Biosynthesis of Complex Polyketides in a Metabolically Engineered Strain of E. coli

See allHide authors and affiliations

Science  02 Mar 2001:
Vol. 291, Issue 5509, pp. 1790-1792
DOI: 10.1126/science.1058092


The macrocyclic core of the antibiotic erythromycin, 6-deoxyerythronolide B (6dEB), is a complex natural product synthesized by the soil bacterium Saccharopolyspora erythraea through the action of a multifunctional polyketide synthase (PKS). The engineering potential of modular PKSs is hampered by the limited capabilities for molecular biological manipulation of organisms (principally actinomycetes) in which complex polyketides have thus far been produced. To address this problem, a derivative ofEscherichia coli has been genetically engineered. The resulting cellular catalyst converts exogenous propionate into 6dEB with a specific productivity that compares well with a high-producing mutant of S. erythraea that has been incrementally enhanced over decades for the industrial production of erythromycin.

Erythromycin, a broad spectrum antibiotic synthesized by the bacterium S. erythraea, is a prototype of a class of complex natural products called polyketides (1). These biomolecules are synthesized from simple building blocks such as acetyl coenzyme A (acetyl-CoA), propionyl-CoA, malonyl-CoA, and methylmalonyl-CoA through the action of large modular megasynthases called PKSs (2). Polyketide structural complexity often precludes the development of practical laboratory synthetic routes, leaving fermentation as the only viable source for the commercial production of these pharmaceutically and agriculturally useful agents. But the challenges associated with developing scalable and economically feasible fermentation processes for polyketide production from natural biological sources (principally theActinomyces family of bacteria) are enormous and represent the most serious bottleneck during polyketide preclinical and clinical development. Analogous to the pivotal role of E. coli in the production of therapeutic proteins, the ability to produce complex polyketides in this exceptionally well-developed heterologous host can be expected to benefit natural product drug and agrochemical development.

Notwithstanding the importance of this objective, three challenges are encountered in the production of complex polyketides inE. coli. The first is the ability to express correctly folded and posttranslationally modified modular PKSs in this heterologous host. For example, deoxyerythronolide B synthase (DEBS), which synthesizes the macrocyclic core of erythromycin, is a 2-megadalton α2β2γ2 complex composed of two equivalent sets of 28 distinct active sites, seven of which must be posttranslationally pantetheinylated. Once coexpressed inE. coli in an active form, the three DEBS subunits would represent the largest and perhaps the most highly elaborated polypeptides in this bacterium. Second, the building blocks for polyketide biosynthesis must be available in E. coli. For example, 6-deoxyerythronolide B (6dEB, 1) is synthesized from one propionyl-CoA–derived primer unit and six (2S)-methylmalonyl-CoA–derived extender units (Fig. 1). Although E. coli synthesizes propionyl-CoA, this substrate is formed under conditions that are incompatible with typical high cell density fermentation conditions (3). Moreover, (2S)-methylmalonyl-CoA has not been observed as a metabolite in E. coli (4). Therefore production of 6dEB in E. coli would require re-engineering its metabolism to synthesize these two precursors. Lastly, in order to design an effective cellular catalyst for 6dEB biosynthesis, intracellular DEBS activity needs to be well-synchronized with precursor biosynthesis to maximize the relevant stoichiometric and kinetic parameters. Here we address each of these three issues.

Figure 1

(A) The 6-deoxyerythronolide B synthase. Catalytic domains: KS, ketosynthase; AT, acyl transferase; ACP, acyl carrier protein; KR, ketoreductase; ER, enoyl reductase; DH, dehydratase, TE, thioesterase. DEBS utilizes 1 mole of propionyl-CoA and 6 moles of (2S)-methylmalonyl-CoA to synthesize 1 mole of 6-deoxyerythronolide B (6dEB, 1). (B) Truncated DEBS1+TE produces the triketide lactone 2. (C) The rifamycin synthetase is a polyketide synthase that is naturally primed by a nonribosomal PKS loading module, comprised of two domains: an ATP-dependent adenylation domain (A) and a thiolation domain (T). Substitution of this A-T didomain in placeof the loading didomain of DEBS yields an engineered “hybrid” synthase that utilizes exogenous acids such as benzoic acid to synthesize substituted macrocycles such as 3 in an engineered strain ofE. coli.

Although the expression of DEBS genes has been achieved in the model actinomycete Streptomyces coelicolor (5), to our knowledge genes of this size have never been functionally expressed in E. coli. Therefore, we first attempted to express the genes encoding each of the three DEBS proteins, followed by in vitro assays of protein activity. DEBS3, DEBS2, and a variant of DEBS1, DEBS1+TE (6, 7), were cloned individually into the pET21c expression vector (8) and introduced via transformation into E. coli BL21(DE3) harboring the sfp phosphopantetheinyl transferase gene (9) on pRSG56 (10). The expression levels of the three DEBS genes (11) were found to be comparable to those reported earlier from S. erythraea (12) orS. coelicolor (13). However, although DEBS3 was found to be active in these lysates, DEBS1+TE and DEBS2 lacked any detectable activity (14). Consistent with these results, recombinant DEBS3 could be purified from these lysates using procedures described earlier (15), but neither DEBS1+TE nor DEBS2 could be purified in detectable quantities. The key parameter that facilitated detection of in vitro activity and subsequent purification of DEBS1+TE and DEBS2 was the incubation temperature after IPTG (isopropyl-β-d-thiogalactopyranoside) induction. Upon lowering the expression temperature from 30° to 22°C, active DEBS1+TE, DEBS2, and DEBS3 proteins could be detected in recombinantE. coli lysates. Low temperature induction conditions were subsequently maintained throughout the course of this study.

Although plasmid-based coexpression of the sfp gene facilitated posttranslational modification of the DEBS proteins inE. coli, integration of this gene into the chromosome was desirable before attempting coexpression of all the DEBS genes. Using homologous recombination, a single copy of the sfp gene under control of the T7 RNA polymerase promoter was integrated in theprp operon of BL21(DE3), yielding E. coli BAP1 (Fig. 2) (16). This site was chosen for sfp gene insertion for two reasons. First, the prp operon is putatively responsible for propionate catabolism in E. coli (17). Because propionate was intended to be the sole source of carbon building blocks for 6dEB biosynthesis, concurrent propionate catabolism and anabolism were deemed undesirable. By deleting prpRBCDin the process of sfp integration, the ability of BAP1 to utilize propionate as a carbon and energy source was eliminated. Second, together with the sfp gene, the prpE gene in BAP1 was also placed under control of an IPTG-inducible T7 promoter. PrpE is thought to convert propionate into propionyl-CoA (18); therefore, in the presence of exogenous propionate, propionyl-CoA can be expected to accumulate inside the cell at the same time as DEBS is expressed in an active form.

Figure 2

Genetic design of E. coli BAP1.

As a final step toward engineering E. coli for 6dEB production, a suitable pathway for (2S)-methylmalonyl-CoA biosynthesis needed to be engineered. Recently, the propionyl CoA carboxylase (pcc) genes from S. coelicolor have been isolated and characterized (19). Preliminary experiments confirmed that the genes encoding the two subunits,pccA and pccB, could be expressed in E. coli in an active form and that the activity of the biotinylated subunit (pccA) could be enhanced upon coexpression of the E. coli birA biotin ligase gene (20). To test whether propionyl-CoA carboxylase can synthesize the correct isomer of methylmalonyl-CoA in vivo, BAP1 was transformed with three plasmids: pRSG32 carrying the DEBS1+TE gene under control of the pT7 promoter (8), pTR132 carrying the pccAB genes (19), and pCY216 carrying the birA gene (21). M9 minimal media cultures of transformed cells were grown at 37°C to mid-log phase, followed by induction at 22°C with 0.5 liter mM IPTG, arabinose (2.5 g/liter), and [1-14C]– or [1-13C]–propionate (26 mg/liter or 250 mg/liter), respectively (22). After 12 to 48 hours, the culture supernatant was extracted and analyzed for formation of the expected triketide lactone (2) product of DEBS1+TE. Formation of triketide lactone under both feeding conditions confirmed the ability of BAP1 to produce polyketides and encouraged us to attempt 6dEB production in the same host.

To coexpress all three DEBS genes and the pcc genes, vectors pET21c and pET28a (Novagen, Milwaukee, Wisconsin) were modified to express two and three genes, respectively (23). When tested individually, protein production was observed from each gene located on both plasmids. BAP1 was transformed with these plasmids together with the birA plasmid. Individual transformants were cultured, induced, and analyzed similar to the experiment for DEBS1+TE using [1-13C]–propionate. Nuclear magnetic resonance (NMR) analysis of the crude organic extract revealed 6dEB as the major propionate-derived metabolite of these recombinant cells. The product was later purified by high-performance liquid chromatography (HPLC) and subjected to mass spectrometry yielding a major peak of the expected mass (24). To quantify the productivity of this newly characterized polyketide cellular system, culture samples were taken periodically, and the concentration of 6dEB was measured (Fig. 3). From this data it can be calculated that the specific productivity of this cellular catalyst is 0.1 mmol of 6dEB per gram of cellular protein per day. This is superior to wild-type S. erythraea and compares well to an industrially relevant strain that overproduces erythromycin (0.2 mmol of erythromycin per gram of cellular protein per day) (25) as a result of a decades-long program of directed-strain improvement based on random mutagenesis.

Figure 3

Production of 6dEB in E. coli. Cellular protein content and 6dEB concentration are plotted versus time.

To demonstrate the power of engineering modular PKSs in this previously unknown heterologous system, we attempted to construct a derivative of DEBS in which a PKS module was fused to a nonribosomal peptide synthetase (NRPS)–like (26) module. The first module of the rifamycin synthetase has recently been shown to be an NRPS-like module comprised of two domains: an adenylation (A) and a thiolation (T) domain (27). The A domain activates 3-amino-5-hydroxybenzoate (as well as benzoate and several benzoate derivatives) (27) in an adenosine triphosphate (ATP)–dependent reaction and transfers the aryl adenylate onto the phosphopantetheine arm of the T domain (Fig. 1). This NRPS-like module was fused upstream of the first condensation module of DEBS in lieu of the loading didomain of DEBS (28). In the presence of exogenous propionate and benzoate, the resulting strain of E. coli produced the expected 6dEB analog (3), as confirmed by NMR and mass spectrometry (Fig. 1) (29).

In summary, we have demonstrated the feasibility of engineeringE. coli to produce complex polyketide natural products. Multiple changes were made to the E. coli genome for relevant 6dEB production, including introduction of the three DEBS genes from S. erythraea, introduction of the sfpphosphopantetheinyl transferase gene from Bacillus subtilis, introduction of genes encoding a heterodimeric propionyl-CoA carboxylase from S. coelicolor, deletion of the endogenousprpRBCD genes, and overexpression of the endogenousprpE and birA genes. When gene expression was coordinately induced at low temperature, propionate could be converted into 6dEB by this metabolically engineered cellular catalyst with excellent kinetic parameters. Given the availability of well-established scalable protocols for fermenting E. colito overproduce bioproducts, the ability to synthesize complex polyketides in this heterologous host bodes well for the practical production of these bioactive natural products. Equally important, as indicated by the hybrid PKS-NRPS described here, it opens the door for harnessing the enormous power of molecular biology in E. colito engineer modular polyketide synthases using directed and random approaches.

  • * Permanent address: Facultad de Ciencias Bioquimicas y Farmaceuticas, Departamento de Microbiologia, Suipacha 531, Rosario 2000, Argentina.

  • To whom correspondence should be addressed. E-mail: ck{at}


View Abstract

Navigate This Article