A dual transacylation mechanism for polyketide synthase chain release in enacyloxin antibiotic biosynthesis

Summary

Polyketide synthases assemble diverse natural products with numerous important applications. The thioester intermediates in polyketide assembly are covalently tethered to acyl carrier protein domains of the synthase. Several mechanisms for polyketide chain release are known, contributing to natural product structural diversification. Here we report a dual transacylation mechanism for chain release from the enacyloxin polyketide synthase, which assembles an antibiotic with promising activity against Acinetobacter baumannii. A non-elongating ketosynthase domain transfers the polyketide chain from the final acyl carrier protein domain of the synthase to a separate carrier protein and a nonribosomal peptide synthetase condensation domain condenses it with (1S, 3R, 4S)-3, 4-dihydroxycyclohexane carboxylic acid. Molecular dissection of this process reveals that non-elongating ketosynthase domain-mediated transacylation circumvents the inability of the condensation domain to recognize the acyl carrier protein domain. Several 3, 4-dihydroxycyclohexane carboxylic acid analogues can be employed for chain release, suggesting a promising strategy for producing enacyloxin analogues.

Introduction

Polyketides are a large and structurally diverse family of specialized metabolites with a wealth of applications in medicine and agriculture, such as the treatment of infectious diseases and cancer, the protection of crops, and animal health.¹ Many industrially-important polyketides are assembled by type I modular polyketide synthase (PKS) biosynthetic assembly lines.² These remarkable molecular machines are capable of constructing a wide range of structurally complex carbon skeletons from acyl thioester building blocks via a series of condensation and modification reactions. With the exception of the loading module, each module in such PKSs is typically responsible for one round of chain elongation and minimally contains an acyl carrier protein (ACP) domain, which is post-translationally modified via attachment of a phosphopantetheine prosthetic “arm” to a conserved serine residue and a ketosynthase (KS) domain, which catalyzes chain elongation.² Modular PKSs fall into two distinct phylogenetic groups that are typically distinguished by the presence or absence of acyltransferase (AT) domains in their modules. In cis-AT PKSs, each module contains an AT domain that is responsible for loading an (alkyl)malonyl extender unit onto the adjacent ACP domain, whereas in trans-AT PKSs individual modules lack AT domains. Instead, a single AT catalyzes the transfer of malonyl extender units from coenzyme A onto the ACP domains in each module of the PKS.

Modular PKSs employ several strategies for generating structural diversity, such as the incorporation of a range of starter and extender units, various α- and β-carbon processing reactions after each round of chain extension and several distinct mechanisms for release of the fully-assembled polyketide chain.¹ The most frequently encountered and well-studied chain release mechanism involves macrolactonization catalyzed by a thioesterase (TE) domain appended to the C-terminus of the final PKS module.³ Examples of TE domains that catalyze chain release via thioester hydrolysis or macrolactamization are also known.^3,4 In addition, several other types of catalytic domains have been reported to catalyze chain release. Examples include thioester reductase domains, which catalyze reductive release to form an aldehyde or primary alcohol,^5,6 α-oxoamine synthase domains, which release the polyketide chain via decarboxylative condensation with the α-carbon of an amino acid,^7,8 and KS-like domains that catalyze condensation of the polyketide chain with a glyceryl-ACP to form a 4-hydroxymethyl-2-acyltetronic acid (Supplementary Fig. 1).⁹

Enacyloxin IIa (1) is a polyketide antibiotic^10,11 with clinically-relevant activity against Acinetobacter baumannii (MIC = 3 μg/ml),¹² a problematic multidrug-resistant Gram-negative pathogen. It selectively inhibits bacterial protein biosynthesis by binding to ribosomal elongation factor Tu.^13–15 We recently identified the enacyloxin biosynthetic gene cluster in Burkholderia ambifaria AMMD and proposed a pathway for enacyloxin biosynthesis, involving construction of the 25-carbon acyl chain by an unusual modular PKS containing a mixture of cis and trans-AT subunits (Bamb_5925-5919).¹² Release of the fully-assembled polyketide chain from the final (Bamb_5919) subunit is proposed to proceed via an unusual dual transacylation mechanism. A non-elongating ketosynthase (KS⁰) domain appended to the C-terminus of Bamb_5919 is hypothesized to transfer the polyketide chain from the adjacent ACP domain to the C-terminal peptidyl carrier protein (PCP) domain of Bamb_5917 (Fig. 1a). Such KS⁰ domains lack a conserved His residue required for chain elongation and are commonly found in trans-AT PKSs (Supplementary Fig. 2). They are proposed to transfer biosynthetic intermediates between ACP domains,^{16, 17} but biochemical evidence for this is limited to a single partially characterized example,¹⁸ and their functional significance remains unclear. Once the fully assembled enacyloxin polyketide chain has been transferred to the PCP domain of Bamb_5917 it is proposed to be released via condensation with the C-3 hydroxyl group of (1S, 3R, 4S)-3, 4-dihydroxycyclohexane carboxylic acid (DHCCA). Bamb_5915, which shows sequence similarity to condensation (C) domains, typically responsible for peptide bond formation between PCP-bound amino acyl thioesters in nonribosomal peptide synthetase (NRPS) multienzymes, is hypothesized to catalyze this reaction (Fig. 1a).¹²

Fig 1.

Proposed mechanism for chain release from the enacyloxin PKS and confirmation that Bamb_5915 and Bamb_5917 are involved in enacyloxin biosynthesis. (a) Proposed dual transacylation mechanism for chain release from the type I modular PKS responsible for enacyloxin IIa (1) biosynthesis. The KS⁰ domain at the C-terminus of Bamb_5919 (the final PKS module) is proposed to transfer the fully-assembled polyketide chain from the upstream ACP domain to the Bamb_5917 PCP domain. Bamb_5915, which shows sequence similarity to NRPS C domains, catalyzes condensation of the resulting thioester with (1S, 3R, 4S)-3,4-dihydroxycyclohexane carboxylic acid. (b) Deletion of the genes encoding Bamb_5915 and Bamb_5917 abolishes enacyloxin production in B. ambifaria BCC0203. Extracted ion chromatograms at m/z = 724.2267 ± 0.005 (corresponding to the [M+Na]⁺ ion for enacyloxin IIa) from UHPLC-ESI-Q-TOF-MS analyses of extracts from agar-grown cultures of wild type B. ambifaria BCC0203 (top), the bamb_5917 (second from top) and bamb_5915 (second from bottom) mutants. The chromatograms for extracts from the bamb_5915 and bamb_5917 mutants complemented by in trans expression of the deleted genes are also shown (bottom and third from bottom, respectively). The peak corresponding to enacyloxin IIa is indicated with an asterisk. The other peaks are isomers resulting from light / acid-promoted isomerization during isolation / analysis.

Here, we report an extensive set of genetic and biochemical experiments that establish the role played by Bamb_5915, Bamb_5917 and the KS⁰ domain of Bamb_5919 in the unusual chain release mechanism employed by the enacyloxin PKS. These experiments demonstrate that the Bamb_5919 KS⁰ domain overcomes the inability of Bamb_5915 to recognize the Bamb_5919 ACP domain by transferring the fully assembled polyketide chain to the PCP domain of Bamb_5917. We also show that Bamb_5915 possesses relaxed substrate specificity, indicating that it has the potential to be exploited for the production of novel enacyloxin analogues.

Results

Bamb_5915 and Bamb_5917 are required for enacyloxin biosynthesis

To establish whether Bamb_5915 and Bamb_5917 are required for enacyloxin IIa biosynthesis, we created in-frame deletions in the corresponding genes. Because our originally identified enacyloxin producer, B. ambifaria AMMD, proved refractory to genetic manipulation using a homing endonuclease-based mutagenesis system¹⁹, these deletions were created in B. ambifaria BCC0203 (also known as B. ambifaria BC-F²⁰). The enacyloxin biosynthetic gene cluster in B. ambifaria BCC0203 has the same organization as that in the AMMD strain and shows a very high degree of sequence similarity (Supplementary Fig. 3). UHPLC-ESI-Q-TOF-MS analyses of extracts from agar cultures showed that enacyloxin production is abolished in the bamb_5915 and bamb_5917 mutants. Complementation of these mutants via in trans expression of deleted genes restored enacyloxin production (Fig. 1b). These results show that Bamb_5915 and Bamb_5917 play an essential role in enacyloxin biosynthesis.

The Bamb_5919 KS⁰ domain is a carrier protein transacylase

To investigate the function of the Bamb_5919 KS⁰ domain, we overproduced Bamb_5917, its C-terminal PCP domain (lacking the N-terminal domain of unknown function), and the ACP and KS⁰ domains of Bamb_5919 (both individually and as an ACP-KS⁰ di-domain) in E. coli as N-terminal His₆-fusion proteins, and purified them to homogeneity using nickel affinity chromatography (Supplementary Fig. 4). The identity of all purified proteins was confirmed by UHPLC-ESI-Q-TOF-MS (Supplementary Fig. 4; note that Bamb_5917, and the ACP/PCP domains of Bamb_5919 and Bamb_5917 are produced in their apo-forms, presumably because the E. coli phosphopantetheinyl transferase is unable to recognize them).

We first sought to establish the function of the KS⁰ domain, which has been hypothesized to catalyze the transfer of the fully assembled polyketide chain from the Bamb_5919 ACP domain to the Bamb_5917 PCP domain. To investigate this hypothesis, we converted the Bamb_5917 PCP domain to its holo-form by incubating it with CoA and the phosphopantetheinyl transferase Sfp (Supplementary Fig. 5). Similarly, an S-acetyl derivative of the holo-ACP domain from Bamb_5919 was created by incubating the apo-protein with Sfp and acetyl-CoA (Supplementary Fig. 6). The acetylated Bamb_5919 ACP domain was incubated with the Bamb_5919 KS⁰ domain and the Bamb5917 holo-ACP domain to examine whether the KS⁰ domain can transfer the acetyl group (which serves as a simple mimic of the fully assembled enacyloxin polyketide chain) from the ACP domain to the PCP domain. UHPLC-ESI-Q-TOF-MS analyses of the resulting mixture showed that 40.2 ± 1.4 % of the PCP domain underwent acetylation (Fig. 2a). Due to the similar bond enthalpies for the linkages broken/formed in this reaction, an approximately 1:1 mixture of the ACP and PCP thioesters is produced. The level of acetylation of the Bamb_5917 PCP domain significantly decreased when the Bamb_5919 KS⁰ domain was omitted from the reaction (8.4 ± 0.9 %), or when it was replaced with a C1988A mutant (in which the thiol group has been removed from the active site Cys residue; 7.9 ± 0.8 %). These data are consistent with the hypothesis that the KS⁰ domain catalyses translocation of the fully assembled polyketide chain in enacyloxin biosynthesis from the Bamb_5919 ACP domain to the Bamb_5917 PCP domain. Substitution of the Bamb_5919 ACP and KS⁰ domains with the ACP-KS⁰ di-domain, or the Bamb_5917 PCP domain with full-length Bamb_5917 in these experiments gave analogous results (Supplementary Fig. 7 and 8). This shows that neither the covalent linkage between the ACP and KS⁰ domains, nor the N-terminal appendage of the PCP domain present in the wild type system are critical for catalysis of the transacylation reaction.

Fig 2.

Acyl transfer and active site acylation assays reveal that the KS⁰ domain of Bamb_5919 functions as a transacylase. (a) Deconvoluted mass spectra from incubation of the holo-Bamb_5917 PCP domain with the acetylated Bamb_5919 ACP domain in the presence of the Bamb_5919 KS⁰ domain (top) and the C1988A mutant of the KS⁰ domain (middle), and in the absence of the KS⁰ domain (bottom). The data demonstrate that the KS⁰ domain is able to transfer an acetyl group from the ACP domain to the PCP domain and that the active site Cys residue of the KS⁰ domain is required for this reaction. Due to the similar bond enthalpies for the thioester linkages broken/formed in the reaction catalyzed by the KS⁰ domain an approximately 1:1 mixture of starting material and product is formed. (b) Deconvoluted mass spectra of the Bamb_5919 ACP domain (left), the Bamb_5919 KS⁰ domain (top right) and the C1988A mutant of the Bamb_5919 KS⁰ domain (bottom right), following incubation of the wild type and mutant KS⁰ domains with a 10-fold excess of acetyl-Bamb_5919 ACP. Transfer of the acetyl group from the ACP domain to the KS⁰ domain is observed for the wild type enzyme (top), but not for the mutant in which the active site Cys residue has been mutated to Ala (bottom). The data shown are from a single measurement and are representative of three independent experiments.

To directly probe the acylation of the active site Cys residue in the Bamb_5919 KS⁰ domain during the transfer of the acetyl group from the Bamb_5919 ACP domain to the Bamb_5917 PCP domain, the KS⁰ domain was incubated with a 10-fold molar excess of the acetylated Bamb_5919 ACP domain. Analysis of the protein mixture by UHPLC-ESI-Q-TOF-MS showed that the KS⁰ domain undergoes acetylation (35.1 ± 2.3 %) (Fig. 2b). In contrast, no acetylation was observed in such analyses when the acetylated Bamb_5919 ACP domain was incubated with the C1988A mutant of the KS⁰ domain.

Bamb_5915 catalyzes chain release from Bamb_5917

The role played by Bamb_5915 (a homologue of NRPS C domains) in enacyloxin biosynthesis was similarly investigated via overproduction in E. coli as an N-terminal His₆ fusion protein, purification using nickel affinity chromatography and confirmation of identity by UHPLC-ESI-Q-TOF-MS (Supplementary Fig. 4). The S-acetyl derivative of the Bamb_5917 holo-PCP domain was created by incubating the apo-protein with acetyl-CoA and Sfp (Supplementary Fig. 6). Incubation of this simple mimic of the Bamb_5917-bound enacyloxin polyketide chain with Bamb_5915 and chemically synthesized racemic DHCCA resulted in the production of a monoacetylated DHCCA derivative, as evidenced by UHPLC-ESI-Q-TOF-MS analyses (Supplementary Fig. 9). The greater thermodynamic stability of the ester bond formed in this reaction than the thioester bond in the acetylated PCP drives product formation. No acetylation of DHCCA was observed when Bamb_5915 was omitted from the reaction. We were unable to confirm the structure of the monoacetylated DHCCA derivative produced in the Bamb_5915-catalysed reaction, due to its propensity to undergo rearrangement during purification and the difficulty of preparing authentic standards. To circumvent this problem, we synthesized racemic (1S, 3R, 4S)-3-amino-4-hydroxycyclohexane carboxylic acid (AHCCA), an analogue of DHCCA containing an amino group in place of the C-3 hydroxyl group. Incubation of AHCCA with the acetylated Bamb_5917 PCP domain and Bamb_5915 resulted in a monoacetylated product (Fig. 3a). The identity of this product was confirmed as N-acetyl-AHCCA by comparison with a chemically-synthesised standard (Fig. 3a). None of this product was formed when Bamb_5915 was omitted from the reaction (Fig. 3a), or when an H205A mutant of Bamb_5915 (in which the active site His residue known to play an important role in catalysis in NRPS C domains has been altered) was employed (Supplementary Fig. 10).

Fig. 3.

Functional characterization of Bamb_5915. Extracted ion chromatograms at m/z = 224.089 ± 0.005 (corresponding to the [M+Na]⁺ ion for N-acetyl-AHCCA) from UHPLC-ESI-Q-TOF-MS analyses of Bamb_5915-calaysed reactions. (a) Incubation of the acetylated Bamb_5917 PCP domain with AHCCA and Bamb_5915 results in a monoacetylated product with the same retention time as a chemically synthesized authentic standard of N-acetyl-AHCCA. (b) No acetylated products are observed when Bamb_5915 and AHCCA are incubated with the acetylated Bamb_5919 ACP domain.

Chain transfer is necessitated by Bamb_5915 carrier protein specificity

Having established that Bamb_5915 is able to catalyse chain release from Bamb_5917, we next investigated whether this enzyme can also offload acyl groups from the Bamb_5919 ACP domain. Thus, the acetylated Bamb_5919 ACP domain was incubated with Bamb_5915 and AHCCA. No products with an m/z corresponding to mono-acetylated AHCCA could be detected by UHPLC-ESI-Q-TOF-MS analysis of the reaction mixture, indicating that Bamb_5915 is unable to recognise acyl groups bound to the Bamb_5919 ACP domain (Fig. 3b). These data are consistent with the ability of the Bamb_5919 KS⁰ domain to transfer acyl groups from the Bamb_5919 ACP domain to the Bamb_5917 PCP domain and strongly indicate that such transacylation is required to circumvent the intrinsic carrier protein specificity of Bamb_5915. To directly test this hypothesis, we incubated the acetylated Bamb_5919 ACP domain with the Bamb_5919 KS⁰ and Bamb_5917 holo-PCP domains, Bamb_5915 and AHCCA. UHPLC-ESI-Q-TOF-MS analyses confirmed that N-acetyl-AHCCA is produced in this reaction (Fig. 4). When the Bamb_5919 KS⁰ domain was omitted from the reaction or replaced by the C1988A mutant, small amounts of N-acetyl-AHCCA were still produced (Fig. 4). This likely results from uncatalyzed transfer of the acetyl group from the Bamb_5919 ACP domain to the Bamb_5917 PCP domain via direct transthioesterifcation, which has been observed in related systems.²¹ It is not the result of the spontaneous transfer of the acetyl group from the acetylated Bamb_5919 ACP domain to AHCCA, because no N-acetyl-AHCCA was detected when Bamb_5915 was omitted from the reaction (Fig. 4). Analogous results were obtained when the acetylated full length Bamb_5917 protein was used in place of the acetylated Bamb_5917 PCP domain (Supplementary Fig. 11). This indicates that the cryptic N-terminal domain of Bamb_5917 plays no specific role in the chain release process and is likely an evolutionary remnant. However, other roles for this domain (e.g. in mediating protein-protein interactions with other components of the biosynthetic machinery) cannot be excluded. To provide further evidence for the relevance of these findings to chain release from the enacyloxin PKS, we repeated the experiments with acyl groups that more closely mimic the fully assembled enacyloxin polyketide chain. Thus, dodecanoylated and (2E, 4E)-2, 4-hexadienoylated-Bamb_5917_ACP domain were incubated separately with the Bamb_5919 KS⁰ and Bamb_5917 holo-PCP domains and Bamb_5915. In both cases, a monoacylated AHCCA derivative was produced (Supplementary Figs 12 and 13). No product could be detected when Bamb_5915 or the Bamb_5919 KS⁰ domain were omitted from reaction, or when the C1988A mutant of the KS⁰ domain was used (Supplementary Figs 12 and 13). Taken together, our data show that Bamb_5915 is unable to interact productively with the Bamb_5919 ACP domain. This necessitates transfer of the fully assembled enacyloxin polyketide chain to the Bamb_5917 PCP domain (catalysed by the Bamb_5919 KS⁰ domain) where it is released by Bamb_5915 via condensation with DHCCA.

Fig. 4.

In vitro reconstitution of chain release from the enacyloxin PKS. Extracted ion chromatograms at m/z = 224.089 ± 0.005 (corresponding to the [M+Na]⁺ ion for N-acetyl-AHCCA) from LC-ESI-Q-TOF-MS analyses of synthetic N-acetyl-AHCCA (top), and the product of the reaction of the acetylated Bamb_5919 ACP domain with the Bamb_5915 KS⁰ domain, the Bamb_5917 holo-PCP domain, Bamb_5915 and AHCCA (second from top). The bottom three chromatograms are from control reactions in which Bamb_5915 has been omitted (third from bottom), the Bamb_5919 KS⁰ domain has been omitted (second from bottom), and the C1988A mutant of the Bamb_5919 KS⁰ domain has been used in place of the wild type enzyme (bottom).

Bamb_5915 tolerates various acyl acceptor and donor analogues

The observation that Bamb_5915 can tolerate an acetyl thioester and AHCCA in place of the natural acyl donor and acceptor, respectively, prompted us to further investigate the substrate tolerance of this enzyme. To probe acyl acceptor specificity, we incubated Bamb_5915 with the acetylated Bamb_5917 PCP domain and several chemically-synthesized and commercially available DHCCA analogues. Analysis of the reaction mixtures by UHPLC-ESI-Q-TOF-MS showed that Bamb_5915 is able to accept a variety of carbocyclic DHCCA analogues with variations in ring size, degree of unsaturation, substitution pattern and stereochemistry (Fig. 5a and Supplementary Fig. 9). Moreover, the enzyme was also able to utilise acyclic DHCCA analogues, such as γ-aminobutyric acid (Fig. 5a). Indeed, the minimal determinant of substrate acceptance by Bamb_5915 appears to be the 1, 3-juxtaposition of a nucleophile and a carboxyl group (Fig. 5a and Fig. 5b). The acyl donor specificity of Bamb_5915 was investigated by assessing its ability to catalyse acylation of AHCCA with several N-acetyl cysteamine (NAC) thioester analogues of the native substrate (Supplementary Fig. 14). While various straight chain thioesters (e.g. acetyl, propionyl, and (2E, 4E)-2, 4-hexadienoyl) were well tolerated by Bamb_5915, little or no acylation of AHCCA was observed when the enzyme was incubated with α-branched substrates, such as isobutyryl, pivaloyl and serinyl-NAC thioesters.

Fig. 5.

Bamb_5915 tolerates a wide range of acyl acceptors. Overview of DHCCA analogues tested as substrates for Bamb_5915 using acetylated Bamb_5917 as an acyl donor. (a) Structures of DHCCA analogues converted to monoacetylated products by Bamb_5915. (b) DHCCA analogues not accepted as substrates by Bamb_5915.

Discussion

The dual transacylation mechanism elucidated here for chain release from the enacyloxin modular polyketide synthase has several unusual features. First, it involves a rare intermolecular esterification reaction catalyzed by a standalone C domain,²² which is a novel mechanism for modular PKS chain release. Second, the standalone C domain appears to possess uncommonly broad substrate tolerance towards both acyl acceptors and acyl donors, and third, the polyketide chain must be translocated by a KS⁰ domain from a carrier protein domain that is not recognized by the C domain, to one that is, for chain release to occur.

KS⁰ domains are a common feature of trans-AT PKSs,¹⁶ but their functional significance has remained unclear. Our data show that the KS⁰ domain appended to the C-terminus of Bamb_5919 functions as a transacylase, consistent with the hypothesis that it shuttles the fully assembled polyketide chain from the upstream ACP domain to the PCP domain of Bamb_5917. This is necessary because Bamb_5915, an NRPS-like C domain that catalyzes polyketide chain release via intermolecular esterification with DHCCA, can recognize acyl donors when they are attached to the Bamb-5917 PCP domain, but not the Bamb_5919 ACP domain. Indeed, KS⁰ domains are commonly found at the interface between PKS and NRPS subunits in hybrid trans-AT PKS/NRPS assembly lines (Fig. 6),¹⁶ where they have been shown to play an essential role,¹⁷ and are proposed to translocate an acyl thioester intermediate from an upstream ACP/PCP domain to a downstream PCP/ACP domain.¹⁶ Thus, the inability of C (and other NRPS catalytic) domains to recognize acyl donors when they are attached to trans-AT PKS ACP domains appears to be a general phenomenon. Our results suggest that KS⁰ domains have been recruited during the evolution of many such hybrid systems as “adapters” to overcome this obstacle. In the accompanying manuscript, we show that protein-protein interactions mediated by mutually-compatible docking domains are primarily responsible for the specific recognition of the Bamb_5917 PCP domain by Bamb_5915.²³ While it seems very likely that such interactions also underlie similar carrier protein incompatibilities in other systems, further experiments will be required to establish their precise nature.

Fig. 6.

Examples of trans-AT PKS architectures containing KS⁰ domains. In each case, we hypothesize that the catalytic domain downstream of the KS⁰ domain is unable to interact with the upstream ACP/PCP domain. The KS⁰ domain thus transfers the substrate from the upstream to the downstream carrier protein domain, which the neighbouring catalytic domain is able to interact with, allowing the reaction it catalyses to proceed. Domains (other than the KS⁰ domain) we propose are able to engage in productive protein-protein interactions are shown in the same colour.

In the broader context, KS⁰ domains are frequently found juxtaposed between up- and downstream ACP/PCP domains in several other types of trans-AT PKS architecture.¹⁶ Common examples include: ACP-KS⁰-PCP-Cy; PCP-KS⁰-ACP-KS; ACP-KS⁰-ACP-TE; ACP-KS⁰-OMT-ACP; ACP-KS⁰-DH-ACP; and ACP-KS⁰-ER-ACP (domain abbreviations not defined previously are as follows: Cy, heterocyclisation; KR, ketoreductase; DH, dehydratase and related domains, including enoyl isomerases and pyran synthases; OMT, O-methyltransferase; ER, enoyl reductase) (Fig. 6). In the all of these cases, it seems likely that the catalytic domains downstream of the KS⁰ domains cannot interact productively with the ACP/PCP domains upstream of them. The KS⁰ domains therefore transfer the acyl chains attached to these ACP/PCP domains to the downstream ACP/PCP domains, which can be recognized by the adjacent catalytic domains, allowing the reactions they catalyze to proceed. Modules that recruit trans-acting oxygenases, which modify the growing polyketide chain during its assembly, also employ KS⁰ domains.¹⁶ These appear to transfer the requisite intermediate onto an ACP domain that can be recognized by the oxygenase. Further experiments will be required to test these hypotheses, but one obvious consequence is that in each case the ACP/PCP domain downstream of the KS⁰ domain must avoid being loaded with a malonyl extender unit by the trans-acting AT. Whether this is controlled solely by protein-protein interactions, as demonstrated for selective ACP recognition by the trans-AT KirCII in kirromycin biosynthesis,²⁴ or more complex phenomena remains to be established. There is no selective pressure to maintain the HGTGT motif in a KS domain when it is positioned upstream of an ACP or PCP domain that cannot be malonylated. Thus KS⁰ domains probably evolved from elongating KS domains via the random accumulation of mutations in the HGTGT motif (Supplementary Fig. 2).

Although enacyloxin IIa has promising activity against A. baumannii,¹² a multi-drug resistant pathogen for which new antibiotics are urgently required, it is unlikely to find direct clinical application. Enacyloxin analogues with enhanced potency, reduced toxicity and greater chemical stability are therefore needed. The broad acyl acceptor tolerance of Bamb_5915 demonstrated by our work indicates that it may be possible to produce enacyloxin analogues with modifications to the DHCCA moiety via biosynthetic engineering. One attractive strategy for doing this involves feeding of DHCCA analogues to mutants of B. ambifaria blocked in DHCCA biosynthesis. Bamb_5915 is also able to accept a range of thioesters as acyl donors, suggesting that it may be possible to make analogues of enacyloxin with modifications to the polyketide chain via the manipulation of tailoring genes and other biosynthetic engineering strategies. Together, such approaches offer the potential to deliver a library of enacyloxin analogues that illuminate the structure-activity relationship of the natural product, which would be an important milestone in its development towards therapeutic application.

In conclusion, our results provide important insights into the mechanism and specificity of chain release from the enacyloxin polyketide synthase. They not only suggest a general role for KS⁰ domains in overcoming an intrinsic incompatibility between several types of catalytic domains and certain types of carrier protein domain in trans-AT PKSs, but also define plausible approaches for the production of enacyloxin analogues via biosynthetic engineering.

Methods

See the Supplementary Information for a description of the methods employed.

Supplementary Material

SUPPLEMENTARY INFORMATION is available in the online version of the paper.

Data availability

The genome sequence of B. ambifaria BCC0203 was deposited in the European Nucleotide Archive (Accession No. ERS782625). All other data are available from the authors upon request.