Phylip Drawgram version 3.6a3 (http://bioweb2.pasteur.fr/phylogeny) was used.

Deletion of A. fumigatus SidI and SidH.

The sidI and sidH genes were deleted using the bipartite marker technique as described in ref. 50 (SI Appendix, Figs. S2A and S6A). Flanking regions of sidI were amplified by PCR. A 1.8-kb fragment at the 5′ region was amplified with primers sidI-1 and sidI-2 (SI Appendix, Table 3) containing add-on restriction sites for ClaI and NotI, respectively, and then was cleaved with NotI. The 1.4-kb flanking region at the 3′ end was amplified with primers sidI-3 and sidI-4 containing add-on BamHI and StuI restriction sites, respectively, and then was cleaved with BamHI. Subsequently, the NotI-BamHI 3kb-fragment from pHYTK containing the hygromycin resistance-thymidine kinase gene (hph-Tk) cassette was ligated with the sidI 5′ and 3′ flanking regions, respectively. Each ligation product then was used as template for the amplification of the bipartite hph-Tk marker fragment and flanking region using sidI-5 and hph-14 for the 5′ flanking region plus partial hph-Tk and sidI-6 and hph-15 for 3′ flanking region plus partial hph-Tk.

To generate ΔsidH, the 5′ flanking region of sidH (1.7 kb) was amplified using sidH-1 and sidH-4 containing add-on restriction sites for NotI, and the 3′ flanking region (1.3 kb) was amplified using sidH-2 and sidH-3 containing add-on restriction sites for BamHI. Subsequently, the NotI–BamHI fragment of hph-Tk released from pHYTK was ligated with the sidH 5′ and 3′ flanking regions, respectively. Each ligation product then was used as template for the amplification of the bipartite hph-Tk marker fragment and flanking region (as described for sidI) using sidH-5 and hph-14 for the 5′ flanking region plus partial hph-Tk and sidH-6 and hph-15 for the 3′ flanking region plus partial hph-Tk.

Bipartite marker constructs for sidI and sidH were used to transform WT A. fumigatus. Transformation was done essentially as described by Punt et al. (51). The transformants were selected on medium containing 200 μg/mL of hygromycin B. Homologous recombination at the desired locus was confirmed by Southern blot analysis (SI Appendix, Figs. S2A and S6A).

Complementation of the SidI and SidH Deletions.

A 3.3-kb fragment containing the sidI WT copy was amplified from genomic DNA using primers sidI-5 and sidI-6 (SI Appendix, Table 3). After cleavage with XbaI-Acc65I, this fragment was inserted at the respective site in the multiple cloning site of pUC18, giving rise to pUC-CoA. A 2.7-kb SspI-HindIII fragment of the pyrithiamine resistance gene (ptrA) from plasmid pSK275 was inserted into SspI/HindIII-digested pUC-CoA to create pSY06. This construct was introduced into the ΔsidI strain by protoplast transformation. Transformants were selected using 0.1 mg/l pyrithiamine. To confirm homologous integration, a Southern blot of XbaI-digested genomic DNA from the complemented strain was probed with exon 2–3 (SI Appendix, Fig. S2B).

To reconstitute the ΔsidH strain with a functional sidH copy, a 7.9-kb EcoRI-HindIII fragment released from the cosmid pANsCosSidD was subcloned into plasmid pPhleo carrying the phleomycin resistance marker gene (ble) (52). The resulting 14.2-kb plasmid psidH^c was linearized with HindIII and used to transform A. fumigatus ΔsidH. Phleomycin-resistant transformants with single homologous reconstitution of the sidH gene (sidH^c strain) were selected by Southern blot analysis (SI Appendix, Fig. S6B).

Overexpression of hmg1.

A 1.6-kb fragment encoding the N-terminal region of the hmg1 gene (the entire gene is 3.6 kb) was amplified from A. fumigatus genomic DNA using oligonucleotides hmg-3 and hmg-4 to create an NcoI site at the hmg1 start codon and a SphI site 1.6 kb downstream from the start codon. The NcoI/SphI-digested amplified fragment was inserted in the respective sites of pGEM(5zf+), yielding plasmid pHMG-GEM. Next, the 1.7-kb fragment of the xylP promoter was amplified using plasmid pXyluidA-P (17) as template and oligonucleotides xylP-R and xylP-F, introducing an NcoI site in the xylP start codon and an NotI site at bp −1674. After digestion with NcoI/NotI, the fragment was inserted into the equally cleaved pHMG-GEM plasmid, resulting in pHMG-xyl. Then, the ptrA-containing the 1.9-kb PstI-NdeI fragment from pSK275 was inserted into the respective site of pHMG-xyl. The resulting plasmid pHMG was used to transform A. fumigatus WT as described in SI Appendix, Fig. S2C.

Analysis of Siderophores and Ergosterol.

TAFC and FC were characterized by reversed-phase HPLC chromatography as described previously (53).

Sterols were extracted following 24-h growth in liquid minimal medium under iron-depleted or iron-replete conditions as previously described (54). To 100–200 mg of freeze-dried mycelia, 3 mL of 25% alcoholic (3:2 methanol:ethanol) potassium hydroxide solution was added. After vortexing for 1 min, the mixture was incubated at 85 °C for 1 h. Sterols were extracted twice with a mixture of 1 mL and 1.5 mL hexane and, after vigorous vortex mixing, the upper hexane layer was transferred to a clean glass tube and evaporated. Samples were spiked with 0.1 mg of androstanol (Sigma) as an internal standard for recovery and quantitation. Trimethylsilyl ether derivatives of fungal sterols dissolved in toluene/N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA; 1:1; Merck) were analyzed on a Trace GC-DSQ-II single-quadrupole mass spectrometer (Thermo Scientific) on a 60 m × 0.25 mm i.d.× 0.25 μm film thickness 5% diphenyl- and 95% dimethyl-polysiloxane (HP-5MS; Agilent) capillary column using He as carrier gas at 1.2 mL/min. Injector, transfer line, and ion source were at 300, 200, and 250 °C, respectively. Sterols were identified by comparison of mass spectra and retention time data (54). Relative sterol composition was expressed as the ratio of raw peak areas with respect to the total sterol content. Each experiment was repeated three times. The significance of content (P < 0.05) was determined by nonparametric Mann–Whitney test.

Northern Blot Analyses.

Total RNA was isolated from mycelia by using RNA-Solv reagent as described by the manufacturer (Omega Bio-Tek). For the Northern analysis 10 μg of total RNA was used. RNA was fractionated in formaldehyde-agarose gels, blotted onto Hybond-N membranes (Amersham Biosciences), and hybridized with digoxigenin-labeled probes (Boehringer Mannheim). Hybridization probes were generated by PCR amplification using oligonucleotides listed in SI Appendix, Table S3.

Analysis of Fungal Burden in Murine Lungs.

Virulence of A. fumigatus strains was assayed in a murine model for invasive pulmonary aspergillosis described previously (55). Briefly, outbred male mice (strain CD1, 20–28 g; Charles River Laboratories) were used for animal experiments. Immunosuppression was carried out by s.c. injection of 112 mg/kg hydrocortisone acetate and i.p. injection of 150 mg/kg cyclophosphamide as described previously (55). Bacterial infections were prevented by adding 1 g/L tetracycline and 64 mg/L ciproxicin to the drinking water. Inocula of 5 × 10⁵ or 9 × 10⁴ conidiospores in 40 μL of saline were prepared by harvesting spores from 5-d-old slices of solid medium followed by filtration through Miracloth (Calbiochem) and washing with saline. Mice were anesthetized by inhalation of isofluorane and infected by intranasal instillation. Infected mice were weighed twice daily after infection. The end point for survival analyses was determined by a weight loss of 20% relative to the day of infection. Murine survival was compared between groups using Kaplan–Meier plots and log rank analysis.

All work using animals in this study was carried out pursuant to Project License PPL/70/6487 authorized by the Home Office pursuant to the Animals (Scientific Procedures) Act 1986 (widely acknowledged to be the most rigorous legislation of its type in the world). Details of the relevant legislation and its application can be found at: http://www.homeoffice.gov.uk/science-research/animal-research/. All infections were performed under isofluorane anesthesia, and all efforts were made to minimize suffering.

Drug Susceptibility Testing.

E-test strips (Biomerieux) impregnated with a gradient of voriconazole or amphotericin B were used according to the manufacturer's instructions. Each strip was placed onto a minimal medium agar plate with different iron availability containing a lawn of conidia. Growth inhibition was measured after 48 h by direct observation of the plates at 37 °C.

Accession Numbers.

The genes and gene products (including accession numbers at http://www.cadre-genomes.org.uk/) studied in this work are hmg1 (AFUA_2G03700), hmg2 (AFUA_1G11230), erg12 (AFUA_4G07780), sidI (AFUA_1G17190), sidH (AFUA_3G03410), sidF (AFUA_3G03400), sidD (AFUA_3G034290), sidA (AFUA_2G07680), and sidC AFUA_1G17200).