ABSTRACT
Coenzyme F420 is a redox cofactor involved in hydride transfer reactions in archaea and bacteria. Since F420-dependent enzymes are attracting increasing interest as tools in biocatalysis, F420 biosynthesis is being revisited. While it was commonly accepted for a long time that the 2-phospho-l-lactate (2-PL) moiety of F420 is formed from free 2-PL, it was recently shown that phosphoenolpyruvate is incorporated in Actinobacteria and that the C-terminal domain of the FbiB protein, a member of the nitroreductase (NTR) superfamily, converts dehydro-F420 into saturated F420. Outside the Actinobacteria, however, the situation is still unclear because FbiB is missing in these organisms and enzymes of the NTR family are highly diversified. Here, we show by heterologous expression and in vitro assays that stand-alone NTR enzymes from Thermomicrobia exhibit dehydro-F420 reductase activity. Metabolome analysis and proteomics studies confirmed the proposed biosynthetic pathway in Thermomicrobium roseum. These results clarify the biosynthetic route of coenzyme F420 in a class of Gram-negative bacteria, redefine functional subgroups of the NTR superfamily, and offer an alternative for large-scale production of F420 in Escherichia coli in the future.
IMPORTANCE Coenzyme F420 is a redox cofactor of Archaea and Actinobacteria, as well as some Gram-negative bacteria. Its involvement in processes such as the biosynthesis of antibiotics, the degradation of xenobiotics, and asymmetric enzymatic reductions renders F420 of great relevance for biotechnology. Recently, a new biosynthetic step during the formation of F420 in Actinobacteria was discovered, involving an enzyme domain belonging to the versatile nitroreductase (NTR) superfamily, while this process remained blurred in Gram-negative bacteria. Here, we show that a similar biosynthetic route exists in Thermomicrobia, although key biosynthetic enzymes show different domain architectures and are only distantly related. Our results shed light on the biosynthesis of F420 in Gram-negative bacteria and refine the knowledge about sequence-function relationships within the NTR superfamily of enzymes. Appreciably, these results offer an alternative route to produce F420 in Gram-negative model organisms and unveil yet another biochemical facet of this pathway to be explored by synthetic microbiologists.
INTRODUCTION
Organic coenzymes are crucial to the bioenergetics and metabolism of microorganisms. Being small molecules required for the catalytic activity of enzymes, they are also frequently required for biocatalytic applications. Flavins (flavin mononucleotide [FMN] and flavin adenine dinucleotide [FAD]), for example, are redox cofactors mostly acting as prosthetic groups in various enzymes, including reductases, hydroxylases, or Baeyer-Villiger monooxygenases (1). Nicotinamides (NAD+ and NADP+) are cofactors of dehydrogenases and often serve as freely diffusible cosubstrates involved in hydride transfer. They form pools of their reduced and oxidized forms in the cell, thus transporting redox equivalents between oxidizing and reducing pathways. In addition to these highly ubiquitous coenzymes, there are a few specialized examples that are restricted to some microbial phyla. One example is the coenzyme F420, which was originally discovered as a cofactor of methanogenesis in archaea (2). Intriguingly, F420 is also widespread in actinobacteria, where it performs versatile roles. In mycobacteria, it is involved in glucose catabolism, respiration (3, 4), detoxification of nitrosative stress (5), and cell wall biogenesis (6, 7), as well as antibiotic resistance (3, 8) and prodrug activation (9), for instance. In Streptomycetes, F420 has been shown to participate in antibiotics biosynthesis (10, 11).
From a chemical point of view, F420 is a deazaflavin with, among other modifications, a nitrogen-for-carbon substitution at C-5 of the redox-active chromophore compared to the well-known flavins (12). Although structurally related to flavins, deazaflavins rather resemble nicotinamides by acting as obligatory hydride carriers mediating two-electron transfers (12). Single-electron transfers and oxygen activation as known from flavins are not supported by deazaflavins. On the other hand, reduced deazaflavins are significantly more stable than flavins under oxygen exposure (12). It is not fully understood why F420 is needed as an additional coenzyme in some organisms, although a physiological explanation could be the option to separate another pool of redox equivalents from commonly used nicotinamides. Especially in anaerobic methanogens, another reason might be the extraordinary redox potential of F420 (E°′ = −340 mV), making it an ideal cofactor to transfer electrons from H2 (−410 mV) to NADP (−320 mV) (13). Since some F420-dependent reductions are chemically challenging or stereospecific, deazaflavin-dependent enzymes have risen increasing interest for applications in biocatalysis (14–16). Unfortunately, F420 is produced in relatively low yields by microorganisms, thus limiting its use for in vitro applications (16). Even engineered strains of Mycobacterium smegmatis generate only a few milligrams per liter of culture (17). The use of an “unnatural” F420 derivative was proposed as a solution recently (18). Consequently, a better understanding of F420 biosynthesis to support attempts for increased production by metabolic engineering is highly desired.
The biosynthesis of F420 (Fig. 1) branches off from the flavin biosynthetic route before the formation of 7,8-dimethyllumazine from 5-amino-6-aminoribityl-uracil (2). The latter intermediate is condensed with l-tyrosine by the key enzyme deazaflavin synthase, a radical S-adenosylmethionine enzyme with two iron-sulfur clusters (19–21). In some organisms, including the archaea, deazaflavin synthase is composed of two subunits (CofG and CofH), while in bacteria the dual-domain FbiC is more common (21). The product of deazaflavin synthase represents the redox-active core moiety ribityldeazaflavin (commonly called FO). Four more steps are required to form the mature F420 molecule, which can be described as FO decorated with an oligo-γ-glutamate tail bridged by a 2-phospho-l-lactate (2-PL) moiety.
Biosynthesis of the “bridge and tail” moieties of coenzyme F420 starting from FO. Recently, it was shown that CofC and CofD activate and transfer PEP in a GTP-dependent reaction to install the bridge moiety containing a phosphate group. The C-terminal domain of FbiB belonging to the FMN-dependent NTR superfamily reduces the resulting dehydro-F420 (DF420-0) to F420-0. The N-terminal domain of FbiB (CofE) acts as glutamyl ligase to synthesize the oligoglutamate tail.
The biosynthesis of the 2-PL “bridge” has recently been revised (22). For a long time, the accepted biosynthetic model proposed that free 2-PL was activated by CofC (2-phospho-l-lactate guanylyltransferase) (23) to form a short-lived, guanylylated intermediate (LPPG) (24). In turn, LPPG is a substrate of CofD (2-phospho-l-lactate transferase), which installs the 2-PL at the ribityl side chain of FO (25). The origin of 2-PL, however, has remained elusive for more than 15 years. Just recently it was shown that, at least for actinobacteria, it is not 2-PL but phosphoenolpyruvate (PEP) that is introduced by CofC/D (22). The resulting unstable dehydro-F420 (DF420) is then reduced to F420 by the C-terminal domain of FbiB (22, 26), a member of the ancient FMN-binding nitroreductase (NTR) superfamily.
This protein superfamily is widespread in nature and evolved a broad spectrum of catalytic activities ranging from the eponymous nitroreduction to dehydrogenation, dehalogenation, and even flavin fragmentation during cobalamine synthesis (i.e., BluB reaction) (27). The N-terminal domain of FbiB is a homolog of the γ-glutamyl transferase CofE and catalyzes the biosynthesis of the oligo-γ-glutamate tail (28). The dual-domain protein FbiB has been observed almost exclusively in actinobacteria, with the one archaeal exception being Lokiarchaeum sp. (22). In archaea or F420-producing Gram-negative bacteria, free-standing CofE homologs are common, while the C-terminal NTR domain is missing. It is not known how the reduction of DF420 is achieved in these organisms or whether it happens at all. We recently showed that in the Gram-negative bacterium Paraburkholderia rhizoxinica, an alternative “bridge” is formed so that reduction of the double bond was not required to obtain a stable F420 derivative. Specifically, in this organism, CofC accepts 3-phospho-d-glycerate instead of PEP to form 3PG-F420, a novel derivative of F420. In addition, we confirmed previous results that 2-PL is the actual substrate of archaeal CofC/D in vitro and could therefore still be the relevant substrate in these organisms.
Therefore, we sought to delineate how F420 biosynthesis might proceed in other F420-producing Gram-negative organisms where FbiB is missing. Here, we combined in vivo and in vitro evidences to show that in thermophilic bacteria belonging to the class Thermomicrobia (phylum Choroflexi), F420 biosynthesis proceeds via DF420, but reduction of DF420 is catalyzed by a stand-alone dehydro-F420 reductase domain that is distantly related to the C-terminal domain of FbiB.
RESULTS AND DISCUSSION
Bioinformatics studies.In order to find genetic hints on how the bridge moiety of F420 might be formed in Gram-negative organisms and archaea, we reanalyzed sequenced genomes of F420-producing species for the presence of the fbiB gene encoding the dual-domain enzyme FbiB. While a (stand-alone or fused) cofE gene is commonly present in F420-producing organisms (29, 30), the fbiB gene encoding a dual-domain protein was missing in the genome of most Gram-negative (putative) F420 producers. In fact, we found only one dual domain FbiB protein (WP_096480527.1 from Pseudomonas frederiksbergensis) in a Gram-negative bacterium by domain architecture search. When probing for homologs of the C-terminal NTR domain of FbiB from M. tuberculosis (DF420 reductase domain) in these organisms by BLAST, we realized that genes encoding related proteins produce hits in almost any genomic background, given the virtual ubiquity of the NTR subfamily in prokaryotes. Consequently, further functional classification was needed to predict whether candidates were possibly involved in F420 reduction. Curiously, three candidates annotated as “nitroreductase” appeared to be organized in one genomic locus together with cofC and cofD in the thermophilic Chloroflexi, as it was shown before (30), namely, WP_012873063.1 from Sphaerobacter thermophilus, WP_051913893.1 from Thermorudis peleae, and WP_052294161.1 from Thermomicrobium roseum (Fig. 2a). Analysis of their primary sequence by an NCBI conserved domain search (Table 1) detected a partial hit to “F420-0–gamma-glutamyl ligase” (PRK13294) suggesting that these proteins might be related to “FbiB.” This functional categorization is of course misleading, since FbiB is a dual-domain protein and the nitroreductase is certainly not catalyzing the oligoglutamylation. To test whether the enzymes in question form any clades that could reflect functional identity, we aligned the sequences in question along with representatives of subfamilies of the nitroreductase superfamily and computed a phylogenetic network (Fig. 2b). Reflecting their phylogenetic relatedness, the thermomicrobial enzymes clustered closely together. Strikingly, however, they were located on a ancient branch together with the C-terminal domains of actinobacterial FbiB representatives. This relationship further substantiated our assumption that these free-standing nitroreductase candidates could indeed participate in the F420 biosynthesis. It should be noted that there is another NTR encoded in the Thermomicrobium roseum genome (WP_012642630.1) that rather belonged to the “Arsenite_oxidase” subfamily (cd02135) and was not clustered with F420 biosynthesis genes.
Phylogenetic analysis of NTR enzymes. (a) Schematic representation of the genomic context (operons) of NTR enzymes from Thermomicrobia. Arrows indicate coding sequences. The proximity of NTR encoding genes to cofC and cofD suggested involvement in F420 metabolism. DMT, drug/metabolite transporter. (b) Phylogenetic networking of selected NTR family enzymes revealing a relationship between thermomicrobial NTRs and C-terminal domains of FbiB enzymes (putative DF420 reductase family). The analysis includes representatives of important subfamilies of the NTR superfamily, as well as NTR enzymes deduced from the genomes of selected F420-producing organisms (Archaeoglobus fulgidus, Methanocaldococcus jannaschii, Oligotropha carboxidovorans, Paracoccus denitrificans, and Paraburkholderia rhizoxinica). Sequences experimentally analyzed in this study are indicated in boldface. The network was created using the neighbor-net algorithm. The scale bar represents the uncorrected pairwise distance. Source organisms, accession numbers, and subfamily assignments of all proteins used are provided in Table 1.
Nitroreductase family enzymes analyzed in this studya
Next, we located the DF420 reductase candidates in a sequence similarity network that was constructed during recent bioinformatics studies of about 25,000 NTR sequences (27, 31). These analyses clustered the NTR superfamily into 22 subgroups, of which only 14 contained biochemically characterized members, including the C-terminal NTR domain of FbiB proteins. However, only dual-domain FbiB proteins were classified as “FbiB,” while the Thermomicrobium and Sphaerobacter NTRs mentioned above were assigned to the “NTR Hub 2nd level subgroup 10.” The Thermorudis peleae NTR was not present in the data set, but the results presented here suggest that it would, most likely, fall into the same subgroup. Akiva et al. (27) proposed four diagnostic residues, namely, W317, H358, Y360, and R365 of the M. tuberculosis FbiB homolog, to be diagnostic for the FbiB subgroup. A multiple sequence alignment (Fig. 3) showed that only three of the four diagnostic residues (W317, Y360, and R365) were conserved in thermomicrobial sequences, whereas H358 was replaced by an aspartate.
Multiple sequence alignment (MUSCLE) of NTR sequences. Residue numbers refer to the reference sequence (M. tuberculosis FbiB). Putative diagnostic F420-binding site residues are marked with asterisks and framed with red boxes. W317, Y360, and R365 are highly conserved in DF420 reducing NTRs (the names are shown in red). H358 is conserved in FbiB homologs but is substituted by an aspartate in thermomicrobial sequences. Amino acids that are identical to the reference sequence in each alignment position are colored. Colors represent different amino acids.
Heterologous coexpression experiments.In order to experimentally examine whether the candidate enzymes were able to reduce DF420, we set up a heterologous in vivo test system employing a two-plasmid technique (Fig. 4a). We previously reported on the construction of plasmid pDB070 harboring among other genes fbiC from Paraburkholderia rhizoxinica, as well as cofC and cofD (both codon-optimized) from Methanocaldococcus jannaschii (32). When expressed in Escherichia coli, this plasmid yielded DF420-0 according to liquid chromatography-mass spectrometry (LC-MS) analyses, but no classical F420. In order to test the activity of the thermomicrobial NTR superfamily enzymes their corresponding codon-optimized genes were coexpressed from a second plasmid (Table 2) together with the F420 biosynthesis genes encoded on pDB070. The resulting E. coli strains were grown under inducing conditions, their metabolites extracted with organic solvents and analyzed by LC-MS (Fig. 4b). All three coexpression strains harboring thermomicrobial NTR genes produced saturated F420, while control strains containing empty vectors did not. As additional controls, we included two genes from F420-producing model organisms belonging to the archaeal domain of life encoding NTR superfamily enzymes, namely, WP_064496814.1 from Methanocaldococcus jannaschii DSM 2661 and WP_010877737.1 from Archaeoglobus fulgidus. Their corresponding enzymes, however, were not related to FbiB according to CD search (Table 1). Accordingly, they did not show any DF420 reductase activity in our test system. The heterologous expression experiment was already a strong indication that the NTR enzymes from Thermomicrobia were indeed able to reduce DF420 to F420 (DF420 reductase activity). However, it should be noted that E. coli also harbors members of the NTR superfamily (Table 1), including NfsA and NfsB, which catalyze nitroreductions, but might serve other physiological functions in the cell (33). Although we never observed DF420 reduction by the control strain, the induction of unknown reductases of E. coli could not be fully ruled out.
In vivo reduction of DF420 by coexpression of NTR encoding genes. (a) Two-plasmid system used for coexpression experiments in E. coli. Plasmid pDB070 (pETDuet backbone) harbors genes for the production of DF420 in E. coli but no factors leading to the formation of mature F420. NTR-encoding genes are coexpressed from a second vector with pCDF backbone. The second plasmid is illustrated with plasmid pDB074, which carries the T. roseum FbiB-like nitroreductase. (b) Formation of F420-0 by coexpression of pDB070 and five individual plasmid-borne NTR encoding genes in E. coli. I, T. roseum (pDB074); II, S. thermophilus (pDB080); III, T. peleae (pDB079); IV, A. fulgidus (pDB082); V, M. jannaschii (pDB073); control, empty pCDFDuet-1.
Plasmids used in this study
In vitro experiments.In order to more rigorously test the hypothesis that the thermomicrobial NTR enzymes catalyze FMN-dependent DF420 reduction, we opted for an in vitro experiment with purified enzyme. First, we procured DF420 as a substrate for the activity assay. Unfortunately, DF420 is produced in very low yields by microorganisms and is inherently unstable. Furthermore, it turned out to be difficult to separate DF420 from F420 species, as well as further background metabolites, by chromatographic techniques. Therefore, we decided to generate DF420 in vitro with the help of CofC and CofD. Starting with FO and PEP as the substrates, this pair of enzymes is able to generate DF420 in a GTP-dependent reaction (22, 32). The reaction product DF420 could be directly purified by solid-phase extraction. Although the amounts obtained were low, they were enough for qualitative enzymatic assays monitored by LC-MS analysis. For this purpose, the NTR WP_052294161 from T. roseum was produced as a hexahistidine-fusion protein in E. coli BL21(DE3) and purified using immobilized metal affinity chromatography (IMAC). The protein eluted as a yellow fraction from the IMAC column, already indicating that it might be a flavoprotein. SDS-PAGE confirmed a high degree of purity and the expected molecular weight of the fusion protein (see Fig. S1 in the supplemental material). Acid denaturation, followed by LC-MS analysis, revealed that the protein contained FMN as a cofactor, as expected for a member of the NTR superfamily (Fig. 5a). To test for DF420 reductase activity, we combined the enzyme with purified DF420 as a substrate, as well as NADH and recombinant flavin reductase (FRE) from E. coli as a regeneration system. Analysis of the reaction products by LC-MS indeed demonstrated that F420 was formed in the presence of all crucial reaction components (Fig. 5b). Controls lacking the NTR were not able to yield any F420 (Fig. 5c). These results proved that WP_052294161 from T. roseum can catalyze DF420 reduction.
In vitro DF420 reductase activity assay of NTR from T. roseum. (a) LC-MS analysis revealed FMN as the flavin cofactor coeluting with the NTR domain from T. roseum. An extracted ion chromatogram (XIC) resulting from sample (lower trace) and the total ion current (TIC) of flavin standard (upper trace) mix are shown. (b) LC-MS analysis of F420 formed by the in vitro reaction of DF420 with NTR in the presence of flavin reductase FRE and NADH as regeneration system. I, extracted ion chromatogram (XIC) of DF420-0 (substrate); II, XIC of F420-0 (product). Expected m/z ([M+H]+, 5 ppm mass tolerance): DF420-0, 514.08574; F420-0, 516.10139. Peak heights above the baseline reflect intensities in arbitrary units. (c) Control reaction lacking the NTR enzyme (description of traces, as in panel b). No formation of F420 was observed in the absence of NTR.
Metabolome and proteome analyses.To further validate this biosynthetic model, we analyzed metabolites from T. roseum by LC-MS. Unexpectedly, we found only FO and traces of DF420-0, but no classical F420, in small-scale cultivations. It was sensible to consider that WP_052294161 was not produced by T. roseum under the growth conditions tested, which led us to analyze protein extracts by peptide mass-fingerprinting. The proteomics approach revealed that the NTR was actually produced under the culture conditions tested (see Data Set S1 in the supplemental material). Therefore, we scaled up the cultivation to a total volume of 2.4 liters and enriched F420 derivatives by solid-phase extraction combined with anion-exchange chromatography. In large-scale extracts, we detected not only F420-n but also DF420-n by LC-MS/MS (see Fig. S2 to S5 in the supplemental material). Surprisingly, the levels of DF420 species according to area under the curve summed up to about 20% of the F420 levels detected (Fig. 6). These unexpectedly high levels of DF420 suggest that the reduction of DF420 might be slow under certain cultivation conditions and/or might be tightly regulated. It is also worth mentioning that both DF420 and F420 were oligo-γ-glutamylated, suggesting that CofE from T. roseum accepts both molecules as substrates.
Distribution of F420-n and DF420-n species from T. roseum. The areas under the curve (arbitrary units) of characteristic ions [M+H]+ with increasing length n of the oligo-γ-glutamate chain are shown. Areas were calculated from the traces depicted in Fig. S2 and S4 in the supplemental material.
Conclusion.Consequently, we conclude that F420 biosynthesis in Thermomicrobia proceeds via DF420, which is reduced by a free-standing FMN-dependent nitroreductase (DF420 reductase) that is distantly related to the C-terminal domain of FbiB. Thus, we shed light on the biosynthetic route to F420 in a class of Gram-negative bacteria. In addition, we add valuable information to the growing body of knowledge about the functional space of NTR superfamily enzymes. (27, 31). Obviously, the FbiB subfamily of NTR enzymes obtained by sequence-similarity networks is not the only one of the currently defined subfamilies that is involved in F420 biosynthesis. The enzymes studied here are members of the “NTR Hub 2nd level subgroup 10.” Furthermore, the four diagnostic residues deduced from 3D structure and alignments of M. tuberculosis FbiB (W317, H358, Y360, and R365) are being refined by our results, where we infer the substitution of H358 by an aspartate in this subfamily. W317 can be accounted as the most informative residue because it is involved in stacking interaction with the deazaflavin ring system of F420. Thus, our study will help future efforts to improve these classifications and thus advance the understanding of sequence-function relationships of NTR superfamily enzymes. Finally, we found a new way to produce coenzyme F420 in E. coli and the enzymes revealed here might be used in the future to engineer coenzyme F420 production in E. coli toward higher yields.
MATERIALS AND METHODS
Bioinformatics analysis of NTR subfamily enzymes.In total, 21 sequences were retrieved from public databases (Table 1) and submitted to the NCBI conserved domain search (34). Sequences were aligned using the MUSCLE algorithm (35) implemented in Geneious Prime (2019.2.3). Alignments were trimmed to remove the N-terminal CofE domain of FbiB homologs (alignment position 330). The remaining alignment was exported to Nexus format and used as input for SplitsTree 4 (36) to construct a phylogenetic network (neighbor-net).
Microorganisms and culture conditions.The E. coli strains Stellar and Top10 were used as hosts for cloning and plasmid propagation, whereas E. coli BL21(DE3) was used for heterologous protein production. In both cases, the bacteria were cultivated at 37°C and 210 rpm in Luria-Bertani medium supplemented with 50 μg/ml of the appropriate antibiotic(s) to ensure selective pressure. T. roseum DSM 5159 was grown at 65°C and 180 rpm in DSMZ medium 592 (1 g/liter yeast extract, 1 g/liter tryptone, 1.3 g/liter [NH4]2SO4, 0.247 g/liter MgSO4⋅7H2O, 0.28 g/liter KH2PO4, 0.074 g/liter CaCl2⋅2H2O, 0.019 g/liter FeCl3⋅6H2O [pH 8.5]; 1 ml/liter trace elements: 0.18 g/liter MnCl2⋅4H2O, 0.44 g/liter Na2B4O7⋅10H2O, 0.022 g/liter ZnSO4⋅7H2O, 0.005 g/liter CuCl2⋅H2O, 0.003 g/liter NaMoO4⋅2H2O, 0.003 g/liter VOSO4⋅2H2O [pH 2.0]) or in 10× CPS (5 g/liter peptone, 2.5 g/liter sucrose, 0.1 g/liter nitrilotriacetic acid, 0.1 g/liter MgSO4⋅7H2O, 0.006 g/liter CaCl2⋅2H2O, 0.008 g/liter NaCl, 0.689 g/liter NaNO3, 0.103 g/liter KNO3, 0.044 g/liter Na2HPO4, 1 ml/liter FeCl2 solution [0.44 g/liter FeCl2⋅4H2O], 1 ml/liter Nitsch trace element solution [0.5 ml/liter concentrated H2SO4, 2.8 g/liter MnCl2⋅4H2O, 0.5 g/liter ZnSO4⋅7H2O, 0.5 g/liter H3BO3, 0.16 g/liter CuSO4⋅5H2O, 0.03 g/liter Na2MoO4⋅2H2O, 0.046 g/liter CoCl2⋅6H2O]).
Construction and cloning of plasmids.All gene sequences, plasmids, and oligonucleotides used in this work are summarized in Tables 1 to 3. Synthetic genes and constructs were optimized for expression in E. coli and were purchased from Biocat. PCRs were carried out using Q5 high-fidelity polymerase (New England Biolabs), except when intended for colony PCR, where DreamTaq polymerase (Thermo Scientific) was used. Diagnostic restriction digests were carried out using NEB restriction enzymes (New England Biolabs). DNA Sanger sequencing was performed by Eurofins Genomics.
Oligonucleotides used in this study
Genes encoding FbiB-like reductases from Methanocaldococcus jannaschii, Thermomicrobium roseum, and Archaeoglobus fulgidus were cloned in pCDFDuet-1 using the FastCloning protocol (37). The vector was linearized by PCR using the primers oDB081/oDB123, whereas the genes were amplified with the primer pairs oDB134/oDB135, oDB136/oDB137, and oDB148/oDB149. The new constructs were named pDB073, pDB074, and pDB082, respectively. Genes encoding FbiB-like reductases from Thermorudis peleae and Sphaerobacter thermophilus were purchased as pCDFDuet-1 constructs, with the coding sequence (CDS) located between the NcoI and EcoRI recognition sites (pDB079 and pDB080, respectively).
The generation of pFS04 and pDB070 was described previously (32) M. jannaschii cofC (codon-optimized CDS of WP_064496647.1) was purchased as a DNA fragment and was subcloned to pET28a(+) by FastCloning with the primer pairs pET28a_FP/pET28a_RP (vector) and oDB122/oDB123b (cofC) to yield pMH03. pMH12 was purchased as a synthetic pET28a(+) construct, where the codon-optimized CDS encoding the FbiB-like nitroreductase of T. roseum was flanked by NcoI and HindIII recognition sites. pMM14 (38) encoding the flavin reductase (FRE, WP_074551252.1) from E. coli was kindly provided by Dirk Hoffmeister from the Friedrich Schiller University (Jena, Germany).
In vivo reduction of DF420.E. coli BL21(DE3)/pDB070, a strain able to produce DF420-n, was individually transformed with compatible plasmids encoding FbiB-like nitroreductases from different microorganisms—pDB073 (M. jannaschii), pDB074 (T. roseum), pDB079 (T. peleae), pDB080 (S. thermophilus), and pDB082 (A. fulgidus)—in order to observe the formation of F420-n upon reduction of DF420-n. The expression of deazaflavin biosynthetic gene clusters and of individual genes encoding NTR enzymes, as well as heterologous protein production and small-scale biosynthesis and purification of deazaflavins, was carried out as described previously (32).
Production and purification of recombinant enzymes.FbiB-like nitroreductase from T. roseum and flavin reductase from E. coli (on pMM14), as well as CofC and CofD from M. jannaschii (on pMH03 and pFS04, respectively), were purified as N-terminal hexahistidine fusion proteins after heterologous expression in E. coli BL21(DE3) as described previously (32). In short, bacteria transformed with corresponding expression plasmids were grown at 37°C until an optical density at 600 nm (OD600) of 0.7 was reached. Afterward, gene expression was induced by addition of IPTG (isopropyl-β-d-thiogalactopyranoside; 1 mM final concentration), and cultivation continued for ca. 20 h at 16°C. After cell lysis by sonification and centrifugation, the resulting protein was purified from the cleared lysate by immobilized-metal affinity chromatography on a Ni-NTA column. His-tagged fusion proteins were eluted by a gradient of increasing imidazole concentration and desalted on a PD-10 column.
Analysis of flavin cofactors.Purified recombinant NTR (100 μl) resulting from a 100-ml culture was heated at 75°C for 30 min, acidified by addition of 5 μl of glacial acetic acid, and finally dried at 60°C in a SpeedVac (Thermo Scientific). The residue was suspended in 100 μl of purified H2O, centrifuged at full speed to remove insoluble particles, and subjected to LC-MS analysis as described. A solution of flavin cofactors (0.15 mM each) was used (FMN, FAD, and riboflavin) as a standard.
In vitro production of DF420-0.DF420-0 was produced in vitro by a coupled enzyme reaction (23) with the purified fractions of hexahistidine fusion proteins of CofC and CofD from M. jannaschii. Reaction setup and conditions were similar to what we reported previously (32). A 1-ml reaction mixture consisted of 50 mM HEPES buffer (pH 7.5), 2 mM GTP, 5 mM MgCl2, 50 ng of FO, 35 μM CofD, and 0.5 mM phosphoenolpyruvate. The reaction started with the addition of 25 μM CofC. The mixtures were incubated at 70°C and 300 rpm for 10 min, quenched with 20% formic acid (FA), and directly loaded onto a C18 SPE cartridge. The cartridge was washed with 20% methanol (MeOH) plus 0.2% FA and then eluted with 30% MeOH plus 0.2% FA. The eluate was dried in a SpeedVac and suspended in 150 μl of water.
In vitro DF420-0 reductase assay.To test the hypothesis that FbiB-like nitroreductases could reduce DF420 to F420 in vitro, enzyme assays analogous to those described by Bashiri et al. (22) were prepared. The 50-μl reaction mixtures consisted of 50 mM HEPES buffer (pH 7.5), 100 mM KCl, 5 mM MgCl2, 15 μM DF420-0, 10 mM dithiothreitol (DTT), 15 μM FMN, and 1.5 mM NADH. Enzymes were added at concentrations of 40 μM (NTR) and 0.2 μM (FRE). The reaction mixture was incubated at 37°C and 300 rpm for 1 h and stopped by the addition of 20% FA (final). Proteins were precipitated by the addition of 200 μl of MS-grade acetonitrile followed by centrifugation (17,000 × g, 30 min). The solvent was removed by evaporation under reduced pressure, and the dried reaction extract was suspended in 100 μl of MeOH and analyzed by LC-MS as described earlier (32).
Analysis of F420 species from T. roseum.Enrichment and analysis of F420 species were performed by anion-exchange chromatography, followed by solid-phase extraction and LC-MS as described previously (32).
Proteomics analysis of T. roseum.For protein identification we followed the approach of in-gel digestion for mass spectrometric characterization of proteins and proteomes (39). Cells from 600 ml of culture at an OD600 of 1.5 were harvested by centrifugation, washed with Tris-HCl buffer (20 mM, pH 8), and resuspended in 6 ml of the same buffer. Cell lysis was achieved by sonication (ten cycles of 1 min with a 30-s break after each cycle, 50% duty cycle, and 50% power) at 4°C. Cell lysate was centrifuged at 5,870 × g and 4°C for 30 min. Cleared cell lysate (20 μl) was mixed with SDS-PAGE loading buffer and heated for 10 min at 95°C. Proteins were separated by SDS-PAGE (40). After Coomassie blue staining and destaining, the gel was washed twice in deionized water for 15 min. The lane with the specific sample was then cut out of the SDS-PAGE gel and divided into 12 fractions. The fractions were further cut into pieces. The pieces were transferred into 1.5-ml reaction tubes and washed with a mixture of 40 μl each of pure acetonitrile and fresh NH4CO3 buffer (50 mM). After 15 min, the supernatant was removed, and 80 μl of acetonitrile was added. Acetonitrile was removed when the gel pieces had decreased in size, and 80 μl of the NH4CO3 buffer was added to rehydrate the gel. After 5 min of incubation, 80 μl of acetonitrile was added, and the mixture was incubated for 15 min. Subsequently, the supernatant was removed, and 80 μl of acetonitrile was added and removed again. The gel pieces were then dried in a SpeedVac. When the gel pieces were completely dry, they were swollen in 10 mM DTT in NH4CO3 buffer at 56°C for 30 min. After cooling to room temperature, the samples were washed three times with 40 μl each of acetonitrile and NH4CO3 buffer with incubation times of 5 min per washing step. Liquid was removed, and the dried gel pieces were rehydrated with 20 μl of sequencing-grade trypsin (Sigma, 10 μg/ml in NH4CO3) and incubated at 37°C. After 1 h, 100 μl of NH4CO3 buffer was added, and the suspension was incubated overnight at 37°C. The supernatant was removed, and 160 μl of extraction buffer, consisting of a 1:1 (vol/vol) mixture of acetonitrile and 0.1% formic acid, was added. The tubes were incubated at 37°C for 1 h and sonicated for 5 min in a water bath. The supernatant was transferred into a fresh tube. The gel pieces were then resuspended in 40 μl of acetonitrile and sonicated for another 5 min. The resulting supernatant was added to the fresh tube, which was then dried completely in a SpeedVac. In the end, 20 μl of the extraction buffer was added to the tube and sonicated for 5 min.
This solution was transferred into high-performance liquid chromatography (HPLC) vials and analyzed by ultra-HPLC coupled with mass spectrometry (UHPLC-MS). The separation was performed with an Aeris Peptide XB-C18 column (150 by 2.1 mm, 1.7 μm, 100 Å; Phenomenex) on a Dionex Ultimate3000 system combined with a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) with a heated electrospray ion source (HESI). Water (A) and acetonitrile (B) served as mobile phase and were both acidified with 0.1% formic acid. The gradient of these two components of the liquid phase was as follows: 0 to 2 min, 3% B; 9 to 38 min, 9% to 55% B; and 39 to 44 min, 97% B, with a flow rate of 200 μl/min at 40°C. Mass spectrometry was carried out within an m/z range of 375 to 2,000, followed by a data-dependent MS2 analysis. MS1 measurements were performed at a resolving power of 70,000 (FWHM at m/z 200), with an injection time of 100 ms and an automatic gain control (AGC) target of 1e6, and MS2 experiments were performed at a 17,500 resolving power. The isolation window was set to 2.0, and the normalized collision energy (NCE) was set to 30 and the dynamic exclusion time to 15 s. A protein search database was obtained by in silico translation of the Thermomicrobium roseum genome (NCBI accession numbers NC_011959 and NC_011961 for the genome and plasmid, respectively). Peptide fragmentation spectra were searched against theoretical mass spectra using MaxQuant (41). Here, up to three missed tryptic cleavages were tolerated by the software. Moreover, methionine oxidation was allowed as variable modification.
Data availability.LC-MS data are freely available via the DRYAD repository (https://doi.org/10.5061/dryad.xd2547ddk).
ACKNOWLEDGMENTS
We thank the Carl Zeiss Foundation, the Deutsche Forschungsgemeinschaft (DFG, project 408113938) and the Leibniz Association for funding.
We declare there are no conflicts of interest.
FOOTNOTES
- Received 23 February 2020.
- Accepted 7 April 2020.
- Accepted manuscript posted online 10 April 2020.
Supplemental material is available online only.
- Copyright © 2020 American Society for Microbiology.