Studies on the Glutathione-Dependent Formaldehyde-Activating Enzyme from Paracoccus denitrificans

Formaldehyde is a toxin and carcinogen that is both an environmental pollutant and an endogenous metabolite. Formaldehyde metabolism, which is probably essential for all aerobic cells, likely proceeds via multiple mechanisms, including via a glutathione-dependent pathway that is widely conserved in bacteria, plants and animals. However, it is unclear whether the first step in the glutathione-dependent pathway (i.e. formation of S-hydroxymethylglutathione (HMG)) is enzyme-catalysed. We report studies on glutathione-dependent formaldehyde-activating enzyme (GFA) from Paracoccus denitrificans, which has been proposed to catalyse HMG formation from glutathione and formaldehyde on the basis of studies using NMR exchange spectroscopy (EXSY). Although we were able to replicate the EXSY results, time course experiments unexpectedly imply that GFA does not catalyse HMG formation under standard conditions. However, GFA was observed to bind glutathione using NMR and mass spectrometry. Overall, the results reveal that GFA binds glutathione but does not directly catalyse HMG formation under standard conditions. Thus, it is possible that GFA acts as a glutathione carrier that acts to co-localise glutathione and formaldehyde in a cellular context.


Introduction
Formaldehyde (HCHO) is produced in cells during redox processes including during the enzymatic demethylation of methylated nucleic acids and proteins [1][2][3]. Above threshold levels, HCHO is toxic to cells and animals including humans; acute exposure either through inhalation or ingestion may result in convulsions and renal failure, whereas chronic exposure is linked to increased risk of cancers including nasopharyngeal cancer and leukaemia [4,5].
A predominant pathway assigned for HCHO metabolism in eukaryotic cells involves its reaction with the tripeptide glutathione (GSH) [6,7]. GSH reacts with HCHO via its nucleophilic thiol group to form the intermediate S-hydroxymethylglutathione (HMG). HMG may Formaldehyde (HCHO) reacts with glutathione (GSH) via its nucleophilic thiol group to form S-hydroxymethylglutathione (HMG), which is a substrate of was observed to bind GSH by NMR and mass spectrometric binding experiments. Overall, although our findings indicate that GSH can bind to GFA, they imply that recombinant GFA does not catalyse HMG formation from GSH and HCHO in any detectable way under standard aqueous conditions.

Materials
Deuterated Tris buffer was from Cortecnet (France). γ-Glutamyl-serinyl-glycine and γ-glutamyl-D-cysteinyl-glycine were from Peptide Synthetics (U. K.). All other reagents were from The Sigma-Aldrich Chemical Company. HCHO solution was prepared by heating paraformaldehyde powder in H 2 O in a glass vial using a heat gun at >100°C until the powder had dissolved. Pre-prepared formalin solution was not used so as to avoid contamination from methanol. [ 13 C]-labelled HCHO ([ 13 C]-HCHO) was purchased as 20 wt % in H 2 O from Sigma-Aldrich.

Protein Production
Recombinant wild-type GFA was expressed from its gene cloned in a modified pET16b plasmid (to encode for recombinant protein with a Tobacco Etch Virus nuclear inclusion-a endopeptidase (TEV protease) cleavage site between the His-Tag and the insert, S. Becker, Göttingen) in E. coli BL21 (DE3) cells grown on 2 × Tryptone Yeast (2TY) Extract medium [20][21][22]. The protein was purified via Ni-affinity before removal of the N-terminal His-Tag using TEV protease [23] and size-exclusion chromatography. Some experiments were carried out using His-Tagged GFA (unless stated, GFA (without His-tag) was used). Aliquots were stored in 50 mM Tris buffer in H 2 O pH 7.5 (His-Tagged GFA was stored in 20 mM HEPES buffer in H 2 O pH 7.5). The GFA C54A variant was generated using the QuickChange sitedirected mutagenesis kit (Stratagene) and the mutation was confirmed by DNA sequencing. This plasmid was then transformed into E. coli BL21 (DE3) and expressed and purified as described above (without His-Tag, stored in 50 mM Tris buffer in H 2 O pH 7.5).

NMR Analyses
NMR experiments were carried out using a Bruker Avance III 700 MHz spectrometer equipped with an inverse TCI cryoprobe optimised for 1 H observation and installed with Topspin 3 software. Each sample was prepared in a microcentrifuge tube (either 75 μL or 160 μL final volume) before being transferred to either a 2 mm or a 3 mm Bruker MATCH NMR tube (Hilgenberg), centrifuged for a few seconds using a hand centrifuge, and transferred to the spectrometer for NMR analysis. Where applicable, mixtures were left in microcentrifuge tubes to reach equilibrium before transfer to NMR tubes (15-30 min). All experiments were carried out at 298 K unless otherwise stated.
For 1 H NMR experiments, the solvent resonance was removed by excitation sculpting using a 2 ms 180°sinc pulse [24]. Time-course experiments were conducted using automated glutathione-dependent alcohol dehydrogenase (ADH, ADH5 in humans). The product, S-formylglutathione, is then further metabolised by Sformylglutathione hydrolase to give formate and GSH. The reaction of HCHO and GSH, i.e. the first step in GSH-dependent metabolism, occurs spontaneously in aqueous solution; however, the reaction might also be catalysed by GFA (and homologues in other organisms, e.g. CENPV in humans [13]). There is also evidence, at least in vitro, that GSH can react with HCHO to form cyclised adducts [10][11][12]. (B) Views of X-ray crystal structures of GFA from Paracoccus denitrificans (PDB IDs: 1X6M and 1XA8 [14]). The GFA domain contains two zinc binding sites; one zinc ion is coordinated by four cysteinyl thiols (C31, C33, C99 and C102) in a tetrahedral geometry, whereas the other zinc ion is coordinated by three cysteinyl thiols (C52, C54 and C57) in a trigonal planar geometry. Crystallographic studies have proposed that GSH binding induces translocation of the second zinc ion (circles) [14].  routines; 20 analyses were performed on each sample, each accumulating 16 transients corresponding to 89 s (or 80 s for the experiments with lysate) of total measurement time. EXSY analyses were performed using 1D NOESY pulse sequences using selective refocusing with Gaussian pulses [13,25,26]. 2D experiments were run accumulating 16 transients with a mixing time (τ m ) of 400 ms. 1D experiments were performed accumulating 64 transients with τ m of 32-400 ms (for comparative studies, experiments were run with a τ m of 80 ms). For 1D EXSY analyses, the resonance at δ H 2.95 ppm (corresponding to one β-cysteinyl proton of HMG) was selectively irradiated.
Saturation transfer difference (STD) analyses were performed using the Bruker pulse sequence, in which selective saturation of the protein resonance was achieved by a train of Gaussian-shaped pulses (50 ms) [27]. A 2 ms sinc pulse was employed for excitation sculpting water suppression. The on-resonance irradiation was selected at 593 Hz and the off-resonance at 35000 Hz. The total saturation time and relaxation delays were 10 s and 18 s respectively.
Water-ligand observed via gradient spectroscopy (waterLOGSY) analyses were performed using the pulse sequence described by Dalvit et al [28]. Solvent water excitation was achieved using a 16 ms selective rectangular shape pulse set at the H 2 O frequency. The relaxation delay was 2 s and the τ m was 1 s. A 2 ms sinc pulse was used for excitation sculpting water suppression. Each experiment accumulated 64-256 transients at 280 K.
Unless otherwise stated, all errors are reported as standard deviations of the mean (n = 3).

Cell Lysate Experiments
For work with cell lysates, untransformed E. coli BL21 (DE3) cells were grown in 2TY medium at 37°C for 16 hours under aerobic conditions (to stationary phase) before centrifugation and re-suspension in 50 mM Tris buffer pH 7.5 (0.5 mg of cells per mL). The cells were lysed by sonication (20 × 2 s) and the supernatant separated from insoluble particulates by centrifugation and then filtration using a 0.45 μm syringe filter (Sartorius Stedim Biotech). NMR experiments (160 μL total volume) were then conducted on the supernatant after addition of [ 13 C]-HCHO and D 2 O (16 μL).

Results
In order to investigate the effect(s) of GFA on the reactions of GSH and HCHO in aqueous solution, NMR studies were carried out using recombinant GFA purified from E. coli (>95% pure by SDS-PAGE, S2-S4 Figs) [13,14]. Our initial experiments focussed on reproducing the previously observed increases in EXSY-correlation intensities between GSH and HMG in the presence of GFA [13]. EXSY, an NMR-based method, utilises the same pulse sequences applied to detect nuclear Overhauser effect (nOe) correlations between spatially close nuclei; EXSYcorrelations (cross-peaks) are classically observed between two species, with distinct chemical shifts, that exist in a slow exchange equilibrium (e.g. chemical or conformational). The initial growth ratios of the EXSY-correlations are proportional to the rate of exchange between the species, and therefore, EXSY may be used as a quantitative measure of equilibrium dynamics, e.g. HCHO/HMG inter-conversion [13]. Mixtures of GSH (15 mM) and HCHO (15 mM) were prepared in deuterated Tris buffer in 90% H 2 O / 10% D 2 O pH Ã 7.5 and were left for 30 minutes to reach equilibrium (note: evidence of a previously characterised Tris-HCHO adduct [29] was observed but was at low levels relative to free HCHO) before addition of either GFA (20 μM final concentration) or buffer (as a control) prior to 2D EXSY analysis. As reported [13], reproducible EXSY correlations between resonances corresponding to the β-cysteinyl protons of GSH and HMG increased in intensity in the sample containing GFA (with both poly-histidine (His)-Tagged and non-tagged GFA batches), relative to the control sample in the absence of GFA (Fig 2A). Similar findings were also observed in 1D EXSY experiments, where the HMG β-cysteinyl resonance at δ H 2.95 ppm was selectively irradiated (an increase of roughly 2 to 3-fold, Fig 2B, 2C and 2D). 1D EXSY analyses were also carried out using a rotating-frame NOESY pulse sequence (ROESY), which revealed identical build-up curves of the intensities of the GSH/HMG cross-peaks at different mixing times to those observed using the NOESY pulse sequence (Fig 2C; for interpretation of the ROESY analyses, see Discussion); no increase in the EXSY-correlation intensity (relative to no-enzyme control) was observed upon incubation with a GFA variant (GFA C54A, S5 Fig), which contains a compromised zinc-binding site, as shown by non-denaturing electrospray ionisation mass spectrometry (S6 Fig). Further, EXSY analysis after incubation of the GFA-containing samples in boiling water to denature the protein for two minutes resulted in a decrease in cross-peak intensities (S5 Fig). Addition of zinc ions (20 μM) to the reaction mixtures did not affect the EXSY cross-peak intensities, either in the absence or presence of GFA (note: purified recombinant GFA contains two zinc binding sites [14]). The ratio of EXSY correlation intensities between the samples with and without GFA was found to be pH-dependent: at low pH (pH Ã 5.5), the ratio was most significant, although the respective intensities were smaller than at higher pH values (S7 Fig). These results suggest a predominance of protein-independent GSH/HMG inter-conversion pathways at higher pH, presumably due to the increased nucleophilicity of the GSH thiolate [12]. Incubation with GFA was also observed to increase cross-peak intensities between GSH and both S-hydroxyethylglutathione (HEG, from reaction of GSH with acetaldehyde, S8  Overall, these NMR analyses reveal that GFA induces an increase in EXSY cross-peak intensities between GSH and hydroxyalkyl-GSH adducts derived from reactions not only with HCHO, but also with acetaldehyde and propionaldehyde, implying binding of GSH / the hydroxyalkyl-GSH adducts to GFA induces an increase in their inter-conversion rates. The binding of GFA to GSH was then investigated. Binding analyses were carried out using two ligand-observed NMR-based methods (saturation transfer difference spectroscopy (STD), and water-ligand observed via gradient spectroscopy (waterLOGSY)) and non-denaturing mass spectrometry. STD and waterLOGSY are NMR methods which monitor the transfer of magnetisation from a selectively irradiated species to a ligand when the two are close in space [27,30]. In the case of STD, 1 H nuclei on the protein of interest are irradiated; magnetisation transfer from the protein to the ligand occurs when the ligand is bound to the protein, and may be used as a measure of ligand binding providing the ligand experiences rapid on-off exchange with the protein over the time course of the experiment. In waterLOGSY, magnetisation transfer is observed between the ligand and irradiated solvent H 2 O; when the small molecule ligand population is free in solution (i.e. not bound to protein), magnetisation transfer results in signals for the ligand with opposite sign to the irradiated H 2 O 1 H resonance, due to the fast tumbling rate of the ligand during the experiment [30]. When the ligand is bound to protein, the net tumbling rate is decreased, which results in perturbation of the signal intensity. Therefore, monitoring the net difference in ligand signal intensities in the presence and absence of protein provides evidence for ligand protein binding. In each of the STD, waterLOGSY and MS analyses, GSH was observed to bind to GFA, with an estimated dissociation constant (K D ) of 500 μM using waterLOGSY (Figs 2E and 2F and S9 and S10). Further, GSH was not observed to bind the GFA C54A variant using waterLOGSY (S11 Fig). Interestingly, GFA was not observed to bind GSH analogues including N-acetyl-cysteine, γ-glutamyl-serinyl-glycine, oxidised glutathione (GSSG) and S-methyl-glutathione using waterLOGSY (S12 Fig). GFA therefore has at least some selectivity for GSH binding. 1D EXSY analyses on mixtures containing HCHO and N-acetyl-cysteine, γ-glutamyl-cysteine, N-acetyl-cysteamine and γ-glutamyl-D- We then investigated whether GFA accelerates the initial rate of HMG formation. Assuming GFA accelerates the rates of both HMG formation and fragmentation (to give GSH and HCHO), the time required for the GSH + HCHO / HMG equilibrium to be established (from mixing of GSH and HCHO) should be accelerated upon addition of GFA, thus resulting in an increase in the initial HMG formation rate. Reaction mixtures of GSH and HCHO were prepared in deuterated Tris buffer as described above, and the samples (either with or without GFA) were monitored over early time-points using 1 H NMR. Interestingly, no detectable differences in initial HMG formation rates were identified in the samples (Fig 2G)-the experiments were attempted on multiple (>10) occasions with different protein batches, including with both His-tagged and non-tagged GFA (S14 Fig). The initial formation rates of HEG (from GSH and acetaldehyde) and HPG (from GSH and propionaldehyde) could not be determined as the reaction mixtures had already reached equilibrium before the first NMR experiment.
The results from the in vitro experiments prompted studies using E. coli cell lysates as a more biologically-relevant model system. E. coli cells are reported to utilise multiple HCHO metabolism pathways including the GSH-dependent pathway [31,32]; however, GFA is not produced. E. coli BL21 (DE3) cells were grown in 2TY medium and were lysed by sonication in 50 mM deuterated Tris buffer in H 2 O pH 7.5 (0.5 mg/mL).  Fig 3A). In E. coli, methanol and formate may be produced from HCHO via direct reduction or oxidation of cysteinyl resonance of GSH, which increased in intensity at longer mixing times. (C) Graph showing the intensity of the GSH cross-peak relative to the inverted HMG resonance in the absence and presence of GFA at different mixing times, using either NOESY or ROESY pulse sequences. τ m = 4-400 ms. (D) Bar graph showing the intensity of the GSH 1D EXSY-correlation relative to the irradiated HMG resonance in the absence (blue) and presence (green) of GFA (τ m = 80 ms). The build-up rates of the 1D EXSY analyses (note: a τ m of 80 ms is within the linear range of the EXSY build-up curves, Fig 2C) correlate with the rates of GSH/HMG exchange at equilibrium. Therefore, the observed increase in correlation intensity in the presence of GFA implies an increase in GSH/HMG inter-conversion rate. (E) Non-denaturing MS analyses of GSH binding to GFA. Two new peaks corresponding to the masses of monomeric GFA (with two zinc ions in complex) bound to one and two GSH molecules respectively were observed upon incubation with GSH (4 equivalents, right). HCHO respectively, reactions which are coupled to oxidation/reduction of NAD + /NADH. Alternatively, formate can be produced via the GSH-dependent pathway via HMG (S3 Scheme) [31,32]. It is also possible that other as yet undefined pathways of formate and methanol formation from HCHO exist in E. coli.
The concentration of [ 13 C]-HMG in the samples, formed via reaction of [ 13 C]-HCHO with GSH, was highest at early time-points, decreasing slowly over the course of the experiment (Figs 3B and S16). Importantly, the rates of both [ 13 C]-methanol and [ 13 C]-formate production were dependent upon the concentration of added [ 13 C]-HCHO (Fig 3C, production  Hence, assuming GFA catalyses the GSH/HMG inter-conversion, addition of recombinant GFA to the lysates should increase the dynamicity of the GSH + HCHO / HMG equilibrium, allowing the equilibrium to respond more quickly to the sequestering of free [ 13 C]-HCHO and/or [ 13 C]-HMG to form [ 13 C]-methanol and [ 13 C]-formate. Therefore, assuming that the rates of the GSH-dependent and non-GSH-dependent HCHO metabolism pathways are different in the lysates, (note: the [ 13 C]-HCHO signal (hydrated form) could not be analysed due to suppression of the nearby water resonance), addition of GFA should induce an increase in formate and methanol formation rates as a consequence of increasing the GSH/HMG inter-conversion rate. However, either with or without added GSH, the presence of GFA in the lysate (added in recombinant form) did not affect the formation rates of [ 13 C]-formate upon addition of [ 13 C]-HCHO (Fig 3C, at 1 mM, [ 13 C]-methanol production was not observed). Overall, the lysate experiments do not support a proposed role for GFA catalysis in HMG formation or degradation.

Discussion
GFA was initially identified as a GSH-dependent HCHO-metabolising enzyme by 2D EXSY analyses and shown to be a zinc-binding enzyme by crystallographic studies [13,14]. In our work, binding experiments using both NMR and mass spectrometric techniques revealed that GFA binds GSH; however, although we observed that GFA can induce increases in EXSY cross-peak intensities between the β-cysteinyl protons of GSH and HMG, our studies demonstrate that GFA does not increase the rate of reaction of GSH and HCHO to form HMG, at least under the tested conditions. The biological relevance of these findings are supported by experiments using cell lysates, which did not reveal an acceleration of HCHO metabolism (by a GSH-dependent pathway or any other pathway) upon addition of GFA. Overall, the evidence suggests that GFA is a GSH-binding protein but does not catalyse HMG formation under standard conditions; although roles in HCHO metabolism cannot be ruled out, any effects would either likely not involve catalysis of HMG formation, or may be restricted to situations / environments where conditions are perturbed. Such assignments require further validation.
The conclusion that changes in EXSY cross-peak intensities induced by GFA are not indicative of GFA-catalysed HMG formation and fragmentation raises the question of how the observed changes arise and of their significance. Typically, EXSY experiments involving small molecules have been applied to study chemical transformations, such as conformational interconversions or ligand exchange processes [26]. In these cases, the mechanism of magnetisation 'transfer' involves the physical interconversion of species that exist at equilibrium, and the observed correlation intensities are dependent on the rate(s) of exchange between these species. Given that GFA binds GSH (with a K D value significantly lower than the GSH concentrations used in the EXSY experiments), and that both denatured GFA and a GFA variant containing a compromised zinc binding site do not induce increased EXSY cross-peak intensities, it seems likely that the observed magnetisation transfer between GSH and HMG is dependent on GFA-GSH binding. One possible exchange pathway may therefore involve binding of both GSH and HMG simultaneously to the GFA active-site; in this case, inter-conversion between the enzyme-bound GSH and HMG by transfer of the hydroxymethyl group may be, in effect, 'catalysed' by GFA, although the GSH/HMG populations would remain constant after interconversion. It is also possible that GSH and HMG are bound within separate 'active-sites' of the GFA homodimer (S17 Fig), which appear sufficiently close to allow transfer of HCHO between the two molecules (S18 Fig). Another plausible explanation for the observed increase in EXSY correlation intensities upon addition of GFA is that magnetisation may be transferred from GSH/HMG to GFA via nOe spin-diffusion processes, followed by re-magnetisation of another molecule of HMG/GSH binding subsequently at the same site. Such processes are utilised in the interligand nOes for pharmacophore mapping (INPHARMA) method for the screening of small-molecule ligands binding to the same protein active site [33]. In this process, NOESY experiments are used to detect NOEs between the two protein ligands that displace one another in the active-site; a process of potential relevance here since the EXSY sequence is essentially identical to that of NOESY, suggesting the INPHARMA process may contribute to EXSY peak intensities. However, both in the absence and presence of GFA, initial build-up curves of the intensities of the GSH/HMG cross-peaks at different mixing times were identical using NOESY and ROESY pulse sequences ( Fig 2C); rotating frame nOes (rOes) are of opposite sign to classical nOes observed for large molecules (such as a GFA-GSH complex) and to EXSY exchange peaks, so the presence of these should act to reduce EXSY cross-peak intensities. These findings suggest that NOE spin-diffusion processes do not contribute to the appearance of exchange crosspeaks between GSH and HMG. Overall our data suggest that the increased EXSY peak intensities are due to binding of GSH/HMG to GFA, possibly due to their mutual inter-conversion in the enzyme bound state.

Conclusions
Cellular studies have provided evidence that both GFA and its human homologue CENPV are involved in formaldehyde metabolism. The combined biochemical and structural results, however, imply that GFA does not catalyse the production of HMG from GSH and HCHO, either in isolated protein form or in crude bacterial cell lysates. However, GFA does bind GSH; thus, one possibility is that GFA / CENPV acts as a 'store' for GSH. Given the established roles of spectra of E. coli BL21 (DE3) cell lysate (0.5 mg/mL in 50 mM Tris buffer pH 7.5) incubated with [ 13 C]-HCHO (6 mM) over time. Production of 13 C-formate (as indicated by the increase in intensity of the doublet resonance at δ H 8.36 ppm, red top left) and [ 13 C]-methanol (as indicated by the increase in intensity of the resonance at δ H 3.36 ppm, red bottom right) are clearly observed. Only one half of the expected doublet resonance is observed for [ 13 C]-methanol in the 1 H spectra due to overlap with the Tris buffer. Note: the relatively high initial [ 13 C]-methanol level (relative to the level of [ 13 C]-formate) is due to contamination of the commercial source of [ 13 C]-HCHO (S15 Fig).  Studies on GFA from Paracoccus denitrificans targeting domains in enzymes involved in eukaryotic epigenetic regulation, including for histone demethylases, it is possible that GSH and HCHO are 'channelled' to one another in a process involving CENPV / GFA. We propose that such channelling is likely to be especially important in the case of the toxic and uniquely reactive carbonyl compound HCHO.
Supporting Information S1 Fig. Alignment of GFA (residues 23-107) from Paracoccus denitrificans and its human homologue CENPV (residues 141-221). Cysteine residues predicted to coordinate zinc ions are highlighted in red (for the tetrahedral coordinated zinc ion) and green (for the trigonal planar coordinated zinc ion) respectively. In this work, full-length GFA was used (GFA WT, residues ; an additional glycine residue and a histidine residue were also present on the Nterminus of this construct as a result of TEV protease-mediated removal of the N-terminal His-Tag after purification.  [14]. The monomers of the GFA homodimer interact via electrostatic and hydrophobic interactions between residues within alpha-helices, which position the partially exposed trigonal planar zinc binding sites together to form a cleft along the dimer interface. There is sufficient space within the cleft to accommodate two molecules of GSH (as observed in a crystal structure of GSH bound to GFA, PDB ID: 1XA8 [14]), and, potentially, one molecule of GSH and one molecule of HMG (note HMG is unlikely to form a disulphide with the protein). (TIF) S1 Scheme. Proposed mechanism of GFA-catalysed formation of S-hydroxymethylglutathione (HMG) from glutathione (GSH) and formaldehyde (HCHO), as proposed by Neculai et al [14]. Incubation of GFA with GSH results in loss of the crystallographically observed trigonal planar coordinated zinc(II) ion, presumably via reaction of C54 and oxidised GSH (GSSG). It is proposed that the zinc ion then binds at another protein binding site, possibly via interactions with GSH molecules as well as protein ligands. The zinc ion can then catalyse reaction of GSH and HCHO to form HMG. After the reaction, the zinc ion returns to the trigonal planar binding site. (TIF) S2 Scheme. Reaction schemes for the initial products of reactions of GSH with HCHO, acetaldehyde and propionaldehyde. For acetaldehyde and propionaldehyde, reaction with GSH forms two diastereomeric hydroxyalkyl adducts. (TIF) S3 Scheme. Possible metabolic pathways of HCHO in Escherichia coli. Upon addition of HCHO, production of formate, and at high HCHO concentration, methanol was observed. However, it is unclear which pathways of formate and methanol production predominate (note: it is also possible that other pathways producing formate and methanol from HCHO are operational). (TIF)