Iterative type I polyketide synthases (PKSs) are megasynthases found in fungi that produce a wide variety of compounds with a diverse range of biological activities1–9 (see Figure 1). These enzymes repetitively use a single set of active domains to assemble complex metabolites by a pathway that is similar to that of fatty acid synthases (FAS).1–5, 10–12 In both polyketide and fatty acid biosynthesis, acetate units undergo repetitive Claisen condensations (see Figure 2). In fatty acid synthesis, the β-keto group undergoes reductive processing after each condensation, resulting in a fully saturated product. In polyketide synthesis, reductive steps may be omitted during a cycle, leading to various levels of oxidation throughout the polyketide. To initiate polyketide synthesis, the β-ketosynthase (KS) domain is covalently attached to an acetyl starter unit by a thioester bond. This starter unit commonly originates from an acetyl coenzyme A (CoA) or malonyl CoA, but other molecules may also be used.1–4, 6, 7, 13 The next acyl unit, usually derived from malonyl CoA,1–4, 6–8, 13 is then attached to the acyl carrier protein (ACP) by the acyl transferase (AT) domain. The KS domain then catalyzes a decarboxylative Claisen condensation, elongating the acyl chain on the ACP. At this stage, processing may occur by the other domains; a ketoreductase (KR) can reduce the β-keto group to a hydroxyl, a dehydratase (DH) can catalyze the dehydration of the hydroxyl group to an alkene, which may be reduced by an enoyl reductase (ER). Other processing domains, such as a methyl transferase (MT) may also be present that can further functionalize the polyketide.1, 3, 7 A critical unanswered question is why the processing domains are active at certain polyketide lengths and not others. This allows an enzyme to produce a particular functionalized structure with great fidelity, but affords access to a great diversity of structures by a family of closely related proteins.1, 4, 5, 14 Once processed, the acyl chain is then transferred to the original KS domain, and a new acyl unit can be attached to the ACP to continue the cycle.1, 3, 5, 7, 8, 11 Once the polyketide is formed, it can be released from the PKS by a thioesterase (TE) domain.
Hybrid PKS nonribosomal peptide synthases (PKS-NRPS) lead to a larger diversity of structures.2, 3, 7, 13, 15 In addition to the KS, MT, and processing domains, a PKS-NRPS hybrid usually contains condensation (C), adenylation (A), and thiolation (T) domains, as well as an off-loading domain such as a TE or reductive (R) domain.2, 6, 7, 13, 15, 16 An amino acid is selected and adenylated by the A domain, and is attached to the T domain. The C domain forms an amide bond between the amino acid and the polyketide. The product is typically released by a TE domain, or by an R domain that reduces the thioester bond to release the product as an aldehyde, which is often processed by downstream enzymes.2, 9, 13
Lovastatin, also known as mevinolin, monacolin K, and mevastatin, was isolated from Aspergillus terreus in 197817, 18 and Monascus ruber in 1979.19, 20 It is used to treat hypercholesterolemia as it inhibits (3S)-hydroxy-3-methylglutaryl-coenzyme A (HMG-CoA) reductase, which catalyzes the rate-limiting step in cholesterol biosynthesis.20 Lovastatin is also used as a precursor for the semi-synthetic drug, simvastatin (Zocor).
The lovastatin gene cluster contains 18 genes potentially involved in the biosynthesis of this metabolite, but five of these have been identified to encode enzymes essential for the formation of lovastatin, namely lovA, lovB, lovC, lovD, and lovF.21, 22 Two of these, lovB and lovF encode PKSs, whereas lovC, lovD, and lovA encode an enoyl reductase, an esterase, and a cytochrome P450 oxygenase, respectively. Two genes, lovE and lovH, have homology with transcription factors. When extra copies of lovE are added, lovastatin production increases, and when either of these genes are rendered inactive via mutations, lovastatin and its intermediates are not detected.21 The gene lvrA was shown to provide lovastatin resistance and is similar to HMG-CoA reductases.21
The 335 kDa enzyme LovB (lovastatin nonaketide synthase) is a highly reducing iterative type I PKS with KS, AT, DH, MT, KR, and ACP domains (see Figure 3).21–23 It also contains an inactive ER domain, so a separate ER enzyme, LovC, is required.22 Together, LovB and LovC catalyze about 35 reactions to form dihydromonacolin L (see Figure 4). The production of dihydromonacolin L requires nine malonyl CoA molecules, NADPH and S-adenosyl-L-methionine (SAM). Once the starter acetyl unit derived from malonyl CoA24 is attached to LovB, each Claisen condensation is catalyzed by the KS domain, followed by a ketoreduction until the nonaketide is formed. The DH domain is also active for the first six elongations, until the heptaketide is formed. The MT domain transfers a methyl group from SAM to the growing polyketide chain during the formation of the tetraketide. Enoyl reduction occurs by LovC at the tetraketide, pentaketide, and heptaketide stages, resulting in fully reduced sections.
In addition to the reductive processing by the various domains after each condensation, it is proposed that an enzymatically catalyzed Diels-Alder cycloaddition occurs at the hexaketide stage, to form the fused rings of the decalin system of dihydromonacolin L.14, 25, 26 To test LovB for Diels-Alderase activity, a triene N-acetylcysteamine (NAC) thioester, 1, was synthesized (see Figure 5).25 In the absence of LovB, the thioester produces the exo and endo adducts 2 and 3 in equal amounts, which arise when the chair-like transition state has the C-6 methyl group in the pseudoequatorial position.26 When the thioester was added to LovB in vitro, a new endo adduct, 4, was detected along with the nonenzymatic products, 2 and 3 in a ratio of 1:15:15, respectively. This new endo product results from the chair-like transition state where the C-6 methyl group is in the pseudoaxial position. The product has the same stereochemistry as the natural polyketide, dihydromonacolin L. To test for nonspecific interactions, the thioester was also added to LovB that had been denatured by heat treatment, and once again only the products 2 and 3 were detected. This suggests that LovB does have Diels-Alderase activity, and that it is needed to produce the proper stereochemistry for dihydromonacolin L.
The 277 kDa enzyme, LovF (lovastatin diketide synthase) is similar to LovB, but its ER domain is active (see Figure 3).22, 23 It condenses two acetyl units, fully reduces them and transfers a methyl group from SAM to form the 2-methylbutyrate side chain of lovastatin. To connect the two polyketides, dihydromonacolin L is oxidized to monacolin J by the putative cytochrome P450 oxygenase, LovA27 and the diketide is covalently attached to monacolin J by LovD, a transesterase, to form lovastatin (see Figure 4).28
The C-termini of LovB and LovF neither contain TE domains nor any other common off-loading domain.22, 23 Homology studies suggest that the C-terminus of LovB contains elements of a NRPS. This section contains a full C domain22, 23 and it may also contain a truncated A domain,29 in which the catalytic region has been deleted. It is not fully understood what causes dihydromonacolin L to detach from the enzyme; when LovB and LovC were utilized to produce this polyketide in vitro, it remained attached to LovB.24
Upon addition of the PKS13 TE domain from Gibberella zeae, dihydromonacolin L was released. It has been suggested that the C domain may catalyze the release of dihydromonacolin L.1 When the C domain is deleted from LovB, no dihydromonacolin L can be detected, even upon addition of a TE domain.24 The C domain appears to be able to function intrans with LovB; when the purified C monodomain was added to the truncated LovB and TE domain, dihydromonacolin L was once again formed.
Similar to dihydromonacolin L and LovB, it has been shown that the 2-methylbutyrate chain remains attached to the ACP of LovF.28, 30 The transesterase, LovD, removes the diketide from LovF and transfers it to monacolin J. It has been shown that LovD is capable of transferring other thioesters, including NAC, ACP, and CoA thioesters, to various decalins that are structurally similar to monacolin J. The activity of LovD is altered by interactions with LovF though; LovD's enzymatic activity is higher for thioesters attached to LovF compared with other thioester substrates, including the LovF ACP monodomain.28 The ability for LovD to utilize various substrates may be useful for the production of lovastatin analogs, as this allows the side chain attached to monacolin J to be varied. For example, simvastatin can be formed utilizing a dimethylbutyryl NAC thioester, LovD, and monacolin J.28, 30
Compactin, also known as mevastatin and ML-236B,31 is very similar to lovastatin both in its structure and how it is biosynthesized. The only structural difference is that compactin lacks the C-6 methyl group. Compactin is also an inhibitor of HMG-CoA and is a precursor for the drug pravastatin.32, 33 The compactin gene cluster contains nine genes, mlcA–mlcH and mlcR.34, 35 The genes mlcA and mlcB encode the PKSs MlcA and MlcB which correspond to LovB and LovF, respectively (see Figure 3).35 There is no ER domain in MlcA and like LovB, it requires MlcG, an ER protein similar to LovC, for the production of the nonaketide. The MT domain is present in MlcA, but it is thought to be inactive.35 The C-terminus of MlcA also has homology with NRPS domains, which supports the hypothesis that this section cleaves the nonaketide to release it from the PKS.1
The ability for the ER proteins MlcG and LovC to be interchanged was examined.24 The activity of LovC requires the presence of the methyl group which is presumed to be added at the tetraketide stage; when SAM is absent in a system normally capable of producing dihydromonacolin L, the desmethyl-dihydromonacolin L intermediate cannot be detected, instead pyrones are produced. In contrast to LovC, the ER protein MlcG is capable of transforming its substrates regardless of the absence or presence of the methyl group. When MlcG is substituted for LovC in a LovB/LovC system and SAM is absent, desmethyl-dihydromonacolin L is produced; when SAM is added, dihydromonacolin L is detected.
The compactin gene cluster also encodes MlcC and MlcH, which are analogous to LovA and LovD, the P450 and transesterase enzymes.35 The gene mlcR, which is homologous to lovE, encodes a transcription factor that regulates the mlcB, mlcF, mlcG, and mlcH genes.36 The compactin genes, mlcC and mlcD, are similar to lovA and lvrA, in the lovastatin gene cluster.35
EQUISETIN AND TRICHOSETIN
Equisetin is an HIV-I integrase inhibitor37 that was first isolated in 1974 from the fungus Fusarium equiseti.38 Equisetin is synthesized by an iterative type I PKS-NRPS hybrid, EqiS.29, 39 The polyketide portion consists of an octaketide with a trans decalin system and the tetramic acid, 2,4-pyrrolidinedione comprises the peptide section. The PKS section of EqiS contains KS, AT, DH, MT, KR, and ACP domains,29 whereas the NRPS consists of C, A, and T domains, as well as a Dieckmann cyclase (DKC) domain that was initially identified as an R domain. EqiS and LovB are very similar and their domains are arranged in the same manner (see Figure 3). It has been suggested that LovB may have originated from a tetramic acid synthase that lost its C-terminus,29 including the T, R/DKC, and part of the A domains. The polyketide portion of equisetin is proposed to be biosynthesized in a similar manner as lovastatin, and a Diels-Alder reaction is believed to occur at the heptaketide stage14 (see Figure 6).
An N-desmethyl homolog of equisetin, trichosetin, has also been isolated.40 The gene cluster for trichosetin has not been examined, but it is likely equisetin and trichosetin utilize analogous biosynthetic pathways. Trichosetin has been shown to incorporate serine into its structure,41 which is covalently attached to the decalin polyketide by the C domain via an amide bond.29, 39 The R domain was originally proposed to catalyze a reduction,29 leading to the release of the aldehyde intermediate, which would be followed by a cyclization. However, when NAC and CoA thioesters of acetylalanine and acetoacetylalanine were added to the R monodomain, no reduced products were detected, rather tetramic acid was formed from the acetoacetylalanine thioesters.39 It was also determined that neither NADH nor NADPH were consumed for this transformation, which would be required for a reduction. The hypothesized reduction mechanism was thus replaced by the Dieckmann condensation, which forms the tetramic acid and releases equisetin or trichosetin from EqiS. The R domain was thus relabeled as a DKC domain. Similar DKC activites have been observed for the biosyntheses of other tetramic acids, including cyclopiazonic acid42 and tenellin.43 Methylation is also required to form equisetin, but it is unknown if this occurs before or after the Dieckmann condensation.29, 39
Cytochalasins are a group of compounds that were named for their ability to inhibit movement and cleavage of cells44 due to their ability to bind to actin filaments.45 Since cytochalasins A and B were first isolated in 1966 from Phoma strain S 298,46 over 80 cytochalasins have been identified.47 The cytochalasins are biosynthesized by iterative type I PKS-NRPSs and consist of a polyketide macrocyclic ring that contains a lactone, carbonate, or carbocycle, that is fused to an isoindolone ring.48 The attached amino acid can vary and examples include phenylalanine in zygosporins, tryptophan in chaetoglobosins, and leucine in aspochalasins.44 Of the cytochalasins, the biosynthetic pathway that forms both chaetoglobosin A and C is best understood, as the gene cluster has been identified and sequenced.44 The gene cluster contains seven genes, cheA to cheG. CheA comprises the hybrid PKS-NRPS enzyme and its domains are arranged in the same manner as LovB and EqiS; the PKS section contains KS, AT, DH, MT, KR, and ACP domains and the NRPS section contains C, A, T, and R domains (see Figure 3). The PKS section of CheA does not contain a functional ER domain and similar to the biosynthetic machinery for lovastatin and equisetin, there is an ER enzyme, CheB, that acts intrans with CheA. It has been shown that the polyketide portion of chaetoglobosin A is a nonaketide that has three methyl groups which originate from SAM (see Figure 7).49 Once the nonaketide section has been assembled, it is thought that the NRPS attaches an activated tryptophan via a condensation reaction. This intermediate likely detaches during reduction by the reductive domain to form an aminoaldehyde. It is proposed that the aminoaldehyde undergoes a Knoevenagel condensation, followed by a Diels-Alder cycloaddition to form the isoindolone fused macrocycle.44 Synthetic studies have been done on other cytochalasins that support the role of a Diels-Alderase,50, 51 although the identity and mechanism of this enzyme is unknown. Once the cyclized intermediate is formed, it is thought that the putative oxygenases, CheD, CheE, and CheG oxidize segments to form the final products, chaetoglobosins A and C.44 Based on inhibition studies, these oxygenases appear to functionalize the chaetoglobosins at positions C-6, C-7, C-19, and C-20.52, 53 When P450 inhibitors were added to chaetoglobosin producing cultures, various prochaetoglobosins45 were produced, which have different functional groups at these positions.
The solanapyrones are phytotoxins produced by fungi associated with chickpea blight.54 The first solanapyrones (A, B, and C) that were discovered were isolated from Alterna solani55 and another ten have been isolated.56–59 This class of compounds was the first shown to utilize a Diels-Alderase in their biosynthesis,60 and to incorporate diene and dienophile precursors into natural products.61 Feeding experiments with labeled sodium acetate and methionine showed that the backbone of solanapyrones are biosynthesized by a PKS, which condenses eight acetate units in a head to tail fashion and transfers two methyl groups,62 to produce the octaketide, prosolanapyrone II (see Figure 8). Prosolanapyrone II must undergo an oxidation to form prosolanapyrone III.63, 64 It is thought that the same enzyme that oxidizes prosolanapyrone II also catalyzes a subsequent Diels-Alder cyclization, to form solanapyrones A and D. The sequence of these reactions was confirmed when prosolanapyrone II was added to the enzyme extract in an argon atmosphere.64 No transformation occurred, but upon exposure to oxygen, activity was recovered. To examine whether the Diels-Alder reaction could occur spontaneously, prosolanapyrones II and III were synthesized and transformations were studied both with and without enzymatic extracts.60, 64 In aqueous solution, prosolanapyrone II was not observed to undergo a cycloaddition; however, when it was added to the enzyme extract, 25% of it reacted to form 6% prosolanapyrone III and 15% of the exo and endo cycloaddition solanapyrone adducts, A and D, in a ratio of 85 to 15, respectively. Prosolanapyrone III does undergo a Diels-Alder cyclization in aqueous solution; 15% of it was transformed into the solanapyrones A and D, in a ratio of 3 to 97, respectively. When prosolanapyrone III was added to the enzyme extract, it reacted 4.1 times faster than in the aqueous solution and it produced the solanapyrones A and D in a ratio of 53 to 47. When this ratio was corrected for the amount of cyclization that occurs spontaneously in aqueous solution, it was found that the enzyme produces the exo adduct (solanapyrone A) and the endo adduct (solanapyrone D) in a ratio of 87 to 13, which is the opposite selectivity of the noncatalyzed reaction. The reduced solanapyrones B and E were not detected in any of these experiments and it is thought that these are formed downstream of solanapyrones A and D.14, 60
The polyketides lovastatin, compactin, equisetin, cytochalasins, and solanapyrones share similarities in their biosyntheses, and information that has been obtained regarding the construction of one of these compounds is often applicable to the others. All of these compounds appear to utilize a Diels-Alderase to cyclize their backbone. In the case of solanapyrones and cytochalasins, this occurs after construction of the complete chain by the PKS. Interestingly, the PKS-bound hexaketide chain leading to the solanapyrones is likely identical to that enroute to compactin, but the former is presumably held by its protein in such a manner that Diels-Alder cyclization is not possible. Equisetin and the polyketide intermediates of lovastatin and compactin contain trans decalins, while the solanapyrones consist of both cis and trans decalin systems. Lovastatin, compactin, and chaetoglobosins A and C utilize an ER enzyme in trans to their PKS megasynthase. Although the compactin ER can substitute for LovC in lovastatin biosynthesis, the presence of such a partner protein is critical for correct completion of decalin biosynthesis. In a number of experiments, absence of a required partner protein or essential cofactors (e.g. NADPH, SAM) leads to chain extension and off-loading of a resulting pyrone.24 Equisetin and the chaetoglobosins A and C contain a full NRPS section connected to their PKS megasynthases; the similarities of the enzymes LovB and MlcA suggest that these two PKSs may have been tetramic acid synthases that lost sections of their NRPSs. The biosyntheses of these compounds are not yet completely understood, but structural information about the PKS proteins will provide fascinating insight into how these complex structures are formed and the control mechanisms that determine their functionality.