Correspondence: Uffe H. Mortensen, Department of Systems Biology, Center for Microbial Biotechnology, Technical University of Denmark, Building 223, 2800 Kgs. Lyngby, Denmark. Tel.: +45 4525 2701; fax: +45 4588 4148; e-mail: firstname.lastname@example.org
Fungi possess an advanced secondary metabolism that is regulated and coordinated in a complex manner depending on environmental challenges. To understand this complexity, a holistic approach is necessary. We initiated such an analysis in the important model fungus Aspergillus nidulans by systematically deleting all 32 individual genes encoding polyketide synthases. Wild-type and all mutant strains were challenged on different complex media to provoke induction of the secondary metabolism. Screening of the mutant library revealed direct genetic links to two austinol meroterpenoids and expanded the current understanding of the biosynthetic pathways leading to arugosins and violaceols. We expect that the library will be an important resource towards a systemic understanding of polyketide production in A. nidulans.
It is well known that filamentous fungi produce a large number of bioactive secondary metabolites (Bérdy, 2005; Cox, 2007; Newman & Cragg, 2007). Polyketide (PK) compounds constitute a major part of these metabolites and have long been recognized as a valuable source of diverse natural compounds of medical importance, for example lovastatin (cholesterol lowering) (Lai et al., 2005), griseofulvin (antibiotic) (Chooi et al., 2010) and mycophenolic acid (immunosuppressant) (Bentley, 2000). However, polyketides also include many toxic compounds that pose a serious threat to human health, for example patulin, ochratoxins, fumonisins and aflatoxin (Frisvad et al., 2004; Månsson et al., 2010). Polyketides are biosynthesized by large multidomain polyketide synthases (PKSs), which besides acyl transferase, β-ketoacyl synthase and acyl carrier domains may also contain keto reductase, dehydratase, cyclization and methyl-transferase domains (Cox, 2007; Smith & Tsai, 2007; Hertweck, 2009). In fungi, the different catalytic activities often work in an iterative manner (fungal type I) and it is generally difficult to predict the exact product formed by a given PKS from its sequence alone (Keller et al., 2005). Product prediction is further complicated by the fact that the resulting polyketide structure may be decorated by tailoring enzymes. Such genes are often physically associated with the PKS gene in a gene cluster allowing for coordinated regulation (Schümann & Hertweck, 2006). The fact that natural products may be of mixed biosynthetic origin, combining elements such as polyketides with terpenes (meroterpenoids) and/or nonribosomal peptide units, adds to the complexity (Chang et al., 2009; Geris & Simpson, 2009; Hertweck, 2009; Scherlach et al., 2010).
As a consequence of their bioactivity, societal importance and also the prospect of reprogramming the biosynthetic machinery for drug development (Cox, 2007), there is tremendous interest in the discovery and understanding of fungal polyketide biosynthesis. The availability of full genome sequences of a number of filamentous fungi has provided a means to address the discovery of polyketides because the PKS genes are large and contain several conserved protein domains. Importantly, analysis of the genomic sequences from filamentous fungi (including Aspergillus nidulans, teleomorph, Emericella nidulans) predict numbers of individual PKS genes that exceeds significantly the number of polyketides that these fungi are known to produce (Galagan et al., 2005). In fact, the genome of A. nidulans (Galagan et al., 2005) appears to contain as many as 32 individual PKS genes (Nierman et al., 2005; Szewczyk et al., 2008; von Döhren, 2009), but until now only nine genes have been linked to eight polyketides (Yamazaki & Maebayas, 1982; Bergmann et al., 2007; Chiang et al., 2008; Szewczyk et al., 2008; Bok et al., 2009; Chiang et al., 2009; Schroeckh et al., 2009) (see Supporting Information, Fig. S1). This may reflect that many polyketides are either produced in small amounts, under special conditions, or in developmental stages that are rarely observed under laboratory conditions.
Towards a more complete genetic mapping of the secondary metabolism in A. nidulans, we first cultivated a reference strain on an array of different growth media to uncover polyketides that were not previously linked to a gene cluster. This analysis revealed several compounds, including austinols, violaceols, arugosins and prenylated xanthones. Next, genetic links to these compounds were established by constructing and screening an A. nidulans mutant library containing individual deletions of 32 putative PKS genes.
Materials and methods
The A. nidulans strain IBT29539 (argB2, pyrG89, veA1, nkuAΔ) (Nielsen et al., 2008) was used as the reference strain and for deletion-strain constructions. Escherichia coli strain DH5α was used for cloning.
Fungal minimal medium (MM) was as described in Cove (1966), but with 1% glucose, 10 mM NaNO3 and 2% agar. Medium for alcA promoter induction consisted of MM supplemented with 100 mM l-threonine and 100 mM glycerol as carbon source instead of 1% glucose. Polyketide screening media variants CYA, CYAs, YES and RT were prepared as per Frisvad & Samson (2004). CY20 medium consisted of CYA with 170 g sucrose; RTO contained RT with 30 g organic oat meal; and YE was made as YES but without sucrose. All media variants were supplemented with 10 mM uridine, 10 mM uracil and/or 4 mM l-arginine where appropriate.
Construction of A. nidulans gene deletion library
Individual PKS gene deletions were carried out as described previously (Nielsen et al., 2006), except that the targeting fragments were assembled using Gateway technology (Hartley et al., 2000) (see Table S1 for PCR oligonucleotide and Fig. S2 for an overview of the procedure). The A. nidulans transformants were streak purified and rigorously screened through three complementing diagnostic PCRs. Subsequently, the Aspergillus fumigatus pyrG marker was eliminated from all strains by selecting on 5-fluoroorotic acid medium before final verification by two additional complementing diagnostic PCRs (see Fig. S3 and Table S2). All strains have been deposited in the IBT strain collection, DTU, (http://www.fbd.dtu.dk/straincollection/).
Construction of the ausA-S1660A strain
The amino acid substitution of serine to alanine, S1660A, in ausA (AN8383) was created by USER fusion (Geu-Flores et al., 2007) according to the method described by Hansen et al. (2011). The allele was transferred to IBT29539 and the pyrG marker was eliminated by direct repeat recombination, creating strain IBT31032 containing only the desired point mutation. The strain was verified to contain the ausA-S1660A allele by sequencing (StarSEQ, Germany). See Table S3 for primer details.
Ectopic integration of ausA into IS1
The gene, ausA, was PCR amplified by USER fusion (Geu-Flores et al., 2007) and inserted into both pU1111-1 and pU1211-1 (Hansen et al., 2011) creating pDH23 (gpdA promoter) and pDH24 (alcA promoter), respectively. For both plasmids, the gene-targeting substrate was excised by NotI digestion and transformed into IBT28738 using A. nidulans argB as a selectable marker. Transformants were streak purified and verified for correct integration into the IS1 site (Hansen et al., 2011) by two complementing diagnostic PCRs.
Chemical analysis of the mutants
Strains were inoculated as three point stabs on solid media and incubated for 7 days at 37 °C in the dark. Metabolite extraction was performed according to the micro extraction procedure (Smedsgaard, 1997). Extracts were analyzed by two methods: (1) Ultra-high performance liquid chromatography-diode array detection (UHPLC-DAD) analyses using a Dionex RSLC Ultimate 3000 (Dionex, Sunnyvale, CA) equipped with a diode-array detector. Separation of 1 μL extract was obtained on a Kinetex C18 column (150 × 2.1 mm, 2.6 μm; Phenomenex, Torrence, CA) at 60 °C using a linear water–acetonitrile gradient starting from 15% CH3CN to 100% (50 ppm trifluoroacetic acid) over 7 min at a flow rate of 0.8 mL min−1. (2) Exact mass, HPLC-DAD-HRMS, was performed on a 5 cm × 3 μm, Luna C18(2) column (Phenomenex) using a water–acetonitrile gradient from 15% CH3CN to 100% over 20 min (20 mM formic acid). LC-DAD-MS analysis was performed on a LCT oaTOF mass spectrometer (Micromass, Manchester, UK) as in Nielsen & Smedsgaard (2003) and Nielsen et al. (2009).
Isolation and structure elucidation of secondary metabolites
3,5-Dimethylorsellinic acid and dehydroaustinol were purified from large-scale ethyl acetate extracts prepared from 100 MM agar plates after 4 days' cultivation in darkness at 37 °C. The compounds were purified using a 10 × 250 mm Phenomenex pentafluorophenyl column (5 μm particles) with a water–acetonitrile gradient from 15% to 100% CH3CN in 20 min using a flow of 5 mL min−1. Arugosin A was isolated from an ethyl acetate extract of the reference strain grown on 200 CYAs agar plates using a Waters 19 × 300 mm C18 Delta Pak column (15 μm particles), gradient from 80% to 90% CH3CN in 10 min, and a flow of 30 mL min−1. The NMR spectra were acquired on a Varian Unity Inova 500 MHz spectrometer using standard pulse sequences. Additional details about the compound identification can be found in the supporting information.
Results and discussion
Selecting media that support secondary metabolite production in A. nidulans
Construction and initial characterization of a genome-wide PKS gene deletion library
To investigate whether any of the compounds observed in Fig. 1 could be genetically linked to a PKS gene, we decided to take a global approach and individually deleted all 32 (putative and known) PKS genes in the A. nidulans genome (see Fig. S4), as defined from the annotation of the genome databases at the Broad Institute of Harvard and MIT and the Aspergillus Genome Database at Stanford (Arnaud et al., 2010). All genes were individually deleted by replacing the entire ORFs using gene-targeting substrates based on the pyrG marker from A. fumigatus for selection. Before analyzing the deletion mutant strains, the pyrG marker was excised by direct repeat recombination (Nielsen et al., 2006) in each case. This was carried out to ensure that the analyses of individual mutant strains were comparable to and not influenced by differences in the primary metabolism due to gene cluster-specific expression levels of the pyrG marker. All 32 deletion mutant strains (see Table S4) were viable and able to sporulate, showing that none of the 32 genes are essential for growth and that no polyketide product is essential for conidiation. As expected, the one strain carrying the wAΔ mutation formed white conidiospores as it fails to produce the naphthopyrone, YWA1, the precursor of green conidial pigment (Watanabe, 1998; Watanabe et al., 1999).
In addition to wA, eight additional PKS genes have previously been linked to metabolites. In our analysis, key compounds representing four of these gene clusters could be detected: monodictyphenone (1) (observed on RTO, YES and CY20), orsellinic acid (2) (observed on YES, CY20, RT, CYAs and CYA), emericellamide (A) (3) (observed on all media) and sterigmatocystin (4) (observed on RTO, CYAs and CYA). To verify the previously published gene links to these compounds, we individually compared the metabolic profiles of the reference strain to the corresponding profiles obtained with the single PKS gene deletion mutant strains. In agreement with previous analyses, these four compounds disappeared in mdpG (Bok et al., 2009), orsA (Schroeckh et al., 2009), easB (Chiang et al., 2008) and stcA (Yu & Leonard, 1995) deletion strains of our library (Fig. S5). Compounds resulting from the remaining four PKS genes were identified by activating the gene clusters by controlled expression of the transcription factor gene in the cluster (Bergmann et al., 2007; Chiang et al., 2009) or by deleting sumO that influences regulation of biological processes at many different levels (Szewczyk et al., 2008). Expression from these clusters is apparently not triggered by growth on any of our media, and natural conditions provoking their activation remain to be discovered.
Next, we performed a comparison of the metabolite profiles from the 32 deletion mutants with those obtained with the reference strain with the aim of uncovering novel genetic links between PKS genes and polyketides. The most significant changes are described below.
mdpG is involved in the biosynthesis of arugosins
First we focused our attention on the most prominent compound produced on RTO, YES, CY20 and RT media, which eluted as a broad peak around 7.2 min. This compound completely disappeared in the mdpGΔ strain (Fig. 2 and Fig. S6). Moreover, it had a characteristic UV spectrum in the UHPLC analysis indicating an arugosin-like metabolite. MS analysis indicated the compound to be arugosin A (m/z 425 a.m.u. for [M+H]+), which to our knowledge has not been reported before from A. nidulans. We therefore decided to confirm the structure of this compound (5). A large-scale extraction was performed and the metabolite was purified. The NMR data in dimethyl sulfoxide are in agreement with the literature (Kawahara et al., 1988) for the hemiacetal form of arugosin A except that the equilibrium was shifted completely to the open form (Fig. 3). In methanol, the NMR data showed that the compound exists in equilibrium between the closed and open ring form (data not shown), explaining the broad peak observed in Fig. 2. A minor peak could be assigned as a mono-prenylated arugosin as [M+H]+ at m/z 357 a.m.u. The MS data of this compound did not indicate loss of a prenyl moiety, suggesting that it is arugosin H (6), a likely immediate precursor of arugosin A (Fig. 3). Hence, our data show that mdpG, which is known for its role in formation of monodictyphenone, is also involved in formation of arugosins.
It is not unusual that one PKS gene cluster is responsible for the biosynthesis of a family of structurally similar compounds (Walsch, 2002; Kroken et al., 2003; Frisvad et al., 2004; Amoutzias et al., 2008). In the original analysis of the mdpG gene cluster, it was activated due to remodeling of the chromatin landscape, which occurs in a cclA deletion strain (Chiang et al., 2010). That study genetically linked the mdpG gene cluster to eight emodin analogues, including several aminated species, which were detected and tentatively identified. In our analyses, we also detected several emodins including 2-ω-dihydroxyemodin (7), ω-hydroxyemodin (8) and emodin (9), as well as the more apolar compounds emericellin (10), shamixanthone (11) and epi-shamixanthone (12) (Fig. 1 and Fig. S7). Like in the original study, all emodins disappear in our mpdGΔ strain.
Recently, it was demonstrated that the polyketide part of prenylated xanthones also could be coupled to mpdG (Sanchez et al., 2011). Our finding that mpdG is involved in formation of arugosins indicates that these compounds serve as intermediates in the conversion of monodictyphenone into xanthones, Fig. 3. In agreement with this, previous studies have reported arugosins to be precursors for emericellin (10) and shamixanthones (11) and (12) (Ahmed et al., 1992; Kralj et al., 2006; Márquez-Fernández et al., 2007), but have not established a genetic link to mpdG.
AN7903 and orsA are involved in violaceol production
Our reference strain produces the antibiotic violaceol I (13) and II (14), in significant amounts (Fig. 4 and Fig. S8). These two diphenyl ethers have been identified in Emericella violacea, Aspergillus sydowi and Aspergillus funiculosus (Taniguchi et al., 1978; Yamazaki & Maebayas, 1982) and recently also in A. nidulans (Nahlik et al., 2010). Our analysis now links the biosynthesis of the two violaceols to orsA as they disappear in our orsAΔ strain. It has previously been shown that orsA (AN7909) is involved in the formation of orsellinic acid (2), lecanoric acid (15), the two colored compounds F-9775A (16) and F-9775B (17), orcinol, diorcinol, gerfeldin and deoxy-gerfeldin. (Schroeckh et al., 2009; Sanchez et al., 2010). Our analysis confirms the link between orsellinic acid, lecanoric acid, diorcinol, F-9775A, F-9775B to orsA as these compounds are missing in the orsAΔ strain. However, we have not been able to detect the gerfeldins in any of our strains, and apparently our conditions favor violaceol and not gerfeldin formation.
The violaceols are formed by dimerization of two C7 monomers of 5-methylbenzene-1,2,3-triol, a compound that we could tentatively detect as [M-H]− at m/z 139 in cultivation extracts. The C7 backbone of 5-methylbenzene-1,2,3-triol, may conceivably be formed by decarboxylation of a C8 aldol intermediate as suggested by Turner 40 years ago (Turner, 1971) (Fig. 5). This C8 intermediate also serves as a branch point towards orsellinic acid.
Interestingly, the same compounds that disappear in the orsAΔ strain also disappear in AN7903Δ, a strain missing a PKS gene separated from orsA by only ∼20 kb (Fig. 4). This result does not contradict the original assignment of orsA as the PKS gene responsible for production of orsellinic acid. Although the enzymes encoded by the two genes are predicted to share many of the same functional domains, AN7903 is larger by around 500 amino acid residues and contains a methyl-transferase domain, which is not required for orsellinic acid production. Moreover, we note that Schroeckh et al. (2009) observed that both AN7903 and orsA were upregulated when orsellinic acid was induced by co-cultivation with Streptomyces hygroscopicus, indicating cross-talk between the two clusters. Surprisingly, what appear to be trace amounts of orsellinic acid can be detected as m/z 167 [M-H]− in both the AN7903Δ and the orsAΔ strains (Fig. 4). The source of this residual orsellinic acid remains elusive, but it could possibly stem from unmethylated byproducts from the PKS, AN8383, that produces 3,5-dimethylorsellinic acid, see below.
AN8383 is responsible for austinol and dehydroaustinol biosynthesis
Interestingly, production of austinol (18) and dehydroaustinol (19) was observed in the reference strain on several media (Fig. 1). Despite the fact that the production of these compounds is known from A. nidulans (Szewczyk et al., 2008), they have not yet been assigned to a specific gene. Only the AN8383Δ strain failed to produce the two austinols on all the media, which triggered austinol production in the reference strain (Fig. 6a). This, phenotype could be rescued by inserting the structural gene of AN8383 under the control of the gdpA promoter into an ectopic locus, IS1 (Hansen et al., 2011) (Fig. 6a). Moreover, a point mutant strain AN8383-S1660A also failed to produce austinols on these six media (Fig. 6a). In this strain, the DSL motif of the AN8383 PKS has been mutated to DAL, preventing the phosphopantetheine moiety of coenzyme A to attach to the acyl carrier protein domain of the PKS, thus disrupting polyketide synthesis (Evans et al., 2008). The lack of austinols can thus be linked directly to an AN8383-encoded function rather than to silencing of another gene caused by chromatin changes provoked by the AN8383 deletion.
To confirm the role of AN8383 in austinol production, we constructed a new strain that expresses the AN8383 ORF controlled by the inducible alcA promoter from the ectopic locus, IS1 (Hansen et al., 2011). On inductive medium, the subsequent LC-MS analysis showed a large novel peak eluting at ca. 6 min (see Fig. S9). The corresponding compound was purified and the structure was elucidated by NMR (Fig. S10), identifying 3,5-dimethylorsellinic acid (20), which is in good agreement with the route of synthesis previously proposed for austinol (Fig. 6b; Geris & Simpson, 2009). In a parallel analysis using a strain expressing AN8383 under the control of the constitutive gpdA promoter we obtained the same result (data not shown).
Together, the results strongly indicate that AN8383 encodes a PKS producing 3,5-dimethylorsellinic acid and that this compound serves as the first intermediate in the complex biosynthesis of austinol and dehydroaustinol that also involves a yet unidentified prenyl transferase(s). Based on these results, we have named the locus AN8383, ausA.
In the present study, we constructed a genome-wide PKS deletion library, which we screened for effects on polyketide production on a variety of media. This analysis has provided novel links between PKS genes and polyketide products demonstrating the strength of this approach. Many PKS genes still remain to be functionally connected to products, as many gene clusters have not yet been activated. As the repertoire of tools and methods to induce gene expression is rapidly increasing, new polyketide compounds will likely soon be uncovered in A. nidulans. To this end, the genome-wide PKS gene deletion library presented here will undoubtedly serve as a useful resource.
This work was supported by the Danish Research Agency for Technology and Production, grant # 09-064967. We thank Lisette Knoth-Nielsen for her dedicated and valuable technical assistance in the laboratory. In addition, we thank Rasmus John Normand Frandsen for suggestions and critical reading of the manuscript.