Synthetic Biology Enabling Access to Designer Polyketides

Abstract

The full potential of polyketide discovery has yet to be reached due to a lack of suitable technologies and knowledge required to advance engineering of polyketide biosynthesis. Recent investigations on the discovery, enhancement, and non-natural utilization of these biosynthetic gene clusters via computational biology, metabolic engineering, structural biology, and enzymology-guided approaches have facilitated improved access to designer polyketides. Here, we discuss recent successes in gene cluster discovery, host strain engineering, precursor-directed biosynthesis, combinatorial biosynthesis, polyketide tailoring, and high-throughput synthetic biology, as well as challenges and outlooks for rapidly generating useful target polyketides.

Graphical Abstract

Introduction

Nature has evolved diverse enzymatic machinery for the assembly of highly complex small molecule natural products. Biosynthesized by polyketide synthases (PKSs), polyketides are a large class of natural products and represent a significant source of new drugs, molecular probes, and bioactive small molecules [1,2]. Many polyketides are blockbuster drugs such as erythromycin (antibiotic), epothilone B (antitumor), and lovastatin (anticholesterol), while others, such as solithromycin, are currently in clinical development [3].

Within canonical type I PKS assembly lines, each module is responsible for the incorporation of a single malonyl-derived extender unit. Minimally, modules are comprised of an acyltransferase (AT), acyl carrier protein (ACP), and ketosynthase (KS) catalytic domains, which enable extender unit selection and subsequent decarboxylative Claisen condensation between the extender unit and the growing chain. Additional in-line tailoring domains, including the ketoreductase (KR), dehydratase (DH), enoylreductase (ER), or more rarely, methyltransferase (MT), can also further site-selectively modify the alkylation and oxidation patterns within the growing polyketide chain. The final elongated chain is then cleaved from the PKS and cyclized by a thioesterase (TE) domain to yield a core macrolactone, which can be further decorated by post-PKS enzymes [4].

Despite their broad chemical and structural diversity, naturally occurring polyketides often require optimization for a given application [5]. For example, the macrolide antibiotic erythromycin was synthetically modified to prevent cyclization in the acidic gastric environment, for increased bioavailability, and for additional molecular interactions with the ribosome, spawning multiple generations of improved antibiotics. While synthetic methods have successfully modified polyketide structures, biosynthetic modifications offer a scalable and potentially facile approach for regio- and stereoselective scaffold diversification. Herein, this review describes the current state-of-the-art in synthetic biology to enable access to designer polyketides, leveraging the modularity of PKS machinery. We highlight recent works that have expanded the polyketide chemical space through pathway discovery, refactoring, and engineering and reflect on the future outlook of synthetic biology approaches to polyketide analoging (Figure 1).

Figure 1.

Accessing designer polyketides leveraging diverse engineering strategies. Native pathways are shown in light grey, while engineered and non-natural pathways are shown in dark grey. Modifications to the PKS and post-PKS tailoring enzymes are highlighted in their corresponding colors.

Polyketide Biosynthetic Gene Cluster Discovery

Though there are many elucidated polyketide synthases (PKSs), these pathways likely represent a portion of the genetic machinery available in Nature [6]. Moreover, high rediscovery rates continue to hinder the identification of novel natural compounds. Therefore, several “omics-guided” methods such as transcriptomics [7], metabolomics [8], and genome mining (metagenomics) [9,10] have been utilized in polyketide biosynthetic gene cluster (BGC) discovery. These approaches expand the polyketide engineering toolbox, but more significant barriers for accessing non-natural polyketide derivatives including host selection, starter and extender unit availability, and pathway design, must be overcome.

Accessing Polyketides via Host Strain Selection and Engineering

Exploring and optimizing polyketide biosynthesis can be completed within native production strains given that the BGC and necessary precursors are often already functionally expressed and produced. Typically, enhancing polyketide production focuses on the redirection and optimization of carbon flux via enhanced precursor availability [11,12], promoter and ribosome engineering [13], removal or overexpression of regulatory elements [14,15], deletion of competing biosynthetic pathways, and combinations thereof to significantly enhance titers above that of the wild-type pathway. Strategies including cooperative induction [9], co-culturing [16], and transcription factor decoys [17] have been explored for the activation of cryptic BGCs (cBGCs). The expression of cBGCs has also been enabled through the deletion of “primary” PKS pathways [18,19]. While these methods have been useful in specific well-studied, genetically tractable hosts, the full potential of polyketide cBGCs will not be fully realized until there is a global strategy for pathway activation under traditional and industrially relevant fermentation conditions.

Heterologous pathway expression in well-characterized chassis has addressed challenges posed by native host expression. Heterologous platforms for the rapid and robust biosynthesis of polyketides derived from type I [20,21], II [22,23], and III [24] synthases, offer key opportunities for synthetic biology including the expression of cBGC, development of chimeric PKSs through “plug and play” modifications, and precursor-directed engineering [22]. Usually, the goal of heterologous expression is to decouple secondary and primary metabolism to enhance polyketide production. A recent omics-guided approach in Streptomyces revealed that primary metabolism derived triacylglycerols can limit polyketide biosynthesis. This was overcome by dynamically degrading triacylglycerol and led to enhanced titers of actinorhodin, jadomycin B, oxytetracycline and avermectin B1a in various Streptomyces strains [25]. However, direct BGC transfer [13] and their expression [22] present numerous obstacles including significant cellular burden, resulting in poor expression or inefficient cellular maintenance of the BGC [20]. Advances in DNA manipulation and genomic integration technologies, including transformation-associated recombination (TAR) [26] and direct pathway cloning (DiPaC) [27], continue to make these BGC transfers more feasible.

Altered Macrolactone Sidechains through Precursor-Directed Biosynthesis

The AT is often a target of engineering given that it dictates large portions of the macrolactone structure. Extender unit specificity of AT domains has been altered through domain and motif swaps while targeted mutagenesis strategies have enabled site-selective integration of non-natural building blocks. Notably, AT mutagenesis of the final two modules of the pikromycin PKS revealed the unprecedented capability to introduce consecutive non-natural extender units into a macrolactone [28]. In another recent example, mutation of a conserved tryptophan switched specificity of an AT from ACP- to coenzyme A (CoA)-linked extenders [29], providing an additional potential route to polyketide diversification.

While many of these efforts have had limited utility to in-line, or cis-ATs, opportunities for expanding polyketide chemical space also exist via complex free-standing trans-AT pathways [30]. Though there are examples of trans-ATs like the orthogonal and promiscuous ZmaF whose activities have been probed, their non-natural utility and ability to be leveraged by enzyme engineering in general has not been well explored [31]. Non-canonical modular junctions within these systems, often between a KS and DH, make their products difficult to predict through traditional genome mining [32]. However, programs such as TransATor, which predicts trans-AT containing PKS products by comparing a PKS to KS domains in BGCs and their associated downstream modular enzymes, are powerful tools for discovery [33].

While native and engineered AT promiscuity facilitates the biosynthesis of novel polyketides, the modest diversity of extender units [34] beyond malonyl-CoA (mCoA) and methylmalonyl-CoA (mmCoA) has restricted in vivo efforts. To expand the existing in vivo repertoire of non-natural extender units, mCoA synthetases have been used for the activation of diverse C2-substituted malonates. A native mCoA synthetase from Streptomyces cinnamonensis was used to synthesize allyl-, propargyl-, and propyl-CoAs for the production of monensin analogues [35]. The native promiscuity of enoyl-thioester carboxylase/reductases (ECRs) [36,37] have also been leveraged to produce non-natural extender units in vitro. Moreover, halogenases such as SalL can diversify precursors to produce chlorinated and fluorinated mCoA analogues (Figure 2) [38,39]. Notably, once incorporated, these halogenated analogues can be further leveraged as chemical handles for downstream cross-coupling reactions for further derivatization, as has been done with other classes of natural products [40].

Figure 2.

The development of malonyl-CoA analogues for introduction into polyketides. Top: Biosynthesis of malonyl-CoA and various C2-substituted malonyl-CoA analogues. These acyl-CoAs are then utilized by PKSs to produce naturally occurring polyketides and their analogues. Bottom: Panel of previously biosynthesized C2-substituted analogues (See text for references).

Access to Designer Polyketides via Combinatorial Biosynthesis of PKSs

The templated biosynthesis of polyketides by type I PKSs implies the modularity of biosynthetic machinery. Leveraging this paradigm, novel designer molecules are accessible through a “plug-and-play” strategy, wherein modifications to the assembly line, including the insertion, deletion, and exchange of key domains/modules afford predictable scaffold diversification. However, significant changes to protein structure often impair or inactivate the chimeric PKS.

Emerging bioinformatic and evolutionary analyses have challenged the canonical KS_n-AT_n-ACP_n module structure, suggesting that the boundaries be re-defined as AT_n-ACP_n-KS_n+1 [41]. Construction of a hybrid pikromycin-venemycin pathway (Figure 3) with these newly delineated boundaries resulted in higher turnover rates than the traditional boundary swaps and incorporation of the non-natural starter unit 3-hydroxybenzoic into a small combinatorial library of molecules, though non-natural extender selectivity was not explored [42]. Others have also independently identified the KS-AT linker site as a key target for homologous recombination in type I PKSs [43].

Figure 3.

Hybrid synthases of the venemycin PKS (top) were constructed with traditional domain boundaries between the KS and ACP (middle). Module swaps were then constructed using boundaries between the KS and AT (gray box). The new boundary definition resulted in a chimera that was several-fold faster than its traditional boundary counterpart. A (adenylation domain), KR (Ketoreductase), ACP (acyl carrier protein), KS (ketosynthase), AT (acyltransferase), TE (thioesterase), KS^Q (ketosynthase-like decarboxylase), KR⁰ (inactive ketoreductase).

Recent advances aside, there have been varying levels of success in generating chemical diversity in polyketides. Small libraries of compounds with improved antifungal activities have been produced through non-natural incorporation of native extender units by removal of enzymes responsible for extender unit synthesis, inactivation of reductive domains, and knock-outs of tailoring enzymes [44]. Likewise, the fusion of 6-methylsalicylic acid synthase (6MSAS) with a PKS from Pseudallescheria boydii produced a novel compound [45]. Lastly, improvements of a chimeric PKS by point mutations in the KR and host engineering increased production of short-chain ketone fuel additives [46].

Additionally, there have been increased efforts to understand the structure and engineering of specific domains and linkers. This has included reductive loop engineering informed by cheminformatics, which led to higher yields of the expected products [47], phosphopantetheinyltransferase swaps [48], the addition of multiple ACPs on the C-terminus of the final ACP in erythromycin biosynthesis [49], and association of chimeric pathways through docking domain engineering [50–52]. Finally, the recently confirmed “extended conformation” of a module via X-Ray crystallography will better inform future engineering efforts [53].

Non-Natural Chain Lengths

Natural breaks in the colinearity paradigm, such as module skipping, stuttering, or stalling, can similarly enable the biosynthesis of unpredictable macrolactone cores. Although the infidelity of these pathways was thought to be highly unusual, a number of pathways capable of producing multiple or unusual biosynthetic products have been identified. These studies have focused on both single domain inactivation [54] and engineered functionality [55], to dramatically alter the product profile by producing polyketides of variable chain length. Additional mechanisms for non-colinear polyketide biosynthesis have been recently discovered, including the native reversal of selectivity of vatiamides, which allows the biosynthetic pathway to produce multiple products of variable chain lengths and subsequent post-processing modifications [56] as well as the pass-back chain extension mechanism of thalassospiramide biosynthesis [57]. Moreover, some of these processes have been accelerated through engineering, leading to seventeen rapalogs as a result of laboratory evolution that mimicked how PKSs might have evolved in Nature [58].

However, colinearity does not apply to iterative PKSs that catalyze different sets of reactions while maintaining exquisite control to produce ‘cryptic’ templated products. Notably, recent work from Yang et al. identified a possible mechanism in which the KR plays a significant role in chain length determination using intrinsic selectivity across multiple iterative PKSs [59]. The relevance of individual domain processing for the specificity of the chain length as determined by starter unit selection has also been probed [60].

Diversification of Polyketide Scaffolds via Non-Native Post-PKS Tailoring

Modifications to the core scaffolds introduce additional complexity and functionality. By leveraging native and engineered promiscuity of post-PKS tailoring enzymes, new-to-nature polyketides can be generated through the introduction of non-natural or non-native moieties.

Macrolactone glycosylation is critical to the biological functionality of polyketides. Owing to native glycosyltransferase (GT) promiscuity, several groups have leveraged GTs for the biosynthesis of non-native polyketides by leveraging combinatorial libraries derived from wide panels of macrolactone cores and NTP-sugars. Notably, the biological functionality of non-natively glycosylated polyketides can display different activity from their naturally occurring counterparts [61], as demonstrated by a series of erythromycin analogues, some of which have been characterized with potent activity against erythromycin A resistant strains [62]. Although these efforts continue to demonstrate the broad capabilities of these enzymes, the ability to diversify aglycone structures continues to pose a challenge.

Polyketide alkylation via biological and chemical syntheses, has similarly expanded biological functionality, as evidenced by the enhanced bioactivity of the daidzein analogue, 4’-O-methyl daidzein [63]. Often catalyzed by S-adenosylmethionine (SAM) dependent methyltransferases (MTs) [64], post-PKS polyketide alkylation has been diversified using rational reprogramming of promiscuous MTs for use in combinatorial biosynthesis reactions [65]. Interestingly, some natural product MTs have displayed promiscuity for non-native SAM analogues for the transfer of ethyl-, propargyl-, allyl-, and benzyl- moieties [66–68]. Notably though, while MTs have been highly successful for the diversification of non-ribosomal peptides, they have not yet been fully explored with polyketide substrates.

Halogenases also contribute to the diversification of natural products [69]. Flavin-dependent halogenases, such as VemK [70], ChmKN [71], and Rdc2 [72] have been demonstrated to halogenate aromatic polyketides, including resveratrol, to produce natural product analogues with potent bioactivities; however, their mechanisms are not conducive for use on non-aromatic substrates, limiting their potential to diversify other polyketides.

High-Throughput Approaches to Polyketide Synthetic Biology

Evaluating large numbers of artificial PKS pathways remains a critical engineering bottleneck due to the low-throughput of traditional analytical methods. Emerging high-throughput strategies, including engineered transcription-factor biosensor platforms and colorimetric assays [24], offer promising tools for engineering polyketide biosynthesis. Recently, a newly refactored FapR-based biosensor system for the detection of a variety of C2-substituted CoA- and N-acetylcysteamine (SNAc)-linked extender units [73] was described, both in vivo and in vitro. This biosensor detected the over-production of mmCoA in an engineered E. coli strain and also detected other extender units, paving the way for high-throughput engineering of native and non-natural extender unit supply in producing organisms.

There has also been much effort toward detection of polyketide pathway biosynthesis end-products and intermediates. Significant enhancements to mass spectroscopy-based screening methods, including laser-assisted rapid evaporative ionization mass spectrometry [74], have enabled the high-throughput direct detection of some natural products, including erythromycin A. Moreover, transcriptional regulators such as MphR are currently being leveraged for sensitive and specific screening of novel non-natural polyketides [75]. It is anticipated that such biosensors can be used to improve access to pathway intermediates and novel products through directed evolution by providing the ability to screen millions of pathway variants.

Future Outlook

The discovery of increasingly diverse polyketide enzymatic machinery offers powerful plug-and-play potential for the diversification of complex bioactive compounds. Yet, the ability to effectively engineer these pathways has been limited. However, efforts underpinned by enhanced mechanistic and structural studies of these megasynthases suggest an optimistic outlook for future PKS engineering to yield tailored polyketides. Probing the promiscuity of this machinery utilizing precursor-directed biosynthesis has enabled the mutasynthesis of a variety of critical targets. Moreover, analysis of product profiles derived from single domain and module modifications have elucidated unanticipated breaks in the co-linearity paradigm to enable cryptic and non-natural products. While engineering efforts were previously throttled by low-throughput screens, current and future engineering will be enhanced with the availability of tailored biosensors that enable high-throughput approaches to solve otherwise challenging problems related to polyketide biosynthetic engineering. An exciting future vision for designer polyketides includes augmenting traditional medicinal chemistry approaches with the application of computational methods to identify potential compounds of interest and value [76]. For example, “PKS Enumerator” [77] and the follow-up “SIME: Synthetic Insight-Based Macrolide Enumerator” [78] are cheminformatics tools that generates virtual libraries of macrolactones with multiple user-defined constraints derived from knowledge of polyketide biosynthesis. Such in silico libraries could then be mined using machine learning and/or 3D docking for potential new bioactive compounds. With this in mind, MacrolactoneDB integrates almost 14,000 existing macrolactones and their bioactivity information from various public databases [79]. Machine learning on this data set led to impressive prediction power for activity against several critical targets. Together, our vision for polyketide synthetic biology melds together high-throughput synthetic biology, advances in structural biology, mechanistic studies, and in silico prediction of bioactive compounds to develop new-to-nature polyketides as potential therapeutics and probes.