Introduction

The rice pathogenic ascomycete Fusarium fujikuroi is a member of the Asian clade of the Gibberella fujikuroi species complex closely related to Fusarium proliferatum (Nirenberg and O’Donnell 1998). It is the causative agent of the bakanae (“foolish seedling”) disease of rice. The most characteristic symptoms of this disease are excessively elongated seedlings with chlorotic stems and leaves due to the ability to produce and secrete gibberellins, a family of plant hormones (Bömke and Tudzynski 2009).

Besides the gibberellins, the fungus produces several other secondary metabolites (SM) including pigments, such as bikaverin and fusarubins (Wiemann et al. 2009; Studt et al. 2012) as well as harmful mycotoxins such as fusarins, fusaric acid, apicidin F, and fujikurins (Niehaus et al. 2013, 2014a, b; von Bargen et al. 2015). The recently sequenced genome comprises many more, so far uncharacterized, gene clusters showing a huge potential of the fungus to produce further secondary metabolites (Wiemann et al. 2013).

One of the gene clusters identified in F. fujikuroi revealed almost perfect homology to the fumonisin gene cluster (FUM) in the maize pathogens Fusarium verticillioides and F. proliferatum (Wiemann et al. 2013). The polyketide metabolites fumonisin B1 (FB1) and fumonisin B2 (FB2) were first isolated and identified in 1988 (Gelderblom et al. 1988). Their chemical structure is characterized by an aminopentol backbone with two tricarballylic acid side chains and one or two hydroxy groups (Bezuidenhout et al. 1988). Besides the fumonisins of the B series (FB1–4), several analogs have been found belonging to the three minor series of A fumonisins (N-acetyl, FAs), C fumonisins (dimethyl, FCs), and P fumonisins (N-3-hydroxypiridinium, FPs) in several Fusarium spp. (Gelderblom et al. 1988; Branham and Plattner 1993; Musser et al. 1996; Lazzaro et al. 2013) (Fig. 1a).

Fig. 1
figure 1

Chemical structure of different fumonisins and fumonisin gene cluster comparison. Produced fumonisins are divided into the B, A, C, and P series referring to different residues added to or missing at the basic aminopentol backbone (a). The fumonisin gene clusters of F. fujikuroi and F. verticillioides are highly conserved (b) (modified after Wiemann et al. 2013)

Full size image

Fumonisins, and particularly fumonisin B1 cause porcine pulmonary edema and equine leukoencephalomalacia in swine and horses, respectively, and are suggested to be potential risk factors for human esophageal cancer and even birth defects (Harrison et al. 1990; Kellerman et al. 1990; Desai et al. 2002; Marasas et al. 2004; Sun et al. 2007). Because of their structural similarity to sphingoid bases and fatty acyl-CoA, fumonisins of the B series inhibit the ceramide synthase leading to perturbation in sphingolipid metabolism which is important for the maintenance of membrane structures and the prevention of apoptosis (Schmelz et al. 1998; Merrill et al. 2001; Desai et al. 2002).

In general, fumonisin production, in particular of FB1 and FB2, has been mainly reported for the maize pathogens F. verticillioides and F. proliferatum, but also for some isolates of Fusarium oxysporum (Rheeder et al. 2002; Munkvold 2003) as well as Aspergillus niger and some related species (Frisvad et al. 2007; Varga et al. 2010). Only traces of N-acetylated fumonisins (FAs) were found in cultures of F. verticillioides and F. proliferatum isolates and contaminated maize samples (Lazzaro et al. 2013; Tamura et al. 2014). F. fujikuroi was shown to produce no or only very low amounts of fumonisins (Wiemann et al. 2013; Cruz et al. 2013).

The key enzyme of fumonisin biosynthesis in F. verticillioides was found to be the polyketide synthase (PKS) Fum1 (Proctor et al. 1999). Subsequent characterization of the surrounding genes resulted in the identification of a cluster consisting of 16 co-expressed genes (FVEG_00315 to FVEG_00329) encoding, beside the PKS, two fatty acid synthases, several monooxygenases, dehydrogenases, transporter proteins, an aminotransferase, and a dioxygenase (Butchko et al. 2006; Brown et al. 2007). Deletion of the genes encoding Fum1 (Proctor et al. 1999), the P450 monooxygenase Fum6, and the aminotransferase Fum8 (Seo et al. 2001) led to total loss of fumonisin production. In contrast, deletion of the genes FUM2, FUM3, FUM7, FUM10, FUM11, FUM13, FUM14, and FUM19 encoding a P450 monooxygenase, a dioxygenase, a dehydrogenase, a fatty acyl-CoA synthase, a tricarboxylate transporter, a short-chain dehydrogenase, a NRPS-like condensation domain-containing enzyme, and an ABC transporter, respectively, only resulted in a variation of the ratio between the end products FB1 and FB2 and their precursors fumonisin B3 (FB3) and fumonisin B4 (FB4), respectively, or slightly modified forms (Butchko et al. 2003a; Butchko et al. 2003b; Proctor et al. 2003; Proctor et al. 2006; Butchko et al. 2006). The homologs of almost all cluster genes are present and well conserved in F. fujikuroi (FFUJ_09240 to FFUJ_09255) except for FUM20 which is absent and FUM17 which is only present as a pseudogene. However, both genes are very unlikely to be involved in fumonisin biosynthesis of F. verticillioides (Proctor et al. 2003; Brown et al. 2005; Wiemann et al. 2013).

Regulation of fumonisin production has been studied mainly in F. verticillioides. Under alkaline conditions, fumonisin biosynthesis is repressed by the pH regulator Pac1 (Flaherty et al. 2003). In general, high nitrogen concentrations inhibit fumonisin production in F. verticillioides (Shim and Woloshuk 1999). The nitrogen-responsive regulator AreA was shown to be involved in this regulation and is essential for fumonisin production (Kim and Woloshuk 2008). Deletion of the Zn(II)2Cys6 transcription factor (TF) encoding gene ZFR1, which was shown to be an activator of fumonisin biosynthesis, indicated an involvement of sugar supply on fumonisin production (Flaherty and Woloshuk 2004; Bluhm et al. 2008). The gene encoding the putative sugar transporter Fst1 was found to be upregulated in the WT during fumonisin production on maize kernels but downregulated in Δzfr1 mutants. An independent deletion of FST1 also confirmed its role as positive regulator of fumonisin biosynthesis (Bluhm et al. 2008). Recently, the global regulator Sge1 was shown to act as a strong activator of the FUM cluster in F. fujikuroi (Michielse et al. 2015).

The cluster-specific Zn(II)2Cys6 TF Fum21 was identified by Brown et al. (2007). Its activity is essential for fumonisin biosynthesis: deletion of FUM21 resulted in low expression of FUM1 and FUM8 in the early stages of growth (Brown et al. 2007). Recently, Montis et al. (2013) identified a putative TF-binding motif (CGGMTA) in F. verticillioides which is relevant for transcriptional activation of FUM1 and is present in several fumonisin producers, but also non-producers (Montis et al. 2013). However, it was not shown which regulator might bind to this motif (Flaherty and Woloshuk 2004; Brown et al. 2007). In F. fujikuroi, this motif is not present in the promoter of FUM1.

In the present work, we gained more insight into the regulation of fumonisin biosynthetic genes in F. fujikuroi. We show that production levels of fumonisins are very low in F. fujikuroi compared to F. verticillioides despite the high homology of the cluster genes. However, constitutive expression of the cluster internal TF Fum21 strongly activated the fumonisin biosynthetic genes and led to 1000-fold elevated fumonisin yields. Similarly to F. verticillioides, fumonisin biosynthetic genes are subject to nitrogen metabolite repression in an AreA-dependent manner. In addition, we provide evidence that the second nitrogen-responsive GATA TF AreB is also essential for fumonisin gene expression and biosynthesis. To study the hierarchy between the impact of the cluster internal TF Fum21 and the two global regulators AreA and AreB, we over-expressed FUM21 in the background of the AREA and AREB deletion mutants. We show that the fumonisin biosynthesis was only partially restored in these double mutants, indicating that the activity of both GATA factors is essential for the action of Fum21.

Material and methods

Strains and cultivation conditions

The F. fujikuroi wild-type (WT) strain IMI58289 (Commonwealth Mycological Institute, Kew, UK), and the mutants ΔareA (Tudzynski et al. 1999) or ΔareB (Michielse et al. 2014), were used as parental strains for gene knock-out and over-expression experiments. For comparative analyses of SM production and SM gene expression, the WT strains of F. fujikuroi and F. verticillioides (strain M-3125, provided by D. Brown, U. S. Department of Agriculture) as well as the F. fujikuroi ΔareA and ΔareB mutants were cultivated in different conditions. For general maintenance, all strains were grown on solid complete medium (CM; Pontecorvo et al. 1953). To grow strains for genomic DNA isolation, a cellophane layer was placed on the surface of solidified CM and the colonies were grown for 2 to 5 days. For fungal transformation, gene expression, and metabolite analyses, the indicated strains were grown in liquid cultures. Two agar plugs from CM plates were pre-cultivated for 3 days in 300-mL Erlenmeyer flasks containing 100 mL Darken medium (Darken et al. 1959) at 28 °C and 180 rpm on a rotary shaker in darkness. As main culture, 100 mL of defined ICI medium (Imperial Chemical Industries Ltd., UK; Geissman et al. 1966) were prepared with different nitrogen sources and concentrations replacing the original nitrogen source, 6/60 mM L-glutamine (Gln), maize flour (2 g/L; Frießinger Mühle GmbH, Bad Wimpfen, Germany) or ground, lyophilized maize seedlings (2 g/L). The cultures were inoculated with 500 μL of the pre-culture and shaken for 3 to 7 days under conditions described above. For the germination of maize (Zea mays) seedlings, the kernels were surface sterilized (Glenn et al. 2008), incubated for 3 days at 28 °C in the dark on water agar (1.5 %) and incubated for an additional 3 days in the presence of a 12 h light–12 h dark cycle at room temperature. Prior to preparation of protoplasts of F. fujikuroi, 500 μL of the pre-culture were transferred to 100 mL ICI medium containing 10 g/L fructose (instead of glucose), and 0.5 g/L (NH4)2SO4 as nitrogen source. These cultures were incubated for no longer than 16 h.

The Saccharomyces cerevisiae strain FY834 (Winston et al. 1995) was used for recombinational cloning of vectors (Colot et al. 2006; Schumacher 2012), and the Escherichia coli strain TOP10F′ (Invitrogen, Groningen, The Netherlands) was used for plasmid propagation.

Standard molecular techniques

Total DNA from F. fujikuroi was isolated following the protocol of Cenis (1992) from lyophilized and ground mycelium. Plasmid DNA from S. cerevisiae and E. coli was isolated using the NucleoSpin® Plasmid Kit (Machery-Nagel, Düren, Germany). Large amounts of plasmid for fungal transformation were isolated from E. coli using the Nucleobond® Xtra Midi Kit (Machery-Nagel, Düren, Germany). For standard DNA amplification by polymerase chain reaction (PCR), BioTherm™ DNA Polymerase (GeneCraft, Lüdinghausen, Germany) was used. Large fragments were amplified using TaKaRa LA Taq® DNA Polymerase (Takara Bio, Saint-Germain-en-Laye, France) and in case proof reading was essential, Phusion® High-Fidelity DNA Polymerase (Finnzymes, Vantaa, Finland) was applied. All polymerases were used according to the manufacturers’ instructions. After isolation from yeast and propagation in E.coli, all vectors used for either over-expression or point mutation were sequenced with the BigDye Terminator v3.1 Cycle Sequencing Kit and the ABI Prism 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA).

Prior to RNA isolation, strains were cultivated for 3 to 5 days in liquid ICI medium, the mycelium lyophilized and then ground under liquid nitrogen. The isolation was carried out using the TRI Reagent™ (Sigma-Aldrich, Deisenhofen, Germany). In order to analyze gene expression levels via Northern blot (Church and Gilbert 1984), 20 μg of isolated total RNA were separated in a 1 % (w/v) denaturating agarose gel (Sambrook et al. 1989) and subsequently transferred on to a nylon membrane (Nytran™ SPC, Whatman, Sanford, USA). The blotted membrane was hybridized with 32P-labeled probes (Sambrook et al. 1989). Fragments of several genes involved in fumonisin biosynthesis were used as template. The following primer pairs were used for amplification: fum1_F/fum1_R for FUM1, fum6_F/fum6_R for FUM6, fum8_F/fum8_R for FUM8, and fum21_F/fum21_R2 for FUM21. Total genomic DNA of F. fujikuroi served as template for PCR reactions. All primers are listed in Table S1.

Plasmid construction

Yeast recombinational cloning was used to generate deletion, over-expression, and mutation vectors (Colot et al. 2006; Schumacher 2012). Directed gene deletion of FUM21 was achieved by homologous recombination using a ca. 1 kb upstream (5′ flank) and downstream region (3′ flank) of the genes that were amplified using the primer pairs fum21_5F/5R and fum21_3F/3R, respectively. The hygromycin resistance cassette encoding the hygromycin B phosphotransferase gene hph under control of the strong trpC promoter from Aspergillus nidulans was amplified with the primer pair hph_F/hph_R from pCSN44 (Staben et al. 1989). S. cerevisiae strain FY834 was co-transformed with the three fragments and the EcoRI/XhoI digested pRS426 shuttle vector (Christianson et al. 1992) resulting in the knock-out vector pΔfum21 (Fig. S1). For constitutive expression of FUM21, the gene was amplified using the primer pair OEfum21_F/OEfum21_R and cloned downstream of the strong oliC promoter of the ATP synthase subunit 9 gene from A. nidulans (Ward and Turner 1986) using the NcoI/NotI digested vector pNDN-OGG (Schumacher 2012). As resistance marker, the nourseothricin resistance cassette, consisting of the nat1 gene (nourseothricin acetyltransferase) under control of the strong trpC promoter from A. nidulans, was present in the resulting over-expression vector pOE::FUM21 (Fig. S2). Constructs for directed point mutations in the FUM1 promoter region were obtained by combination of the fragments amplified with the following primer pairs: MutPFum1_5’F/MutPFum1_5’R_hph (1), MutPFum1_PromF_hph/Mut2u3_R (2), Mut2u3F/MutPFum1_3’R (3), MutPFum1_PromF_hph/Mut3_R (4), Mut3_F/MutPFum1_3’R (5), Mut_F1/MutPFum1_3’R (6), MutPFum1_PromF_hph/Mut_R1 (7), and MutPFum1_PromF_hph/MutPFum1_3’R (8) with the EcoRI/XhoI-digested pRS426 shuttle vector and the hygromycin resistance cassette. For pMut2/3, carrying point mutations in the second and third motif sequence, fragments (1), (2), and (3) were combined. The vector pMut3, carrying point mutations in the third motif sequence, combines the fragments (1), (4), and (5), while pMut_denovo, carrying point mutations in the de novo motif sequence, combines fragments (1), (6), and (7). For the negative control vector pMut0 without mutation, fragments (1) and (8) were combined (Fig. S3 and Fig. S6). Used primers are listed in Table S1.

Fungal transformation and confirmation of transformants

F. fujikuroi was transformed by protoplast transformation as published previously (Tudzynski et al. 1999). The FUM21 deletion cassette was amplified from the isolated yeast DNA containing the vector pΔfum21 using the primer pair fum21_5F/3R. The PCR product or, in the case of over-expression, 10 μg of circular vectors were used for transformation of ca. 107 protoplasts. For selection of transformants, 100 μg/mL hygromycin B (Calbiochem, Darmstadt, Germany) or 100 μg/mL nourseothricin (Werner-Bioagents, Jena, Germany) were applied. The putative transformants were confirmed by diagnostic PCR. Δfum21 knock-out mutants were tested for homologous integration of the deletion cassette and absence of the WT gene using the primer pairs fum21_5F_dia/pCSN44_trpC_T (5′ flank), fum21_3R_dia/pCSN44_trpC_P2 (3′ flank), and fum21_seq1/fum21_seq2 (WT genotype), respectively (Fig. S1). Possible additional ectopic integration of the resistance cassette was excluded in a Southern blot analysis using the 5′ flank (primer pair: fum21_5F/5R) as labeled probe on HindIII digested gDNA (Fig. S1). The presence of the over-expression construct of FUM21 in the WT, ΔareA, and ΔareB backgrounds was tested with the primer pair PoliC_seqF2/fum21_seq1 (Fig. S2). The successful FUM21 over-expression was verified by Northern blot analysis. For generation of point mutation mutants, 30 μg of the respective vector was digested with PvuII and used for transformation of the F. fujikuroi WT and the OE::FUM21 mutant. Point mutation of the putative DNA binding motifs in the FUM1 promoter and absence of the WT sequence were verified by diagnostic PCR using the primer pairs: PromFum1_seq1/pCSN44_trpC_T (5′ flank), fum1_R/pCSN44_trpC_P2 (3′ flank) and PromFum1_seq1/fum21_F (WT genotype) and sequencing of the relevant region which was amplified using the primer pair PFum1_Mut_seqF1/PFum1_Mut_seqR1. All primers are listed in Table S1.

Prediction and identification of DNA binding sites

In order to identify over-represented promoter motifs in the gene cluster, we applied the de novo prediction by use of the algorithms MEME (Bailey et al. 2006), Weeder (Pavesi et al. 2004), and PhyloCon (Wang and Stormo 2003) and scanned for previously published binding motifs of the TRANSFAC database (Wingender et al. 2001). To specifically identify palindromic 5′-CGG[Xn]CCG-3′ binding sites, which are associated with Zn(II)2Cys6 transcription factors (MacPherson et al. 2006), we wrote a script that scans for CGG palindromes with a spacer of 0 to 30 bp between the CGG/CCG triplets. We selected the intergenic upstream sequence of genes with a maximum length of 1 kb as promoter sequence. Additionally, orthologous promoters of the closely related F. verticillioides were scanned for conserved motifs.

We determined the genome-wide occurrence of all candidate motifs and selected statistically over-represented motifs on the cluster promoters using Fisher’s exact test (Fisher 1922). All resulting p values were corrected for multiple testing using the Benjamini-Hochberg procedure (Benjamini and Hochberg 1995). Motifs with a corrected p value below 0.05 were selected as significantly over-represented promoter motifs.

Synthesis of N-acetylated fumonisin FB1

For synthesis of FA1, 10 mg (1.38 mmol) of FB1 (isolated according to Hübner et al. (2012)) were incubated with acetic anhydride (2 mmol) using trimethylamine (1.42 mmol) as chemical catalyst in 10 mL 1:1 methanol/chloroform (v/v). The reaction mixture was incubated under a protective argon atmosphere for 48 h at room temperature under constant stirring. The reaction product was evaporated, resolved in water, and analyzed by HPLC-MS/MS.

HPLC-MS/MS analysis of fumonisins

To prepare samples for HPLC-MS/MS measurements, the supernatants of fungal cultures were filtered through 0.45-μm membrane filters (BGB Analytik, Schloßböckelheim, Germany). For quantification of fumonisins, isotopically labeled FB1 (d 6 -FB1) was used as internal standard (Lukacs et al. 1996). Measured FBx and FAx levels were adjusted to the dry weight of the fungal mycelium. An API 4000 QTrap mass spectrometer (Sciex, Darmstadt, Germany) coupled to an Agilent 1100 series HPLC and a QTRAP 5500 mass spectrometer (Sciex, Darmstadt, Germany) coupled to a Nexera™ system (Shimadzu, Duisburg, Germany) were used for the detection of different fumonisins in multiple reaction monitoring (MRM) mode. Data acquisition was performed with Analyst 1.6.2 software (Sciex). Chromatographic separation was done on a 150 × 2 mm, 3-μm, Hyperclone C8 3 μm BDS 130 Å column with a 4 × 3 mm i.d. guard column (Phenomenex, Aschaffenburg, Germany) using a binary gradient of 1 % (v/v) formic acid in acetonitrile (solvent A) and 1 % (v/v) formic acid in water (solvent B), similar to the method published by Bergmann et al. (2013). The injection volume was 10 μL, and the autosampler was cooled to 4 °C. The flow rate was set to 250 μL/min with the following gradient: 35 % solvent A (0 min), 95 % A (8 min), 95 % A (10 min), subsequent return to the start conditions of 35 % A (10.5 min), and equilibration (15 min). For electrospray ionization, the ion voltage was set to +5500 V in the positive mode, and nitrogen was used as curtain gas (20 psi) and collision gas (5 × 10−5 Torr for API 4000 or medium for QTRAP 5500). Zero-grade air was used as a nebulizer gas (35 psi) and as a drying gas (45 psi) heated to 350 °C. Two MRM transitions were measured for identification and quantitation (first transition mentioned) of each analyte of interest with a duration of 50 ms: FB1 [M + H]+ 722 → 334 and 722 → 352, FB2 and FB3 [M + H]+ 706 → 336 and 706 → 318, FB4 [M + H]+ 690 → 320 and 690 → 338, FA1 [M + H]+ 764 → 394 and 764 → 334, FA2 and FA3 [M + H]+ 748 → 378 and 748 → 336, as well as d 6 -FB1 [M + H]+ 728 → 358 and 728 → 340. According parameters for declustering potential (DP), entrance potential (EP), collision cell entrance potential (CEP), collision energy (CE), and collision cell exit potential (CXP) were the following in Volt: DP 121, EP 10, CEP 35, CE 55, CXP 8.

Results

F. fujikuroi and F. verticillioides differ in quantity and quality of produced fumonisins

Previously, we have shown that F. fujikuroi produces only small amounts of fumonisin under laboratory conditions compared to F. verticillioides (Wiemann et al. 2013). The reasons for these differences are not known as the FUM cluster organization is almost the same (Fig. 1b), and the level of protein sequence identity is very high between both species (Fum1 86 %, Fum8: 83 %, Fum6 85 %, Fum21: 69 %).

In order to study the functionality of the FUM gene cluster in F. fujikuroi in more detail, we first compared the fumonisin production of both Fusaria in liquid culture. For F. verticillioides, it has been shown that FUM gene expression is subject to nitrogen metabolite repression (Kim and Woloshuk 2008). To confirm the nitrogen-dependent fumonisin production also for F. fujikuroi, the two WT strains were grown in standard synthetic ICI medium with high (60 mM) and low (6 mM) concentrations of nitrogen in the form of glutamine (Gln). Total B-type fumonisin levels (FBx) were measured by HPLC-MS/MS after 7 days of cultivation as a sum of the two main fumonisins, FB1 and FB2 (Fig. 2). Only traces of FB3 and FB4 were detected in most samples (data not shown). As expected, highest levels of FBx were produced under nitrogen-limiting conditions in both strains, although F. fujikuroi produces only about 1–10 % of the FBx level of F. verticillioides (Fig. 2). However, while F. verticillioides consistently produces FB1 as the main product, F. fujikuroi synthesizes FB1 and FB2 in a more balanced ratio mostly tending to FB2 as main compound (Fig. 2).

Fig. 2
figure 2

Comparison of fumonisin production of F. fujikuroi and F. verticillioides. The two wild-type strains were cultivated for 7 days in liquid ICI with 6 and 60 mM Gln, respectively, as sole nitrogen source. Fumonisin levels (FB1 + FB2 = ΣFBx), as well as FB1, FB2, and FA1 and FA2 separately, were analyzed in triplicate via HPLC-MS/MS from the culture supernatant using d 6 -FB1 as internal standard. Both strains show repression of FBx under high nitrogen, and under inducing conditions (6 mM Gln) F. fujikuroi only produces 1–10 % compared to F. verticillioides (a). F. verticillioides clearly produces FB1 as main compound whereas F. fujikuroi shows a more balanced ratio mostly tending to FB2 as main compound (b). In all samples measured in 6 mM Gln, the ratio of FA1 to FA2 was about the same as FB1 to FB2 (c). F.f. F. fujikuroi, F.v. F. verticillioides

Full size image

HPLC-MS/MS measurements revealed also signals of N-acetylated forms of fumonisins (FA1 and FA2) in the supernatants as, e.g., described for F. verticillioides (Lazzaro et al. 2013). Mass spectrometric transitions were verified in comparison to FA1 which was synthesized from FB1. Peak ratios representing ∑FAx (FA1 + FA2) reflect the ratio of ∑FBx comparing different strains and were therefore not included in further descriptions (Fig. 2; Fig. S4). Measurements in the WT compared to the mutant strains are in general presented as the sum of the two major fumonisins FB1 and FB2 (∑FBx).

Maize supplements induce fumonisin production in F. verticillioides, but not in F. fujikuroi

Recently, it has been shown that different plant extracts induce fumonisin production in F. proliferatum (Stępień et al. 2015). To establish whether the different host plants (maize versus rice) make the difference in FBx production yields between F. verticillioides and F. fujikuroi, we studied the effect of maize-derived supplements on FBx production in the low-producing rice pathogen F. fujikuroi and the high-producing maize pathogen F. verticillioides. Equal amounts of maize flour and germinated maize seedlings, respectively, were added as sole nitrogen source to synthetic ICI liquid medium. The FBx contents of the culture filtrates were measured by HPLC-MS/MS after 7 days of cultivation and compared to those produced in the presence of 6 mM Gln (inducing conditions). Interestingly, the addition of germinated maize seedlings led to about a fivefold increase in FBx levels in F. verticillioides, whereas the addition of maize flour had no effect (Fig. 3). It was notable that the addition of maize seedlings resulted also in production of detectable amounts of FB3 in F. verticillioides of about 4–7 % compared to FB1 (data not shown). In F. fujikuroi, none of the supplements had any effect on produced FBx concentrations (Fig. 3). Although the strains showed varying growth in the different media, no direct correlation of growth rate to produced FBx levels was found (data not shown).

Fig. 3
figure 3

Comparison of F. fujikuroi and F. verticillioides fumonisin production in media containing different maize supplements. The F. fujikuroi and F. verticillioides wild-type strains were cultivated for 7 days in liquid ICI with 6 mM Gln and 2 g/L maize flour or maize seedlings as sole nitrogen source, respectively. Fumonisin levels (FB1 + FB2 = ΣFBx) of the three strains were analyzed in triplicate via HPLC-MS/MS from the culture supernatant using d 6 -FB1 as internal standard

Full size image

Cluster-specific transcription factor FUM21 crucially regulates fumonisin biosynthesis

In F. verticillioides, the narrow-domain Zn(II)2Cys6 TF Fum21 plays an essential role for activating the expression of the FUM cluster genes. Deletion of the FUM21 gene led to almost complete inhibition of fumonisin production (Brown et al. 2007). One possibility for differing fumonisin production levels between the two species might be a low activity of Fum21 in F. fujikuroi. However, sequence comparison revealed no significant differences. The predicted intron/exon structure of the FUM21 homolog in F. fujikuroi (FFUJ_09240) was confirmed experimentally by cDNA sequencing revealing a 2067-bp cDNA encoding 688 amino acids (aa) compared to 722 aa in F. verticillioides. Both proteins contain a Zn(II)2Cys6 domain (Zn_clus, PF00172) and a domain typical for fungal TFs (Fungal_trans, PF04082) identified using the Pfam algorithm (Finn et al. 2014).

To study the role of Fum21 in F. fujikuroi, FUM21 deletion and over-expression strains were generated and analyzed for FBx production and FUM gene expression. FBx production was completely abolished in the Δfum21 mutant. In contrast, the constitutive high expression of FUM21 strongly induced gene expression of FUM cluster genes (shown for FUM1, FUM6, and FUM8) and resulted in up to 1000-fold elevated FBx levels under low nitrogen (6 mM Gln) and even under repressing conditions (60 mM Gln) (Fig. 4, Fig. S5). Interestingly, FUM21 is also expressed in the WT though slightly slower than in OE::FUM21. However, this WT expression level seems not to be sufficient for proper expression of the other FUM cluster genes. To study this phenomenon in more detail, a time course experiment was performed showing that the constitutive oliC promotor causes high expression of FUM21 from day 1 until day 7, whereas in the WT FUM21 expression starts only at day 2, peaks at day 4, and then decreases again (Fig. S5). Accordingly, FUM6 and FUM8 were also highly expressed from day 1 in OE::FUM21 compared to the weak expression in the WT (Fig. S5).

Fig. 4
figure 4

Analysis of the cluster-specific Zn(II)2Cys6 transcription factor Fum21. The F. fujikuroi wild type (WT), Δfum21, and OE::FUM21 were cultivated for 3 or 7 days, respectively, in liquid ICI medium (6 mM Gln). The expression of FUM21 itself and the essential cluster genes FUM1 (PKS), FUM6 (P450 monooxygenase), and FUM8 (aminotransferase) is not detectable in Δfum21 and strongly induced in OE::FUM21 (a). Fumonisin levels (FB1 + FB2 = ΣFBx) of the three strains were analyzed in triplicate via HPLC-MS/MS from the culture supernatant using d 6 -FB1 as internal standard (b)

Full size image

Despite the high FBx levels, growth of the OE::FUM21 was not significantly inhibited compared to the WT, neither in liquid culture nor on solid ICI medium, indicating that fumonisins have no toxic effect on the producing fungus (data not shown).

To gain more insight into the mode of action of the Zn(II)2Cys6 TF Fum21 in F. fujikuroi, we searched for potential Fum21 DNA binding sites. Recently, a conserved motif has been reported in the promoters of the FUM cluster genes in F. verticillioides (Montis et al. 2013). Although an identical motif (CGGATA|TATCCG) is not present in the promoter region of FUM1 in F. fujikuroi, we identified a very similar significantly over-represented (p value <0.05, Fisher’s exact test with Benjamini-Hochberg correction for multiple testing) motif with a consensus of 5′-ATCCGA-3′. The motif was predicted independently by the two algorithms MEME (Bailey et al. 2006) and Weeder (Pavesi et al. 2004) within the promoter regions of all 15 FUM genes in F. fujikuroi (Fig S6). Comparing the locations of these similar consensus sequences in the 5′-upstream regions of the FUM genes in both fungi revealed that only 7 of the 28 sites of the newly identified motif match with the sites found by Montis et al. (2013) limiting the promotor region to the intergenic 5′-upstream regions with a maximum of 1 kb.

Furthermore, many studies show that Zn(II)2Cys6 proteins recognize highly related elements containing trinucleotide sequences such as CGG triplets in single or repeat forms, in either a symmetrical or an asymmetrical format (MacPherson et al. 2006). Searching for those typical 5′-CGG[Xn]CCG-3′ motifs with repeated CGG triplets, we found the motif 5′-CGG[X17]CCG-3′ upstream of 7 out of the 15 FUM genes. We found this motif significantly enriched compared to the genome-wide distribution (p value <0.05, Fisher’s exact test with Benjamini-Hochberg correction for multiple testing) and even present twice in the promoter region of FUM1 (Fig. S6). In F. verticillioides, the same consensus was also identified in most of the FUM gene promoters (Fig S6).

To prove their functionality, the postulated Fum21 binding motifs were mutated by directed single nucleotide exchange in the promoter of FUM1. 5′-ATCCGA-3′ was changed to 5′-ATCAGC-3′. The two 5′-CGG[X17]CCG-3′ motifs were mutated separately and also both simultaneously (−572: CGG➔AGT and/or −543: CCG➔ACT), and the mutated fragments were homologously integrated into the locus of the FUM1 promoter in the WT as well as the OE::FUM21 mutant strains. To be sure that potential changes were not due to homologous integration of the transformed fragment carrying the hygromycin resistance cassette, the same construct without any mutation was also inserted as a control (Fig. S3).

FBx levels of two independent mutant strains per mutation were measured in triplicates and compared to the recipient WT and OE::FUM21 strains. Surprisingly, both the integration of the non-mutated control constructs and the mutated promoter fragments led to about 40–50 % reduced production of FBx (Fig. S6). Therefore, the different mutations seem to have no specific influence on fumonisin biosynthesis in F. fujikuroi (Fig. S6). The general reduction of FBx levels in all transformed strains seems to result from the insertion of the resistance cassette upstream of the FUM1 promoter what clearly hinders efficient activation of the FUM cluster and metabolite biosynthesis.

Both AreA and AreB are critical for the action of Fum21 and efficient fumonisin production in F. fujikuroi

As FBx production in F. fujikuroi is strictly regulated by nitrogen availability, we studied the potential role of the major GATA-type nitrogen regulator AreA and the recently characterized second nitrogen-responsive GATA TF AreB (Michielse et al. 2014) in fumonisin regulation. Both, FUM gene expression (shown for FUM1, FUM6 and FUM8) as well as FBx production were completely inhibited in both the ∆areA and ∆areB deletion mutants, indicating that the two global GATA-type TFs act as positive regulators of fumonisin biosynthesis (Fig. 5). Indeed, several double GATA sequence elements are present in the regions upstream of the FUM genes, which might represent binding motifs for AreA and AreB. In the promoter region of the cluster-specific TF FUM21 five double GATA motifs are located, and two are present in the promoter region of FUM1 (Fig. S6).

Fig. 5
figure 5

Influence of the global GATA transcription factors AreA and AreB on regulation of FBx biosynthesis. The F. fujikuroi wild type (WT), OE::FUM21, ΔareA, ΔareB as well as OE::FUM21/ΔareA and OE::FUM21/ΔareB were cultivated for 3 or 7 days, respectively, in liquid ICI medium (6 mM Gln). The expression of FUM21 (transcription factor) and the FUM cluster genes FUM1 (PKS), FUM6 (P450 monooxygenase) and FUM8 (aminotransferase) and was analyzed (a). Additionally, fumonisin levels (FB1 + FB2 = ΣFBx) of the strains were measured in triplicate via HPLC-MS/MS from the culture supernatant using d 6 -FB1 as internal standard (b)

Full size image

To gain insight into the hierarchy between the cluster-specific TF and the two global GATA regulators, all three having a strong influence on fumonisin regulation, FUM21 was constitutively over-expressed in the backgrounds of ΔareA and ΔareB, respectively. The expression of FUM genes (shown for FUM1, FUM6, and FUM8) was only partially restored in the OE::FUM21/ΔareA and OE::FUM21/ΔareB double mutants compared to the single ΔareA and ΔareB mutants. Both the expression of FUM genes and FBx levels were significantly less than in the OE::FUM21 mutant (Fig. 5).

Discussion

According to bioinformatic analyses, the F. fujikuroi genome contains 47 key genes for biosynthesis of SM, demonstrating the great potential of the fungus to produce a large variety of these bioactive compounds (Wiemann et al. 2013). One of them is the FUM gene cluster which exhibits a high level of conservation to that of the main fumonisin producer F. verticillioides. Fumonisins, especially those belonging to the B series fumonisins, are very potent mycotoxins responsible for food and feed contamination of maize products. They have been associated with severe diseases in humans, including esophageal cancer and neural tube defects (Desai et al. 2002; Marasas et al. 2004). To get a deeper insight into factors affecting FBx production in the low-producing fungus F. fujikuroi, we studied the complex regulation of fumonisin biosynthesis, and in particular, the role of pathway-specific and broad regulators and the interplay between them.

F. fujikuroi and F. verticillioides differ in quantity and quality of produced fumonisins

In F. fujikuroi, all 15 FUM cluster genes are highly conserved compared to the F. verticillioides cluster genes except for the missing FUM17 and FUM20 genes. FUM17 is predicted to encode a longevity assurance factor, but was shown to be not essential for fumonisin biosynthesis in F. verticillioides (Proctor et al. 2003), and FUM20 has no predicted function or even an open reading frame (Brown et al. 2005; Wiemann et al. 2013). Therefore, strong similarities in FBx biosynthesis and its regulation have been expected. However, F. fujikuroi produces only 1–10 % of the FBx levels of F. verticillioides under inducing low-nitrogen conditions.

Nevertheless, nitrogen repression of FBx biosynthesis, as reported for F. proliferatum and F. verticillioides (Shim and Woloshuk 1999; Kohut et al. 2009), was also true for F. fujikuroi because it produced significantly less FBx in the presence of high (60 mM Gln) nitrogen compared to low nitrogen (6 mM Gln).

Most fumonisin-producing Fusarium field isolates produce FB1 and FB2 as main products (Summerell and Leslie 2011). In contrast to F. verticillioides M-3125, which produces mainly FB1, the ratio in F. fujikuroi was more balanced, tending to a higher level of FB2. This fumonisin was found to be comparably toxic to FB1 in jimsonweed (Datura stramonium), and even more toxic in rat hepatoma cells and turkey lymphocytes (Abbas et al. 1993; Gutleb et al. 2002). High levels of FB2 were also reported for some Aspergillus and several Fusarium isolates, including some of F. verticillioides (Frisvad et al. 2007; Matić et al. 2013). In Aspergillus niger, the absence of FB1 can result from the lack of the monooxygenase encoding gene FUM2 (Susca et al. 2010).

Although F. fujikuroi and F. verticillioides are similar plant pathogens, they have evolved different host specificities: whereas F. fujikuroi infects rice (Leslie and Summerell 2006; Wiemann et al. 2013), F. verticillioides is mainly a maize pathogen (Nelson 1992). As might be expected, FBx production was only stimulated in the maize pathogen F. verticillioides by addition of maize supplements. One component contained in maize kernels, amylopectin, was shown to activate FB1 biosynthesis in F. verticillioides (Bluhm and Woloshuk 2005). However, in our experiments, only ground maize seedlings showed a stimulating effect, suggesting a specific signal probably emerging during germination.

Fum21 strongly regulates fumonisin biosynthesis in F. fujikuroi

The cluster-specific narrow-domain Zn(II)2Cys6 TFs Fum21 in F. fujikuroi and F. verticillioides are 69 % identical at the protein level. Both show an identical domain structure consisting of the typical Zn(II)2Cys6 domain (PF00172) and a second conserved domain (PF04082), which is typical for fungal TFs. Deletion of FUM21 in F. fujikuroi resulted in the expected loss of FBx production, as it was first described for Δfum21 mutants in F. verticillioides (Brown et al. 2007). On the contrary, over-expression of cluster-specific TFs, e.g., the apicidin F TF Apf2, is a common approach to activate silent or only faintly expressed gene clusters (Niehaus et al. 2014a). APF2 over-expressing strains produce up to six times more product than the WT under inducing conditions and show even slightly enhanced product levels under repressing conditions (Niehaus et al. 2014a). For FBx biosynthesis, such an effect was not reported so far. In fact, constantly high expression of FUM21 in F. fujikuroi led to a strong activation of cluster gene expression (shown for FUM1, FUM6, and FUM8) and a 1000-fold elevation of FBx levels compared to the WT under inducing (6 mM Gln) as well as repressing (60 mM Gln) conditions. It is surprising that FUM21 transcripts are clearly detectable in the WT while the Fum21-regulated cluster genes are not or only weakly expressed. Accordingly, only low FBx levels are detectable in the WT. Apparently, the constitutive strong expression of FUM21 in OE::FUM21 starting already from day 1 is the reason for the significantly elevated FBx level. In addition, a different transcription start introduced by the constitutive oliC promotor or changed transcript stability due to the used Botrytis cinerea glucanase (gluc) terminator (Schumacher 2012) could have an impact on production yield.

Hence, even the low-level producer F. fujikuroi can be a potent producer of the mycotoxins fumonisins, leading to potentially high contamination of rice products. Fum21 activators or activating conditions are crucial for high fumonisin contamination, making Fum21 and its regulation a suitable target for further studies dealing with fumonisin control.

In order to better understand the regulatory mechanism underlying the FUM cluster gene activation by Fum21 in F. fujikuroi, putative binding motifs were analyzed. Though the exact motif previously reported by Montis et al. (2013) is not present in the promoter of FUM1 in F. fujikuroi, a very similar motif (5′-ATCCGA-3′) was found in the 5′-upstream regions of all cluster genes. The positions of both motifs overlap only in one fourth of the predicted sites. Additionally, a typical Zn(II)2Cys6 binding motif, 5′-CGG[X17]CCG-3′, was identified in seven FUM genes, including the FUM1 promoter. However, the mutation of each of these motifs in either OE::FUM21 or WT background did not have any inhibitory effect on produced FBx levels. Already the introduction of the negative control construct without mutation led to a reduction of the FBx content of about 40–50 % which also occurred in all tested mutants. Recently, Visentin et al. (2012) showed by GFP fusion that the genomic localization and thereby probably the chromatin status of the promoter regions of the key genes FUM1 and FUM21 in F. verticillioides is important for activation of the gene expression. We assume that the interaction of further regulators involved in the regulation of fumonisin biosynthesis with the FUM1 promoter region is hindered by the introduction of the resistance cassette leading to the overall FBx reduction in the transformants. However, additional experiments will have to test this hypothesis, though from other fumonisin-producing strains several eligible regulators are known (reviewed in Woloshuk and Shim 2013).

Beside the two motifs described above, a conserved motif with a significantly different consensus sequence (5′-AAATTGTCA-3′) was found in the promotor region of all cluster genes except for FUM14 and FUM15 and would be promising for future mutation experiments (motif3, Fig. S6a).

Fumonisin biosynthesis is regulated by a complex regulatory network in F. fujikuroi

In general, fumonisin biosynthesis has been shown to involve several levels of control, and FUM genes are the target of several regulators (reviewed in Woloshuk and Shim 2013). Among other factors, histone H4 acetylation of promoter regions of FUM cluster genes is important for their transcription in F. verticillioides (Visentin et al. 2012), and a role of the HOG MAP kinase pathway in regulation of fumonisin biosynthesis was described in F. proliferatum (Kohut et al. 2009). In F. verticillioides and F. fujikuroi, mutations in the VEL1 gene, a homolog of the velvet gene VEA in A. nidulans, abolished fumonisin production on cracked maize (Myung et al. 2009; Wiemann et al. 2010). In addition, the pH regulator Pac1 was shown to repress the expression of FUM genes in F. verticillioides (Flaherty et al. 2003). Furthermore, deletion of ZFR1, encoding a Zn(II)2Cys6 TF of unknown function, resulted in a mutant strain producing only trace amounts of FB1 (Flaherty and Woloshuk 2004). Studies in F. verticillioides and F. fujikuroi revealed the essential role of the cluster-specific TF Fum21 (Brown et al. 2007; this work). There are no reports describing the putative cis-element in the promoters of the fumonisin pathway genes that is specifically recognized by Fum21 so far. Bioinformatic analysis revealed two potential binding motifs for Fum21 in the promoters of the F. fujikuroi FUM genes. However, targeted replacement of the FUM1 WT promoter for the mutated promoter fragments did not result in significant reduction of fumonisin production, neither in the WT nor in the OE::FUM21 backgrounds. Systematic mutations of more putative binding motifs will be necessary to identify the correct Fum21 binding motifs.

Besides being regulated by the cluster-specific TF Fum21, fumonisin biosynthesis in F. fujikuroi was shown to be subject of nitrogen metabolite repression. Therefore, the two prominent GATA-type TFs involved in nitrogen-dependent regulation, AreA and AreB, were analyzed to establish their role in fumonisin regulation. In F. verticillioides, ∆areA mutants produced little to no fumonisin on cracked maize, while constitutive expression of AREA led to production of FB1 even under nitrogen-sufficient conditions (Kim and Woloshuk 2008). Here, we have shown for the first time that not only AreA, but also the second nitrogen-responsive GATA TF, AreB, is essential for activation of the fumonisin gene cluster in F. fujikuroi. Deletion of either AREA or AREB resulted in total loss of fumonisin production. According to this observation, several double GATA sequence elements are present in the regulatory regions of FUM cluster genes. Fumonisins are only the second example, besides gibberellins, for a secondary metabolite gene cluster being a common target for both TFs under nitrogen-limiting conditions (Michielse et al. 2014). There are other gene clusters which are regulated only by one of the two GATA factors. Thus, AreB but not AreA is essential for apicidin F and fusaric acid biosynthesis under inducing nitrogen-sufficient conditions in F. fujikuroi (Niehaus et al. 2014a, b).

In addition to the global regulators AreA and AreB, we showed the strong activating effect of the pathway-specific TF Fum21. To gain insight into the role of AreA and AreB in the hierarchical gene regulatory network of global and pathways-specific TFs, double mutants were generated. In these double mutants, the FUM21 over-expression was combined with the deletion of either AREA or AREB. Interestingly, loss of either AreA or AreB had a much stronger impact on fumonisin production than over-expression of FUM21. Compared to the 1000-fold elevation of FBx in the OE::FUM21 mutant, both double mutants produced only 1 % of the FBx level measured with the sole over-expression of FUM21.

Studies on the nitrate reductase cluster in A. nidulans revealed a function of AreA in chromatin remodeling and showed its epistatic effect over the specific TF NirA at the promoters of the nitrate utilization gene cluster (Berger et al. 2008). A similar interconnection is conceivable for the mode of action of AreA and AreB at the fumonisin cluster. However, a putative link has to be further investigated.

Recently, a global regulator of secondary metabolism, Sge1, has been identified in F. verticillioides and F. fujikuroi which affects biosynthesis of multiple secondary metabolites including fumonisins (Brown et al. 2014; Michielse et al. 2015). In F. fujikuroi, deletion of SGE1 resulted in significantly reduced expression levels of FUM, gibberellin, fusaric acid, fusarin, apicidin F, and bikaverin biosynthetic genes. Conversely, over-expression of Sge1 led to a 98-fold increase in FBx production. Interestingly, no FBx production was recorded in the ∆areA/OE::SGE1 double mutant underlining the dominating role of AreA in the complex fumonisin regulatory network (Michielse et al. 2015).

In summary, we showed that the FUM gene cluster and FBx biosynthesis are strongly inducible by the cluster-specific TF Fum21. Using this approach, the naturally low fumonisin-producing rice pathogen F. fujikuroi was able to produce 1000-fold more FBx than the WT. However, the binding motif for Fum21 still remains to be elucidated. Furthermore, we showed that nitrogen metabolite repression of fumonisin biosynthesis is mediated not only by the major nitrogen regulator AreA, but also a second GATA-type TF AreB. Both are essential for full activation of the fumonisin gene cluster and metabolite biosynthesis, and the loss of one of them obstructs the pathway-specific TF Fum21 from fully activating expression of FUM cluster genes.