Recent advancements in the biosynthetic mechanisms for polyketide-derived mycotoxins

Authors


Abstract

Polyketides (PKs) are a large group of natural products produced by microorganisms and plants. They are biopolymers of acetate and other short carboxylates and are biosynthesized by multifunctional enzymes called polyketide synthases (PKSs). This review discusses the biosynthesis of four toxic PK, aflatoxins, fumonisins, ochratoxins (OTs), and zearalenone. These metabolites are structurally diverse and differ in their mechanisms of toxicity. However, they are all of concern in food safety and agriculture because of their toxic properties and their frequent accumulation in crops used for food and feed. The focus is on the recent advancements in the understanding of the molecular mechanisms for the biosynthesis of these mycotoxins. Several of the mycotoxin PKSs have been genetically and biochemically studied while other PKSs remain to be investigated. Multiple post-PKS modifications are often required for the maturation of the mycotoxins. Many of these modification steps for aflatoxins and fumonisins are well established while the post-PKS modifications for zearalenone and OTs remain to be biochemically characterized. More efforts are needed to completely illustrate the biosynthetic mechanisms for this important group of PKs. © 2010 Wiley Periodicals, Inc. Biopolymers 93: 764–776, 2010.

INTRODUCTION

Mycotoxins are low-molecular weight, toxic secondary metabolites that are produced by fungi, including many important pathogenic and food-spoilage species of Aspergillus, Fusarium, and Penicillium.1, 2 These toxins have been seen as causal factors associated with sickness and death in both animals and humans.2–6 The chemical toxicity and associated diseases, collectively termed mycotoxicoses, indirectly result from the ability of fungi to infect crops species, thereby contaminating the food ingested by both livestock and humans. Pathogenesis by the fungi usually occurs either before or after harvest, under optimal growth conditions, such as high moisture and high temperature. Five classes of mycotoxins are most significant in agriculture and food industry.2–6 They are aflatoxins (1, aflatoxin B1), fumonisins (2, fumonisin B1), ochratoxins (3, OT; ochratoxin A), zearalenone (4), which are derived from polyketide (PK) metabolism, and the trichothecenes (5, deoxynivalenol), whose biosynthetic pathway is of terpenoid origin (see Figure 1). The topic of this review is about the PK-derived mycotoxins, with a focus on their biosynthesis and recent advancements thereof. It should be pointed out that aflatoxins, fumonisins, OTs, and zearalenone are grouped together here mainly because of their PK origin and common importance in food safety and agriculture. There are groups of PKs that are structurally more similar and have a more similar mechanism of toxicity than these four.

Figure 1.

Chemical structures of representative mycotoxins discussed in this review including aflatoxin B1 (1), fumonisin B1 (2), ochratoxin A (3), zearalenone (4), and deoxynivalenol (5, an example of trichothecenes).

PKs are metabolites derived by the repetitive condensation of acetate units or other short carboxylic acids, via an enzymatic mechanism that is similar to that responsible for fatty acid synthesis.7, 8 These reactions are catalyzed by a group of enzymes termed polyketide synthases (PKSs). Similar to fatty acid synthases (FASs), PKSs contain several catalytic subunits (domains) that are responsible for the extension of the PK chain: β-ketoacyl synthase (KS), acyltransferase (AT), and an acyl carrier protein (ACP). Unlike fatty acid biosynthesis, where the nascent poly-β-keto chain is completely reduced with a set of β-keto processing enzymes: β-ketoacyl reductase (KR), dehydratase (DH), and enoylreductase (ER), PKs are variably reduced depending on the type and number of β-keto processing subunits (domains) present in PKS.

Based on protein architecture, PKSs are classified into three basic types.7, 8 Type I PKSs are large modular proteins containing multiple functional domains, type II PKSs are complexes of several separate protein subunits, each having a single catalytic active site, and type III PKSs are a distinct group of synthases that do not include an ACP domain. Each domain of the type I PKS typically acts only one time during the biosynthesis, whereas each of the subunits of the type II PKS acts iteratively. Fungal PKSs are unique in that they have a modular organization like Type I PKSs, but each domain acts iteratively. Thus, fungal PKSs are termed iterative type I PKSs.9 The fungal PKS, depending on the types of accessory domains present within the module, can produce various PKs with one of three reduction states: nonreduced (NR), partially reduced (PR), or highly reduced (HR) (see Figure 2). The general domain organization of a NR-PKS from the N-terminus to the C-terminus is the starter unit ACP transacylase (SAT), KS, AT, product template (PT), ACP, and a region which variably could contain, for example, thioesterase/Claisen cyclase (TE/CLC), or other chain-modification domains.9 The SAT domain is responsible for loading a starter unit that is donated by a dedicated FAS, another PKS, or an acyl CoA, for further chain elongation by the KS, AT, and ACP domains with further modifications from the processing domains. The PR-PKS contains an incomplete set of β-keto-processing domains, in addition to having the chain-extension domains: KS, AT, and ACP. The general domain organization of a PR-PKS from the N-terminus to the C-terminus is KS, AT, DH, Core (thought to take part in subunit-subunit interaction), KR, and ACP. The fungal HR-PKS contains a complete set of β-keto-processing domains (DH, KR, and ER) and yield products that are highly reduced. In some HR-PKSs, a C-methylation (C-MeT) domain can also be present. The general domain organization of an HR-PKS from the N-terminus to the C-terminus is KS, AT, DH, C-MeT, ER, KR, and ACP.9

Figure 2.

The general domain organization of (A) nonreducing PKS (NR-PKS), (B) partially reducing PKS (PR-PKS), and (C) highly reducing PKS (HR-PKS).

AFLATOXINS

Aflatoxins are the most toxic and carcinogenic compounds among the known mycotoxins.3 They are produced by several Aspergillus species and consist of at least 16 structurally related furanocoumarins, of which AFB1, AFB2, AFG1, and AFG2 are the four most abundant aflatoxins. Certain Aspergillus (A. flavus, A. ochraceoroseus, and A. pseudotamarii) species are able to produce only the B-type, whereas the other species can produce both B- and G-type aflatoxins (Aspergillus parasiticus).10 The B-type aflatoxins are characterized by a cyclopentanone E-ring (Figure 3A), whereas G-type aflatoxins have a six-membered lactone ring (Figure 3B) in place of the cyclopentanone. Under long-wave UV light, the B- and G-type aflatoxins fluoresce differently, blue and green respectively. The B2- and G2-type aflatoxins differ from AFB1 and AFG1 in that their bisfuranyl ring is saturated (Figure 3C) as opposed to the unsaturated ring found in AFB1 and AFG1, whereas AFB2a and AFG2a contain a hydrated bisfuranyl structure (Figure 3D). Some Aspergillus species (e.g. A. nidulans) produce sterigmatocystins instead of aflatoxins. However, in aflatoxin-producing species, sterigmatocystins are the penultimate precursors in aflatoxin biosynthesis.

Figure 3.

Chemical structures of aflatoxins and analogs. A, B: General structures; (C, D) functional moieties which have a saturated bond at C8/C9, additionally (D) has a hydroxyl group at C8.

Many reviews have described the aflatoxin biosynthetic genes and their regulations.11–17 The genes required for aflatoxin biosynthesis are clustered within a 70-kb region. The 3′ end of the cluster is delineated by a well-defined sugar utilization gene cluster.15 The expression of all genes (except for aflJ) is regulated by aflR, which encodes a transcription factor. aflR positively regulates the gene expression involved in aflatoxin and sterigmatocystin synthesis by binding to the consensus motif TCGN5CGR in the promoter regions of aflatoxin biosynthetic genes and activating their expression.18, 19

The biosynthesis of aflatoxins begins with the formation of a 6-carbon acyl chain (hexanoate) formed by two FASs (HexA/HexB), encoded by aflA and aflB.12 The resulting hexanoate is used as the starter unit of the type I iterative NR-PKS (PksA) encoded by aflC and is elongated to a 20-carbon chain (see Figure 4). Although still attached to the PKS, cyclization of the chain occurs that eventually leads to the formation of the first stable intermediate in aflatoxin/sterigmatocystin biosynthesis, norsolorinic acid (6). This FAS–PKS complex is sometimes referred to as norsolorinic acid synthase (NorS).20 NorS had been isolated from A. parasiticus as a 1.4 MDa protein complex. It was shown that the complex was able to form (6) in vitro in the presence of acetyl CoA, malonyl CoA, and NADPH. In this process, a hypothetical C20 PK chain is cyclized to an anthrone (hexanoyltetrahydroxyanthrone, 7), which subsequently oxidized to the anthraquinone, norsoloric acid (6). It is unclear whether the oxidation is spontaneous or enzyme-catalyzed.

Figure 4.

Biosynthetic mechanism for the early steps of aflatoxins.

One of the most exciting recent advancements in aflatoxin biosynthetic studies came from the enzymatic and structural investigations of PksA, which was undertaken by the Townsend group.21–24 These studies have advanced our understanding of how fungal NR-PKS control initiation, extension, folding, and cyclization of the nascent PK chain. The approach they took was a “deconstruction” approach, which disassembles and reassembles functional fragments of the PKS to unveil the biosynthetic roles of each of the domains. The process was facilitated by a bioinformatics method called UMA (Udwary–Merski algorithm) that predicts interdomain linker regions of PKS.25 They identified two previously unrecognized domains within fungal NR-PKS, the starter unit ACP transacylase (SAT) and the PT domain.21–24 The data obtained from site-directed mutagenesis, radioactive labeling, and kinetics experiments carried out with individual domains showed that SAT is responsible for selecting an appropriate starter unit (hexanoate in this case) to initiate PK biosynthesis. They expressed several mono- (ACP, PT, and TE/CLC), di- (PT-ACP and ACP-TE/CLC), and tri-domains (SAT-KS-AT and PT-ACP-TE/CLC) in Escherichia coli and used in combinatorial reconstitution experiments.22 Upon incubating the domains in various combinations, with hexanoylCoA and malonylCoA substrates, evidence was obtained for the roles of the individual catalytic domains (see Figure 4). The SAT-KS-AT + ACP combination, which lacks PT and TE/CLC, produced a minute quantity of the incorrectly cyclized naphthopyrone (8) and the correctly cyclized anthraquinone (6). Although a low yield was obtained, these four domains together make a correct chain length. Adding the TE/CLC domain to this mixture caused a decrease in the amount of (8) and increased the amount of the expected product (6). This result indicated that the TE/CLC domain is responsible for chain release. When the PT domain was added, either as a stand-alone domain or as a PT-ACP didomain fusion and in the absence of the TE/CLC domain, the amount of (8) and the expected product (6) increased dramatically, suggesting that the PT domain is responsible for the formation of rings A and B, but not ring C (Figure 4). Because the TE/CLC domain is not present, the result also showed that the napththopyrone (8) is formed via a spontaneous C[BOND]O bond formation, further verifying the function of the TE/CLC domain. Finally, when both the PT and TE/CLC were added, the correct product (6) was increased while the shunt product (8) was decreased proportionally. Together, the results showed that PT is responsible for the formation of rings A and B and TE/CLC releases the final product by catalyzing the formation of ring C (see Figure 4). More recently, the Townsend–Tsai team resolved the crystal structure of the PT domain.23 The PT has a “double hot dog” fold, similar to DH domains in animal FASs. The crystal structures of PT domain, bound with either palmitate or a mimic of the putative PKS-bound octaketide intermediate, enabled the discovery of a pocket that is about 30 Å deep and is made up of three distinct regions. The first region accommodates the ACP-bound linear poly-β-keto chain where, in a regiospecific manner, the catalytic dyad residues (Asp1543/His1345) act on it. Here, two aldol condensations followed by two dehydrations occur, yielding the bicyclic intermediate (A and B ring). Once the two rings are formed, the product is then moved into the second region, the cyclization chamber. The third region makes up the hexyl-binding region. To further verify that Asp1543 and His1345 were important players in the cyclization reactions, they were independently mutated to an alanine residue. Both mutations resulted in loss of activity, confirming the role of the catalytic dyad. Because the PT domain is found in many fungal NR-PKS, the knowledge that the PT domain controls the specific aldol condensation and aromatization of the nascent PK chain will be valuable in dissecting biosynthetic mechanisms for other fungal aromatic PKs.

The steps in aflatoxin biosynthesis that occur after norsolorinic acid formation have been reviewed in several outstanding articles.11–17 Below is a brief summary of these multistep enzymatic conversions. Norsolorinic acid is reduced to (1′S)-averantin by a short-chain alcohol dehydrogenase/reductase encoded by aflD. Disruption of aflD causes an accumulation of 6, and a large decrease in aflatoxins. aflG encodes a monooxygenase that is responsible for hydroxylation of averantin in the presence of NADPH, forming (1′S, 5′S)- and (1′S,5′R)-hydroxyaverantin. Another short-chain alcohol dehydrogenase/reductase encoded by aflH oxidizes hydroxyaverantin to 5′-oxoaverantin, which further reacts yielding (1′S,5′S)-averufin via intramolecular acetal formation by the oxoaverantin cyclase. The exact identity of the cyclase involved in this step is not clear, but the enzyme appears to be a homodimer of a 79 kDa protein that does not require a cofactor. Averufin is accumulated when aflI was disrupted; therefore, aflI, a monooxygenase gene, is responsible for hydroxylating averufin to hydroxyversicolorone. Versiconal hemiacetal acetate is formed through a Baeyer–Villiger reaction from hydroxyvericolorone. The gene responsible for this reaction has not been identified. Versiconal hemiacetal acetate is converted to versiconal by more than one esterase; however, aflJ is the primary esterase in the aflatoxin gene cluster. Versiconal then undergoes intramolecular cyclization to give versicolorin B by the versiconal cyclase encoded by aflK, which determines the configuration of aflatoxins and sterigmatocystin.26, 27 At this point, the aflatoxin pathway consists of a metabolic grid that involves three reductive reactions, from hydroxyversicolorone to versicolorone, from versiconal hemiacetal acetate to versiconol acetate, and from versiconal to versiconol.16 The Yabe group recently demonstrated that the same reductase encoded by the vrdA gene of A. parasiticus catalyzes the three reactions.28 Interestingly, the gene is not located in the aflatoxin gene cluster and is not regulated by aflR. It appears that vrdA is not essential for aflatoxin biosynthesis although it participates in the conversion of several aflatoxin intermediates.

Versicolorin B is similar to AFB2 and AFG2 in that it contains a tetrahydrobisfuran ring. It is desaturated by AflL to versicolorin A, which requires NADPH and probably O2 for activity.29 This is the branching point between the tetrahydrobisfuran-containing AFB2 and AFG2 and the dihydrobisfuran-containing AFB1 and AFG1, via the desaturation of versicolorin B to versicolorin A. Versicolorin B leads to the type II aflatoxins, and versicolorin A leads to type I aflatoxins, using the same enzymes to achieve this. The next step is the conversion of versicolorin A to demethylsterigmatocystin and versicolorin B to dihydrodemethylsterigmatocystin. However, the enzyme(s) needed to catalyze this reaction have not yet been fully characterized, and this step is likely catalyzed by more than one enzyme. It is likely that the aflM-encoded dehydrogenase and aflN-encoded monooxygenase are required for this conversion.30, 31 The aflO-encoded O-methyltransferase catalyzes the transformation of demethylsterigmatocystin and dihydrodemethylsterigmatocystin to sterigmatocystin and dihydrosterigmatocystin, respectively.32 This is the last step in Aspergillus nidulans as sterigmatocystin is the final metabolic product in this pathway. aflP is a second O-methyltransferase that transfers another methyl group to form O-methylsterigmatocystin and dihydro-O-methylsterigmatocystin, respectively.33 These represent the last stable intermediates of aflatoxins and are the precursors of AFB1/AFG1 and AFB2/AFG2. Finally, O-methylsterigmatocystin is converted to AFB1, whereas dihydro-O-methylsterigmatocystin is converted to AFB2, by a P450 monooxygenase encoded by by aflQ.34–36 Alternatively, O-methylsterigmatocystin and dihydro-O-methylsterigmatocystin are converted to AFG1 and AFG2, respectively. It appears that both cytosolic and microsomal enzymes are involved in this conversion.37 A recent study reported the potential functions of two previously uncharacterized genes, aflF and aflZ in G-type aflatoxin biosynthesis.38aflF was predicted to encode a NAD+ or NADP+-dependent alcohol dehydrogenase, and aflZ was predicted to encode an OYE-FMN-binding domain reductase. The analyses of the metabolites accumulated in the gene disruption mutants showed that AFG1 biosynthesis involves NadA reduction and NorB oxidation. The involvement of the aflZ gene in formation of G-group aflatoxins in A. parasiticus was also supported by the results from the Yabe group.39

FUMONISINS

Fumonsins are mycotoxins produced by several agriculturally important fungi, including Fusarium verticillioides, which is a common fungal contaminant of corn and maize-derived products worldwide.40–44 Fumonisin contamination in corn has been associated with livestock loss45–48 and human health risks, including esophageal cancer49–53 and neural tube defects.40, 54 To date, at least 28 fumonisins have been isolated from fungi, and they are classified into four groups, fumonisin A, B, C, and P series.55 The A-series is structurally similar to the B-series (FB1, 2 in Figure 1) with the exception of a C2 amino-acylated functionality. The P-series fumonisins have a 3-hydroxypyridinium moiety at the C2 position in contrast to the free amino group found in the B and C-series. The C-series is distinct from the A, B, and P-series in that it has a C19 backbone resulting from the condensation of an acyl chain with glycine as opposed to an alanine residue. B-series fumonisins are the most abundant, and fumonisin B1 (FB1, 2) constitutes ∼70% of the total fumonisin content found in naturally contaminated maize and is the most toxic fumonisin analog.43, 56 Recently, Aspergillus niger was also found to produce fumonisins (i.e., fumonisins B2 and B4),57–60 and a new B-series fumonisin, FB6, was identified from this fungus.61 FB6 is an isomer of FB1, having hydroxyl groups at C3, C4, and C5, instead of at C3, C5, and C10 in FB1.

Proctor and coworkers62–64 identified the fumonisin biosynthetic gene cluster, which is located in a 75-kb region of DNA from F. verticillioides. They first isolated a HR-PKS gene FUM1 from F. verticillioides. FUM1 was predicted to encode a seven-domain PKS, consisting of KS, AT, DH, C-MeT, ER, KR, and ACP (see Figure 2). The FUM1 disruption mutants did not produce any fumonisins, showing that this gene is required for fumonisin biosynthesis. So far, FUM1 has not been biochemically characterized, but available evidence supports that this PKS assembles the 18-carbon chain (C3–20) of fumonisins from one molecule of acetyl-CoA, eight molecules of malonyl-CoA, and two molecules of S-adenosyl methionine (Figure 5A).65–70

Figure 5.

Biosynthetic mechanism for fumonisins. A: The biosynthetic pathway for FB1; (B) the PLP-dependent polyketide chain-releasing mechanism in fumonisin biosynthesis.

The next gene in the cluster is FUM8, encoding a product (Fum8p) homologous to the class II α-aminotransferases, which belong to a group of pyridoxal 5′-phosphate (PLP)-dependent enzymes that catalyze the decarboxylative condensation between amino acids and acyl-CoA thioester substrates. The disruption of FUM8 led to the elimination of fumonisin production, indicating that this gene is required for fumonisin production.63 In fumonsin biosynthesis, Fum1p does not contain a TE/CLC domain that is required for the product release in fungal NR-PKS. Fum8p was proposed to function in the release of the nascent 18-carbon PK chain assembled by Fum1p, bound to the 20 Å phosphopantetheine arm of ACP.68 This has recently been proven by biochemical approaches (Figure 5B).71 In this study, Gerber et al.71 heterologously expressed FUM8 in yeast and prepared Fum8p as a microsomal fraction. The in vitro results showed that Fum8p was able to offload the acyl chain from the acyl-S-ACP of Fum1p and at the same time incorporate two carbons and an amino group from alanine to yield the expected 3-keto fumonisin precursor (Figure 5B). The activity was dependent on PLP, and 18C-S-ACP was the best substrate among the acyl-S-ACP tested. This is an unprecedented PK chain-releasing mechanism. It uses PLP cofactor, rather than a catalytic triad as seen in the common TE/CLC system. The releasing enzyme uses a carbon (α-carbon of alanine), rather than the typical oxygen (hydroxyl group) or nitrogen (amino group), as a nucleophile to attack the terminal carbonyl of acyl-S-PKS to release the PK chain. Finally, the substrate specificity of this PLP-dependent enzyme is important in controlling the length of the PK chain.72

Another study from the Du group was the investigation on the functional complementation of a FUM1 disruption mutant of F. verticillioides using ALT1, the PKS from Alternaria alternata required for the biosynthesis of AAL-toxins, which are mycotoxins structurally similar to fumonisins.69 Both Alt1p and Fum1p belong to the HR-PKS group and have the same domain organization. However, Fum1p synthesizes an 18-carbon backbone, whereas Alt1p assembles a 16-carbon backbone. When ALT1 was introduced into the fumonisin-nonproducing strain, one of the transformants produced metabolites that coincided with fumonisins from HPLC and MS analysis. The result showed that ALT1 can in trans complement fumonisin production in the FUM1 disruption mutant. It also demonstrated that even though the native programming of Alt1p is to produce a 16C-chain in A. alternata, it is able to form an 18C-chain in F. verticillioides transformant. This supports the notion that the fungal HR-PKS alone does not determine the structure of the final products, but rather, the chain-releasing enzyme (Fum8p in this case) is more critical to forming a distinct product. In addition to the chain length, the methylation pattern differs between the Fum1p- and Alt1p-catalyzed reactions. In fumonisins, the methylation occurs during the second and fourth cycle of PK chain elongation. In AAL toxins, the methylation occurs during the first and third cycle of PK chain elongation. A closer look into the methylation pattern of the metabolites obtained from the ALT1 transformant by NMR showed that these compounds had a fumonisin-type methylation pattern (C12/C16 methylation) as opposed to an AAL-toxin methylation pattern (C14/C18).73 This suggests that the timing and regioselectivity of the C-MeT domain of the PKS could be reprogrammed.

The remaining FUM genes take part in decorating the 20-carbon backbone. FUM13 is predicted to encode a short-chain dehydrogenase/reductase. FUM13 disruption mutants yielded the three-keto form of FB3 and FB4.74 Yi et al.75 expressed the gene in E. coli and used the purified Fum13p to show the three-keto reduction activity of the enzyme. The results showed that Fum13p is an NADPH-dependent ketoreductase. The vicinal diol at C14 and C15 was initially proposed to be generated by hydrolysis of an epoxide.76 Current evidence suggests that it is more likely that the hydroxyls are formed via the activity of one or a combination of P450 monooxygenases, such as the enzyme encoded by FUM6.63FUM6 disruption mutants did not accumulate any detectable fumonisins. However, co-cultures of FUM6 mutant with FUM1 mutant or FUM6 mutant with FUM8 mutant restored fumonisin production.66 Co-culture of FUM1 mutant with FUM8 mutant did not produce any fumonisin. The results suggest that the FUM6 mutant produced the intermediate compound(s) that was converted into the final products by FUM1 mutant or FUM8 mutant (Figure 5A).

Fumonisins contain two tricarballylic esters on C14 and C15, which are rare structural features in natural products. Four genes, FUM7, FUM10, FUM11, and FUM14, are involved in the formation of the tricarballylic esters (Figure 5A).77FUM11 is predicted to encode a tricarboxylate transporter, which could be involved in the in vivo substrate transportation across cellular compartments. The other three genes may encode a nonribosomal peptide synthetase (NRPS)-like complex, where Fum10p is analogous to the adenylation (A) domain, Fum14p contains the peptidyl carrier protein and condensation (C) domain, and Fum7p may be regarded as a reductase domain. Fum10p-14p-7p could represent an unusual “stand-alone” NRPS-like complex. So far, the function of FUM7 and FUM10 has not been biochemically characterized, but the reaction catalyzed by Fum14p has been demonstrated in vitro.78 Both the intact Fum14p and the C domain have been expressed in E. coli, and the purified proteins were used to test their activity. Fum14p, as well as the C domain alone, is able to convert HFB1 (containing the two hydroxyl groups at C14 and C15) to FB1 when the N-acetylcysteamine monothioester of tricarballylic acid was used as the acyl group donor. This shows that Fum14p catalyzes the esterification of fumonisins. Interestingly, Fum14p represents the first example of an NRPS-like enzyme catalyzing a C[BOND]O bond formation instead of the typical C[BOND]N bond.78

The esterification at C14 and C15 produces FB4, which is further oxidized to FB2, FB3, and FB1 (Figure 5A).70 The steps are catalyzed by two different types of oxygenase. FUM2 encodes a P450 monooxygenase. Mutants with a deleted FUM12 produced only FB2 and FB4, suggesting that this gene is required for C10 hydroxylation.79 The mutants with deleted FUM3 produced FB4 and FB3, but not FB2 and FB1, suggesting that this gene is required for C5 hydroxylation.80 The function of FUM3 was demonstrated in vitro by expressing the gene in yeast and using the purified enzyme to convert FB3 to FB1.81 The C5 hydroxylase activity of Fum3p requires the presence of 2-ketoglutarate, as well as iron(II), ascorbic acid and catalase, which are characteristics of 2-ketoglutarate-dependent dioxygenase.

The gene cluster also contains a pathway specific regulator, FUM21, which was predicted to encode a Zn(II)-2Cys6 DNA-binding transcription factor.82FUM21 deletion mutants do not produce fumonisins. Real-time polymerized chain reaction results showed that FUM21 positively regulates the expression of other FUM genes. Previously, a number of genes, outside the FUM cluster, had been shown to regulate fumonisin production.83–86 Recently, a new gene, FvVE1, was also shown to be involved in the production of fumonisins (as well as fusarins).87FvVE1 deletion completely suppressed fumonisin production. More importantly, FvVE1 was shown to be necessary for the expression of the pathway-specific regulatory gene FUM21 and biosynthetic enzyme-encoding genes. Together, these results suggest that the regulation of fumonisin biosynthesis is controlled by multiple regulators at multiple levels.

Within the FUM cluster, there are a number of genes whose functions remain unclear. These genes are FUM15, FUM16, FUM17, FUM18, FUM19, and the newly identified FUM20.88 Because these genes are clustered with other fumonisin biosynthetic genes and their expression is also coregulated, it is probable that they play a role in fumonisin production, such as self-protection.70

OCHRATOXINS

Ochratoxins (OT) are potential nephrotoxic and carcinogenic mycotoxins produced by a number of Aspergillus and Penicillium species.89–91 They are found in food products such as cereals, coffee, wine, beer, and spices. At least 20 different OT analogs have been detected, where ochratoxin A (OTA, 3) is produced at the highest levels and is the most toxic derivative. The structure of OTA consists of a pentaketide dihydroisocoumarin linked by an amide bond to a molecule of the amino acid phenylalanine. The carbonyl group of the amide is derived by oxidation of the methyl side chain of the pentaketide (see Figure 6). A chlorine atom on one of the dihydroisocoumarin rings contributes to its toxicity, such that dechlorination of OTA to ochratoxin B (OTB) results in a 10-fold decrease in toxicity. Several major ochratoxin analogs exist, including hydroxylated OTA, OTα, OTβ, and ochratoxin C (OTC). OTα and OTβ are formed by the cleavage of the amide bond of OTA or OTB. OTC is the ethyl ester form of OTA. OTA and OTB can also exist as the open lactone form (Figure 6C).

Figure 6.

Chemical structures of ochratoxin analogs. A: General structures; (B) the polyketide originated portion; (C) the open lactone form.

Although several mycotoxin biosynthetic pathways have been elucidated, compared to aflatoxin and fumonisin biosynthesis, relatively little is known about the molecular mechanisms of OT biosynthesis. Early feeding experiments and chemical degradation showed that phenylalanine was responsible for the phenylalanine moiety of OTA, five acetate units were incorporated into the dihydroisocoumarin moiety of OTA, C7 originates from methionine of ochratoxin α that is oxidized to a carboxyl group and becomes the link between the phenylalanine moiety and the dihydroisocoumarin moiety.92 Mellein is a metabolite also produced by ochratoxigenic species. Its structure is the same as the dihydroisocoumarin portion of ochratoxins, but without the C7 carboxyl group. In spite of the similarity, experiments with 14C-labeled precursors did not support an intermediary role for mellein. Instead, the pentaketide OTβ is the most likely intermediate as it was biotransformed very efficiently into both OTA and OTB. The chlorination is probably the penultimate step in the biosynthesis of OTA, because the chlorinated OTα was only biotransformed significantly into OTA. This is supported by the very low conversion of radiolabeled OTB into OTA, which implies that OTB may arise by dechlorination of OTA.

A proposed biosynthetic pathway for OTA is illustrated in Figure 7. It includes a PKS for the synthesis of the PK dihydroisocoumarin, a methyltransferase (or a C-MeT domain within the PKS), and a P450 type oxidation enzyme for formation of the carboxyl group at C7, a NRPS to catalyze the ligation of phenylalanine with the PK, and a halogenase (chloroperoxidase) to incorporate the chlorine atom. In spite of the many years of study, none of the biosynthetic steps have been genetically and biochemically established. So far, only the PKS gene has been partially characterized, including PKS from Aspergillus ochraceus,93Aspergillus westerdijkiae,94Aspergillus carbonarius,95 and Penicillium nordicum.96 One fragment of a PKS gene was initially cloned from A. ochraceus, and the disruption of this gene resulted in mutants that were unable to produce OTA.93 This result confirms the PK origin of OTA biosynthesis. In another producer, P. nordicum, several DNA fragments obtained by differential display RT-PCR showed homology to genes encoding enzymes that are potentially involved in OTA biosynthesis, including PKS, NRPS, halogenase, phenylalanine-tRNA synthetase, methylase, and ABC transporter.97 Later, the Geisen group sequenced a 10-kb fragment from P. nordicum. It contained three open reading frames, a partial sequence of a PKS homolog otapksPN, a complete sequence of a NRPS homolog npsPN, and a complete sequence of an alkaline serine protease homolog aspPN.96 The otapksPN disruption mutants lost the capacity to produce OTA, and the expression of otapksPN correlated with OTA biosynthesis. The data showed the relevance of this PKS gene in OTA biosynthesis. It appeared that otapksPN is present only in P. nordicum, but not in the related Penicillium species or ochratoxigenic Aspergillus species, whereas npsPN was detected in another Penicillium species but not in other species. This suggests the existence of a convergent pathway for OT biosynthesis in different fungal species.

Figure 7.

A proposed biosynthetic mechanism for ochratoxin A.

Recently, another PKS gene, aoks1, from A. westerdijkiae was investigated for its role in OTA biosynthesis.94 Ochratoxin production in A. westerdijkiae over an 18-day period was studied and compared to expression levels of aoks1 in RT-PCR studies. The results show that OTA production was at its highest at day twelve, and this closely correlated with aoks1 transcript levels that were at their highest between days 5 and 10. The aoks1 disruption mutants (aoΔks1) lost the ability to produce OTA. These results confirm that aoks1 is required for OTA biosynthesis. Similar experiments were also carried out in another ochratoxin producer, A. carbonarius.95 In summary, the OTA biosynthetic genes remain to be characterized, and the proposed biosynthetic steps in Figure 7 need to be established by detailed genetic and biochemical studies.

ZEARALENONE

Zearalenones are a group of mycotoxins produced by several species of Fusarium, such as Fusarium graminearum and Fusarium culmorum.98, 99 They cause estrogenic disorders in domestic animals feeding on contaminated feeds.100 Zearalenones have a resorcinol (m-benzenediol) moiety fused to a 14-member macrocyclic lactone. There are five main metabolites, α-zearalenol, β-zearalenol, α-zearalanol, β-zearalanol, and zearalenone (see Figure 8). α- and β-Zearalenol as well as α- and β-zearalanol contain a C6′ hydroxyl group compared to the keto group of zearalenone and zearalanone. In addition, zearalanone and α- and β-zearalanol do not contain the C1′–C2′ double bond that is present in zearalenone.

Figure 8.

Biosynthetic mechanism for zearalenone. A: The structure of zearalenone and its five main analogs; (B) a proposed biosynthetic pathway for zearalenone.

The PK origin of the zearalenone backbone was established by isotope-feeding experiments.101 The experiments also showed that α- and β-zearalenol are precursors of zearalenone.102 Although the PK biosynthesis had been proposed earlier on, the zearalenone PKS genes were identified only a few years ago. Three laboratories independently reported the genes required for zearalenone biosynthesis.103–105 Kim et al.103 first reported two different PKSs (ZEA1, i.e., PKS13 and ZEA2, i.e., PKS4) that are required for zearalenone biosynthesis in G. zeae (anamorph F. graminearum). They reasoned that a nonreducing PKS could be involved in the biosynthesis because of the presence of ketone functional groups (as enol in resorcinol ring) in zearalenone. Taking advantage of the annotated PKS genes in Gibberella zeae genome, they deleted each of the NR-PKS genes and found that ZEA1 is required for zearalenone production. Subsequence analyses of the flanking regions of ZEA1 revealed a gene cluster in a 50-kb segment of DNA. Further targeted gene deletions showed that four genes of the cluster are essential for zearalenone production, including ZEA2, ZEB1, ZEB2, in addition to ZEA1. They showed that these genes were coexpressed in the fungus. ZEA2 was predicted to encode a HR-PKS, ZEB1 a putative isoamyl alcohol oxidase, and ZEB2 a leucine zipper domain-containing regulator. Their results also suggested that ZEB1 is responsible for the oxidation of zearalenol to zearalenone, and ZEB2 controls the transcription of other genes in the zearalenone pathway. Lysoe et al.106 disrupted ZEA2 of F. graminearum and found that this gene is essential for zearalenone production, which is consistent with the results obtained by Kim et al.103

In the meantime, the Trail group deleted each of the 15 PKS genes in the genome of G. zeae and characterized the mutant strains.104, 105 The mutants of two of the genes, ZEA1 and ZEA2, lost the ability to produce zearalenone. ZEA1 and ZEA2 share the common promoter and are transcribed divergently. Based on the gene function and expression studies, they proposed a biosynthetic pathway, in which two PKSs, one HR-PKS and one NR-PKS, work collaboratively to assemble the backbone of zearalenone. Two PKSs (LovB and LovF) are also required for the biosynthesis of another fungal PK, lovastatin.107 However, in lovastatin, each of the PKSs synthesizes an independent PK chain, and the two chains are linked via an ester bond, which is catalyzed by a dedicated AT (LovD).108 In the proposed zearalenone pathway (Figure 8B), the HR-PKS (ZEA2p, i.e., PKS4) catalyzes the formation of the largely reduced hexaketide chain, which is transacylated to the NR-PKS (ZEA1p, i.e., PKS13) and further elongated for three more cycles to form the nonaketide chain of zearalenone. In the latter three-cycle elongations, the β-keto group after each condensation reaction remains unreduced, and seven carbons of the nonaketide form the resorcylate unit via an intramolecular C2–C7 aldol condensation (Figure 8B). In essence, this mechanism is similar to the FAS-PksA complex (NorS) in aflatoxin biosynthesis, where FAS synthesizes the highly reduced hexanoyl starter that is further elongated by PksA to form the 20-carbon PK chain that is cyclized and oxidized to norsolorinic acid.20 Another analogous system is the T-toxins produced by Cochliobolus heterostrophus.109, 110 Two PKSs, PKS1/PKS2, are required to synthesize the linear carbon chain, although the mechanistic details of this system have not been demonstrated.

ZEA2p is a typical HR-PKS containing six domains, KS-AT-DH-ER-KR-ACP, whereas ZEA1p is a typical NR-PKS also containing six domains, SAT-KS-AT-PT-ACP-TE. The Tang group expressed ZEA1 in E. coli and successfully purified the 222-kDa enzyme for in vitro demonstration of the regiospecific cyclization to form the resorcylate core.111 A chemically synthesized substrate 11-hydroxyundecanyl-S-acetylcysteamine was used as the mimic of the starter hexaketide-S-PKS4 in the ZEA1p reactions. The results showed that ZEA1p is a resorcylic acid macrolactone megasynthase that catalyzes the iterative chain elongation, aromatic cyclization, and macrolactone esterification. ZEA1p exibited a relatively relaxed starter-unit specificity toward C6–C16 acyl-CoAs, with C10-CoA being the best substrate. The results showed that ZEA1p possesses a significantly broader starter-unit specificity than other fungal PKS, such as aflatoxin PksA.21 This broad specificity was further exploited in E. coli/ZEA1 for precursor-directed biosynthesis of PKs.

In a study by the Boddy group, the thioesterase domain of ZEA1p was further investigated for its functional role in macrocyclization.112 The TE domain of ZEA1p was expressed, purified, and tested in vitro. It was shown that the TE catalyzed the macrocyclization of a mimic of the thioester–precursor of zearalenone to produce the corresponding macrolactone. The TE domain was also tested for cross-coupling activity by incubating the domain with benzoyl-SNAC. As a result, butyl benzoate was formed. This result supported the acyl-enzyme intermediate as a functional step in the reaction mechanism. Kinetic data supported that the substrates of choice were aromatic thioesters, which correspond to the resorcylate intermediate. This data suggest that the TE domain, because of its specificity for the resorcylate intermediate, quickly offloads only when this correct intermediate is formed. Thus, the TE domain may play a role in controlling the number of the iterative cycles of ZEA1p.

The PK programming rules exhibited by zearalenone ZEA1p/ZEA2p are very intriguing and different from many other fungal systems. The two enzymes must be able to interact with each other to facilitate the delivery of a properly sized starter unit from ZEA2p to ZEA1p. The role of the SAT domain of ZEA1p in selecting the starter unit that is transferred from ZEA2p remains unclear, as a S121A mutation at the putative active site of the SAT domain did not have a detectable effect on activity.111 Furthermore, in spite of the relaxed specificity observed in vitro, ZEA1p accepts only a hexaketide starter and controls the iterative cycles to three rounds in vivo. The recent work of the Boddy group suggests that the TE domain may play an important role in the timing of macrocyclization and in the choice of the substrate for cyclization.112

FINAL REMARKS

Fungi produce numerous PK metabolites. This review focuses on the four major groups of mycotoxins that are derived from PKs. Among these mycotoxins, aflatoxins and zearalenone are the best studied in terms of the biochemical characterizations of their PKSs. Fumonisin PKS have been genetically studied, but the enzyme has not been produced for in vitro studies, whereas ochratoxin PKS remains to be fully cloned and characterized. A great deal of knowledge has been obtained from the post-PKS modifications of aflatoxins and fumonisins, while not much is known about the post-PKS modifications of ochratoxins and zearalenone. In addition to these “traditional mycotoxins” (including trichothecenes, 5), more mycotoxins, both PK-derived and nonpolyketide-originated, are emerging as important food and feed contaminants, such as fusaproliferin, beauvericin, enniatins, and moniliformin.113 Many studies have been carried out for their occurrence, toxicity, and methods of detoxification. More efforts are needed to fully understand the molecular mechanisms for the biosynthesis of both “traditional mycotoxins” and “emerging mycotoxins,” so that rational approaches may be developed for mycotoxin reduction/elimination in agriculture and food industry.

Acknowledgements

We thank the authors whose work is cited in the individual references of this review. We thank Lili Lou for critical reading of this manuscript.