Application of Synthetic Biology to the Biosynthesis of Polyketides

3.1. Enzyme Engineering for New Structures of Polyketides The differences in catalytic mechanisms and substrate specificities of PKSs result in the diversity in polyketide chain synthesis. A series of post-modification reactions further increase the structural diversity of polyketides, making it possible to engineer both PKSs and post-modification enzymes to produce new non-natural polyketides. 3.1.1. PKSs The number and function of PKS modules and domains determine substrate selection, the extent of reduction and the stereochemistry of the products in the polyketide synthesis pathway. The number of modules defines the size of the backbone, and the domain composition controls the degree of functionalization [62]. Therefore, PKS modification can often enhance the structural diversity and biological activity of polyketides. This is mainly achieved through site-directed mutagenesis, as well as the insertion and replacement of domains and modules (Figure 5, Table 3). Type I PKSs are the most typical and complex of the three PKSs, therefore, this section focuses on Type I PKSs. Site-directed mutation has minimal impact on the overall protein structure, however, it requires specific conditions regarding the size of the side chain and its flexibility [63,64]. In the biosynthesis pathway of FK506, the AT4_FkbB domain can use both allylmalonyl and emthylmalonyl to produce the target product FK506 and its analog FK520. Molecular dynamics simulations revealed that the V187K mutation enhances substrate specificity and reduces the production of the by-product FK520 [65]. This strategy is often combined with domain and module engineering. The starter and extender units largely determine the structural diversity and chemical complexity of polyketides. For example, the AT domain is responsible for CoA-based starter and extender units, and shows greater substrate promiscuity than other domains. Sheehan et al. replaced the loading module of the spinosad PKS with the loading module from the erythromycin PKS and added a range of carboxylic acids exogenously, resulting in 16 new spinosad derivatives. Among them, 21-cyclobutyl-spinosyn A along with its semisynthetic 5,6-dihydro derivative exhibited improved insecticidal activity [66]. Sirirungruang et al. introduced an F190V mutation into the wild-type trans-AT from the disorazole biosynthetic pathway (DszAT) and loaded it onto a specific PKS module in which the cis-AT was inactivated, and selectively introduced fluorine into the polyketide backbone and producing two fluorinated 6dEB analogs, 2-fluoro-2-desmethyl 6dEB and 4-fluoro-4-desmethyl 6dEB [67]. Domain engineering strategies are also applied to KR (responsible for the formation of stereocenters in polyketides) [68], DH [69], and TE in the unloading modules [70], although research in these areas remains relatively limited. Another PKS engineering strategy for producing new compounds involves modifying domains or modules (including deletion, replacement, and addition), which directly influences product synthesis. Using RedEx technology, the integration of five domains (KS 1b, AT 1b, DH 1b, KR 1b, and ACP 1b) from module 1b of butenyl-spinosyn into the spinosad synthesis pathway between the KS and AT domains of module 1 successfully achieved butenyl group modification [71,72]. However, inserting heterologous domains often disrupts PKS stability, resulting in loss of activity and protein misfolding [73]. Appropriate linker regions and splice sites can retain protein interactions between donor and acceptor units during domain or module exchanges, reduce the impact of conformational changes during the polyketide chain elongation, and maintain overall PKS structure. Thus, splice sites are typically chosen within the AT and KS sequences or in linker regions upstream of ACP. However, the success rate of module exchanges is often low [74,75,76]. Yuzawa et al. systematically analyzed segments of AT domains and associated linkers in AT domain exchanges, identifying the module boundaries that maintain protein stability and enzyme activity when exchanging AT domains [77]. They subsequently developed a fluorescence-based solubility biosensor to rapidly detect AT-exchanged PKS hybrids with randomly assigned domain boundaries, thereby minimizing structural disruption [78]. Besides directly engineering domains, another approach involves starting with enzymes that exhibit high catalytic activity and broad substrate specificity. Zhang et al. discovered a unique β-subunit (Arm13) in acyl-CoA carboxylase (ACC) that combined with the α and ε subunits of propionyl-CoA carboxylase, forming a highly catalytic ACC. This ACC can recognize acyl-CoAs of different chain lengths and with various functional groups, efficiently converting them into corresponding alkylmalonyl-CoAs. By exogenously adding carboxylate precursors, a series of reactive functional groups such as alkenyl, alkynyl, and phenyl were introduced at the C6 position of the armeniaspirol backbone [79]. Zheng et al. found that the naturally lacking Lys-A10 residue in the acyl-CoA synthetases (ACSs) UkaQ in the UK-2A biosynthetic pathway enabled the synthesis of diverse acyl-CoAs. These, combined with permissive CCR and ACC, produced various malonyl-CoA extender units, resulting in the formation of new antimycin analogs [80]. This demonstrated the critical role of the catalytic activity of ACCase and ACSs in synthesizing polyketide extender units and new compounds. 3.1.2. Post-PKS Enzymes After the synthesis of the polyketide chain, a series of post-synthetic modification reactions (such as methylation, glycosylation and acylation) are required to form the final polyketides. Modifying the post-PKS enzymes or related genes can alter the biological activity and structural diversity of end products. The types of sugar moieties attached to the polyketide backbone can affect the biological activity, stability, and solubility of the compounds. Rhamnose not only is involved in the synthesis of spinosyn but also serves as a component in the synthesis of cell wall lipopolysaccharides. When rhamnose biosynthesis genes from S. spinosa are absent, different Streptomyces species cannot heterologously produce spinosyn [95]. Amphotericin B without mycosamine cannot trigger fungal membrane permeability, resulting in the loss of antifungal activity and toxicity [96]. Engineering natural product GTs can modify donor sugar, acceptor substrate, and acceptor position specificities to obtain natural products with unnatural glycosylation patterns [97,98,99]. For example, by replacing key amino termini of the C-glycosyltransferase encoded by the urdGT gene broadens and alters substrate specificity during the synthesis of urdamycin, resulting in the addition of glycosyl groups to unnatural substrates and the production of 28 new active urdamycins [100,101]. By deleting the genes responsible for the biosynthesis and attachment of TDP-D-desosamine in S. venezuelae ATCC 15439 and overexpressing genes encoding deoxysugar biosynthesis and glycosylation, various macrolide glycosyl derivatives were obtained [102]. Song et al. knocked out the spnK gene responsible for rhamnose methylation in the spinosyn biosynthetic gene cluster, thereby achieving heterologous biosynthesis of spinosyn JL [71]. Hydroxylation or other modifications by CYP450 also have the potential to lead to structural diversity [103]. The substrate-flexible CYP450 (pikC) from the S. venezuelae HK954 mutant produced new hydroxylated analogs of oleandomycin [104]. Modifying key enzymes is an effective way to generate new compounds. With the explosive development of big data and AI, the development of polyketides is not limited to substances with known synthetic pathways. In 2022, to further understand the abundance and diversity of aromatic polyketides from bacteria, Chen et al. constructed a phylogenetic tree and discovered a correlation between CLF enzyme amino acid sequences and compound structures. Using CLF as a marker, they identified 3254 bacterial Type II PKS gene clusters from databases, creating the first global atlas of bacterial aromatic polyketides. This allowed for the prediction of compound types synthesized from CLF gene information. Consequently, researchers identified a novel aromatic polyketide, oryzanaphthopyran, from Streptacidiphilus, therefore, making a breakthrough in the synthesis of new aromatic polyketides from bacterial Type II PKSs [105]. Due to the colinearity of the PKSs, tools like Alphafold and RoseTTAFold can combine natural PKS sequences to reversibly design specific PKSs biosynthetic pathways, generating millions of organic molecules and enabling the retrosynthesis of polyketides [106]. 3.2. Synthesis Pathway Optimisation Enhances Adaptation for Polyketides Production Eriko Takano proposed several design principles for synthetic biology of secondary metabolites. First, the biosynthetic gene clusters of secondary metabolites are identified through genome analysis. Then, the genes in the biosynthetic pathway are modularly designed and assembled based on the existing component library. Next, these genes are expressed in suitable screening hosts, with compatibility analyses conducted at the transcriptional level (e.g., promoters) and translational level (e.g., RBS). Finally, the best combinations are transferred into production hosts for production [107] (Figure 6). Based on these principles, deep mining of efficient functional parts, reconstructing metabolic pathways, and optimizing the adaptation of components and chassis to finely regulate metabolic flux to achieve high yields of target compounds. 3.2.1. Optimization of Biological Parts Using synthetic biology to assemble biological parts in an orderly manner allows for easier and faster acquisition of compounds with desired properties. However, many parts exhibit incompatibility in different host systems, significantly reducing the biosynthetic efficiency of polyketides. Utilizing well-characterized genetic control elements can regulate gene expression at the transcriptional and translational levels. Promoter engineering is a primary strategy to enhance the yield of natural products by regulating the expression of key genes. Using the CRISPR/cas9 system to replace the native promoters of limiting genes in the erythromycin biosynthetic gene cluster with the heterologous weak promoter PermE^∗_s23 increased erythromycin yield by 6.0-fold, while the insertion of the strong promoter PkasO did not affect erythromycin yield, indicating that the use of compatible promoters to regulate gene expression can maximize the yield of target products [108]. Wang et al. obtained a strong endogenous promoter 5063p based on transcriptome analysis, which increased the expression of geldanamycin PKS by 4-141 times and geldanamycin yield by 39% when replacing the native PKS promoter gdmA1p [109]. Wei et al. mined the promoters of transcription factors related to secondary metabolism in A. nidulans, and, combined with single-cell fluorescence detection using flow cytometry, they obtained a natural promoter library with relative expression levels up to 37 times higher. Among them, the expression levels of PzipA and PsltA were 2.9 and 1.5 times higher than the constitutive promoter PgpdA, respectively [110]. Combining random mutagenesis of promoters and high-throughput screening is an efficient method to obtain promoters with a wide range of strengths. Tu et al. focused on droplet microfluidic-based high-throughput screening technology in Streptomyces, rapidly constructing promoter mutant libraries for subsequent fine regulation of metabolic pathways [111,112]. Artificial neural networks and computer-aided design can rationally design promoter sequences with specific requirements, avoiding interference from host regulatory mechanisms. The AI-assisted promoter sequence optimization method DeepSEED developed by Zhang et al, successfully designed promoters from various sources and types [113]. Terminators have less impact than promoters and are less studied and applied in Streptomyces. Horbal et al. mined a series of terminators from databases, including the Mycobacteria-derived terminator tt_sbiB, which has a read-through of less than 4% [114]. Regulation alone at the transcriptional level is insufficient to construct highly efficient gene expression systems and optimization of translation efficiency is also required [115]. Translational regulation is primarily determined by the ribosome binding site (RBS) and the 5′ untranslated region (5′-UTR). In Streptomyces, an inappropriate RBS can reduce gene expression efficiency to zero. Researchers developed an in vivo RBS-selector using gusA as a reporter system to rapidly select optimal RBS for target genes, thereby rationally controlling protein expression levels [114]. The selection of RBS and the 5′ UTR is often combined with promoters. For example, by randomizing the strongest RBS sequence in S. venezuelae, a series of synthetic RBS with varying strengths were obtained and combined with promoters, identifying the optimal promoter-RBS combination for gene expression. This approach replaced the indigenous promoter and RBS sequences, achieving activation of silent gene clusters at different levels and mass production of target compounds [116]. Yi et al. screened two promoters and four 5′-UTR sequences from multiomics data of S. coelicolor, whose combination of which resulted in protein expression levels ranging from 0.03 to 2.4 times that of the strong promoter ermE^∗p combined with the Shine-Dalgarno sequence [117]. The 5′-UTR limits the direct application of σ⁷⁰-dependent promoters in Streptomyces. Replacing E. coli-derived 5′-UTR with Streptomyces-derived 5′-UTR converts σ⁷⁰-dependent promoters into σ^hrdB-dependent promoters that efficiently are expressed in Streptomyces. RBS libraries were also constructed to optimize hybrid-5′-UTR. This strategy expands the sources of Streptomyces promoters and facilitates the construction of Streptomyces cell factories [118]. By combining the advantages of transcriptional and translational parts to design and control key enzymes in biosynthetic pathways can greatly promote the production of secondary metabolites. By integrating the development of biological parts with omics technologies that can mine important biological features and explore complex relationships between features, and by using machine learning and deep learning algorithms for model prediction, the accuracy and controllability of these elements have been significantly improved [119] (Figure 7). 3.2.2. Reconstruction of Synthetic Pathway Reconstructing biosynthetic pathways to improve the synthesis of secondary metabolites is one of the core challenges for synthetic biology. Promoter replacement of rate-limiting genes in biosynthetic pathways with compatible ones is an effective strategy for the efficient production of target products. Tylosin, a 16-membered macrolide antibiotic used in veterinary medicine, consists of four structurally similar components: tylosin A, tylosin B, tylosin C, and tylosin D. By overexpressing the rate-limiting enzyme genes tylF and tylI with the strong promoter stnYp from Streptomyces flocculus CGMCC4.1223, the yield of tylosin A was elevated to 10.30 g/L, which was 1.7-fold higher than the industrial strain Streptomyces fradiae, the purity and yield of tylosin A were both improved [120]. Using a genome-scale metabolic network model (GSMM), overexpression of the four predicted key targets with kasO^∗p increased pikromycin production in S. venezuelae [121]. The same strategy has also been applied to spinosyn [122] and actinorhodin [123]. BGCs display highly modular organization, and biosynthetic pathways typically involve the co-expression of multiple genes. According to the functions of the synthesis genes and the characteristics of the intermediates, they can be rearranged into different modules for fine regulation and coordinated expression, enhancing compatibility from module to module and from module to host. Song et al. took advantage of ExoCET multi-fragment assembly technology, divided the 79 kb biosynthetic gene cluster of spinosyn into seven operons and strengthened them with strong constitutive promoters, achieving efficient expression of spinosyn in S. albus J1074, with a 328-fold increase in yield compared to the native gene cluster [124]. Jiang et al. constructed a 106-kb multioperon artificial gene cluster, including five operons involved in natural salinomycin synthesis and five fatty acid β-oxidation genes into a single operon, driven by strong constitutive promoters, achieving efficient heterologous expression [125]. Shao et al. divided the BGC of type I polyketide spectinabilin into promoter modules, gene modules, and helper modules based on the modular design principle in synthetic biology. For promoter modules, strong promoters recognizable by the expression host were used to initiate transcription of exogenous genes, ensuring the expression of exogenous gene clusters; the gene modules contained eight genes required for spectinabilin biosynthesis; and the helper modules introduced elements that maintain the replication and selection of exogenous DNA within the host. Finally, the re-engineered and synthetically obtained gene cluster was cloned downstream of the promoter in the promoter module and introduced into S. lividans, resulting in a spectinabilin yield of 100 μg/L [126]. 3.2.3. Adaptation of Chassis Hosts Many hosts that synthesize polyketides face challenges such as slow growth, unsuitability for laboratory cultivation, or difficulty in genetic manipulation; these factors lead to extremely low yields or failure to synthesize many valuable PKs. Understanding biosynthetic mechanisms is critical to the rational selection and utilization of heterologous hosts are key to expressing BGCs for secondary metabolites. Many efficient and practical tools used for cloning and assembling large size-BGCs have been developed to capture the PKs synthesis pathway for heterologous expression in different hosts, e.g., TAR-cloning [127], AssemblX [128], MoClo [129], etc. Common heterologous expression strains include Streptomyces coelicolor, Streptomyces lividans, Streptomyces albus, Streptomyces avermitilis [130,131] and Aspergillus oryzae [132]. E. coli and S. cerevisiae are the most commonly used cell factories in synthetic biology. Although E. coli itself has difficulty in efficiently translating and folding key synthetic enzymes (such as PKS and CYP450) and cannot synthesize some acyl-CoA compounds, rational design and modification of the synthetic pathway can still achieve high-efficiency expression [133,134]. Curcumin is synthesized by a six-enzyme reaction, which is typically heterologously expressed in E. coli. Kang et al. used the multiplex automatic genome engineering (MAGE) tool in E. coli to mutate the 5′-UTR of each of the six genes, optimizing the expression balance of each enzyme. This led to a variant (6M08rv) that improved the curcumin yield by 38.2-fold [135]. P450 EryF is a key enzyme in the synthesis of erythronolide B (EB). After identifying the optimal SaEryF for EB biosynthesis and designing the I379V mutant based on its crystal structure, the EB yield in E. coli reached 131 mg/L [136], significantly reducing the synthesis cycle. S. cerevisiae has been used to produce various simple polyketides, such as resveratrol and naringenin, and is also commonly used for the expression of fungal-derived BGCs [137,138]. Zhao et al. used promoter replacement and enzyme fusion strategies in S. cerevisiae to achieve the synthesis of bikaverin, with a yield of 202.75 mg/L [139]. These results demonstrate the enormous potential of E. coli and S. cerevisiae in producing complex polyketides.   3.3. Efficient Strategies for Production of Polyketides Polyketide synthesis often involves balancing primary and secondary metabolism, and the biosynthetic pathways and regulatory mechanisms are complex. Simple metabolic pathway modifications may not significantly increase yields and may disrupt the metabolic network balance, leading to reduced target product yield or the generation of by-products. Therefore, to maximize the yield of target natural products, a global approach is adopted, enhancing precursor supply, relieving feedback inhibition, and employing dynamic balancing strategies. 3.3.1. Enhancing Precursor Supply Primary metabolism provides precursors for secondary metabolism. By optimizing pathways related to precursor synthesis and removing competitive pathways, metabolic flux can be directed towards polyketide synthesis (Figure 8a). Liu et al. used transcriptome analysis to knock out the PEP phosphonomutase gene, leading to pyruvate accumulation and consequently increasing the precursor supply for spinosyn biosynthesis, increasing spinosyn A production by 2 times [140]. Genome-scale metabolic models (GEM) can predict metabolic flux distributions related to secondary metabolite precursors under specific conditions [141]. Under the guidance of GEM , knocking out the pfk gene encoding 6-phosphofructokinase resulted in upregulated pentose phosphate pathway flux, increasing NADPH levels and enhancing rapamycin production [142]. Oviedomycin, an anticancer polyketide synthesized by Type II PKS, Gu et al. used GEM to predict and overexpress the precursor and cofactor genes associated with enhanced oviedomycin synthesis. They also reconstructed the gene cluster by replacing the natural promoters of the oviedomycin BGC, the oviedomycin yield increased to 670 mg/L [143]. Fu et al. conducted a metabolomic analysis on the high-yield spinosyn strain ADE-AP and found that the high expression of PKS genes and acyl-CoA synthesis genes led to significant precursor accumulation, highlighting the critical role of precursor supply in yield improvement [144]. Malonyl-CoA, a crucial precursor for polyketide synthesis, is formed by the oxidative decarboxylation of pyruvate catalyzed by pyruvate dehydrogenase (PDH) to produce acetyl-CoA, followed by acetyl-CoA carboxylation. The natural PDH-ACC pathway has issues related to low catalytic efficiency and poor carbon utilization. Li et al. developed a non-carboxylative malonyl-CoA (NCM) pathway with a C3 (pyruvate)-C3 (3-oxopropanoate)-C3 (malonyl-CoA) catalytic mode, avoiding carbon loss, energy consumption and interference with the natural metabolic network, achieving high-efficiency synthesis of malonyl-CoA and its derivatives in different hosts, ultimately increasing spinosyn yield to 4.6 g/L [145]. Advancements in omics technologies have facilitated the discovery and analysis of unknown BGCs in genomes, a streptomyces genome typically encodes 25–50 secondary metabolite BGCs. The knockout of non-essential gene clusters can simplify the metabolic background of the strain and prevent precursor diversion to competitive secondary metabolite pathways. The S. pogona genome contains 32 gene clusters, including six polyketide gene clusters. The knockout of clu13 (a flaviolin-like gene cluster) increased butenyl-spinosyn yield by 4.06-fold [146]. Using the Latour gene editing system to knock out a ~20 kb NRPS-T1 PKS gene cluster increased butenyl-spinosyn yield by 4.72-fold [147]. Efficient heterologous expression of naringenin in Streptomyces albidoflavus J1074 can also be achieved through the deletion of competitive gene clusters [148].   3.3.2. Relieving Feedback Inhibition The excessive intracellular accumulation of polyketides can impose metabolic burdens on the host or cause feedback inhibition of their synthesis pathways, thereby limiting yield increases. Some genes within the BGC can encode efflux proteins that can expel secondary metabolites from the cell, alleviating cytotoxicity and feedback inhibition, thereby increasing target product yield (Figure 8b). For example, the oxytetracycline (OTC) cluster encodes the membrane-bound protein OtrB, responsible for OTC efflux. Together with OtrA, a ribosomal protection protein, and OtrC, an ABC exporter, expelling the accumulation of excessive oxytetracycline, protecting the host strain from OTC toxicity, thereby enhancing the resistance level of Streptomyces rimosus to oxytetracycline and increasing oxytetracycline yield [149]. Some products also serve as mechanisms of relieving feedback inhibition. When intracellular rifamycin B levels are low, the regulatory factor RifQ negatively regulates the efflux protein RifP. As rifamycin B accumulates to a certain threshold, RifQ dissociates from rifP, avoiding excessive rifamycin B toxicity to the host. Overexpression of rifP and deletion of RifQ increased the yield of rifamycin more than 2-fold [150]. ATP-binding cassette transporters (ABC transporters) are crucial components of the Streptomyces transport system [151]. Daunorubicin (DNR) and its hydroxylated derivative doxorubicin (DXR), Type II polyketides produced by Streptomyces peucetius, exert anticancer effects by inserting into DNA and inhibiting type II topoisomerase. When DNR and DXR accumulate intracellularly, they inhibit the expression of transcriptional activators while inhibiting type II topoisomerase activity, leading to feedback inhibition of their synthesis. The DrrAB proteins encoded by S. peucetius drrAB, belonging to the ABC transporter family, can expel DNR, DXR ,and their structural analogs extracellularly, thus improving strain tolerance and increasing DNR and DXR yields [152,153,154]. The AvtAB proteins encoded by the avermectin BGC perform similar functions, resulting in a 50% increase in total avermectin B1a yield [155]. More and more transport proteins are discovered to promote the effective efflux and further yield enhancement of polyketides. Chu et al. developed a rational transporter screening process combined with a tunable plug-and-play exporter (TuPPE) module, discovering three new ABC transporters that universally increase polyketide yields [156]. These studies demonstrate that enhancing the efflux of end products in synthetic pathways is an effective strategy for relieving product feedback inhibition, increasing polyketide synthesis yields, and enhancing host strain tolerance. 3.3.3. Dynamic Regulation The synthesis of secondary metabolites occurs in two stages. During the primary metabolism phase, cells consume external nutrients for rapid growth and reproduction; when external nutrients become limited, cell growth ceases, and the cells enter the secondary metabolism phase, where they begin producing secondary metabolites. How to increase SMs production without affecting normal growth is an important question. Dynamic regulation plays a crucial role in finely controlling biosynthesis (Figure 8c). The key lies in discovering and applying regulatory elements that respond to metabolic products. Using signals such as light, metabolites, or chemical molecules to design biosensors, cells can maintain a good production state, thereby improving the production performance of chassis cells [157]. Triacylglycerols (TAGs) are key metabolites that link primary and secondary metabolism, and regulate the metabolic switch [12,158]. The degradation of TAGs not only provides precursors and reduces power for the synthesis of polyketide but also redirects more carbon flux toward polyketide synthesis through changes in reducing power levels. Building on this insight, researchers developed the dynamic degradation of TAG (ddTAG) engineering strategy, which selectively controls the timing and strength of TAG degradation, significantly increasing the yield of various polyketides. The yield of avermectin B1 reached 9.31 g/L [12]. Besides using key metabolites for dynamic regulation, quorum sensing (QS) systems independent of metabolic pathways can achieve a dynamic balance. For example, Tian et al. combined QS with CRISPRi driven by γ-butyrolactones (GBLs)-responsive promoter to construct a novel dynamic regulation system called EQCi. This system can simultaneously regulate multiple targets. When integrated into S. rapamycinicus, it increased rapamycin production by approximately 660% without affecting cell growth, effectively balancing the metabolic flux between primary metabolism (cell growth) and secondary metabolism (product production) [159]. Currently, reconstructing biosynthetic pathways for secondary metabolites and heterologous expression are common metabolic engineering approaches. Changes in metabolic flux during this process may adversely affect cell growth and production. Combining GSMM and flux balance analysis (FBA) with secondary metabolic pathways can accurately simulate and predict the production flux of secondary metabolites [160]. From these results, biosensors can be designed, making the dynamic regulation system more efficient.