A Three Enzyme Pathway for 2-Amino-3-Hydroxycyclopent-2-Enone Formation and Incorporation in Natural Product Biosynthesis

Abstract

A number of natural products contain a 2-amino-3-hydroxycyclopent-2-enone five membered ring, termed C₅N, which is condensed via an amide linkage to a variety of polyketide-derived polyenoic acid scaffolds. Bacterial genome mining indicates three tandem ORFs that may be involved in C₅N formation and subsequent installation in amide linkages. We show that the protein products of three tandem ORFs (ORF33-35) from the ECO-02301 biosynthetic gene cluster in Streptomyces aizunenesis NRRL-B-11277, when purified from E. coli, demonstrate the requisite enzyme activities for C₅N formation and amide ligation. First, succinyl-CoA and glycine are condensed to generate 5-aminolevulinate (ALA) by a dedicated PLP-dependent ALA synthase (ORF34). Then ALA is converted to ALA-CoA through an ALA-AMP intermediate by an acyl-CoA ligase (ORF35). ALA-CoA is unstable and has a half-life of ~10 minutes under incubation conditions for off-pathway cyclization to 2,5-piperidinedione. The ALA synthase can compete with the nonenzymatic decomposition route and act in a novel second transformation, cyclizing ALA-CoA to C₅N. C₅N is then a substrate for the third enzyme, an ATP-dependent amide synthetase (ORF33). Using octatrienoic acid as a mimic of the C₅₆ polyenoic acid scaffold of ECO-02301, formation of the octatrienyl-C₅N product was observed. This three enzyme pathway is likely the general route to the C₅N ring system in other natural products, including the antibiotic moenomycin.

Introduction

A variety of natural products with a vast range of biological activities have polyketide and nonribosomal peptide fragments joined together. Frequently, these scaffolds arise from hybrid nonribosomal peptide synthetase (NRPS)-polyketide synthase (PKS) assembly lines. In some cases, the polyketide backbone predominates as in rapamycin and FK506,^1,2 where a single NPRS-derived pipecolate is embedded in a polyketide framework. The reverse can also occur as in bleomycin and its congeners,³ where a single polyketide fragment interrupts the nonribosomal peptide backbone. During the biosynthesis of some NRPS/PKS derived natural products non-NRPS/PKS machinery is enlisted to carry out the condensation between scaffold fragments. In one example, a polyketide acid, coronofacic acid, is enzymatically ligated to a nonproteinogenic amino acid, coronamic acid, by a trans-acting amide synthetase to yield the phytohormone antagonist coronatine.^4,5

Of special note are cyclic five membered nitrogen-containing ring structures that generate conformational constraints and offer hydrogen bonding possibilities for interaction with target proteins. Examples of these ring systems include pyrrolidine-2,5-diones as in the methylsuccinamide terminus of andrimid,⁶ and the pyrrolidine-2,4-dione (tetramic acid) moieties in a wide range of natural products including equisetin⁷ and cyclopiazonate⁸. In these instances the pyrrolidine-diones are generated by the chain termination domains of hybrid PKS-NRPS assembly lines during product release. Typically the ketone at C4 is enolized with the 4-hydroxy-3-ene tautomer predominating.

A distinct type of nitrogen-containing cyclic dione serves as a hydrogen bond donor/acceptor pharmacophore in more than 30 members of the manumycin family.⁹ Termed the C₅N unit, this is formally a 2-aminocyclopentanedione unit, but it too predominates as the enol tautomer, 2-amino-3-hydroxycyclopent-2-enone. In most cases, including the manumycins, limocrocin¹⁰, Sch725424¹¹ and ECO-02301¹², the amino group of the C₅N unit is acylated through an amide bond to a polyenoic acid component of polyketide origin (Figure 1). This suggests an analogy to the biosynthesis of coronatine and the aminocoumarin antibiotics, where the acid and amine components are connected late in the biosynthetic pathway by an amide-forming ligase.^13,14 An unusual variant is found in moenomycin biosynthesis where the C₅N unit is in amide linkage to a hexuronic acid ring.¹⁵

Figure 1.

Examples of natural products containing C₅N (shaded). Candidate ORFs and a putative biosynthetic pathway for C₅N formation during ECO-02301 biosynthesis are boxed. ALAS: 5-aminolevulinate synthase, ACL: acyl-CoA ligase, AMS: amide synthetase.

Initial inspection of the C₅N scaffold suggested it could arise from 5-aminolevulinic acid (ALA) by an unusual cyclization process (Figure 1).^16–18 ALA is a primary metabolite in many organisms as the immediate precursor to the monopyrrolic building block, porphobilinogen, for heme and corrin biogenesis.¹⁹ While ALA can arise from reduction of glutamate on glutamyl-tRNA_glu, ALA can also be generated via the Shemin pathway in which PLP-dependent condensation of glycine and succinyl-CoA results in net loss of CO₂ and CoASH to form the five carbon 5-amino-4-keto-levulinic acid.²⁰ In 2006, genetic studies of the asukamycin producer Streptomyces nodosus subsp. asukaensis revealed genes for both pathways to ALA.²¹ The knockout of gene hemA-asuA abolished asukamycin production. Furthermore, expression and purification of the encoding PLP-binding enzyme in E. coli gave low but detectable activity of ALA synthesis. These results validated that the glycine and succinyl-CoA condensing enzyme is responsible for ALA formation and likely involved in the biosynthesis of C₅N.

Two attributes of assembly of C₅N-containing amide natural products intrigued us most: 1) the unusual cyclization of ALA to C₅N and 2) the ability of the amide synthetase to activate the presumably weak nitrogen nucleophile of C₅N as well as its specificity for the polyenoic or hexuronic acid partner substrates. To this end we have examined the biosynthetic gene cluster of antifungal agent ECO-02301 for possible enzyme candidates responsible for the formation and ligation of the C₅N unit to the C₅₆ polyenoic acid moiety.¹² The production of ECO-02301 had been predicted by genomic analysis of Streptomyces aizunensis NRRL B-11277, and then ECO-02301 was successfully identified as a 1297 Da linear polyketide product from the fermentation broth. Notable to our interests is the terminal C₅N moiety and the prediction by bioinformatic analysis that ORF33-35 in the ECO-02301 gene cluster could be involved in C₅N formation and ligation (Figure 1). ORF33 could be a putative amide synthetase (AMS, ligation of C₅N to polyenoate), while ORF34 has high sequence similarity to a PLP-dependent ALA synthase (ALAS) and ORF35 has signature sequences predicted for an acyl-CoA ligase (ACL). Through in vitro biochemical characterization of all three enzymes, we present experimental validation of the C₅N biosynthetic pathway, including identification of the novel C₅N cyclase activity.

Materials and Methods

Bacterial Strains, Plasmids, Materials, and Instrumentation

S. aizunensis NRRL B-11277 was obtained from USDA ARS Culture Collection. Oligonucleotide primers were synthesized by Integrated DNA Technologies, and PCR was performed with Phusion High-Fidelity PCR Master Mix (NEB). Cloning was performed using the Xa/LIC Cloning Kit (Novagen). Recombinant plasmid DNA was purified with a QIAprep kit (Qiagen). DNA sequencing was performed at the Molecular Biology Core Facilities of the Dana Farber Cancer Institute. E. coli BL21 Gold (DE3) Cells (Stratagene) transformed with pET-derived vectors were used for overexpression of proteins in LB medium supplemented with kanamycin. Nickel-nitrilotriacetic acid agarose (Ni-NTA) superflow resin and SDS-PAGE gels were purchased from Qiagen and Biorad respectively. Protein samples were concentrated using 30 kDa MMCO Amicon Ultra filters (Millipore). DNA and Protein concentrations were determined by Nanodrop 1000 spectrophotometer (Thermo Scientific). All chemicals were purchased from Sigma-Aldrich unless stated otherwise. [4-¹³C]5-aminolevulinic acid and NMR solvents were purchased from Cambridge Isotope Laboratories, Inc. 2,5-piperidinedione was purchased from Small Molecules. Boc-5-aminolevulinic acid was purchased from AnaSpec Inc.

NMR spectra were recorded on a Varian Inova 500 (500 MHz ¹H, 125 MHz ¹³C) instrument in CDCl₃. LC-MS analysis was performed on an Agilent Technologies 6520 Accurate-Mass Q-TOF LC-MS instrument or an Agilent Technologies 6210 Accurate-Mass TOF LC-MS instrument. HPLC analysis was performed on a Beckman Coulter System Gold with an Inertsil ODS-4 C18 column (4.6 × 250 mm) from GL Sciences Inc. A Phenomenex Luna C18 column (21.2 × 250 mm) was used for preparative HPLC. A SPECTRAmax plus 384 96-well plate reader from Molecular Devices was used for continuous spectrophotometric assays.

Cloning, Overexpression, and Purification of ORF33-35

ORF33-35 were PCR amplified from genomic DNA extracted from S. aizunensis. For ORF33, the forward primer was 5’-GGTATTGAGGGTCGCATGACCCCGCAGGACCATTGGTG-3’ and the reverse primer was 5’-AGAGGAGAGTTAGAGCCTTAGGAGTCGAGCAGCTGCAGCC-3’. For ORF34, the forward primer was 5’-GGTATTGAGGGTCGCATGAACCTGCACCTGGAATCGTA-3’ and the reverse primer was 5’-AGAGGAGAGTTAGAGCCTTACGAAAGCCAGTTCCTGTCGG-3’. For ORF35, the forward primer was 5’-GGTATTGAGGGTCGCATGACCCGGTCGGTGGCGGCCGT-3’ and the reverse primer was 5’-AGAGGAGAGTTAGAGCCTTACGCGTAGCGGTGTGCCAGCT-3’. Purified PCR products were ligated to pET-30 Xa/LIC following the standard protocol and confirmed by DNA sequencing. The resulting expression constructs were transformed into E. coli BL21 cells for protein expression. Expression and purification for all three proteins followed the same general procedure and is detailed as follows. In 1 L of liquid culture, the cells were grown at 37 °C in LB medium with 50 µg/mL kanamycin to an OD₆₀₀ of 0.4. The cells were cooled on ice for 10 min and then induced with 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) for 16 h at 16 °C. The cells were harvested by centrifugation (6000 rpm, 6 min, 4 °C), resuspended in 30 mL lysis buffer (20 mM HEPES pH 8.0, 0.5 M NaCl, 5 mM imidazole) and lysed by sonication on ice. Cellular debris was removed by ultracentrifugation (35000 rpm, 30 min, 4 °C). Ni-NTA agarose resin was added to the supernatant (1 mL/L of culture) and the solution was nutated at 4 °C for 1 h. The protein resin mixture was loaded into a gravity flow column, and proteins were eluted with increasing concentrations of imidazole in Buffer A (50 mM HEPES, pH 8.0, 2 mM EDTA). Purified proteins were concentrated and buffer exchanged into Buffer A + 10% glycerol using Amicon Ultra filters. Gel filtration chromatography was performed on a Superdex 200 10/300 column connected to an Amersham automated FPLC system at 4 °C (50 mM phosphate buffer, 150 mM NaCl, pH 7.8). The final proteins were flash-frozen in liquid nitrogen and stored at −80 °C.

HPLC/LC-MS Product Assays

All product assays were performed in 50 µL of 50 mM HEPES (pH 8.0) at 25 °C. LC-MS analysis was normally performed with a linear gradient of 0 to 20% CH₃CN (v/v) over 20 min, 20% to 95% CH₃CN (v/v) over 5 min, and 95% CH₃CN (v/v) for a further 15 min in H₂O supplemented with 0.1% (v/v) formic acid, at a flow rate of 0.5 mL/min. HPLC analysis was normally performed with a linear gradient of 2 to 12% CH₃CN (v/v) over 30 min, 12% to 95% CH₃CN (v/v) over 5 min, and 95% CH₃CN (v/v) for a further 10 min in H₂O supplemented with 0.1% (v/v) TFA at a flow rate of 1 mL/min. For analysis of ALAS activity, 20–50 µM ORF34 was incubated with 5 mM glycine and 1 mM succinyl-CoA for 2 h. The protein was removed by 5 kDa MMCO filter tubes, and the filtered reaction mixture was subjected to LC-MS analysis. OPTA derivatization of amines in the filtered reaction mixture was further performed following the reported protocol.²² For analysis of ACL activity, 20–50 µM ORF35 was incubated with 2–5 mM acid substrates, 2 mM ATP, 2 mM CoA and 2 mM MgCl₂ for 30 min. The reactions were stopped by addition of tricholoacetic acid (TCA) to a final concentration of 5% (v/v), and the supernatant was subjected to both LC-MS and HPLC analysis. For analysis of cyclase activity, 20–50 µM ORF34 was added to the reaction mixture of ORF35 and incubated for ~60 min. The reactions were stopped by addition of 100 µL of CH₃CN, and the supernatant was subjected to LC-MS and HPLC analysis. Alternatively, 20–50 µM ORF34 was added to buffer containing synthesized ALA-CoA substrate and reacted for 10 min prior to LC-MS and HPLC analysis. For analysis of AMS activity, 20–50 µM ORF33 was added to buffer containing synthesized 2 mM C₅N, 2 mM 2,4,6-octatrienoic acid, 5 mM ATP, and 2 mM MgCl₂. The reactions were stopped after 1 h by addition of CH₃CN and the supernatant was subjected to LC-MS and HPLC analysis. For reconstitution of the entire pathway, reaction mixtures contained 50 µM ORF33-35, 5 mM glycine, 1 mM succinyl-CoA, 5 mM ATP, 2 mM CoA, 2 mM MgCl₂, and 2 mM 2,4,6-octatrienoic acid. The reactions were stopped after 2–3 h by addition of 100 µL of CH₃CN and the supernatant was subjected to LC-MS and HPLC analysis.

Preparation of ALA-CoA

Boc-ALA [25 µmol, 1 equivalent (eq.)], CoA sodium salt (25 µmol, 1 eq.), PyBOP (Novobiochem, 50 µmol, 2 eq.) and K₂CO₃ (100 µmol, 4 eq.) were dissolved in tetrahydrofuran/H₂O [1/1 (v/v), 1 mL] and stirred for 2 h at room temperature. The reaction mixture was directly purified by preparative HPLC (two injections, 0.5 mL each) using a linear gradient of 2 to 60% CH₃CN (v/v) over 30 min in H₂O supplemented with 0.1% (v/v) TFA at a flow rate of 10 mL/min. The HPLC-purified mixture was concentrated on a rotary evaporator and lyophilized to yield Boc-ALA-CoA (30% yield) as a white solid. The Boc protecting group was removed by redissolving Boc-ALA-CoA in 1 mL of 5–10% TFA in H₂O and stirring for 30 min at room temperature. The reaction mixture was again purified by preparative HPLC (same method as above), concentrated and lyophilized, yielding ALA-CoA (85% yield) as a white solid.

Preparation of 2,4,6-octatrienyl-C₅N

To a cooled solution (0 °C) of (2E,4E,6E)-octa-2,4,6-trienoic acid²³ (97 mg, 0.700 mmol) in CH₂Cl₂ (7 mL) and DMF (3 drops) was added a solution of oxalyl chloride (100 mg, 0.840 mmol, 1.2 equiv) in CHCl₂ (0.5 mL) drop-wise over 30 minutes. The resulting solution was stirred at 0 °C for two hours, then slowly canulated into a cooled solution (0 °C) of freshly prepared 2-amino-3-hydroxy-2-cyclopenten-1-one hydrochloride²⁴ (103 mg 0.700 mmol), N,N-dimethylaminopyridine (5 mg, 0.041 mmol, 5 mol %) in pyridine (15 mL) . The reaction mixture was slowly warmed to room temperature and maintained stirring overnight. The reaction was then concentrated (azeotroping with toluene) and purified by flash column chromatography with 98 : 2 (CH₂Cl₂ : MeOH) to afford a yellow solid (55 mg 0.236 mmol, 34%). ¹H NMR (500 MHz, CDCl₃) δ_H 13.6 (s, 1H), 7.45 (broad s, 1H), 7.34 (dd, J = 14.5, 11.5 1H), 6.58 (dd, J = 15.0, 11.0 1H), 6.22 (dd, J = 14.5, 12.0 1H), 6.15–6.20 (m, 1 H), 6.00 (dd, J = 14.5, 9.5 1H), 5.95 (d, J = 15.0 1H), 2.60–2.62 (m, 2H), 2.52–2.55 (m, 2 H), 1.85 (d, J = 7.0 3 H); ¹³C NMR (125 MHz, CDCl₃) δ_C 197.3, 177.6, 173.8, 145.0, 142.5, 136.3, 131.3, 127.2, 119.2, 32.3, 25.7, 18.8.

Kinetic Investigations of ORF34

ORF34 activity was measured at 25 °C in a 100 reaction volume containing 100 mM sodium phosphate (pH 7.8), 0.1 mM 5,5’-dithio-bis(2-nitro-benzoic acid) (DTNB), 0.2 mM PLP and 20 µM ORF34. Kinetic parameters for the succinyl-CoA substrate were determined with the concentration of glycine maintained at 10 mM and the concentration of succinyl-CoA varied from 0 to 425 µM; kinetic parameters for the glycine substrate were determined with the concentration of succinyl-CoA maintained at 0.5 mM and the concentration of glycine varied from 0 to 5 mM. Kinetic parameters for the ALA-CoA substrate in the cyclization reaction were determined by varying the concentration of ALA-CoA from 0 to 400 µM. For each concentration a control reaction was carried out without enzyme. Assays were initiated by the addition of acyl-CoA. Activity was monitored continuously by following the increase in absorbance at 412 nm resulting from the reaction between the free thiol of CoASH and DTNB. Initial velocities (v_ini) were calculated using the extinction coefficient of 14150 M⁻¹ cm⁻¹ (v_ini = v_ini[with enzyme] - v_ini[without enzyme]). The v_ini data were fitted to the Michaelis-Menten equation in GraphPad Prism to obtain estimates for k_cat and K_m.

ATP-[³²P]PP_i Exchange Assays for ORF33 and ORF35

A typical assay contained, in a total volume of 800 µL, 5 mM acid substrate, 5 mM ATP, 10 mM MgCl₂, 1 mM Na[³²P]pyrophosphate(PP_i) (~2.5 × 10⁶ cpm/mL), and 50 mM HEPES pH 8.0. Reactions were initiated by the addition of enzyme (1 µM of ORF35 or 0.8 µM of ORF33). At regular time intervals, 100 µL aliquots were quenched with 500 µL of a charcoal suspension (100 mM NaPP_i, 350 mM HClO₄, and 16 g/L charcoal). The mixtures were vortexed and then centrifuged at 13000 rpm for 6 min. The pellets were washed twice with 500 µL of wash solution (100 mM NaPP_i and 350mM HClO₄). Charcoal-bound radioactivity was measured on a Beckman LS 6500 scintillation counter. Turnover was calculated as (% incorporation of [³²P]PP_i)[total PP_i]/[Enz].

ATP-PP_i Release Assays for ORF35

The inorganic pyrophosphate released by enzymatic reaction was measure d continuously using the EnzChek Pyrophosphate Assay Kit (Invitrogen). A typical assay contained, in a total volume of 100 µL, 0–5 mM acid substrate, 5 mM ATP, 2 mM MgCl₂, and 5 µM ORF35 in 50 mM HEPES, pH 8.0. MESG substrate, purine nucleoside phosphorylase and inorganic pyrophosphatase were added according to the protocol. Reactions were initiated by the addition of acid substrate and monitored at 360 nm. Initial velocities were calculated using the standard curve for inorganic pyrophosphate. The data were fitted to the Michaelis-Menten equation in GraphPad Prism to obtain estimates for k_cat and K_m.

Kinetic Investigations of ORF33

The amide synthetase activity of ORF33 was measured at 25 °C in a 100 µL reaction volume containing 50 mM HEPES, pH 8.0, 5 mM ATP, 2 mM MgCl₂, 5 µM ORF33 and acid and amine substrates. Activity was monitored continuously by following the increase in absorbance at 355 nm resulting from the production of the ligated product. Kinetic parameters for the 2,4,6-octatrienoic acid substrate were determined with the concentration of C₅N maintained at 1 mM and the concentration of acid varied from 0 to 5 mM. Higher acid concentrations could not be achieved due to poor solubility. Kinetic parameters for the C₅N substrate were determined with the concentration of 2,4,6-octatrienoic acid maintained at 2 mM and the concentration of C₅N varied from 0 to 1.1 mM. Initial velocities (v_ini) were calculated using a standard curve of synthetically prepared 2,4,6-octatrienyl-C₅N. The v_ini data were fitted to the Michaelis-Menten equation in GraphPad Prism to obtain estimates for k_cat and K_m.

Results

Expression and Purification of ORF33-35 in E. coli

ORF33-35 from S. azuinensis were amplified and cloned into an expression vector encoding an N-terminal His₆-tag. The corresponding recombinant proteins (ORF34, 49 kDa; ORF35, 59 kDa; ORF33, 60 kDa) were overproduced in E. coli and purified using Ni-NTA affinity chromatography (Figure S1). The approximate protein yields were 10 mg/L for ORF34, 26 mg/L for ORF35 and 30 mg/L for ORF33.

ORF34 Is a PLP-dependent Aminolevulinate Synthase (ALAS)

Examination of the UV absorption spectra of ORF34 identified a characteristic absorbance maximum at 420 nm, indicating the presence of an enzyme-bound PLP Schiff base (Figure S2).²⁵ Titration with PLP indicated approximately 60% of the protein was in the holo-form (PLP bound) upon isolation (Figure S3). Incubation of purified ORF34 with glycine and succinyl-CoA led to the formation of a new compound detected by LC-MS, with retention time and mass (m/z = 132 [M + H]⁺) matching those of the commercial ALA standard (Figure 2A and 2B). The identity of the product ALA was further confirmed using OPTA derivatization and LC-MS analysis (Figure S4). The product assays confirmed that ORF34 was an aminolevulinate synthase (ALAS), which is in agreement with the previous report that ALAS is essential for asukamycin biosynthesis in Streptomyces nodosus subsp. Asukaensis. Notably, the gene encoding the S. nodosus subsp. Asukaensis ALAS is also adjacent to genes encoding AMS and ACL homologs.²¹

Figure 2.

Characterization of ORF34 as a 5-aminolevulinate synthase. (A) Schematic of the condensation of glycine and succinyl-CoA to yield 5-aminolevulic acid catalyzed by ORF34. (B) Extracted ion chromatograms (m/z = 132 [M + H]⁺) showing ORF34-catalyzed production of ALA (trace ii) as compared to the ALA standard (trace iii). ALA was not detected in controls without ORF34 (shown, trace i) or without any of the substrates (not shown). (C) Determination of ORF34 kinetic parameters by assay with DTNB. Parameters were determined for succinyl-CoA (top panel) by fixing the glycine concentration at 10 mM while parameters were determined for glycine (bottom panel) by fixing the succinyl-CoA concentration at 500 µM. Error bars represent standard deviations from at least three independently performed experiments.

To determine the kinetic parameters of ORF34, the DTNB assay monitoring the free CoA formation was performed (Figure 2C). The K_m values of the enzyme were determined to be 120 µM for succinyl-CoA and 2.5 mM for glycine, which is within range of the reported values for ALASs originating from different organisms (17–257 µM for succinyl-CoA and 1.9–9.7 mM for glycine).^26,27 The turnover number was determined to be ~0.4 min⁻¹, which was ~10 fold less than those of ALASs dedicated to heme biosynthesis, but ~10 fold greater than that of the ALAS involved in the biosynthesis of asukamycin.²¹

ORF35 Is an Acyl-CoA Ligase (ACL)

ORF35 shows high sequence similarity to acyl-CoA ligases, and was proposed to ligate ALA synthesized by ORF34 to CoASH. To probe the role of ORF35, in vitro assays containing ALA, CoASH, ATP, Mg²⁺ and purified ORF35 were performed in HEPES buffer at pH 8.0 and monitored by LC-MS. The protein readily precipitated at lower pH values (<7.5). ALA-CoA was detected as a new product (m/z = 881.1723 [M + H]⁺, Δ = 2.1 mmu) (Figure S5), albeit in low yields. Full sets of control experiments confirmed that ORF35 catalyzed the formation of ALA-CoA in an ATP-dependent manner (Figure 3A and 3B). To determine the substrate specificity of ORF35, the classical ATP-[³²P]PP_i exchange assay was used to monitor the reversible formation of acyl-AMP as the first half reaction. The enzyme demonstrated specificity towards ALA over substrate analogs such as levulinic acid and 5-aminovaleric acid (Figure 3C), with a k_cat of 6.6 min⁻¹ for ALA. Interestingly, ORF35 recognized octanoic acid (k_cat = 17.5 min⁻¹) better than ALA. A coupled spectrophotometric assay for PP_i release was further used to kinetically characterize ORF35, and the substrate specificity was in agreement with the ATP-PP_i exchange assays (for ALA, k_cat = 1.24 min⁻¹, K_m = 478 µM) (Figure 3D, Table 1 and Figure S6). It is notable that the k_cat values obtained from the PP_i release assays were relatively low for all the acids tested, which is likely due to slow, off-pathway release of the tightly-bound acyl-AMP intermediates. Regardless of assay used, however, ORF35 showed better specificity towards medium-chain fatty acids, which may indicate that the ancestor of this enzyme is a fatty acid acyl-CoA ligase.

Figure 3.

Characterization of ORF35 as an acyl-CoA ligase. (A) Schematic of the ligation of ALA to CoASH catalyzed by ORF35. (B) HPLC traces (260 nm) of ORF35 reaction (trace i) and controls with no ATP (trace ii), no CoA (trace iii), and no enzyme (trace iv). (C) ATP-[³²P]PP_i exchange assays monitoring the reversible formation of acyl-AMP as the first half reaction. Calculated rates: k(ALA) = 6.6 min⁻¹, k(octanoic acid [C8]) = 17.5 min⁻¹, k(levulinic acid [LA]) = 5 min⁻¹, k(5-aminovaleric acid [AV]) = 1.7 min⁻¹. (D) Kinetic parameters for the ALA substrate determined by ATP-PP_i release assays. Error bars represent standard deviations from at least three independently performed experiments.

Table 1.

ATP-PP_i release assays for ORF35 on selected acids. ALA: 5-aminolevulinic acid; LA: levulinic acid; AV: 5-aminovaleric acid; C6: hexanoic acid; C8: octanoic acid.

Acids	Vmax (mM min−1)	Km (µM)	[ORF35](µM)	kcat (min−1)	kcat/Km (min−1 mM−1)
ALA	6.19 ± 0.14	478 ±43	5	1.24 ± 0.03	2.594
LA	5.43 ± 0.08	1087 ± 46	5	1.08 ± 0.02	0.999
AV	8.07 ± 0.22	7431 ± 370	5	1.61 ± 0.04	0.217
C6	6.56 ± 0.15	187 ± 15	5	1.31 ± 0.03	7.016
C8	6.62 ± 0.23	181 ± 21	5	1.32 ± 0.05	7.316

To confirm the acyl-CoA formation catalyzed by ORF35, in vitro reactions followed by LC-MS analyses were performed with selected acids (Figure S7). As expected, under the same reaction conditions, hexanoyl-CoA was quickly formed in good yield and levulinyl-CoA was formed less efficiently. However, very little ALA-CoA could be detected as noted above, and no 5-aminovaleryl-CoA could be detected by UV or mass ion extraction. Instead, 2,5-piperidinedione and 2-piperidinone were observed as products respectively, which were reasoned to be formed through non-enzymatic cyclization of the aminoacyl-CoAs. Particularly, at pH 8.0 the -NH₂ group of the product aminoacyl-CoAs is likely reactive enough to serve as an internal nucleophile and attack the activated carbonyl of the CoA thioester terminus, leading to spontaneous formation of the six membered ring (Figure 4). ALA-CoA was found to be stable in acidic conditions, which is consistent with our hypothesis. Considering the instability of ALA-CoA at physiological pH, the mechanism underlying the cyclization of the activated ALA to C₅N instead of the default 2,5-piperidinedione was intriguing. Because no C₅N was detectable in the ORF35 in vitro assays, we reasoned that an additional dedicated cyclase was required to direct the synthesis of C₅N from ALA-CoA.

Figure 4.

Instability of ALA-CoA. (A) Extracted ion chromatograms (m/z = 114 [M + H]⁺) showing the formation of PDO in the reaction mixture of ORF35 (trace ii). PDO standard is shown in trace iii and the no enzyme control is shown in trace i. (B) Schematic of the spontaneous cyclization of ALA-CoA to PDO.

PLP-dependent ORF34 Is a Bifunctional ALAS and ALA-CoA Cyclase

It has been reported that the keto-form of ALA was in large excess compared to the enolic tautomers at neutral pH. Using ¹³C and ¹H NMR, no direct observation of the enol forms of ALA had been achieved,28 arguing against a ready supply of the C5 carbon nucleophile required for spontaneous intramolecular capture of ALA-CoA to C₅N. Therefore, to favor C₅N formation, it seemed likely that activation of C5 adjacent to the amine and sequestration of the reactive terminal amine of ALA-CoA might be required for cyclization. As no putative cyclase was found in the gene cluster, we hypothesized that the PLP-dependent ORF34 ALAS was the best candidate for “cyclase” considering its catalytic mechanism. The ALAS reaction involves binding of glycine to PLP to yield a PLP-glycine complex, and a resonance-stabilized nucleophilic quinonoid intermediate is formed following pro-R proton subtraction.²⁹ Some ALASs can also catalyze labilization of the pro-R proton at C5 of the ALA product and generate detectable PLP-ALA quinonoid upon exposure of the enzyme to ALA.³⁰ These chemical roles of coenzyme PLP in both the forward and reverse reactions meet the prerequisites for C₅N cyclization as mentioned above, particularly the ability to labilize the C5 hydrogen.

To test our hypothesis, purified ORF34 was added to the reaction mixture of ORF35, and the reaction was quenched with acetonitrile and analyzed by LC-MS. C₅N was successfully detected as a new product with UV absorbance maximum at 250 nm (m/z = 114.0557 [M + H]⁺, Δ = 0.7 mmu), and its mass was shifted by +1 using [4-¹³C]ALA as an alternative substrate (Figure S8). The product identity was further confirmed by comparison to authentic C₅N synthesized as a standard by reduction of the corresponding nitro-cyclopentanedione as described in the methods section.²⁴ Control experiments demonstrated that the formation of C₅N from ALA relied on the presence of both enzymes (ORF34 and ORF35) and all cofactors (CoA, ATP and Mg²⁺) (Figure 5A). Prior reports indicated synthetic C₅N is unstable upon concentration.²⁴ We determined by HPLC analysis that under enzymatic reaction conditions C₅N displayed a usable half-life of 140 min at pH 8.0. HPLC and LC-MS detection of C₅N from enzymatic reactions validated our hypothesis, demonstrating that the PLP-dependent ORF34 can direct the cyclization of ALA-CoA to yield C₅N (Figure 5B).

Figure 5.

Characterization of PLP-dependent ORF34 as an ALA-CoA cyclase. (A) UV traces (250 nm) during LC-MS analysis of ORF34/ORF35 activity showing reaction with ORF34 only (trace i), ORF35 only (trace ii), ORF34+ORF35 with [4-¹²C]ALA (trace iii), ORF34+ORF35 with [4-¹³C]ALA (trace iv), and C₅N synthetic standard (trace v). (B) Schematic of cyclization of ALA-CoA to C₅N catalyzed by ORF34 via the PLP-quinonoid intermediate. (C) HPLC trace (250 nm) of ORF34 reaction producing C₅N and CoA using synthetic ALA-CoA as the sole substrate (trace ii). No enzyme control is shown in trace i. (D) Kinetic parameters for the ALA-CoA substrate determined by DTNB assays. Error bars represent standard deviations from at least three independently performed experiments.

To further verify ALA-CoA as the substrate for ORF34 and kinetically characterize the cyclization step, ALA-CoA was synthesized chemically. The synthesis was undertaken with Boc-5-ALA and unblocked CoA, with PyBOP as a condensation reagent to yield Boc-5-ALA-CoA. The Boc group was then removed by TFA and the ALA-CoA product was purified by preparative HPLC. DTNB assays indicated that ALA-CoA had a half-life of ~10 min at pH 8.0, which is consistent with the difficulty in detecting the product ALA-CoA in ORF35 LC-MS assays. C₅N was successfully generated after the incubation of the synthetic ALA-CoA with ORF34, while no C₅N was detectable in the control without enzyme (Figure 5C). These results unequivocally confirmed the role of ORF34 as an ALA-CoA cyclase to afford C₅N. The enzyme demonstrated substantially higher catalytic efficiency for cyclization (k_cat = 2.6 min⁻¹, K_m = 36.7 µM, k_cat /K_m = 71 min⁻¹ mM⁻¹) than for ALA synthesis according to the DTNB assays (Figure 5D).

ORF33 Is a C₅N Amide Synthetase

The condensation presumably between a carboxylic acid and the amine of C₅N constitutes the final step for C₅N incorporation into the polyketide backbone as a chain terminator. This is putatively promoted by ORF33, which shows reasonable sequence similarity (<40%) to amide synthetases (NovL, CloL, CouL, SimL) in the aminocoumarin antibiotic family.¹⁴ To probe the role of ORF33, the purified enzyme was incubated with C₅N, ATP, MgCl₂, and 2,4,6-octatrienoic acid (OTEA), which was used to mimic the C-terminal polyolefinic moiety of the polyketide substrate (Figure 6A and 6B). LC-MS analysis showed production of a new compound with a UV absorbance maximum at 321 nm (λ_max = 300 nm for OTEA) (Figure S9 and S10), along with the formation of AMP and attenuation of both amine and acid substrate peaks. HRMS (m/z = 234.1108 [M + H]⁺, Δ = 1.7 mmu) and mass fragmentation (two major mass fragments: 114.0553, 121.0643) suggested the new product to be 2,4,6-octatrienyl-C₅N (OTEA-C₅N). The identity of the product was further confirmed by comparison to the synthetic standard, produced from synthetic C₅N and OTEA as described in the methods section (Figure S11). Notably, ORF33 was not able to ligate ALA to OTEA in our assays. These results confirmed the role of ORF33 as a C₅N amide synthetase.

Figure 6.

Characterization of ORF33 as an amide synthetase (AMS). (A) Schematic of ligation of C₅N to 2,4,6-octatrienoic acid (OTEA) catalyzed by ORF33. (B) UV traces (321 nm) during LC-MS analysis showing OTEA-C₅N formation using ORF33 and synthetic C₅N (trace i), the entire reconstituted pathway using ORF33-35, glycine, succinyl-CoA and CoA (trace ii). OTEA-C₅N synthetic standard is shown in trace iii and no enzyme control is shown in trace iv. It is notable that all of the no enzyme controls, including the control without ORF33 for trace i and the controls without either of the three enzymes for trace ii were the same as the one shown in trace iv. (C) ATP-[³²P]PP_i exchange assays monitoring the reversible formation of acyl-AMP as the first half reaction of ORF33. Calculated rates: k(OTEA) = 0.30 min⁻¹, k(2,4-hexadienoic acid [HDEA]) = 0.21 min⁻¹, k(2-butenoic acid [BEA]) = 0.009 min⁻¹, k(octanoic acid [C8]) = 0.005 min⁻¹. (D) Kinetic parameters measured for ORF33 by monitoring the formation of OTEA-C₅N at 355 nm. ORF33 concentration was maintained at 5 µM and concentrations of OTEA and C₅N were fixed at 2 mM and 1 mM respectively. Error bars represent standard deviations from at least three independently performed experiments.

ATP-[³²P]PP_i exchange assays demonstrated that ORF33 adenylated carboxylate substrates as predicted on the basis of its homology to AMP-forming enzymes (Figure 6C). The kinetic data further revealed that the enzyme preferred acids of longer chain length and higher degree of unsaturation (OTEA > 2,4-hexadienoic acid >> 2-butenoic acid > octanoic acid) consistent with resemblance to the presumed C₅₆ native substrate. It is notable that the turnover rates were slow (k_cat = 0.3 min⁻¹ for OTEA) in the assays, perhaps limited by slow PP_i off rates. The kinetic parameters of the enzyme towards C₅N and OTEA were then further examined by monitoring the rate of OTEA-C₅N formation at 355 nm (Figure 6D). This wavelength was selected to minimize the inference of the substrate OTEA on UV absorption (Figure S9 and S10). With a C₅N concentration of 1 mM, the enzyme showed a K_m of ~7.8 mM for OTEA with an apparent k_cat of ~10.2 min⁻¹, indicating the poor affinity of this substrate analog. In contrast, ORF33 had a low K_m of ~95 µM for C₅N and an apparent k_cat of ~1.8 min⁻¹ at an OTEA concentration of 2 mM (poor OTEA solubility precluded reproducible assays at higher concentrations). These results demonstrated the capability of ORF33 in recognizing the presumably weak amine nucleophile of C₅N. The overall turnover rate was an order of magnitude greater in the product assays than in the ATP-PP_i exchange assays performed without the amine donor C₅N, indicating that acid activation and amine binding likely accelerate release of the PP_i co-product.

Reconstitution of the Entire C₅N Biosynthetic Pathway in vitro

All three enzymes, ORF33, ORF34, and ORF35, were incubated together with the essential cofactors (Mg²⁺, ATP, CoA) and substrates (glycine, succinyl-CoA and OTEA), and the reaction mixture was monitored by LC-MS. The OTEA-C₅N was successfully produced (Figure 6B trace ii), and was also detected using ALA as a starting material instead of glycine and succinyl-CoA (data not shown). Omission of any of the three enzymes completely abolished the production of OTEA-C₅N (Figure 6B trace iv), confirming the necessity of all three enzymes in the biosynthesis and incorporation of the C₅N moiety. In addition, no protein complexes were observed between any of the three enzymes by gel filtration chromatography (data not shown), suggesting that all of the intermediates produced during each transformation are freely diffusible. This is in agreement with the above results that every step can be reconstituted individually in vitro.

Discussion

In this work we have studied the biosynthesis of 2-amino-3-hydroxycyclopent-2-enone (C₅N) and its incorporation via an amide linkage to a partial polyenoate scaffold of the antifungal natural product ECO-02301. The condensation between glycine and succinyl-CoA catalyzed by ORF34 (ALAS) leads to formation of ALA, which is then ligated to CoASH yielding ALA-CoA promoted by the ATP-dependent ORF35 (ACL). ALA-CoA is not stable and can undergo spontaneous intramolecular cyclization with the amine group acting as an internal nucleophile to form the shunt product 2,5-piperidinedione. Alternatively, ALA-CoA can be directed by the PLP-dependent ORF34 to form C₅N through promotion of a carbon nucleophile at C5. Finally, ORF33 (AMS) catalyzes ligation of C₅N to a polyenoic acid in an ATP-dependent manner (Figure 7). We suspect that this three enzyme pathway constitutes a general biosynthetic route for C₅N formation and incorporation into other natural products. This supposition is supported by the presence of the same tri-gene cassette encoding ALAS, ACL and AMS in reported gene clusters of C₅N-containing natural products, including ECO-0501 and moenomycin.^15,31

Figure 7.

Characterized biosynthetic pathway for C₅N involving ORF33-35 from ECO-02301 gene cluster.

The discovery of ALAS-catalyzed cyclization of ALA-CoA to C₅N solves the mechanistic problem regarding the formation of C₅N from ALA. Interestingly, the ALAS enzyme catalyzes two distinct reactions in this pathway: (1) ALA synthesis from glycine and succinyl-CoA, and (2) cyclization of ALA-CoA to afford C₅N. ALAS belongs to the family of α-oxoamine synthases and, because of its long known role in providing building blocks for heme biosynthesis, its PLP-dependent catalytic mechanism has been studied extensively. Catalysis involves two bond cleavage steps (C_α-H and C_α-COO) and one bond forming step (between C_α and the thioester carbonyl of succinyl-CoA).³² To initiate catalysis binding of glycine induces transaldimination, leading to the formation of Gly=PLP aldimine complex after deprotonation of the C_α-H_R of glycine by the ε-NH₂ of a lysine residue in the active site. This generates a C_α=PLP carbanion stabilized as the para-quinonoid species. The electrophilic thioester carbonyl is then attacked by the C_α carbanion of the activated Gly=PLP, which yields a proposed α-NH₂-β-ketoadipyl=PLP intermediate upon C-C bond formation from a thio-Claisen type condensation with concomitant release of CoASH. The subsequent decarboxylation of the β-keto acid via C_α-COO⁻ cleavage and C_α reprotonation lead to the formation of PLP bound ALA that is finally released by transaldimination with the active site lysine (Figure 8A).

Figure 8.

Proposed PLP-dependent catalytic mechanism of ORF34. (A) Mechanism for ALA synthesis. In the substrate-free state, PLP is bound by enzyme (internal aldimine). The incoming glycine induces transaldimination, leading to the PLP=Gly adduct (external aldimine). The electrophilic substrate succinyl-CoA is attacked by the PLP-activated glycine, leading to C-C bond formation with the release of CoASH. Decarboxylation and reprotonation result in formation of the PLP=ALA adduct that is finally released to regenerate the internal aldimine. (B) Mechanism for C₅N synthesis. ALA-CoA induces transaldimination, yielding the ALA-CoA=PLP complex. The thioester carbonyl in ALA-CoA is then attacked by the PLP-activated ALA moiety intramolecularly, leading to C-C bond formation with the release of CoASH. The PLP=C₅N complex is finally released to regenerate internal aldimine. (C) Three-dimensional structural model of the ORF34 active site. The structure was modeled on the R. capsulatus ALAS using the online program HHpred, and the substrates were docked using the software GOLD.

We propose a related mechanism for the novel cyclization of ALA-CoA (Figure 8B). The CoA moiety of ALA-CoA is assumed to be a key recognition element, and after its binding in the well-defined CoA binding site,³³ the terminal -NH₂ of ALA-CoA can reach the PLP-Lys internal aldimine and undergo transaldimination (see SI figure 12). The C5 proton of the ALA moiety can be abstracted (presumably by the same lysine residue) as proposed for the Gly=PLP intermediate in the first step of ALA formation.³⁰ Now the C5-ALA=PLP stabilized carbanion can intramolecularly attack the thioester carbonyl, releasing CoASH as a leaving group. The immediate result is the cyclization of the ALA chain to the C₅N group in aldimine linkage to the PLP in the active site. Transaldimination by lysine would release free C₅N and regenerate the resting form of the enzyme. Compared to ALA synthesis by this same PLP enzyme, cyclization of ALA-CoA does not have the decarboxylation step, and the nucleophilic attack is intramolecular instead of intermolecular, which may account for the observed higher catalytic efficiency of the enzyme for cyclization. This is an exquisite example of a PLP-dependent mechanism being employed by Nature to sequester and deactivate the reactive terminal amine of ALA and to simultaneously activate the adjacent carbon to generate a competent carbon nucleophile.

X-ray structures of the Rhodobacter capsulatus ALAS have been recently reported, showing the positions of both Gly=PLP and of succinyl-CoA.³³ The electron density for the 3’,5’-ADP portion of the CoA and the carboxylate of the succinyl moiety was well resolved, allowing the placement of the C_α of the Gly=PLP adduct within 3 angstroms of the thioester carbonyl when modeled into the active site. Using the solved ALAS structure as the template (PDB: 2BWN), the structure of ORF34 was modeled. It showed high similarity to and shared almost identical active site with ALAS from R. capsulatus (Figure S12). ALA-CoA was docked into the active site of ORF34 with the CoA adenosine binding site constrained at the same location as that of succinyl-CoA in the R. capsulatus ALAS structure. In the modeled ALA-CoA=PLP complex, C5 of the ALA moiety was positioned approximately 3 angstroms from Lys₂₄₈ and the thioester carbonyl, demonstrating that the ALA moiety can readily fit into the active site pocket for cyclization while retaining the same CoA adenosine binding site at the protein surface (Figure 8C). Based on the high similarity of protein structures, we further propose that the “cyclase” activity of the ALAS may be general for all of the ALASs, including those from primary metabolism involved in heme biosynthesis. Such an activity would, however, be revealed only in a cell where ALA has been processed to the ALA-CoA thioester.

The two other enzymes in the pathway (ORF35 and ORF33) are both confirmed to be AMP-forming enzymes. ORF35 catalyzes the ligation between acid and CoA substrates to form the C–S thioester bond, while ORF33 generates the connectivity between acid and amine substrates to afford the C–N amide bond found in the natural product. ORF35 showed substrate promiscuity towards various acids and demonstrated higher specificity toward fatty acids than ALA, indicating that the enzyme might have evolved from a fatty acid acyl-CoA ligase. In contrast, ORF33 had low affinity for all the acid substrate analogs tested, suggesting the strong preference of the enzyme towards longer chain polyketide acids. In addition, unlike many AMP-forming enzymes including acyl-CoA ligases and adenylation domains of NRPSs which activate acids to yield acyl-AMP intermediates independently in the first half reaction, ORF33 activates acids much more efficiently upon the binding of the amine, suggesting that C₅N binding likely induces conformation change to render the enzyme in a more favorable state for acid activation.

In addition to the manumycin family compounds, many bioactive metabolites, including reductiomycin (antitumor)³⁴, senacarcin (antitumor)³⁵, limocrocin (reverse transcriptase inhibitor)¹⁰, virustomycin (antivirus)³⁶, enopeptin (antibiotic)³⁷ and Sch 725424 (antibiotic)¹¹ also contain the C₅N ring system. Although there has been no direct study of the structure-activity relationship of C₅N to date, the prevalence of this functionality in natural products with a broad range of biological activities suggests a critical role in efficacy. This is especially apparent in the cases such as limocrocin, Sch 725424 and reductiomycin (Figure 1), where C₅N is the major functional group in these structures. It is possible that C₅N serves as a Michael acceptor for covalent inactivation of enzymes, which is a common inhibition mechanism of many other natural and synthetic molecules, including resorcylic acid lactones and acrylamides.^38,39 Our work presented here successfully closes the knowledge gap around the biosynthesis of C₅N, reveals a novel cyclization activity of ALA synthase, and defines a minimal, presumably portable cassette for the formation and amide-mediated incorporation of this unique pharmacophore.