Dissecting and Exploiting Intermodular Communication in Polyketide Synthases

See allHide authors and affiliations

Science  16 Apr 1999:
Vol. 284, Issue 5413, pp. 482-485
DOI: 10.1126/science.284.5413.482


Modular polyketide synthases catalyze the biosynthesis of medicinally important natural products through an assembly-line mechanism. Although these megasynthases display very precise overall selectivity, we show that their constituent modules are remarkably tolerant toward diverse incoming acyl chains. By appropriate engineering of linkers, which exist within and between polypeptides, it is possible to exploit this tolerance to facilitate the transfer of biosynthetic intermediates between unnaturally linked modules. This protein engineering strategy also provides insights into the evolution of modular polyketide synthases.

Since the discovery of the modular architecture of certain polyketide synthases (PKSs), several reports have highlighted the functional versatility of these multienzyme assemblies by experiments involving domain inactivation, substitution, or addition (1). Although these empirical gene fusion approaches have led to the biosynthesis of diverse “unnatural” natural products, they have usually resulted in decreased in vivo productivity (2). The reasons for the lower productivity are poorly understood but could include structural instability of the engineered protein, suboptimal chemistry within the altered module, or inefficient processing of the nonnatural polyketide intermediates by downstream modules.

An alternative strategy for combinatorial biosynthesis would be to recombine intact modules from the vast natural repertoire of PKSs. Such an approach would benefit from the use of highly evolved modules as intact catalytic units, thereby eliminating unwanted perturbations in module structure or chemistry. Along with the diverse chemistry observed in polyketide biosynthetic pathways, greater degrees of freedom could be combinatorially accessed through the interweaving of PKS modules with those of nonribosomal peptide synthetases, as found in nature (2). However, productive chain transfer between heterologous PKS modules has not been reported, perhaps because of the lack of understanding of the molecular basis for intermodular communication. Furthermore, although many bacterial genomes encode more than one PKS, each composed of multiple modules, there is no evidence for crosstalk between modules belonging to different dedicated PKS assemblies. A variety of experiments and observations have reinforced the view that individual modules show considerable selectivity toward their cognate substrates and that this intrinsic selectivity places serious constraints on harnessing the combinatorial potential of PKSs (1, 2). Here we present several lines of evidence that challenge this viewpoint. Our studies on four individually expressed, catalytically functional modules of the erythromycin PKS reveal that these modules have very similar kinetic parameters for extending a given diketide substrate into the corresponding triketide. Surprisingly, short intermodular segments of variable amino acid sequence, referred to as linkers, were found to play a crucial role in the assembly of functional modules as well as in the intermodular polyketide chain transfer. By appropriate engineering of these linkers, we show that it is possible to facilitate communication between heterologous modules. Our results suggest that whereas the chemistry within PKS modules may be substrate- and stereo-selective, chain transfer between intact modules is permissive as long as the evolutionarily optimized linkers can provide the connectivity between adjacent modules.

The remarkable overall selectivity exhibited by naturally occurring multimodular PKSs suggests that each module possesses significant substrate specificity. As a direct test of this hypothesis, we sought to functionally express and kinetically characterize representative modules of 6-deoxyerythronolide B synthase (DEBS) (Fig. 1), which synthesizes the macrocyclic core 1 of the antibiotic erythromycin (3). For technical convenience,Escherichia coli was chosen as the expression host. The thioesterase (TE) domain, which ordinarily occurs at the COOH terminus of module 6 of DEBS, was fused to the COOH-terminal end of each module to facilitate substrate turnover (4). A key barrier to the functional expression of PKS modules in E. coli is the inability of the host's phosphopantetheinyl transferases to posttranslationally modify the acyl carrier protein (ACP) domains of PKS modules. To overcome this problem, we coexpressed thesfp phosphopantetheinyl transferase from Bacillus subtilis in E. coli BL21(DE3) cells (5). Coexpression of the sfp gene was both necessary and sufficient for functional expression of NH2-terminal modules such as module 3+TE (M3+TE) and M5+TE (6), as assayed by their ability to convert the diketide thioester2 into the expected triketide products 3 and4, respectively (Fig. 2). However, no activity could be detected from similarly expressed COOH-terminal modules such as M2+TE and M6+TE (6), even though these proteins were homodimeric and chromatographically similar to their NH2-terminal counterparts. This lack of activity was unexpected because M2+TE and M6+TE are known to accept diketide thioester 2 when presented in their natural bimodular contexts (7). Detailed analysis of each domain of recombinant M2+TE indicated that whereas its acyltransferase (AT) domain was catalytically competent (as assessed by selective labeling of methylmalonyl-CoA) and its ACP domain was pantetheinylated, the ketosynthase (KS) domain could not be acylated with radiolabeled diketide thioester 2 (8). Because the KS domain is present at the NH2-terminal ends of these recombinant proteins, we suspected a problem at this end of the protein for these COOH-terminal modules. Amino acid sequence comparisons of the six DEBS modules had led to the identification of linker regions of variable sequence that spanned the boundaries between the highly conserved ACP domain at the COOH-terminal end of an upstream module and the KS domain at the NH2 terminus of the following module (9). Our analysis of these linker regions between DEBS modules, and those of other PKSs, revealed a marked contrast between the sequences separating covalently connected modules (for example, sequences between M1 and M2, M3 and M4, or M5 and M6 of DEBS) and those separating noncovalently connected modules (for example, sequences between M2 and M3 or M4 and M5 of DEBS). Specifically, the sequences separating covalently connected modules (hereafter referred to as intermodular linkers) are short and are characterized by the presence of a relatively conserved proline residue, whereas those upstream of NH2-terminal modules and downstream of COOH-terminal modules (hereafter referred to as interpolypeptide linkers) are longer and relatively hydrophilic (Fig. 3). The NH2-terminal ends of all KS domains contain a highly conserved stretch of four amino acids, PIAI (10), whose coding sequence often encompasses a Bsa BI restriction site (4 of 6 modules in DEBS and 8 of 10 modules in the rifamycin PKS possess this restriction site). Thus, construction of gene fusions at this site is a straightforward matter. Using this strategy, we replaced the DNA encoding the first 18 amino acids at the 5′ end of the inactive M2+TE construct described above with DNA encoding the first 39 amino acids from M5+TE (11). The resulting M2+TE protein could now accept diketide thioester2 and convert it into the expected triketide lactone (Fig. 2). Likewise, the activity of the above-mentioned M6+TE gene was restored by a similar replacement. These results indicated that intermodular linkers play a crucial role in the assembly of functional PKS modules. Moreover, the development of this strategy for heterologous expression of single modules provided us with convenient access to the reagents needed for evaluating module specificity.

Figure 1

Schematic representation of the modular organization of DEBS, which catalyzes the synthesis of the macrolide aglycone of erythromycin, 6-deoxyerythronolide B (6-dEB, 1). DEBS is a hexameric protein complex (α2β2γ2). Each constituent polypeptide chain (DEBS 1, 2, and 3; shown here in yellow, blue, and green, respectively) is composed of two modules: an NH2-terminal (lighter shade) and a COOH-terminal (darker shade) module that are covalently connected by an intermodular linker (shown in red). Each module contains a set of domains beginning with a KS domain, followed by an AT domain, and ending with an ACP domain. All modules except M3 have active ketoreductase (KR) domains, and M4 contains an additional dehydratase domain and an enoyl reductase domain. The TE domain (shown in pale blue) follows M6, which is responsible for the cyclization of the heptaketide intermediate to form 1. The loading domain (LD, shown in pink) is present at the beginning of M1. During the course of reaction, all the intermediates are covalently sequestered as acylthioesters of the corresponding ACP and KS domains. The growing chain gets passed on from one module to next, like a baton in a relay race.

Figure 2

Cell-free synthesis of triketides catalyzed by individual modules. All of the protein assays were done with varying concentrations of diketide thioester 2 (0.5 to 10 mM), 2.5 mM 14C-methylmalonyl-CoA, and 100 pmol of purified protein in a 100-ml reaction. Because the KR domain is inactive in M3+TE, 1 mM of NADPH was only used in the assay mixtures for M2+TE, M5+TE, and M6+TE proteins. The reaction mixtures were quenched and extracted by ethyl acetate and separated by means of thin-layer chromatography (TLC). The products were confirmed by simultaneously running standards on TLC plates. Time courses for the formation of triketide lactone 4 and triketide ketolactone3 were performed for 30 min. Quantitative measurements were done on a Packard InstantImager. All of these experiments were performed in triplicate.

Figure 3

Amino acid sequence comparisons (10) of representative intermodular (A) and NH2-terminal interpolypeptide (B) linkers from erythromycin (ery), rifamycin (rif), and rapamycin (rap) PKSs. These linker sequences have low sequence similarity. However, interpolypeptide linker residues have a higher occurrence of charged residues [bold residues in (B)], whereas nearly all of the intermodular linkers have a conserved proline residue [bold residues in (A)].

In addition to highlighting the important role of linkers in module connectivity, the above experiments showed that at least four distinct modules of DEBS were able to accept and extend diketide thioester 2, which is the natural substrate of M2 only. To quantify this unexpected substrate tolerance, the steady-state kinetic parameters of the purified individual modules were measured. As seen inTable 1, the apparent catalytic rate constant (k cat) and Michaelis constant (K M) for 2 are very similar for every module except M3+TE. The value of k cat for M3+TE is five times lower than that for the other three modules. A possible explanation for this difference is that the TE domain, which is present at the COOH-terminal ends of all four modules, discriminates between the ACP-bound β-hydroxythioester substrate generated by M2+TE, M5+TE, and M6+TE, and the β-ketothioester substrate generated by M3+TE. Recent studies with the isolated TE domain of DEBS have shown that it possesses significant specificity toward acylthioester substrates that have a (2R, 3S)-2-methyl-3-hydroxy stereochemistry (12), as is the case for the normal products of M2, M5, and M6. To assess the contribution of the TE domain to the apparent k cat and K M of M6+TE, these parameters were measured for 2 in the absence of NADPH (the reduced form of nicotinamide adenine dinucleotide phosphate), which eliminates the ketoreductase-catalyzed reaction from the catalytic cycle. Although the apparent K Mremained virtually unchanged (4.1 mM), the apparentk cat decreased by half. We therefore conclude that individual modules of multimodular PKSs have an unexpectedly broad tolerance toward incoming ketide chains and that these modules faithfully process incoming substrates.

Table 1

Kinetic parameters of individual modules for the synthesis of triketide.

View this table:

Because the acyl moiety of 2 is the product of the first module of DEBS, the above results suggest that it should be possible to synthesize novel polyketides by fusing M3, M5, or M6 downstream of M1. Although the previous attempts in our lab to construct a nonnative module fusion were unsuccessful (13), the results from heterologous expression of individual PKS modules inE. coli suggested that linkers might play a crucial role in facilitating chain transfer between adjacent modules. To assess the significance of the linkers, the fusions M1-M3+TE and M1-M6+TE were reconstructed in a manner that preserved the natural intermodular linker between M1 and M2 through use of the conserved Bsa BI restriction site described above (Fig. 4, A and B) (14). Remarkably, the recombinant strain ofStreptomyces coelicolor CH999 harboring M1-M3+TE synthesized the expected triketide ketolactone 3, as judged by nuclear magnetic resonance (NMR) spectroscopy, in quantities similar to the triketide synthesized by strains harboring DEBS1+TE (15 mg/liter). Likewise, M1-M6+TE also produced comparable quantities of the ketolactone 3, as judged by NMR spectroscopy (15). As a more demanding test of the role of linkers in polyketide chain transfer, we replaced M2 of DEBS with M5 from the rifamycin (rif) PKS from Amycolatopsis mediterranei (16). The α-methyl and β-hydroxyl stereocenters generated by M5 of the rif PKS are the same as those generated by M2 of DEBS, although the incoming acyl chain differs substantially in the two cases. The bimodular PKS generated by linking rif M5 downstream of DEBS M1 produced the expected triketide lactone 4 in vivo, as judged by NMR spectroscopy (Fig. 4C) (17). Again, the productivity of the recombinant strain of S. coelicolor expressing this PKS is comparable to that for the triketide lactone synthesized by DEBS1+TE (15 mg/liter). The successful engineering of a DEBS M1–rif M5 junction prompted us to test whether the interpolypeptide linker between M2 and M3 of DEBS could also facilitate chain transfer from rif M5 to DEBS M3. For this experiment, the COOH-terminal linker that ordinarily occurs downstream of ACP2 in DEBS1 was fused in place of the TE domain in the bimodular construct described in Fig. 4C (18). Upon coexpression of the resulting bimodular protein with DEBS2 and DEBS3, the recombinant strain of S. coelicolor produced 15 mg/liter of1, as judged by NMR spectroscopy (Fig. 4D). These studies demonstrate the feasibility of functionally rewiring PKS modules. Thus, the linker hypothesis, which was developed using an intermodular linker between M1 and M2 of DEBS, is also applicable to interpolypeptide linkers (as illustrated by the linker between M2 and M3 of DEBS). The linker hypothesis has two important implications. First, it suggests a simple model for the evolution of modular PKSs and possibly for nonribosomal peptide synthetases (2, 19), because gene duplication is both necessary and sufficient for the evolution of multimodular systems as long as linkers provide suitable module connectivity. Second, it provides a fundamentally new strategy for combinatorial biosynthesis, in which modules, rather than individual enzymatic domains, are the building blocks for genetic manipulation.

Figure 4

Engineered heterologous modular fusions. Constructs (A) and (B) are bimodular constructs engineered from the erythromycin PKS, where chain transfer is observed between M1 and an unnatural downstream module (M3 and M6, respectively), as long as the intermodular linker (shown in red) that is naturally present between M1 and M2 is maintained. Communication between two modules derived from different PKSs (erythromycin and rifamycin) is illustrated in (C). The ery M1-M2 linker is also maintained in this construct. The construct shown in (D) demonstrates successful interpolypeptide chain transfer between heterologous modules. In addition to retaining the ery M1-M2 linker (shown in red), the intermodular linkers that are naturally present between M2 and M3 are also preserved (shown in yellow).


View Abstract

Navigate This Article