Cellular and structural basis of synthesis of the unique intermediate dehydro-F420-0 in mycobacteria

F420 is a low-potential redox cofactor used by diverse bacteria and archaea. In mycobacteria, this cofactor has multiple roles, including adaptation to redox stress, cell wall biosynthesis, and activation of the clinical antitubercular prodrugs pretomanid and delamanid. A recent biochemical study proposed a revised biosynthesis pathway for F420 in mycobacteria; it was suggested that phosphoenolpyruvate served as a metabolic precursor for this pathway, rather than 2-phospholactate as long proposed, but these findings were subsequently challenged. In this work, we combined metabolomic, genetic, and structural analyses to resolve these discrepancies and determine the basis of F420 biosynthesis in mycobacterial cells. We show that, in whole cells of Mycobacterium smegmatis, phosphoenolpyruvate rather than 2-phospholactate stimulates F420 biosynthesis. Analysis of F420 biosynthesis intermediates present in M. smegmatis cells harboring genetic deletions at each step of the biosynthetic pathway confirmed that phosphoenolpyruvate is then used to produce the novel precursor compound dehydro-F420-0. To determine the structural basis of dehydro-F420-0 production, we solved high-resolution crystal structures of the enzyme responsible (FbiA) in apo, substrate, and product bound forms. These data show the essential role of a single divalent cation in coordinating the catalytic pre-complex of this enzyme and demonstrate that dehydro-F420-0 synthesis occurs through a direct substrate transfer mechanism. Together, these findings resolve the biosynthetic pathway of F420 in mycobacteria and have significant implications for understanding the emergence of antitubercular prodrug resistance.


56
Factor 420 (F420) is a deazaflavin cofactor that mediates diverse redox reactions in bacteria and 57 archaea (1). Chemically, F420 consists of a redox-active deazaflavin headgroup (derived from 58 the chromophore Fo) that is conjugated to a variable-length polyglutamate tail via a 59 phosphoester linkage (2). While the Fo headgroup of F420 superficially resembles flavins (e.g. 60 FAD, FMN), three chemical substitutions in the isoalloxazine ring give it distinct chemical 61 properties more reminiscent of nicotinamides (e.g. NADH, NADPH) (1). These include a low 62 standard potential (-350 mV) and obligate two-electron (hydride) transfer chemistry (3,4). The 63 electrochemical properties of F420 make it ideal to reduce a wide range of otherwise 64 recalcitrant organic compounds (5-7). Diverse prokaryotes are known to synthesize F420, but 65 the compound is best characterised for its roles in methanogenesis in archaea, antibiotic 66 biosynthesis in streptomycetes, and metabolic adaptation of mycobacteria (1,(8)(9)(10)(11). In 67 mycobacteria, F420 is involved in a plethora of processes: central carbon metabolism, cell wall 68 synthesis, recovery from dormancy, resistance to oxidative stress, and inactivation of certain 69 bactericidal agents (7,(12)(13)(14). In the human pathogen Mycobacterium tuberculosis, F420 is also 70 critical for the reductive activation of the newly approved clinical antitubercular prodrugs 71 pretomanid and delamanid (15)(16)(17). 72 Following the elucidation of the chemical structure of F420 in the 1970s, the F420 biosynthesis 73 pathway in archaea was determined through a combination of in situ biochemistry and 74 recombinant protein analysis (1, 2). Described briefly, the deazaflavin fluorophore Fo is 75 synthesized through condensation of 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione and 76 L-tyrosine by the SAM-radical enzymes CofG and CofH (18). The putative enzyme CofB 77 synthesizes 2-phospholactate (2PL), which links Fo to the glutamate tail of mature F420 (19). 78 Subsequently, the nucleotide transferase CofC condenses 2PL with GTP to form the reactive 79 intermediate L-lactyl-2-diphospho-5ʹ-guanosine (LPPG) (20). The phosphotransferase CofD 80 then transfers 2PL from LPPG to Fo, leading to the formation of F420-0 (i.e. F420 with no 81 glutamate tail) (21). Finally, the GTP-dependent glutamate ligase CofE adds a variable-length 82 γ-linked glutamate tail to produce mature F420 (22,23). With the exception of the putative 83 lactate kinase CofB, the enzymes responsible for F420 biosynthesis in archaea have been 84 identified and characterized to varying extents (1). Crystal structures have been obtained for 85 CofC, CofD and CofE from methanogenic archaea, providing some insight into how these 86 enzymes function, but questions surrounding their catalytic mechanisms remain unresolved 87 (23-25). For example, the crystal structure of CofD from Methanosarcina mazei was solved in 88 the presence of Fo and GDP; however, no divalent cation(s) required for catalysis were present 89 in the structure and the ribosyl tail group of Fo, which receives the 2PL moiety from LPPG was 90 disordered, precluding an understanding of the catalytic mechanism of this step in F420 91 biosynthesis (21,25). 92 It was assumed that the biosynthesis pathway for archaeal F420 was generic to all F420 producing 93 organisms (1). However, recent studies have shown that the structure and biosynthesis of F420 94 varies between producing organisms (24,26). F420 produced by the proteobacterial fungal 95 symbiont Paraburkholderia rhizoxinica was found to incorporate 3-phospho-D-glycerate (3PG) 96 in the place of 2PL, producing a chemically distinct F420 (26). In parallel, analysis of purified F420 97 biosynthesis enzymes from mycobacteria indicated that the central glycolytic and 98 gluconeogenic intermediate phosphoenolpyruvate (PEP), rather than 2PL, is a precursor for 99 F420 biosynthesis (24). In contrast to P. rhizoxinica, in mycobacteria, mature F420 is chemically 100 analogous to that produced by archaea (27) the presence and absence of its substrate and product compounds. These data resolve long-130 standing questions about the catalytic mechanism of FbiA and CofD in F420 biosynthesis.

131
Moreover, they provide a target for therapeutic intervention through the inhibition of F420 132 biosynthesis, as well as insight into potential mechanisms for the emergence of delamanid and 133 pretomanid drug resistance through mutations in FbiA.

136
Phosphoenolpyruvate is the substrate for the biosynthesis of F420 in mycobacterial cells

137
To determine whether PEP or 2PL is the substrate for F420 biosynthesis in mycobacteria ( Figure   138 1A), we spiked clarified cell lysates from M. smegmatis with GTP and either PEP or 2PL, and 139 monitored the synthesis of new F420 species through HPLC coupled with fluorescence detection.

140
In cell lysates spiked with PEP, a species corresponding to F420-0 in the F420 standard was 141 present, which was absent from both the untreated and 2PL spiked lysates ( Figure 1B). The 142 formation of this F420-0-like species in PEP spiked lysates corresponded to a decrease in Fo 143 levels, suggesting that synthesis of DH-F420-0 from PEP is occurring ( Figure 1B). These data 144 strongly suggest that PEP, not 2PL, is the precursor for F420 biosynthesis in M. smegmatis.
While the lysate spiking experiment establishes that PEP is specifically utilised for F420 synthesis 146 in M. smegmatis, the fluorescent detection method utilised does not chemically differentiate 147 between F420-0 or DH-F420-0. As PEP is utilised, it would be expected that DH-F420-0 is produced.

148
However, DH-F420-0 may be rapidly reduced to F420-0 rather than accumulating in the cell. To  Figure S1).

158
The proposed biosynthetic intermediate DH-F420-0 was detected only in cell lysates of the ΔfbiB 159 strain ( Figure 2D, Figure S1). No F420-0 was detected in wildtype or mutant strains.

160
The presence of DH-F420-0 (and absence of detectable F420-0) in whole cells demonstrates that 161 it is the central physiological intermediate in mycobacterial F420 biosynthesis. This also lends 162 support to the biochemical and cellular assays indicating that PEP, not 2PL, is the substrate for 163 this pathway in mycobacteria. Furthermore, in addition to its role as the F420 glutamyl-ligase, 164 structural and biochemical analysis suggests that FbiB is responsible for the reduction of DH-165 F420-0 (24, 28). The detection of DH-F420-0 only in the ΔfbiB strain demonstrates that this 166 intermediate is rapidly turned over in the cell and supports the hypothesis that FbiB and not 167 another enzyme performs this step in mycobacterial F420 biosynthesis.

170
In order to determine the catalytic mechanism for the synthesis of the novel intermediate DH-  Figure S2A). The nature of this substrate was investigated using fluorescence 174 spectroscopy, with purified FbiA found to have a broad absorbance peak at 400 nm and a corresponding emission peak at 470 nm ( Figure S2B), which is consistent with the presence of 176 a deazaflavin with a protonated 8-OH group (16). We then utilized LC/MS to identify the 177 deazaflavin species associated with FbiA and found the major species was its product DH-F420-178 0 ( Figure S2C). In addition, significant quantities of mature F420 species were also associated 179 with FbiA, suggesting that it also binds to mature F420 present in the cytoplasm ( Figure S2C). In order to resolve the catalytic mechanism of DH-F420-0 synthesis, purified FbiA was 183 crystallized and its structure was determined at 2.3 Å by X-ray crystallography (Table S1). FbiA 184 crystallized as a dimer mediated by the interaction of three α-helices and a β-sheet ( Figure 3A).

185
This dimer is predicted to be stable by the protein-interaction prediction program PISA (29) 186 and the molecular weight of FbiA determined by SEC-MALS shows it forms a dimer in solution 187 (Table S2, Figure S2D). The structure of CofD from M. mazei, a homologous enzyme that 188 instead utilizes LPPG derived from 2PL as its substrate, also crystallised as a dimer with an   In the GDP and Fo bound structure, the coordination of Ca 2+ is analogous to the GDP-only 232 structure. However, the glycerol molecule and one of the H2O molecules observed in the GDP 233 only structure are displaced by the ribosyl chain of Fo, resulting in a coordination number of 234 six with octahedral geometry ( Figure 4B, Figure S4B). The terminal hydroxyl group of Fo is 235 significantly closer to the Ca 2+ ion (2.6 Å) and to the β-phosphate of GDP (2.8 Å from O, and 3.0 236 Å from P) than the coordinating hydroxyl of glycerol, which is not within bonding distance of 237 GDP. These bond distances between the hydroxyl of Fo and GDP, as well as the central 238 orientation of the hydroxyl of Fo towards the β-phosphate of GDP, place it in an ideal position 239 to act as the acceptor substrate for the transfer of PEP catalysed by FbiA ( Figure 4B). In the Fo 240 and DH-F420-0 bound structures, no density corresponding to a Ca 2+ ion was observed ( Figure   241 4C, Figure S4C). This suggests that binding of FbiA to its catalytic metal ion is contingent on 242 complex formation with enolpyruvyl-diphospho-5ʹ-guanosine (EPPG) (Figure 2A), which is 243 substituted for GDP in our structures due to the instability of the F420 pathway intermediate  Integrating these findings with other recent literature, it is now clear that the substrate for the 264 initial stage of F420 tail biosynthesis differs between F420 producing organisms (19,24,26 Phylogenetic analysis of FbiD/CofC and FbiA/CofD suggests that these proteins were 283 horizontally transferred from bacteria and archaea (8). Based on this analysis, it is curious that 284 mycobacteria reduce DH-F420-0 produced via PEP to F420, rendering it chemically identical to 285 that produced with 2PL. The redox properties of the deazaflavin group of DH-F420 and F420 are 286 identical and chemically the molecules are very similar, posing the question: why is reduction 287 of DH-F420-0 is required? A plausible explanation is that actinobacteria originally utilized 2PL 288 for F420 synthesis, with a switch to PEP occurring at a later stage in evolution. As a result, the 289 F420-dependent enzymes present in mycobacteria evolved to recognize the non-planar 2PL 290 moiety of F420, requiring reduction of DH-F420 to maintain compatibility after the substrate 291 switch. Previous structural and biochemical analysis suggests the C-terminal domain of 292 mycobacterial FbiB is responsible for the reduction of DH-F420-0 (24, 28). This domain is present 293 in all mycobacterial species, but is absent from FbiB/CofE in most other F420 producing 294 organisms, including M. mazei and P. rhizoxinica, that produce F420 through pathways that do 295 not require this reductive step (8,26). This conclusion is supported by our cellular analysis of 296 F420 biosynthesis in M. smegmatis, which shows that DH-F420-0 accumulates in the ΔfbiB strain 297 (Figure 2A, D).

299
The structural analysis of FbiA that we present in this work provides unprecedented insight 300 into the catalytic mechanism for the novel phosphotransferase reaction employed at this step  The resolution of the F420 biosynthesis pathway also has implications for tuberculosis 319 treatment. It has been proposed that F420 biosynthesis represents a promising target for the  (Table S3). Individual deletion mutants of fbiA, fbiB and fbiD were confirmed by 369 Southern blot following digestion with PvuII and then genome sequencing. grown in LB broth + 0.05% Tween 80 at 37 °C with shaking to an OD600 of ~2.0 before the 386 expression of the fbi genes was induced with 0.2 % acetamide. Cells were grown for an 387 additional 72 hours before harvesting by centrifugation at 10,000 × g for 20 minutes. Cells were 388 resuspended in 50 mM Tris (pH 7.5) at a ratio of 10 ml of buffer per 1 gram of wet cells and 389 lysed by autoclaving. The autoclaved cell suspension was clarified by centrifugation at 20,000 390 × g for 20 minutes. The clarified supernatant was applied to a High Q Anion Exchange Column 391 (Biorad) equilibrated in 50 mM Tris (pH 7.5). Bound species were eluted with a gradient of 0-392 100 % of 50 mM Tris, 1 M NaCl (pH 7.5). Fractions containing F420 were identified via visible 393 spectroscopy based on their distinctive absorbance peak at 420 nm. Fractions containing F420 394 were pooled and applied to a C18-silica column equilibrated in dH2O. F420 was eluted with 20 % 395 methanol in H2O, vacuum evaporated and stored at -20 °C for further analysis.

396
DH-F420-0 was extracted from purified FbiA expressed in M. smegmatis as described below.

397
Purified concentrated FbiA (~20 mg ml -1 ; prior to cleavage of the poly-histidine tag) was 398 denatured in a buffer containing 50 mM Tris and 8 M Urea (pH 7.0). This solution containing 399 denatured FbiA and free DH-F420-0 was applied to a nickel-agarose column, with denatured 400 FbiA binding to the column due to its hexahis-tag and DH-F420-0 eluting in the flowthrough. The  The DNA coding sequence corresponding to FbiA from M. smegmatis was amplified by PCR 410 using primers outlined in Table S3, resulting in a DNA fragment with 5' NcoI and 3' HindIII sites 411 respectively. This fragment was cloned into pMyNT by restriction enzyme cloning using the 412 aforementioned sites, yielding pMyNTFbiAMS, which expresses FbiA with a TEV cleavable N-

417
Colonies from this transformation were used to inoculate 50 ml of LBT media +50 μg ml -1 418 hygromycin B, which was grown with shaking at 37 °C until stationary phase (2-3 days). This 419 starter culture was used to inoculate 5 liters of Terrific Broth + 0.05 % Tween 80 (TBT) giving a 420 1:100 dilution of the starter culture. Cells were grown with shaking at 37 °C for 24 hours until 421 approximately mid-log phase and protein production was induced through the addition of 0.2% 422 acetamide. Cells were grown with shaking at 37 °C for an additional 72 hours before they were 423 harvested via centrifugation at 5,000 × g for 20 minutes. Harvested cells were either lysed 424 immediately or stored frozen at -20 °C.    min-5%, 24 min-80%, 32 min-80%) on a Dionex RSLC3000 UHPLC (Thermo). The flow rate was 481 maintained at 300 μl min -1 . Samples were kept at 4 °C in the autosampler and 10 μl was injected 482 for analysis. The mass spectrometric acquisition was performed at 35,000 resolution on a Q-483 Exactive Orbitrap MS (Thermo) operating in rapid switching positive (4 kV) and negative (−3.5 484 kV) mode electrospray ionization (capillary temperature 300 °C; sheath gas 50; Aux gas 20; 485 sweep gas 2; probe temp 120 °C). The resulting LC-MS data were processed by integrating the 486 area below the extracted ion chromatographic peaks using TraceFinder 4.1 (Thermo Scientific).

487
All species were detected in negative mode as the singly deprotonated anion (Fo and DH-F420-488 0) or in the case of the F420-n species the double deprotonated dianion.  To detect F420 synthesis in spiked M. smegmatis lysates, 500 ml cultures were grown in LBT for 503 3 days at 37 °C with shaking. Cells were harvested by centrifugation at 8,000 × g for 20 min, 504 4 °C. The pellet was resuspended in 50 ml lysis buffer (50 mM MOPS, 1 mM 505 phenylmethylsulfonyl fluoride (PMSF), 1 mM DTT, 5 mM MgCl 2 , 2.5 mg.ml -1 lysozyme, 2.5 mg 506 Deoxyribonuclease I). An M-110P Microfluidizer (Fluigent) pressure-lysis maintained at 4 °C 507 was used to lyse the cells. The lysate was centrifuged at 10,000 × g at 4 °C for 20 min. 1 ml 508 aliquots of lysate were spiked with either 1 mM phosphate buffer (pH 7.0) plus GTP and 2PL, 509 or GTP and PEP. These spiked samples, along with a 'no spike' control, were incubated at 4 510 hours at 37 °C. To terminate the reaction, the aliquots were heated at 95 °C for 20 min, then 511 centrifuged at 16,000 × g for 10 min. The supernatants were filtered through a 0.22 μm PVDF 512 filter and moved to analytical vials. 513 F420 biosynthetic intermediates present in the filtered M. smegmatis cell lysates were analysed 514 by separation and detection using an Agilent 1200 series HPLC system equipped with a 515 fluorescence detector and a Poroshell 120 EC-C18 2.1 x 50 mm 2.7 μm column. The system 516 was run at a flow rate of 0.3 ml min -1 and the samples were excited at 420 nm and emission 517 was detected at 480 nm. A gradient of two buffers were used: Buffer A, containing 20 mM 518 ammonium phosphate, 10 mM tetrabutylammonium phosphate, pH 7.0. Buffer B, 100% 519 acetonitrile. A gradient was run from 25% to 40% buffer B as follows: 0-1 min 25%, 1-10 min 520 25%-35%, 10-13 min 35%, 13-16 min 35-40%, 16-19 min 40%-25%.

521
This research was undertaken in part using the MX2 beamline at the Australian Synchrotron, 524 part of ANSTO, and made use of the Australian Cancer Research Foundation (ACRF) detector. 525 We would like to thank the Monash Crystallisation Facility for their assistance with sample 526 characterization, crystallographic screening, and optimization. pMyNT was a gift from Dr.