Enzymatic Synthesis of Diverse Heterocycles by a Noncanonical Nonribosomal Peptide Synthetase

Abstract

Nonribosomal peptide synthetases (NRPSs) are typically multimodular enzymes that assemble amino acids or carboxylic acids into complex natural products. Here, we characterize a monomodular NRPS, PvfC, encoded by the Pseudomonas virulence factor (pvf) gene cluster that is essential for virulence and signaling in different bacterial species. PvfC exhibits a unique adenylation-thiolation-reductase (ATR) domain architecture that is understudied in bacteria. We show that the activity of PvfC is essential in the production of seven leucine-derived heterocyclic natural products, including two pyrazines, a pyrazinone, and a rare disubstituted imidazole, as well as three pyrazine N-oxides that require an additional N-oxygenation step. Mechanistic studies reveal that PvfC, without a canonical peptide-forming domain, makes a dipeptide aldehyde intermediate en route to both the pyrazinone and imidazole. Our work identifies a novel biosynthetic route for the production of pyrazinones, an emerging class of signaling molecules and virulence factors. Our discovery also showcases the ability of monomodular NRPSs to generate amino acid- and dipeptide-aldehydes that lead to diverse natural products. The diversity-prone biosynthesis by the pvf-encoded enzymes sets the stage for further understanding the functions of pvf in bacterial cell-to-cell signaling.

Graphical Abstract

INTRODUCTION

Nonribosomal peptide synthetases (NRPSs) are a major family of biosynthetic enzymes responsible for making a wide range of natural products with important biological activities and pharmaceutical applications, including antibiotics vancomycin,¹ chloramphenicol,^2,3 and penicillin,⁴ as well as siderophores enterobactin⁵ and pyochelin.⁶ NRPSs are modular assembly lines consisting of repeating sets of domains that organize into modules, and each module conducts reactions to incorporate an amino acid into the growing peptide.^7,8 A canonical module includes an adenylation (A) domain that activates an amino acid,^9,10 a thiolation (T) domain that tethers the amino acid as a thioester,¹¹ and a condensation (C) domain that forms a peptide bond between two T domain-tethered amino acids.¹² Once an NRPS fully assembles the peptide, it cleaves the T domain-linked thioester using a terminal thioesterase (TE) domain,^13,14 releasing the nonribosomal peptide from the assembly line. A growing number of NRPSs are found to employ a terminal reductase (R) domain instead of a TE domain to couple the thioester cleavage with a two- or four-electron reduction, releasing the nonribosomal peptide with a terminal aldehyde or alcohol, respectively, instead of a carboxylic acid.^15-18 The reductive release provides additional functional groups on nonribosomal peptides for late-stage decoration, such as transamination to form an amine and cyclization to form an imine or a hemiaminal.¹⁸

A unique group of R domain-containing enzymes harbor a single module that exhibits an adenylation-thiolation-reductase (ATR) domain architecture without any C or TE domains. These ATR enzymes play essential roles in the biosynthesis of both primary metabolites and natural products by activating, loading, and reducing amino acids or carboxylic acids as aldehydes. For example, the yeast ATR enzyme, Lys2, activates 2-amino-6-oxohexanoate and generates an α-aminoadipate semialdehyde in lysine biosynthesis.¹⁵ Most ATR enzymes in natural product biosynthesis are also of fungal origin, such as HqlA in herquline A biosynthesis.¹⁹ These fungal enzymes release amino acids as amino aldehydes that dimerize to a dihydropyrazine, which is further reduced to a piperazine in herquline A¹⁹ or oxidized to pyrazines and pyrazine derivatives.^19,20 CmlP in chloramphenicol biosynthesis, one of the few bacterial ATRs, activates and loads p-amino benzoic acid that is hydroxylated on the T domain, followed by the proposed reductive release by the R domain post hydroxylation.^2,3 A few ATR variants also exist including the ATRR in fungal choline biosynthesis, where two consecutive R domains convert glycine betaine to the alcohol choline,²¹ and the carboxylic acid reductases, which share an adenylationthiolation-reductase architecture, but adenylate carboxylic acids instead of amino acids.²² Here, we report in vitro reconstitution of all three domains of a bacterial ATR enzyme, PvfC_Pf0-1, encoded by the Pseudomonas virulence factor (pvf; Figure 1AC),²³ which enabled us to identify several compounds, including pyrazines, a pyrazinone, and an imidazole.

Figure 1.

Pseudomonas virulence factor (pvf) cluster contains the gene encoding the ATR enzyme PvfC. (A) pvf gene clusters from P. entomophila L48 (top) and P. fluorescens Pf0-1 (pvf_Pf0-1, bottom). (B) Structures of the pvf-dependent, valine-derived (d)PNOs identified from P. entomophila L48. (C) PvfC_Pf0-1 from P. fluorescens Pf0-1 activates leucine. PvfC from P. entomophila L48 activates valine.

The pvf cluster is widely conserved in Pseudomonas and other proteobacteria^23,24 and is important for signaling, virulence, and biocontrol.^24-27 In addition to the ATR enzyme PvfC, pvf also encodes a diiron N-oxygenase (PvfB) and two proteins of unknown functions (PvfA and PvfD; Figure 1A). Enzymes encoded by pvf produce extracellular signaling molecules (PVF autoinducers) that regulate the proteome and metabolome of the model insect pathogen Pseudomonas entomophila L48.²⁸ While the structures of the PVF autoinducers are undetermined, the pvf cluster is already responsible for the biosynthesis of a variety of natural products in which PvfC plays a key role. PvfC from P. entomophila L48 activates and incorporates l-Val (Figure 1BC) into the-(dihydro)pyrazine N-oxides ((d)PNOs),²⁹ while the homologue, HamD, from Burkholderia cenocepacia H111 incorporates l-Val into the diazeniumdiolates, valdiazen and fragin.³⁰ In response to antibiotic stress, B. cenocepacia overproduces fragin, as well as other metabolites, some of which likely correspond to the (d)PNOs, suggesting a role for these compounds in bacterial stress response.³¹ Previously, we showed that the identity of amino acid activated by PvfC or homologues correlates with pvf signaling activity.^23,29 While PvfC from P. entomophila is specific for l-Val, the PvfC homologue from Pseudomonas fluorescens Pf0-1 (PvfC_Pf0-1) selectively activates l-Leu (Figure 1C).²³ Here, we used the pvf homologue from P. fluorescens Pf0-1 (pvf_Pf0-1) to identify l-Leu-derived pvf-dependent molecules. We show that pvf_Pf0-1-encoded enzymes produce an unexpectedly large number of leucine-derived molecules including l-Leu-(dihydro)pyrazine N-oxides (Leu-(d)PNOs), the pyrazinone flavacol, a fungal natural product,^32-35 a rare imidazole that we named leucinazole, as well as 2,6-diisobutylpyrazine (PZ-1) and 2,5-diisobutylpyrazine (PZ-2). By reconstituting the activity of PvfC_Pf0-1 in vitro, we show that PvfC_Pf0-1 is the only enzyme required to produce flavacol, leucinazole, PZ-1, and PZ-2. Our findings reveal unexpected chemistry catalyzed by ATR enzymes using amino aldehyde intermediates.

RESULTS AND DISCUSSION

Identification of Metabolites Produced by pvf_Pf0-1.

To identify molecules produced by enzymes encoded in pvf_Pf0-1, we overexpressed the cluster in two heterologous hosts, P. aeruginosa PAO1 and E. coli BAP1 (Figures 2AB and S1). As pvfA is not strictly necessary for the production of (d)PNOs or the signaling activity,^26,29 we conducted metabolomics on PAO1 expressing the partial cluster lacking pvfA (pvfBCD_Pf0-1-PAO1). Using liquid chromatography-coupled high-resolution mass spectrometry (LC-HRMS), we detected a number of species uniquely produced by pvfBCD_Pf0-1-PAO1 and focused on seven compounds for structural characterization. The masses and molecular formulae of three compounds (1, 2, 4) correspond to the leucine-derived (d)PNOs (1, [M + H]⁺ = 227.1764, C₁₂H₂₃N₂O₂⁺; retention time (RT) = 8.5 min, 2, [M + H]⁺ = 225.1604, C₁₂H₂₁N₂O₂⁺, RT = 9.9 min; 4, C₁₂H₂₁N₂O⁺, [M + H]⁺ = 209.1639, RT = 11.5 min; Figures 2A, S1, and S4). Four compounds, 3, 5, 6, and 7, were also produced by E. coli BAP1 overexpressing pvf_Pf0-1 (Figures 2AB, S1, and S2). Among these compounds, 3 and 5 exhibit identical masses as 4 ([M + H]⁺ = 209.1648, C₁₂H₂₁N₂O⁺) but elute at different retention times (10.2 and 11.6 min, respectively). Similarly, compounds 6 and 7 exhibit identical masses as each other ([M + H]⁺ = 193.1699, C₁₂H₂₁N₂⁺) and elute at retention times of 11.9 and 12.3 min, respectively, suggesting that pvf_Pf0-1 produces two sets of isomeric compounds: 3/4/5 and 6/7 (Figures S1-S3, Extended Data Set).

Figure 2.

pvf_Pf0-1 cluster encodes proteins that produce seven different heterocyclic molecules. (A) Comparative metabolomic analyses of PAO1 that contains pvfBCD_Pf0-1-pPSV35 (top) and pPSV35 empty vectors (bottom). (B) Comparative metabolomic analyses of BAp1 that contains pvfABCD_Pf0-1-pPSV35 (top) and pPSV35 empty vectors (bottom). HPLC traces at 270 nm UV absorbance in milli absorbance unit (mAU) are shown in panels (A) and (B). (C) Key COSY (bolded bonds) and HMBC correlations (blue arrows) of isolated Leu-dPNO (1), Leu-PNO A (2), leucinazole (3), Leu-PNO B (4), flavacol (5), PZ-1 (6), and PZ-2 (7). See also Tables S1-S7 and Figures S1, S2, and S6.

Isolation of Leu-(dihydro)pyrazine N-Oxides (1, 2, 4).

To identify the structures of compounds 1, 2, and 4, we cultured pvfBCD_Pf0-1-PAO1 on a 5 L scale, isolated 1, 2, and 4 using high-performance liquid chromatography (HPLC), and analyzed the isolated molecules by UV–vis spectroscopy and one- and two-dimensional NMR. All three compounds exhibit proton shifts and COSY correlations that are consistent with two isobutyl groups (Tables S1-S3, Extended Data Set). Otherwise, the UV and NMR spectra of 1, 2, and 4 are similar to the valine-derived (d)PNO counterparts produced by P. entomophila,²⁹ supporting the structures of 2,5-diisobutyl-2,3-dihydropyrazine N,N′-dioxide (Leu-dPNO), 2,5-diisobutylpyrazine N,N′-dioxide (Leu-PNO A), and 2,5-diisobutylpyrazine N-oxide (Leu-PNO B), respectively (Figures 2C and S1 and Tables S1-S3, Extended Data Set). The NMR spectra of 2 and 4 are also consistent with those of synthetic Leu-PNO A and Leu-PNO B, respectively (Tables S2 and S3).^36,37

Isolation and Structural Characterization of Flavacol (5) and Leucinazole (3).

Compounds 3 and 5 were isolated from a 4 L culture of E. coli BAP1 that expresses pvfABCD_Pf0-1 (Figure 2B; 1.8 and 1.0 mg per 1 liter of culture, respectively) and characterized by COSY, HMBC, and HSQC-TOCSY NMR experiments. The splitting patterns and correlations of both compounds support the presence of isobutyl substituents around a heterocyclic core (Tables S4 and S5, Extended Data Set). A single proton shift was observed in the aromatic range for 3 and 5 (Tables S4 and S5, Extended Data Set), in contrast to the two aromatic protons in 4 (Leu-PNO B). Compounds 3 and 5 do not interconvert under ambient conditions by UV–vis spectroscopy (λ_max (flavacol) = 228, 325 nm, λ_max (leucinazole) = 299 nm; Figures S1 and S4) and exhibit different tandem MS patterns (Figure S5), suggesting that they are distinct species (Extended Data Set).

NMR analysis of 3 and 5 shows key distinctions between the two isomers. The ¹H NMR spectrum in CDCl₃ reveals broad singlets at δ9.87 for 3 and at δ11.50 for 5, suggesting different chemical environments of the exchangeable protons (Extended Data Set).³⁵ The ¹H NMR spectrum of compound 5 matches that of the pyrazinone flavacol.^33-35 In addition, the retention time and MS fragmentation pattern of 5 are identical to a flavacol standard that we chemically synthesized by adapting a reported procedure (Scheme S1 and Figure S6, Extended Data Set).³⁸ Therefore, we assigned 5 to be 3,6-diisobutyl-pyrazin-2(1H)-one or flavacol (Figure 2C and Table S5). Flavacol has been identified from fungal extracts in several reports and has been assigned by different authors as either the pyrazinone or its enol tautomer.^32-35 X-ray crystallography allowed unambiguous assignment of flavacol as a pyrazinone.³⁴ Compound 5 and all other isolated flavacol species exhibit the same UV spectrum (Figure S1; λ_max = 228, 325 nm) typical of a pyrazinone core,³⁹ further confirming the structural assignment of 5 and suggesting that all reported flavacol species are likely based on pyrazinone cores.

The structure of 3 was more difficult to identify. A useful spectroscopic clue was a ¹³C NMR signal at δ186.6, which strongly suggests the presence of a ketone, as opposed to the amide or imine in flavacol. Reasoning that a ketone in a compound isomeric with flavacol would likely be placed outside of the heterocyclic ring, we proposed a structure containing an isobutyl ketone attached to a 4(5)-isobutyl-1H-imidazole, which is supported by HSQC-TOCSY analysis (Extended Data Set). Thus, we assigned 3 as 1-(4(5)-isobutyl-1H-imidazol-2-yl)-3-methylbutan-1-one and named it leucinazole. We devised a route to synthesize an authentic sample of the proposed structure of leucinazole through a 4(5)-isobutyl-1H-imidazole intermediate (Scheme S2) based on relevant literature procedures^40-42 and obtained a compound that is identical to isolated leucinazole, confirming the structural assignment (Figure S6 and Table S5, Extended Data Set).

Isolation of 2,6-Diisobutylpyrazine (PZ-1, 6) and 2,5-Diisobutylpyrazine (PZ-2, 7).

The second set of isomers, 6 and 7, were isolated from a 6 L culture of E. coli BAP1 that expresses pvfABCD_Pf0-1 (Figure 2B; 0.2 and 1.4 mg per 1 liter of culture, respectively). They exhibit almost identical ¹H shifts and correlations (Tables S6 and S7, Extended Data Set). HMBC and COSY correlations indicate that both 6 and 7 contain isobutyl substituents around a pyrazine core. Their UV spectra also match that of pyrazines (Figure S1).⁴³ To distinguish and elucidate the structures of 6 and 7, we prepared synthetic standards of 2,3-, 2,5-, and 2,6-diisobutylpyrazines and subjected them to LC and tandem MS analyses alongside 6 and 7 (Schemes S3-5, Tables S6 and S7, and Figure S7, Extended Data Set). Compound 6 exhibits identical retention time and MS fragmentation pattern to those of 2,6-diisobutylpyrazine, while 7 shows similar retention time and fragmentation to those of 2,5-diisobutylpyrazine (Figures S6 and S7). Coinjection of 6 or 7 with each synthetic standard confirmed that 6 coelutes with 2,6-diisobutylpyrazine and 7 with 2,5-diisobutylpyrazine (Figure S6). The synthetic standard of 2,3-diisobutylpyrazine elutes in between 6 and 7 (Figure S6) and exhibits a significantly different fragmentation pattern from 6 or 7 (Figure S7). These LC and tandem MS results support that 6 and 7 are 2,6- and 2,5-diisobutylpyrazine, respectively. Further, the NMR spectra of 6 and 7 show very similar splitting patterns and correlations with each other and the synthetic standards; thus, we mixed 6 with 2,6-diisobutylpyrazine and 7 with 2,5-diisobutylpyrazine in a 1:1 ratio for ¹H NMR experiments. Only one set of resonances were observed for each mixture (Extended Data Set), unequivocally demonstrating that 6 is 2,6-diisobutylpyrazine (PZ-1) and 7 is 2,5-diisobutylpyrazine (PZ-2).

Identification of pvf_Pf0-1 Genes Necessary for the Production of 1–7.

We compared the metabolomic profiles of PAO1 expressing the full (pvfABCD_Pf0-1) or partial clusters of pvf_Pf0-1 (pvfBCD_Pf0-1, pvfACD_Pf0-1, and pvfABC_Pf0-1) by LC-HRMS. The Leu-(d)PNOs, 1, 2, and 4, were produced by PAO1 that expresses pvfABCD_Pf0-1, pvfBCD_Pf0-1, and pvfABC_Pf0-1, but not by PAO1 that expresses pvfACD_Pf0-1, suggesting that both pvfC_Pf0-1 and pvfB_Pf0-1 are necessary for the production of the Leu-(d)PNOs, consistent with our previous report on the production of the Val-(d)PNOs.²⁹ Compounds 3 and 5–7 were produced by all four strains, indicating that only pvfC_Pf0-1 is required for the production of these metabolites (Figure S1). In contrast, E. coli BAP1 that expresses the full or partial clusters of pvf_Pf0-1 only produced 3 and 5–7 but not the N-oxygenated Leu-(d)PNOs. We found that PvfB_Pf0-1 was insoluble when produced in E. coli, supporting that pvfC_Pf0-1 is responsible for the production of the imidazole leucinazole (3), the pyrazinone flavacol (5), and the 2,6- and 2,5-diisobutylpyrazines (6 and 7, respectively).

In Vitro Reconstitution of PvfC_Pf0-1 and Production of 3 and 5–7.

We studied the biosynthesis of leucinazole (3), flavacol (5), PZ-1 (6), and PZ-2 (7) by reconstituting PvfC_Pf0-1 in vitro. Previously, we showed that the A domain of PvfC_Pf0-1 is specific for leucine.²³ Here, we show that l-Leu is activated and loaded on the phosphopantetheine (PPant) arm of holo-PvfC_Pf0-1-T in trans using full-length PvfC_Pf0-1 (Figures 3AB and S8). LC-HRMS of intact proteins as described⁴⁴ reveals a 113 Da mass increase from holo-PvfC_Pf0-1-T, corresponding to the loading of leucine on the thiolation domain (Figures 3AB and S8). The activity of the reductase domain of full-length PvfC_Pf0-1 was monitored using UV–vis spectroscopy based on the absorption of NAD(P)H at 340 nm. NADPH but not NADH was consumed in the presence of PvfC_Pf0-1, ATP, l-Leu, the promiscuous phosphopantetheinyl transferase Sfp,⁴⁵ and coenzyme A (CoA) (Figures 3CD and S8), suggesting that the R domain of PvfC_Pf0-1 is active and specific for NADPH. NADPH consumption by PvfC_Pf0-1 also depends on the presence of l-Leu (Figure S8), suggesting that the reduction by PvfC_Pf0-1 is substrate-dependent. Together, these results suggest that PvfC_Pf0-1 activates (A domain) and loads (T-domain) leucine and then releases a modified leucine (R domain).

Figure 3.

PvfC_Pf0-1 activates and loads leucine and releases a modified leucine. (A) PvfC_Pf0-1 activates and loads leucine to PvfC_Pf0-1-T in trans. (B) Deconvoluted mass spectra for apo-PvfC_Pf0-1-T (left), holo-PvfC_Pf0-1-T (middle), and PvfC_Pf0-1-T-Leu (right). amu, atomic mass unit. (C) Scheme for PvfC_Pf0-1-R domain-catalyzed reductive release of leucine aldehyde. (D) PvfC_Pf0-1 consumes NADPH over time in a reaction including all substrates and cofactors. The red line is the no leucine control reaction at 45 min. See also Figure S3.

Combining the reaction conditions for each PvfC_Pf0-1 domain, we produced leucinazole (3), flavacol (5), PZ-1 (6), and PZ-2 (7) in vitro using full-length PvfC_Pf0-1. PvfC_Pf0-1 was incubated with ATP, l-Leu, Sfp, CoA, MgCl₂, and NADPH in one pot. LC-HRMS analysis of the reaction reveals the production of two compounds with masses and retention times that correspond to isolated leucinazole and flavacol and two that correspond to isolated PZ-1 and PZ-2 (Figures 4AB and S9-S11). This observation provides biochemical evidence to support that PvfC_Pf0-1 is necessary and sufficient to produce flavacol, leucinazole, PZ-1, and PZ-2 (Figure S1). All four compounds are only produced in the presence of NADPH, suggesting that the reductive release is required for their production (Figures 4AB, S10, and S11). Using ¹⁵N-Leu as a substrate results in a 2 Da mass increase in each compound, indicating that two leucine molecules are incorporated into flavacol, leucinazole, PZ-1, and PZ-2 (Figure S8).

Figure 4.

PvfC_Pf0-1 produces flavacol, leucinazole, PZ-1, and PZ-2 in vitro via a dipeptide aldehyde intermediate. (A) Structures and MS analyses of leucinazole (3) and flavacol (5). Extracted ion chromatograms (EICs) of leucinazole (purple) and flavacol (blue) ([M + H]⁺ = 209.1648) isolated from E. coli cultures (i and ii), produced in an enzymatic assay containing PvfC_Pf0-1, ATP, leucine, and NADPH (iii), and not produced in a control reaction excluding NADPH (iv). (B) Structures and MS analyses of PZ-1 (6) and PZ-2 (7). EICs of PZ-1 (yellow) and PZ-2 (green) ([M + H]⁺ = 193.1699) isolated from E. coli cultures (i and ii), produced in an enzymatic assay containing PvfC_Pf0-1, ATP, leucine, and NADPH (iii), and not produced in a control reaction excluding NADPH (iv). (C) Pentafluorophenylhydrazine (PFPH) derivatization of the Leu-Leu dipeptide aldehyde intermediate, 8, yielding 9. (D) EICs of 8 (green, [M + H]⁺ = 229.1911) produced in a reaction containing PvfC_Pf0-1, ATP, Leu, and NADPH without (top) or with PFPH added (bottom). (E) EICs of 9 (pink, [M + H − H₂O]⁺ = 391.1916) produced in a reaction containing PvfC_Pf0-1, ATP, Leu, and NADPH without (top) or with PFPH added (bottom). Reactions in panels A, D, and E were separated on a C18 column, and the reactions in panel B were separated on a Hypercarb column.

Identification of Dipeptide Aldehyde Intermediate and Elucidation of Flavacol Biosynthesis.

Pyrazinone natural products are usually biosynthesized by dimodular NRPS pathways, such as the Staphylococcus aureus NRPSs in leuvalin, tyrvalin, and phevalin synthesis,⁴⁶ and the NRPSs found in gut microbe Ruminococcus bromi.⁴⁷ In these systems, a condensation (C) domain in the dimodular NRPS catalyzes peptide bond formation and the terminal R domain catalyzes the reductive release of the dipeptide aldehyde that is cyclized and oxidized to the pyrazinone core.^39,46,48,49 In contrast, PvfC_Pf0-1 does not contain any C domains; thus, the biosynthesis of flavacol and leucinazole likely undergoes an alternative route. We hypothesized that flavacol and leucinazole production involves a key dipeptide aldehyde (8) intermediate formed independently of a C domain (Figure 4C). Indeed, LC-HRMS analyses reveal masses that correspond to a Leu-aldehyde and a Leu-Leu-aldehyde (8) in the PvfC_Pf0-1 reaction mixture that produced flavacol, leucinazole, PZ-1, and PZ-2 (Figures 4C-E, S12, and S13). Derivatization of the reaction mixture with pentafluorophenylhydrazine (PFPH), which readily reacts with aldehydes,^50,51 resulted in masses that correspond to the derivatized product of the Leu-Leu-aldehyde (8) (Figures 4C-E and S13), as well as that of Leu-aldehyde (Figure S12). Further, the addition of PFPH to the reaction at the same time as NADPH led to a decreased production of flavacol and leucinazole, providing evidence for the intermediacy of the Leu-Leu-aldehyde (8) to flavacol and leucinazole (Figure S12). The identification of a single Leu-aldehyde suggests that the dipeptide product may be formed off the T domain between Leu-aldehyde and an activated form of leucine.

To elucidate the mechanism of peptide bond formation by PvfC_Pf0-1 without a C domain, we sought to identify components of the PvfC_Pf0-1 reaction necessary for peptide bond formation. The reaction by PvfC_Pf0-1 also produces a Leu-Leu dipeptide side product independent of NAD(P)H (Figure 5), suggesting that the reductive release by the R domain is not required for peptide bond formation. The Leu-Leu dipeptide production was abolished when Sfp and CoA were excluded from the one-pot reaction, which contains apo-PvfC_Pf0-1 and adenylated l-Leu formed by the A domain (Figure 5), indicating that peptide bond formation requires the loading of leucine on the T domain and does not involve leucyladenylate. We also investigated the possibility of the Leu–Leu dipeptide being formed while attached to the T domain. MS fragmentation of the reaction that contains leucine-loaded PvfC_Pf0-1-T yields PPant ejection ions containing a single Leu but not Leu-Leu dipeptide (Figure S8), suggesting that peptide bond formation does not occur on the T domain. Thus, the biosynthesis of flavacol likely involves the reductive release of a Leu-aldehyde, the amine of which conducts the nucleophilic attack of another PvfC_Pf0-1-Leu thioester, generating the Leu-Leu-aldehyde intermediate (8) (Figure 6).

Figure 5.

Formation of the Leu-Leu dipeptide bond requires the holo-T-domain of PvfC_Pf0-1. (A) Proposed biosynthesis of Leu-Leu dipeptide. (B) Structure of Leu-Leu dipeptide and mass spectrum in a reaction with PvfC_Pf0-1, ATP, Sfp, CoA, MgCl₂, NADPH, and l-Leu. (C) EICs of Leu-Leu dipeptide ([M + H]⁺ = 245.1860) in the following reactions (top to bottom): full reaction with NADPH, reaction excluding NAD(P)H, reaction excluding l-Leu and NAD(P)H, and reaction excluding Sfp and CoA.

Figure 6.

Proposed biosynthesis of molecules 1–7. (A) Proposed biosynthesis of leucinazole (3), flavacol (5), and PZ-2 (7). Compounds 3 and 5 share intermediates 8 and 10. (B) Proposed biosynthesis of PZ-1 (6). (C) Proposed biosynthesis of Leu-(d)PNOs, 1, 2, and 4.

To gain additional insights into the mechanism of flavacol and leucinazole formation, we analyzed the pH profile for the production of these compounds by PvfC_Pf0-1 and examined potential conditions under which flavacol might convert to leucinazole. We examined the pH dependence of the production of leucinazole and flavacol in vitro by PvfC_Pf0-1 across a pH range of 6.0–9.0. Production of both leucinazole and flavacol increased with increasing pH, but their abundance ratio remained the same (Figure S4). Incubation of synthetic flavacol with PvfC_Pf0-1, ATP, l-Leu, Sfp, CoA, MgCl₂, and NADPH results in no leucinazole production, suggesting that flavacol is not a biosynthetic intermediate for PvfC_Pf0-1 to leucinazole (Figure S4). Neither acid nor base treatment could interconvert the two compounds (Figure S14), suggesting that both leucinazole and flavacol are produced enzymatically and rather than one being a direct precursor to the other, and that they share a common leucine-derived biosynthetic precursor.

pH Dependence of PZ-1 and PZ-2 Production and Proposed Biosynthesis.

The production of 2,5-disubstituted pyrazines has been reported for a few ATR enzymes, both fungal and bacterial,^19,20,52 which likely occurs through dimerization of two amino aldehydes. Similarly, the 2,5-disubstituted PZ-2 is likely formed between two Leu-aldehydes (Figure 6). In contrast, the biosynthesis of 2,6-disubstituted pyrazines such as PZ-1 has been little explored. Based on the position of isobutyl side chains and incorporation of two leucines to the product (Figure S8), an alkyl shift may occur to yield the 2,6-diisobutyl product (Figure 6). We examined the pH dependence on the production of PZ-1 and PZ-2 in vitro by PvfC_Pf0-1. Across a pH range of 6.0–9.0, the ratio of PZ-1 to PZ-2 increased at a higher pH (Figure S4), suggesting that the production of PZ-1 is more favorable under basic conditions and may require a key deprotonation step.

PvfC_Pf0-1 Homologues Produce Valine Derivatives of Compounds 3 and 5–7.

We showed that PvfC homologues, such as the one from P. entomophila L48 (PvfC_L48), activate valine.^23,29 Thus, we explored the possibility for PvfC_L48 to produce valine-derived analogues for 3 and 5–7. The one-pot reaction of PvfC_L48 was conducted using l-Val as a substrate and analyzed by LC-HRMS. We identified four new valine-derived compounds whose production depends on PvfC_L48 (Figure S15), including compounds 11 and 12 ([M + H]⁺ = 181.1335) and 13 and 14 ([M + H]⁺ =165.1386). MS/MS analysis of 11 and 12 shows that the two compounds exhibit a similar fragmentation pattern to the leucine-derived flavacol (5) and leucinazole (3), respectively, suggesting that these species are derivatives of 5 and 3 with isopropyl side chains (Figure S16). MS/MS analysis of 14 shows a similar fragmentation pattern to that of the leucine-derived pyrazines PZ-1 (6) and PZ-2 (7), suggesting that 14 is the isopropyl derivative of 6 or 7. Finally, valine dipeptide aldehyde was only detected within these reactions in the presence of NADPH (Figures S17 and S18), suggesting that the R domain of PvfC_L48 is responsible for the reductive release to yield the aldehyde intermediate en route to 11–14. Thus, the biosynthesis of these compounds uses similar mechanisms as leucinazole, flavacol, PZ-1, and PZ-2.

DISCUSSION

We identified seven leucine-derived molecules (Leu-PNO A, Leu-PNO B, Leu-dPNO, flavacol, leucinazole, PZ-1, and PZ-2) produced by pvf_Pf0-1-encoded enzymes. We showed that a single ATR enzyme PvfC_Pf0-1 produces four different molecules including the pyrazinone flavacol, the imidazole leucinazole, and both 2,6- and 2,5-disubstituted pyrazines. A number of pyrazinone natural products have been isolated, including flavacol from fungi,^32,35 tyrvalin and phevalin (or aureusimine A and B) and analogues from S. aureus,^39,53 pyrazinones from the gut microbiota,⁴⁶ autoinducer-3 from E. coli,⁵⁴ and DPO from Vibrio cholerae (first represented as 3,5-dimethylpyrazin-2-ol and later revised to a pyrazinone).^54,55 Natural products with a 2,4(5)-disubstituted imidazole core are rare, however, with only a few reports on the topsentins isolated from a marine sponge, which contain bisindole substituents.^56,57 The diisobutyl-substituted leucinazole has not been isolated or synthesized to the best of our knowledge.

Two biosynthetic routes have been reported for the disubstituted pyrazinones, whereas the biosynthetic origins of the topsentins are unknown. By in vitro characterization of the bacterial ATR enzyme PvfC_Pf0-1, we identified a new biosynthetic route to the pyrazinones and uncovered a possible biosynthetic origin for the disubstituted imidazole. The most common route for pyrazinone biosynthesis depends on a dimodular NPRS including a C domain and a terminal R domain. The C domain condenses two amino acids loaded on each module into an NRPS-attached dipeptide, which is released by the R domain as a dipeptide aldehyde that is then cyclized and oxidized to a 3,6-disubstituted pyrazinone.^39,48,49 More recently, an NRPS-independent route was reported for autoinducer-3 and its 3,5-disubstituted pyrazinone analogues. This route involves a threonine dehydrogenase that catalyzes the formation of aminoacetone or 1-amino-3-methylbutan-2-one. Spontaneous condensation of two aminoacetones followed by tautomerization and oxygenation was proposed to yield 3,6-dimethyl pyrazinone (autoinducer-3).⁵⁴ Condensation of the aminoacetone or 1-amino-3-methylbutan-2-one with an activated amino acid forms a linear dipeptide ketone intermediate that is then cyclized and oxidized to a 3,5-disubstituted pyrazinone.⁵⁴ Our work revealed a third and novel biosynthetic route where a monomodular ATR enzyme bypasses the need for a C domain by generating an amino aldehyde that directly condenses with a T domain-tethered leucine to form a dipeptide aldehyde en route to the pyrazinone (Figure 6). We show that the biosynthesis of the imidazole leucinazole by PvfC_Pf0-1 also involves a Leu-Leu dipeptide aldehyde intermediate (8; Figures 4 and S12). The biosynthesis of leucinazole and flavacol likely diverges on subsequent steps to form the imidazole or pyrazinone core. The biosynthesis of leucinazole may involve hydration of a dihydropyrazinone followed by ring contraction (Figure 6); mechanistic analysis is underway.

Our work also demonstrated the ability of an ATR enzyme to biosynthesize diverse molecules by generating amino aldehydes. In addition to flavacol and leucinazole, PvfC_Pf0-1 also forms 2,5- and 2,6-diisobutylpyrazines by reductively releasing a Leu-aldehyde. Formation of 2,5-disubstituted pyrazines occurs as a result of dimerization of two amino aldehydes and has been shown for a few ATR enzymes (Figure 6);^19,52 however, no in vitro biosynthesis of 2,6-disubstituted pyrazines has been reported to the best of our knowledge. We show that PvfC_Pf0-1 produces the 2,6-disubstituted pyrazine both in vitro and in cells, albeit at a lower level than the 2,5-disubstituted isomer. A possible biosynthetic mechanism involves a condensation between the amine of one Leu-aldehyde and the aldehyde of the other; protonation of the resulting imine leads to a 1,2 alkyl shift, which forms a second imine at an adjacent carbon (Figure 6). Subsequent N-to-C cyclization of the second imine with the terminal aldehyde followed by dehydration and oxidation would form 2,6-diisobutylpyrazine (Figures 6 and S4). A similar mechanism was proposed for a 2,6-diisopropylpyrazine analogue isolated from Chondramyces crocatus,⁵⁸ but no biosynthetic enzymes were identified. Further, we also identified three (dihydro)pyrazine N-oxides, the Leu-(d)PNOs, whose biosynthesis requires both PvfC_Pf0-1 and PvfB_Pf0-1 and likely employs a similar mechanism as the biosynthesis of the Val-(d)PNOs in P. entomophila L48 that involves an N-hydroxylated amino aldehyde intermediate (Figure 6).²⁹ Finally, the ability of the PvfC_L48 homologue to generate valine derivatives of flavacol, leucinazole, and pyrazines suggests that the biosynthetic versatility may be broadly applicable to different PvfC homologues and other ATR enzymes.

Single natural product biosynthetic pathways often generate multiple products.⁵⁹ For example, a single enzyme in cyanobacteria can produce ~30 lanthionine-containing peptides,⁶⁰ the prochlorosins; a phenazine biosynthetic pathway can generate 20–112 structurally diverse phenazines.^61,62 While the function of prochlorosins remains elusive,⁶³ the phenazine antibiotics have been postulated to serve synergistic or complementary functions in bacteria–bacteria competition and bacteria–host interactions.⁶² Our characterization of PvfC paves the way for understanding the biological significance of the diversity-prone biosynthesis by pvf-encoded enzymes.

Pyrazine-related compounds display wide-ranging bioactivities. The pyrazinone flavacol is cytotoxic,⁶⁴ and other pyrazinones including phevalin and the lumizinones inhibit the protease calpain.^65,66 Roles for pyrazinones in pathogen–host interactions have also emerged: phevalin production in S. aureus is required for efficient phagosomal escape, killing of host cells, and full virulence in S. aureus-induced pneumonia;⁶⁷ autoinducer-3 and DPO are quorum-sensing molecules reported from E. coli and Vibrio cholerae, respectively, which regulate the expression of virulence genes via bacterial cell-to-cell signaling.^54,55 Pyrazines are insect pheromones.⁶⁸ Imidazoles exhibit antimicrobial, anticancer, and anthelmintic activities:^69,70 the bisindole imidazole topsentin inhibits sortase;⁵⁶ synthetic analogues of deoxytopsentin inhibit S. aureus kinases.⁷¹ While the Leu-(d)PNOs, flavacol, leucinazole, PZ-1, and PZ-2 do not inhibit the growth of Bacillus subtilis or E. coli (Figure S19), their valine-derived counterparts are overproduced by the pvf-containing opportunistic pathogen B. cenocepacia under antibiotic stress, suggesting that these pyrazine-related molecules are relevant to the biology of the Pseudomonas virulence factor pathway and could play a role in bacterial stress response. Studies are underway to investigate how the production of pyrazine-related molecules contributes to the biosynthesis of the PVF autoinducers.

Genes encoding ATR enzymes are widely distributed; they have been identified in over 18 bacterial genera, as well as algal and fungal species,⁷² but the functions for most of these ATRs are unknown. Our study of PvfC_Pf0-1 expands the functions of bacterial ATRs beyond the biosynthesis of chloramphenicol³ and the recently discovered guanipiperazines⁷² and showcases the ability of PvfC and homologues to synthesize diverse pyrazine-related compounds. Consistent with our findings, the heterologous expression of an ATR from a Xenorhabdus bacterium also results in a putative pyrazinone product.⁵² Thus, the production of pyrazine-related compounds by ATR enzymes may be widespread in nature and mediate chemical signaling and interactions between organisms.

In conclusion, we show that the ATR enzyme PvfC_Pf0-1 of the Pseudomonas virulence factor pathway directs the formation of seven diverse heterocyclic molecules, including a pyrazinone, an imidazole, two pyrazines, and three pyrazine N-oxides, via amino aldehyde intermediates. Production of flavacol by PvfC_Pf0-1 represents a novel, C domain-independent biosynthetic route to pyrazinones, an emerging class of signaling molecules and virulence factors. The wide distribution of bacterial ATR enzymes indicates potential for identifying additional bioactive molecules.

METHODS

Large-Scale Growth and Extraction of Metabolites 1–7 for Isolation.

To isolate Leu-(d)PNOs (1, 2, 4), three 1 L cultures in LB that contained gentamycin were inoculated with 5 mL overnight cultures of P. aeruginosa PAO1 that contains pPSV35-pvfBCD_Pf0-1. The P. aeruginosa cultures were incubated at 30 °C and induced with 0.5 mM IPTG when OD₆₀₀ reached 0.4. To isolate leucinazole (3) and flavacol (5), four 1 L LB-gentamycin cultures were inoculated with 5 mL overnight cultures of E. coli BAP1 that contains pPSV35-pvf_Pfo-1. To isolate PZ-1 (6) and PZ-2 (7), six 1 L LB-gentamycin cultures were inoculated with 5 mL overnight cultures of E. coli BAP1 that contains pPSV35-pvf_Pf0-1. The E. coli cultures were incubated at 37 °C and induced with 0.5 mM IPTG when OD₆₀₀ reached 0.4. The temperature was dropped to 30 °C after induction. Both the P. aeruginosa and E. coli cultures were grown at 30 °C for 24 h, and then the culture supernatant was separated from bacterial cells by centrifugation. The supernatant was extracted twice with one volume of ethyl acetate each time. The organic layers were combined, dried with sodium sulfate, filtered, and evaporated to dryness. The extracted metabolites were resuspended in 5 mL of methanol, transferred to a scintillation vial, dried down, and stored at −20 °C. See the Supporting Information for detailed methods for purification of 1–7.

PvfC_Pf0-1 In Vitro Reactions to Produce Leucinazole, Flavacol, PZ-1, and PZ-2.

PvfC_Pf0-1 (10 μM) in 25 mM HEPES (pH = 7.5) was incubated with 2 mM MgCl₂, 4 μM Sfp, and 0.2 mM coenzyme A for 30 min at room temperature. Samples of 1 mM ATP and 5 mM l-Leu (or ¹⁵N-l-Leu) were then added and incubated for 30 min at room temperature. A sample of 2 mM NADPH was added, and the reaction was incubated for 2 h at room temperature. The reaction was quenched using acetonitrile (one volume of enzyme reaction) and incubated at −20 °C for at least 20 min. Samples were centrifuged at 20,800 x g for 10 min to remove the precipitated protein and analyzed by LC-HRMS in positive mode. For PZ-1 production, the reaction was set up similarly with the exception of incubation for 2.5 h after the addition of NADPH.

To separate PZ-1 and PZ-2, in vitro reactions were analyzed using the method described below on a Hypercarb column. All other in vitro reactions were analyzed using the method described in the LC-HRMS of In Vitro Reactions Separated on a C18 Column (Supporting information) section.

LC-HRMS of In Vitro Reactions Separated on a Hypercarb Column.

A sample of 10 μL of PvfC_Pf0-1 in vitro reaction was analyzed by LC-HRMS and separated on a Thermo Scientific Hypercarb column (5 μm, 100 ×4.6 mm). Solvent A consisted of 0.1% formic acid in water, and solvent B consisted of 0.1% formic acid in acetonitrile. Using a flow rate of 0.6 mL min⁻¹, samples were separated at 2% B for 2 min, followed by a linear gradient of 2–80% B over 16.8 min, 80–95% B over 6 min, and 95% B for 2 min. ESI mass spectrometry was carried out under positive ion mode using the following parameters: gas temperature 350 °C, drying gas 5 L min⁻¹, nebulizer 20 psi, fragmentor 175 V, and skimmer 65 V.

Derivatization of PvfC_Pf0-1 In Vitro Reactions with Pentafluorophenylhydrazine (PFPH) and Reaction to Produce the Leu-Leu-Aldehyde (8).

PvfC_Pf0-1 (10 μM) in 25 mM HEPES buffer (pH = 7.5) was incubated with 2 mM MgCl₂, 4 μM Sfp, and 0.2 mM coenzyme A for 30 min at room temperature. Samples of 1 mM ATP and 5 mM l-Leu (or ¹⁵N-l-Leu) were then added and incubated for 30 min at room temperature. A sample of 2 mM NADPH was added, and the reaction was incubated for 2 h at room temperature. Reactions were quenched using acetonitrile (one volume of enzyme reaction), and then 25 mM PFPH was added and the mixture was incubated for 20 min at room temperature. The mixture was subsequently incubated at −20 °C for at least 20 min. Samples were centrifuged at 20,800 x g for 10 min to remove the precipitated protein and analyzed by LC-HRMS in positive mode. To identify the Leu-Leu-aldehyde (8) intermediate en route to leucinazole/flavacol formation, the sample was set up as described above with the exception that PFPH was added at the same time as NADPH and incubated for 2 h before quenching of the reaction. The reaction was separated using the method described in the LC-HRMS of In Vitro Reactions Separated on a C18 Column (Supporting information) section.

Additional detailed experimental procedures are provided in the Supporting Information.