Pathogenic bacteria remodel central metabolic enzyme to build a cyclopropanol warhead

Bacteria of the Burkholderia pseudomallei (BP) group pose a global health threat, causing the infectious diseases melioidosis, a common cause of pneumonia and sepsis, and glanders, a contagious zoonosis. A trait of BP bacteria is a conserved gene cluster coding for the biosynthesis of polyketides (malleicyprols) with a reactive cyclopropanol unit that is critical for virulence. Enzymes building this warhead represent ideal targets for antivirulence strategies but the biochemical basis of cyclopropanol formation is unknown. Here we describe the formation of the malleicyprol warhead. We show that BurG, an unusual NAD+-dependent member of the ketol-acid reductoisomerase family, constructs the strained cyclopropanol ring. Biochemical assays and a suite of eight crystal structures of native and mutated BurG with bound analogues and inhibitors provide snapshots of each step of the complex reaction mechanism, involving a concealed oxidoreduction and a C–S bond cleavage. Our findings illustrate a remarkable case of neofunctionalisation, where a biocatalyst from central metabolism has been evolutionarily repurposed for warhead production in pathogens.


BurG catalyses a transient redox reaction.
The formation of a strained cyclopropane ring markedly differs from the biotransformations known for canonical KARIs. Notably, the genomes of BP bacteria harbour additional KARI genes (ilvC) within the operons for branched-chain amino acid biosynthesis. A phylogenetic analysis ( Fig. 2j and Supplementary Fig. 4) of BurG-like enzymes and native KARIs revealed a separate clade for tentative cyclopropanol synthases within the KARI family. We identified nine other organisms (Fig. 2k) that harbour burG orthologues along with all genes required for gonydiol biosynthesis.
HHpred analyses 22 predicted the typical KARI Rossmann fold domains for these BurG-type enzymes, pointing to the binding of a nucleotide cofactor. Indeed, HR-LCMS analysis of a denatured BurG preparation revealed the presence of the nucleotide cofactor NAD + ( Supplementary Fig. 5). This finding is surprising because KARIs are known to prefer NADPH as cofactor. In this context it is noteworthy that considerable effort has been directed towards engineering KARIs to utilize NADH instead of NADPH 23 and to discover rare NADH-preferring KARIs for biotechnological purposes 24 .
To test whether NAD + is essential for the BurG-mediated cyclopropanol formation we removed the cofactor from the enzyme preparation by using an extensive dialysis protocol. In the absence of NAD + , BurG is incapable of forming 6, yet the activity can be restored by addition of NAD + . Addition of NADH, however, cannot restore the reaction (Fig. 3a). These results provide evidence that bound NAD + is essential for the BurG-catalysed reaction.
To gain more insight into the reaction mechanism and the role and position of NAD + in the enzyme, we performed protein crystallography. Due to the close phylogenetic relation between BurG and the KARIs, we turned to surrogates that had proven successful in previous studies in obtaining ligand-bound three-dimensional (3D) structures such as the KARI inhibitor cyclopropane-1,1-dicarboxylate 7 (ref. 19 ). We raised crystals of holo-BurG in the absence or presence of 7 as surrogate and solved the structures at a resolution of 1.8 Å and 1.9 Å, respectively. The obtained 3D structure of the complex ( Fig. 3b; PDB 7PCG) clearly displays 7 bound between two magnesium ions with NAD + in close   Fig. 3 for chromatogram. g, BurG-catalysed transformation of 5 to 6. h, Monitoring of enzymatic turnover of isolated 5 and synthetic rac-syn-5 or rac-anti-5 into 6 via HR-LCMS. i, Steady-state kinetics of formation of 6 from rac-anti-5; mean and standard deviation obtained from two independent preparations (n = 2). j, Phylogenetic analysis of BurG, orthologues and canonical KARIs. The scale bar indicates amino acid substitutions per site. k, Alignment of biosynthetic gene cluster encoding for BurG orthologues clustered with the genes associated with biosynthesis of 5.
proximity, thus enabling its participation in biocatalysis (Fig. 3b). The amino acid residues Asn48 and Asp51 exclude the negatively charged phosphate group of NADP + from binding in the active centre of BurG (Extended Data Fig. 5). A co-crystallized glycerol molecule (used as cryoprotectant) maps the entire specificity pocket next to the catalytic centre. The positioning of NAD + in the active site of BurG indicates that an oxidation event occurs during cyclopropanol formation. Monitoring the BurG assays by HR-LCMS, however, did not show the presence of any oxidized derivatives of 5 or 6 (or any other pathway intermediates). Therefore, BurG is either not redox proficient or a transient redox step occurs during its catalytic cycle. Such a cryptic (or covert) redox reaction would first form NADH by substrate oxidation, followed by NAD + regeneration through ligand reduction.
To investigate the redox proficiency of BurG and its ability to regenerate NAD + from NADH we devised an absorbance-based assay with the substrate surrogate hydroxypyruvate 8. BurG was incubated with 8 and NADH and the light absorption of the dinucleotide at 340 nm was monitored spectrophotometrically. A clearly observed signal decrease corresponds to the formation of NAD + and glyceric acid 9, the reduced form of 8 (Fig. 3c). In substrate saturation kinetics we determined a k cat of 16.3 ± 1.1 min −1 (Fig. 3c and Supplementary Fig. 6) for this reaction, which resembles the k cat for ketone reduction by other KARIs 19 . Further evidence for the redox proficiency of BurG was gleaned from protein crystallization experiments with the previously co-crystallized 19 KARI inhibitor N-hydroxy-N-isopropyloxamic acid (IpOHA, 10) in the presence of NADH ( Fig. 3d; PDB 7PCL). The 3D structure at 2.05 Å resolution unexpectedly shows that the hydroxamate 10 was reduced to the corresponding hemiaminal 11, as evidenced by the tetrahedral geometry of the surrogate in the 3D structure (Fig. 3e). Notably, the stereochemical course of the reduction is identical with the native product of BurG. Taken together, BurG is a redox-proficient enzyme that promotes cyclopropanol formation via a transient and thus concealed oxidation event. This finding is surprising because characterized enzymes that recycle their NAD + cofactors are scarce [25][26][27] .
To provide further experimental evidence for the proposed transient redox cycle (Extended Data Fig. 6) we captured the enzyme in a state with bound NADH after the oxidation event occurred. Therefore, we synthesized the carba analogue of gonydiol, rac-12, which cannot be converted into a cyclopropanol (Fig. 3f), and performed protein crystallography. Analysis of the obtained 3D structure with a resolution of 2.05 Å of BurG with rac-12 revealed that the surrogate had been converted into the oxidized form 13 that is complexed between the two magnesium atoms and NADH in the active site of the enzyme ( Fig. 3g; PDB 7PCM). The planar geometry of the complexed ligand indicates an sp 2 hybridization of both the α and β carbons, indicating that 13 is stabilized in the enolic form. The transformation of rac-12 into the trapped enolate 13 provides evidence for an intermediary oxidation during the catalytic cycle of BurG. We obtained similar results in co-crystallization assays with BurG, NAD + and the substrate surrogate hydroxypyruvate 8. By analogy to 13, the crystal structure depicts a planar geometry of 8 indicating it is bound in its enolized form to the active site of BurG (Extended Data Fig. 7; PDB 7PCI).
Mechanism of cyclopropanol formation. Our structural analysis revealed that the carboxylate of a glutamic acid side chain (E232) is in close proximity (3 Å) to the β carbons of the substrate surrogates and thus probably facilitates the deprotonation event necessary for enol formation. To probe the function of this putative catalytic residue we generated a point-mutated variant (E232Q) of BurG and assayed its activity with the native substrate gonydiol. As predicted, the mutation hampers the transformation of gonydiol into the cyclopropanol derivative 6 (Fig. 3h) for the NADH-mediated reduction of 8 ( Supplementary Fig. 7). Co-incubations of BurG E232Q with rac-12 yielded crystals for which we solved a 3D structure at a resolution of 1.55 Å. In contrast to native BurG, rac-12 was not oxidized by BurG E232Q, as indicated by the tetrahedral geometry of the complexed surrogate ( Fig. 3i; PDB 7PCO). The absence of oxidized 13 in BurG E232Q may indicate a role of E232 in shifting the internal redox equilibrium towards the oxidized state by deprotonating the substrate and forming the enolate. By analogy to glycerate oxidation, the oxidized state is probably energetically less favoured. 28 The obtained near-atomic resolution structure with BurG E232Q and 12 enabled us to analyse the bound ligand and determine its absolute configuration (2R,3R). The observed chirality of the bound substrate confirmed our stereochemical analysis that gonydiol had the 2R,3R configuration. Apart from its function as a catalytic base, E232 also plays an active role in substrate recognition and positioning as we gleaned from crystallization assays with gonydiol and BurG. We solved a 3D structure (at 1.6 Å) for which the electron density map shows a mixed population of two subsets of the converted substrate ( Fig.  4a and Extended Data Fig. 8; PDB 7PCN). In one subset the active centre is populated by the enzymatic reaction products 6 and DMS. The second subset clearly depicts the backbone of gonydiol, which was oxidized into its enolate 14. Intermediate 14 is firmly held in place by ionic interaction of its sulfonium moiety with E232 and cation-π interaction with Y133 ( Fig. 4b). In contrast, the neutral carba analogues 12 and 13 show a higher degree of freedom in their tail regions (Fig. 4c) due to the lack of ionic placement as visible in their crystallographic B factor.
The head region (C1-C3) of all co-crystallized ligands was consistently found to be firmly locked in a conserved binding mode that is facilitated by interactions with the two Mg 2+ ions of the enzyme (see Fig. 4d,e for unliganded holo structure). In this head region, BurG shows remarkable similarity with regards to substrate structure and binding when compared to canonical KARIs (Fig. 4f). The 3D structure obtained for the holo version of BurG (PDB 7PCC) illustrates that the binding sites of the Mg-bound oxygen functionalities are populated by aqueous residues when no ligand is present (Fig. 4e). In addition to similarities in substrate binding and redox capability, canonical KARIs are thought to contain a cyclopropane transition state 19 in their catalytic cascade (Fig. 4g). In the case of BurG, formation of 6 is probably driven by a gain in entropy and removal of the volatile DMS from the chemical equilibrium after cyclopropanol synthesis. Moreover, the enzyme surrounds the cationic sulfonium group of the substrate with a hydrophobic binding pocket lined with Met, Tyr and Val residues, which favour loss of the charge and thus promote catalysis (Fig. 4b).
Based on these results, we devised a plausible reaction mechanism for the BurG-mediated cyclopropanol formation (Fig. 4h). First, by analogy to the initial step in the catalytic cycle of canonical KARIs 29 , the hydroxide bridged between the magnesium centres (shown in the ligand-free holo structure of BurG; Fig. 4e), would deprotonate the α-hydroxy moiety of 5, thus forming the corresponding alkoxide. Next, the two magnesium ions would complex the alkoxide and position it above the pyridinium moiety of the NAD + cofactor. In addition, the glutamate residue (E232) would ionically interact with the positively charged sulfonium group to firmly position the substrate in the binding pocket. Subsequent hydride transfer from the C-α of gonydiol to NAD + (yielding  through interaction with Mg 2+ ions in the active centre of BurG. W apo indicates the binding sites of water molecules as found in the apo structure that are replaced by ligand atoms in the depicted structures. e, Structure of unliganded holo-BurG illustrates a hydroxide ion bridged between two Mg 2+ centres (PDB 7PCC). f, Comparison of BurG and KARIs displays a similar binding mode of substrates in the Mg 2+ cluster. g, Proposed reaction mechanism 19 for KARIs with cyclopropane transition state. h, Reaction mechanism proposed for BurG involving a concealed (transient) redox reaction. NADH), and deprotonation at the β carbon by E232 would result in an enolate intermediate (14), as mimicked by hydroxypyruvate and the oxidized carba-gonydiol analogue 13. The ideal positioning of C-α relative to the hydride-accepting NAD + -carbon (Extended Data Fig. 9) renders an alternative mechanism involving oxidation at C-β unlikely. Quenching of the negative charge at E232 through proton transfer would disrupt the ionic interaction, destabilize the sulfonium and trigger cyclization. Next, an intramolecular nucleophilic attack at the δ position would result in cyclopropyl ring formation and C-S bond cleavage followed by hydride transfer from NADH to the si face of the C-α atom. The proposed stereochemical course of hydride transfer is in line with the determined stereochemistry of the final product 6 and the orientation found of all surrogates in the active site of the enzyme (Fig. 4a). This key reductive step would regenerate the NAD + cofactor, leaving the overall redox reaction neutral and retaining the configuration at C2. Our structural analysis indicates that the outlined reaction sequence takes place in a closed conformation of the active site (see Extended Data This enzymatic envelope would protect the oxidized and highly reactive intermediates from unwanted side reactions. Hence, final expulsion of the reaction products from the active site would prepare the enzyme for the start of a new catalytic cycle.

Discussion
We reason that the KARI-fold is ideally positioned as enzymatic ancestor from primary metabolism to be repurposed for gonydiol binding, a transient redox reaction, and cyclopropanol construction by elimination of DMS. The fact that canonical KARIs facilitate their reaction sequence to branched carbon backbones through a cyclopropane transition state hints to the evolutionary trajectory (neofunctionalisation) towards the cyclopropanol synthase BurG. Inheritance of the redox capability of KARIs allows BurG to activate its substrate gonydiol as the electrophilic intermediate 14. This activation is vital for intramolecular electrophilic displacement of DMS as leaving group and subsequent cyclopropanol formation. Most characterized biosynthetic pathways towards cyclopropyl 30-35 moieties, as in the genotoxin colibactin 36,37 , for example, also employ electrophilic displacement reactions 38 . Generation of the enolic nucleophile, however, is commonly achieved via a thioester linkage between enzyme and substrate, and in all studied cases either methylthioadenosine or chloride are employed as leaving groups. The KARI-mediated reaction discovered here may inspire chemical engineers who seek to incorporate highly strained rings in a biocatalytic fashion 39 . Our work offers the prospects of finding new cyclopropanol-forming enzymes and the corresponding natural products. More importantly, our mechanistic and structural study of the key enzyme responsible for malleicyprol biosynthesis sets the basis for selectively inhibiting the formation of virulence-conferring toxins from pathogenic B. pseudomallei complex bacteria.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41557-022-01005-z.  Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/. © The Author(s) 2022
Suicide plasmid preparation for B. thailandensis mutant. A gene fragment (2,349 bp) containing burG was PCR-amplified using primers Isomerase-fw and Isomerase-rv with 5 Prime Extender Polymerase (VWR International). The amplicon was purified using the GFX PCR DNA and Gel Band Purification Kit (GE Healthcare), and the obtained DNA was cloned into the pGEM T-easy vector (Promega). The resulting plasmid (pGEM-Isomerase) was verified by sequencing (GATC biotech) and restricted with the restriction enzyme SmaI and treated with SAP (Shrimp Alkaline Phosphatase, New England Biolabs). pGEM-Kan was restricted by NotI and blunted by Klenow treatment. This blunted kanamycin resistance cassette gene was then cloned into the SmaI-restricted plasmid, generating the Kan-inserted suicide plasmid pGEM-burG::Kan.
Preparation of gene insertion mutant B. thailandensis Pbur burG::Kan. B. thailandensis E264 Pbur was precultured in LB medium supplemented with tetracycline (45 μg ml −1 ) for 16 h at 30 °C. The cultured cells were inoculated in LB medium with tetracycline (45 μg ml −1 ) and grown at 30 °C until an OD 600 of 0.4-0.6 was reached. Subsequently, the culture broth was centrifuged and the supernatant was removed. Precipitated cells were resuspended in sucrose solution (300 mM) and centrifuged. After repeating this washing step three times, the washed cells were resuspended in 300 mM sucrose and subjected to electroporation (200 kV, Eporator, Eppendorf) with pGEM-burG::Kan (1-10 µg). The transformed cells were then cultured in LB medium (1 ml) for 4 h at 30 °C with shaking and subsequently plated on LB-agar plates supplemented with tetracycline (45 µg ml −1 ) and kanamycin (150 μg ml −1 ). After 3 days, positive colonies were observed and confirmed by colony PCR with the primer pair Isomerase fw2 and Isomerase rv2 using Taq Polymerase (New England Biolabs). See Supplementary Fig. 9 for PCR verification.
Preparation of His 6 -tagged BurC. The gene encoding BurC was amplified by PCR using the primer pair burC4-fw-NdeI and burC-rv-EcoRI. The obtained PCR product was subsequently ligated into the pCR2.1 vector. After sequence analysis (GATC biotech), the gene fragment was restricted with NdeI and EcoRI and then cloned into the NdeI and EcoRI sites of pET28a(+) vector, generating pET28a-burC. To prepare a seed culture, E. coli BL21 (DE3) pET28-burC was grown overnight in LB medium with 50 μg ml −1 of kanamycin. An aliquot (2.5 ml) of this seed culture was used to inoculate 250 ml LB medium supplemented with kanamycin (50 μg ml −1 ) in a 1 litre baffled Erlenmeyer flask. The culture was grown at 37 °C with orbital shaking until an OD 600 of 0.5-0.7 was reached. Protein production was induced by addition of isopropylthiogalactoside (IPTG; 0.5 mM as a final concentration) followed by incubation at 15 °C for 18 h. Cells were harvested by centrifugation (9,000g, 4 °C, 10 min) and frozen at −20 °C until further use. For protein extraction, the cells were resuspended in 50 mM Tris-HCl (pH 8.0), 200 mM NaCl, 10% glycerol, 0.5 mM dithiothreitol, lysozyme (200 μg ml −1 ), and phenylmethylsulfonyl fluoride (100 μM), followed by incubation at 37 °C for 30 min. Then, the cells were lysed using a sonicator (Bandelin Sonopuls HD2200), and centrifuged (9,200g, 4 °C, 35 min) to obtain a clear lysate. The cleared lysate was loaded onto a Ni-IDA agarose (Biontex) column with manual handling. Equilibration and washing buffers were 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 10 mM imidazole and 30 mM imidazole, respectively; at least 10 column volumes were used for each step. The elution was carried out using a stepwise increase to 500 mM imidazole in the same buffer. The obtained fractions were analysed for target protein by SDS-polyacrylamide gel electrophoresis (SDS-PAGE; molecular weight of His 6 -BurC, 39.9 kDa). Protein containing fractions were dialysed against 20 mM sodium phosphate buffer (pH 7.2) and 200 mM NaCl, and concentrated using an Amicon Ultra-30K concentrator (Merck Millipore). See Supplementary  Fig. 10 for SDS-PAGE analysis.
Preparation of His 6 -tagged BurG. The gene encoding BurG was amplified by PCR using the primer pair burG-fw-NheI and burG-rv-HindIII. The obtained amplicon was ligated into the pJET1.2 vector and verified by sequence analysis (GATC biotech). Subsequently, the gene fragment was restricted with NheI and HindIII and then cloned into the NheI-and HindIII-restricted pET28a(+) vector, generating pET28a-burG. To prepare a seed culture, E. coli BL21 (DE3) pET28a-burG was grown overnight in LB medium supplemented with 50 μg ml −1 of kanamycin. An aliquot (2.5 ml) of this seed culture was used to inoculate 250 ml LB medium supplemented with kanamycin (50 μg ml −1 ) in a 1 litre baffled Erlenmeyer flask. The culture was grown at 37 °C with orbital shaking until an OD 600 of 0.5-0.7 was reached. Protein production was induced by addition of IPTG (0.5 mM as final concentration), followed by incubation at 15 °C for 18 h. Cells were harvested by centrifugation (9,000g, 4 °C, 10 min) and frozen at −20 °C until further use. For protein extraction, the cells were resuspended in 50 mM Tris-HCl buffer (pH 8.0) supplemented with 200 mM NaCl, 10% glycerol, 0.5 mM dithiothreitol, lysozyme (1 mg ml −1 ) and PMSF (100 μM), followed by incubation at 37 °C for 30 min. Then, the cells were lysed using a sonicator (Bandelin Sonopuls HD2200), and centrifuged (9,200g, 4 °C, 35 min) to obtain a clear lysate. The cleared lysate was applied to a Ni-IDA agarose (Biontex) column with manual handling. Equilibration and washing buffers were 50 mM Tris-HCl (pH 7.5), 500 mM NaCl, 10 mM imidazole or 30 mM imidazole, respectively; at least 10 column volumes were used for each step. The elution was carried out using a stepwise increase to 500 mM imidazole in the same buffer. The obtained fractions were analysed for target protein by SDS-PAGE (molecular weight of His 6 -BurG, 40.6 kDa). Protein containing fractions were dialysed against 50 mM Tris-HCl (pH 7.5), 200 mM NaCl buffer and concentrated using an Amicon Ultra-30K concentrator (Merck Millipore). See Supplementary Fig. 10 for SDS-PAGE analysis.
Generation of pET28a-burG_E232Q. A codon-optimized, point-mutated synthetic burG E232Q gene fragment was obtained from Integrated DNA Technologies. This gene fragment was ligated into a pJET1.2 vector (Thermo Fisher Scientific) and introduced into E. coli XL1 Blue by electroporation. The obtained plasmid pJET-burG_E232Q was verified by sequencing (GeneWiz) followed by restriction with HindIII and NheI. This obtained gene fragment was cloned into an HindIII-and NheI-restricted pET28a vector, yielding pET28-burG_ E232Q. (See Supplementary Information for sequence.) Thermal stability of BurG and BurG E232Q. The thermal stability of BurG and BurG E232Q protein preparations was determined by using a Tycho NT.6 (NanoTemper Technologies) monitoring the fluorescence of each protein preparation at 330 and 350 nm during a thermal ramp. Inflection temperatures (T i BurG, 60.5 ± 0.1 °C; T i BurG E232Q, 61.85 ± 0.05 °C) were obtained in technical duplicates at a concentration of 70 µM enzyme supplemented with 2 mM MgSO 4 and 4 mM NAD + .

Dialysis of BurG preparations.
A preparation of His 6 -tagged BurG was diluted with denaturing buffer (pH 8.0) to a final concentration of 8 M urea, 100 mM Tris-HCl and 200 mM NaCl. The resulting solution (1 ml) was pipetted into a dialysis bag (Spectra-Por, 3,500 dalton molecular weight cut-off, Spectrum Labs) and the bag was incubated in 3 litres of the same buffer as mentioned above at 4 °C for 24 h. Regular buffer exchange resulted in a final dilution factor of 6.4 × 10 17 . Next, the denaturing buffer was exchanged for refolding buffer (0.5 M l-arginine, 50 mM Tris-HCl and 100 mM NaCl, pH 8.0) followed by incubation for 12 h at 4 °C with regular buffer exchange to remove residual urea. This was followed by the removal of l-arginine through a further buffer exchange step to reaction buffer (50 mM Tris-HCl and 100 mM NaCl, pH 8.0) and incubation for 5 h at 4 °C with regular buffer exchange. Complete removal of NAD + was validated by analysis of the refolded protein preparation via LC-HRMS. The resulting NAD + -free solution was subjected to further gonydiol transformation assays.
Substrate saturation kinetics (spectrophotometry). The reduction activity of BurG (E232Q) was measured with hydroxypyruvate and NADH/NADPH as substrates in varying concentrations in semi-micro disposable ultraviolet cuvettes (Brand) with a final reaction volume of 0.9 ml. Purified tag-free BurG (E232Q) as prepared for protein crystallization (1 µM final concentration) was added to Tris-HCl buffer (100 mM, pH 7.5) containing MgSO 4 (10 mM) and NADH/ NADPH and the resulting mixture incubated at 30 °C for 5 min in a UV-1800 spectrophotometer (Shimadzu). Subsequently, hydroxypyruvate was added, and the absorption at 340 nm was monitored spectrophotometrically during the incubation at 30 °C. Each substrate concentration was measured in duplicate with enzyme obtained from independent preparations. No decrease in signal was observed when NADPH was used instead of NADH under similar conditions. The molarity of NADH at each time point measured in 1 s intervals was calculated by using a calibration curve obtained on the same instrument. Michaelis-Menten parameters were determined with one substrate at a fixed concentration (2 mM hydroxypyruvate or 0.1 mM NADH) while varying the second substrate at concentrations ranging from 0.020 to 1.5 mM for hydroxypyruvate (0.020-9 mM in the case of BurG E232Q) and from 6.68 to 75 µM for NADH. Notably, substrate inhibition occurred when concentrations higher than 2.25 mM hydroxypyruvate were used with wild-type BurG. Initial velocities were determined from the slope of

Substrate saturation kinetics (LC-HRMS).
The kinetic parameters of the formation of 6 from 5 as catalysed by BurG were determined with the same Orbitrap LC-HRMS system as described below. Enzyme reactions consisted of Tris-HCl buffer (100 mM, pH 7.5), MgSO 4 (10 mM), NAD + (1 mM) and purified tag-free BurG as prepared for protein crystallization (1.5 µM final concentration), and were incubated in 1.6 ml glass vials with a glass insert at 30 °C in the autosampler of the ultra-high-performance liquid chromatography (UHPLC) system. To start the enzymatic reaction, synthetic rac-anti- 5  Metabolite extraction for comparative LC-HRMS analysis. B. thailandensis Pbur and B. thailandensis Pbur burG::Kan were independently grown in 300 ml baffled Erlenmeyer flasks filled with 100 ml of MM9 medium that had been supplemented with either tetracycline (45 μg ml −1 ) or tetracycline (45 μg ml −1 ) plus kanamycin (150 μg ml −1 ) at 30 °C with orbital shaking for 24 h. Subsequently, the cells were harvested from a 50 ml aliquot by centrifugation (8,000g, for 15 min). The resulting cell pellet was resuspended in 25 ml methanol, sonicated and subsequently incubated for 60 min at room temperature. The centrifugation and extraction steps were repeated a second time and both obtained fractions were combined and concentrated in vacuo to yield crude extracts. These extracts were dissolved in 4 ml of methanol, and 100 µl of the resulting solution was then diluted with 200 µl methanol and filtered through a PTFE syringe filter. Subsequently the extracts were subjected to LC-HRMS analysis for a metabolomics analysis using the software Compound Discoverer 2.1 SP1 (Thermo Fisher Scientific). Both genotypes were analysed using a metabolomics workflow with and without a pattern scoring node to filter for sulfur-containing metabolites. The obtained metabolic profiles were compared using differential analysis.
Chemical complementation of B. thailandensis Pbur burG::Kan. Trigonic acid 6 (400 µM final concentration) was added to MM9 medium (75 ml in a 300 ml baffled Erlenmeyer flask) supplemented with tetracycline (45 μg ml −1 ) and kanamycin (150 μg ml −1 ) and inoculated with a preculture of B. thailandensis Pbur burG::Kan to a final OD 600 of 0.1. Controls consisted of B. thailandensis Pbur (omitting the kanamycin) and B. thailandensis Pbur burG::Kan without addition of 6 prepared in the same way. All cultures were incubated at 30 °C with orbital shaking for 24 h. The obtained cultures were extracted with EtOAc three times, and the organic phase dried over Na 2 SO 4 and concentrated under reduced pressure. The obtained crude extracts were redissolved in methanol, filtered through a PTFE syringe filter and subjected to LC-HRMS. Similar results were obtained on independent days with independent starting cultures (biological triplicates in total).
In vitro hydroxylation of gonyol. Purified His 6 -BurC (400 nM final concentration) was added to Tris-HCl buffer (50 mM, pH 7.5) containing ascorbate (0.1 mM), α-ketoglutarate (1 mM) and FeSO 4 (0.1 mM). The reaction was started by addition of R-or S-gonyol (4; each as TFA salt; 150 µM final concentration). As a control, purified His 6 -BurC was heat-inactivated at 80 °C for 25 min and used in the same way. The resulting mixtures were incubated at 30 °C and aliquots taken at various time points (20,50,180, 210 and 360 min). Each aliquot was diluted with the same volume of methanol (typically 12.5 µl), filtered through a PTFE syringe filter and directly subjected to LC-HRMS analysis (Extended Data Fig. 2). Similar results were obtained on independent days with independent protein preparations (biological triplicates in total).
In vitro cyclopropanol synthesis. Purified His 6 -BurG or His 6 -BurG E232Q (5 µM final concentration) were added to Tris-HCl buffer (100 mM, pH 7.5) containing MgSO 4 (2 mM) and NAD + (1 mM) and the resulting mixture incubated at 30 °C for 5 min. Subsequently anti-, syn-or natural gonydiol (5; each as TFA salt; 1.3 mM final concentration) were added followed by incubation at 30 °C with aliquots taken at various time points (typically after 60 min). Each aliquot was diluted with the same volume of methanol, then filtered through a PTFE syringe filter and directly subjected to LC-HRMS analysis.
Mg 2+ can be substituted with Co 2+ (in the form of CoSO 4 ) to achieve formation of trigonic acid. However, Mg 2+ cannot be substituted with Mn 2+ or Zn 2+ . Sole addition of Mg 2+ suffices (when no dialysis under denaturing conditions is performed during enzyme purification) to observe enzymatic transformations and no addition of NAD + is necessary. However, NAD + was always detected via LC-HRMS in BurG preparations unless dialysis under denaturing conditions followed by refolding of the protein were performed (see above). Dialysed and thus NAD + -free BurG was used in gonydiol transformation assays as outlined above with and without addition of NAD + or NADH. Assays performed with tag-free BurG/BurG E232Q as prepared for protein crystallography (Protein crystallography) yielded similar results as observed for His 6 -tagged BurG. Similar results were obtained on independent days with independent protein preparations (assay repeated more than three times in biological replicates).

Headspace SPME-GC-MS detection of DMS.
Purified tag-free BurG as prepared for protein crystallization (5 µM final concentration) was added to Tris-HCl buffer (100 mM, pH 7.5) containing MgSO 4 (2 mM) and NAD + (1 mM). Subsequently synthetic rac-anti-gonydiol 5 (as TFA salt, 2 mM final concentration) was added to give a final volume of 200 µl. The resulting reaction mixture was pipetted into a 1.5 ml Eppendorf tube without lid which was placed inside a 20 ml headspace vial. The vial was sealed and incubated at 30 °C for 80 min. A control reaction was carried out in the same way as detailed above with enzyme preparation that had been denatured at 80 °C for 20 min. For headspace SPME a carbon wide range/polydimethylsiloxane fibre 95 µm thick and 10 mm long (Thermo Fisher Scientific) was used. The fibre was incubated at 260 °C for 4 min prior to headspace extraction. The headspace vial was incubated at 70 °C for 5 min followed by headspace extraction with the fibre for 5 min at 70 °C. For analysis, a TRACE 1310 gas chromatograph (Thermo Fisher Scientific) fitted with a BPX5 capillary column, 30 m × 0.25 mm × 0.25 µm (Trajan) coupled to a TSQ 900 mass spectrometer (Thermo Fisher Scientific) was used. The SSL injector was equipped with a topaz split liner (1 mm × 6.3 mm × 78.5 mm, Restek) heated to 250 °C and used in splitless mode. Helium was used as carrier gas at a flow rate of 0.6 ml min −1 . The GC oven was set to a temperature of 30 °C with a hold time for 4 min followed by heating to 200 °C for 3.4 min and a subsequent hold time at 200 °C for 2.6 min. The source and transfer line temperatures of the mass spectrometer were 230 and 250 °C, respectively. The scan range of the mass spectrometer was set to 10-300 m/z. DMS was detected via the extracted ion chromatogram at an m/z of 62, and its identity validated through a database search (NIST 2.4) of the obtained electron ionization spectrum and through its identical retention time when compared with a commercial standard.
Isolation of gonydiol 5. Cells obtained from a 2 litre culture of B. thailandensis Pbur burG::Kan were extracted with methanol (2 × 200 ml) and the solvent subsequently removed under reduced pressure. The obtained crude extract was then redissolved in H 2 O (300 ml) to form a turbid solution which was pelleted by centrifugation at 6,000g for 45 min to yield a clear yellow solution. This solution was then loaded onto 100 g of the cationic exchange resin Amberlite IR-120 in the H + state. After washing with 300 ml of H 2 O, bound cations were eluted with 1 litre of HCl (1 M) and the solvent removed through lyophilization. The concentrated eluate was subsequently redissolved in water and further purified via repeated rounds of preparative HPLC using a Phenomenex Synergi Fusion-RP C 18 column (80-4, 250 × 21.2 mm) with H 2 O + 0.1% TFA as mobile phase at a flow rate of 20 ml min −1 . Fractions containing the target molecule (as judged by NMR and LC-HRMS analysis) were pooled, concentrated in vacuo and further purified via repeated rounds of preparative HPLC with a SeQuant ZIC-HILIC column (200-5, 250 × 10 mm, Merck) (solvent A, 98% H 2 O, 2% CH 3 CN + 0.1% HCOOH, pH 2.7; solvent B, 10% H 2 O, 90% CH 3 CN, 5 mM NH 4 OAc; isocratic conditions: 71% B; flow rate, 5 ml min −1 ). Fractions containing the target molecule were pooled, concentrated under reduced pressure and analysed via 1 H NMR spectroscopy. We noted a distinct difference in the 1 H NMR spectra when fractions that were obtained with TFA in the mobile phase were compared to fractions that were obtained with NH 4 OAc/HCOOH as additive. While the TFA salt of 5 showed two signals resonating at 4.19 ppm (1H) and 4.01 ppm (1H), respectively, both signals merged to one signal resonating at 3.91 ppm (2H) when no TFA was present. As it seemed reasonable that a separation of these signals would aid structure elucidation, we converted the obtained substance to its TFA salt by repeatedly (3×) redissolving it in 1 ml of 1% TFA and subsequent removal of the solvent in vacuo. Through this we obtained 2.  LC-HRMS. LC-HRMS measurements were carried out with an UltiMate 3000 UHPLC (Thermo Fisher Scientific) coupled to a Thermo Fisher Scientific QExactive HF-X Hybrid Quadrupole-Orbitrap equipped with an electrospray ion source.
Detection of NAD + (Supplementary Fig. 5 Phylogenetic analysis. Sequences for KARI family proteins were obtained from the NCBI server (https://www.ncbi.nlm.nih.gov/) by Protein BLAST using the databases of non-redundant protein sequences, UniProtKB/Swiss-Prot and PDB. The respective amino acid sequences were aligned using the MAFFT server (https://mafft.cbrc.jp/alignment/server/) 40 with MAFFT v.7 and default settings. An unrooted fast maximum-likelihood-based tree was generated using IQ-TREE (https://www.hiv.lanl.gov/content/sequence/IQTREE/iqtree.html) 41 with automatic model selection mode (ModelFinder) 42 , where the LG+G4 was selected. Phylogenetic bootstrap analysis was performed by ultrafast approximate bootstrap with 1,000 bootstrap replicates 43 . The obtained tree was displayed using MEGA7 44 . See Supplementary Fig. 4 and Table 6 for detailed information.
Reagents, chemical synthesis and compound characterisation. Unless stated otherwise, all reagents were purchased from Sigma-Aldrich or TCI. The procedures for the preparation of synthetic compounds used in this study, and the spectral data, can be found in the Supplementary Information. For preparation of dimethyldioxirane, see ref. 45 .
Chemical complementation of gene inactivation mutant with trigonic acid and gene inactivation mutant analysis: biological triplicates (Fig. 2b).
In vitro cyclopropanol synthesis: assay repeated more than three times with independent protein preparations (Fig. 2h).
Protein concentration determination: technical triplicates for each preparation.
Protein crystallography. Culture conditions. For protein production to be used in protein crystallization experiments, E. coli transformed with a plasmid containing BurG or BurG E232Q was grown in Fernbach shaking flasks at 37 °C containing 3 litres of lysogeny broth supplemented with kanamycin (50 mg l −1 ). At an OD 600 of 0.6, IPTG was added (final concentration 1 mM), and incubation was continued overnight at 20 °C. Subsequently, the cells were harvested by centrifugation ( Table  2). Interestingly, the high-resolution electron density map at 1.6 Å in PDB 7PCN depicts a mixture of 14 and NADH as well as 6, DMS and NAD + at the active site.
Note that 14 and NADH as well as 6, DMS and NAD + were incorporated with an occupancy of 0.5 each. In the data set PDB 7PCO (BurG_E232Q mutant), enol-oxaloacetate 15 was co-purified from the E. coli lysate. These findings are in contrast to BurG (wild-type) and show that the E232Q BurG mutant has a high affinity for the metabolite. All crystal structures have been deposited in the RCSB Protein Data Bank. See Supplementary Table 2 for molprobity 52 clash scores.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article. Fig. 9 | Cofactor and ligand orientation in the active site of BurG and KaRIs. The structural comparison between BurG and KARIs illustrates the preferred reaction trajectories of hydride transfer in the distinct enzymes. The interactions of the ligand with the two Mg 2+ ions (black dots) indicate that C α is the preferred carbon atom for catalysis. a, Structure of the catalytically inactive BurG E232Q mutant in complex with surrogate 12. The ligand adopts ideal orientation for hydride transfer to NAD + (2.3 Å distance, angle between C(NAD + )HC α : 174°, PDB code: 7PCO). b, Structure of BurG co-crystallized with 12 and NAD + . The electron density shows the ligand oxidized to 13 resulting in NADH. Glu232 acts as base and is key to hydride transfer. Distance and angle between the cosubstrate and C α (2.8 Å, C(NADH)HC α : 123°) are unfavourable for the back reaction. The replacement of sulfur by carbon in surrogate 13 prevents the elimination reaction and keeps the metabolite in its position (PDB code: 7PCM). c, Structure of BurG co-crystallized with 10 and NADH. The electron density shows the C α -atom of the ligand in sp 3 configuration. Therefore, the molecule represents the carbinolamine 11 and the cosubstrate is oxidized to NAD + . Distance and angle between the cosubstrate and C α allow for a back reaction (2.1 Å, C(NAD + )HC α : 168°). However, formation of the alkoxide is favoured by the interactions with both Mg 2+ ions (PDB code: 7PCL). d, Active site architecture of the KARI from Staphylococcus aureus (PDB code: 6AQJ). Crystallization of the enzyme was carried out in presence of 10 and NADPH. It is reported 19 that hydride transfer from NADPH to the nitrogen of 10 takes place, hereby forming NADP + and converting the ligand to 16 via water (W) release. The complex structure lacks defined H-bonds between cosubstrate and ligand.

March 2021
Data Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy All data are available in the main text, the Supplementary Information, the source data or via the RCSB Protein Data Bank. The corresponding PDB numbers are 7PCC, 7PCE, 7PCG, 7PCI, 7PCL, 7PCM, 7PCN, 7PCO, 7PCT. NMR raw files of natural gonydiol and MS raw files used for comparative metabolomics analysis (Fig. 2a, Supplementary Fig. 1) are deposited on Zenodo (https://doi.org/10.5281/zenodo.6554506) and available without restrictions.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.
Sample size N/A Data exclusions N/A Replication See the Statistics and Reproducibility Section of the manuscript for detailed information on how often each experiment was replicated.
In the case of protein preparations and the corresponding assays, biological replicates are defined as independent protein preparation which were obtained on independent days from independent protein producing cultures; technical replicates refer to repetition of similar experiments on independent days with protein preparations obtained from the same purification with storage at cryonic conditions.
For analysis of metabolite production by various mutants, biological replicates are defined as culture extract that were obtained from independent bacterial cultures grown from independent cryonic stocks; technical replicates refer to the analysis of the same culture extract in independent chromatographic runs.