Refactoring the Pikromycin Synthase: A Modular Platform for Combinatorial Macrolide Antibiotic Biosynthesis in Escherichia coli — A Review

Figure 1. Traditional versus updated definitions of the polyketide synthase module boundary. (A) The traditional boundary, drawn upstream of the ketosynthase (KS) domain, groups KS with the downstream acyltransferase (AT) of the next module. (B) The updated boundary, drawn downstream of KS, groups KS with the upstream processing domains it evolutionarily co-migrates with. Original schematic prepared for this review; not a reproduction of any published figure.

Building on this updated module boundary, the Keatinge-Clay laboratory previously developed a single-plasmid, BioBricks-like cloning platform and used it to combinatorially assemble more than 150 triketide-to-pentaketide synthases from modules of the pikromycin PKS. That earlier work stopped short of assembling complete, antibiotic-producing synthases. The study reviewed here extends the approach to a two-plasmid system capable of housing the larger genetic payload required for a full, seven-module pikromycin synthase, and couples it to the downstream tailoring enzymes needed to convert the resulting macrolactones into bioactive, glycosylated macrolide antibiotics. This review is organized to follow the logical progression of the underlying study: Section 2 describes the two-plasmid cloning architecture itself; Section 3 covers the sequence of titre-optimization interventions; Section 4 addresses the combinatorial module-swapping experiments and the ketosynthase gatekeeping phenomenon that constrained them; Section 5 discusses chain-length modification; Section 6 covers downstream tailoring (glycosylation and hydroxylation) and the resulting antibacterial activity data; and Sections 7 through 9 offer a broader discussion, a comparative assessment against the historically dominant erythromycin heterologous-expression platform, and concluding remarks.

2. The Two-Plasmid Refactoring Platform

2.1 Design Rationale

Because a complete pikromycin synthase substantially exceeds the practical size limit for a single E. coli plasmid (roughly 30 kilobases), the authors split the assembly line across two compatible plasmids. One plasmid carries the first pikromycin module together with the N-terminal portion of the fourth module; the second carries the C-terminal portion of the fourth module together with the seventh, terminal module. The two halves of the fourth module are re-joined non-covalently in the assembled protein complex through a heterologous docking-domain pair borrowed from the erythromycin synthase (also known as 6-deoxyerythronolide B synthase, or DEBS). Splitting the assembly line in this way is not merely a size accommodation; it also means that new combinatorial synthases can be built by independently varying the module content of either plasmid, effectively doubling the combinatorial reach of the system relative to a single, monolithic construct.

2.2 Module Cassette Construction and Insertion Logic

Restriction sites flanking each module-encoding cassette allow additional modules to be ligated in sequentially, so that a working synthase such as module-1–module-A–module-B–module-4–module-C–module-D–module-7 can be built up piece by piece, with each junction encoding two additional serine residues in an already-permissive loop region unlikely to disturb folding. A HindIII-N12-SpeI insertion site is preserved after each successful ligation, meaning that the cloning process is iterative: a researcher can insert one module, verify the resulting construct, and then use the same site to insert the next, rather than needing to assemble the entire multi-module cassette in a single step.

Figure 2. Architecture of the two-plasmid refactored pikromycin synthase expression system. Plasmid 1 encodes module 1 and the N-terminal half of module 4; Plasmid 2 encodes the C-terminal half of module 4 and module 7. Additional modules are ligated sequentially into the preserved insertion sites on each plasmid, and the two half-synthases are reunited non-covalently through a heterologous docking-domain pair. Original schematic prepared for this review.

Reconstituting the seven-module pikromycin assembly line this way required six heterologous docking-domain pairs distributed across the synthase and produced a functional enzyme complex composed of six separate polypeptides, in contrast to the four polypeptides of the native S. venezuelae pathway. Expression in the engineered E. coli strain K207-3 — which supplies a phosphopantetheinyl transferase to activate carrier-protein domains and a propionyl-CoA carboxylase to generate the methylmalonyl-CoA extender unit — yielded narbonolide, the core heptaketide macrolactone of the pikromycin pathway, confirming that the refactored synthase functioned end to end. Intermediate constructs producing shorter-chain triketide and pentaketide products were also verified by mass spectrometry, establishing that each stage of the assembly — rather than only the final seven-module product — behaved as expected.

3. Optimizing Titre: Docking Domains, Editing Thioesterase, and Metabolic Engineering

Initial narbonolide titres from the refactored synthase were modest and accompanied by numerous shunt products — truncated or mis-assembled polyketides consistent with acyl carrier proteins escaping their intended downstream docking partner. Replacing the heterologous docking-domain pairs at three module junctions with the native pikromycin pairs improved narbonolide production by roughly 30%, suggesting some degree of co-evolution between docking domains and the modules that naturally carry them, even though the domains function adequately across pathway boundaries. A second, larger gain came from co-expressing PikAV, the pikromycin pathway's editing type II thioesterase, which is thought to remove acetyl or propionyl groups mistakenly loaded onto acyl carrier proteins. A single extra copy of this enzyme increased narbonolide titre by 1.6-fold, bringing the docking-domain-optimized system to 49 mg per litre of culture in shake-flask fermentation — substantially above the roughly 1 mg per litre reported for heterologous erythromycin A production in E. coli.

Table 1. Summary of narbonolide titre optimization steps. Values as reported in the primary source; mean ± standard deviation from biological replicates.

Figure 3. Sequential narbonolide titre optimization across three engineering interventions, plotted from the values in Table 1. Original chart prepared for this review.

The largest titre improvements, however, came from host-strain engineering rather than synthase engineering itself. Because the sugar precursor required for the pikromycin-associated deoxysugar desosamine is shared with E. coli's own lipopolysaccharide and cell-surface glycolipid biosynthetic pathways, competing consumption of this precursor limited desosamine attachment to the macrolactone. Using CRISPR–Cas9, the authors sequentially inactivated the competing rmlC, wecD/wecE and vioA/vioB loci in a strain already carrying a genomically integrated copy of the desosamine biosynthesis and transfer operon. The resulting engineered strain, in combination with a plasmid-supplied second copy of the same operon, achieved a 58-fold increase in narbomycin titre relative to the unmodified starting strain, reaching 37 mg per litre. This progression across engineered strains is summarized in Figure 5 in Section 6, alongside the tailoring-enzyme data it depends on.

4. Combinatorial Module Swapping and Ketosynthase Gatekeeping

With a working, optimizable synthase established, the authors turned to the combinatorial question that motivated the study: could modules be freely exchanged to generate new macrolide scaffolds? Systematic single-module swaps were performed at four positions within the pikromycin synthase, each position tested against every other extension module available from the same pathway — sixteen swapped synthases in total. Only one of these, in which the third module was replaced by the fifth, produced its anticipated macrolactone product, alongside an equal quantity of an alternative six-membered lactone side product resulting from incomplete macrocyclization. This low success rate illustrates what the authors term ketosynthase gatekeeping. Although the updated module boundary guarantees that a ketosynthase domain and the acyl chain it must accept are compatible at the positions immediately adjacent to the point of attachment, ketosynthases evidently remain selective toward more distal structural features of their substrate, including bulkier or differently oriented substituents further along the chain. This selectivity was substantial enough that even the fifth pikromycin module — independently identified as the pathway's most substrate-tolerant ketosynthase in earlier combinatorial studies — still showed strong preference for its natural upstream substrate when tested downstream of a full-length pentaketide intermediate in an octaketide-generating synthase, where the bulk of flux bypassed the newly inserted module altogether. When gatekeeping could not be overcome using modules native to the pikromycin pathway itself, the authors turned to naturally promiscuous ketosynthases from unrelated pathways. Modules from the erythromycin synthase performed unevenly — one substitution retained 81% of wild-type narbonolide titre, while a second essentially abolished product formation — while modules borrowed from the rapamycin synthase, previously identified as unusually substrate-tolerant, successfully generated two new macrolactone scaffolds, one of which was structurally confirmed by X-ray crystallography.

Table 2. Selected outcomes from single-module swapping experiments. "Approx. titre vs. parent heptaketide" reflects overall product titre (macrolactone plus any δ-lactone side product) relative to narbonolide production by the unmodified refactored synthase.

5. Extending Ring Size: Hexaketide and Octaketide Derivatives

Beyond substituting individual modules, the platform was used to alter the overall chain length of the assembled polyketide. A single point mutation inactivating the sixth module's ketosynthase reactive cysteine redirected the synthase toward selective production of the six-membered-ring precursor of methymycin, 10-deoxymethynolide, at a titre exceeding that of the unmodified, heptaketide-producing synthase. Complete deletion of the third module produced a different six-membered lactone product through an analogous chain-truncation strategy. An attempt to extend the pathway to an eight-ketide, sixteen-membered macrolactone — of the type found in tylosin and mycinamycin — by duplicating the fifth module was only partially successful: the engineered synthase produced primarily the original seven-membered narbonolide, alongside a smaller quantity of an unanticipated eight-ketide side-chain lactone, rather than the intended sixteen-membered ring, indicating that the native pikromycin thioesterase domain is not well suited to macrocyclizing rings of that size.

Table 3. Chain-length-modified synthase products and their relative titres.

6. Glycosylation, Hydroxylation, and Antibacterial Activity

Macrolactones alone are not antibiotics; biological activity in this class depends on attachment of the amino-sugar desosamine and, for some family members, subsequent hydroxylation by the cytochrome P450 enzyme PikC. Co-expression of the desosamine biosynthetic and transfer genes converted narbonolide quantitatively into narbomycin, and addition of PikC converted a fraction of the narbomycin pool into pikromycin, with conversion efficiency limited by comparatively weak PikC activity in the E. coli host — a longstanding challenge for heterologous expression of cytochrome P450 tailoring enzymes.

Figure 4. Branch points converting pikromycin synthase products into four distinct macrolide antibiotics. Narbonolide (heptaketide) and 10-deoxymethynolide (hexaketide) are each glycosylated by DesVII/DesVIII to give narbomycin and YC-17, respectively; subsequent hydroxylation by PikC yields pikromycin and methymycin. Original schematic prepared for this review.

The same glycosylation machinery was applied to several of the module-swapped macrolactones, generating a small set of new, desosaminylated narbomycin analogues; three were purified and structurally confirmed by nuclear magnetic resonance spectroscopy.

Figure 5. Narbomycin titre across sequential host-strain engineering steps. TM1 through TM7 refer to CRISPR–Cas9-engineered E. coli strains in which pathways competing with desosamine biosynthesis for a shared sugar precursor were progressively inactivated. Values are approximate, derived from the fold-changes reported relative to the final 37 mg/L titre. Original chart prepared for this review from reported data.

Antibacterial testing against Gram-positive organisms (Bacillus subtilis, Staphylococcus epidermidis, Micrococcus luteus) and Gram-negative organisms (E. coli, Pseudomonas putida) showed that narbomycin and one new analogue retained potency comparable to a commercial pikromycin standard against Gram-positive strains, while a second new analogue showed only weak activity and a third was inactive. Only the parent compound, pikromycin, showed any activity against the Gram-negative panel.

Table 4. Antibacterial activity (minimum inhibitory concentrations) of purified compounds against Gram-positive (B. subtilis, S. epidermidis, M. luteus) and Gram-negative (E. coli, P. putida) test panels.

The authors relate the loss of activity in some analogues to the absence or altered geometry of the enone moiety that normally engages the bacterial ribosome, drawing on prior structural work describing the interaction between pikromycin and the 70S ribosome.

DISCUSSION

7.1 Significance of the Platform

Taken as a whole, this study represents a meaningful step toward realizing modular PKSs as genuinely combinatorial platforms for macrolide antibiotic discovery, and it does so in a genetically tractable, fast-growing host rather than in the native, slower-growing actinomycete producers. Several features distinguish the pikromycin platform favorably from the more extensively studied erythromycin system: fewer tailoring enzymes are required, at least two bioactive family members can be reached without relying on cytochrome P450 chemistry, and achievable titres are an order of magnitude higher than historical benchmarks for heterologous erythromycin A production.

7.2 Ketosynthase Gatekeeping as the Central Bottleneck

At the same time, the module-swapping results make clear that ketosynthase gatekeeping, not vector capacity or expression, is now the principal obstacle to unrestricted combinatorial synthase design. Only one of sixteen attempted intra-pathway module substitutions succeeded, and even the most substrate-tolerant native ketosynthase showed clear discrimination against non-native substrates several carbons removed from the site of attachment. The authors' strategy of borrowing promiscuous ketosynthases from unrelated pathways, exemplified by the rapamycin-derived substitutions, is a pragmatic workaround, but a more systematic solution — for instance, rational or evolved relaxation of ketosynthase substrate specificity — would likely be needed before combinatorial libraries could be generated at a scale comparable to the modular chemical synthesis platforms referenced at the start of this review.

7.3 Titre and Scale-Up Considerations

Titres, while a substantial improvement over earlier heterologous macrolide systems, remain far below what would be needed for preparative-scale medicinal chemistry follow-up without further optimization, such as fed-batch bioreactor cultivation (previously shown to increase related polyketide titres several-fold over shake-flask conditions) or additional host-strain metabolic engineering. The mechanistic basis for several observations — including incomplete macrocyclization in favor of competing lactone side products, and the unexplained interaction between genomic and plasmid-based desosamine pathway expression — also remains to be resolved, and the authors note that their platform itself provides a convenient experimental system for investigating these open questions in PKS enzymology.

7.4 Implications for Antibiotic Discovery Pipelines

For antibiotic discovery specifically, the demonstration that structurally modified narbomycin analogues can be produced, purified and directly screened for antibacterial activity within a single heterologous host establishes a practical, if still limited, discovery pipeline. Expanding the accessible chemical space will likely depend on the combined application of ketosynthase engineering, expanded docking-domain libraries, and improved cytochrome P450 tailoring activity in E. coli, each of which represents a clear and tractable direction for follow-up work.

8. Comparative Assessment Relative to the Erythromycin Platform

Because the erythromycin synthase remains the most extensively engineered heterologous PKS system in E. coli, it provides the most informative point of comparison for judging how much practical ground the pikromycin platform covers. Table 5 summarizes the principal differences discussed throughout this review.

Table 5. Comparative summary of pikromycin versus erythromycin heterologous E. coli platforms, based on figures and statements reported in the primary source and cited historical literature

Reading the pikromycin gene cluster's individual components alongside these comparisons helps clarify why the platform behaves as it does. Table 6 lists the principal genes referenced throughout this review and the function each contributes to the overall pathway, from chain assembly through tailoring.

Table 6. Principal genes referenced in this review and their functional roles in macrolactone assembly, tailoring, and host-strain engineering.