Introduction

Soil-borne ascomycete fungi belonging to the genus Fusarium have high impact on health and agriculture (Nucci and Anaissie 2007; Dean et al. 2012). Like other filamentous fungi, Fusarium has the ability to produce small-specialized compounds, secondary metabolites (SMs), not associated directly with growth or reproduction, although hypothesized to contribute to the fitness of the fungal producer (Brakhage 2013; Ding et al. 2018). Major research achievements have provided an understanding of the biochemical and molecular machinery which controls the formation of these chemical compounds (Keller 2019). Specialized SMs may act as plant hormones or virulence factors, as mycotoxins dangerous to humans and animals, or even as weapons against other microbes (Desjardins and Proctor 2007; Brown et al. 2014). This makes characterization of the compounds and the underlying genetics an important priority. Here, we aim to highlight the technologies that have been used to unravel the secondary metabolism in Fusarium and encourage further initiatives to link biosynthetic gene clusters to their products.

Fusarium comprises more than 100–500 (Leslie and Summerell 2006) species capable of causing infection in plant, humans, and domesticated animals (Summerell et al. 2010). Members of this genus are found in warm and temperate ecosystems throughout the globe, often as plant pathogens contributing to major economic losses due to infected crops (Mcmullen et al. 1997; Windels 2000; Michielse and Rep 2009). Many species are harmless, but species such as F. graminearum and F. oxysporum infect cereals and produce high amounts of mycotoxins rendering entire harvests unfit for consumption (Windels 2000). The speciation of Fusarium has always posed a challenge to researchers due to the lack of distinguishing morphological features. Historically, the number of recognized species has varied between 9 and  > 1000, depending on the implemented identification scheme. Currently, many species concepts represent polyphyletic clades comprising several individual species (Leslie and Summerell 2006; Watanabe et al. 2011). Therefore, careful naming of strains is important for this genus. In some cases, the term forma specialis (f. sp.) has been introduced, e.g., categorizing strains of F. oxysporum after the plant they infect (Gordon and Martyn 1997).

Fusarium SMs exhibit an extreme diversity in function and chemical structures. They are usually formed by multi-domain core synthases often cooperating with several decorating enzymes in a pathway to generate the final product. The genes encoding these enzymes are co-regulated and commonly found as neighbors to the core-synthase-encoding gene, and together, they form a biosynthetic gene cluster (BGC) (Keller et al. 1997; Yu and Keller 2005). In addition, gene-encoding transcriptional regulators, transport proteins, and even product detoxification proteins can be encoded by BGCs. Unfortunately, many of these potential BGCs shows little to no expression when grown under the standard laboratory conditions (Gaffoor et al. 2005). Therefore, the potential new SMs may not be produced at all or are present at levels too low to be detected by the standard chemical methods (Wiemann and Keller 2014). Although many molecules have been isolated and described, the full biochemical potential of the collected Fusarium secondary metabolome is yet to be explored. In this review, we cover techniques that have been successful in Fusarium to link gene clusters to biosynthetic compounds and pathways, and suggest new possibilities to be explored in future endeavors.

Classes of secondary metabolites

Secondary metabolites are synthesized from small metabolites such as short-chain carboxylic acids and amino acids from the primary metabolism. These precursors are polymerized by large synthase/synthetase enzymes such as iterative polyketide synthases (PKS, types I and III), non-ribosomal peptide synthetases (NRPS), or terpene cyclases (TC). Fusarium species are capable of producing many terpenes (Brock et al. 2013; Burkhardt et al. 2016), some of which are important virulence factors, for example, nivalenol and deoxynivalenol (Marasas et al. 1979; Yoshida and Nakajima 2010; Yang et al. 2018) or plant hormones such as gibberellins (Bömke and Tudzynski 2009; Troncoso et al. 2010). However, the majority of characterized SMs belong to the chemical groups of polyketides (reduced and non-reduced), non-ribosomal peptides, or hybrid compounds (Sieber et al. 2014; Hoogendoorn et al. 2018) (Fig. 1). PKSs are large multi-domain enzymes that as a minimum contain beta-ketosynthase (KS), acyl-transferase (AT), and acyl-carrier protein (ACP) domains which work together in an iterative manner to elongate a polyketide chain with one ketide unit at a time (McDaniel et al. 1994; Bentley and Bennett 1999). Fungal PKSs work in different ways to create structural diversity. Usually, the biosynthesis will start from an acetyl-CoA unit, which is then elongated with malonyl-CoA units through Claisen condensation performed by the KS domain. However, in some cases, the starter unit can stem from another PKS or a fatty acid synthase (Brown et al. 1996). In addition to the KS-AT-ACP domains, PKSs may contain tailoring domains which add to the chemical diversity, e.g., reductase, dehydrogenase, or methyltransferase domains (Meier and Burkart 2009). The PKS type I, which is most predominant in Fusarium (Brown and Proctor 2016), can be further subdivided into reducing or non-reducing PKSs, yielding either fatty acid like or aromatic products. Finally, the tailoring domains can skip an iteration as seen for zearalenone, where only four out of five ketones are fully reduced (Gaffoor and Trail 2006). In addition, similar PKSs may produce different products. It is not surprising that the prediction of the final product based on amino-acid sequence alone has proven impossible.

Fig. 1
figure 1

Identified polyketide and non-ribosomal peptide metabolites from Fusarium

Full size image

NRPSs are multi-modular assembly lines catalyzing the formation of small peptides from amino-acid monomers. A full NRPS module contains an adenylation (A), a peptide acyl-carrier (T), and a condensation (C) domain. An NRPS is thus composed of one or more elongation modules (A–T–C) which catalyze the formation of a polypeptide chain. In addition, each module may contain tailoring domains, e.g., epimerization or N-methylation domains, that contribute to the chemical diversity of non-ribosomal peptides (Finking and Marahiel 2004). The compound is released from the synthetase by cyclization, reduction, or hydrolysis, and can be further modified by additional tailoring enzymes encoded by the gene cluster such as cytochrome P450 monooxygenases and dehydrogenases.

At least 500 different NRPS substrates have been reported in filamentous fungi, which comprise non-proteinogenic amino acids, d and l forms, and even hydroxyl acids (Strieker et al. 2010). The A domain contains a binding pocket that recognizes a specific amino-acid substrate (Conti et al. 1997), and substrate prediction algorithms have been developed, first for bacterial NRPSs (Stachelhaus et al. 1999; Challis et al. 2000) and further modified to include eukaryotic NRPSs (Rausch 2005; Bachmann and Ravel 2009; Röttig et al. 2011; Khayatt et al. 2013; Knudsen et al. 2016). The feasibility of using these tools to predict Fusarium NRPS substrate accurately remains to be demonstrated (Wollenberg et al. 2017).

In the case for both NRPS and PKS BGCs, the linking of biosynthetic metabolites to their respective genes relies on experimental evidence. In the following sections, we will cover the current pre-requisites for linking Fusarium metabolite–BGC pairs.

Genomic resources

Knowledge of the genetic basis for SM biosynthesis is essential for genetic manipulation and genome-mining strategies. Genome sequencing has been carried out for several species representative of the genus Fusarium (Cuomo et al. 2007; Ma et al. 2010, 2014; Al-Reedy et al. 2012; Gardiner et al. 2012, 2014; Wiemann et al. 2013; Moolhuijzen et al. 2013; Lysøe et al. 2014; King et al. 2015; Vanheule et al. 2016; Brown and Proctor 2016; Ponts et al. 2018; Walkowiak et al. 2016), revealing a potential for SM production that exceeds the expected (Kroken et al. 2003; Sieber et al. 2014). Comparative analyses of biosynthetic genes show their distribution across the Fusarium metagenome and provide insight into the evolution of BGCs (Ma et al. 2010). Different species of Fusarium are able to produce structurally similar compounds, e.g., fusarubins (PKS3), gibepyrone A (PKS8), and ferricrocin (NRPS2) (Wiemann et al. 2013; Hansen et al. 2015) and, therefore, carry alleles with high levels of homology. To detect BGCs in Fusarium genomes, different bioinformatic complementary tools are used. Protein prediction tools such as SMURF (Khaldi et al. 2010) can predict SM-related genes (Ma et al. 2010; Wiemann et al. 2013) and InterPRO (Apweiler et al. 2000) can determine the protein domain functions (Frandsen et al. 2006). The BLASTP alignment-based algorithm (Altschul et al. 1990) has become a cornerstone in every genetics study and was used in the first Fusarium genetic analyses (Proctor et al. 1999; Linnemannstöns et al. 2002; Kim et al. 2005a; Malz et al. 2005; Varga et al. 2005). CASSIS (Wolf et al. 2016) was developed based on the hypothesis that BGC genes that are co-localized and co-expressed within the same cluster will contain common regulatory patterns in the cluster promoters. SMIPS/CASSIS has been used to identify cluster-specific promoter motifs (Sieber et al. 2014). Finally, the AntiSMASH (Blin et al. 2017) platform combines some of the above methods and includes co-localization comparison analysis (cluster orthology) data to identify BGCs (Wiemann and Keller 2014). These bioinformatic analyses can be combined and aligned with experimental data to form strong evidence in scientific studies.

Three recent studies have analyzed 31 available Fusarium genomes for the presence and distribution of BGCs (Hansen et al. 2012b, 2015; Brown and Proctor 2016; Hoogendoorn et al. 2018) (Fig. 2). Prediction of SM gene clusters (and pseudo-genes) has been carried out and a numbering nomenclature was introduced (Hansen et al. 2012b, 2015). This has been extended to provide a simple system to identify all the PKS and NRPS genes by a number (Brown and Proctor 2016). So far, 67 PKS and 52 NRPS gene clusters have been identified distributed across the Fusarium metagenome. Only 16 out of 67 PKS and 13 out of 52 NRPS Fusarium genes have been linked to their respective biosynthetic product (Table 1).

Fig. 2
figure 2

Frequency of polyketide synthase gene clusters in 31 species of Fusarium

Full size image
Table 1 Fusarium PKS and NRPS gene clusters linked to natural products
Full size table

Some biosynthetic gene clusters are found in the majority of species of Fusarium (PKS3, 7, 8), while others are restricted to a single phylogenetic clade, e.g., PKS29, 30, 31, 32, 33, and 35 from the F. solani complex. The distribution of Fusarium BGCs does not always follow a strict phylogenetic pattern and evidence for horizontal gene transfer events has been reported (Oide et al. 2006; Ma et al. 2010; Gardiner et al. 2012; Sieber et al. 2014).

Activation through cultivation

Although Fusarium does produce many metabolites under cultivation, the majority of BGCs remain silent and their products cryptic. The nature and function of Fusarium SMs probably require specific physiochemical conditions to trigger their activity (Brakhage 2013). To maximize chance of observing fungal metabolites, a popular strategy is to use different cultivation parameters (Bode et al. 2002; Hemphill et al. 2017). However, the activation of cryptic biosynthesis pathways is never guaranteed (Gaffoor et al. 2005). Properties such as pH and nitrogen source are important parameters to control for some metabolite pathways (Linnemannstöns et al. 2002; Kim et al. 2005b; Gomez-Gil et al. 2018). Substituting the nitrogen source glutamine with sodium nitrate in ICI medium leads to the formation of fusarubins instead of bikaverin pigmentation in F. fujikuroi (Studt et al. 2012), which emphasizes the importance of standardized growing medium recipes to strengthen reproducibility (Wiemann et al. 2009; Sørensen and Sondergaard 2014). Biological challenges in the form of co-cultivation with other microorganisms may activate silent biosynthetic pathways leading to an increased metabolite and mycotoxin production as well as changes in growth rate (Müller et al. 2012; Netzker et al. 2015). For instance, F. demicellulare shows enhanced production of fusaristatin A which inhibits the growth of its competitor (Li et al. 2016). The co-cultivation of F. tricinctum and B. subtilis enhances the formation of enniatins and fusaristatin A drastically, and induces the formation of three novel compounds: macrocarpon C, 2-(carboxymethylamino)benzoic acid and (−)-citreoisocoumarinol (Ola et al. 2013). This demonstrates the utility of this approach and confirms the role of SMs as competitive agents.

Transformation methods

The majority of studies mentioned in Table 1 have used genetic manipulation to create a link between genes and the formation of a specific biosynthetic metabolite. A prerequisite for the use of this approach in Fusarium metabolomics was to develop transformation protocols and tools. Protoplast-mediated transformation (PMT) is the most commonly used transformation system in filamentous fungi. Freshly germinated hyphae (Fig. 3a) are treated with commercially available enzymes to remove complex cell-wall components, resulting in the formation of protoplasts (Rodriguez-Iglesias and Schmoll 2015). The protoplasts are usually mixed in an osmotic stabilizing solution such as 1.2 M KCl or 1.2 M sorbitol containing CaCl2 (Fig. 3b). Protoplasts can be mixed with both circular plasmid or linearized double-stranded DNA. Calcium ions are added to open channels in the cytomembrane and thus promote uptake of nucleotides (Olmedo-Monfil et al. 2004). Polyethylene glycol (PEG) is added to promote fusion between exogenous nucleotides and protoplasts (Fig. 3c) (Becker and Lundblad 2001). Transformed protoplasts often require regeneration in osmotically stabilized medium before they are selected. PMT has been applied to transform several Fusarium species with high levels of success (Table 2). PMT has been observed to result in multicopy and ectopic integration events for some species of filamentous fungi (de Groot et al. 1998; Meyer 2008). Indeed, ectopic integration events have been reported in transformants from F. verticillioides (Proctor et al. 1999) and F. fujikuroi (Tudzynski et al. 1996) protoplasts. However, the frequency of homologous recombination-guided integration events depends not only on the transformation host strain in question, but also on the transformation protocol applied (Fernández-Martín et al. 2000).

Fig. 3
figure 3

Overview of protoplast-mediated transformation (a–c) and Agrobacterium tumefaciens-mediated transformation of Fusarium spp. (d). a Mycelial tissue comprising a thick cell wall. b Enzymatic treatment of mycelium releases protoplasts encapsulated by cytomembrane and no cell wall. c Protoplast transformation. Polyethylene glycol (PEG) can form a molecular bridge between cell and nucleotide. DNA uptake is possible through a porous membrane. d Overview of transfer-DNA delivery to the nucleus of Fusarium spp. conidia

Full size image
Table 2 List of studies exemplifying development and application of protoplast-mediated protocols
Full size table

The Gram-negative bacterium Agrobacterium tumefaciens is known for its ability to infect plants and during this process transfer the transfer-DNA (T-DNA) region of the Ti plasmid to the genome of the colonized host. The T-DNA regions are bordered by two imperfect inverted repeats (left and right border), and it is possible to introduce exogenous DNA by inserting it between the two border sites (Citovsky et al. 2007). A. tumefaciens is also capable of infecting filamentous fungi when induced by acetosyringone (Fig. 3d) (Idnurm et al. 2017), and a vast arsenal of binary vectors has been developed for this purpose (Frandsen 2011). The T-DNA is usually integrated in the fungal genome as a single copy by homologous recombination (Mullins et al. 2001; Michielse et al. 2005), and has been proven to be effective for targeted gene deletion mediated by homologous recombination (Kistler and Benny 1988; Idnurm et al. 2017). The major bottlenecks in this technique include the preparation of binary vectors and testing of various technical parameters, as an optimized protocol has to be developed for every species (de Groot et al. 1998; Utermark and Karlovsky 2008; Sørensen et al. 2014b). As for PMT, ATMT protocols have been developed for several representatives of Fusarium (Table 3).

Table 3 List of studies exemplifying development and application of Agrobacterium tumefaciens-mediated transformation protocols
Full size table

Other less used transformation strategies for species of Fusarium include glass bead transformation (Singh and Rajam 2013), electroporation (Liang et al. 2014), and restriction enzyme-mediated integration (Linnemannstöns et al. 1999; Shim and Woloshuk 2001; Inoue et al. 2001; Han et al. 2004). For manipulation of novel strains, we recommend using or expanding the already developed vector systems and protocols reported in the literature.

Low hanging fruits: known secondary metabolites produced under laboratory conditions, identified through knock-out/gene disruption

In this section, we will show how metabolites identified in laboratory cultures can be linked to their underlying genetic machinery. To identify biosynthetic genes, a simple but effective strategy has been to identify putative gene candidates, deleting or disrupting the genes, and then determining the SM complement. The concept of gene replacement, sometimes referred to as knock-out, is based on genomic insertion of a DNA cassette guided by one or two border sequences homologous to the respective target gene or region. The amount of homologous nucleotides required for homologous recombination varies from species to species, but most Fusarium spp. can perform homologous recombination between insert and genome if the segments are 800–1500 bp long (Gaffoor et al. 2005; Oide et al. 2006; Frandsen et al. 2012; Wollenberg et al. 2017; Bahadoor et al. 2018). However, heterologous or ectopic recombination is a common phenomenon (Proctor et al. 1999; Malz et al. 2005) and genetic validation is required to confirm the correct integration of the deletion cassette, e.g., via Southern blot or diagnostic PCR analyses.

Disruption vectors containing a single segment homologous to the target gene are rapidly prepared and introduced into the fungal genome by a single cross-over recombination event guided by either PMT (Gaffoor et al. 2005; Gaffoor and Trail 2006; Brown et al. 2012) or ATMT (Malz et al. 2005) (Fig. 4a). Gaffoor et al. 2005 identified 15 PKS genes in F. graminearum and used RT-PCR to determine under what growth conditions 14 of the genes were expressed. The authors developed a single cross-over gene disruption vector system targeting 500–1750 bp downstream of the start codon of each PKS gene. The mutants provided evidence for PKS12 being responsible for the formation of the red mycelial pigment aurofusarin, PKS10 was responsible for the formation of fusarin C, and PKS3 provided the basis for perithecial pigments, later identified as the fusarubins (Studt et al. 2012) and bostrycoidin (Frandsen et al. 2016). PKS4 and PKS13 were shown to collectively synthesize the mycotoxin zearalenone (Gaffoor and Trail 2006; Lysøe et al. 2006). However, the remaining 11 PKS genes could not be correlated with a phenotype or metabolite (Gaffoor et al. 2005). Meanwhile, a similar approach was used for targeted gene disruption with a vector carrying a 833 bp fragment homologous to the KS domain of PKS12 in F. graminearum, which was introduced via ATMT with A. tumefaciens AGL1 (Malz et al. 2005). The resulting FgΔPKS12 strain showed no PKS12 expression and had an albino phenotype. In addition, F. graminearum was disrupted in the promoter region of aurR1, which turned out to act as a transcriptional activator of several genes in the PKS12 cluster. Later, a similar disruption vector was prepared, linearized, and transformed into protoplasts targeting the KS domain of PKS10 in F. verticillioides (Brown et al. 2012). Disruption mutant FvΔPKS10 provided evidence that this PKS was responsible for formation of fusarin C.

Fig. 4
figure 4

Targeted disruption and gene replacement strategies. a Gene disruption by integration with targeting plasmid-containing segment homologous to part of biosynthetic gene. b Gene replacement by two homologous recombination events replacing the entire open-reading frame with an insertion cassette. c Split-marker gene replacement in protoplasts transformed with two nucleotide fragments

Full size image

A more popular disruption strategy is based on vectors containing two homologous segments to the target gene separated by a selection marker gene (Fig. 4b). This enables replacing a large portion or the entire biosynthetic gene with the selection marker. Gene replacement has been carried out in most Fusarium spp. guided by either PMT of F. fujikuroi and F. venenatum (Proctor et al. 1999; Song et al. 2004; Wiemann et al. 2009; Niehaus et al. 2014a; Janevska et al. 2016; Studt et al. 2016a) or ATMT of F. graminearum, F. pseudograminearum, F. solani, F. avenaceum, and F. semitectum (Frandsen et al. 2006, 2016; Tobiasen et al. 2007; Ma et al. 2010; Sørensen et al. 2014a, b; Romans-Fuertes et al. 2016; Wollenberg et al. 2017; Bahadoor et al. 2018).

In a study of linking a biosynthetic gene to the formation of fusarin C, Song et al. 2004 produced a knock-out vector able to replace PKS10 from F. venenatum through a double cross-over recombination event. However, Southern blot analysis of the transformed mutants showed that four different types of recombination events between circular plasmid and genome had occurred, all resulting in gene replacement or disruption and the inability to produce fusarin C (Song et al. 2004). The four recombination possibilities between genome and a vector carrying two segments homologous to target genes are: double cross-over leading to gene replacement, integration of the plasmid in the 5′ end of the gene by a single cross-over, integration in the 3′ end of the gene by a single cross-over event, or plasmid integration in both ends of the gene. This agrees with studies in F. verticillioides, where different recombination events have been observed. In one study, only one out of 16 mutants displayed correct gene replacement through double recombination, while 15 mutants had experienced integration of the entire vector in one or more copies, disrupting FvPKS24 (Proctor et al. 1999). Overall, 14% of screened F. pseudograminearum protoplast mutants carried a deletion of the virulence-related FpAH1 gene resulting from double cross-over events (Gardiner et al. 2012). Similarly, successful knock-outs of NRPS32 led to the identification of the novel lipopeptide W493 A and B (Sørensen et al. 2014a).

In contrast to transforming fungi with circular or linearized copies of knock-out vectors, some researchers choose to use the PCR-amplified knock-out cassette directly to minimize the risk of ectopic integration of plasmid backbone elements. This PCR-based PMT approach was used to study the zearalenone gene cluster in F. graminearum (Kim et al. 2005b) and the fusarubins BGC in F. fujikuroi (Studt et al. 2012). Catlett et al. (2003) introduced the split-marker system, where two PCR products each comprised a 3′ or 5′ homologous target region with each two-thirds of the selection marker, together capable of forming an intact deletion cassette when combined through homologous recombination (Fig. 4c) (Catlett et al. 2003; Chung and Lee 2015). Brown et al. (2012) demonstrated targeted gene replacement of FvPKS21 by split marker and PMT. This approach was utilized in the investigation of F. graminearum siderophore biosynthesis pathways (Oide et al. 2006, 2007, 2014) to create knock-outs of NRPS1, 2, and 6, and, furthermore, used for investigation of the aurofusarin pigmentation pathway in F. graminearum (Kim et al. 2005a). Extensive work has been performed describing biosynthetic pathways in F. fujikuroi, and gene replacement experiments have been the key to establishing links between BGCs and the pigment bikaverin (Wiemann et al. 2009), the plant mycotoxin fusaric acid (Studt et al. 2016a), and the peptide drug lead apicidin F (Niehaus et al. 2014a). In one study, knock-out mutants combined with isotopically labeled precursors led to the identification of FfPKS8 as the progenitor of gibepyrone biosynthesis (Janevska et al. 2016).

ATMT with A. tumefaciens strain LBA4404 is used for gene replacement experiments in Fusarium spp. with great success (Idnurm et al. 2017). Based on the ATMT protocol developed by Malz et al. (2005), a vector system for targeted gene deletion was established for F. graminearum (Frandsen et al. 2006, 2008, 2012), allowing characterization of the aurofusarin (PKS12) and fusarubins (PKS3) pigment biosynthesis. Furthermore, ATMT and gene replacement in F. graminearum enabled characterization of ferricrocin (NRPS2), fusaristatin (PKS6-NRPS7), chrysogine (NRPS14), and recently gramillins biosynthesis (NRPS8) (Tobiasen et al. 2007; Sørensen et al. 2014a; Wollenberg et al. 2017; Bahadoor et al. 2018). The F. graminearum ATMT protocol was adapted for F. avenaceum by testing different transformation parameters to identify optimal concentration of spores, co-inoculation time, and incubation temperature. Both F. graminearum and F. avenaceum displayed a high gene-targeting efficiency guided by homologous recombination between T-DNA and the gene of interest. In rare cases (1–10%), the T-DNA integrated ectopically in the recipient genome through non-homologous end-joining (Frandsen et al. 2012; Sørensen et al. 2014b). For gene disruption in F. solani and F. semitectum, ATMT protocols are available relying on the A. tumefaciens AGL-1 strain (Jin et al. 2010; Romans-Fuertes et al. 2016).

To cement the function of a gene, complementation can be carried out by reintroduction of the target gene into the disrupted or replaced gene mutant strains (Proctor et al. 2008; Studt et al. 2016a). Complementation of fumonisin production was carried out by transforming a deficient F. verticillioides strain with a gDNA cosmid library clone carrying PKS24, yielding mutants producing wild-type titers of fumonisins (Proctor et al. 1999). A feasible complementation technique for F. graminearum and F. semitectum has been to PCR amplify a wild-type gene allele including native promoter and terminator sequence in a single fragment which can be co-transformed into protoplasts together with a plasmid containing a selection marker different to what was used to disrupt the target gene originally (Kim et al. 2005a, b; Jin et al. 2010).

Gene replacement is a powerful tool to link genes to function and entire pathways can be resolved in this way. Not only the core synthase can be identified, but the contribution of the other genes in the same cluster to the final product can be determined (Frandsen et al. 2006, 2016; Wiemann et al. 2009; Studt et al. 2012, 2016a; Kakule et al. 2013). However, it is important to bear in mind that for a successful outcome of this strategy, the fungus must produce the target compound under the cultivation conditions used.

Targeted activation

With the introduction of sequencing, the identification of gene clusters has become trivial—but their silence is still a challenge. In consequence, targeted gene activation is used to discover new biosynthetic pathways in Fusarium spp. Core-synthase genes such as PKS and NRPS genes make ideal targets for targeted gene activation. A vector is prepared containing a constitutive promoter and a selection marker between two segments for targeted integration upstream of the biosynthetic gene in question (Fig. 5a). USER cloning has been demonstrated to enable quick assembly of such vectors for targeted promoter replacement in F. graminearum (Frandsen et al. 2008). The pRF-HU2E vector can be easily equipped with suitable homologous sequences upstream from the target gene, enabling promoter swapping to the constitutive A. nidulans PgdpA in front of PKSs and NRPSs. This activated production of gibepyrones A, B, D, and G and polypyrone B (PKS8) (Westphal et al. 2018a), chrysogine (NRPS14) (Wollenberg et al. 2017), orsellinic acid and orcinol (PKS14) (Jørgensen et al. 2014), and three novel bostrycoidin anthrones (PKS3) (Frandsen et al. 2016). Overexpression of FgNRPS4 leads to an increase in surface hydrophobicity, but no specific SM responsible for this phenotype could be identified by chemical analyses (Hansen et al. 2012a). Comparison of knock-out mutants to the wild type in the F. heterosporum PKS69 pathway failed to identify differences in the SM profile on different growing media. However, fusing a copy of the fsdS (PKS69) gene with the constitutive equisetin synthase (PKS18) promoter in a mutant construct resulted in formation of fusaridione A, which is likely the first intermediate in the biosynthetic pathway (Kakule et al. 2013).

Fig. 5
figure 5

Targeted activation strategies utilizing Agrobacterium tumefaciens-mediated transformation. a Targeted promoter replacement. b Targeted ectopic overexpression of transcription factor gene. Here shown targeting PKS12 locus to identify mutants from lack of pigmentation

Full size image

Targeted activation can also aim to activate transcriptional regulator genes. Biosynthesis gene clusters often contain a Zn(II)2Cys6-domain gene that acts as a cluster-specific transcription factor (Brown et al. 2007; Brakhage 2013). Examples are the Gip2, Bik5, and Fsr6 transcription factors controlling pigment biosynthesis in F. graminearum and F. fujikuroi (Kim et al. 2006; Studt et al. 2012; Wiemann et al. 2013). Exchanging the native promoter of putative transcription factor APS2 for the β-tubulin promoter in F. semitectum resulted in upregulation of NRPS31 cluster genes and increased formation of apicidin (Jin et al. 2010). Likewise, overexpression of the fusaric acid cluster intrinsic Zn(II)2Cys6 transcription factor Fub10 upregulated gene expression of all cluster proteins including PKS21 (Fub1) and the NRPS-like enzyme Fub8, resulting in product formation. Deletion of the second cluster-specific transcription factor gene FUB12 abolished product derivatization (Studt et al. 2016a). Analysis of BGC promoter regions with the Regulatory Sequence Analysis Tool (RSAT) can reveal conserved transcription factor binding motifs, suggesting that expression is regulated by a single Zn(II)2Cys6 binuclear transcription factor (van Helden et al. 2000; Sørensen et al. 2012a; Sieber et al. 2014; Frandsen et al. 2016).

To ease the process of identifying overexpression mutants, one method has been to amplify the transcription factor genes including the native terminator and fusing it to the pRF-HUEA expression cassette with homologous targeting segments in the F. graminearum PKS12 locus, resulting in albino mutants (Fig. 5b) (Frandsen et al. 2008, 2016). This system was used to overexpress the putative transcription factor Fsr7, resulting in increased formation of three novel toxins: fusarielins F, G, and H (Sørensen et al. 2012a). To ensure high expression, targeted integration into a non-coding locus adjacent to the β-tubulin gene in F. graminearum (Josefsen et al. 2012) has been used for AurR1 overexpression, enabling overproduction of aurofusarin biosynthesis metabolites including novel putative shunt products (Westphal et al. 2018b). Combined overexpression of PKS39 and the cluster-specific transcription factor gave a tenfold increase in metabolite production and enabled characterization of a novel group of metabolites: fujikurins B, C, and D (Wiemann et al. 2013; Von Bargen et al. 2015). Fungal metabolites and their intermediates can be toxic to the producer and it may be necessary to use an inducible expression system. In F. fujikuroi, controlled overproduction of the silent trichosetin gene cluster was obtained by placing the cluster-specific transcription factor gene TF22 under regulation of the inducible, tetracycline-dependent tet-on promoter (Janevska et al. 2017). A novel strategy for production of silent SM genes was developed with elements of the highly producing equisetin polyketide BGC in F. heterosporum. The equisetin synthase eqxS is under regulation of the cluster-specific transcription factor eqxR. eqxR was fused with the inducible/leaky alcA promoter, and the bidirectional promoter peqxS was fused to lovB and lovC from the A. terrus lovastatin nonaketide biosynthesis cluster. Transformation of F. heterosporum with this multi-gene vector construct succeeded in production of the expected lovastatin precursor metabolites. This approach was similarly used to express an uncharacterized biosynthetic pathway in the endophytic fungus NRRL 50135, resulting in the isolation of the anti-tuberculosis agents pyrrolocins A and C (Kakule et al. 2015).

Activation through global regulators and histone modification

The secondary metabolism of filamentous fungi is controlled by complex regulatory network of proteins responding to environmental conditions such as substrate, pH, light and temperature, excellently reviewed by Axel A. Brakhage (2013). In contrast to cluster-specific transcriptional regulator proteins activating a small number of genes, global regulatory proteins control expression of secondary metabolism on higher level (Wiemann and Keller 2014). In approximately 40% of fungal gene clusters, there are no identifiable TF present (Brakhage 2013), and therefore, targeted activation strategies cannot be performed. In such cases, alternative strategies can be used such as manipulation of histone-modifying enzymes or global transcriptional regulator genes (Bok and Keller 2004; Butchko et al. 2012; Giese et al. 2013).

Deletion of the COMPASS protein Ccl1 in F. graminearum and F. fujikuroi significantly altered secondary metabolism (Studt et al. 2017). In both species, SMs produced by genes localized near telomeres were upregulated. These chromosomal regions often have low gene expression mediated by trimethylation of the H3K4 (H3K4me3) and proteins from the COMPASS complex (Palmer and Keller 2010; Zhao et al. 2014). Disruption of the heterochromatin methyltransferase Kmt6 led to transcriptional activation of four novel putative BGCs in F. fujikuroi, leading to isolation of a novel sesquiterpene (Studt et al. 2016b). Likewise, F. graminearum Δkmt6 displayed a drastic change in secondary metabolism profile (Connolly et al. 2013). In F. fujikuroi, the global regulator protein Sge1 is responsible for activation of a number of SM pathways including gibberellic acids, bikaverin, fumonisins, apicidin F, fusarins, and fusaric acid (Michielse et al. 2014, 2015; Studt et al. 2016a). Bikaverin biosynthesis is repressed in nitrogen-rich and ambient pH conditions. Expression of bik genes was repressed by pH-related transcription factor PacC, and a knock-out mutant ΔpacC displayed significant increase in expression of bikaverin cluster genes in comparison with wild-type F. fujikuroi (Wiemann et al. 2009). In addition, the overexpression of the global nitrogen regulator AreA in F. fujikuroi resulted in higher titers of bikaverin, even under repressing conditions (Linnemannstöns et al. 2002). However, global regulators do not result in activation of all biosynthetic genes. In the hunt for the FgNRPS5-NRPS9 product, the overexpression of the known regulator of secondary metabolism FgLaeA did not results in activation of the BGC. Instead, Jia et al. 2019 showed that the overexpression of the cluster-specific transcription factor fgm4 ectopically was required for the formation of the novel virulence factor fusaoctaxin A (Jia et al. 2019). Global regulators may suppress activated gene clusters, but probably cannot in themselves activate clusters. To ensure effectiveness, both types of activation are probably required for a successful outcome.

Chemical analysis

Identification of changes to the secondary metabolome by the aforementioned approaches relies on robust and accurate chemical analyses. High-performance liquid chromatography (HPLC) is often a chosen chemical separation of polyketides and non-ribosomal peptides due to their chemical properties, although gas chromatography was recently used in linking PKS8 to gibepyrones in F. fujikuroi (Janevska et al. 2016). Through the advances in high-resolution mass spectrometry (HRMS), several automated methods have been developed, which have become very helpful in deciphering which fungal SMs are produced (Nielsen et al. 2010). The identification of novel compounds from a complex fungal extract can be accelerated by fast and accurate identification of already known and characterized compounds. This can be achieved through dereplication, where chromatographic and spectroscopic methods are coupled with searches in existing databases (Nielsen et al. 2011). The usefulness of this approach was demonstrated by Klitgaard et al. (2014), who developed a method for automated identification of up to 3000 fungal secondary metabolites (Klitgaard et al. 2014).

Although dereplication holds great potential for working with SM discovery, it has not been widely adapted to studies of Fusarium. MS-based dereplication was used to determine the SM profiles among selected strains from F. solani species complex isolated from human infections (Short et al. 2013). This study led to the identification of several SMs, including citreoisocoumarin, YCM1008A, and lucilactaene, which had not been reported from members of the species complex before. NMR-based dereplication has been used on Fusarium strains isolated from the rhizosphere of Senna spectabilis in which fusaric acid and beauvericin were discovered in F. oxysporum and the depsipeptide HA23 in F. solani (Selegato et al. 2016).

The advances in chemical analyses have also enabled increased use of stable isotopes for elucidation of biosynthetic pathways. A feeding experiment in A. niger with fully labeled 13C8 6-methylsalicylic acid (6-MSA) was used to propose a biosynthetic route ending with yanuthone D as the end product (Holm et al. 2014). A similar approach was used in F. avenaceum, which was fed with 13C14-YWA1 (Klitgaard et al. 2015), the first stable intermediate formed during biosynthesis of aurofusarin. The feeding experiment in F. avenaceum showed that aurofusarin was derived from YWA1, but more interestingly, antibiotic Y (avenacein Y) was also identified as being derived from YWA1. This compound is well known from F. avenaceum, but had not been linked to a biosynthetic pathway. Subsequent genome analyses showed that F. avenaceum contains a gene (aurE, FAVG1_08663) located centrally in the aurofusarin gene cluster, which is not present in the aurofusarin gene cluster in F. graminearum. This gene is predicted to encode an epoxide hydrolase, which could be involved in antibiotic Y biosynthesis.

Outlook

This review has summarized the recent advances in deciphering biosynthetic pathways in Fusarium. The quest is far from finished, as the products are currently unknown for the majority of gene clusters. Ongoing global gene-expression analyses will offer further insight into the biosynthesis of SMs through identification of co-regulated genes (Brown et al. 2012; Jørgensen et al. 2014). Proteomics may supplement this data, as shown in an analysis of an overexpression mutant in F. graminearum (Westphal et al. 2018b). The majority of metabolites are now identified through gene orthology rather than chemical isolation and analysis (Brown and Proctor 2016; Hoogendoorn et al. 2018). Gene homology analyses are now an inherent part of modern metabolomics research due to the strength and swift application of tools such as BLASTP (Altschul et al. 1990) and AntiSMASH (Blin et al. 2017), and resource collections, e.g., the Minimum Information about a Biosynthetic Gene cluster (MIBiG) repository (Epstein et al. 2018). Bioinformatic approaches have helped researchers to predict the product of a Fusarium BGC (Varga et al. 2005) before it was experimentally deduced (Tobiasen et al. 2007; Oide et al. 2014). Currently, a handful of putative metabolites has been assigned to species of Fusarium based on homology (Gaffoor et al. 2005; Hansen et al. 2012b, 2015; Wiemann et al. 2013; Brown and Proctor 2016; Hoogendoorn et al. 2018; Janevska and Tudzynski 2018). Metabolites assigned to Fusarium spp. through homology comprise: depudecin (PKS17), solanapyrone (PKS44), tenellin/fumosorinone (PKS45), alternapyrone (PKS52), 3-methylorsellinic acid (PKS54), oxononal benzaldehyde (PKS55-PKS64), mellein (PKS56), hexadehydro-astechrome (NRPS42), and fumarylalanine (NRPS43) (Hansen et al. 2015; Brown and Proctor 2016; Hoogendoorn et al. 2018). Although prediction tools have proven reliable, many of the predicted metabolites remain to be detected in the fungal organism by chemical analyses. Gene comparisons may be useful in risk assessment as exemplified by the observation of a putative mycotoxin producing synthase in the genome of the biological control strain F. oxysporum Fo47 (Hoogendoorn et al. 2018).

Looking ahead, the majority of detected Fusarium PKS and NRPS BGCs remain to be characterized (Hansen et al. 2015) and their product pathways elucidated. Deletion of biosynthetic gene-producing unwanted metabolites may increase the flux of precursors to a target pathway, resulting in higher yields of the desired compounds (Chiang et al. 2013). Elimination of endogenous biosynthesis pathways may also decrease metabolic background noise in LC–MS, and ease purification of novel products. To control the regulation of endogenous metabolism, two recent studies report novel Cre-loxP-based systems for F. graminearum enabling rapid and effective gene-targeting strategies and selection marker recycling (Connolly et al. 2018; Twaruschek et al. 2018). Major efforts have until now focused on key species such as F. graminearum and F. fujikuroi. The genomes of F. solani and F. avenaceum comprise several untapped PKS and NRPS BGCs not found elsewhere in the metagenome. With the introduction of optimized ATMT protocols for both (Sørensen et al. 2014b; Romans-Fuertes et al. 2016), we expect to contribute new information on SM products to the collected Fusarium metabolome. In addition, we expect that expression in a non-natural host (Munawar et al. 2013; Rugbjerg et al. 2013) will contribute to assessment of enzymatic function.

We hope that this review has shown that data and molecular tools are now available to get insight into the vast number of SMs in Fusarium and evaluate their potential as leads in the biomedical sciences and their impact in the environment.