Crystal structure, modeling, and identification of key residues provide insights into the mechanism of the key toxoflavin biosynthesis protein ToxD

Abstract

Toxoflavin, a toxic secondary metabolite produced by a variety of bacteria, has been implicated as a causative agent in food poisoning and a virulence factor in phytopathogenic bacteria. This toxin is produced by genes encoded in the tox operon in Burkholderia glumae, in which the encoded protein, ToxD, was previously characterized as essential for toxoflavin production. To better understand the function of ToxD in toxoflavin biosynthesis and provide a basis for future work to develop inhibitors of ToxD, we undertook the identification of structurally and catalytically important amino acid residues through a combination of X-ray crystallography and site directed mutagenesis. We solved the structure of BgToxD, which crystallized as a dimer, to 1.8 Å resolution. We identified a citrate molecule in the putative active site. To investigate the role of individual residues, we used Pseudomonas protegens Pf-5, a BL1 plant protective bacterium known to produce toxoflavin, and created mutants in the ToxD-homolog PFL1035. Using a multiple sequence alignment and the BgToxD structure, we identified and explored the functional importance of 12 conserved residues in the putative active site. Eight variants of PFL1035 resulted in no observable production of toxoflavin. In contrast, four ToxD variants resulted in reduced but detectable toxoflavin production suggesting a non-essential role. The crystal structure and structural models of the substrate and intermediate bound enzyme provide a molecular interpretation for the mutagenesis data.

Graphical Abstract

Introduction

The genus Burkholderia consists of a monophyletic group of the β-proteobacteria and are obligate aerobic, Gram-negative, rod shaped bacteria that are found in a wide variety of habitats.^1–3 These include non-pathogenic soil strains such as B. ambifaria and the more widely known pathogenic strains that can infect humans (B. pseudomallei, melioidosis;⁴ B. cepacia complex in cystic fibrosis patients),⁵ horses and humans (B. mallei, glanders)⁶, and plants (B. gladioli and B. glumae, rice rot).^{7, 8}

The genomes of Burkholderia isolates are typically large but range from 1–11.5 Mbp, and the large genome size is thought to help in genetic adaptability owing to the diversity of habitats that Burkholderia occupy and the extensive number of biotransformations they can catalyze.¹ These large genomes also harbor a sizeable potential for the production of natural products (secondary metabolites),⁹ and in the case of B. pseudomallei, it was shown that the encoded natural products act as virulence factors.^10–12 These include compounds produced by both polyketide synthases and non-ribosomal peptide synthetases. B. gladioli pathovar cocovenenans (formally Pseudomonas cocovenenans), which was identified as the causative agent in rash of food poisonings on Java in the 1930s, produces the polyketide synthase-derived respiratory toxin bongkrekic acid^{13, 14} and additionally the 1,2,4-triazine-containing small molecule toxoflavin (1).^{15, 16} This latter compound is also produced by phytopathogenic strains of B. gladioli⁸ and B. glumae⁷ and is a known virulence factor.¹⁷ Therefore, investigations into the biochemistry of natural product formation in Burkholderia species should have an impact on the development of new anti-infective and anti-virulence strategies for the treatment of Burkholderia infections.

Although 1 is a key virulence factor in the infection of B. glumae and B. gladioli of rice,^{17, 18} we have previously shown that it is also produced by the plant protective bacterium Pseudomonas protegens Pf-5.¹⁹ Recent studies have also identified 1 as a product of Streptomyces hiroshimensis ATCC 53615,²⁰ while chemical activation strategies in the Seyedsayamdost laboratory have demonstrated the production of related metabolites, taylorflavin A (2) and B (3) by S. hiroshimensis A18.²¹ Related metabolites include reumycin (1-demethyltoxoflavin, 1-DMT, 4) and fervenulin (5). Reumycin was shown to be a biosynthetic precursor of toxoflavin during in vitro characterization of ToxA from B. glumae^{22, 23} and could also be an isolation by-product derived from demethylation of toxoflavin. Previous work has also identified the diastereomers 2096A (6) and 2096B (7) from Streptomyces sp. IM 2096²⁴ and pyrizinostatin (8) from Streptomyces sp. strain SA228²⁵ as oxidized and elaborated 1,2,4-triazine natural products (Figure 1).

Figure 1.

1,2,4-triazine natural products isolated from bacteria.

Early investigations into the biosynthetic origins of toxoflavin demonstrated that the azapteridine core was derived from guanine (presumed to be through guanosine triphosphate (GTP)) and glycine²⁶ and validated with stable isotope labeling of the taylorflavins.²¹ This hypothesis was validated by the identification of the toxoflavin biosynthetic gene cluster (BGC) in B. glumae which contains five putative biosynthetic genes (toxA-E, Figure 2).^{27, 28} The putative biosynthetic pathway begins with ToxB, a GTP cyclohydrolase II homologous to RibA, excising the C8 position and cleaving off pyrophosphate to generate 2,5-diamino-6-(5-phospho-D-ribosylamino)pyrimidin-4(3H)-one (9),²⁹ which is then converted to 5-amino-6-(5′-phosphoribosylamino)uracil (10) by ToxE, a bifunctional deaminase-reductase homologous to RibD.^{30, 31} Compound 10 is a known intermediate in riboflavin and F₀ biosynthesis.^32–35 The intermediacy of 9 and 10 are supported by the complementation of a toxB knockout strain by overexpression of ribA in B. glumae³⁶ and the continued, but reduced production of toxoflavin in ΔtoxB (PFL_1034) and ΔtoxE (PFL_1032) deletion strains of P. protegens Pf-5.¹⁹ Compound 10 or 11 is then coupled to the N2-C2 moiety of glycine through an unknown mechanism by ToxC to form ribityl-didemethyltoxoflavin (12) and ToxD is responsible for the oxidative cleavage of the ribityl side chain to form 1,6-didemethyltoxoflavin (1,6-DDMT, 13) as recently described by Song et al. (Figure 2).³⁷

Figure 2. Toxoflavin biosynthetic gene clusters and pathway in B. glumae BGR1 and P. protegens Pf-5.

A. Diagrams of the toxoflavin biosynthetic gene clusters found in B. glumae BGR1 (top) and P. protegens Pf-5 (bottom). The numbers under the P. protegens cluster refer to the last two digits of the locus tag (PFL_10XX) assigned during genomic sequencing (e.g., 35=PFL_1035). Putative orthologs in the two clusters are depicted with identically colored arrows. Colors denote predicted or experimentally verified gene functions: red, efflux or resistance; green, regulation; blue, biosynthesis (genes that encode proteins with identical functions have the same color); grey, transposition; orange, unknown function. B. Proposed toxoflavin biosynthetic pathway.

1,6-DDMT marks an intriguing branchpoint for natural products in BGCs containing ToxB-E.²³ It was first proposed to be the product of either BgToxC or BgToxD and the substrate of BgToxA by Suzuki et al.²⁸ Working backwards from toxoflavin, we provided biochemical evidence that 1,6-DDMT is a kinetically competent intermediate in the ToxA-catalyzed reaction that can be enzymatically converted into toxoflavin.²² We also showed that 1,6-DDMT is a substrate of a novel methyltransferase encoded by a related but distinct gene cluster in Burkholderia thailandensis that lacks ToxA but contains ToxBCDE.²³

In the present study, we aimed to gain insight into how 1,6-DDMT is formed through characterization of the active site structure of ToxD. We determined a high-resolution crystal structure of BgToxD and generated docking models that include the peroxy group. We also performed an extensive site-directed mutagenesis study accompanied by toxoflavin production assays using the ToxD ortholog in P. protegens Pf-5, PFL1035, as our testbed system. The sites were selected based on a combined bioinformatics and structural analysis. Our selections were made prior to the recent substrate and product identification and cover a range of sites that extend beyond the putative reaction center, enabling the identification of both structurally and catalytically important residues.

Methods and Materials

General methods and kits.

The GenCatch plasmid DNA mini-prep kit and EconoSpin columns were purchased from Epoch Life Sciences (Missouri City, TX). PrimeSTAR GXL polymerase was purchased from Takara Bio USA (Mountain View, CA). NdeI, HindIII-HF, DpnI, EcoRV-HF, T4 DNA ligase, and Escherichia coli NEB5α were obtained from New England Biolabs (Ipswich, MA). Oligonucleotides were ordered from Integrated DNA Technologies (Coralville, IA) and purified with standard desalting. Isopropyl β-D-1-thiogalactopyranoside (IPTG) and 5-Bromo-4-chloro-3-indolyl β-D-galactopyranoside (X-gal) were purchased from VWR and a stock solution of 10 mg/mL and 20 mg/mL respectively in dimethylformamide and plates were supplemented with 50 μL for blue/white selection. Kanamycin and ampicillin were purchased from VWR and supplemented at 50 μg/mL and 100 μg/mL, respectively, for both E. coli and Pseudomonas protegens Pf-5. Lysogeny broth (LB), Miller was purchased from MilliporeSigma (Burlington, MA). The vector pBluescript SK(−) was purchased from Agilent Technologies (Santa Clara, CA). The Genomic Wizard DNA purification kit was purchased from Promega (Madison, WI). All kits and enzymes were used according to the manufacturer’s supplied protocols unless noted. All molecular biology procedures and bacterial manipulations were performed according to standard protocols.²¹ Sanger sequencing was performed with the BigDye terminator v. 3.1 cycle sequencing kit (AbSciex) at the Center for Quantitative Life Sciences (CQLS, Oregon State University). All other chemicals were purchased from VWR and HPLC solvents were HPLC grade and used without further purification.

Protein crystallography.

ToxD of B. glumae was expressed from a modified pET-28 vector that supplies an N-terminal fusion tag MGSDKIHHHHHHSSGENLYFQGH. E. coli B834(DE3) cells were transformed with the recombinant vector and selected on Luria-Bertani (LB) agar plates supplemented with kanamycin (40 mg/L) after overnight growth at 37 °C. A few colonies containing the recombinant plasmid were inoculated into 50 mL of LB medium supplemented with kanamycin (40 mg/L) and grown at 37 °C and 200 rpm for 14−18 h. A portion of the overnight culture (10 mL) was inoculated into shaker flasks containing minimal medium containing 11.3 g/L M9 salts, 4 g/L dextrose, 50 mg/L L-selenomethionine, the remaining 19 L-amino acids (40 mg/L), 2 mM MgSO₄, 0.1 mM CaCl₂, 25 mg/L FeSO₄·7H₂O, a 1× minimal essential medium vitamin solution, and 30 mg/L kanamycin. The cells were grown at 37 °C and 200 rpm until an optical density at 600 nm of 0.5−1.0 was reached. The cultures were then cooled to 22 °C for 30 min and then to 15 °C for 1 h. Isopropyl 1-β-D-galactopyranoside was then added to the cultures to a final concentration of 0.1 mM, and the flasks were shaken at 15 °C and 200 rpm for 18 h. Following induction, the cells were pelleted and stored at −80 °C. Frozen cells were defrosted at room temperature and then resuspended in lysis buffer [50 mM tris(hydroxymethyl)aminomethane (Tris), 300 mM NaCl, and 5 mM imidazole (pH 8.0)], lysed via sonication on ice (8 × 30 s rounds consisting of 1.5 s on/off cycles), and subjected to centrifugation at 20,000 × g (4 °C for 20 min). The supernatant was applied to a 5 mL HisTrap HP column (Cytiva, Wilmington, DE) using an AKTA Explore (Cytiva). The column was then washed with approximately 100 mL of lysis buffer. The protein was eluted with elution buffer [50 mM Tris, 300 were tested for the presence of protein mM NaCl, 250 mM imidazole, and 2 mM dithiothreitol (DTT) (pH 8.0)]. Fractions using the Bradford assay and sodium dodecyl sulfate−polyacrylamide gel electrophoresis analysis. Fractions containing ToxD were then combined, and the His-tag was cleaved with TEV protease and subjected to subtractive Ni²⁺-chelate chromatography. ToxD was then applied to size exclusion chromatography using a HiLoad Superdex 200 26/600 column (Cytiva) equilibrated with 15 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 125 mM NaCl, and 1 mM DTT (pH 7.5). ToxD that eluted at a volume consistent with a dimer (Figure S1) was collected and then was concentrated and buffer exchanged into 5 mM HEPES and 20 mM NaCl (pH 7.0) at a final concentration of approximately 25 mg/mL.

Purified ToxD was crystallized at 18 °C using the method of hanging drop vapor diffusion. Drops were prepared by combining protein and reservoir solutions (Wizard crystal screen I, formulation 6) in a 1:1 ratio without further optimization. Cryoprotected crystals were looped and plunged into liquid nitrogen and then shipped to beamline 24-ID-C of the APS. Single crystals of both forms of ToxD were irradiated at 100 K using X-rays having a wavelength of 0.97918 Å. Diffraction images were recorded with an ADSC Quantum 315 CCD detector. The native ToxD crystal was positioned 258.5 mm from the detector and sixty images were recorded using an oscillation width of 1°. The SeMet labeled ToxD crystal was positioned 259.2 mm from the detector and 75 images were recorded using an oscillation width of 0.8°. Images were processed using HKL2000 and initial phases were determined using SHELX.^{38, 39} The model was iteratively built using COOT⁴⁰ and then refined using Phenix⁴¹. X-ray diffraction and structural refinement statistics are listed in Table S1. Structural models were analyzed using Chimera,⁴² PyMol⁴³ and the CAVER 3.0.3 PyMol plugin.⁴⁴

LK230 mutant creation.

Pseudomonas protegens Pf-5 strain LK230 contains an in-frame deletion mutant of PFL_1035 and construction of LK230 was previously described.¹¹

Microbial strain source.

Pseudomonas protegens Pf-5 strain LK230 was a kind gift of Dr. Joyce Loper and is maintained in the Philmus bacterial stock collection.

Creation of pBJP0102.

For complementation assays, PFL_1035 was amplified from P. protegens Pf-5 genomic DNA purified using the Genomic Wizard DNA purification kit (Promega). The PCR was performed using PrimerSTAR GXL polymerase according to the manufacturer’s supplied instructions. The reactions contained PFL1035_Fwd3 (5′-ATATATCCATATGACCGCCCCTCTCCCCTC-3′), PFL1035_Rev3 (5′- TAGATATAAGCTTTCAGAGGCTCTCGGCGAGCCT-3′) and 100 ng of genomic DNA. The underlined bases in PFL1035_Fwd3 and PFL1035_Rev3 correspond to the restriction sites for NdeI and HindIII, respectively, while the italicized bases are random bases added to the primer to increase restriction endonuclease efficiency. The PCR was performed using the following program. 98°C, 3 min; 98°C, 10 s, 60°C, 15 s, 68°C, 1:30 min (30 cycles); 68°C, 10 min. The PCR product was purified using the EconoSpin columns and then digested with NdeI and HindIII-HF in CutSmart buffer. The digested PCR product was then purified using an EconoSpin column and ligated into pMEKm12 (digested with the same enzymes and agarose gel-purified) using T4 DNA ligase. The ligation mixture was transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with kanamycin to select for plasmid-containing cells. Three colonies were randomly chosen, grown in 3 mL LB media supplemented with kanamycin (overnight, 37°C, 200 rpm) and the plasmids were isolated with the GenCatch plasmid DNA mini-prep kit. This plasmid and others described below were sequenced using Sanger sequencing to verify the correct insert sequence.

Creation of pBS-PFL1035–1.

The PFL_1035 gene was amplified as described above and following purification was digested with HindIII. The digested PCR product was then purified using an EconoSpin column and cloned into pBluescript-SK(−) that had been previously digested with EcoRV and HindIII using T4 DNA ligase. The ligation mixture was transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with ampicillin, IPTG and X-Gal to select for cells containing plasmid. Three white colonies were randomly chosen, grown in 3 mL LB media supplemented with ampicillin (overnight, 37°C, 200 rpm) and the plasmids were isolated with the GenCatch plasmid DNA mini-prep kit.

Creation of the PFL1035 Ala screening library.

The GCC codon (most preferred codon for alanine in P. protegens Pf-5) was introduced at the desired locations through splice overlap PCR for the K143A, D303A, D324A, H332A, and R315A variants. Using the oligonucleotides listed in Table S2, two PCR reactions were performed for each PFL1035 alanine⁴⁵ variant desired. The polymerase chain reactions were performed using PrimeSTAR GXL polymerase according to the manufacturer’s supplied instructions. The PCR amplifying the 5′-portion of PFL_1035 utilized PFL1035_Fwd3 (5′-ATATATCCATATGACCGCCCCTCTCCCCTC-3′) and the appropriate “PFL1035_X###A_Rev” oligonucleotide (Table S2), while the PCR amplifying the 3′-portion of PFL_1035 utilized PFL1035_Rev3 (5′- TAGATATAAGCTTTCAGAGGCTCTCGGCGAGCCT-3′) and the appropriate “PFL1035_X###A_Fwd” oligonucleotide (Table S2).

The PCRs were performed using the following program: 98°C, 3 min; 98°C, 10 s; 60°C, 15 s; 68°C, 1 min (repeat with either 25 (from plasmid) or 30 (from genomic DNA) cycles, vide infra); 68°C, 10 min. A portion of the reactions were then loaded on a 1% agarose gel to verify the synthesis of a fragment of the desired length. Reactions that showed a single band were purified with the EconoSpin columns. Reactions that showed multiple bands were first separated on a 1% agarose gel and the excised band was purified using the EconoSpin columns. Initial reactions were performed with genomic DNA using 100 ng of genomic DNA. Reactions that did not produce a band of the expected size were repeated using pBJP0102 as a PCR template (35 ng). Following the PCR reaction with pBJP0102 as a template, the reaction mixture was exposed to DpnI at 37°C for 2–3 h and then the PCR product was purified as above.

The matching 5′-portion and the 3′-portion of PFL_1035 amplified with the alanine codons were then used in a splice overlap PCR using GXL polymerase with oligonucleotides PFL1035_Fwd3 and PFL1035_Rev3 along with 10–90 ng of the first two PCR products. The PCRs were performed using the following program. 98°C, 3 min; 98°C, 10 s, 60°C, 15 s, 68°C, 1 min (30 cycles); 68°C, 10 min. The resulting products were purified after excising from a 1% agarose gel. The purified products were ligated into pBluescript SK(−), previously digested with EcoRV-HF using T4 DNA ligase. The ligation mixtures were transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with ampicillin, IPTG, and X-Gal. Two to three white colonies from each transformation mixture were grown in 3mL LB media supplemented with ampicillin and the plasmids purified.

A fragment of the PFL_1035 gene (between the SphI site and stop codon and encoding a 3′-HindIII site) encoding the W140A, E144A, E284A, D302A, R321A, C329A, and R331A variants were purchased from Genscript as synthetic DNA fragments cloned into the pUC57 vector. The plasmids were digested with SphI and HindIII and the fragments were isolated after separation using a 1% (w/v) agarose gel using EconoSpin columns. The fragments were ligated into pBS-PFL1035–1 (previously digested with SphI and HindIII and purified using agarose gel electrophoresis) using T4 DNA ligase. The ligation mixture was transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with ampicillin. Three from each transformation plate were randomly chosen, grown in 3 mL LB media supplemented with ampicillin (overnight, 37°C, 200 rpm) and the plasmids were isolated with the GenCatch plasmid DNA mini-prep kit.

Plasmids containing the correct insert were then digested with NdeI and HindIII and the PFL_1035 variant genes were purified after excision from a 1% agarose gel. The products were then ligated into pMEKm12 (digested with NdeI and HindIII and purified from an agarose gel) using T4 DNA ligase. The ligation mixture was transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with kanamycin to select for plasmid containing cells. Three colonies were randomly chosen, grown in 3 mL LB media supplemented with kanamycin (overnight, 37°C, 200 rpm) and the plasmids were isolated with the GenCatch plasmid DNA mini-prep kit. Plasmids are listed in Table S3.

Construction of the His₆-tagged PFL_1035 expression plasmids.

For complementation assays, PFL_1035 was amplified from P. protegens Pf-5 genomic DNA purified using the Genomic Wizard DNA purification kit (Promega). The PCR was performed using PrimerSTAR GXL polymerase according to the manufacturer’s supplied instructions. The reactions contained PAGE purified HisPFL1035Fwd (5′- CATATGGGCAGCAGCCATCATCATCATCATCACAGCAGCGGCCTGGTGCCGCGCGGCAGCATGACCGCCCCTCTCCCCTC-3′), PFL1035_Rev3 (5′- TAGATATAAGCTTTCAGAGGCTCTCGGCGAGCCT-3′) and 100 ng of genomic DNA. The PCR was performed using the following program. 98°C, 3 min; 98°C, 10 s, 60°C, 15 s, 68°C, 1.5 min (30 cycles); 68°C, 10 min. The PCR product was purified using the EconoSpin columns and then digested with NdeI and HindIII-HF in CutSmart buffer. The digested PCR product was then purified using an EconoSpin column and ligated into pMEKm12 (digested with the same enzymes and agarose gel purified) using T4 DNA ligase. The ligation mixture was transformed into chemically competent E. coli NEB5α cells. The transformation mixture was plated on LB agar supplemented with kanamycin to select for plasmid containing cells. Three colonies were randomly chosen, grown in 3 mL LB media supplemented with kanamycin (overnight, 37°C, 200 rpm) and the plasmids were isolated with the GenCatch plasmid DNA mini-prep kit.

An analogous protocol was followed for the construction of the C-terminal His₆-tagged PFL1035, with reactions containing PFL1035_Fwd3 (5′-ATATATCCATATGACCGCCCCTCTCCCCTC-3′) and PAGE purified PFL1035C-HisRev (5′-AAGCTTTCAGCAGCCGGATCTCAGTGGTGGTGGTGGTGGTGGAGGCTCTCGGCGAGCCTGAA -3′), (pMEKm-PFL1035-CHis-6).

Introduction of plasmids to P. protegens Pf-5.

The plasmids listed in Table S3 were introduced into P. protegens LK230 via electroporation. To prepare the electrocompetent cells, a single colony was inoculated from a fresh King’s media B (KMB)²² agar plate at 28°C, 200 rpm for 14–16 h. A portion of the overnight culture (1 mL) was inoculated into fresh KMB liquid (25 mL), and the culture grew at 28°C at 200 rpm until the OD₆₀₀ was between 0.4–0.6. The cells were then placed on ice for 15 minutes. The cells were pelleted (3,200 x g, 20 minutes, 4°C) and the supernatant was decanted. The cells were resuspended in 25–30 mL 10% (v/v) glycerol (pre-cooled to 4°C) and the cells were pelleted by centrifugation (3,200 x g, 20 minutes, 4°C). This process was repeated two additional times, and the cells were resuspended in 1 mL 10% (v/v) glycerol (pre-cooled to 4°C) and aliquoted (50 μL) into pre-chilled 1.5 mL microcentrifuge tubes. Cells were mixed with plasmid (5 μL, ~250 ng) and placed in a pre-chilled electroporation cuvette (0.2 cm gap, Bulldog Bio). Electroporation was accomplished using a BioRad MicroPulser Electroporator and the ‘Ec2’ settings (2.5 kV). Electroporated cells were immediately mixed with room temperature LB media (950 μL) and transferred to a 1.5 mL microcentrifuge tube. The cells were shaken at 28°C, 200 rpm for 2 h and then transformed cells were selected on LB agar plates supplemented with kanamycin. The plates were incubated at 25°C for 36 h and single colonies were randomly selected and grown overnight in KMB media supplemented with kanamycin (22°C, 200 rpm).

Cultures of P. protegens Pf-5.

In a 24-well plate, the previously prepared colonies were inoculated in 1.5 mL of KMB media supplemented with kanamycin and iron sulfate (0.1 mM). The cultures were then incubated at 22°C without shaking for 72 h at which time all liquid was aseptically removed followed by pelleting the cells by centrifugation (21,130 x g, 22°C, 15 min. The supernatant was transferred to a new glass vial for extraction of toxoflavin and the pelleted cells were stored at −75°C.

Extraction of toxoflavin.

Toxoflavin was extracted from the samples described above with chloroform (1.5 mL) for a total of three rounds. Between each round, the vials were centrifuged for 30 minutes to help separate the two liquids. The remaining chloroform was transferred to a new glass vial and after the third round, was removed via MiVac Quattro rotary concentrator (SP Scientific). The remaining toxoflavin was diluted with MQ water (125 μL) and transferred to an amber HPLC analysis vial.

HPLC analysis of toxoflavin production.

Toxoflavin production was quantified using an Agilent 1100 HPLC instrument consisting of a vacuum degasser, quaternary pump, autosampler (cooled to 4°C), column thermostat (maintained at 30°C), and diode array detector. Separation was achieved using a Luna C18 column (4.6 × 150 mm, 5 μm, Phenomenex, Torrance, CA) with a flow rate of 1 mL/min where line A was water + 0.1 % (vol/vol) formic acid, and line B was methanol + 0.1 % (vol/vol) formic acid with the following program. The column was pre-equilibrated in 100% A/0% B and upon injection (100 μL) this composition was held for 2 min. The composition of the mobile phase was then changed to 85% A/15% B over 10 min utilizing a linear gradient followed by changing to 0% A/100% B over the next 1 min. This composition was held for 7 min followed by changing to 100% A/0% B over 2 min. The column was equilibrated in 100% A/0% B for 6 min prior to the next injection. Under these chromatographic conditions, toxoflavin eluted at 10.5 min. The HPLC was operated with, and data were viewed using ChemStation (version B.04.03, Agilent, Santa Clara, CA). Quantitation of toxoflavin production was accomplished through the analysis of three replicate samples and compared to a standard curve generated with synthetic toxoflavin. Data were processed with Microsoft Excel.

Molecular dynamics simulations of BgToxD

Ligands were docked into the BgToxD crystal structure using AutoDock Vina⁴⁶ in UCSF Chimera.⁴² Ligand binding was restricted to within the inside of the protein near the hypothetical active site, and conformations were picked based on the lowest energy binding state. Molecular dynamics simulations of BgToxD were carried out with the GROMACS⁴⁷ simulation package v2022 using the CHARMM36 force-field (FF) for proteins.^48–50 Water was simulated using CHARMM TIP3 parameters. Potassium and chloride ions were added to a concentration of 150 mM. The combined protein-ligand system was set up using the CHARMM-GUI webserver.^{51, 52} Four different simulations were set up with the wild-type protein and D279A substitution with compound 12 present and not present for each protein. FF parameters for the two substrates were generated using the CHARMM-GUI interface and CGenFF.^53–55 Systems were simulated for 100 ns after initial energy minimization and equilibration using position restraints. Equations of motion were integrated with a leap-frog algorithm using a 2-fs time step using the default LINCS constraint algorithm. Van der Waals interactions were computed using a force-switched Lennard-Jones potential between 1.0 and 1.2 nm. Electrostatic interactions were computed using the particle-mesh Ewald method⁵⁶ with a real space cutoff of 1.2 nm and a Fourier grid spacing of 0.12 nm. Temperature was held constant using a velocity rescaling algorithm⁵⁷ with a time constant of 1 ps, and pressure was held constant with a stochastic cell-rescaling algorithm using a time constant of 5 ps.⁵⁸ Particle positions were saved in 5 ps intervals for trajectory analysis.

Computational Docking and Energy Minimization of BgToxD Reaction Intermediates

Compound 12 was docked into the BgToxD crystal structure using Schrödinger Extra Precision Glide docking software.⁵⁹ Ligand docking was restricted to the area around the putative active site of the enzyme. The best docked structure by the glide docking score was used for the next step. The docked structure was edited into the hypothesized peroxide intermediate for the BgToxD reaction. The entire system, BgToxD and the intermediate, was energy minimized in Schrödinger Maestro.

Computational Docking of BgToxD Reaction Substrate 12 and Comparison to Peroxide Intermediate.

Proposed intermediates 14, 15, or 16 were docked into the BgToxD crystal structure using Autodock Vina.⁴⁶ Charges for docked compound were assigned using Schrödinger Maestro.⁶⁰ The ligand was restricted to the same area around the putative active site as the previous method with the docking of compound 12. The best conformation was chosen based on the conformational similarity to the docking of 12. To compare the energy differences between the two intermediates, the peroxide intermediate from the previous step was redocked in the minimized BgToxD structure to obtain a consistent method of scoring.

Results and Discussion

ToxD orthologs

To identify ToxD-like proteins involved in the biosynthesis of triazine natural products and characterize their sequence conservation, we performed a phylogenetic analysis of Pfam PF03781 (14,664 sequences), of which BgToxD is a member (Figure S2), which shows the ToxD-like proteins cluster off a single branch. To identify putative biosynthetic gene clusters (BGCs), we scanned deposited genome sequences for the co-occurrence of genes encoding proteins with similarity to ToxC and ToxD within five kilobases of each other using MultiGeneBlast⁶¹ and EFI-EST GNN.^{62, 63} We identified over 100 BGCs in pathogenic Burkholderia or Pseudomonas strains as well as Actinobacteria, Proteobacteria, and cyanobacteria. The BGCs met three criteria: they were identified in the ToxD-like clade, are proximal to genes encoding ToxB, ToxC, and ToxE-like proteins, and are from phylogenetically distinct bacterial genera (Figure S3).

We chose 21 of the ToxD-like proteins from the identified BGCs to use in a multiple sequence alignment (MSA) which revealed 82 highly conserved residues across the 21 protein sequences (Figure S4 and Table S4). The majority of conserved residues are found in the C-terminal half of these proteins (Figure S4–S5 and Table S5). The CxxxxC motif important for Cu(I) binding and O₂ activation in formylglycine-generating enzymes (FGEs)^{20, 23–25} is absent in this region, supporting the lack of exogenously added metal ions and insensitivity of the in vitro reaction to EDTA.³⁷ Two of the aligned sequences (Saccharopolyspora erythraea NRRL 2338, NCBI accession number CAM02164.1 and Streptomyces autolyticus strain CGMCC0516, NCBI accession number AQA11602.1) contain a single cysteine residue, but cysteine is not universally conserved at any position. FGEs also contain two Ca²⁺ binding sites as well as a peptide substrate binding cleft.^64–66 Of the several conserved stretches of two or more residues, the “R₃₀₆RHG” motif and “K₁₁₉E” dipeptide (B. glumae numbering) are both unique to the ToxD family of enzymes and predicted to cluster near the FGE active site. BgTrp116 is also strongly conserved. We discuss these features, including a cocrystallized cation at one of the calcium sites in more depth in the context of the BgToxD crystal structure in the next section.

Crystal structure of ToxD.

BgToxD contains 326 residues and has a molecular weight of 35.5 kDa and its structure was deposited in the Protein Data Bank (PDB: 7SPQ). The BgToxD crystal belongs to the space group I23, and the asymmetric unit contains one ToxD molecule (residues 11–326 visible in electron density map), a buried cation, two citrate ions, three polyethylene glycol molecules, and 470 waters (Figure S6). The structure contains the FGE core fold, an 82 residue N-terminal extension that wraps partly around the core, and several loops and helical elements (Figures 2A and B). The identity of the cation is unknown but was tentatively assigned as calcium (Figure S6). It is located at the first Ca²⁺ binding site of human FGE and is coordinated by four carbonyl groups (Asp254, Met255, Gly257, and Val259), one carboxylate (Glu261), and one water molecule. The ion also interacts electrostatically with the carbonyl of Gly293 (3.4 Å).

A search for component oligomers within the crystalline lattice reveals a dimer, having a buried surface area of 4,115 Ų (theoretical ΔG_diss = 27.4 kcal/mol)⁶⁷ (Figure 3C). The interface is formed largely by the N-terminal extension and a dimerization loop (residues 151–177), but additionally contains the ‘RRHG’ loop (Figure S5). The lattice also reveals an octamer comprising of four ToxD dimers, having a buried surface area of 23,750 Ų (theoretical ΔG_diss = 2.7 kcal/mol) (Figure 4A). While we cannot rule out the possibility that the octamer is purely crystallographic, the two oligomeric forms are consistent with size exclusion chromatograms displaying both low and high molecular weight peaks (Figure S1).

Figure 3. Crystal structure of ToxD.

(A) Asymmetric unit, containing one protein chain, one cation, and two citrate ions. Structural elements are color labeled as follows: N-terminal extension, salmon; FGE core fold, light blue; ‘WIxKE’ helix, green; dimerization loop, purple; ‘RRHG’ loop, khaki; ‘DL’ helix, yellow; side loop, orange. (B) Associated topology diagram. Helices and β-strands are depicted as cylinders and arrows, respectively. (C) ToxD dimer. The dimer is shown looking down the axis of two-fold rotational symmetry (left) and in a rotated view with structural element coloring applied to one subunit (right).

Figure 4. Cleft assembly: ToxD-specific structural motifs.

(A,B) Crescent-shaped cleft formed through interactions between side loop and two adjacent protomers in octameric assembly in crystalline lattice. (A) Octamer (tetramer of ToxD dimers) within lattice. (B) Protomer interface. Inset shows interactions involving side loop. (C) ‘RRHG’ loop and ‘WIxKE’ and ‘DL’ helices form remainder of cleft, in which citrate is deeply bound. (D) Tunnel connecting center of cleft to surface through ‘RRHG’ loop. The tunnel was delineated using a 1.45 Å probe radius.

The active site is predicted to reside within a crescent-shaped cleft having a length of about 15 Å. A section of the cleft is formed through interactions between the side loop (Val263-Val289) and two adjacent protomers within the octamer (Figure 4B). Within this interface, a predominantly hydrophobic α-helix (residues 280–283) in addition to Ala275 and Ile276 pack against hydrophobic surface residues in the two neighboring chains; two hydrogen bonds are also observed (Figure S7).

The more deeply bound of the two citrate ions modeled in the asymmetric unit is located at the center of the cleft, which is assembled by ToxD-specific structural elements containing the aforementioned ‘RRHG’ and ‘KE’ motifs, as well as Trp116 (Figure 4C). The latter three residues form a section of the outer wall of the cleft and are introduced by a helical extension, the ‘WIxKE’ helix, of the characteristic curved β-hairpin of the FGE fold. Trp116 forms a planar surface against which citrate and the modeled azapteridine substrate stack.³⁷ Helix α1 within the N-terminal extension positions Arg11 15 Å from the center of the cleft (central carbon of citrate) and it remains to be determined if the first ten residues introduce additional sites for binding substrates and intermediates.

The ‘RRHG’ loop is preceded by a β-strand of the conserved FGE fold and the side loop that shapes the cleft. Arg310 within the ‘RRHG’ loop interacts with the side loop via stacking with Tyr287 and a salt bridge with Asp279. It is noted, however, that Arg310 is not a strongly conserved residue. The highly conserved Arg306 and Lys119, which both bind citrate, are poised to donate hydrogen bonds or form salt bridges inside the cleft near Trp116. Asp300 of the DL helix orients the side chain of Arg306 via ionic bonding. His308 places its Nε2 atom into the cleft where it can play a role in acid-base chemistry as Nδ1 forms a hydrogen bond with Glu260. Gly309 lines part of a tunnel, plausibly utilized by the enzyme for oxygen transport, connecting the citrate site to the surface through the RRHG loop (Figure 4D). π-π-stacking between the Gly309-Arg310 backbone and the side chain of Tyr287 may be important for shaping the tunnel structure; the backbone dihedral angles of Gly309, (φ=110.5, ψ=2.8), reside only within the energetically allowed region of Ramachandran space for glycine. Arg307 and Glu120 play a structural role, as both are deeply buried and form salt bridges with residues of the FGE core.

Mutagenesis strategy

Given the location of the citrate ion within the putative active site cleft and its close proximity to conserved, catalytically competent residues, we initially considered for site-directed mutagenesis all conserved residues residing within a radius of 8 Å of the buffer component. The number of residues were further reduced by intersecting with amino acids with polar or aromatic side chains given the multitude of polar sites in the substrate as well as the azapteridine ring system. The resulting 12 residues were then targeted for substitution with alanine as described in the experimental section.

Development of a chassis strain and vector for complementation of mutants.

For our in vitro complementation experiments, we chose P. protegens strain LK230,¹⁹ which contains an in-frame deletion of PFL_1035 (toxD homolog). A lack of production of toxoflavin was previously shown in the P. protegens strain LK230 grown under standard conditions.¹⁹ We utilized pMEKm12,²⁶ an E. coli-Pseudomonas shuttle vector for complementation studies. The PFL1035 gene was cloned between the NdeI and HindIII sites of pMEKm12 to create pBJP0102. This results in production of native PFL1035 protein (no N- or C-terminal tags) from the tac promoter. Introduction of pBJP0102 or pMEKM12 into strain LK230 was accomplished by electroporation and the resulting strains were tested for complementation of toxoflavin production. pBJP0102 showed toxoflavin production, while pMEKm12 showed no production as expected (Figure 5A).

Figure 5. Toxoflavin production by PFL1035 wild-type and variants in P. pseudomonas Pf-5.

A. Toxoflavin production in cultures of P. protegens Pf-5 strain LK230 (contains an in-frame deletion of PFL_1035). (i) LK230 containing pMEKm12 (empty vector),). (ii) LK230 containing pBJP0102 (PFL1035 wild type), (iii) toxoflavin standard. B, Amount of toxoflavin produced in a 1.5 mL culture of P. protegens strain LK230 containing plasmid encoding wild-type and variant PFL1035 proteins. Toxoflavin amount was determined via comparison to a standard curve generated from synthetic toxoflavin.

Testing the reproducibility of the extraction procedure was vital to ensure consistent toxoflavin production. To test the variability in the extraction procedure P. protegens LK185 was used as it was previously shown to produce toxoflavin at easily detectable levels. To examine technical variability, P. protegens LK185 was cultivated in triplicate (1.5 mL in 24-well plate) and the resulting media was combined, mixed, and split evenly (3 × 1.5 mL samples). The three identical media samples were extracted in parallel and HPLC quantification of toxoflavin resulted in an average of 9.13 ± 0.5 nmole toxoflavin being identified in the three cultures. To measure biological variability P. protegens LK185 was cultivated in triplicate (1.5 mL in 24-well plate) and the media from each well was extracted with chloroform. HPLC analysis of the toxoflavin content showed that the samples contained an average of 8.4 ± 1.5 nmole toxoflavin. These two experiments demonstrated the reproducibility and accuracy of our extraction procedure, and that biological variability contributes more to error than our extraction protocol.

Alanine substitution mutagenesis.

The twelve residues identified via MSA (described above) were analyzed for toxoflavin production. Eight of the alanine substitutions (P. protegens H332A, R330A, R315A, D303A, D324A, W140A, R321A, and R331A) were shown to abolish toxoflavin production while three substitutions (E284A, E144A, and K143A) produced reduced amounts of toxoflavin (50.7 ± 8.3%, 41.0 ± 12.3%, and 31.1 ± 8.2%, respectively) compared to wild-type PFL_1035 (Figure 5 and Figure S8). The alanine substitution of Asp302 resulted in a neglible change, producing 96.0±12.6% of the toxoflavin generated by the wild-type enzyme. By comparison, the corresponding residues of PpR321, PpD324, and PpH332 in ToxD cloned from B. gladioli were recently demonstrated to be essential for 1,6-DDMT production during in vitro characterization.³⁷

Potential mechanisms for ToxD

In reviewing the crystal structure and Ala-scanning results we posited three mechanisms that were possible. Mechanism A, was that proposed by Song et al³⁷ in which the 4a-peroxo intermediate (14, Figure 6) is formed. Intermediate 14 degrades to 16 and hydrogen peroxide. Mechanism B involves a single electron transfer to directly form 16 and hydrogen peroxide, while Mechanism C is postulated to proceed through the N8-peroxo intermediate 15, which is converted to 16. The N8-peroxo intermediate (15) is analogous to the flavin-N5-oxide^{68, 69} proposed for Enc, DszA, and RutA.⁶⁸ and the FAD_N5OO intermediate recently proposed for TdaE, which is involved in tropone biosynthesis.^{70, 71} In all three mechanisms, nucleophilic attack by water at C1´ generates 17, which upon deprotonation of the alcohol generates didesmethyltoxoflavin (13) and D-(−)-ribose (18) (Figure 6).

Figure 6. Potential mechanism for the conversion of 12 to 13 and 18 by ToxD.

Mechanism A involves the formation of a 4a-peroxy intermediate, which in silico binding calculations estimated to have a binding energy of −7.5 kcal/mol. Mechanism B proceed through an intermolecular electron transfer to generate hydrogen peroxide and 16, which is has a binding energy of −8.2 kcal/mol. Mechanism C is proposed to proceed through a N8-peroxy intermediate with a calculated binding energy of −6.8 kcal/mol.

We calculated the binding energies of the ToxD substrate 12, and the intermediates (14, 15, and 16). The substrate had a calculated binding energy of −7.8 kcal/mol. Intermediate 14 had a similar calculated binding energy of −7.5 kcal/mol, while 15 had a higher binding energy (−6.8 kcal/mol), while 16 has the lowest binding calculated binding energy. Coupling this with the fact that to undergo Mechanism C, molecular oxygen would have to bind prior to 12, which seems unlikely,³⁷ we posit that either Mechanism A or B is most likely.

Peroxo intermediate model: molecular basis for mutagenesis results

The absolutely conserved histidine, BgH308, was proposed previously to act as a catalytic base to generate an iminium intermediate, which would then form a peroxo intermediate (see Figure 3 in reference³⁷). The essentiality and side chain orientation of His308 and its disposition relative toTrp116 provide a rational starting point for docking 12 into the crystal structure. In the model based on the crystal structure (Figure 7), the side chain is well-positioned to abstract a proton from the C1´ methylene of the ribityl group via Nε2, and the azapteridine ring system stacks against the indole group of Trp116. Additionally, the guanidinium of Arg306 (PpR330) binds the pyrimidine C7 carbonyl and N8. The side chains of Arg297 (PpR321), Asp300 (PpD324), and Arg306 through a hydrogen bonding network bind the ribityl side chain, which displaces several water molecules in the crystal structure.

Figure 7. Putative active site of BgToxD with substrate (12) positioned by molecular docking.

Compound 12 is shown in salmon, while the residues previously investigated by Song et al⁵ are shown in orange. The additional residues investigated in this work are shown in blue. The residue is numbered according to the position in BgToxD.

A model containing the proposed peroxo intermediate (14) overlays well with 12 (Figure S9). One major difference is that His308 is rotated by approximately 30° and is now in a position to form a hydrogen bond with, and thence protonate, the peroxy anion. We did not observe any necessary significant changes in the loop surrounding the peroxy group for binding of the intermediate. Notably, the tunnel illustrated in Figure 4D overlaps partially with the putative binding site of the peroxy group.

Alanine substitution of all of the above residues resulted in no toxoflavin production. Of the remaining residues producing a similar outcome, the conserved residues Arg291 (PpR315) and Arg307 (PpR331) are not predicted to bind directly to the ribityl-didemethyltoxoflavin, and the guanidino groups are oriented away from the substrate. Arg307 forms hydrogen bonds with Lys119, Glu201, and Arg268 while Arg291 forms hydrogen bonds with Glu201, Glu205, and Cys304. Thus, the loss of activity is most probably the result of structural instabilities. Asp279 (PpD303) resides in the side loop and directs its side chain into the crescent shaped cleft on the side opposite to that expected to bind the ribityl side chain. Its carboxylate forms a salt bridge with Arg310 of the ‘RRHG’ loop and thus the loss of activity may be related to structural changes in the loops. A second hypothesis was that the aspartate could be involved in binding and stabilization of the peroxo intermediate proposed by Song et al.³⁷ In their model, which is based on AlphaFold2, the side loop containing this residue adopts a different conformation and the aspartate is closely positioned to the docked substrate. During the simulation time frame, using one protomer from the crystal structure (instead of the octamer) we did not observe the side chain of D303 being located within 5 Å of the bound peroxo intermediate (data not shown). We did note that the average RMSD of the main chain atoms of the D303A substituted ToxD variant increased more during the course of the simulation compared to the wild-type (Figure S10), however the RRHG motif does not move in any greater degree compared to the protein as a whole. These results suggest that D303 is involved in protein stability.

Mutagenesis of the ‘KE’ motif residues, despite their conservation, only partially reduce toxoflavin production. Lys119 (PpK143) is predicted to bind the substrate C7 carbonyl oxygen atom via its ε-amino group. In contrast, Glu120 (PpE144) is not expected to bind the substrate. Instead, its side chain forms a conserved ion pair with Arg303, bridging the ‘WIxKE’ helix and ‘RRHG’ loop. The last residue of this grouping, Glu260 (PpE284), forms a hydrogen bond with the imidazole ring of His308. It is envisioned that this increases the basicity of histidine allowing it to perform the proposed catalytic deprotonation at the C1´ atom of the ribityl group.

Of the residues that partially reduce toxoflavin production, the ε-nitrogen of Lys143 (BgK119) is suggested to bind the C7 carbonyl oxygen atom; its removal probably weakens the substrate binding affinity and/or allows small orientational changes that reduce the catalytic efficiency. Glu284 (BgE260) forms a hydrogen bond with the imidazole ring of His332 (BgH308). It is envisioned that this increases the basicity of histidine allowing it to perform the proposed catalytic deprotonation at the C1´ of the ribityl group. It also likely orients the imidazole ring for proton abstraction because the rotamer observed in the crystal structure differs to some degree from ones in the high probability set (Figure S11, Supplementary Files 1–3).³⁷

Conclusion

The primary aim of this structure-function study was to provide additional insights into the ToxD catalyzed reaction. Being the synthase of 1,6-DDMT, ToxD exists at a vital branching point in BGCs that are hypothesized to generate a variety of azapteridine natural products.²³ ToxD contains several unique structural elements relative to its parent FGE scaffold, including some involved in the formation of higher oligomers revealed by the crystalline lattice, and others that shape the presumptive active site. The structure readily accommodates the substrate and peroxy intermediate via computational docking and thus provides clues into the catalytic mechanism as well as a molecular basis for the alanine scanning mutagenesis data. The latter results in turn pinpoint a set of sites within the structure that play a pivotal role in the catalysis and structural integrity.