Current State-of-the-Art Toward Chemoenzymatic Synthesis of Polyketide Natural Products

Abstract

Polyketide natural products have significant promise as pharmaceutical targets for human health and as molecular tools to probe disease and complex biological systems. While the biosynthetic logic of polyketide synthases (PKS) is well-understood, biosynthesis of designer polyketides remains challenging due to several bottlenecks, including substrate specificity constraints, disrupted protein-protein interactions, and protein solubility and folding issues. Focusing on substrate specificity, PKSs are typically interrogated using synthetic thioesters. PKS assembly lines and their products offer a wealth of information when studied in a chemoenzymatic fashion. This review provides an overview of the past two decades of polyketide chemoenzymatic synthesis and their contributions to the field of chemical biology. These synthetic strategies have successfully yielded natural product derivatives while providing critical insights into enzymatic promiscuity and mechanistic activity.

Graphical Abstract

Graphical Abstract

Polyketides represent an area of significant interest for drug discovery efforts. Leveraging a combination of chemical synthesis and enzymology, researchers have routinely demonstrated that the best of both strategies are conferred and result in improved outcomes. This review highlights the success of such an approach and provides an outlook for polyketide synthetic biology.

1. Introduction

Polyketides are often used as therapeutic agents for various diseases and conditions and include compounds such as erythromycin A (antibiotic),^[1] rapamycin (immunosuppressant),^[2] and lovastatin (anti-cholesterol)^[3]. Natural products and their second-generation analogs comprised 24% of the approved small molecule therapeutics between 1981 and 2006, with polyketides contributing more than 50 approved drugs from various scaffolds and sales of blockbuster drugs totaling more than $15 billion.^[4]

However, natural products often require additional derivatization to enhance their biological activities, minimize off-target activities, or combat the growing emergence of antibiotic resistance.^{[5, 6]} While some total syntheses of polyketides have been reported,^[7–9] first made possible by the synthesis of erythromycin A by Woodward in 1981,^[10–12] they carry formidable challenges to high-volume production due to the high step count and low yields of these stereoselective reactions. Nonetheless, numerous research groups are actively searching for new ways to derivatize existing polyketide scaffolds for novel antibiotic discovery. The design and synthesis of natural product analogs allows for the incorporation of unnatural functional groups that enhance native bioactivity or produce new activity altogether.^[13] Notably, the development of a combinatorial synthetic strategy has led to the gram-scale construction of more than 300 macrolides with unnatural synthetic moieties, many of which display improved antibiotic activity compared to common blockbuster drugs.^[14] Biosynthetic strategies have also been explored for analoging; however, without the employ of high-yielding production strains, the low concentrations of many natural products in producing microbes–often reduced even further after enzymatic modification–severely limits the scale at which unnatural polyketide derivatives can be produced without metabolic engineering efforts. Given the possibilities provided by both biological and synthetic production of polyketides, a combinatorial approach offers a customizable and higher-yielding method of polyketide assembly.

Combining the best of biological and chemical synthetic methods, chemoenzymatic synthesis enables the biotransformation of synthetically accessible biomimetics of precursors and/or intermediates to rapidly access natural products and their derivatives in a targeted manner. This approach enables traditional chemical synthesis to shine by focusing on smaller fragments and takes advantage of the high level of stereoselectivity and specificity afforded by biosynthetic machinery. Utilized in the generation of derivatives of numerous natural products, including terpenes,^[15] alkaloids,^[16] and phenylpropanoids,^[17] chemoenzymatic synthesis is potentially well suited to produce polyketides, given the modularity of their biosynthesis.

In Nature, polyketides are biosynthesized by large enzymatic assembly lines, aptly named polyketide synthases (PKSs), which catalyze the extension of a growing polyketide chain via the decarboxylative thio-Claisen condensation of a malonic thioester or derivative thereof. While several variations of the PKS machinery exist, type I PKSs are noted for their structural diversity of their products, and are the primary focus of this review.These are megadalton assembly lines comprised of multiple polypeptides that are neatly organized into modules of highly conserved catalytic domains that, together, each catalyze a single chain extension step. It is important to note that this modularity is organizational rather than functional, as the swapping of modules will likely not result in product due to intermediate selectivity constraints. Minimally, chain extension requires a β-ketosynthase (KS), an acyltransferase (AT), and an acyl carrier protein (ACP). The beginning of this assembly line is typically initiated with a loading didomain (LDD), which loads the initial acyl-CoA building block, typically referred to as the starter unit. The acyl starter unit is then passed to the KS, while the AT domain selects an appropriately substituted malonyl-CoA building block, typically referred to as an extender unit. Once selected, this extender unit is transferred to the ACP domain, which is then interfaced with by the KS to catalyze the critical C-C bond formation. Following chain extension, reductive domains like the β-ketoreductase (KR), dehydratase (DH), or enoylreductase (ER) (which may or may not be included in a module) reduce the ketide unit to the desired oxidation state. Typically, this process is then repeated, once per module, until the elongated chain reaches the terminal thioesterase (TE) domain, where the product is cyclized and released before undergoing further functionalization by post-PKS tailoring enzymes (Figure 1).^[18] Alternatively, a minority of type I PKSs leverage hydrolysis rather than TE-mediated cyclization.^[19] Given that many of the active sites of domains use cysteine residues and trans-thioesterification reactions, research has often leveraged synthetic thioester substrates to introduce targeted modifications to polyketide products.

Figure 1:

Polyketide chemoenzymatic production. A: Enzymatic pathway for Erythromycin A (1) production using the DEBS pathway comprising EryA – EryH. Wavy lines represent ppant thioester handles. Use of synthetic substrates such as diketide-SNAC (2) enables portions of the pathway to be bypassed, or alternatively, enables the introduction of alternative functional groups. Synthetic thioester NAC enables uptake by PKS modules due to its structural similarity to the ppant handle of ACP domains. ACP: acyl carrier protein; AT: acyltransferase; DH: dehydratase; ER: enoylreductase; KR: ketoreductase; KS: ketosynthase; LDD: loading didomain; Ppant: phosphopantetheine; SNAC: N-acetylcysteamine; TE: thioesterase.

In contrast to type I PKSs, type II PKSs conserve similar mechanistic systems and outcomes, but via discrete monofunctional enzymes, rather than the connected megadalton polypeptides employed in type I. Additionally, type II PKSs tend to leverage heterodimeric KS dimers instead of the homodimeric structure employed by type I. Type II PKSs tend to react iteratively and without reduction, resulting in large, polycyclic aromatic compounds.^{[20, 21]} Additionally, type III PKSs differ from types I and II by acting as standalone KS domains that require no ACP-linked substrate to function, enabling the use of CoA thioester compounds in solution.^{[22, 23]} Types II and III are mentioned as relevant in this review, and the reader is directed to more detailed reviews for further information.^{[24, 25]}

This review surveys the experimental chemoenzymatic strategies utilized to produce polyketides and the contributions these efforts have made to the field (Figure 2). The leveraging of substrate promiscuity for the attachment of unnatural extender units to synthetic biomimetic intermediates is also discussed. While this review will primarily discuss type I PKS systems given that their linear ‘assembly line’ style pathway has inspired a ‘plug-and-play’ approach, other PKS scaffolds are also highlighted briefly as relevant.

Figure 2:

Representative transformations utilized in PKS chemoenzymatic syntheses and mechanistic investigations. Examples shown represent transformations accomplished using the archetypal DEBS system to make erythromycin analogs, although other transformations have been performed using additional PKS systems.

2. Synthesis of SNAC and Thioesters

An invaluable tool for polyketide chemoenzymatic synthesis is the synthetic thiol, N-acetylcysteamine (SNAC). SNAC is available commercially or via in-house amide formation by treating cysteamine with an acetyl electrophile (such as acetic anhydride or acetyl chloride), which can be completed, and the product purified in a few hours (Figure 3).^{[26, 27]} First popularized by Cane and Yang, SNAC offers an accessible thioester handle that mimics the far more expensive coenzyme A (CoA). In theory, SNAC can be leveraged as a substitute substrate for any enzyme that natively utilizes the CoA or phosphopantetheinyl (Ppant) thioester bonds due to the structural similarity, allowing for rapid experimentation and inquiry into biological systems. Additionally, the facile nature of trans-thioesterification renders these molecules excellent substrates for these enzymes, which leverage thioester handles in their native mechanisms. Other thioester replacements for CoA have been used (thiophenol, pantetheinyl, ppant, etc.); however, due to its ease of construction and installation, low price point, and lack of pungent odor, SNAC remains the most common.^[29] Commonly, SNAC is incorporated into the desired substrates by coupling of the carboxylic acid of the substrate with the free thiol of SNAC using diimide reagents like ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC),^[30] dicyclohexylcarbodiimide (DCC),^[31] and diisopropylcarbodiimide (DIC) (Figure 3B).^[31] Alternatively, thioesterification can be performed using thiophenol as a precursor to performing a trans-thioesterification with SNAC or CoA under mild conditions.^{[32, 33]} Similar strategies have also been employed to exchange SNAC with CoA via trans-thioesterification.^[34] This strategy offers the advantage of incorporating SNAC or CoA to substrates that have proven to be fragile or resistant to direct installation. Additionally, β-keto thioesters are often constructed using pyrolysis of Meldrum’s Acid (2,2-dimethyl-1,3-dioxane-4,6-dione) adducts to release carbon dioxide and acetone, which are readily removable (Figure 3B).^{[35, 36]} Furthermore, the use of trimethylaluminium enables the direct transformation from ethyl ester to thioester. This approach is often used for constructing thiophenyl thioesters,^[33] but more research is needed to probe the robustness of this method for direct NAC installation. While other coupling reagents and methods exist, the above approaches are most common and pertinent to polyketide chemical biology. The high efficiency and efficacy of these methods have led to their widespread implementation for the creation of synthetic thioesters for chemoenzymatic research.

Figure 3:

Synthesis and integration of N-acetylcysteamine (HSNAC) into synthetic polyketide intermediates. A: Typical synthesis of HSNAC through free-basing of the amine and acetylation. B: Methods of SNAC incorporation into acyl precursors. Free carboxylic acid can be coupled with the free thiol using coupling reagents. Thiophenol precursors enable trans-thioesterification. Use of Meldrum’s Acid allows for rapid thioesterification through decarboxylation

Complementing chemical methods, several enzymatic and chemoenzymatic methods of accessing thioester building blocks represent working approaches for biosynthesis of these building blocks. Commonly, malonates are chemically appended with the side chain of choice (if not commercially available) before enzymatic ligation with a malonyl-CoA ligase, such as MatB,^[37] which has demonstrated success in both in vitro and in vivo settings.^[38–42] This approach also lends itself to a broad array of monocarboxylic acids, which can be ligated to CoA using one of many available promiscuous acyl-CoA ligases, such as AcsA, FACL, or UkaQ, before serving as a precursor in polyketide biosynthesis.^[43–45] With this, an effective understanding of how various thioester building blocks are interconverted has enabled the construction of a chemoenzymatic roadmap between them. Leveraging synthetic coupling or acylation methods in tandem with acyl-CoA dehydrogenases (AcDHs), enoyl-CoA carboxylase/reductases (ECRs), and MatB, recent efforts have identified and coupled enzymes with high substrate tolerance to access a broad array of thioester substrates.^[46] Beyond the hypothetical, this approach was recently pushed to the extreme, resulting in the anaplerotic biosynthesis of 6-deoxyerythronolide B (6-DEB)— that is, no catalytic intermediates were introduced exogenously—purely from carbon dioxide.^[47] With such a diverse array of precursors and pathways available for polyketide products, researchers need only identify their target and an approach with the appropriate level of promiscuity or chemical reactivity, quickly and effectively enabling access to valuable compounds.

3. Stereoselective Biosynthesis of Building Blocks for Chemical Synthesis

Due to the high level of stereoselectivity exhibited by PKSs, chemoenzymatic methods that use them offer a reliable pathway to expand the chiral pool by creating new biologically sourced building blocks. Initial investigation demonstrated that ketoreductase (KR) domains could be neatly organized into four categories (A1, A2, B1, and B2) that reduced synthetic β-keto diketide substrates to the corresponding β-hydroxyl products in the (2R,3S), (2S,3S), (2R,3R), or (2S,3R) conformations, respectively, as well as two categories (C1 and C2) that do not reduce.^[48] This research also began to highlight that PKS domains contain some level of native promiscuity, enabling the utilization of such an approach on an array of compounds. Using this approach, a stereopure triketide lactone (TKL) was chemoenzymatically synthesized, trimming seven synthetic steps from the previously established synthesis of trans-9,10-dehydroepothilone D (Figure 4A).^[49] Using modules from the 6-deoxyerythronolide B synthase (DEBS), geldanamycin, soraphen, leptomycin, rifamycin, epothilone, and rapamycin PKS assembly lines, the degree of saturation and stereochemistry of the TKL product was controlled through module selection based on the presence of domains that dictate saturation (KR, dehydratase (DH), and enoylreductase (ER)).^[49] To quickly and effectively choose a module with the desired product, modules are categorized into types A through H, where the stereochemical and oxidative outcome of the product determines the module’s type.^[50] Leveraging specific domains for in vitro studies, researchers have also been able to complete stereospecific biotransformations using the KR domain. Using six KR domains from the DEBS or Tylosin (TYL) pathways, a panel of ketones were reduced to the corresponding alcohols, revealing a broad promiscuity in PKS KR domains, including a surprising selectivity for cyclohexyl moieties.^[51] A similar study was performed using KR domains from mycolactone (MLS) PKSs, resulting in a similar range of activities for accessing stereoenriched synthetic alcohols.^[52] More recently, studies have utilized the inherent stereochemical control of two KR domains (TylKR2 and MycKR6) to generate stereopure linear triketide carboxylic acids (Figure 4B).^[53] Additionally, a library of 51 KR enzymes was recently used to provide stereo-enriched pools of pharmaceutically relevant secondary alcohols from ketone substrates also containing halides, esters, and aryl moieties, effectively providing a robust platform to obtain useful building blocks in gram-scale quantities.^[54] While significant optimization is still required to improve yields in vivo, chemoenzymatic synthesis affords highly stereo-enriched products with the potential to be used as building blocks in future critical advances in the fields of synthetic methodology and chemical biology with potential applications in medicinal chemistry and drug development.

Figure 4:

Construction of polyketide derived building blocks for further modification. A: Chemoenzymatic construction of lactone 3 reduces the synthetic pathway to aldehyde 4 by 7 steps, readily enabling the production of trans-9,10-dehydroepithilone D (5). B: Targeted use of known KR stereospecific outcomes enables production of stereopure triketide products. Reduction of β-keto diketide thioester 6 using different KR domains results in β-hydroxy thioesters 7a and 7b. Following polyketide chain extension, the new β-keto products 8a and 8b are reduced to stereopure products 9a, 9b, 9c, and 9d.

4. Chemoenzymatic Synthesis for Polyketide Derivatization

Perhaps the most attractive use of a chemoenzymatic approach to polyketide synthesis is the ability to obtain natural product analogs with desired variations quickly and effectively in vitro. Assuming the discrimination of downstream PKS machinery for unnatural substitutions can be overcome, synthetic biomimetic fragments can be readily produced and subsequently utilized by PKS assembly lines via “precursor-directed biosynthesis.” In this approach, the synthetic precursor replaces the intermediate that would otherwise be produced by preceding machinery.^[55] SNAC-linked analogs of various chain lengths have been leveraged for targeted regioselective derivatizations (Figure 5). For example, diketide-SNAC mimics of the diketide acceptor substrate have been utilized by the DEBS pathway to yield C14-substituted derivatives of erythromycin.^[56] Remarkably, the enzymatic pathway incorporated numerous unnatural starter units, including diketides with altered stereochemistry, aryl moieties, and long saturated and unsaturated aliphatic chains at this position. Moreover, this system produced an analog with a terminal alkyne tag, which could be leveraged in biorthogonal “click chemistry” reactions. Such a modification enabled visualization and quantification of tagged species with a fluorescent tag, which has been shown to be possible using sulforhodamine B azide^[57] or coumarin^[58] derivatives (Figure 5A). These strategies have also been applied to the iterative type III csypyrone B1 (CsyB) synthase, wherein a panel of non-natural aliphatic diketide-SNAC species was utilized to produce sixteen different pyrones with side chains of differing lengths.^[59] Longer chain SNAC analogs have also been utilized for diversification strategies, including a recent study where a hexaketide-SNAC was used to repurpose the hypothemycin synthase (Hpm3) to produce (6’S, 10’S)-trans-7’,8’-dehydrozearalenol (11), as well as other, shorter unnatural resorcylic acid lactones (12–14) (Figure 5B).^[60]

Figure 5:

A: Use of alkyne-tagged building blocks enables fluorescent tagging of polyketide intermediates, hydrolyzed shunt products, and product lactones via copper-catalyzed click chemistry. B: Condensation of acetyl-CoA and malonyl-CoA catalyzed by hypothemycin PKS enzymes Hpm3 and Hpm8 results in production of resorcylic lactones (12 – 14) and (6′S, 10′S)-trans-7′,8′-dehydrozearalenol (11). Feeding synthetic hexaketide-SNAC precursor 10 eliminates lactone shunt product formation and yields the target macrolide 11 alone.

Leveraging synthetic SNAC mimics of natural chain intermediates, numerous research efforts have sought to regioselectively modify existing polyketide scaffolds via non-native chain extensions. While each acyltransferase (AT) domain has evolved to select a single extender unit, these enzymes have not evolved to discriminate against unnatural or uncommon extender units that the bacterial producer does not naturally encounter. Using DEBS Mod6TE, Koryakina et al. demonstrated the synthesis of several unnatural triketide lactones via the extension of a synthetic diketide-SNAC with a panel of C2-substituted malonyl-CoA extender units (Figure 6A).^[39] While the natural extender unit was preferentially incorporated in competition studies, which could pose a challenge to selective in vivo incorporation, the work clearly illustrated the inherent promiscuity of a polyketide module for unnatural substrates. The phoslactomycin (PnC) PKS has also been demonstrated to have significant promiscuity towards various small-to-medium C2-substituted malonyl-CoA analogs, including methyl-, ethyl-, and butylmalonyl-CoA. However, it was unable to successfully incorporate analogs with larger substituents (e.g. hexyl- and benzylmalonyl-CoA) in high yields, underscoring the importance of steric accommodation within the active site of the AT domain (Figure 6B).^[61] Conveniently, this selectivity trend has been observed in many pathways, especially those with trans-AT domains. For context, most type I PKS modules employ their AT domain, which is integrated into the modular polypeptide, (cis-ATs). Some exceptions, however, leverage separately expressed AT domains (trans-ATs), which allow for alternate protein-protein interactions, and alternate patterns of substrate selectivity.^[62] The promiscuity of the loading domains has also been probed using the orthogonal type III PKS stilbene synthase (STS) found in Pinus sylvestris, which has been leveraged to produce novel pyrones from seven synthetic starter units (Figure 6C).^[34] In the same vein, Hansen et al. tested a panel of synthetic pentaketide-SNACs to determine the promiscuity of the pikromycin (Pik) PKS TE.^[63] By changing the length and stereochemistry of groups at the end of the SNAC analogs, this work assessed the tolerance of the TE for upstream structural effects, analogous to using different starter units or an unnatural first extender unit. As anticipated, the native substrate displayed increased conversion over derivatives. Not limited to large synthetic substrates or NAC thioesters, this approach also lends itself to the construction and implementation of substituted carboxylates that have yet to be converted to the corresponding thioester. Using a synthetic panel of terminally fluorinated aliphatic potassium carboxylates, the incorporation of fluorine into the core scaffold of ErA (1) was demonstrated for the first time using an engineered promiscuous strain (ERMD1) of Saccharopolyspora erythraea.^[64] Importantly, this research demonstrates that some targeted modifications are tolerated by downstream machinery, enabling early installation of desired chemical groups into bioactive drug leads (Figure 6D). Encouragingly, the ability of an unnatural substrate to be accepted represents an entry point for natural product diversification, with enzymatic, metabolic, or strain engineering offering promising pathways to improving unnatural product yields.

Figure 6:

PKS promiscuity enables diversification of product structures. A: DEBS Module 6 incorporated a panel of C2-substituted malonyl-CoA substrates, enabling the production of C2-substituted pyrones. R¹ = H, Me, Et, Ppg, All, Bu, EtPh, OMe, EtN₃. NADPH was not used to bypass the activity of the KR, resulting in aromatic pyrone products. B: An aliphatic panel of C2-substituted malonates was accepted by the iteratively acting phoslactomycin Synthase PnB, containing modules 2 and 3, to generate a panel of hexaketide products. R² = Me, Et, Bu, (3-Me)-Bu, Hex, Bn. C: The stilbene PKS, STS, accepts a panel of acyl-CoA starter units, generating a panel of C5-substituted pyrones. R³ = 4-methylthiazole-5-carboxyl, 2-methylthiazole-5-carboxyl, 2-chloro-1,3-thiazole-5-carboxyl, 5-(4-fluorophenyl)-pentanoyl, 6-phenylhexanoyl, 7-phenylheptanoyl, and 4-biphenylacetoyl. D: Synthetic production of potassium 4-fluorobutanoate enabled production of a fluorinated precursor of 1.

Alternatively, mutasynthesis offers an exciting route for polyketide diversification in a targeted manner. By inactivating portions of the native pathway to favor the uptake of introduced analogs, selectivity concerns can be largely mitigated. This approach was leveraged in the construction of a panel of FK506 derivatives, replacing the natively produced allylmalonyl-CoA building block with three alternative side chains.^[65] By knocking out the genes responsible for allylmalonyl-CoA construction, concerns about the selectivity of AT4 were removed, as it would be impossible for the AT to encounter its native substrate. Additionally, the above-mentioned examples of trans-AT supplementation are also examples of mutasynthesis, as the native cis-ATs were inactivated via active site mutagenesis. While attractive, mutagenesis relies on native promiscuity and genetic engineering, and is therefore best implemented when coupled to domain and module engineering platforms. Other reviews have highlighted mutasynthesis more extensively, and the reader is directed to these for more examples.^{[66, 67]}

To target unnatural products, the selectivity of these pathways may be engineered to enhance the incorporation of a target substrate, while shifting selectivity away from the typical, natural substrate. Given the gatekeeping function of the AT for extender unit selection, it has often been the primary target of such efforts. For example, Koryakina et al. successfully inverted the specificity of DEBS Mod6AT with a V187A/Y189R mutation, which selected synthetic propargylmalonyl-CoA over the native methylmalonyl-CoA in competition studies (Figure 7A).^[40] Utilizing these mutations, C2-substituted analogs of 10-deoxymethynolide (10-dml) and narbonolide were preferentially produced in competition studies with the native substrate. Building on this work, Kalkreuter et al. leveraged site-selective mutations in PikA3-AT and PikA4-AT to produce the first example of consecutive installation of multiple unnatural extensions, opening the possibility of integrating unnatural extender units at any position in the PKS pathway.^[41] Alternatively, leveraging molecular dynamics and genomics guided analyses, a panel of Val295 mutants of DEBS AT6, including V295L, V295G, and V295A, were utilized to successfully incorporate the propargyl group at the C2 position of erythromycin A.^[68] Notably, the V295A mutation enabled the incorporation of three unnatural extender units (ethylmalonyl-, allylmalonyl-, and propargylmalonyl-SNAC) to produce three C2-substituted erythromycin A analogs (Figure 7B).^[69]

Figure 7:

Incorporation of non-natural extender units in the core scaffold of methynolide- and narbonolide-type structures. A: Using AT mutants in 1- or 2-module systems with synthetic pentaketide substrate enabled inversion of extender unit specificity and subsequent installation of a second non-natural extender unit. B: The introduction of an AT mutation in the production pathway of erythromycin A enabled the formation of C2-substituted natural product analogs.

Natural promiscuity and specificity can also be leveraged to obtain regio- and chemo-selectively modified polyketide derivatives. For example, the promiscuous monensin (Mon) Mod5AT is capable of taking up three unnatural extender units (allyl, propyl, and propargyl) in addition to the typically accepted malonate extender units (methyl and ethyl), leading to three novel analogs of monensin,^[70] which were then fully functionalized by post-PKS machinery. Alternatively, highly-selective domains can be introduced into nonnative contexts to produce chimeric modules and domains, allowing for transplatation of their specificity and substrate scope.^[71] More broadly, recent work has shown that exchanging AT domains with those from orthogonal PKSs enables the incorporation of alternative groups, from small (H or Me) to large (Ph, ⁿhex, etc).^[72] Leveraging this approach, a large panel of TKLs was quantified, enabling robust structural diversification. More specifically, targeting key residues which differ between AT active sites has been shown to enable a wider substrate scope in naturally specific modules. By swapping the selectivity motifs of the DEBS AT6 domain with selectivity motifs from the cinnabaramide (Cin1) and thailandin (Tha13) AT domains, Kalkreuter et al. generated a chimeric AT domain with increased selectivity for unnatural malonates, namely propargylmalonate, butylmalonate, and pentylmalonate (Figure 8).^[73] This combinatorial protein engineering approach builds on other PKS engineering advances to open the door for the construction of PKS systems via a “plug-and-play” strategy, enabling targeted functionalization. Alternatively, this type of incorporation has enabled structural modifications that lead to next-generation antibiotics. Indeed, the goal of increasing access to solithromycin and other similarly substituted designer fluorinated macrolides has attracted attention recently.^[74] Two recent, contemporaneous studies aimed to demonstrate and optimize fluorine incorporation into the core scaffold of narbonolide-based polyketide products. Leveraging a synthetic 2-fluoro-2-methylmalonyl-CoA precursor, a chimera of DEBS Mod6TE containing a promiscuous murine fatty acid synthase (FAS) AT domain (specifically, a malonyl/acetyl transferase, or MAT) enabled the regioselective incorporation of the gem-fluoro-methyl motif at the C2 position of a 14-membered lactone, with the stereochemical outcome computationally confirmed to be in the desired (S) outcome matching solithromycin.^[75] In parallel, investigation of the installation of fluorinated extender units demonstrated that a single amino acid mutation (F190V) to the trans-AT from the disorazole biosynthetic pathway (DszAT) enabled a specificity shift away from malonyl-CoA and toward fluoromalonyl-CoA.^[76] Targeted replacement of DEBS AT5 or AT6 enabled the regioselective incorporation of fluorine at C2 or C4 of the corresponding 6-DEB analogs. While it is believed that this approach can be leveraged at any AT domain in the 6-module pathway, further research is needed to demonstrate this effect within more upstream modules. Advances in genetic engineering and knowledge of module substrate specificities and promiscuity will enable a more targeted approach to the design of novel polyketides.

Figure 8:

Generation of AT chimeras enabled the site-selective incorporation of unnatural extender units, effectively diversifying the core polyketide scaffold. Targeted chimera construction using selectivity motifs from modules with differing activity profiles confers the activity toward non-natural groups.

An emerging branch of chemoenzymatic synthesis research aims to derivatize products from natively occurring non-ribosomal peptide synthetase (NRPS)-PKS hybrid pathways, which produce several natural products including tenuazonic acid (15) and swainsonine (16) (Figure 9A).^[77,78] Typically, the NRPS catalyzes the formation of the starter unit, which is then passed to the PKS for typical malonate-driven extension.^[79] In an effort to use this type of assembly line for the production of natural product derivatives, synthetic phenylalanine analogs were leveraged as starter units for the Aspergillus niger fungal nonribosomal peptide synthase non-reducing polyketide synthase (NRPS-NRPKS) hybrid enzyme (AnATPKS) to obtain analogs of pyrophen (17) (Figure 9B).^[80] These type I PKSs are unique in their repetitive, iterative, monomodular nature, and still produce a wide range of products.^[81] While the primary focus of this study was a phenylalanine panel of derivatives, it is likely that similar natural products such as campyrone A, B, and C, which structurally resemble pyrophen are naturally produced in the same manner. Likewise, a chemoenzymatic approach enabled the stereodivergent production of NRPS-PKS hydrid-like products that highlighted the molecular promiscuity of a panel of TE domains.^[82] Combining polyketide and macrocyclic peptide motifs, the researchers hoped to created a platform for the construction of potent antibiotic scaffolds. This research illustrates the promising capacity of NRPS-PKS hybrid systems to be used for the chemoenzymatic synthesis of natural product derivatives. Equally encouraging results have recently been demonstrated outside of macrolide-like polyketides, demonstrating the broad utility of this approach. Using a synthetic panel of polyketide-derived β-orcinaldehyde derivatives, a new-to-Nature panel of stipitaldehyde compounds was constructed due to the promiscuity of a library of non-heme iron(II)-dependent oxygenases like TropC.^[83] If coupled with known patterns of PKS engineering, this platform may offer a route for in vivo production of such compounds for further study, as well as aiding in complex natural product retrosynthetic analyses.

Figure 9:

Utilizing promiscuity afforded by polyketide scaffolds to derivatize NRPS-PKS products. Portions of structures provided by PKSs are highlighted in red, while NRPS portions (typically amino acids) are highlighted in teal. A: Example products generated using NRPS-PKS systems. B: Use of phenylalanine derivatives enabled the production of substituted pyrophen analogs.

5. Probing Inherent PKS Promiscuity via Chemoenzymatic Synthesis

Synthetic substrates can also be leveraged to probe the promiscuity of PKSs for cross-system functionality. Despite subtle differences in the macrolactone produced by PKSs such as DEBS and Pik, similarities in chain length, electronic character, and module sequence have enabled orthogonal extensions when incubated with pentaketide- and hexaketide-SNAC.^[84] The Pik pathway yields both the 12- (10-dml) and 14-membered (narbonolide) lactones when the proteins are incubated in vitro with a native pentaketide precursor, indicating the ability of the pathway to perform either a single or double extension, respectively.^[85] high promiscuity of PikA4, resulting in two products–one as the result of PikA3 and PikA4 extension and one using only PikA4.^[86] Similar activity has been noted when in vitro evaluation of DEBS3 (containing modules 5 and 6) has occurred. This promiscuity is mostly mitigated in the DEBS assembly line, however, as the final modules are linked as a single polypeptide, reducing the likelihood that unfinished intermediates will reach the promiscuous final module.^[86] Recently, TE flexibility was evaluated through the construction of hybrid PKSs, which contain homologous or heterologous TE domains. Overall, the results demonstrated the tendency of these domains to act as bottlenecks for product formation.^[87] By comparing the promiscuity of different PKS pathways across systems, the similarities and orthogonality of various substrate specificities allow for the design of logical pathways to new natural product derivatives.

6. Conducting Mechanistic Studies of PKS Machinery via Chemoenzymatic Synthesis

The chemo-enzymatic synthesis of probes has contributed to elucidating PKS mechanisms via the introduction of biomimetic intermediates and subsequent observation of product formation through various analytical techniques.^[88] Due to the clinical success of solithromycin, a fluorinated analog of erythromycin, a recent investigation determined the mechanistic impacts of introducing a fluorinated malonate extender unit into PKS machinery.^[27] In an attempt to specifically bypass the native AT specifically, the cis-AT was inactivated leading to a mechanistic change that drastically reduced the participation of the ACP domain in the mechanism, limiting the amount of extension that could take place. However, introducing a trans-AT from a different pathway—the disorazole biosynthetic cluster (DszAT or DszS) significantly increased the amount of fluorinated extension that could occur, even permitting consecutive chain elongation cycles with fluorinated extender units. While cis-AT domains are ATs embedded within the PKS polypeptide, trans-AT domains are separately encoded ATs.^[89] Recent work with the trans-AT domain used in the biosynthesis of zwittermicin (ZmaF) explored the molecular and mechanistic requirements for the recruitment of the enzyme to a cognate ACP.^[90] While this integration did not involve chemoenzymatic synthesis per se, the increased promiscuity of the trans-AT may enable future extensions that prove difficult to incorporate via native cis-AT domains. This work ultimately led to the conclusion that trans-acylation of the ACP is likely the rate-limiting step in the use of trans-ATs, and simultaneously demonstrated the wide promiscuity of the domain toward unnatural extension. This was also shown to be the case in the acylation of the trans-acting AT, KirCII, despite its broad substrate scope that enabled the diversification of the kirromycin structure.^[58] While acylation of the trans-AT itself has been shown to be relatively facile, the offloading of the acyl group to the ACP presents a mechanistic bottleneck. Cumulatively, these studies highlight the more narrow range of extender unit specificities of trans-ATs, and the lack of knowledge about how they interact with their cognate ACPs hinder further use. An additional hurdle to consider when using this approach is that a known trans-AT with high specificity is needed before such a transformation can be completed. Similarly, the mechanism of dehydratase (DH) domains was investigated in Pik module 2 as products with olefin moieties were produced after the introduction of β-hydroxy diketide SNACs, resulting in extension by the reducing module. DH function was probed by investigating H3611F mutants of Pik DH2, resulting in loss of DH activity.^[91] This study leveraged their findings to further illustrate the mechanistic importance of the widely-conserved HxxxGxxxxP motif across PKS DH domains.

Beyond canonical macrolide examples, mechanistic insights into alternative PKS product pathways have also been obtained. Effectively, Heine et al. leveraged a triketide-SNAC to demonstrate that the rhizoxin PKS sets the stereochemistry of two branches simultaneously using the enone motif of the substrate as a Michael acceptor in rare fashion.^[92] Using several reference compounds, a panel of β-branching polyketide analogs was obtained. A similar study was performed with a recombinant type III PKS derived from Scutellaria baicalensis leveraging various SNAC species to elucidate the mechanism of chalcone synthesis.^[93] In another example, an iterative extension by the native PKS of both malonic and fluoromalonic precursors were reported in the formation of thiotetronate natural products.^[94] While slightly different from their macrolide cousins, other polyketide subclasses’ products have the same capacity to help researchers elucidate assembly line mechanistic details like their counterparts. Likewise, another study used pentaketide-SNAC to clearly illustrate the cyclization mechanism of the elaiophylin (Ela) PKS, which catalyzes iterative chain ligation to afford a macrocyclic diester.^[95] These results were further verified with an additional unnatural polyketide-SNAC, and a novel diolide was observed.

7. Diversification via Post-PKS Chemical and Biological Transformations

While the majority of chemoenzymatic synthesis of polyketides has focused on the PKS, bioactivity often results from post-PKS modification.^[96] An early investigation into post-PKS tailoring consisted of an in vivo system containing no endogenous PKS assembly line and only the post-PKS enzymes DesVII and PikC for C5-glycosylation and C12-hydroxylation of narbonolide, respectively.^[97] The results demonstrated that post-PKS tailoring enzymes operate independently of the the PKS machinery, as synthetic narbonolide was used to test the activity of the tailoring enzymes in vitro. Since these tailoring enzymes operate independently, their capacity to act on non-native substrates can be readily evaluated through in vitro experiments, enabling the rapid diversification of drug-like molecules produced in research studies.

Similarly, the identification of the A242V/S132F/P26T (ASP) triple mutant of the glycosyltransferase from the oleandomycin post-PKS pathway enabled the glycosylation of a broad panel of substrates with differing structural motifs, including macrolide polyenes, enediynes, aromatic systems, and heterocycles.^[98] The identification of this mutant may enable glycosylation to be performed on future designer substrates whose biosynthesis is not associated with an established glycosyltransferase system. In a later study, Hansen et al introduced the native biomimetic pentaketide-SNAC to the final portion of the pikromycin PKS pathway to obtain both narbonolide and 10-deoxymethynolide (10-dml), before carrying them through the post-PKS tailoring pathway to obtain several bioactive macrolides (Figure 10A).^[26] This study highlighted the promiscuity of the tailoring enzymes in the Pik pathway via conversion of unnatural substrates into biologically active final products. Likewise, a hexaketide-SNAC was introduced into the juvenimicin (Juv) pathway to obtain tylactone, and then the researchers used the post-PKS tailoring enzymes MycC1, JuvD, Tyl1 as well as a TEMPO-mediated oxidation of a primary alcohol to an aldehyde to obtain 10 distinct macrolides.^[99] Furthering this approach, synthetic research efforts led to the development of solithromycin (26), a fourth-generation macrolide stemming from clarithromycin (27), the 6’-O-methyl derivative of erythromycin A; groups typically not observed in natural PKS products, such as a cyclic carbamate, an aryl-1,2,3-triazole ring, and a fluorine atom, were installed on the canonical PKS product scaffold using chemical transformation strategies, enabling the construction of a new macrolide with increased antibiotic activity and increased stability.^[100] While solithromycin was contrived in a semi-synthetic manner, established precursor-directed and chemoenzymatic approaches (such as unnatural malonate synthesis and uptake) have been utilized to increase accessibility to derivatives en route to this designer antibiotic. Additional chemical techniques have enabled the construction of other analogs such as telithromycin and nafithromycin (28) which leverage unique structural variations to interact with bacterial ribosomes in novel ways.^[101] Outside of macrolide core diversification, acyltransferase engineering has enabled the construction of lovastatin analogs. By first assessing the native promiscuity to ensure a way forward, subsequent directed evolution afforded a shift in substrate specificity, allowing the AT LovD to produce simvastatin instead of lovastatin.^{[102, 103]} Given the demonstrated broad substrate tolerance of post-PKS tailoring enzymes and the robust nature of semi-synthetic chemistry, their ability to assist in the production of new-to-Nature drugs represents a compelling argument. Together, chemoenzymatic and semi-synthetic strategies synergistically provide exciting avenues to pursue clininally-relevant compounds, hopefully enabling widespread industrial use.

Figure 10:

A: Module-catalyzed extension upon synthetic pentaketide 18 using methylmalonyl building blocks results in 10-deoxymethynolide (10-DML) 19 and narbonolide 20 products, which are then both decorated by tailoring enzymes to yield bioactive macrolides 21 methymycin (R¹=OH, R²=H), 22 neomethymycin (R¹=H, R²=OH), 23 novamethymycin (R¹=OH, R²=OH), 24 ketomethymycin (R¹=H, R²=O), and 25 pikromycin. B: Examples of next-generation semi-synthetic macrolides stemming from erythromycin. Portions of structures appended in synthetic steps are highlighted in teal.

8. Conclusions and Future Outlooks

The current state-of-the-art in chemoenzymatic polyketide synthesis has critically elucidated the synthesis of natural products and the derivation of substrates. These studies have enabled the construction of natural product analogs, probed the inherent promiscuity and specificity of new and existing PKSs, and unveiled how these assembly lines function at the molecular level. While this article highlights numerous novel approaches for growing the collective understanding of the field of such systems, the complexity of these synthases and their products continues to pose challenges, which require the ongoing development of new methods. The advantages of computational chemistry are poised to provide new insights into polyketide scaffolds. Leveraging the power afforded by computational storage and modeling, a database of nearly 14,000 macrolactones has been compiled and organized using filters (e.g., molecular weight, number of esters in the macrocycle, number and identity of sugars bound, etc.) to furnish the web-based application, MacrolactoneDB.^[104] This online database uses modeling and chemical space visualization to identify target macrolactones for specific purposes and offers promise as new compounds are sought as medicinal agents. On a much larger scale, the online database SIME (Synthetic Insight-Based Macrolactone Enumerator) allows users to set the parameters for the generation of macrolactones with certain features and has been used to generate V1B, an in silico library of 1 billion theoretical macrolides.^[105] The generation and modeling of theoretical structures hold the promise of identifying promising targets of chemoenzymatic methods. Once an area of untapped potential, recent advances in computational power and the ability to access designer macrolides have made this method of drug development a reality; a recent study utilized in silico modeling to generate a panel of C3-O-acyl derivatives of the macrolide clarithromycin and validated their antibacterial efficacy via in vivo tests against bacterial cell lines.^[106] This method of drug development resulted in the construction of several new compounds which reduced the IC₅₀ values by as much as 3-fold. Further goals involve the ability to completely derivatize macrolides at every position, the uptake of new synthetic feedstocks, and the use and synthesis of novel extender units. To that end, PKSs are poised to have their repertoire expanded with the use of more synthetic efforts. Furthermore, continued probing with synthetic substrates consisting of larger and more complicated chains, the collective understanding of both the capacities and limitations of PKSs to produce increasingly complex natural product derivatives will grow. While screening the integration of new extender units is traditionally performed using a “one-substrate-one-enzyme” approach, new methods have shown that testing multiple substrates simultaneously against multiple mutants allows for the rapid discovery of useful MatB mutations, even “hidden” mutations that are unable to be discovered in a single-substrate screen and enable the conversion of a broad panel of malonate substrates.^[107] Additionally, the use of an in vitro CRISPR-Cas system may enable a more rapid approach to PKS mutagenesis.^[108] Partnered with continued biological assays and testing against cell lines, the rapid chemoenzymatic synthesis of polyketides may uncover new potential pharmaceuticals.

Biographies

Thaddeus Paulsel obtained his B.Sc. in Chemistry in 2019 from the University of Georgia, where he studied chemical methods for the decarboxylation of amino acids and assisted in the development of pedagogy for undergraduate laboratory experiments. In 2019, he joined the Williams lab at NC State University, where his research focuses on the precursor-directed biosynthesis of new-to-Nature polyketides.

Gavin Williams received his Ph.D. in chemical biology from the University of Leeds. He completed postdoctoral research at the University of Leeds with Prof. Adam Nelson and Prof. Alan Berry where he created tailored aldolase enzymes for the synthesis of sugars. He then moved to the University of Wisconsin at Madison as a research scientist with Prof. jon Thorson where he engineered enzymes involved in natural product glycosylation. He joined NC State University in 2009 where his research group uses protein engineering, metabolic engineering, and synthetic biology to reprogram the biosynthesis of secondary metabolites, including polyketides and terpenes.