ToxB encodes a canonical GTP cyclohydrolase II in toxoflavin biosynthesis and ribA expression restored toxoflavin production in a ΔtoxB mutant

Abstract

Burkholderia glumae synthesizes toxoflavin, a phytotoxin that contributes the virulence of this phytopathogen. The toxoflavin biosynthetic gene cluster contains a tox operon composed of five genes, toxABCDE, and toxB is predicted to encode a GTP cyclohydrolase II, which is the first enzyme (RibA) in riboflavin biosynthesis. ToxE is also homologous to RibD, the bifunctional deaminase/reductase acting on the RibA product. This suggests that toxoflavin and riboflavin biosynthesis share the first two steps in their respective biosyntheses. In this study, we demonstrated that ToxB and B. glumae RibA (Bglu-RibA) both displayed GTP cyclohydrolase II activity with comparable kinetic parameters. When toxB was inactivated, toxoflavin production was abolished, and introduction of a plasmid copy of Bglu-ribA restored toxoflavin production in a ΔtoxB mutant. ToxB and Bglu-RibA can thus be defined as GTP cyclohydrolase II isozymes, even though Bglu-ribA is not a genetic equivalent of toxB because the chromosomal copy of Bglu-ribA is unable to support toxoflavin production in the absence of toxB. In LB agar culture, toxAB is incapable of complementing ΔtoxB though toxABC induced toxoflavin accumulation in the ΔtoxB mutant up to 80 % of the WT level. This indicates that toxBC co-expression is a critical factor for toxoflavin biosynthesis in this condition, suggesting that the WD repeat protein ToxC acts as a scaffolding protein in a toxoflavin biosynthetic metabolon. In LB liquid culture, toxAB successfully restored toxoflavin production, suggesting that a role of toxBC co-expression is dependent upon growth condition.

Introduction

Toxoflavin, fervenulin, and reumycin are 7-azapteridine antibiotics with bacterial origins (Machlowitz et al. 1954). Toxoflavin was later identified as a key player in the virulence of Burkholderia glumae, a rice pathogen (Sato et al. 1989; Yoneyama et al. 1998; Kim et al. 2004). An early biosynthetic study on toxoflavin in Pseudomonas cocovenenans substantiated its origin of a purine derivative (Levenberg and Kaczmarek 1966; Levenberg and Linton 1966). Together with these results, the knowledge of riboflavin biosynthesis (Fischer and Bacher 2005; Gräwert et al. 2013) guides us to deduce the involvement of GTP cyclohydrolase II and deaminase/reductase activities in toxoflavin biosynthesis (Fig. 1A). GTP cyclohydrolase II (RibA) that mediates a committed step in riboflavin biosynthesis converts GTP into an anomeric pair of 1, involving a guanine ring opening and release of formate and pyrophosphate (Ritz et al. 2001). Structural and mechanistic analysis of RibA revealed a catalytic arginine residue that is involved in the generation of a covalent guanylyl-RibA intermediate with pyrophosphate release (Ren et al. 2005). This enzyme-bound intermediate undergoes an imidazole ring-opening event and formate is subsequently released (Ritz et al. 2001; Schramek et al. 2001). The resulting product of 1 was then hydrolyzed from RibA. Kinetics study of RibA reaction indicated that the first chemical event of the reactions is the rate-determining step, suggesting that a slow formation of guanylyl-RibA intermediate controls the overall turnover number of RibA (Schramek et al. 2001); pre-steady state quenched flow analysis of RibA demonstrated that GTP was consumed at a rate constant of 0.064/s and the final product was formed at an apparent rate constant of 0.062/s.

Fig. 1

Toxoflavin biosynthetic pathway. (A) The proposed toxoflavin biosynthetic pathway with comparisons to that of riboflavin. (B) Genetic organization of tox operon

Transposon mutagenesis and subsequent targeted gene inactivation studies identified the toxoflavin biosynthesis locus (toxABCDE) from B. glumae and the presence of toxB, a gene homologous to that of GTP cyclohydrolase II asserted the GTP origin of toxoflavin (Kim et al. 2004; Suzuki et al. 2004) (Fig. 1B). The ribA candidate (bglu_2g14580) of B. glumae bears a modest homology to toxB (52, 61, and 2 in the percentage values of identity, homology, and gap, respectively, between the deduced amino acid sequences). The toxE gene is predicted to encode a RibD isozyme (a deaminase/reductase activity that converts 1 into 2 in riboflavin biosynthesis) (Magalhães et al. 2008) and B. glumae RibD is encoded in bglu_1g08000 (homology values of 70, 82, and 0 between ToxE and RibD). Riboflavin biosynthesis is processed through 3 (Fischer and Bacher 2005) (Fig. 1A). The tox operon composition supports the idea that the toxoflavin biosynthetic route shares its early two steps with that of riboflavin. However, the low level of similarity (61 %) between ToxB and B. glumae RibA (Bglu-RibA) and the lack of a sugar moiety in the toxoflavin structure suggest that ToxB intrinsically differs from RibA and utilizes an unidentified purine derivative, instead of GTP for its substrate.

ToxA, a S-adenosyl-l-methionine-dependent methyltransferase, is predicted to be an N-methyltransferase that mediates the final step in the biosynthesis of toxoflavin and its congeners. The biochemical mechanism involved in the conversion of 2–4 is totally veiled, but ToxC and ToxD can be tentatively assigned as the enzymes involved in this conversion. ToxC is a WD-40 repeat domain protein, and a biochemical role of ToxD is elusive. WD-40 repeat domain proteins are widely known for their roles in protein–protein interactions (Stirnimann et al. 2010). The conversion of 2–4 involves the formation of an N–N bond that is found in diverse natural products; a biochemical mechanism of this bioconversion is yet enigmatic (Blair and Sperry 2013). Toxoflavin biosynthesis can be regarded as an excellent template for elucidating the N–N bond formation mechanism due to the relative simplicity of its structure and the catalyst candidates (ToxC and ToxD) at hand. Understanding the molecular mechanisms involved in toxoflavin biosynthesis can also provide a biocatalytic tool to generate derivatives of 7-azapteridine, which can be explored as herbicides (Kim et al. 2012) or new drug candidates (Nagamatsu and Yamasaki 2001; Choi et al. 2013).

Materials and methods

Strains, culture conditions, and plasmids

Burkholderia glumae BGR1 was used in this study. Escherichia coli DH5α and BL21 (DE3) were used for routine DNA cloning experiments and recombinant protein expression, respectively. E. coli S-17 [recA pro (RP4-2Tet::Mu Kan::Tn7)] (Simon et al. 1983) was used for intergeneric conjugation to introduce the plasmids into B. glumae. All of the bacterial strains were grown or maintained in LB broth or on LB agar plates at 37 °C. Liquid cultures were maintained in a shaking incubator at 250 rpm. The antibiotics and their working concentrations used in this study were rifampicin (Rif), 50 µg/mL; kanamycin (Km), 50 µg/mL; and gentamycin (Gm), 10 µg/mL. The Burkholderia suicide vector pGSV3 (gentamicin resistance, Gm^r) (DeShazer et al. 2001) was used for a targeted inactivation of toxB. Complementation plasmids were constructed in pBBR1MCS-5 (Gm^r) (Kovach et al. 1995), except that nptII promoter (nptIIp)-toxB was prepared in pPROBE-GTkan (Gm^r) plasmid that harbors a 131 bp nptIIp from Tn5 in its multi-cloning site (Miller et al. 2000). pBBR1MCS-5 replicates as a low-copy number plasmid, with approximately 10 copies per cell.

Preparation and GTP cyclohydrolase II activity assay for the RibA and ToxB recombinant proteins

Bglu_ribA and toxB were amplified from B. glumae total DNA by polymerase chain reaction (PCR) with the ribA-F/-R and toxB-F/-R primer pairs, respectively (Table 1). The resulting PCR products are digested with NdeI and HindIII and cloned into the same sites in pET28b. The resulting expression constructs were introduced into E. coli BL21 (DE3) cells. The resulting transformants were cultured in 500 mL LB medium at 37 °C until OD₆₀₀ reached to 0.5; then, the incubation temperature was lowered to 14 °C. Next, recombinant protein synthesis was induced by the addition of isopropyl-β-d-thiogalactopyranoside to a final concentration of 1 mM and the incubation was continued overnight at 14 °C. Purification of N-terminal six histidine-tagged proteins was achieved with a nickel nitrilotriacetic acid agarose resin (Qiagen, the Netherlands) according to the manufacturer’s protocol. The GTP cyclohydrolase II activity assay reaction buffer contained 5 mM MgCl₂ and 50 mM Tris–HCl, pH 8.5. For the high-performance liquid chromatography (HPLC) analysis, 1 mM GTP was incubated with either 2.3 μM RibA (A) or 1.7 μM ToxB (B) in 100 μL at 37 °C for the indicated time period and 25 μL each of the samples was applied for HPLC separation. HPLC elution was performed on a Gemini C-18 column (150 × 2 mm, particle size of 3 μm, pore size of 11 nm; Phenomenex, USA) with a ProStar system (Varian, USA). The column was eluted with isopropanol/triethylamine/85 % phosphoric acid/water (8:10:3:979, v/v, pH not adjusted) and monitored at 280 nm as previously reported (Ritz et al. 2001). For the ultraviolet–visible (UV–Vis) spectroscopic measurements, 0.1 mM GTP was incubated with either 1.0 μM RibA (A) or 1.2 μM ToxB (B) at the ambient temperature. A UV–Vis scanning spectrophotometer (Cary50, Varian) was used for the spectroscopic measurements. The spectroscopic method was also used to determine the kinetic parameters with the varying GTP concentrations. A molar absorption coefficient of 6,280/cm at 310 nm for the GTP cyclohydrolase II product (2,5-diamino-6-β-ribosylamino-4(3H)-pyrimidinone 5′-phosphate) was used to calculate the activity (Grill et al. 2007).

Table 1 Oligonucleotide primers used in this study

Inactivation of toxB

The toxB inactivation construct was designed to insert kanamycin resistance gene (kan) inside toxB. The primers used in this study are shown in Table 1. The B. glumae genomic DNA encompassing toxB, its 1,223 bp upstream, and 2,106 bp downstream region were amplified with the dtoxB-F/-R primer pair. The PCR product was cloned into XbaI and KpnI sites in pBluscript II KS (Stratagene, USA), generating pKS-toxB. kan gene was amplified with the Kan-F/-R primer pair from pMKm2 (Um et al. 2011) and inserted into the BglII site inside toxB in pKS-toxB through the In-Fusion cloning method (Clontech, USA) to generate pKS-dtoxB. Then, pKS-dtoxB was used as the template for ΔtoxB::kan DNA amplification with the dtoxB-pGSV3-F/dtoxB-pGSV3-R primer pair. The ΔtoxB::kan fragment was cloned into the XbaI and DraI sites in pGSV3 to generate pTBKG. We have introduced pTBKG into B. glumae for toxB inactivation but failed in getting kanamycin resistant (Km^r) and gentamicin-sensitive phenotype. All of the exconjugants displayed Km^r and Gm^r phenotype, which is indicative of the single crossover-mediated plasmid integration event. One of these single crossover exconjugants was subjected to plasmid curing procedure but it failed. A genotyping study revealed that the single crossover-mediated plasmid integration exclusively happened at the toxB downstream region. Thus, we modified pTBKG by extending the size of the upstream region. Additional 2,242 bp upstream region was amplified with the Up-tox-F/-R primer pair. The PCR product was cloned into pTBKG through the In-Fusion cloning method to generate pTBKG-2.2. Intergeneric conjugation method with E. coli S17-1 was used to introduce pTBKG-2.2 in B. glumae. In details, 50 μL each of overnight cultures of E. coli S17-1/pTBKG-2.2 and B. glumae was inoculated in fresh 5 mL LB broth and was grown for 4 h. Cells from each culture were washed using fresh LB broth twice and resuspended in 60 µL of LB broth. The resulting cell suspensions were mixed at the varying ratios, spotted on LB agar plate, and grown overnight. The cell mixture was recovered and resuspended in 400 µL of LB broth. A 200 µL of the resulting resuspension was overlaid on LB agar plate with Km. The Km^r exconjugants were selected after 24–48 h to be tested for the sensitivity toward Gm.

B. glumae expression plasmids

The toxB fragment was rescued as the XbaI and HindIII fragment from the pET28a expression construct and cloned into the same sites of pLitmus28 (New England Biolabs) to generate pLitmus28-toxB. The toxB fragment was rescued as a KpnI and EcoRI fragment from pLitmus28-toxB and ligated in the same sites in pPROBE_GTkan for the toxB expression construct of nptIIp-toxB. The toxAB fragment including the 696 bp toxA upstream region (toxAp) and the 85 bp toxB downstream region was amplified with the toxAB-F/-R primer pair. The PCR product was cloned in pMD20, rescued as the XbaI and HindIII fragment, and ligated in the same sites in pBBR1MCS-5 to generate the toxAB expression construct. The toxABC fragment including the toxAp sequence and the 90 bp toxC downstream region was amplified with the toxAB-F/toxABC-R primer pair. The PCR product was cloned into pMD20, rescued as the XbaI and HindIII fragment, and ligated in the same sites in pBBR1MCS-5 to generate the toxABC expression construct. To generate the toxAB-toxC construct that has a gap between toxAB and toxC, the toxC fragment including the 83 bp upstream and the 90 bp downstream of toxC was amplified with the toxC-F/toxABC-Bm-R primer pair. The PCR product was digested with HindIII and BamHI and cloned into the XbaI and BamHI sites in pBBR1MCS-5 through combination with the XbaI and HindIII fragment of toxAB. In the resulting toxAB-toxC expression construct, there is 174 bp gap between toxB and toxC; the 177 bp toxB 3′-regions were in tandem. For the ribA expression construct, the ribA fragment including the 390 bp upstream and the 81 bp downstream of ribA was amplified with the ribA-F/-R primer pair. The PCR fragment was cloned into pMD20, rescued as the XbaI and HindIII fragment, and ligated in same sites in pBBR1MCS-5 to generate the ribA expression construct. To generate the ribA:toxC construct in which ribA is translationally coupled with toxC like the toxBC context, the ribA fragment including ribAp was amplified with the ribA-F/ribA(141224)-R primer pair and the toxC fragment including its 90-bp downstream region was amplified with the toxC(141224)-F/toxABC-R primer pair. Overlapping PCR was performed using the resulting PCR products and the ribA-F/toxABC-R primer pair. The final PCR product was cloned into pBBRMCS-5 after XbaI and HindIII digestion. All of the expression constructs were introduced into the ΔtoxB::kan mutant through the intergeneric conjugation as describe above.

Extraction and analysis of toxoflavin

LB liquid culture was routinely prepared in 50 mL culture volume in 250 mL flask. The culture was maintained for 24 h. The culture supernatant was obtained by centrifugation, and the 20 mL portion of the supernatant was subjected to organic extraction. The 20 mL supernatant extracted with 30 mL of the chloroform/methanol mixture (v/v 2:1). The chloroform layer was used for further experiments. The LB agar culture (~30 mL) was prepared by streaking method and maintained for 48 h. For the toxoflavin extraction, agar was cut into small pieces with a razorblade and chopped agar was mixed with 100 mL of chloroform to be kept overnight. The chloroform extract was then filtered through cotton. The resulting chloroform extracts were then concentrated in reduced pressure. The residue was dissolved in 1 mL of chloroform for spectroscopic analysis after proper dilution. Spectroscopic measurement of absorbance at 393 nm was previously used to determine the relative amount of toxoflavin for the samples in 80 % methanol (Chen et al. 2012). In the present spectroscopic analysis of toxoflavin, each sample was dissolved in chloroform and absorbance at 415 nm was adopted in evaluating the relative level of toxoflavin, guided by the UV–Vis absorption spectrum of the authentic toxoflavin in chloroform (see Fig. 4B). Authentic toxoflavin was kindly provided by Prof. Ingyu Hwang from Seoul National University.

Results and discussion

To determine whether ToxB is a RibA isozyme, ToxB and Bglu-RibA proteins were prepared using through E. coli expression (Fig. 2A, B). HPLC analysis of the reaction samples was performed using a previously described method for E. coli RibA characterization (Ritz et al. 2001). As shown in Fig. 2C and D, both the ToxB and Bglu-RibA recombinant proteins displayed time-dependent GTP consumption and formation of new products, which were assigned as the anomeric pair of 1, by referring to the previous E. coli RibA characterization study (Ritz et al. 2001). UV–Vis absorption spectra changes in both the ToxB and Bglu-RibA reactions matched the previously published data for RibA, displaying a maximum decrease and increase of absorbance at near 255 and 300 nm, respectively (Fig. 2E, F). The substrate (GTP) and the product (1) of GTP cyclohydrolase II have the characteristic UV absorption at near 255 and 300 nm, respectively. Thus, the UV–Vis absorption spectra change (a decrease at 255 nm and an increase at 300 nm) was a characteristic feature of GTP cyclohydrolase II reaction and has been used to monitor GTP cyclohydrolase II activity (Spoomamore et al. 2006). These enzyme experiments confirmed that ToxB is a biochemical equivalent to Bglu-RibA, at least when considering GTP cyclohydrolase II activity. The ToxB and Bglu-RibA kinetic parameters were 2.2 ± 0.27 and 1.6 ± 0.16 for k _cat (min⁻¹) and 21.6 ± 10.3 and 5.6 ± 2.4 for K _m (μM), respectively. These values are comparable to the previously reported data for GTP cyclohydorlase II enzymes (Ritz et al. 2001; Spoonamore et al. 2006). To reassure the GTP cyclohydrolase II activity, formate release was also measured with a coupled assay using formate dehydrogenase and NAD⁺. In this coupled assay, both ToxB and Bglu-RibA were also found to be active (data not shown). When guanine was tested for a cyclohydrolase activity, neither ToxB nor Bglu-RibA displayed the relevant activity (data not shown).

Fig. 2

Purification of the recombinant RibA (A) and ToxB (B) and their activity assay (C–F). (A) and (B) Sodium dodecyl sulfate–polyacrylamide gel electrophoresis of the six histidine-tagged recombinant RibA (A) and ToxB (B) proteins purified by a nickel affinity chromatography. The lanes indications are as follows: (M) molecular weight protein markers; (1) total soluble proteins; and (2–5) four-step imidazole elution. The 17 and 26 kDa markers are indicated with bars at the left. (C–D) HPLC traces of the RibA (C) and ToxB (D) GTP incubation products. (E–F) UV–Vis spectral changes of the RibA (E) and ToxB (F) reactions in a time-dependent manner. The zero time point was set as a blank

Although we substantiated that ToxB exerts a GTP cyclohydrolase II activity similar to RibA, we were unable to conclude that the early steps of toxoflavin biosynthesis were parallel to those for riboflavin biosynthesis. This is because ToxB could also have other unpredictable activities in addition to its RibA activity, making it specifically involved in toxoflavin biogenesis. To clarify this issue, we assessed whether a toxB is a redundant gene for Bglu-ribA in vivo via marker replacement gene inactivation of toxB. If a loss in toxB results in the loss of toxoflavin, then the role of toxB is not redundant for Bglu-ribA, thus implying that the biological role of toxB differs from that of ribA. Although toxB knockout has been previous demonstrated (Kim et al. 2004; Suzuki et al. 2004), the genetic methods used in these studies, vector-integrating gene disruption and transposon insertion, may have a suppressive effect on downstream gene transcription. Thus, we achieved toxB inactivation through the insertion of a kanamycin resistance gene (kan) lacking transcriptional terminator (Um et al. 2011) (Fig. 3A). In PCR analysis, 0.63 kb fragment, which was expected for the wild-type (WT) toxB, was evident in the PCR sample of the WT DNA (Fig. 3B, lane 1), and a 1.6 kb fragment, but not the 0.63 kb fragment, was amplified from the ΔtoxB::kan mutants (Fig. 3B, lane 2–4). Biosynthesis of toxoflavin was abolished in a ΔtoxB::kan mutant as found in spectrophotometric analysis of the organic extracts obtained from both agar and liquid cultures (Figs. 4, 5). This observation was initially interpreted as indicating that ribA was incapable of complementing for toxB, but failure to restore toxoflavin production in the ΔtoxB::kan strain with nptIIp-toxB suggested that the ΔtoxB::kan genotype may be incorrect, requiring further investigations. We have therefore prepared several complementation plasmids and tested their abilities to restore toxoflavin production in the toxB::kan mutant. Because the initial trial with nptIIp-toxB was unsuccessful in restoring toxoflavin production, native promoters upstream of toxA and ribA were included (coined toxAp and ribAp, respectively) in the following complementation constructs. The native toxAp region was previously determined by Kim et al. (2009). Bglu-ribA and an inosine-5-monophosphate dehydrogenase gene are divergently arranged with a 150 bp intergenic region. We intuitively adopted a 200 bp region upstream of ribA as ribAp. Using these sequences, we have prepared toxAB, toxABC, and ribA expression constructs. The translational coupling nature of toxBC (the starting codon of toxC overlaps with the termination codon of toxB, as ATGA) suggested the possibility that the co-translational nature of toxBC is critical to toxoflavin biosynthesis. To assess this issue, ribA:toxC (the TGA termination codon of ribA was engineered to overlap with the ATG start codon of toxC, to be ATGA) and toxAB-toxC (artificial interval of 200 bp between toxB and toxC) were also prepared.

Fig. 3

The targeted inactivation of toxB in B. glumae BGR1. (A) Genetic map of the toxB inactivation construct pTBKG-2.2, B. glumae WT, and the ΔtoxB::kan strain. (B) Analytical PCR of toxB with the toxB-F/R primer pair in WT (1) and three progeny of the ΔtoxB::kan strains (2–4). M indicates the DNA marker. The 0.5, 0.65, 0.8, 1.0, 1.5, and 2.0 kb fragments are highlighted with bars at the left

Fig. 4

Toxoflavin production of the B. glumae ΔtoxB::kan mutant and its gene complementation derivatives in LB agar plate cultures. (A) Absorbance at 415 nm (A₄₁₅) of the extracts from WT (1), ΔtoxB::kan/pBBR1MCS-5 (2), ΔtoxB::kan/nptII-toxB (3), and ΔtoxB::kan/the pBBR1MCS-5 derivatives with toxAB (4), toxABC (5), toxAB-toxC (6), ribA (7), and ribA:toxC (8). The absorbance value was calculated to be that in the original culture volume, i.e., A₄₁₅/mL of the culture. All of the experiments were performed in triplicate, and the standard derivations were used to draw the error bars. (B) UV–Vis absorption spectra of the extracts from WT (line), ΔtoxB::kan/pBBR1MCS-5 (unfilled triangles), and the ΔtoxB::kan transformants of nptII-toxB (filled triangles), toxAB (unfilled circles), toxABC (filled circles), and approximately 0.15 mM toxoflavin (grey filled diamonds)

Fig. 5

Toxoflavin production of the B. glumae ΔtoxB::kan mutant and its gene complementation derivatives in LB liquid plate culture. (A) A₄₁₅ of the extracts of WT (1), ΔtoxB::kan/pBBR1MCS-5 (2), ΔtoxB::kan/nptII-toxB (3), and ΔtoxB::kan/the pBBR1MCS-5 derivatives with toxAB (4), toxABC (5), and toxAB-toxC (6). (B) A₄₁₅ of the extracts from WT (1), ΔtoxB::kan/pBBR1MCS-5 (2), and the ΔtoxB::kan/the pBBR1MCS-5 derivatives with ribA (3), and ribA:toxC (4). The absorbance value was calculated to be that in the original culture volume, i.e., A₄₁₅/mL of the culture. All of the experiments were performed in triplicate, and the standard derivations were used to draw the error bars. The experiments shown in A and B were performed separately

In LB agar cultures prepared via a streaking method, neither nptII-toxB nor toxAB restored toxoflavin production, while both toxABC and toxAB-toxC did. The toxoflavin levels in the toxABC complementation were significantly higher than in toxAB-toxC and were comparable to that of WT (Fig. 4A). UV–Vis absorption spectra of selected samples are shown in Fig. 4B. This observation indicates that a co-expression of toxB and toxC is critical for toxoflavin biosynthesis and that translation coupling fortifies the co-expression effects. Considering that ToxC is a WD repeat domain protein, association of ToxB and ToxC can be envisioned. Contrary to the case of toxAB expression, ribA alone successfully induced toxoflavin biosynthesis in the ΔtoxB::kan mutant, with ribA:toxC having higher toxoflavin levels. This suggests that ribA with its plasmid copies (approximately 10 copies) can complement for the loss of toxB, even though the chromosomal copy was incapable of doing so. To check the possibility that the ribA 3′-UTR region negatively regulated ribA expression, 600 bp of the 3′-UTR region was included in another ribA expression plasmid. This resulting plasmid also induced toxoflavin production in the ΔtoxB::kan mutant (data not shown). Thus, we concluded that the multiple copies of ribA enhanced RibA levels and that this unusually high level of RibA could substitute for ToxB activity in the ΔtoxB::kan mutant. The transcription of toxAp may be tightly controlled; thus, an increase in the copy number did not result in an increase in expression level. The relatively slow growth on agar plates may be related to a tight regulation on toxAp expression. Toxoflavin biosynthetic genes have been very recently characterized in Pseudomonas protegens Pf-5 (Philmus et al. 2015). The authors found a trace amount of toxoflavin in an in-frame toxB deletion mutant and deduced that ribA complemented for the loss of toxB. With this observation, the authors suggest that ToxB is a RibA isozyme. However, there is no evidence that shows that ribA is truly responsible for toxoflavin production in the ΔtoxB mutant. In our experiment, we confirmed this notion by introducing ribA into the ΔtoxB::kan mutant and detecting substantial toxoflavin levels in the resulting strain. When toxE was inactivated in P. protegens Pf-5, toxoflavin was readily found, suggesting that ribD efficiently substituted for toxE in toxoflavin production (Philmus et al. 2015). It is important to note that the spectroscopic quantification may not only reflect toxoflavin but also its congeners, such as fervenulin and reumycin. However, we used the terminology toxoflavin production for simplicity.

In LB liquid cultures, all three of the toxB-containing constructs (toxAB, toxABC, and toxAB-toxC) partially restored toxoflavin production (Fig. 5A). The production of toxoflavin in the toxAB expression transformant is rather surprising because this strain failed to generate toxoflavin in the agar solid cultures (Fig. 4A). An evident difference between solid and liquid cultures is the growth rate; a liquid culture should display faster growth. We speculate that tox operon expression is highly induced at an idiophase of liquid culture and that co-expression of toxB and toxC is not a critical determinant for the integrity of toxoflavin biosynthetic machinery in this condition. Cellular protein synthesis dynamics or other cellular factors during growth in liquid culture may neutralize the requirement for toxBC co-expression. ribA successfully restored toxoflavin biosynthesis at the higher level with the ribA:toxC construct (Fig. 5B), which was also observed in the agar culture experiment (Fig. 4A).

Biochemical characterization, toxB inactivation, and the gene complementation experiments demonstrated that ToxB is a functional GTP cyclohydrolase II. This substantiated that toxoflavin biosynthesis shares the early steps with riboflavin biosynthesis, providing a platform to trace the unique biochemistry involved in toxoflavin biosynthesis. Hypothetically, 5,6-diaminouracil may not be a true intermediate; N–N bond formation may be coupled with a removal of the ribitol moiety from 2. ToxD might be a novel enzyme that mediates the enigmatic N–N bond formation. Operon structure and translational coupling between toxB and toxC appear to control toxoflavin productivity, implying the presence of a toxoflavin biosynthetic metabolon. It was recently demonstrated that the RibA product (1) and RibD deaminase product are toxic to E. coli and Arabidopsis thaliana and that a directed overflow mechanism by an N-glycosidase (COG3236) removes these reactive metabolites (however, a COG3236 gene could not be identified in the B. glumae genome) (Frelin et al. 2015). Frelin et al. (2015) assert that the formation of a metabolon, a multienzyme complex achieving a high metabolic flux by channeling intermediates from one enzyme to another, is a damage control mechanism against reactive intermediates in riboflavin biosynthesis. We propose that a toxoflavin biosynthetic metabolon is constituted by a scaffolding role of ToxC.