Malleilactone, a polyketide synthase-derived virulence factor encoded by the cryptic secondary metabolome of Burkholderia pseudomallei group pathogens

Abstract

Sequenced bacterial genomes are routinely found to contain gene clusters that are predicted to encode metabolites not seen in fermentation based studies. Pseudomallei group Burkholderia are emerging pathogens whose genomes are particularly rich in cryptic natural product biosynthetic gene clusters. We systemically probed the influence of the cryptic secondary metabolome on the virulence of these bacteria and found that the disruption of the MAL gene cluster, which is natively silent in laboratory fermentation experiments and conserved across this group of pathogens, attenuates virulence in animal models. Using a promoter exchange strategy to activate the MAL cluster we identified malleilactone, a polyketide synthase-derived cytotoxic siderophore encoded by this gene cluster. Small molecules targeting malleilactone biosynthesis either alone, or in conjunction with antibiotics, could prove useful as next-generation therapeutics for combating melioidosis and glanders.

Pseudomallei group pathogens including Burkholderia pseudomallei (BP), Burkholderia mallei (BM) and Burkholderia thailandensis (BT) are a closely related collection of Gram negative bacteria¹. BP is the causative agent of melioidosis in humans, BM is the causative agent of glanders in horses, and while BT is not generally considered to be a human pathogen, it is infectious in a number of model laboratory organisms. Melioidos is is endemic in parts of Southeast Asia and Northern Australia and is the third most frequent cause of mortality from infectious disease after HIV and tuberculosis in Northeast Thailand.² Cur rent therapies for this neglected disease are inadequate as mortality rates from melioidos is have been reported to approach 50% even with antibiotic treatment.² Additionally, these pathogens have been classified as potential bio-terrorism threats by the CDC, and in spite of the considerable attention paid to this group, their virulence determinants are still not well understood. Small molecules, including signaling molecules, siderophores, and toxins, are known to play important roles in both the establishment and propagation of bacterial infections.³ Biosynthetic gene clusters to which no small molecule has yet been assigned are frequently encountered in bacterial genome sequencing projects.⁴ These cryptic gene clusters represent the pool of biosynthetic pathways from which additional small molecule virulence factors might be characterized. In an attempt to better understand the virulence factors used by pseudomallei group pathogens, we have investigated the influence of the cryptic secondary metabolism on Burkholderia pathogenicity by systematically disrupting individual cryptic biosynthetic pathways in BT and assessing these mutants for changes in virulence. Disruption of the MAL gene cluster, a polyketide synthase based-cluster that is conserved across this group of pathogens, was found to attenuate virulence. Here we describe the structural and functional characterization of malleilactone, a MAL gene cluster encoded cytotoxic siderophore and virulence factor.

The pangenome of the pseudomallei group of pathogens⁵ is rich in polyketide synthase (PKS) and nonribosomal peptide synthetase (NRPS) gene clusters, two common biosynthetic systems that are easily identified bioinformatically in sequenced genomes. Many clusters are shared among these three pathogens and the majority of molecules encoded by these clusters remain cryptic in structure and/or biological function (Fig. 1a and Tables S1 and S2). BM and BP are classified as BSL3 level pathogens, and due to their potential use in biological warfare, there are significant regulatory issues governing their genetic manipulation. BT is strikingly similar genetically to BP,^5b,c possessing the majority of gene clusters found within BP (Table S1); however there are no similar restrictions on its genetic manipulation.

Figure 1.

(a) Venn diagram showing the relationship of number of NRPS/PKS gene clusters shared among pseudomallei group pathogens B. pseudomallei K96243, B. mallei ATCC 23344 and B. thailandensis E264 (BP, BM and BT respectively). (b) C. elegans after 24 h co-culture with: i. wild type BT (dead), ii. BT(ΔmalF), a PKS2 gene deletion mutant and iii. BT(ΔmalR), a transcription factor deletion mutant (40x magnification). (c) C. elegans survival after 24 h exposure to wild type BT, BT(ΔmalF), BT(ΔmalR) and BT(ΔmalR/malA), a malR-PKS1 double mutant. ***P<0.001, two-tailed t-test. (d) D. discoideum co-culture at 120 h with: i. wild type BT and ii. BT(ΔmalF). Arrows highlight D. discoideum aggregation and differentiation into fruiting bodies with MAL cluster disruption (white box - close-up of a fruiting body). (e) MAL clusters from BP, BM and BT with predicted functions for each MAL protein.

To address the potential role of secondary metabolites in Burkholderia pathogenesis, we created a collection of BT mutants wherein individual cryptic PKS/NRPS clusters were disrupted either by suppressing transcription through promoter exchange or by deleting one of the large modular PKS/NRPS genes within the gene cluster itself (Table S1). These strains were then assessed for virulence using a Caenorhabditis elegans nematode co-culture model, which has been successfully used to assess pathogenicity and identify virulence factors in bacterial pathogens.⁶ In this co-culture virulence model, both BT and BP elicit a rapid toxicity to C. elegans, which is thought to arise from specific in vivo host-bacteria interactions.⁷ In co-culture assays using wild-type BT worms display either no movement or lack any significant loco-motor behavior within 24 hours (Fig. 1b and c and Video S1). With one exception, C. elegans grown on cryptic gene cluster disruption strains were similarly dead or dying within 24 hours. The disruption, either by promoter exchange or gene deletion, of a cryptic gene cluster that we have termed the MAL cluster results in worms that uniformly continue to move and forage beyond 24 hours (Fig. 1b and c). Worms feeding upon these MAL knockout strains are larger than worms fed wild type BT, retain their active feeding and foraging behavior, and grow into adulthood as evidenced by the accumulation of eggs and hatched larvae on assay plates (Fig. 1b and c and Video S1).

To assess the generality of the importance of the MAL gene cluster to pseudomallei group virulence, we also tested MAL disruption mutants in an amoeba co-culture model.⁸ The response of the social amoeba Dictyostelium discoideum to pathogenic bacteria is hypothesized to closely mimic the response of the mammalian immune system to these organisms, as D. discoideum similarly feeds upon bacteria through phagocyticingestion.^8,9 In this model, D. discoideum cells foraging on pathogenic bacteria often fail to develop past their unicellular state into multicellular aggregates and fruiting bodies. In many instances, this inhibition has be observed to be dependent on the same set of virulence factors that are required for mammalian infections.^8,9 When co-cultured with wild-type BT amoebae die as unicellular organisms, preceding any observable social coordination. In identical experiments using MAL gene cluster knockout strains amoebae aggregated as expected and developed completely through the formation of terminal fruiting bodies (Fig. 1d). As observed in the C. elegans model, the MAL gene cluster appears to be a key component of BT's virulence toward D. discoideum.

The MAL gene cluster spans ∼35 kb and is predicted to contain 13 open reading frames, two of which (malA and malF) encode large modular PKSs (Fig. 1e). It is one of only three PKS/NRPS gene clusters that is conserved across the pseudomallei group, making it a conspicuous candidate for encoding a communal virulence factor. MAL gene clusters from all three pathogens have the same gene organization and content (Fig. 1e), with the exception of a hypothetical gene (malB_b) found in BM and BP but not in BT. Bioinformatics comparisons to previously characterized gene clusters revealed no close relatives that could provide insight into the structure or function of the MAL-encoded metabolite.

Culture broth extracts from MAL gene cluster knockout strains of BT did not show any obvious differences when compared to wild type control extracts (Fig. 2a) indicating that in the laboratory setting this gene cluster is silent. Many secondary metabolite gene clusters are thought to have remained cryptic even after extensive investigation of the organisms in which they reside, be cause these clusters are only expressed in response to specific environmental cues and therefore silent under standard laboratory fermentation conditions. In previous studies with BT, we found that silent gene clusters could be activated via transcriptional activation strategies.¹⁰ The most universal of such approaches is likely to be promoter exchange, where a native silent promoter governing gene cluster expression is replaced with a model inducible promoter. The MAL gene cluster is predicted to be com posed of a set of thirteen genes that are unidirectional and potentially organized into a single largeoperon (Fig. 1e). We reasoned that inducing the transcription of this operon could lead to the production of detectable levels of the product encoded by the MAL gene cluster. The promoter region in front of the first ORF in the MAL operon (malA) was therefore replaced by homologous recombination with the rhamnose inducible promoter, P_RhaB,¹¹ to yield BT:P_RhaB-MAL (Figs. 2b and Fig. S1). Ethyl acetate extracts from cultures of BT:P_RhaB-MAL induced with rhamnose contained one major metabolite that was not seen in extracts from similarly treated wild type cultures (1, Fig. 2c). This peak is not seen in extracts from wild-type BT cultures, nor is it seen in extracts from rhamnose induced cultures of BT:P_RhaB-MAL(ΔmalF), a BT:P_RhaB-MAL strain where the PKS2 gene (malfF) is disrupted, thus directly linking MAL operon induction to the production of 1.

Figure 2.

(a) HPLC traces (total diode array: 210-450 nm) of BT culture broth ethyl acetate extracts. i. wild type BT, ii. BT(ΔmalF), iii. BT(ΔmalR), iv. BT(ΔmalR/ΔmalA). (b) Promoter exchange strategy used to induce cryptic gene cluster expression. In brief, BT cells are transformed with a cassette designed to hybridize and recombine with the native promoter region resulting in the insertion of the rhamnose-inducible promoter P_RhaB directly upstream of the cryptic biosynthetic gene cluster. Small molecule production is then induced with the addition of 0.2% L-rhamnose to the culture media. (c) HPLC traces of culture broth extracts, showing that the induction of the MAL cluster leads to the production of the novel metabolite (malleilactone, 1). Cultures were grown with or without 0.2% L-rhamnose: v. wild type BT with rhamnose, vi. BT:P_RhaB-MAL, no rhamose, vii. BT:P_RhaB-MAL with rhamnose, viii. BT:P_RhaB-MAL(ΔmalF) with rhamnose.

Compound 1, which we have named malleilactone, was purified from ethyl acetate extracts of large-scale (6L) cultures of rhamnose-induced BT:P_RhaB-MAL using a modified Kupchan scheme followed by three rounds of silica gel chromatography. The structure of 1 was determined through a combination of HRMS (HRMS-TOF m/z: [M+H]⁺ calcd for C₁₈H₂₇O₄, 307.1910; found 307.1903), NMR and UV data (Fig. 3a). All 18 of the carbons predicted by HRMS to be present in 1 are seen in the ¹³C spectra. Empirical chemical shift data and ¹H-¹³C HMQC exper iments indicated the presence of 2 carbonyl/enol (194, 189 ppm), 5 olefin (140, 139, 137, 123, 100 ppm), 1 ester (174 ppm), 7 methylene and 3 methyl carbons (Fig. 3a). Three spin systems could be resolved in the ¹H–¹H COSY spectrum, two of which are joined by HMBC correlations into an unbroken saturated 7-carbon chain that is connected to the methyl substituted C-8/C-9 olefin through HMBC correlations. The trans geometry of this double bond is supported by the chemical shift of the C-8a tertiary methyl (13.9 ppm).¹² HMBC correlations from the C-9 methine proton and the C-8a methyl protons to C-7 connect this substructure to the predicted C-7 enol. At the other end of the molecule, the final 2-carbon ¹H-¹H COSY spin system can be connected by HMBC correlations to the C-3 carbonyl and C-4, and HMBC correlations from H-5 to C-3 and C-4 define the position of the C-4/C-5 olefin. These two substructures are linked through C-6 based on the large collection of HMBC correlations involving H-5 that is shown in Fig. 3a. A final three-bond correlation from H-5 to the C-6a carbonyl yields the complete 18 carbon skeleton of 1. Based on the un saturation index and empirical chemical shift data, the final unsaturation is satisfied by closing the γ-butyrolactone ring at C-4 to give 1. The presence of the conjugated unsaturated system seen in the final structure of 1 is supported by the UV_max observed at 373 nm.¹³

Figure 3.

(a) Key NMR arguments used to define the structure of malleilactone (1). (b) Retro-biosynthetic analysis of 1. (c) Biosynthetic proposal for 1. AT, acyltransferase; KS, ketosynthase; DH, dehydratase; KRⁱ, inactive ketoreductase; C, condensation domain; R, reductase domain; T, thiolation domain.

A detailed examination of 1 suggested that it likely arise from the condensation of two separate polyketide chains, one running from the C-16 methyl through the C-6a carbonyl and a second running from the C-1 methyl through the C-5 olefin (Fig. 3b). The formation of two such polyketide precursors and their subsequent condensation into 1 can be easily rationalized based on the collection of domains found in the MAL PKSs, MalA and MalF (Fig. 3b, c). In our biosynthetic proposal, one polyketide precursor is produced on MalA from propionic acid and a hydroxylmalonyl extender unit. The use of the rare oxidized extender unit, which is necessary for the subsequent lactone formation, is supported by the presence of an FkbH homolog (MalH) in the MAL cluster (Fig S5). FkbH proteins are predicted to load a glycolytic pathway intermediate onto a thiolation domain (T), which then serves as a substrate for the biosynthesis of oxidized extender units (Fig S5).¹⁴ The second of the two required polyketide pre cursors is predicted to be synthesized on MalF from a caprylic acid starter unit, followed by two rounds of elongation: one involving incorporation, reduction and dehydration of a methylmalonyl extender unit and the second involving incorporation of a non-reduced malonyl extender. While MalA contains an acyltransferase (AT) domain, MalF is predicted to be a trans AT system that use the AT activity of MalL. The formation of 1 from these two PKS precursors requires that they be linked through an ester bond and an additional C-C bond to form the central five membered lactone. There are now a handful of NRPS condensation (C) domains that are predicted to form esters instead of am ides, and MalF contains the conserved “HHXXXDD” active-site motif shared among this set of domains.¹⁵ We therefore propose that the diketide from MalA is transferred to the final T-domain in MalF and linked via a MalF C-domain catalyzed ester bond. MalA contains two thioesterase (TE) domains, both of which are predicted to be type II, or proof reading, TE domains and there fore not responsible for releasing the polyketide. Release of the polyketide from MalF is predicted to instead occur through the action of the terminal reductase (R) domain, which have increasingly been shown to catalyze the terminal release and cyclization of polyketides.¹⁶ Reductive cleavage of the condensed intermedi ate found on the T₄ domain of MalF, followed by intramolecular cyclization, provides an intermediate that at upon dehydration would yield 1. The best of our knowledge 1 has not been previously reported as a natural or synthetic compound.

In all three pseudomallei group pathogens, the MAL cluster resides directly adjacent to a predicted LuxR-type transcription factor, malR(btaR4¹⁷). LuxR homologs function as receptors for Nacylhomoserine lactones (AHLs) in quorum sensing circuits and are often responsible for regulating virulence gene expres-sion.^3a As seen in experiments with the MAL PKS deletion strain BT(ΔmalF), C. elegans, and D. discoideum co-cultured with malR-deletion strains show extended lifespans (Fig 1b and Supporting data), suggesting that MAL expression is mediated through MalR. A consensus Lux Rlike binding element (lux box) is present in the promoter region immediately upstream of malA (Fig. S5). In rare cases, LuxRs are ligand independent;¹⁸ however, in cases where a native ligand has been defined, they are con trolled by AHLs. Thus, malleilactone production, like the production of many virulence factors, appears to be governed by AHL-dependent quorum sensing.

Malleilactone was assayed for toxicity against a panel of cell types. It shows low micromolar cytotoxicity to human cell lines (e.g. IC₅₀ 19 μM) and in disk diffusion assays it inhibits the growth of Gram-positive bacteria at as low as 2.5 μg per disk (Table S4). When incorporated in growth media (>100 μg-mL^-1), malleilactone had no obvious deleterious effects upon C. elegans/E. coli co-cultures, nor did worms co-cultured with MAL mutants show increased distress upon malleilactone reintroduction, suggesting that its bioactivity might extend beyond localized cytotoxicity. The central core of malleilactone closely resembles a tetronic (or tetramic) acid substructure, and while reported bio-activities for tetronic acids vary widely, they are predicted to share the common phenomenon of chelating cations, in particular iron.¹⁹ Using a chromogenic iron chelation assay, we observed iron-binding activity with malleilactone (Fig. S6). Malleilactone may serve therefore both as an ironchelator to assist with iron acquisition as well as a toxin when secreted at the site of infection. It is also possible that in vivo malleilactone serves an additional undetermined role beyond either of these functions, as so far the MAL cluster has proven recalcitrant to disruption in BP and BM.²⁰ Given that malleilactone contains a γ-butyrolactone core, a structural motif found in a variety of bacterial signaling molecules including AHLs, it is also plausible that malleilactone is a signaling molecule capable of activating downstream virulence factors.

As iron is tightly sequestered in eukaryotes, bacterial pathogens have in many cases evolved ironchelation systems in order to survive in these low iron environments. Two additional PKS/NRPS gene clusters predicted to encode siderophores (malleobactin/ornibactin and pyochelin) are shared between BP and BT genomes (BM lacks pyochelin). It has been theorized that the production of multiple siderophores with different structures and polarities serves to benefit pathogens by providing alternate routes for securing iron and possibly circumventing host defense mechanisms that neutralize conserved microbial siderophores.²¹ The therapeutic importance of this class of secondary metabolites as potential anti-infective targets is highlighted by the use of the siderophore (mycobactin) biosynthesis inhibitor, p-aminosalicylate, as an anti-tuberculosis drug.²² Small molecules targeting malleilactone biosynthesis could similarly prove useful as next-generation therapeutics for combating both melioidosis and glanders.

The systematic gene cluster knockout/promoter exchange strategy we used to identify malleilactone is easily generalizable, and should permit the functional characterization of small molecule virulence factors encoded by silent gene clusters found in the genomes of a diverse collection of additional bacterial pathogens.