Refactoring the pikromycin synthase for the modular biosynthesis of macrolide antibiotics in E. coli

Abstract

While engineering modular polyketide synthases (PKSs) using the recently updated module boundary has yielded libraries of triketide-pentaketides, this strategy has not yet been applied to the combinatorial biosynthesis of macrolactones or macrolide antibiotics. We developed a 2-plasmid system for the construction and expression of PKSs and employed it to obtain a refactored pikromycin synthase in E. coli that produces 85 mg of narbonolide per liter of culture. The replacement, insertion, deletion, and mutagenesis of modules enabled access to hexaketide, heptaketide, and octaketide derivatives. Supplying enzymes for desosamine biosynthesis and transfer enabled production of narbomycin, pikromycin, YC-17, methymycin, and 6 derivatives thereof. Knocking out pathways competing with desosamine biosynthesis and supplying the editing thioesterase PikAV boosted the titer of narbomycin 55-fold to 37 mgL^-1. The replacement of the 3rd pikromycin module with its 5th yielded a new macrolide antibiotic and demonstrates how libraries of macrolide antibiotics can be readily accessed.

Figure 1. Pikromycin PKS and 2-plasmid platform for constructing engineered PKSs.

a, Pikromycin/methymycin biosynthetic pathway (PKS colored by updated modules). b, Construction of the BioBrick-like units encoding pikromycin modules. T7 promoters and terminators are shown. c, The 2-plasmid system enables the BioBrick-like assembly of PKSs such as P1-A-B-P4-C-D-P7. Docking domains (DDs) are specific for each position in the constructed synthase. The ligation of HindIII-XbaI fragments into the HindIII-N₁₂-SpeI insertion sites yields XbaI/SpeI scars (*) encoding 2 serines at module boundaries.

Figure 2. Evaluation of the 2-plasmid system and narbonolide biosynthesis by refactored pikromycin PKS.

Expected products are observed from P1-P4-P7, P1-P2-P3-P4-P7, P1-P4-P5-P6-P7, and P1-P2-P3-P4-P5-P6-P7.

Figure 3. Module swapping of the refactored pikromycin PKS.

While 16 swaps with pikromycin modules were attempted, only the replacement of P3 with P5yielded an active PKS. This synthase generates anticipated macrolactone 10as well as d-lactone 11. Swapping P2 and P6 for E2and E6 (from the erythromycin PKS), respectively, yielded hybrid synthases that generate narbonolide (1). Swapping P2 for S2(from the spinosyn PKS) yielded a hybrid synthase that generates macrolactone 12. Titers are reported in comparison to that of narbonolide by the refactored pikromycin PKS.

Figure 4. Construction of hexaketide and octaketide synthases.

Hexaketide synthases P1-P2-P3-P4-P5-P6*-P7, P1-P2-P5-P4-P5-P6*-P7, P1-S2-P3-P4-P5-P6*-P7, and P1-P2-P4-P5-P6-P7, respectively produce 4 (10-dml), 13, 14, and 15. The asterisk (*) next to the P6 KS indicates the C->A mutation. The octaketide synthase P1-P2-P3-P4-P5-P5-P6-P7 does not produce the anticipated octaketide macrolactone but does produce octaketide d-lactone 16 and twice as much heptaketide macrolactone 1 (narbonolide). Docking domain (DD) motifs from A1 of the amphotericin synthase were used for the upstream P5.

Figure 5. Production of macrolide antibiotics in E. coli.

a, Plasmids pDes, pDesPikC, and pDesPikAV encode desosamine biosynthesis/transfer and macrolide resistance enzymes. Plasmids pDesPikC and pDesPikAV additionally encode the P450 monooxygenase PikC and the editing thioesterase PikAV, respectively. PikAV is controlled by a second T7 promoter in pDesPikAV. b, With the exception of 13, each of the macrolactones in this study is glycosylated by DesVII/DesVIII and hydroxylated by PikC to yield known macrolide antibiotics or derivatives thereof.

Figure 6. CRISPR/Cas9 engineering helps boost titers of macrolide antibiotics.

a, Biosynthetic pathways in E. coli compete with desosamine biosynthesis for TDP-4-keto-6-deoxy-d-glucose. b, The TM1 strain was engineered by inserting the desosamine biosynthesis/transfer genes into the E. coli K207–3 genome. The TM2-TM4 and TM5-TM7 strains were engineered through sequentially inactivating rmlC, wecD/E, vioA/B in E. coli K207–3 and TM1, respectively (Supplementary Table 1). c, Narbomycin production is reported for engineered E. coli strains transformed with pP1-P2-P3-^NP4 & p^CP4-P5-P6-P7, encoding the refactored pikromycin synthase, as well as an additional plasmid or BAC that encodes enzymes from the pikromycin biosynthetic pathway (Fig. 5a). TM7 cells with pP1-P2-P3-^NP4, p^CP4-P5-P6-P7, and pDesPikAIV yield a narbomycin (2) titer of 37.1 mgL⁻¹.