Aspergillus flavus Secondary Metabolites: More than Just Aflatoxins

Abstract

Aspergillus flavus is best known for producing the family of potent carcinogenic secondary metabolites known as aflatoxins. However, this opportunistic plant and animal pathogen also produces numerous other secondary metabolites, many of which have also been shown to be toxic. While about forty of these secondary metabolites have been identified from A. flavus cultures, analysis of the genome has predicted the existence of at least 56 secondary metabolite gene clusters. Many of these gene clusters are not expressed during growth of the fungus on standard laboratory media. This presents researchers with a major challenge of devising novel strategies to manipulate the fungus and its genome so as to activate secondary metabolite gene expression and allow identification of associated cluster metabolites. In this review, we discuss the genetic, biochemical and bioinformatic methods that are being used to identify previously uncharacterized secondary metabolite gene clusters and their associated metabolites. It is important to identify as many of these compounds as possible to determine their bioactivity with respect to fungal development, survival, virulence and especially with respect to any potential synergistic toxic effects with aflatoxin.

1. General Introduction

Chemical analysis of fungal developmental structures such as hyphae, conidia, fruiting bodies and sclerotia has uncovered a plethora of secondary metabolites, many of which are secreted into the environment¹⁾. Secondary metabolites (SMs) are low molecular weight natural products that often display bioactive properties and are produced from precursors derived from primary metabolism. Secondary metabolites are, by definition, not essential to the producing organism and for the most part are characterized based on their impact on human and animal health. Fungal SMs are quite diverse in their chemistry and their bioactivities, being sources of beneficial therapeutic compounds such as penicillins and lovastatin as well as harmful toxins such as aflatoxins and fumonisins. The evolutionary advantages of the production, sequestration and secretion of bioactive SMs by fungi has been the subject of decades of scientific speculation. Though not essential for growth like primary metabolites, SMs do impart advantages to the producing organism. Fungi are in a constant battle with both predator and competitor organisms that they must be able to coexist, compete and interact with in their particular environmental niche. These niches, be they the human body or an agricultural field are dynamic environments subject to change where fungi must be able to adapt in order to compete for available nutrients allowing them to reproduce and disseminate thus ensuring survival. Fungi have evolved the capability to produce a wide array of SMs in response to their environments many of which play important biological roles such as virulence factors, chemical defense agents, insect attractants, protection from abiotic stressors, developmental regulators, and as chemical signals for communication with other organisms.

Aspergillus flavus is a saprophyte in soils worldwide where it survives by obtaining nutrients from dead or decaying organic matter. In soils, A. flavus can survive as propagules in the form of as asexual conidia or overwintering structures termed sclerotia that can germinate when environmental conditions become conducive to growth to form hyphae and conidiophores. With the discovery of a sexual stage, A. flavus sclerotia were shown to function as fruiting bodies, termed stromata, that harbor the sexual ascospores²⁾. A. flavus is a frequent contaminant of agricultural crops such as maize, peanut, cottonseed, and a number of tree nuts, both pre- and post-harvest, where it produces the toxic and carcinogenic group of SMs known as aflatoxins (AFs)³⁾. Contamination of susceptible crops with AFs poses serious health risks in developing countries and significant economic losses in the U.S. and other developed countries^4,5). AF contamination of crops usually results from opportunistic invasion by the fungus due, in most cases, to mechanical or insect damage of the plant cell wall⁶⁾.

The A. flavus genome is estimated to be approximately 37 Mb and capable of encoding about 13,000 predicted functional genes⁷⁾. Though best known for its production of AFs, in silico analysis of the A. flavus 3357 genome sequence utilizing the web-based Secondary Metabolite Unique Regions Finder (SMURF) algorithm (http://www.jcvi.org/smurf) has shown that it has the potential to produce numerous other SMs. Using SMURF, Georgianna et al.⁸⁾ predicted the genome of A. flavus to contain 55 SM gene clusters. The total number of SM clusters in A. flavus can be increased to 56 when including the kojic acid (KA) gene cluster identified originally in A. oryzae by Marui et al.⁹⁾. However this may be an underestimation, as other SM gene cluster prediction programs have identified even greater numbers of gene clusters in A. flavus¹⁰⁾. Although about 40 SMs have been identified from A. flavus cultures^11–13) (Table 1), metabolites have only been experimentally assigned to twelve of the predicted 56 A. flavus SM clusters (see section 5). Elucidation of the products of the remaining SM gene clusters is confounded by the fact that many are transcriptionally silent, inactive due to gene mutations or expressed at extremely low levels thus making identification of their respective products problematic under laboratory growth conditions. These types of gene clusters are termed cryptic or orphan and identification of their metabolite(s) cannot be achieved using classical knockout of a cluster gene followed by comparative metabolic analysis of the mutant and the wild-type by liquid chromatography-mass spectrometry (LC-MS). Analysis of ribonucleic acid sequencing (RNA-Seq) data from a number of A. flavus strains growing on artificial media indicates that approximately half of the gene clusters are silent or expressed at very low levels¹⁸⁾. A number of methods have been used to overcome silent gene clusters thereby enabling identification of some previously unknown cluster metabolites^19–21). Determination of the chemical structure and biological activities of metabolites produced by these uncharacterized gene clusters will undoubtedly provide invaluable insights into the ecology of A. flavus as well as other fungal plant pathogens harboring similar gene clusters. It is important to identify as many of these SMs as possible to determine their bioactivity with respect to fungal development, survival, virulence and especially with respect to any potential synergistic toxic effects with AF.

Table 1. A. flavus secondary metabolites.

Compound	Class	Reference
3-nitropropionic acid	propionate	12
4-hydroxymethyl-5-hydroxy-2H-pyran-2-one	pyranone	13
aflatoxins	difurocoumarolactone	12
aflatrem	indole diterpene	12
aflavarin	bicoumarin	13
aflavazole	indole diterpene	12
aflavinines	indole diterpene	12
asparasone	anthraquinone	13
asperflavin	anthrone	12
aspergillic acids	pyrazinone-hydroxamic acid	12
asperipin-2a	cyclic peptide	14
aspirochlorines	epithiodiketopiperazine	13
cladosporin	isocoumarin	12
cyclo(D-N-methyl-Leu-L-Trp)	diketopiperazine	13
cyclopiazonic acid	indole-tetramic acid	12
ditryptophenaline	diketopiperazine	12
ferriaspergillin	pyrazinone-hydroxamic acid	12
flavoglaucin	quinol	13
flufuran	furan	13
gliotoxin	epithiodiketopiperazine	13
indole	indole	13
isoflavipucine	pyridone	12
itaconic acid	succinate	13
kojic acid	pyranone	13
kotanin	bicoumarin	13
leporins	pyridone	15
lna and lnb piperazines	piperazine	16
lovastatin	nonaketide	13
mellein	isocoumarin	13
miyakamides	didehydrotryptamine-dipeptide	13
oxylipins	fatty acid	13
parasiticol	difurocoumarolactone	13
parasiticolide A	terpene	13
paspaline	indole diterpene	12
paspalinine	indole diterpene	12
penicillin	beta-lactam	13
sterigmatocystins	difuroxanthone	13
ustiloxin B	cyclic peptide	17

In this review, we present information on the identification and characterization of SM gene clusters and their concomitant metabolites from A. flavus. A number of gene clusters in A. flavus are highly homologous to gene clusters from other filamentous fungi for which the associated metabolite has been identified. However, this review will only discuss those SM gene clusters for which the attendant metabolite(s) have been experimentally confirmed in A. flavus. Information will be presented on the genomic, biochemical, bioinformatic and molecular methods used to elucidate the genes and their encoded enzymes/proteins that are responsible for biosynthesis of SMs in A. flavus. Due to the many comprehensive reports available on AFs^3,22,23) and cyclopiazonic acid (CPA)^24,25), these two SMs will not be covered in this review. Instead, we will focus on recently characterized A. flavus SM gene clusters and their metabolites. In addition to discussion of the genetics and chemistry of A. flavus SMs and their gene clusters, the potential roles these metabolites play in the biology of the fungus will also be presented.

2. Functional Genomics and Secondary Metabolism

Research approaches in fungi utilizing functional genomics may entail a range of molecular, biochemical or bioinformatic tools, however, most often the term describes research aimed at determining gene or protein function and interaction utilizing a genome-wide analysis of data. Such experiments include a bottom-up approach where specific genes are targeted for disruption or ectopic expression and the resulting effect on transcriptional regulation or biological processes is determined^26–29). Other common approaches include the use of chemical treatment, environmental conditioning, host-pathogen interaction or other stimuli relevant to the investigation to identify novel genes, proteins, interactions, expression profiles, and entire processes responding to the stimuli^30–33). Alternatively, functional genomics can utilize genome-wide information to elucidate gene function with techniques such as such as gain-of-function cDNA screens³⁴⁾, small interfering RNA (siRNA)-based knockdown genome libraries³⁵⁾, or clustered regularly interspaced short palindromic repeats (CRISPR)-Cas9 approaches designed to target a majority of the transcriptome³⁶⁾. In addition, proteomics and metabolomics have incorporated functional genomics as a component of large scale screening approaches, such as mass-spec analysis of 2D protein gels combined with genome data^37,38), high-throughput yeast two hybrid interaction assays³⁹⁾, carbon flux algorithms correlated with gene expression data⁴⁰⁾ and transcriptome-metabolome co-expression analysis⁴¹⁾.

Through the use of knockout technologies, microarray analysis, RNA-sequencing, and in silico analysis, SM cluster prediction is continually evolving to become more targeted and efficient in its predictive power. Section 2 of this review will focus on the functional genomic tools using genome-wide expression analysis to elucidate secondary metabolic gene cluster activation and regulation in A. flavus.

2-1. Microarray Technology

Microarray technology was among early approaches characterizing A. flavus gene function⁴²⁾, natural antisense transcripts⁴³⁾, and secondary metabolism, with an initial focus on AF regulation^44–48). Subsequent to this, the examination of other SM gene clusters using customized microarrays were reported^8,29,49,50). An early effort at this was by Georgianna et al.⁸⁾ whereby a collection of microarray data from 28 different treatments was used to comprehensively associate cluster expression patterns depending on treatment and environmental conditions. Here four clades, or expression patterns, of the 55 SMURF-identified clusters was described, with about half of the clusters determined to be low or non-expressing at all conditions analyzed. Another clade was characterized by high expression at most conditions and included clusters 23 (leporins), 55 (CPA), 26 (unknown) and 36 (unknown). Cluster 26 was also identified by Ehrlich and Mack¹⁸⁾ and Wang et al.⁵¹⁾ as differentially expressed on solid media and during pathogen-host interaction in peanut, respectively. The study by Georgianna et al.⁸⁾ was limited in that it could not account for the effect of accessory enzymes, but provided one of the first comprehensive profiling of secondary metabolic genes in A. flavus.

In 2013, Reverberi et al.⁴⁹⁾ examined 24 SMURF-identified gene clusters that were described to be up or down regulated in two separate comparisons. In the first, A. flavus grown in liquid media with and without autoclaved maize kernels indicated up regulation of what is now known to be the gene clusters producing a pyrazinone (cluster 11), aflatrem (cluster 15), leporins (cluster 23), and down regulation of the AF (cluster 54) cluster. A second comparison whereby A. flavus was grown on autoclaved kernels and live kernels in the field identified effects of 16 different clusters, including the pyrazinone, leporin, CPA and AF clusters. Cary et al. observed the altered co-expression of clustered genes using microarray analysis of a veA knockout strain to identify the sclerotia-related clusters that synthesize aflavarin²⁹⁾ and asparasone⁵²⁾. Both aflavarin (cluster 39) and asparasone (cluster 27) were among the clusters predicted to be sclerotia-related in the Georgianna et al.⁸⁾ microarray analysis and by Ehrlich and Mack¹⁸⁾ based on the temporal correlation of sclerotia development and cluster expression.

2-2. RNA Sequencing

RNA-sequencing technologies are quickly replacing most other means of high-throughput transcriptomics, given their decreasing cost, high accuracy, and the increased availability of bioinformatic services^53–55). Ehrlich and Mack¹⁸⁾ was among the first to utilize RNA-sequencing to characterize secondary metabolic gene clusters in A. flavus, primarily for annotation purposes in identifying absolute expression activity and defining cluster boundaries. The authors identified eight clusters demonstrating high expression levels under their conditions and eleven clusters correlating with sclerotial development. In addition to asparasone and aflavarin mentioned previously, the cluster for leporin production was also among those predicted¹⁵⁾.

To date there has not been numerous use of RNA-seq combined with gene knockout technology in A. flavus, however RNA-seq was conducted on the small sclerotia producing A. flavus strain AF70 to examine the effects of knocking out the zinc-finger transcription factor nsdC. Inactivation of nsdC resulted in decreased aflatrem production, as well as decreased expression of genes encoding the backbone enzymes for penicillin, asparasone, ustiloxin B, and AF biosynthesis²⁶⁾. Clusters 5, 20, 41, 44, and others also showed altered transcriptional profiles in the nsdC knockout strain. Metabolites produced by these clusters remain to be identified.

In addition to gene knockout strategies, chemical treatment and the manipulation of environmental conditions combined with RNA-seq technology has successfully characterized various gene clusters in A. flavus. Treatment of A. flavus with resveratrol, a naturally occurring phytoalexin and antifungal compound, affected 17 genes in 13 clusters, 2 of which were nonribosomal peptide synthetases (NRPSs) or NRPS-like enzymes (cluster 3 and 45, respectively)⁵⁶⁾. Although fungi are thought to not undergo DNA methylation⁵⁷⁾ the DNA methyltransferase inhibitor 5-azacytidine has been speculated to interfere with toxisome formation⁵⁰⁾. Upon treatment of A. flavus with 5-azacytidine, RNA-seq revealed an increase in expression of several genes of cluster 27 (asparasone), cluster 35 (piperazine), cluster 55 (CPA) and moderately decreased expression of AF cluster genes⁵⁸⁾. In another RNA-seq study, the plant volatile, decanal, impacted expression of genes present in 26 of the 55 clusters⁵⁹⁾. Finally, because drought-stress leads to increases in reactive oxygen species, a RNA-seq study examined the effects of oxidative stress (exposure to H₂O₂) on 3 strains of A. flavus synthesizing varying levels of AF. The focus of this study on secondary metabolic gene products besides AF was limited, however the authors did indicate genes in the aflatrem and KA gene clusters were affected⁶⁰⁾.

Recently the use of functional genomics, and in particular RNA-sequencing, to measure the genome-wide response to changing environmental conditions has become more widespread^30,61), and whereas early experiments focused on the AF gene cluster, secondary metabolic cluster prediction and transcriptome-wide profiling has allowed for more complete analysis. The influence of temperature or water activity combined with RNA-seq has been used to examine small non-coding RNAs⁶²⁾, development⁶³⁾ and SM production^64,65). Yu et al.⁶⁵⁾ showed 10 of the 55 gene clusters had higher expression at 30°C as compared to 37°C (cluster 1, 10, 11, 19, 20, 21, 24, 45, 54, and 55) and cluster 3 decreasing at lower temperatures. Medina et al.⁶⁴⁾ found genome-wide, a decreased transcriptome response to changing environmental conditions correlating with low water activity or high temperature, with water being the primary predictor. As an example, the CPA cluster (cluster 55), was lower in every condition examined due to low water availability, regardless of temperature conditions. Preliminary or small-scale studies have indicated that other environmental conditions, for example, those related to climate change also impact secondary metabolism^61,66,67). Studies are ongoing to examine this impact on a genome-wide scale.

2-3. Bioinformatic Tools Used for Cluster Prediction in A. flavus

The most commonly used bioinformatic tools for RNA-seq analysis such as KEGG (Kyoto Encyclopedia of Genes)⁶⁸⁾, gene co-expression network analysis⁶⁹⁾ and gene ontology (GO) enrichment analysis⁷⁰⁾ have the potential to facilitate the analysis of large data sets and have been used in functional genomics analysis of A. flavus^26,52,56,58). While these methods will categorize genes, determine categories significantly enriched, and identify similar gene expression patterns, the identification of cluster function is generally beyond these tools. To facilitate cluster prediction in A. flavus multiple computational approaches have been developed⁷¹⁾(Table 2). During development and testing, A. flavus clusters were used as test sequences for the motif-dependent program SMURF, and the motif-independent programs: Motif-Independent De Novo Detection of Secondary Metabolite Gene Clusters (MIDDAS-M), and Motif-Independent Prediction without Core Genes (MIPS-CG). Among the more popular programs, SMURF, developed by Khaldi et al.⁷²⁾ at the J. Craig Venter Institute, uses hidden Markov Model (HMM) searches to predict the presence of TIGRFAM and Pfam conserved protein domains associated with various SM backbone enzyme genes. These include NRPSs, polyketide synthases (PKSs), hybrid NRPS-PKS enzymes and prenyltransferase enzymes (DMATS, dimethylallyltryptophan synthases). SMURF predictions are limited in the type of cluster predicted, but are generally regarded as consistent with other forms of prediction^73,74). While SMURF has demonstrated significant effectiveness as evidenced by the publications it has contributed to, the last update was April 2015 and technical support to resolve discrepancies appears unavailable^75,76).

Table 2. Comparison of categories of A. flavus predicted secondary metabolite gene clusters using different web-based computational programs.

Category	SMURFa	antiSMASHb	MIDDAS-Mc
PKS (Type I and III)	25	20	15
NRPS	18	11	12
PKS-NRPS hybrid	2	3	3
DMAT/Terpenesd	8	13	7
NRPS-Like	14	−	20
PKS-Like	3	2	−
Indolee	−	5	4
RiPS	−	−	1f

^aPre-computed data from www.jcvi.org/smurf. Differences from Georgianna et al. (2010)⁸⁾ data could be contributed to changes in the SMURF algorithm or sequences used.

^bObtained by running antiSMASH v. 4.0.0 (www.fungismash.secondarymetabolites.org) using A. flavus NRRL 3357 genome sequence (EQ963472).

^cSupplementary data obtained from Umemura et al. (2013)¹⁰⁾.

^dantiSMASH categorizes DMAT's as terpenes, while SMURF defines them independently.

^eIndole clusters in antiSMASH determined by hidden Markov modeling using the StaD-like chromopyrrolic acid synthase domain (Medema et al. 2011)⁷⁸⁾.

^fMIDDAS-M identified the ustiloxin B cluster. A. flavus genome analysis has identified an additional 7 potential Ri

Alternatively, the popular algorithm antibiotics and Secondary Metabolite Analysis Shell (antiSMASH), is subject to frequent upgrades and user input, with versions made available as recently as April 2017^77–79). The software serves as a pipeline that integrates multiple published analysis types to predict a wide range of cluster types and enzyme domains, including PKS/NRPS’s, terpenes, indoles, siderophores, and in recent versions, Ribosomally-synthesized and Post-translationally modified Peptides (RiPPs). At the core of antiSMASH, like SMURF, amino acid sequence translations are identified as containing signature domains using hidden Markov Models, and regions up to 20kb away are searched for accessory enzymes. Other programs within antiSMASH utilize trained and phylogentic-based algorithms to predict substrate structure based on motifs within acyltransferase or adenylation domains, from which a final structure is derived based on other domains present. In early versions the “greedy” approach of boundary prediction led to excessively large clusters being predicted, however recent versions have attempted to optimize this element by incorporating sequence-based algorithims (CASSIS and ClusterFinder) that use known regulatory motifs and protein domains to assist in border prediction. Additionally, antiSMASH incorporates data from the Minimum Information about a Biosynthetic Gene cluster (MiBIG) database which attempts to incorporate experimentally-determined SM data.

Finally, MIDDAS-M does not utilize characterized motifs as a basis for prediction¹⁰⁾. Rather, it utilizes both genomic and transcriptomic data to calculate expression ratios of neighboring genes, and large induction ratios to indicate cluster activity. This is beneficial when attempting to identify novel clusters that do not contain motifs searched for by other programs. The authors claim that even SM biosynthetic genes not co-localized, such as those synthesizing KA in A. oryzae can be detected⁸⁰⁾. The ustiloxin B biosynthetic gene cluster is unique in that the peptide is ribosomally synthesized. The ustiloxin B cluster was detected by MIDDAS-M, which was later experimentally-validated¹⁷⁾. A comprehensive comparison of MIDDAS, antiSMASH and SMURF using A. flavus was conducted by Umemura et al⁸⁰⁾. We have not found other software programs utilized extensively for cluster predictions in A. flavus. In our own experience, this is sometimes due to difficulty of use, inaccurate cluster predictions, a failure to maintain basic but essential updates beyond publication, or technical defects (bugs) with no technical support.

2-4. Cluster Prediction in the Age of Genomics

With the recent increase in large SM and genomic data sets and databases such as ClusterMine360, IMG-ABC, and MIBiG, strategies using comparative and computational genomics and transcriptomics for cluster prediction are becoming more effective^73,74,81). These approaches must go beyond using genes in other species of Aspergillus to predict A. flavus SM gene clusters, primarily because the degree of homology needed and role of accessory proteins and “hypotheticals” is unknown. Certainly more than backbone gene identity is important. For example, PKS enzymes often share high degrees of similarity with unrelated PKSs. The PKS-encoding gene afvB (AFLA_108550) in A. flavus shares identity (E value =0, 43.8% aa identity) to the A. nidulans PKS encoded by mdpG (AN0150), however the A. nidulans product monodictyphenone is structurally dissimilar to aflavarin⁸²⁾, and none of the aflavarin cluster accessory enzymes are thought to be located near mdpG. Thus Basic Local Alignment Search Tool (BLAST) or bioinformatic approaches must involve a multigene approach.

In A. flavus, one example of such an approach is the discovery that ustiloxin B is ribosomally synthesized¹⁷⁾ and that two genes in the cluster, ustYa/Yb, are found in 19 loci in the A. flavus genome, 9 of which are associated with a ustiloxin-like precursor sequence (ust-RiPS)¹⁴⁾. In the case of one of these loci, knockout of the ustYa/Yb genes led to the identification of the novel compound, asperipin-2a¹⁴⁾. The authors here developed a pipeline utilizing PSI-BLAST, which iteratively refines searches based on conserved amino acids, signal peptide detection software, and BLASTClust to further sort precursor proteins. Another method utilized in A. flavus by our group involves using known eukaryotic clusters in the MIBiG database⁸³⁾ as queries against sequenced strains of A. flavus. The software program multigeneblast⁸⁴⁾ will search a genome or set of sequences using multiple sequences as the query, thus allowing for gene rearrangement and differing degrees of identity. Hits can then be associated with the known metabolite from the database. Using this approach we can putatively identify several uncharacterized clusters in A. flavus. For example, all 22 of the genes in the experimentally verified A. oryzae aspirochlorine cluster are present as a cluster in A. flavus (cluster 21). Similar results are obtained for tenellin, chaetoglobosin, and trans-resorcyclide with varying levels of identity. As databases such as this increase in user contribution, predictive powers of such tools is likewise expected to increase.

3. Regulation of Secondary Metabolism in A. flavus

3-1. Global Regulation

The genetic mechanisms that regulate secondary metabolism in fungi have been the subject of intense investigation^21,85). In this section, we focus on the regulation of SM production in A. flavus along with recently published literature that has not been presented in previous review articles. The role of LaeA, a putative histone methyltransferase, as a global regulator of development and SM production in A. flavus as well as other Aspergillus species, has been the subject of several reports^86,87). Deletion of laeA in A. flavus resulted in complete loss of aflatoxin B₁ (AFB₁), aflatoxin B₂ (AFB₂), and CPA production, whereas, over-expression of laeA did not have a significant effect on production of these SMs⁸⁶⁾. The presence of other SMs including paspaline, paspalinine, aflatrem, and aflavinines (see section 5–5) were difficult to detect in wild-type A. flavus, however, in the A. flavus laeA over-expressor strain, levels were increased such that they could be readily detected⁸⁶⁾. Kojic acid (KA) production was lost in A. flavus ΔlaeA mutants and levels were not increased due to over-expression of laeA. Kojic acid production was also abrogated in an A. oryzae ΔlaeA mutant due to down-regulation of KA biosynthesis genes⁸⁸⁾. In addition, laeA has also been shown to regulate the A. flavus lna and lnb gene clusters (see section 5–4) involved in the production of piperazine and morpholine SMs¹⁶⁾.

The light-responsive protein, VeA is a component of a heterotrimeric, nuclear-localized protein complex, designated the velvet complex, which also includes LaeA and the velvet-like protein VelB. The velvet complex is responsible for regulation of development and SM production in numerous filamentous fungi and was shown to positively regulate the production of AF, CPA, and aflatrem in A. flavus^89,90). Production of CPA was reduced by 48–66% in the ΔveA mutant (vs. control) depending upon growth in light or dark conditions. An A. flavus ΔvelB mutant was unable to produce sclerotia and AFs²⁸⁾. A transcriptomic study in A. flavus showed that 28 SM gene clusters contain at least one gene whose expression is differentially regulated by the presence or absence of veA²⁹⁾. Characterization of the aflavarin gene cluster (see section 5–3) showed that veA directly regulates the production of aflavarin with all of the cluster genes significantly down-regulated in a ΔveA mutant, demonstrating direct regulation of this cluster by VeA. Functional characterization of the asparasone A gene cluster (see section 5–1) in A. flavus showed a significant down-regulation of the polyketide synthase gene (pks27) and the Zn(2)-Cys(6)-type transcription factor gene znf27 in a ΔveA mutant resulting in loss of asparasone production⁵²⁾.

Another global regulatory gene, nsdC, mainly known for its role in development in A. nidulans⁹¹⁾, was also shown to be required for AF production in A. flavus²⁷⁾. In addition to loss of AF biosynthesis, A. flavus ΔnsdC mutants also demonstrated reduced aflatrem content compared to a control strain. RNA-seq analysis of the ΔnsdC mutant revealed differential expression of genes associated with development and production of SMs such as AFs, penicillin, and aflatrem²⁶⁾. The studies presented here suggest a coordinated action between global regulators and cluster specific transcription factor genes in regulating SM production in A. flavus.

The Far transcription factor was shown to play an important role in fungal virulence and AF production through its involvement in lipid metabolism. The A. nidulans farA and farB genes were shown to regulate expression of genes associated with fatty acid catabolism⁹²⁾. Further characterization of far genes in A. flavus showed their essential roles in mitochondrial beta oxidation of fatty acids and regulation of genes involved in unsaturated fatty acid biosynthesis⁹³⁾. Deletion mutants of A. flavus farA and farB showed aberrant growth phenotypes when grown on medium-chain fatty acids. Aflatoxin production was significantly decreased in a ΔfarB mutant (vs. control) both in vitro and during maize seed infection accompanied by reduced sporulation, suggesting a role of farB in lipid metabolism contributing to fungal pathogenesis and AF production. Another Zn(2)-Cys(6) transcription factor gene, aswA, has been demonstrated to affect sclerotial production and associated sclerotial SMs in A. flavus⁹⁴⁾. The ΔaswA mutant produced higher numbers of deformed, non-pigmented sclerotia and no asparasone A, aflatrem, dihydroxyaflavinine, or aflavarin could be detected from sclerotia as compared to the control strain. Aflatoxin and CPA content were not affected by the aswA mutation. In a recently published study, Cary et al.⁹⁵⁾ demonstrated that a homeobox transcription factor gene, hbx1, is required for production of conidiophores, sclerotia, and the SMs AF, CPA, and aflatrem in A. flavus. Upon forced overexpression, a globally-acting C₂H₂ zinc-finger transcription factor, MtfA, was shown to be required for normal development and AF production in A. flavus⁹⁶⁾.

Besides DNA-binding transcription factors, other regulatory factors with diverse functions have been found to regulate secondary metabolism in A. flavus. Examples include RmtA, a putative arginine methyltransferase, shown to be required for production of normal levels of AFB₁ and other uncharacterized SMs⁹⁷⁾. Similar regulation of SM production in A. flavus was also observed by RtfA, a putative RNA-Pol II transcription elongation factor, in coordination with VeA and LaeA⁹⁸⁾. The ΔrtfA mutant exhibited a significant decrease in sclerotia production along with reduction in AF and other unknown SMs.

3-2. Cluster-specific Regulation

In addition to global regulation of SM gene clusters, cluster-specific transcription factors are also involved in regulation of SM production in A. flavus. One of the most studied cluster specific transcriptional regulators is the aflR gene present in the AF gene cluster. The aflR gene is a DNA-binding, Gal4-type Zn(2)-Cys(6) transcriptional activator that regulates expression of most AF cluster genes^99–101). A. flavus aflR deletion mutants do not make AFs due to lack of expression of cluster genes. AflR binds to the palindromic sequence 5ʹ-TCGN5CGA-3ʹ present in the promoter region of most AF biosynthetic genes^102,103). Aflatoxin biosynthesis is also regulated by aflS (aflJ) located adjacent to and divergently transcribed from aflR¹⁰⁴⁾. The exact role of aflS in AF biosynthesis is not clear despite intense investigation. Deletion of aflS in A. flavus did not appear to affect expression of AF cluster genes however aflS deletion mutants did not produce AF¹⁰⁴⁾.

Inactivation of the A. flavus leporin cluster (see section 5–2) Zn(2)-Cys(6) transcription factor gene, lepE, resulted in significant down-regulation of all the cluster genes¹⁵⁾. Over-expression of lepE produced orange-red pigmented hyphae due to the synthesis of the 2-pyridone, leporin B. The asparasone A gene cluster contains a Zn(2)-Cys(6) transcription factor gene, znf27, that when inactivated resulted in transformants that produced lighter, reddish-brown sclerotia compared to the dark brown sclerotia of the control strain. The reduced sclerotial pigmentation was due to lower levels of asparasone A production in the znf27 mutant. Gene expression analysis showed a significant reduction in expression of asparasone cluster genes in the ∆znf27 mutant compared to the control indicating that znf27 functions as a pathway-specific transcription factor. A pathway-specific C6-type transcriptional regulator gene present in the ustiloxin B gene cluster (see section 5–6), encoded by ustR, was identified and its overexpression resulted in an increase of ustiloxin B production by about 5-fold while deletion abrogated ustiloxin production^14,17). Disruption of KA gene cluster (see section 5–8) Zn(2)-Cys(6) transcriptional activator gene, kojR, resulted in a significant down-regulation of expression of KA biosynthetic genes kojA and kojT accompanied by a complete loss of KA production⁹⁾. Over-expression of kojR resulted in >3-fold higher expression of KA biosynthetic genes along with >3-fold increase in KA production compared to the control strain demonstrating direct regulation of KA biosynthesis by KojR.

3-3. Methods to Activate Silent Secondary Metabolite Gene Clusters

In fungi and other micro-organisms, considerable interest lies in understanding the production of SMs by cryptic or silent gene clusters that only produce their associated metabolites when exposed to specific environmental conditions not normally duplicated during growth on standard laboratory media. Several methods have been used to activate silent SM gene clusters in fungi and these have been summarized in a number of excellent reviews^{21,105–107)}. These include over-expression of global or pathway-specific transcriptional regulators, expression of SM cluster genes in heterologous hosts, co-culture of a fungal strain with another micro-organism, proteomics, exposure to epigenetic modifiers and antibiotics and alteration of media composition. In the genus Aspergillus, most experiments to activate silent SM gene clusters have been conducted in the model species, A. nidulans, so there is a need to extend these approaches to A. flavus.

4. Chemistry of Secondary Metabolites

4-1. Chemistry of Cluster Backbone Genes

Secondary metabolites produced by fungi exhibit a broad range of chemical diversity. Several classes of biosynthetic backbone genes give rise to a plethora of complex chemical structures that impact human health and agriculture. These backbone genes encode for enzymes, most of which are multifunctional megasynthases, that biosynthesize the metabolites’ core structure. Tailoring enzymes then operate on the core structure to afford a diverse suite of SMs with varying functions, many of which have been shown to enhance the fitness of the fungi. Table 3 provides a list of features associated with experimentally characterized secondary metabolite gene clusters in A. flavus.

Table 3. Features of A. flavus secondary metabolite gene clusters.

Metabolitea	Cluster numberb	Backbone gene classification	Cluster-specific transcriptional regulator	Number of genes in clusterc	Putative biological function	Representative structure	Reference
Aflatoxin	54	PKS	Yes	30	antiinsectan/oxidative stress resistance?		3,22,23
Aflatrem	15 and 32	DMATs/GGPPs	No	5 and 3	antiinsectan/ antifeedant		108,109
Aflavarin	39	PKS	No	5	antiinsectan/sclerotial development		29
Asparasone A	27	PKS	Yes	4	antiinsectan/sclerotial pigment/abiotic stress resistance		52
Cyclopiazonic acid	55	DMATs/hybrid PKS-NRPS	No	5	mammalian antifeedant?		24,25
Ditryptophenaline	4	NRPS	No	3	?		110
Kojic acid	NA	oxidoreductase	Yes	3	insect antifeedant/ antioxidant?		111
Leporin	23	hybrid PKS-NRPS	Yes	9	insect antifeedant/ sclerotial development		15
Piperazine	35 and 48	NRPS-like	No	6 and 6	sclerotial development		16
Ustiloxin B	31	RiPPS	Yes	15	?		10,17

^aSome clusters produce more than one metabolite.

^bBased on Georgianna et al.⁸⁾ numbering system. NA-not applicable as the kojic acid cluster was not identified by SMURF analysis.

^cAs defined experimentally or as defined in Georgianna et al.⁸⁾.PS (Nagano et al. (2016)¹⁴⁾.

4-1-1. Polyketide synthases (PKSs)

Fungal polyketide metabolites are biosynthesized by iterative, type I PKSs and have highly diverse chemical architecture^112–114). A type I PKS contains multiple modular catalytic domains encoded by one gene. Three domains are necessary for polyketide biosynthesis: the acyltransferase (AT) domain binds acetyl-CoA or malonyl-CoA starter units, the ketosynthase (KS) domain catalyzes a Claisen condensation that extends the polyketide chain with additional malonyl-CoA or methylmalonyl-CoA units, while the acyl carrier protein (ACP) tethers the elongating polyketide chain to the PKS through a phosphopantetheine linkage. Iterative PKSs use the same AT and KS catalytic domains to add multiple malonate units to the growing polyketide. Additional reductive tailoring of the growing polyketide chain can occur if the PKS harbors functional ketoreductase (KR), dehydratase (DH), or enoyl reductase (ER) domains. Typically, a thioesterase (TE) domain severs the thioester-linked PKS-polyketide bond, liberating the nascent polyketide as a linear molecule or a macrolactone. After release from the PKS, the polyketide is further transformed by oxidases and oxygenases resulting in SMs with abundant structural complexity¹¹⁵⁾. PKSs in A. flavus produce AFs, aflavarin, and asparasone A (see sections 5–1 and 5–3).

4-1-2. Nonribosomal peptide synthetases (NRPSs)

Fungal nonribosomal peptides are biosynthesized by NRPSs from proteinogenic and non-proteinogenic amino acids^113,116). NRPSs contain multiple catalytic domains arranged in modules. Each module contains an adenylation (A) domain that activates a specific amino acid using ATP, a peptidyl carrier (P) domain, which utilizes a phosphopantetheine linker analogous to ACP domains in PKSs, and a condensation (C) domain that catalyzes peptide bond formation. Multiple modules are arranged in an assembly-line fashion whereby each module adds one amino acid to the growing peptide chain. Many NRPSs contain a TE termination domain that catalyzes peptide release. In some fungal NRPSs, the peptide chain is liberated from the NRPS by a condensation-like termination domain (C_T), which affords a cyclic peptide^117,118). “NRPS-like” proteins contain A and T domains but lack C domains and terminate with a thioester reductase (R) domain that liberates the tethered amino acid or peptide as an aldehyde. NRPSs in A. flavus produce ditryptophenaline and NRPS-like enzymes produce several piperazine natural products (see sections 5–4 and 5–7).

4-1-3. Hybrid backbone genes

Hybrid PKS-NRPSs biosynthesize SMs the have both polyketide and amino acid components. A. flavus utilizes a hybrid PKS-NRPS enzyme (KS-AT-ACP-C-A-T-R) that forms an intermediate in CPA acid biosynthesis from a mevalonate and two acetate moieties using PKS domains and a tryptophan residue using NRPS domains¹¹⁹⁾. A. flavus also produces leporins using a hybrid PKS-NRPS (see section 5–2).

4-1-4. Dimethylallyl tryptophan synthases (DMATs)

Fungal dimethylallyl tryptophan synthases alkylate tryptophan or other aromatic moieties with the isoprenoid substrate dimethylally pyrophospate¹²⁰⁾. A DMATS in A. flavus prenylates an intermediate in the biosynthesis of CPA^119,121). The biosynthetic pathway of aflatrem in A. flavus also contains a DMATS-catalyzed prenylation (see section 5–5).

4-1-5. Meroterpenes

Meroterpenes are SMs that are partially derived from terpenoid biosynthesis pathways. Indole diterpenes are meroterpenes that contain a cyclic diterpene skeleton derived from geranylgeranyl diphosphate (GGPP) and an indole moiety derived from indole-3-glycerol phosphate¹²²⁾. Biosynthesis of these molecules is complex and involves a GGPP synthase as well as variety of oxidases. A. flavus produces an assortment of indole diterpenes including aflatrem and aflavinines (see section 5–5).

4-2. Methods of Analysis

Robust chromatographic techniques have been used extensively to detect mycotoxins and other SMs produced by fungi. Thin-layer chromatography (TLC) is the most accessible and, due to its simplicity and low cost, the most widely-used technique, especially for preliminary metabolite investigation¹²³⁾. Normal-phase TLC is commonly used for mycotoxin analysis and utilizes non-polar solvent systems to separate metabolites over silica gel adsorbent. TLC plates pretreated with a fluorescent indicator allow UV absorbing compounds to be easily detected. Non-UV active and non-fluorescent compounds can be detected by treating the plate with various developing reagents¹²⁴⁾. High-performance liquid chromatography (HPLC) affords better separation, usually on a C18 reverse-phase column, and when paired to a diode-array detector (DAD) provides diagnostic UV spectra for each peak^125,126). Fungi can be identified using chemotaxinomic data acquired from organic fungal extracts run on an HPLC-DAD instrument. Coupling HPLC-DAD to a mass spectrometer (MS) provides more information for identification^127,128). High resolution mass spectrometry (HRMS) allows molecular formula estimation and simplifies dereplication¹²⁹⁾. Databases of characterized natural products searchable by monoisotopic mass, such as Antibase, can assist in dereplication¹³⁰⁾. HPLC-DAD-MS dereplication of SM production produced by genome-sequenced strains of A. oryzae (RIB40) and A. flavus (NRRL 3357) revealed that, although bioinformatics analysis shows 99.5% gene homology between the strains, the overall chemical profiles differed significantly¹³¹⁾. Orbitrap HRMS and multiple stage mass spectrometry (MSⁿ) was used in conjunction with SIEVE software to identify differentially expressed metabolites in control and mutant A. flavus strains (see section 5–1)¹³²⁾. Similar analytical methodology was developed using ultra-high-performance liquid chromatography (UHPLC)-Orbitrap HRMS that allows untargeted determination of fungal metabolites and led to the characterization of A. flavus cluster 23 metabolites (see section 5–2)¹³³⁾.

One-dimensional (1D)-nuclear magnetic resonance (NMR) data, usually ¹H and ¹³C spectra, are often used to confirm metabolite identification. Structure elucidation, particularly atom connectivity, of new natural products is reliant on various 2D-NMR techniques¹³⁴⁾. Purification of metabolites is usually required for structure elucidation, however, recently, several metabolites have been identified and structurally characterized using NMR without having been isolated in pure form¹³⁵⁾. 2D-NMR analysis of unpurified extracts can also be used to investigate an organisms’ metabolome¹³⁶⁾. Differential analysis by 2D-NMR spectroscopy (DANS)^137,138) was employed to examine the gliotoxin biosynthesis pathway (gli) in A. fumigatus, which revealed many new gli-dependent metabolites¹³⁹⁾.

5. Experimentally Identified A. flavus Secondary Metabolite Gene Clusters

5-1. Asparasone Cluster

Asparasone A, a PKS-derived 1,3,6,8-tetrahydroxy-2-(1′-hydroxy-3′-oxobutyl)-anthraquinone was shown to be a precursor molecule required for the formation of the dark pigment associated with the outer rind of sclerotia⁵²⁾. Asparasone A was first described in A. parasiticus extracts as a pigment that was structurally related to versicolorin intermediates involved in the biosynthesis of AFs and demonstrated to possess antiinsectan properties¹⁰⁸⁾. The asparasone A gene cluster consists of four genes, a polyketide synthase (AFLA_082150, pks27), a Zn(2)-Cys(6)-type transcription factor (AFLA_082140, znf27) and a putative high-affinity glucose transporter (AFLA_082160, mfs1) and MFS transporter (AFLA_082170, mfs2)⁵²⁾. A. flavus Δpks27 mutants produced greyish-yellow sclerotia rather than having the dark brown pigment presented by control strains (Fig. 1). Complementation of the ∆pks27 mutant with a wild-type copy of pks27 restored production of dark-pigmented sclerotia. Production of sclerotia was significantly reduced in the ∆pks27 mutant compared to that of the wild-type. Ultra-high performance liquid chromatography and mass spectrometry revealed that asparasone A (m/z (M-H)=357.0612) was still being produced by the ∆znf27 mutant though at significantly reduced levels which probably explained the reduction in observed sclerotial pigmentation. While function of the putative mfs1 and mfs2 transporters was not confirmed, they were considered to be a part of the asparasone cluster due to their coordinate regulation with pks27 by znf27.

Fig. 1.

Microscopic examination of sclerotia. Sclerotia production in A. flavus SRRC1582 wild-type (WT) and a ∆pks27 mutant was observed using an Olympus SZH10 stereomicroscope (x96 magnification) and Nikon DS-Qi1 camera. Strains were grown on GMMS agar for 10 days in the dark at 30°C. Note loss of dark pigmentation in sclerotia of the ∆pks27 mutant due to abrogation of production of asparasone A.

The characteristic dark pigmentation associated with most fungal sclerotia was previously thought to be due to a process of melanization. However, the study of Cary et al.⁵²⁾ showed the dark brown pigments present in A. flavus sclerotia are the product of PKS27-derived anthraquinones rather than typical tetrahydronapthalene-based melanins (Fig. 2). The presence of the asparasone A-based anthraquinone pigment in A. flavus sclerotia was shown to provide resistance to both biotic and abiotic stressors. In both choice and no choice feeding assays with the sap beetle Carpophilus freemani, a preference for consumption of sclerotia produced by the ∆pks27 mutant compared to control sclerotia was observed indicating that the dark pigment functions as an insect antifeedant. Exposure of sclerotia to UV irradiation or high temperatures resulted in decreased viability of sclerotia from the ∆pks27 mutant compared to the control strain.

Fig. 2.

Proposed biosynthesis of asparasone A in A. flavus. Pks27 assembles the linear nonaketide from an acyl started unit and eight malonyl units. Cyclization and release from the PKS affords the anthrone. The anthrone is oxidized to the anthraquinone and one ketone moiety reduced by unknown enzymes that are predicted to be similar to those in the aflatoxin biosynthesis gene cluster. Adapted from Cary et al⁵²⁾.

5-2. Leporin Cluster

Of the 55 A. flavus SM gene clusters identified by SMURF analysis, two were found to harbor hybrid PKS-NRPS backbone genes. One of these clusters was shown to be responsible for the production of the indole tetramic acid, cyclopiazonic acid¹⁴⁰⁾ while the second cluster produced a family of 2-pyridones known as leporins¹⁵⁾. TePaske et al.¹⁴¹⁾ reported on the production of the antiinsectan compound, leporin A, in the fungus Aspergillus leporis, however there had been no previous reports of leporin production in A. flavus. In addition to the PKS-NRPS (AFLA_066840, lepA), the leporin cluster was predicted to encode an additional 14 genes that could be involved in metabolite production¹⁵⁾. An A. flavus ∆lepA mutant showed no significant differences in metabolite production compared to that of an isogenic control strain, indicating that the genes responsible for leporin biosynthesis were not being expressed or expressed at extremely low levels. However, overexpression of a putative Zn(2)-Cys(6) cluster-specific transcriptional regulatory gene, lepE (AFLA_066900) resulted in transformants producing a distinctive orange-red pigment (Fig. 3). Liquid chromatography and high resolution mass spectrometry (LC-HRMS) analysis of the OE::lepE extracts revealed a metabolite of m/z (M+H)=352.1902 that represented a demethylated leporin A precursor designated leporin B¹⁵⁾. Additionally, a metabolite of m/z (M+H)=336.1959 was also produced that represented the N-demethoxy derivative of leporin A, designated leporin C. While leporin B had been previously reported to be produced by an unidentified fungus¹⁴²⁾, this was the first report of production of leporin C. Interestingly, metabolites were identified in the OE::lepE strain that were consistent with homodimers of both leporin B and C.

Fig. 3.

Production of leporins by A. flavus. Images of broth cultures of an A. flavus CA14 OE::lepE strain and CA14 wild-type (WT) following 2 days growth on Czapek-Dox broth with agitation in the dark at 30°C. Note orange-red pigmentation of the OE::lepE culture broth indicative of the formation of leporin B-iron complexes.

The biosynthesis of leporin B and C was elucidated using LC-HRMS analysis of extracts from knockout mutants of the leporin cluster genes in the A. flavus OE::lepE background¹⁵⁾ (Fig. 4). The products of both lepA (PKS-NRPS) and lepG (enoyl reductase) are responsible for fusion of phenylalanine with a hexaketide and subsequent release of the stable tetramic acid precursor, pre-leporin C. The putative cytochrome P450, LepH, an enzyme demonstrating significant homology to the ring expansion cytochrome P450s found in the 2-pyridones, tenellin and bassianin, was shown to be required for the ring expansion step. LepF, a putative short chain dehydrogenase/reductase, was hypothesized to reduce the acyl carbonyl to an alcohol¹⁴³⁾. Interestingly, although lepI was predicted to encode a putative methyltransferase, elegant in vitro studies by Tang and coworkers¹⁴³⁾ showed that LepI catalyzed an S-adenosyl-L-methionine (SAM)-dependent inverse electron demand hetero-Diels–Alder (HDA) reaction forming leporin C. In addition to the HDA reaction, LepI was shown to possess retro-Claisen activity, which allows the enzyme to convert a byproduct of the Diels–Alder reactions to leporin C. Leporin C is then further oxidized by the N-hydroxylase, LepD, to form leporin B. The function of LepA, G, H, F, I and D were confirmed by heterologous reconstitution in A. nidulans; the function of LepF and LepI were investigated in vitro using recombinant protein expressed from S. cerevisiae and E. coli, respectively¹⁴³⁾.

Fig. 4.

Proposed biosynthesis of leporin B. The hybrid PKS-NRPS LepA and the LepG enoyl reductase form a tetramic acid intermediate (not shown), which is converted to a 4-hydroxypyridinone intermediate by the ring-expanding P450 LepH. LepF reduces the ketone in the intermediate to an alcohol. The S-adenosyl-L-methionine (SAM)-dependent enzyme LepI catalyzes a stereoselective dehydration of the alcohol followed by an inverse electron demand hetero-Diels–Alder reaction forming leporin C. The LepD P450 forms the hydroxamic acid moiety through N-hydroxylation resulting in leporin B. Adapted from Cary et al.¹⁵⁾ and Ohashi et al.¹⁴³⁾.

While 2-pyridones are normally colorless, it is believed that leporin B, in the presence of an atom of Fe(III) is oxidized to form a highly stable leporin B trimer with an atom of iron (0) (m/z = 1107, iron(0)-trioxyleporin B) which results in the observed orange-red colony pigmentation. It was not determined if oxidation of leporin B to form the trimer is self-catalyzed or catalyzed by another enzyme residing within or outside of the leporin gene cluster. Such a trimeric iron chelate had not been previously described for a 2-pyridone and unlike typical iron-chelating siderophores, the leporin B-iron complex was highly soluble in organic solvents thus representing the first siderophore-like molecule with this property to be produced by a PKS-NRPS.

Though leporin A from A. leporis was shown to have antiinsectan properties there were no previous reports on the insecticidal properties of leporins B or C. In no-choice insect feeding studies performed with a preparation of the leporin B-iron chelate (LIC) compound extracted from the OE::lepE strain, a significant reduction in growth of first instar larvae of Fall armyworms (Spodoptera frugiperda) and Corn earworms (Helicoverpa zea) was observed¹⁵⁾. In choice assays, both caterpillar species were observed feeding more frequently on the control disks than 0.1% LIC disks, and consumed significantly less of the LIC diet disks than the control diet disks. These results indicated that the LIC was acting as an antifeedant with little evidence of antiinsectan properties. Alterations in the developmental programs were also noted between the OE::lepE strain compared to the control¹⁵⁾. A 10-fold reduction in conidia production was observed in the OE::lepE strain and sclerotial production was either reduced or delayed depending on the growth medium used compared to the control. Studies have shown that conidiation and sclerotial differentiation in filamentous fungi including A. flavus are induced by oxidative stress^144,145). Cary et al.¹⁵⁾ speculated that the large quantities of leporin B produced by the OE::lepE strain may have led to an increase in sequestration of iron, in the form of the leporin B-iron chelate, thus reducing oxidative stress caused by excess free iron. The reduction in oxidative stress levels, in turn, resulted in a decrease in the production of both conidia and sclerotia.

5-3. Aflavarin Cluster

Comparative transcriptomics of an A. flavus ΔveA mutant and isogenic control strain led to the identification of the aflavarin gene cluster in which all but one of the genes were significantly down-regulated in the mutant in at least one of the conditions assayed²⁹⁾. Like many previously characterized A. flavus SMs, aflavarin was isolated from sclerotia and shown to exhibit antiinsectan activity however the biosynthetic genes responsible for its production were unknown¹⁴⁶⁾. SMURF analysis predicted the aflavarin cluster to be composed of a backbone PKS gene (AFLA_108550, afvB) encoding an N terminal beta-ketoacyl synthase (KS) domain with an adjacent acyl transferase (AT) domain and a C-terminal malonyl CoA-acyl carrier protein transacylase (ACP) domain. Other veA-dependent genes present in this cluster included an O-methyl transferase (AFLA_108560, afvC), methyl transferase (AFLA_108570, afvD), cytochrome P450 (AFLA_108580, afvE), and a NADH oxidase (AFLA_108540, afvA). A blastp search of the NCBI and the Broad Aspergillus Comparative Genome databases indicated that the genes constituting the aflavarin gene cluster in A. flavus were highly conserved in A. oryzae, A. niger and partially conserved in A. clavatus and A. terreus.

Comparative analysis of the metabolomes of an A. flavus control and ∆afvB PKS mutant using high resolution mass spectrometry allowed identification of a series of bicoumarins that were present in the control and absent in the mutant²⁹⁾. The major differentially produced metabolite had a calculated mass of 455 which was identical to the mass of aflavarin and this was subsequently confirmed by LC/MS comparison to an authentic sample of aflavarin. Knockout of the cluster biosynthetic genes allowed determination of their roles in aflavarin synthesis (Fig. 5). Deletion of the O-methyltransferase (afvC) gene, involved in the O-methylation of the PKS-derived pentaketide, didesmethyl-siderin, totally abolished the biosynthesis of the monomeric siderin precursors that are subsequently dimerized to form the final bicoumarin products. Disruption of the cytochrome P450 monooxygenase (afvE) led to accumulation of the monomeric coumarins and loss of bicoumarin production, proving that the cytochrome P450 monooxygenase is responsible for phenolic coupling reactions. Inactivation of the afvD methyltransferase was found to be responsible for the final O-methylation steps following formation of the bicoumarins. The putative NADH oxidase (afvA) was proposed to be involved in hydroxylation of siderin and 7-O-demethyl siderin to their respective hydroxylated analogues.

Fig. 5.

Proposed biosynthesis of aflavarin. The PKS AfvB assembles 4,7-didesmethyl-siderin from an acyl started unit and four malonyl units. AfvC can perform subsequent methylations affording siderin. Hydroxylation by AfvA gives hydroxy-siderin, which can be oxidatively coupled with siderin by the AfvE P450 to form aflavarin. An alternate pathway to aflavarin is possible by an AfvE-catalyzed coupling of hydroxydesmethyl-siderin and siderin followed by a methylation of desmethylaflavarin by AfvD. Adapted from Cary et al.²⁹⁾.

In addition to its previously demonstrated antiinsectan properties, it was also shown that production of aflavarin is necessary for normal levels of sclerotial production. While no cluster-specific transcriptional activator gene was present in aflavarin cluster, there were a number of putative binding sites for stress-responsive DNA binding proteins as well the AF cluster-specific transcriptional activator, AflR. This suggested that there may be regulatory cross-talk between these 2 clusters. (Fig. 6)

Fig. 6.

Proposed biosynthesis of piperazines. The NRPS-like enzymes LnaA and LnaB catalyze the reduction of tyrosine and form an imine dimer. LnaA and LnaB can then reduce the imine dimer to a piperazine. LnaB and LnbB can form a morpholine analog from the imine dimer intermediate, while LnaC and LnbC are proposed to oxidize the imine dimer to a pyrazine analog (not shown). Adapted from Forseth et al.¹⁶⁾.

5-4. Piperazine Clusters

Two clusters were identified in A. flavus named lna (chromosome 6) and lnb (chromosome 8) that were shown to be primarily involved in the biosynthesis of two piperazines as well as a number of other derivatives¹⁶⁾. The genetic makeup of the two clusters is quite similar being composed of genes encoding a putative NRPS protein (lnaA/lnbA), and a number of genes likely to encode tailoring enzymes such as an NMR-like protein (lnaB/lnbB), cytochrome P450s (lnaC/lnbC), alcohol dehydrogenase/NADH:flavin oxidoreductase (lnaE/lnbE), and a transporter (lnaF/lnbF). The lna cluster also contains a second putative P450 enzyme gene (lnaD) while the lnb cluster contains a gene encoding a hypothetical protein not present in the lna cluster. The LnaA and LnbA NRPSs share a fairly high level of amino acid identity (58%) and canonical adenylation, acyl carrier protein and thioester reductase domains but lack a condensation domain. No cluster-specific transcriptional regulator genes were found in either the lna or lnb clusters. However, both clusters were shown to be under the control of LaeA, a global regulator of fungal development, secondary metabolism and virulence^86,147). Searches of fungal genome databases indicated that LnaA-like proteins are present a number of Aspergillus species (though not specified) and in the genomes of the basidiomycetes, Serpula lacrymans and Heterobasidion annosum¹⁶⁾.

Comparative metabolomics of A. flavus wild-type, deletion, and lnaA overexpression (OE) strains allowed identification of two major piperazine metabolites and eight other lna cluster-derived metabolites that had not been previously reported in aspergilli or other fungi¹⁶⁾. Both LnaA and LnbA proteins were shown to activate tyrosine but because they lacked condensation domains, there was no peptide bond formation as confirmed by their inability to be activate an L-Tyr-L-Tyr dipeptide. Instead it was postulated that the C-terminal reductase domains of LnaA and LnbA act to reduce L-Tyr to its aldehyde. This suggested that the lna cluster NRPSs function as amino acid reductases characteristic of Lys2-type reductases involved in fungal L-lysine biosynthesis¹⁴⁸⁾. The resulting L-Tyr aldehyde was proposed to dimerize and be subsequently further reduced to piperazines and morpholine compounds by the activities of the NMR-like reductase LnaB/LnbB. While the function of the LnaC, LnaD or LnbC oxidases was not determined it was speculated that one or all may be involved in the conversion of the Tyr dimer to a pyrazine. Of particular interest was the finding that LnaB can participate in the reduction of LnaA-derived intermediates indicating that the lna and lnb clusters encode partially redundant biosynthetic pathways capable of sharing biosynthetic intermediates.

Suppression of lnaA and lnbA expression in A. flavus resulted in significant reduction in sclerotial formation compared to the wild-type suggesting a role for the SM products of these clusters in fungal development¹⁶⁾.

5-5. Aflatrem Clusters

Similar to piperazine biosynthesis, the genes required for the synthesis of the potent tremorgenic toxin, aflatrem, are also located at loci present on two different chromosomes (5 and 8) in A. flavus¹⁴⁰⁾. Aflatrem and its isomer, β-aflatrem, are examples of indole-diterpenes (IDTs), a class of structurally-diverse SMs that are found mainly in Aspergillus and Penicillium spp. as well as Epichloë. Identification of the genes required for aflatrem biosynthesis (atm) in A. flavus and A. oryzae was made possible through comparative genomics with genes from Penicillium paxilli required for the production of the indole-ditepene, paxilline^109,149). As observed in A. flavus, the genes required for paxilline biosynthesis (pax) are clustered in 2 loci present on separate chromosomes. A total of three pax gene homologs (atmG, atmM and atmC) comprise the A. flavus ATM1 locus while atmD, atmQ, atmB, atmA and atmP comprise the ATM2 gene cluster locus. The presence of these pax gene homologs in A. flavus provided support for the biosynthesis of aflatrem proceeding via a paxilline intermediate.

Indole-diterpenes are all derived from a common core structure consisting of a tetracyclic diterpene derived from the activities of a geranylgeranyl diphosphate synthase (GGPPS, PaxG/AtmG) and geranylgeranyl transferase (GGT, PaxC/AtmC) that combines GGPP with an indole moiety derived from indole-3-glycerol-phosphate to form 3-geranylgeranyl indole (GGI)¹⁴⁰⁾. GGI is then converted to an epoxide by the action of an epoxidase (PaxM/AtmM) and subsequently transformed by an IDT cyclase (PaxB/AtmB) to the first stable intermediate, paspaline^150,151). In A. flavus, paspaline is then converted to paspalinine by the action of the cytochrome P450 monooxygenases AtmP and AtmQ followed by subsequent conversion to aflatrem by the prenyltransferase AtmD. Interestingly, the chemical diversity that is characteristic of IDTs is due in large part to the IDT cyclases (IDTCs). Using a yeast host, Tang et al.¹⁵¹⁾ were able to propose mechanisms by which A. flavus utilizes unique IDTCs to produce both aflatrems and the IDT, aflavinine (Fig. 7). They proposed that AtmG/C/M produces a common epoxide intermediate which is then converted to paspaline by the aflatrem cluster IDTC, AtmB, thus leading to subsequent formation of aflatrem. Based on the ability of A. flavus to also produce aflavinine, Tang et al.¹⁵¹⁾ performed a phylogenetic analysis of all atmB homologs in A. flavus to identify candidate IDTCs that may be responsible for formation of aflavinine. When they introduced the standalone (not associated with a typical SM gene cluster) atmB homolog, afB, into a yeast host expressing atmG/C/M, the major product was shown to be aflavinine. These findings indicated that the standalone IDTC, AfB, catalyzes an alternate set of cyclization reactions leading to the production of aflavinine. How the fungus regulates expression of IDTCs such as atmB and afB to selectively control the ratios of aflatrems to aflavinines produced is unclear but it may have to do with the desired biological activity inherent in the particular IDT. Interestingly, the yeast studies indicate that the product of the atmA gene is not required for aflatrem or aflavinine biosynthesis.

Fig. 7.

Proposed biosynthesis of indole diterpenes in A. flavus. Two indole diterpene cyclases (AtmB and AfB) work in conjunction with the epoxidase AtmM to produce diverse chemical structures from a common intermediate. Genes encoding for biosynthesis of the epoxide intermediate (AtmG, AtmC, and AtmM) are clustered together (ATM1). Genes encoding for biosynthesis of paspaline, paspalinine, and aflatrem (AtmB, AtmP, AtmQ and AtmD) are also clustered together (ATM2), while the gene encoding AfB, responsible for aflavinine biosynthesis, is not clustered with either ATM1 or ATM2. Adapted from Nicholson et al.¹⁴⁰⁾ and Tang et al.¹⁵¹⁾.

Expression of atm genes at both ATM loci were shown to be coordinately controlled despite their presence on separate chromosomes. However, no cluster-specific transcriptional regulatory genes were found in or near either of the ATM loci and this was also the case for the pax loci in P. paxilli¹⁴⁰⁾. This indicated that control of expression of atm genes probably occurs at the global level. In fact, it was shown that the global regulators of A. flavus development and secondary metabolism, VeA and LaeA, are both required for production of wild-type levels of aflatrem^86,90).

Aflatrems and aflavinines are both associated with sclerotia in A. flavus¹⁵²⁾. Aflatrem was shown to be a potent mammalian tremorgen¹⁵³⁾ while aflavinines demonstrated antiinsectan properties¹⁵²⁾. The biological advantage that aflatrem imparts on the fungus is not clear though β-aflatrem has been shown to have some antiinsectan activity¹⁴⁶⁾. The mechanism of action of aflatrems and aflavinines remains to be determined but their antifeedant/antiinsectan properties suggest a role to increase survival of sclerotial propagules in the field from insect predation, thus allowing eventual germination to produce more fungal spores for dissemination.

5-6. Ustiloxin B Cluster

The proliferation of next generation sequencing of fungal genomes coupled with web-based bioinformatics tools (e.g. SMURF and antiSMASH) has been a major driver in the identification of secondary metabolic gene clusters in filamentous fungi. However, these prediction algorithms are based in large part on the presence of typical SM cluster backbone genes such as PKSs and NRPSs. A novel algorithm created by Umemura et al.¹⁰⁾, designated MIDDAS-M was able to predict fungal SM gene clusters lacking a backbone gene based on integration of annotated genome and transcriptomic data. Using this approach, they were able to identify a gene cluster in A. flavus predicted to consist of 18 genes, none of which were identified as a typical cluster backbone gene. Comparative metabolomics of gene knockout mutants of this cluster compared to a control strain identified a metabolite of m/z (M+H)=646.240 that corresponded to the antimitotic cyclic peptide, ustiloxin B.

Ustiloxins were originally identified in the rice pathogen, Ustilaginoidea virens, but were not known to be produced by A. flavus^154,155). These compounds are tetrapeptides with ustiloxin B consisting of the amino acids Tyr-Ala-Ile-Gly (YAIG). Enzymes encoded by the gene cluster cyclize the tetrapeptide at the side chains of Tyr and Ile and catalyze side chain modifications and linking of a non-proteinogenic amino acid, norvaline, to the Tyr moiety¹⁷⁾ (Fig. 8). Ustiloxin B was initially thought to be the product of a NRPS-like enzyme, but no NRPS-like genes were present in or near the gene cluster. Upon further examination, the cluster gene ustA, was predicted to encode a protein containing an N-terminal signal sequence and 16-fold repeated peptide region containing YAIG repeats which represent the exact tetrapeptide sequence of ustiloxin B. Biosynthesis was proposed to proceed via processing of the UstA precursor protein by a Kex2 protease not present in the cluster that cleaves at nearby Lys-Arg residues on the C-terminal side of the YAIG tetrapeptide¹⁷⁾. Subsequent work in A. oryzae confirmed that inactivation of the kexB endopeptidase gene resulted in almost complete loss of ustiloxin B¹⁵⁶⁾. The partially processed pre-protein is then believed to be processed in a stepwise manner by the products of ustP and ustH, both encoding peptidases, resulting in generation of the mature YAIG tetrapeptide. Prior to removal of the N-terminal leader peptide by UstP/UstH, the pre-protein is believed to be modified and cyclized by the action of the monooxygenases UstF1/UstF2 and UstQ (tyrosinase), respectively, to yield the intermediate ustiloxin F. The intermediate is then further modified by methylation (UstM), linkage of the norvaline to the Tyr residue (UstF and UstS) and oxygenation of the sulfur group (UstC). The functions of a number of genes considered as part of the ustiloxin B gene cluster were not determined in the study of Umemura et al.¹⁷⁾. The absence of an NRPS-like gene in the cluster along with a predicted protein (UstA) harboring 16 short YAIG repeat peptide regions and an N-terminal signal peptide suggested that ustiloxin B is biosynthesized through a ribosomal peptide synthetic pathway (RiPS) not previously described in A. flavus or other Ascomycetes¹⁷⁾.

Fig. 8.

Proposed biosynthesis of ustiloxin B. UstA is processed by the endoproteinase Kex2 at the C-terminus in the Golgi apparatus resulting in 16 copies of the precursor peptide. The C-terminus of each precursor peptide is then cleaved by peptidases UstP and UstH to afford a nonapeptide containing a leader and core peptide (underlined). While still attached to the leader peptide, the core YAIG peptide is cyclized. The leader peptide is then cleaved by UstP and UstH affording ustiloxin F. Methylation by UstM is follow by attachment and modification of norvaline by UstF, S, D, and C yielding ustiloxin F. Adapted from Umemura et al.¹⁷⁾.

A subsequent study by Nagano et al.¹⁴⁾ showed that deletion mutants of ustQ or ustYa/ustYb failed to produce the pathway intermediate ustiloxin F, suggesting that all three gene products are involved in cyclization of the intermediate. Interestingly, Nagano et al.¹⁴⁾ screened Aspergillus genome sequences for the presence of the ustYa/ustYb and ustA gene homologs and were able to predict the presence of 94 ribosomal peptide precursor candidate genes that could be classified into 40 types having a variety of core peptide sequences. From this computational analysis they were able to experimentally identify a novel bicyclic peptide designated asperipin-2a in A. flavus.

Despite their ability to inhibit polymerization of purified porcine microtubule proteins as well as mitosis in human cell lines but not bacteria or fungi, the role of ustiloxin B in the biology of A. flavus remains to be determined¹⁵⁴⁾.

5-7. Ditryptophenaline Cluster

The ditryptophenaline (DTP) cluster is responsible for the production of the dimeric diketopiperazine alkaloid, (−)-ditryptophenaline, that is derived from two molecules of tryptophan and phenylalanine¹⁵⁷⁾. Comparative metabolomics of putative A. flavus DTP cluster gene (dtp) mutants with a control strain enabled determination of the function of the cluster genes. The cluster was found to be composed of three genes encoding a NRPS (dtpA) that condenses L-tryptophan and L-phenylalanine together to form the diketopiperazines core; an N-methyltransferase (dtpB) that methylates the core structure; and a cytochrome P450 (dtpC) that is likely responsible for cyclization and dimerization of the methylated diketopiperazine to form DTP (Fig. 9). Though DTP has been shown to have weak activity as an inhibitor of the neurotransmitter, substance P, in animal models, its function in the biology of A. flavus is not known¹¹⁰⁾.

Fig. 9.

Proposed biosynthesis of dityrptophenaline. The NRPS DtpA forms the diketopiperazine cyclo(L-Phe-L-Trp) from phenylalanine and tryptophan. DtpB methylates cyclo(L-Phe-L-Trp) and the P450 DtpC catalyzes dimerization affording ditryptophenaline. Adapted from Saruwatari et al.¹⁵⁷⁾.

5-8. Kojic Acid (KA) Cluster

Though not originally confirmed in A. flavus, the KA gene cluster was experimentally identified in its close relative, A. oryzae¹¹¹⁾. Use of comparative transcriptomics of A. oryzae cultures grown under KA-inducing and non-conducing conditions, allowed the identification of a limited number of genes that may be involved in KA biosynthesis. Deletion of these candidate genes in A. oryzae followed by growth on media that is diagnostic for production of KA allowed identification of two deletion mutants that could no longer produce KA. The genes knocked out in these two mutants mapped within 3 kb of one another on the A. oryzae genome suggesting that a small gene cluster was responsible for KA biosynthesis. One of the genes (AO090113000136; A. flavus AFLA_096040) designated kojA, encodes a predicted FAD-dependent oxidoreductase while the other, kojT (AO090113000138; A. flavus AFLA_096060) is a major facilitator superfamily transporter. KojA is believed to function in the oxidation of glucose to KA though other as of yet unidentified genes may be required for complete conversion of glucose to KA¹¹¹⁾ (Fig. 10). Located between kojA and kojT is a gene designated kojR (AO090113000137; A. flavus AFLA_096050) that encodes a Zn(2)-Cys(6) transcription factor.

Fig. 10.

Proposed biosynthesis of kojic acid. The putative FAD-dependent oxidoreductase KojA is predicted to form kojic acid from glucose, possibly in conjunction with other enzymes not encoded in the gene cluster. Adapted from Terabayashi et al.¹¹¹⁾.

The earlier studies of Terabayashi et al.¹¹¹⁾ had shown when sodium nitrate was added to KA production medium, KA biosynthesis was inhibited in A. oryzae. The negative impact of sodium nitrate on KA production was supported by deletion of nrtA, encoding a transporter essential for sodium nitrate uptake in A. oryzae¹⁵⁸⁾. Deletion strains of nrtA were still be able to produce KA even in the presence of sodium nitrate and the expression of kojA and kojT were significantly higher in the nrtA mutant than the control in the presence of sodium nitrate.

Though the contribution of KA to the biology of the producing fungus is not clear at this time it appears that it may be involved as an insect antifeedant or perhaps as an antioxidant that reduces the ability of insect oxidative enzymes to detoxify AFB₁ or host plant-based antiinsectan compounds such as nicotine¹⁵⁹⁾. Compared to AFB₁, kojic acid was shown to be relatively non-toxic to mice but in insect feeding studies in combination with AFB₁, it appeared to act synergistically causing increased mortality compared to feeding of either metabolite alone¹⁵⁹⁾. The ability of KA to chelate iron and other cations may indicate that it is involved in homeostasis of these metals.

6. Conclusions

The advent of fungal genomics, bioinformatics and advanced analytical chemical techniques has resulted in a rapid expansion of the numbers of identified A. flavus SM gene clusters and their affiliated metabolites. While it took the better part of two decades to identify the aflatoxin gene cluster and the enzymes required for production of aflatoxins, genome sequencing and annotation coupled with modern molecular biological, bioinformatic and analytical techniques have allowed an additional nine A. flavus metabolites and their associated gene clusters and enzymes/regulatory genes to be described, all within a period of less than a decade. Though metabolites have been linked to most of the more readily tractable SM gene clusters in A. flavus, there remains a number of silent gene clusters for which characterization of their SM product’s form and function is lacking.

There is a growing body of evidence that points to significant roles of SMs in A. flavus biology and therefore it is critical that approaches are developed to activate silent SM gene clusters so that their impact on the physiology of the fungus can also be ascertained. While a number of techniques have been used successfully to activate silent gene clusters in the model fungus, A. nidulans, as well as other Aspergillus spp., experiments of this nature are lacking for the most part in A. flavus. More information must be gleaned on regulatory mechanisms of SM production in A. flavus such as global and cluster-specific regulators, signaling pathways, epigenetic regulation, and how these respond to the fungus’ environment. Evidence for regulatory cross-talk between A. flavus gene clusters and their ability to share biosynthetic intermediates has opened up new avenues of investigation into the regulation of SM gene cluster biosynthetic machinery. It is imperative that laboratory observations on the biological function of SMs also be translated to the field with respect to the interaction of A. flavus with the environment, especially during the host plant-fungus interaction. Additionally, little is known about the role of SMs on the interaction of A. flavus with other microbes that constitute the microbiome of the host plant and surrounding soil environment. These types of studies will provide invaluable insights into the roles of A. flavus SMs with respect to fungal virulence, survival, toxigenic potential and modulation of microbial and insect communities, all with the goal of controlling contamination of susceptible crops with A. flavus mycotoxins, particularly aflatoxins.

Abbreviations: ACP: acyl carrier protein, AFs: aflatoxins, AFB₁: aflatoxin B₁, AFB₂: aflatoxin B₂, antiSMASH: antibiotics and secondary metabolite analysis shell, AT: acyltransferase, BLAST: Basic Local Alignment Search Tool, CASSIS: Cluster Assignment by Islands of Sites, CRISPR: clustered regularly interspaced short palindromic repeats, CPA: cyclopiazonic acid, DAD: diode-array detector, DANS: differential analysis by 2D-NMR spectroscopy, DH: dehydratase, DMATS: dimethylallyltryptophan synthases, DTP: ditryptophenaline, ER: enoyl reductase, GGI: 3-geranylgeranyl indole, GGPP: geranylgeranyl diphosphate, GGT: geranylgeranyl transferase, GO: gene ontology, HAD: hetero-Diels–Alder, HMM: hidden Markov Model, HPLC: high-performance liquid chromatography, HRMS: high resolution mass spectrometry, IDT: indole-diterpene, IDTC: indole-diterpene cyclase, KEGG: Kyoto Encyclopedia of Genes, KA: kojic acid, KR: ketoreductase, KS: ketosynthase, LC-HRMS: liquid chromatography and high resolution mass spectrometry, LC-MS: liquid chromatography-mass spectrometry, LIC: leporin B-iron chelate, MiBIG: Minimum Information about a Biosynthetic Gene cluster, MIDDAS-M: motif-independent de novo detection of secondary metabolite gene clusters, MIPS-CG: motif-independent prediction without core genes, MS: mass spectrometer, NMR: nuclear magnetic resonance, NRPS: nonribosomal peptide synthetase, OE: overexpression, PKS: polyketide synthase, PSI-BLAST: Position Specific Iterated-BLAST, RiPPs: ribosomally-synthesized and post-translationally modified peptides, RNA-Seq: ribonucleic acid sequencing, SAM: S-adenosyl-L-methionine, siRNA: small interfering RNA, SMs: secondary metabolites, SMURF: secondary metabolite unique regions finder, TE: thioesterase, TLC: thin-layer chromatography, UHPLC: ultra-high-performance liquid chromatography, YAIG: Tyr-Ala-Ile-Gly