Introduction
In vitro multienzyme synthesis is a powerful approach for constructing complex biomolecules and offers an alternative to conventional synthesis and metabolic engineering1. To explore the feasibility of an in vitro synthesis of a polyketide natural product derived from a variety of enzyme activities in a single reactor, we set out for the full in vitro reconstitution of enterocin (Compound 1). We previously demonstrated that enterocin is derived from one molecule of phenylalanine-derived benzoic acid (Compound 2) and seven molecules of malonyl coenzyme A (malonyl-CoA, Compound 3) through the oxidative derailment of an aromatic polyketide synthase (PKS) pathway (Scheme 1)2. The genes encoding enterocin biosynthesis, regulation and resistance proteins are clustered in a 20-open-reading-frame locus (enc) in S. maritimus3 and have been heterologously expressed to yield enc-based polyketides in vivo4.
Scheme 1: Structures and proposed biosynthesis of enc-based polyketides.
(a,b) Biosynthesis primed with acetate via the decarboxylation of malonate (a) and benzoate (b).
Full size image (58 KB)We purified the recombinant enc enzymatic components (Supplementary Fig. 1 online) and assembled them sequentially in the buildup of the fully functional pathway (Supplementary Table 1 and Supplementary Methods online). We first reconstituted the minimally functional enc PKS, which consists of the ketosynthase chain-lengthfactor heterodimer EncA-EncB5, acyl carrier protein (ACP) holo-EncC5 and SgFabD (malonyl-CoA:ACP transacylase from Streptomyces glaucescens)6, with malonyl-CoA in the presence or absence of benzoic acid to yield a series of all malonate–derived polyketides of unknown composition (Scheme 1a and Fig. 1a,b). We explored their chain length by reincubating the enzyme mixture with [2-13C]malonyl-CoA7 (Compound 4) and analyzed the products by HPLC-MS. This characterization revealed a 9-AMU mass shift from m/z 342 (compounds a and b) and 298 (compound c) for the unlabeled samples to 351 and 307 m/z, respectively, which suggests that a, b and c are nonaketides. We confirmed this identification and determined their molecular formulas as C18H14O8 (a and b) and the decarboxylated C17H14O6 (c) by high-resolution FT-MS. These nonaketides, which were not observed from expression of the same genes in vivo4, harbor an additional malonate extender unit in comparison to the benzoate-primed enterocin and wailupemycin octaketides (Scheme 1b), which suggests that the nature and size of the primer unit dictates polyketide chain length.
Figure 1: HPLC analysis at 254 nm of organic extracts from different reconstituted enc PKS enzyme sets.
(a) EncA/B and EncC. (b) EncA/B and EncC plus benzoic acid. (c) EncA/B, EncC, EncD and EncN plus benzoic acid, ATP/Mg2+ and NADPH. (d) EncA/B, EncC, EncD, EncN and EncM plus benzoic acid, ATP/Mg2+ and NADPH. (e) EncA/B, EncC, EncD, EncN, EncM and EncK plus benzoic acid, ATP/Mg2+, NADPH and SAM. (f) EncA/B, EncC, EncD, EncN, EncM, EncK and EncR plus benzoic acid, ATP/Mg2+, NADPH, SAM, ferredoxin, ferredoxin-NADP+ reductase and catalase. (g) Authentic standards used for product analysis. All assays were performed in Tris-HCl buffer containing SgFabD and malonyl-CoA.
Full size image (24 KB)Previous experiments have demonstrated that the benzoate:ACP ligase EncN and the ketoreductase EncD are required in vivo for a minimally functional biosynthetic system4. To evaluate this phenomenon in vitro, the previous enzymatic reaction was supplemented with recombinant EncN8 and EncD together with their cofactors ATP (Compound 5) and NADPH (Compound 6), respectively. Upon incubation with benzoic acid and malonyl-CoA, the benzoate-primed natural products wailupemycin F (Compound 7) and wailupemycin G (Compound 8) predominated (Fig. 1c). This enzymatic conversion was dependent on all constituents being present, as even in the absence of NADPH the reaction reverted to the production of only the acetate-primed nonaketides a, b and c. This experiment confirmed the unprecedented interplay between the priming mechanism and ketide reduction previously documented in vivo4. Unlike other non-acetate-primed type II PKS systems, which often involve a primer-specific ACP and a polyketide-extension-specific ACP9, EncC is the only ACP used in enterocin biosynthesis. Given that Compound 7 and Compound 8 are reduced at C9, whereas a, b and c are not (Supplementary Table 1), EncD may additionally function as a gatekeeper that accepts benzoate-primed substrates while rejecting those primed with acetate.
The diversion of the aromatic PKS pathway to the oxidatively rearranged enterocin that harbors the tricyclic caged core was shown in vivo to be mediated by the "favorskiiase" flavoprotein EncM2. Overexpression of EncM in Escherichia coli gave soluble apoprotein, but by switching hosts to Streptomyces lividans, hexahistidyl-tagged EncM was expressed at similar concentrations yet was brightly pigmented yellow. Major UV absorption peaks at 375, 455 and 480 nm suggested the presence of the protein-bound flavin cofactor in its oxidized state10 (Supplementary Fig. 2 online). Addition of holo-EncM to the previous incubation harboring EncA/B, EncC, EncD, EncN and SgFabD gave desmethyl-5-deoxyenterocin (Compound 9) in addition to Compound 7 and Compound 8 (Fig. 1d), thereby confirming that EncM alone is responsible for catalyzing the Favorskii-like oxidative rearrangement and for facilitating the two aldol condensation and two heterocycle-forming reactions (Scheme 1). Though the mechanism of this rearrangement has not yet been clarified, the substrate of EncM is likely protein-bound via EncC, given that the released products Compound 7 and Compound 8 did not serve as reaction substrates.
The next reaction in the pathway is putatively catalyzed by the methyltransferase EncK2. Whereas expression in E. coli gave largely insoluble protein, expression in S. lividans provided soluble hexahistidyl-tagged EncK that was pigmented red with a mixture of prodiginines (Supplementary Fig. 3 online). Further expression in the prodiginine-deficient 
red host Streptomyces coelicolor YU105 (ref. 11) provided pigmentless EncK. Its addition to the EncA/B, EncC, EncD, EncM, EncN and SgFabD enzyme mixture together with the cofactor S-adenosyl-L-methionine (SAM, Compound 10) provided 5-deoxyenterocin (Compound 11) as the only methylated product (Fig. 1e). The identical 4-hydroxy-2-pyrone moiety in Compound 7 and Compound 8 was not methylated, even when pure compounds were directly assayed against EncK, thereby suggesting important substrate recognition features in Compound 9, which is consistent with our previous in vivo observations2.
Oxidation at C5 by the cytochrome P450 hydroxylase EncR is the final enzymatic reaction in the pathway and was previously documented in vitro with recombinant maltose-binding-protein-fused EncR, spinach ferredoxin, and ferredoxin-NADP+ reductase3. However, the enzymatic conversion involving the EncA/B, EncC, EncD, EncM, EncN, EncR, SgFabD, ferredoxin, ferredoxin-NADP+ reductase and catalase mixture stalled at Compound 11 with low turnover to enterocin (Compound 1). Further analysis revealed that EncR is inhibited by SAM, as previously reported for other cytochrome P450 enzymes12. To circumvent the inhibitory effect of SAM on EncR, the nonpolar reaction products, including Compound 11, were extracted under acidic conditions and then subjected to the final EncR-catalyzed reaction to yield the natural product Compound 1 (Fig. 1f,g).
The in situ generation of malonyl-CoA by Rhizobium leguminosarum malonyl-CoA synthetase MatB13 from malonic acid (Compound 12) and coenzyme A (CoA, Compound 13) provided the opportunity to use simple organic substrates for the ex vivo synthesis of Compound 11. Incubation of benzoic and malonic acids with EncA/B, EncC, EncD, EncK, EncM, EncN, SgFabD and RlMatB and associated cofactors similarly provided Compound 8 and Compound 11 as the major polyketides, albeit in lower overall yield in comparison to the incubation with commercial malonyl-CoA (Supplementary Table 1). CoA levels could be reduced to 0.1 equivalents malonate without adversely affecting product throughput.
In summary, the total syntheses of the enterocin and wailupemycin natural products have been demonstrated using purified enzymes. The longest linear series of reactions from benzoic acid to enterocin (Compound 1) results in the formation of ten C-C bonds, five C-O bonds and seven chiral centers, uses nine recombinant and three commercial proteins with 
25% overall yield, and consumes seven equivalents of malonyl-CoA and two equivalents NADPH per molecule of benzoic acid, SAM and ATP. Key features of this ex vivo synthesis are the sole use of biocatalysts, the simplicity of a single reaction vessel, and the involvement of the flavoprotein EncM, which catalyzes the unprecedented Favorskii-like rearrangement of the linear polycarbonyl precursor to the tricyclic enterocin skeleton.
Note: Supplementary information and chemical compound information is available on the Nature Chemical Biology website.
Author contributions
Q.C., L.X. and M.I. prepared recombinant proteins, Q.C. performed the biochemical assays and with B.S.M. discussed the results and wrote the paper, and D.M. conducted FT-MS analyses.
