Abstract
Coenzyme F420 is a specialized redox cofactor with a highly negative redox potential. It supports biochemical processes like methanogenesis, degradation of xenobiotics or the biosynthesis of antibiotics. Although well-studied in methanogenic archaea and actinobacteria, not much is known about F420 in Gram-negative bacteria. Genome sequencing revealed F420 biosynthetic genes in the Gram-negative, endofungal bacterium Paraburkholderia rhizoxinica, a symbiont of phytopathogenic fungi. Fluorescence microscopy, high-resolution LC-MS, and structure elucidation by NMR demonstrated that the encoded pathway is active and yields unexpected derivatives of F420 (3PG-F420). Further analyses of a biogas-producing microbial community showed that these derivatives are more widespread in nature. Genetic and biochemical studies of their biosynthesis established that a specificity switch in the guanylyl transferase CofC re-programmed the pathway to start from 3- phospho-D-glycerate, suggesting a rerouting event during the evolution of F420 biosynthesis. Furthermore, the cofactor activity of 3PG-F420 was validated, thus opening up perspectives for its use in biocatalysis. The 3PG-F420 biosynthetic gene cluster is fully functional in E. coli, enabling convenient production of the cofactor by fermentation.
Significance statement Cofactors are crucial helper molecules for enzymatic reactions. Coenzyme F420 is an unusual redox cofactor found in prokaryotes involved in valuable enzymatic processes like biogas production, bioremediation, and antibiotics biosynthesis. Thus, it has attracted increasing interest in the field of biocatalysis and metabolic engineering. We discovered derivatives of F420 from bacteria living within fungal mycelia. Surprisingly, these molecules have been long overlooked in biogas-producing organisms. Our studies on their biosynthesis make an important contribution to validating and refining the just recently revised biochemical pathway of F420 formation. The novel cofactor is highly active and easily accessible, thus representing an ideal system for biotechnological applications. Its discovery will guide projects to unearth bioprocesses and enzymes involving F420-derivatives in the future.
Author contributions and conflict of interest Daniel Braga performed research, analyzed data (molecular biology, malachite green assay, mass spectrometry) and contributed to writing the manuscript, Daniel Last performed research and analyzed data (structure elucidation, Fno assay, biogas plant studies), Mahmudul Hasan performed research (CofC/D enzyme assays), Huijuan Guo performed research and analyzed data (structure elucidation), Daniel Leichnitz performed research (chemical synthesis), Zerrin Uzum performed research (microscopy), Ingrid Richter performed research (microscopy), Felix Schalk performed research (cofE constructs), Christine Beemelmanns designed research, acquired funding, analyzed data (structure elucidation, synthesis) and edited the manuscript, Christian Hertweck designed research, acquired funding and edited the manuscript, Gerald Lackner designed the study, acquired funding and wrote the original manuscript. The authors declare no conflict of interest.
Introduction
Cofactors complement the catalytic power of many enzymes and thus play a key role in virtually all metabolic pathways. Knowledge of their catalytic functions and biosynthesis is highly important for the understanding of biochemical reactions as well as their application in biocatalysis and biotechnology. An important subclass comprises redox-cofactors that mediate electron transfer between molecules. The deazaflavin coenzyme F420 (Fig. 1) is a specialized redox cofactor with a lower redox potential (−350 mV) than NAD (1). This feature makes F420 an ideal electron carrier between H2 and NAD(P) in methanogenesis and renders it a strong reducing agent for challenging reactions in biocatalysis (2). For instance, enzymatic processes involving F420 facilitate the degradation of pollutants like aromatic nitro compounds (3) or the carcinogen aflatoxin (4). Furthermore, F420-dependent enzymes are valuable for asymmetric ene reductions (5, 6). In actinomycetes, coenzyme F420 is involved in the biosynthesis of antibiotics like oxytetracycline (7), pyrrolobenzodiazepines (8), or thiopeptins (9). Additionally, F420 has attracted considerable interest as a fitness factor of the human pathogen Mycobacterium tuberculosis, being involved in nitrosative stress response (10) or prodrug-activation (11).
A key step during the biosynthesis of F420 is the formation of the deazaflavin fluorophore FO (Fig. 1A), a stable metabolic precursor of F420 originating from tyrosine and an intermediate of riboflavin biosynthesis. This chemically challenging step is catalyzed by the radical SAM enzyme complex CofG/H in archaea or the homologous dual-domain protein FbiC in actinobacteria (12). FO is then further processed by CofC and CofD. This pair of enzymes is responsible for the biosynthesis of the 2-phospho-L-lactate (2-PL) moiety (Fig. 1B). Previous studies have shown that CofC from Methanocaldococcus jannaschii directly activates 2-phospho-L-lactate (2-PL) by guanylation (13) resulting in the formation of the short-lived metabolite lactyl-diphosphoguanosine (LPPG). CofD then forms F420-0 by transfer of the 2-PL moiety from LPPG to FO (14). The enzymes producing 2-PL, however, have remained elusive in all F420 producers so far. Recently, Bashiri et al. proposed a revised biosynthetic pathway by demonstrating that phosphoenolpyruvate (PEP) instead of 2-PL can serve as a substrate of CofC in Mycobacteria (15). The resulting dehydro-F420 (DF420) is reduced to F420 by a flavin-dependent reductase domain present in the FbiB protein (Fig. 1C). In mycobacteria, FbiB is a dual-domain protein consisting of a γ-glutamyl ligase domain (CofE-like) and the C-terminal DF420 reductase domain (16). The γ-glutamyl ligase CofE finally decorates F420-0 with a varying number of (oligo-)γ-glutamate residues (17). F420 is not ubiquitous in prokaryotes, but it is associated with certain phyla (18, 19). First discovered in methanogenic archaea (20, 21), it was extensively studied as a potential drug target of pathogenic mycobacteria (10) or as a cofactor enabling antibiotics biosynthesis in streptomycetes (2). Genome sequencing revealed that some Gram-negative bacteria have acquired F420 genes by horizontal transfer (19, 22). However, virtually nothing is known about the biosynthesis and role of F420 in these organisms. By genome mining, we found a biosynthetic gene cluster (BGC) homologous to those previously implicated in the biosynthesis of F420 in the endofungal bacterium Paraburkholderia rhizoxinica HKI 454 (Fig. 2A and SI Appendix, Table S3). This organism is an intracellular endosymbiont of the phytopathogenic fungus Rhizopus microsporus supplying its host with antimitotic toxins that act as virulence factors during infection of rice plants (23-25). We hypothesized that genes from Gram-negative bacteria related to F420 biosynthesis could facilitate F420 production in E. coli or could reveal novel biosynthetic routes towards this valuable molecule. Therefore, we set out to investigate if the BGC is active and if it can be refactored to produce F420 in E. coli.
Here, we show that P. rhizoxinica produces unexpected F420-derivatives (3PG-F420) both in symbiosis as well as in axenic culture. Heterologous expression and large-scale production in E. coli allowed for elucidation of their chemical structure. By comparative analyses, we discovered related metabolites in a biogas-producing microbial community, thus indicating their broader abundance and relevance. Enzyme assays showed that a switch in substrate specificity of CofC is responsible for the biosynthesis of 3PG-F420 and proved that it can serve as a substitute for F420 in biochemical reactions.
Results
Discovery and structure elucidation of 3PG-F420
To test whether P. rhizoxinica is capable of producing deazaflavins, we investigated axenic cultures of symbiont (P. rhizoxinica) and host (R. microsporus) as well as symbiotic cultures by fluorescence microscopy (Fig. 2B-I and SI Appendix, Fig. S1-4). Indeed, symbiotic P. rhizoxinica and the cytosol of colonized mycelia emitted strong fluorescence characteristic of deazaflavins. Notably, even fluorescence of bacteria present inside of fungal spores was observed. Axenic fungi, however, showed no fluorescence, whereas only low signals were measured from axenic bacteria under the same conditions. To corroborate the results obtained by microscopy, we extracted metabolites from axenic P. rhizoxinica, axenic R. microsporus as well as fungal host containing endosymbionts and analyzed the extracts by LC-MS/MS. To our surprise, only FO could be detected in axenic P. rhizoxinica and in the fungal host containing endosymbionts, but none of the expected F420-n species.
To further test the biosynthetic capacity of the full BGC, we refactored it to obtain a single operon under the control of a T7 promoter for heterologous expression yielding E. coli/pDB045 (Fig. 2J). Examination of transformed bacteria by fluorescence microscopy revealed strong fluorescence as in the endofungal bacteria (Fig. 2D), but LC-MS analyses again yielded only mass traces of FO (mainly found in the culture supernatant), but not of F420-n. Since the impeded production could be attributed to nonfunctional proteins, we analyzed proteins by SDS-PAGE and found that CofD was poorly soluble in E. coli. Replacement of the cofD gene with the corresponding M. jannaschii homolog (14) provided soluble protein (E. coli/pDB060), however, again no trace of F420 could be detected. Therefore, we reexamined the metabolome of E. coli/pDB045 for characteristic MS/MS fragments derived from the FO moiety (m/z 230.06, 346.10, 364.11). Surprisingly, the analysis revealed spectra with a similar fragmentation pattern, yet derived from precursor ions that had a mass shift of 15.995 compared to F420, indicating the presence of an additional oxygen atom (Fig. 3A-B and SI Appendix, Figs. 5-8). Further analyses revealed an (oligo-)γ-glutamate series of the oxygenated compound suggesting these species are congeners. According to MS/MS fragmentation, the additional oxygen was present in the “phospholactyl” moiety of F420-n thus forming a “phosphoglyceryl” moiety. Extensive 1D- and 2D-NMR experiments (Fig. 3C and SI Appendix, Section 2.2) and comparison to classical F420 (20) corroborated that this moiety corresponds to 3-phosphoglycerate (3-PG). Therefore, we named the molecules 3PG- F420. This finding was unexpected because 2-phosphoglycerate is structurally more similar to PEP and 2-PL than 3-PG. Chemical degradation followed by chiral UHPLC-MS finally substantiated that the additional stereocenter of 3PG-F420 is R-configured (Fig. 3D and SI Appendix, Fig. S51). Large-scale cultivation of E. coli/pDB045 also revealed traces of dehydro-F420 (DF420), but the yields were too low for NMR studies. To investigate if 3PG-F420 is produced by wild-type P. rhizoxinica, we reanalyzed LC-MS data for the presence of corresponding mass signals. Indeed, 3PG-F420 species were found in samples containing bacteria (axenic culture, symbiosis), but not in symbiont-free host mycelia (SI Appendix, Figs. S9-14). In extracts of P. rhizoxinica, no traces of F420 and DF420 were detected. The presence of 3PG-F420 was restricted to the cell pellet, whereas FO was abundant in culture supernatants. We thus conclude that the fluorescence observed in bacterial cells of P. rhizoxinica is derived from 3PG-F420 and FO.
Biosynthesis of 3PG-F420
In order to rationalize how the biosynthetic pathway was redirected to form 3PG-F420 instead of F420, we examined key steps of the biosynthesis more closely. We observed that production of 3PG-F420 was not abolished by the exchange of cofD from P. rhizoxinica by cofD from M. jannaschii (plasmid pDB060). Hence, the phospholactyl transferase CofD could not be held accountable for the switch towards 3PG-F420. According to the existing biosynthetic model, the most plausible scenario was that CofC incorporated 3-phospho-D-glycerate (3-PG), an intermediate of glycolysis, instead of 2-PL to form 3PG-F420-0 and, to a minor extent, PEP to form DF420-0. To test this hypothesis, we exchanged cofC and cofD by the corresponding M. jannaschii homologs. The resulting strain E. coli/pDB070 (Fig. 2J and SI Appendix, Fig. S18) produced neither 3PG-F420 nor F420 but traces of DF420. To further investigate the substrate specificity of CofC, we performed an in-vitro assay using CofC and CofD (13). Genes cofC of P. rhizoxinica as well as cofC and cofD from M. jannaschii were cloned, corresponding proteins produced as hexahistidine fusions in E. coli and purified by metal affinity chromatography for in-vitro assays. The physiologically relevant isomers 2-phospho-D-glycerate (2-PG) and 3-phospho-D-glycerate (3-PG), as well as PEP and 2-PL, served as substrates. Reaction products were monitored by LC-MS. Indeed, when CofC from P. rhizoxinica was tested, the mass of 3PG-F420-0 appeared after reaction with D-3-PG eluting at the same retention time as the in-vivo product (Fig. 4A and SI Appendix, Section 2.4). In addition, the formation of DF420-0 and F420-0 was detected, when the enzymes were incubated with PEP and 2-PL, respectively. In contrast, reaction with 2-PG yielded mass signals close to the noise level. Controls lacking CofC did not generate any of these products. In a direct substrate competition assay, 3-PG was found to be the preferred substrate with a relative turnover of ca. 72% (2-PL: 23% PEP: 4%). This finding is in agreement with the structure of 3PG-F420 and the occurrence of DF420 as a minor biosynthetic product in bacteria. Note that CofC from M. jannaschii displayed a strong turnover of 2-PL (96.5%), weak turnover of PEP (3.5%) and no turnover of 3-PG. This finding supports the notion that the CofC of P. rhizoxinica has undergone a substrate specificity switch during evolution.
To address the question if the γ-glutamyl ligase CofE represents another specificity layer that directs the biosynthesis of F420 towards 3PG-F420, we individually co-expressed cofE genes from P. rhizoxinica, M. jannaschii and Mycolicibacterium smegmatis (fbiB) together with a minimal gene cluster consisting of fbiC, cofC, and cofD in a two-plasmid system. Extraction of metabolites and LC-MS/MS revealed that all three CofE homologs elongated 3PG-F420-0 to oligo-glutamate chain lengths up to n=6 (SI Appendix, Figs. S63-65). Thus, we conclude that CofE does not act as an additional specificity filter during chain elongation of 3PG-F420.
Cofactor role of 3PG-F420
The successful isolation of 3PG-F420 and reconstitution of its biosynthesis in E. coli motivated us to address the question of whether 3PG-F420 could substitute F420 in biocatalysis. To this end, we cloned a gene encoding Fno (F420:NADPH oxidoreductase), an enzyme that serves as a regeneration system for F420H2 using NADPH/H+ as an electron donor (26). We first examined if Fno can accept 3PG-F420 as a substrate. Indeed, we observed an efficient reduction of 3PG-F420 by recombinant Fno as mirrored by a rapid decrease of characteristic UV absorption. An examination of kinetic parameters (Fig. 5A) revealed that the apparent KM of Fno for F420 was 3.6 ± 0.7 µM. This value is similar to the reported KM of 10 µM (27). Under identical assay conditions, the KM for 3PG-F420 was only slightly higher (5.1 ± 1.0 µM). The vmax values were in a similar range as well (F420: 1.3 ± 0.2 µmol min-1; 3PG-F420: 0.88 ± 0.07 µmol min-1) pointing towards only a minor reduction of maximal turnover. Encouraged by the finding that 3PG-F420 can substitute F420, we aimed at an in-vivo application of the cofactor for malachite green reduction as a proof of principle. To this end, we combined the fno gene with a minimal BGC producing 3PG-F420-0 (Fig. 5C and SI Appendix, Fig. S62) on a single vector (pDB071). Additionally, the F420-dependent malachite green reductase gene MSMEG_5998 (28) from M. smegmatis was cloned and expressed from a compatible vector backbone (pDB061). Finally, co-expression of all components in E. coli yielded a strain (pDB061/pDB071) that was able to decolorize malachite green significantly faster than control strains expressing the reductase or the cofactor alone (Fig. 5B). We, therefore, conclude that 3PG-F420 can substitute F420 as a redox cofactor in this case.
Occurrence of 3PG-F420 in nature
Finally, we set out to test if 3PG-F420 is restricted to fungal endosymbionts or if it might be more widespread in the environment. We examined M. jannaschii and M. smegmatis for the presence of any F420 congeners. Only F420-n was found from extracts of M. jannaschii, while F420-n and DF420-n were detected in extracts of M. smegmatis. (SI Appendix, Figs. S15-16). None of the 3PG-F420-derivatives were found in the reference organisms. Next, we analyzed (two independent) sludge samples from a local biogas production plant. To our surprise, extraction of the microbial community present in the biogas-producing sludge followed by LC-MS/MS eluted, besides classical F420, a compound with identical retention time, exact mass and MS/MS fragmentation pattern as 3PG-F420 (SI Appendix, Fig. S17). Therefore, we conclude that microbes present in the biogas-producing microbial community produce 3PG-F420-derivatives or highly similar congeners.
Discussion
Unexpected biological sources of F420-derivatives
In the course of this study, we discovered a BGC consisting of genes related to F420-biosynthesis in the Gram-negative bacterium P. rhizoxinica. We could show that the BGC is highly active and mediates production of 3PG-F420, a derivative of coenzyme F420, thus revealing an untapped terrain of deazaflavin-based biochemistry in Gram-negative bacteria. So far, there is only preliminary evidence for the occurrence of deazaflavin cofactors in a few Gram-negative bacteria, e.g., Oligotropha carboxidivorans and Paracoccus dentrificans (19) as well as in the uncultured, but biosynthetically highly prolific ‘Candidatus Entotheonella factor’ (22). The exact structure and function of F420 in most Gram-negative bacteria that harbor corresponding biosynthetic genes, however, is unknown. The fact that P. rhizoxinica produced 3PG-F420 under symbiotic cultivation conditions allows for the conclusion that it provides a fitness benefit in its natural habitat. Notably, none of the well-characterized F420 dependent enzyme families (4, 18) are encoded in the P. rhizoxinica genome according to BLAST and conserved domains searches, not even any of the widespread regeneration systems like Fno or F420 dependent glucose-6-phosphate dehydrogenase. Therefore, future investigations of 3PG-F420 producing organisms are likely to reveal novel enzymes, regeneration systems, and cellular pathways depending on this cofactor. Intriguingly, substantial amounts of phosphoglycerate-derived deazaflavins were also extracted from a microbial community of a biogas-production plant. Since fluorescence and UV-based detection does usually not resolve classical F420 and 3PG-F420 these derivatives might have been misidentified as F420 in the past. Since neither P. rhizoxinica nor its host R. microsporus are able to grow under anaerobic and thermophilic (temperatures >42 °C) conditions, they can be excluded as the source of these cofactors. The high complexity and variability of biogas-producing microbiomes (29) do not allow for an educated guess of the producer, although methanogens would be reasonable candidates. In any case, phosphoglycerate-derived deazaflavins are more abundant in nature than anticipated.
Structural diversification of deazaflavin cofactors
The structure and occurrence of 3PG-F420 has not been reported before. To date, the only known derivatives of F420-n are factor F390A and F390-G, 8-OH-AMP and 8-OH-GMP esters of F420, respectively (30). In methanogens, they are formed reversibly, e.g., during oxygen exposure, acting as a reporter compound for hydrogen starvation (31). In contrast, the modifications seen in 3PG-F420 are not temporary. Rather, 3PG-F420 seems to replace F420 as a natural deazaflavin-cofactor in P. rhizoxinica. At least in this organism, it does not coexist with classical F420. This situation is reminiscent of mycothiol, a specialized thiol cofactor that replaced glutathione in actinobacteria (32). Notably, the redox-active deazaflavin moiety of 3PG-F420 remains unchanged and, therefore, redox activity of 3PG-F420 was retained. The phospholactate and glutamate-derived tail moiety of F420 appears to have two functions. Firstly, the tail avoids loss of the cofactor, since FO, in contrast to F420, is crossing the bacterial cell membranes and is found mainly in culture supernatants. This function is also fulfilled by the altered tail moiety of 3PG-F420. Secondly, the tail of F420 provides additional contacts for non-covalent interactions (33). Our first results show that the apparent KM of Fno for 3PG-F420 is in a similar low micromolar range as the KM for F420, thus indicating comparable binding affinity. Importantly, 3PG-F420 can be used as a cofactor in E. coli to reduce malachite green. A similar result was just recently obtained using classical F420 and a different malachite green reductase (15).
Rerouting of F420 biosynthetic pathways
In addition to the question if 3PG-F420 can act as a cofactor, we clarified how a F420-derivative can be produced in a metabolic background lacking DF420 reductase and 2-PL. Obviously, a substrate specificity switch in CofC directed the biosynthesis towards 3-PG. Recently, Bashiri et al. claimed that PEP is the substrate of CofC in prokaryotes (15). Our results partially confirm their hypothesis, since we observed the turnover of PEP as a potential substrate of both CofC homologs we tested. However, in contrast to Bashiri et al., 2-PL was the best substrate of M. jannaschii CofC in our assay. No DF420 was measurable in cell extracts of M. jannaschii, while it was readily detected in M. smegmatis cells. Since Graupner and White detected significant amounts of 2-PL in methanogenic archaea and observed the conversion of lactate into 2-PL by isotope labeling (34), we conclude that 2-PL might still be the relevant substrate in archaea. From an evolutionary perspective, our results suggest that multiple metabolic re-wiring events occurred in the evolution of F420 biosynthesis. While actinobacteria evolved the DF420 reductase (C-terminal domain of FbiB), archaea accomplished to produce the (unusual) metabolite 2-PL. Other organisms, as exemplified by P. rhizoxinca, rerouted the biosynthesis to the ubiquitous metabolite 3- PG. Notably, the altered pathway offered another route to heterologous production of F420-derivatives in E. coli. The production of classical F420 in E. coli has just recently been achieved in moderate yields using Mycobacterium genes including the DF420 reductase domain (15).
Prospects for biotechnology
In recent years, there has been increasing interest in F420-dependent enzymes for biocatalytic applications (5, 6, 35, 36). Therefore, the convenient production in E. coli along with the anticipation to discover novel enzymes and regeneration systems specialized for 3PG-F420 opens intriguing prospects for biocatalysis and metabolic engineering. Here, as a proof of concept, 3PG-F420 has been shown to be easily reduced by Fno and to substitute F420 for malachite green reduction in vivo. Besides future in-vitro applications, for instance enantioselective biotransformations, metabolically engineered microbes can be envisioned to use 3PG-F420 to promote challenging reactions of industrial appeal, to help in bioremediation or to produce high-value compounds like antibiotics in a sustainable manner.
Materials and methods
Materials and methods are summarized in SI Appendix, Section 1.
Acknowledgments
We thank Prof. Dr. Dina Grohmann and Dr. Ghader Bashiri for kindly providing M. jannaschii and M. smegmatis / fbiABC, respectively. We thank Biogas Jena GmbH and Co. KG for the kind donation of biogas plant samples. We thank Heike Heinecke for conducting NMR experiments. G.L. thanks the Deutsche Forschungsgemeinschaft (DFG Grant LA 4424/1-1) and the Carl Zeiss Foundation for funding. Financial support by the DFG (CRC 1127 ChemBioSys) to C.H. and C.B., and, BE-4799/2-1 to C.B., and Leibniz Award to C.H., by the ERC (MSCA-IF-EF-RI Project 794343, to I.R.) and the JSMC to Z.U. is gratefully acknowledged.