TFM (L-trifluoromethionine), a potential prodrug, was reported to be toxic towards human pathogens that express MGL (L-methionine γ-lyase; EC 18.104.22.168), a pyridoxal phosphate-containing enzyme that converts L-methionine into α-oxobutyrate, ammonia and methyl mercaptan. It has been hypothesized that the extremely reactive thiocarbonyl difluoride is produced when the enzyme acts upon TFM, resulting in cellular toxicity. The potential application of the fluorinated thiomethyl group in other areas of biochemistry and medicinal chemistry requires additional studies. Therefore a detailed investigation of the theoretical and experimental chemistry and biochemistry of these fluorinated groups (CF3S− and CF2HS−) has been undertaken to trap and identify chemical intermediates produced by enzyme processing of molecules containing these fluorinated moieties. TvMGL (MGL from Trichomonas vaginalis) and a chemical model system of the reaction were utilized in order to investigate the cofactor-dependent activation of TFM and previously uninvestigated DFM (L-difluoromethionine). The differences in toxicity between TFM and DFM were evaluated against Escherichia coli expressing TvMGL1, as well as the intact human pathogen T. vaginalis. The relationship between the chemical structure of the reactive intermediates produced from the enzymatic processing of these analogues and their cellular toxicity are discussed.
- methionine γ-lyase (MGL)
- Trichomonas vaginalis
The protozoan parasite Trichomonas vaginalis is a sexually transmitted infection that is reported to infect 147 million people a year. In endemic areas, the parasite poses a serious health problem to the general population [1,2] and is considered a major factor in certain birth complications [3,4] and a risk factor in predisposition to HIV . Current treatment of T. vaginalis infections utilizes the nitroimidazole-based compounds metronidazole and tinidazole ; however, the misuse and overuse of these drugs has led to an increase in acquired resistance and, as a result, creates a demand for alternative chemotherapeutic strategies .
MGL (methionine γ-lyase), a PLP (pyridoxal-5′-phosphate)-dependent enzyme, catalyzes an α,γ-elimination reaction on L-methionine to produce methanethiol, ammonia and α-oxobutyrate [7,8] (Figure 1). Previous studies have proposed that the methionine analogue TFM (L-trifluoromethionine) may act as a potential prodrug in organisms that express a γ-eliminating enzyme, such as MGL . The processing of TFM is hypothesized to produce trifluoromethanethiol that spontaneously decomposes to thiocarbonyl difluoride, a chemically reactive molecule that results in cellular toxicity. The bioactivity of the compound was later confirmed in antimicrobial screens against micro-organisms that contain MGL, such as Entamoeba histolytica, T. vaginalis and Porphyromonas gingivalis [10–12].
Previous investigations on the effects of fluorination on methionine analogue biochemistry indicated alterations in its electronic, as well as steric, properties when incorporated within enzymes . Therefore it was of interest to evaluate the effects of increasing fluorination at the thiomethyl moiety of the methionine analogue on the efficiency of enzymatic processing by MGL, and the effect of fluorination on antiparasitic activity. For example, the processing of the analogue DFM (L-difluoromethionine) would probably serve as a thioformylating agent. Thus it would be of interest to determine if this type of conversion could also be lethal to a parasite containing MGL.
To date, the production of a thioacylating agent upon the processing of TFM and DFM by MGL has not been experimentally confirmed. To identify these reactive agents, the fluorinated methionine analogues were activated by using a model pyridoxal system which is known to mimic the γ-elimination reaction of the enzyme, as well as the TvMGL1 enzyme itself. To investigate the differences in cytotoxicity between TFM and DFM, the levels of cell growth inhibition for the two fluorinated methionine analogues were examined with Escherichia coli expressing TvMGL1 as well as with the intact human pathogen T. vaginalis.
Chemicals were purchased from Bioshop and EMD Chemicals, unless otherwise stated. The cell line M15pREP4 E. coli that contains the gene encoding mgl1 (from Trichomonas vaginalis) was reported previously by McKie et al. .
The recombinant enzyme TvMGL1 was purified from M15pREP4 E. coli as described by McKie et al.  with some minor changes. Following the purification of the enzyme from the Ni-NTA (Ni2+-nitrilotriacetate) superflow resin (Qiagen), the protein was dialysed four times against buffer [100 mM sodium phosphate (pH 7.8), 10 mM EDTA, 0.2 M NaCl, 4 mM DTT (dithiothreitol) and 100 μM PLP], over a period of 15 hr at 4 °C. The samples were concentrated to 25 mg/ml with a 10k MWCO (molecular mass cut-off) centrifuge tube (Sartorius Stedim Biotech & Corning), flash frozen in liquid nitrogen and stored at −80 °C for future use (stable for more than a month).
In the LDH (lactate dehydrogenase)-coupled assay, the 300 μl of reaction mixture consisted of 60 mM Hepes/NaOH (pH 8.0), 5.0 units of porcine heart LDH (Calbiochem; catalogue number 427211), 400 μM NADH (molar absorption coefficient of 6.22 mM−1 · cm−1; Sigma–Aldrich), 0.1 μM TvMGL1, 10 μM PLP and 0.1–1.2 mM substrate. The reaction were incubated for 2 min at 37 °C on a 96-well plate (Greiner Bio One) and initiated with the addition of TvMGL1. The results were collected with a Spectramax M5 plate reader (Molecular Devices), and the kinetic values were determined with Microcal™ Origin® v6.0 using the Michaelis–Menten equation.
The processing of TFM and DFM were detected by 19F-NMR and 1H-decoupled 19F-NMR respectively using a 600 MHz NMR (Burker) spectrometer. The reaction mixture consisted of 20 μM PLP, 30 mM Hepes/NaOH (pH 7.2), 4.2 mM substrate and 10% (v/v) 2H2O. The reactions were performed at 25 °C and initiated with the addition of 0.5 μM TvMGL1. The 19F-NMR results (externally locked with CFCl3) δ: −93.3 p.p.m. for DFM, −41.4 p.p.m. for TFM and −119.8 p.p.m. for the fluoride anion.
Theoretical calculation of the pKa values for the thiol compounds
The pKa values for the thiols were calculated using the Jaguar v4.1 pKa module (Schrödinger), which involved geometry optimization for the protonated and deprotonated species at the HF/B3LYP/6-31G* level of theory, accurate single point energies for the species at the DFT/cc-pVTZ(+) level of theory and the computation of the solvation free energy of the gas phase geometry optimized species. The theoretical pKa values for the reference compounds CH3SH (10.6) and CH3CH2SH (10.6) were similar to the reported literature values (10.3 and 10.6 respectively); and HOCH2CH2SH (10.2) and HSCH2CH2SH (10.2) were ~1 pKa greater than the literature pKa values (9.4 and 9.1 respectively).
Cross-linking of MGL protein upon processing of fluorinated analogues
For the MGL intermolecular cross-linking experiments, the reaction mixture consisted of 100 mM Hepes (pH 8.0), 5.0 mM EDTA/NaOH (pH 8.0), 4.2 mM of the fluorinated methionine analogue (or 4.2 mM methionine in the presence or absence of 40 mM formaldehyde) and 20 μM TvMGL1. The reactions were quenched by the addition of 2-fold loading buffer [100 mM Tris/HCl (pH 6.8), 2% (v/v) 2-mercaptoethanol, 4% (v/v) SDS, 0.2% Bromophenol Blue and 20% (v/v) glycerol], and 10 μl aliquots from the reaction mixture were resolved by SDS/PAGE (13% gel) at 200 V. For the inhibition of cross-linking MGL, a final concentration of 50 mM CA (cysteamine) at pH 8.0 was added to the reaction mixture. The integrated density value for the stained protein bands were calculated with AlphaEase® FC software v6.0 (Alpha Innotech Corporation).
Model processing of fluorinated analogues
For the PLP model system, 1 mM fluorinated methionine analogue and 0.1 mM MPAL (N-methylpyridoxal chloride) were incubated in a 2 mM KOH/methanol solution for 1 h at 21 °C. Afterwards, a final concentration of 0.1 mM Al(NO3)3 and 900 mM morpholine were added sequentially to the reaction mixture, and incubated for 15 h at 21 °C. The product of the TFM reaction was extracted with ethyl acetate and dried under vacuum, whereas the DFM reaction was extracted with CHCl3. The extracted products for DFM and TFM were developed by TLC with CH2Cl2 and ethyl acetate/ethanol (4:1) respectively, visualized at 320 nm and with iodine stain and compared with the authentic standards. The yields for N,N′-bismorpholino thiourea and fluoride based on NMR spectroscopy were 50% and 82% respectively for the processing of TFM (Supplementary Table S1 at http://www.BiochemJ.org/bj/438/bj4380513add.htm). The yields for N-morpholino thioformamide and fluoride, based on NMR spectroscopy, were 16% and 83% respectively for the processing of DFM (Supplementary Table S1).
The synthesis of the fluorinated methionine analogues was adapted from Houston and Honek  and Tsushima et al. . Briefly, 1 mmol of L-homocystine was stirred in liquid ammonia, and the disulfide bonds were reduced with a slight excess of sodium metal (until a purple colour remained). The reaction was carefully quenched by the addition of ammonium chloride, and liquid ammonia was then evaporated off. The reaction was dissolved in deoxygenated ethanol/potassium t-butoxide solution, and chlorodifluoromethane or iodotrifluoromethane (under UV light) was bubbled into the solution to yield DFM or TFM respectively.
The synthesis of N,N′-bismorpholino thiourea was performed as described by Beerheide et al. . Briefly, the mixture consisted of 1 mmol of morpholine, 0.5 mmol of 1,1′-thiocarbonyl diimidazole and 0.5 mmol of triethylamine in CH2Cl2 at 25 °C. The Rf value on TLC was 0.39 in ethyl acetate/hexane (1:1). 1H-NMR (CDCl3) δ: 3.59 (t, J 4.9 Hz, 8 H), 3.72 (t, J 4.9 Hz, 8 H). ESI–MS (electrospray ionization MS) calculated for C9H17N2O2S+ [M+H]+ 217.31; found 217.0899.
The synthesis of N-morpholino thioformamide was performed as described by Mills . Briefly, the reaction mixture consisted of 1.5 mmol of N-dimethyl thioformamide and 1.5 mmol of morpholine in 5 ml of toluene, and the mixture was refluxed overnight at 100 °C. The Rf value on TLC was 0.34 in ethyl acetate/CH2Cl2, 1:4. 1H-NMR (C6D6) δ: 2.33 (t, J 4.9 Hz, 2 H), 2.80 (t, J 4.9 Hz, 2 H), 3.03 (t, J 4.9 Hz, 2 H), 3.61 (t, J 4.9 Hz, 2 H), 8.76 (s, 1 H). ESI–MS calculated for C5H10NOS+ [M+H]+ 132.2; found 132.0432.
The synthesis of MPAL was adapted from Heyl et al.  and Johnston et al. . Briefly, 1 g of pyridoxal hydrochloride was dissolved in methanol to produce monomethylacetal. Subsequently, 200 ml of benzene and 15 ml of methyl iodide was added to the solution step wise, and refluxed in the dark to yield N-methyl pyridoxal monomethylacetal iodide. The solution was distilled to dryness, and the yellow residue was crystallized from methanol-ether. The iodide was removed from the sample with 250 mg of silver chloride in 7 ml of water. After filtration, the sample was titrated to pH 2.0 with 1 M HCl in order to hydrolyse the monomethylacetal. The solution was evaporated to dryness under reduced pressure, and crystallized from water acetone. 1H-NMR (2H2O) δ: 2.60 (s, 3 H), 4.16 (s, 3 H), 5.19 (dd, J 47.0, 13.8 Hz, 2 H), 6.66 (s, 1 H), 8.13 (s, 1 H). ESI–MS calculated for C9H12NO3+ [M]+ 182.2; found 182.0847.
Cloning of mgl1 into the pET vector and cell inhibition assay
The mgl1 gene from the pMGL4100 plasmid was cloned into the pET28b vector (Novagen) at the NcoI and XhoI sites with the appropriate restriction enzymes (NEB). The gene was amplified by overlap-extension PCR with Pwo SuperYield DNA polymerase (Roche Applied Science) in order to replace an internal XhoI site with a non-degenerate codon, and to introduce a linker sequence at the C-terminus of the sequence. The primer pair (Invitrogen) termed MGL1 forNcoI (5′-GATATACCATGGCGCACGAGAGAATGAC-3′) and MGL1 revC222G (5′-TTCTGTTTTCTCCAgGAAGGCGATCTTGC3′) generated the N-terminal fragment, which introduced the NcoI restriction site (underlined) and non-degenerate codon (lower case letter) respectively. The primer pair termed MGL1 forC222G (5′-ATCGCCTTCCTgGAGAAAACAGAAAGCATG-3′) and MGL1 rev1aTHB (5′-cacgtggcaccagacctAAAAGAGCGTCAAG-3′) generated the C-terminal fragment, which introduced the non-degenerate codon (lowercase lettering) and linker region (lower case lettering and underlined) respectively. The N-terminal fragment and C-terminal fragment were combined and amplified with the primer pair termed MGL1 forNcoI and MGL1 rev1bXhoI (5′-GTAAATCTCGAGgctaccacgtggcaccagac-3′), which introduced the XhoI site (uppercase lettering and underlined) and the linker region (lowercase lettering and underlined); and thereby, generating the ORF (open reading frame) of mgl1. Amplification of the PCR products was carried out under the following conditions: denatured at 95 °C, 1 min; 18 cycles (95 °C, 1 min; 63 °C, 1 min; 72 °C, 1 min 45 s); and 11 cycles (95 °C, 1 min; 63 °C, 1 min; 72 °C, 1 min 35 s) and 72 °C, 5 min. The PCR products were recovered using a Silica Bead DNA Gel Extraction Kit (Fermentas). The PCR products and the pET28b vector were doubled digested with NcoI and XhoI restriction enzymes (NEB), purified using the Silica Bead DNA Gel Extraction kit, and ligated with T4 DNA ligase (NEB). The ligation mixture was transformed into electrocompetent XL-1 Blue cells (Stratagene) and selected on LB (Luria–Bertani)/agar medium with 35 mg/l KAN (kanamycin). The plasmids with the insert were purified from the transformed cells using a Qiagen miniprep kit, and then transformed into electrocompetent BL21 (DE3) (Stratagene) cells. This recombinant enzyme exhibited similar kinetics to TvMGL1 that was expressed from the pMGL4100 vector (results not shown).
In the cell-growth-inhibition assay, the BL21 (DE3) cells containing the recombinant MGL (as described above) were grown overnight at 37 °C in the presence of 35 mg/l KAN, inoculated into 10 ml of LB medium (2×106 cells/ml) containing 35 mg/l KAN and induced with 0.1 mM IPTG (isopropyl β-D-thiogalactoside) for 30 min at 37 °C with shaking. The cells were diluted 2-fold in a 96-well plate (Falcon), which contained 35 mg/l KAN, 0.1 mM IPTG and the compound of interest. The plate was sealed with gas-permeable tape, the cells were grown at 37 °C, and the readings were taken at 600 nm when the negative control cells (in the absence of the compound) reached an attenuance of 0.15–0.16. The percentage growth inhibition for each compound was normalized to the uninhibited growth of the negative control cells. The EC50 values of the compounds were determined with Microcal™ Origin® v6.0 using the Hill equation.
In the growth inhibition of T. vaginalis, the procedures were followed as described by McMillan et al. . Briefly, the cells were diluted 2-fold in a 96-well plate (to a final concentration of 105 cells/ml) in Modified Diamond's medium with 10% (v/v) heat-inactivated horse serum and the compound of interest. The plates were sealed with gas-permeable tape and incubated for 18–22 h at 37 °C in a humidified box. Aliquots (100 μl) were transferred to a black 96-well luminometer plate and luminescence was measured with the Cell Titer-Glo® Assay (Promega). The EC50 values were determined with Microcal™ Origin® v6.0 using the standard dose–response equation: where A1 is the lowest value, A2 is the highest value, x is the concentration of compound and n is the Hill slope.
Detecting the products from the enzymatic processing of the fluorinated methionine analogues
The recombinant TvMGL1 was previously cloned into a pQE vector possessing an inducible promoter for overexpression, and with a hexahistidine tag incorporated at the C-terminus of the enzyme . The enzyme was overexpressed in E. coli and purified by a Ni2+-chelating resin. The processing of DFM and TFM by the recombinant enzyme was monitored by 1H decoupled 19F-NMR and 19F-NMR respectively for changes in the chemical shift of the fluorine nuclei of the compounds, as described previously by Alston and Bright  for the processing of TFM by cystathionase (a γ-eliminating PLP-dependent enzyme). A total of 144 scans were taken with a sweep width of +100 to −125 p.p.m., in order to cover the region where the calculated and experimental fluorine signals were expected to appear (Supplementary Table S2 at http://www.BiochemJ.org/bj/438/bj4380513add.htm). In the negative control experiments (i.e. incubation without enzyme), a 19F-NMR resonance signal was observed at−41.4 p.p.m., which corresponded with the fluorine nuclei in TFM (Figure 2a), and in a separate experiment, at −93.3 p.p.m. for the fluorine nuclei in DFM (Figure 2c). Upon the addition of the enzyme, a second 19F-NMR resonance was observed at −119.8 p.p.m. for both TFM and DFM samples (Figures 2b and 2d), which coincided with the chemical shift of the fluoride ion under similar conditions (Figure 2e). All of the 19F-NMR chemical shifts in the current experiments were consistent with the literature values for TFM, DFM and the fluoride ion previously reported [22–25].
No other 19F-NMR chemical shifts were observed during the timeframe and sweep width of the experiments for the intermediates proposed for the enzymatic processing of TFM and DFM. The reported literature values for the 19F-NMR chemical shifts for the fluorine nuclei of trifluoromethanethiol and thiocarbonyl difluoride (i.e. the proposed products in the enzymatic processing of TFM) have been reported to occur at −31.4 and +41.2 p.p.m. respectively in CFCl3 [26,27]. However, there are no known reports of the 19F-NMR chemical shifts for the fluorine nuclei of difluoromethanethiol and thioformyl difluoride (i.e. the proposed products in the enzymatic processing of DFM).
To examine whether the decomposition of the thioacylating agents were thermodynamically favourable, the heats of formation were calculated using T1 and G3(MP2) high level thermochemical methods [28,29]. The nucleophiles, methylamine and water were chosen for the calculations as a model for the ϵ-amines in lysine found on the surface of the homodimer (45 out of a total 52 lysine residues; PDB code 1E5F) and the water present in the reaction mixture respectively. The theoretical calculations indicated that the heats of formation for the reactions between the thioacylating agent and the nucleophiles (methylamine or water) were thermodynamically extremely favourable (Supplementary Figure S8 at http://www.BiochemJ.org/bj/438/bj4380513add.htm).
To examine the stabilities of the thiol products generated from the enzymatic reaction, the theoretical pKa values of the thiols were calculated using previously reported computational methodologies . The pKa values were found to decrease as follows: methanethiol (10.6) >, difluoromethanethiol (5.2)> trifluoromethanethiol (2.8) which was consistent with the previously reported trend for their theoretical gas-phase acidities . Thus analysis of the predicted pKa values indicated that the thiolate forms of trifluoromethanethiol and difluoromethanethiol would be the predominant ionized forms at physiological pH. However, these forms were expected to collapse by ejecting a fluoride ion.
The difficulty in confirming the existence of the thioacylating agent, produced from the enzymatic processing of the fluorinated methionine analogues, was the heterogeneity of potential nucleophiles within the reaction mixture (e.g. buffer, water and enzyme). To trap the thioacylating products, the complexity of the reaction was reduced by using a PLP model system; thus avoiding protein and extra buffers. It has been demonstrated previously that MPAL in a KOH/methanol solution can mechanistically mimic the γ-elimination reaction catalysed by the intact MGL enzyme [20,32]. It was hypothesized that the fluorinated methionine analogues would be processed by MPAL to produce the reactive thioacylating agents (Figure 3, a and b). Morpholine was selected as the trapping agent in the present study because it is a restrained strong nucleophilic amine . The conditions for the PLP model system were followed according to the methods described previously by Karube and Matushima  for normal substrates. The reaction mixture containing synthetic MPAL and the fluorinated methionine analogues were treated with morpholine. Any reaction products trapped by morpholine were extracted by washing the aqueous reaction mixture with ethyl acetate or CHCl3. The major products for the DFM and TFM reactions were consistent with the authentic reference standards (N-morpholino thioformamide and N,N′-bismorpholino thiourea respectively) in terms of their Rf values by TLC (see Experimental section), 1H-NMR and ESI–MS (Supplementary Table S1, and Supplementary Figures S1 and S2 at http://www.BiochemJ.org/bj/438/bj4380513add.htm). Thus the intermediates trapped from the catalysis of DFM and TFM by MPAL were thioformyl fluoride and thiocarbonyl difluoride respectively. A low but detectable additional 19F-NMR resonance at +11.6 p.p.m. during the reaction of TFM with MPAL was consistent with the predicted 19F-NMR resonance for the reaction product of attack of one morpholine on thiocarbonyl difluoride to produce N-morpholinothiocarbonyl fluoride (Supplementary Figure S7 at http://www.BiochemJ.org/bj/438/bj4380513add.htm).
Kinetic parameters for the enzymatic processing of the fluorinated methionine analogues
The activity of the enzyme was initially assayed spectrophotometrically at 320 nm after incubation of the enzymatic products with the derivatizing agent MBTH (3-methyl-2-benzothiazalone hydrazone hydrochloride) as described by Soda et al. . However, the kinetic parameters of TvMGL1 upon processing of DFM at high concentrations (>1.2 mM) could not be accurately obtained due to extensive precipitation upon inclusion of the derivatization reagent (results not shown). To circumvent this issue, another approach was developed (see below) as we hypothesized that the derivatizing agent might not be compatible with the product that was generated from the enzymatic reaction (i.e. thioformyl fluoride).
A LDH-coupled assay was developed in order to obtain the kinetic parameters for the processing of methionine and its fluorinated analogues by TvMGL1, similar to the method described for the γ-lyase activity of cystathionine γ-synthase . The product α-oxobutyrate produced by the activity of TvMGL1, was coupled to the oxidation of NADH by LDH to produce 2-hydroxybutyrate, and the resulting loss in NADH was monitored spectrophotometrically at 340 nm (Supplementary Figure S3a at http://www.BiochemJ.org/bj/438/bj4380513add.htm). The catalytic rate constant (kcat) for TvMGL1 was found to increase as follows: methionine < DFM < TFM; whereas the Km was found to increase as follows: DFM < TFM < methionine (Table 1 and Supplementary Figures S3b and S3c). Analysis of the enzyme's catalytic efficiency indicated that DFM was processed more efficiently than TFM and methionine (Table 1).
Possible modes of inhibition of TvMGL1
The evidence presented so far would suggest that the processing of fluorinated methionine analogues will generate thioacylating agents that have the potential to react with cellular nucleophiles as well as covalently modifying protein. For example, the processing of TFM by TvMGL1 would probably generate thiocarbonyl difluoride, which has the potential to thiocarbamoylate an amine or cross-link another amine on the surface of the protein (Supplementary Figure S4, a at http://www.BiochemJ.org/bj/438/bj4380513add.htm). Alternatively, the processing of DFM by TvMGL1 would generate thioformyl fluoride, which has the potential to thioformylate only a single amine on the surface of the protein since only one fluorine atom is present (Supplementary Figure S4, b). Either one of the two scenarios might interfere with the activity of the enzyme or inactivate it. The possibility of this occurring in the LDH-coupled assay was uncertain, as the decrease in the initial rate of the enzyme after 80 s upon processing of TFM or DFM might also result from product inhibition (Supplementary Figure S3c).
Evidence for protein modification was examined by the extent of cross-linking of MGL, since the processing of DFM by MGL might lead to the sequential reaction of thioformyl fluoride with two primary amines. For this to occur, the sulfur atom in thioformyl fluoride would have to be eliminated as hydrogen sulfide upon reacting with two primary amines (Supplementary Figure S4, c). Precedence for this alternative reaction has indeed been observed for the decomposition of a thioformamide to an amidine as a result of an excess of a primary amine . To examine this possibility, an intermolecular cross-linking experiment was performed by incubating TvMGL1 with the substrate of choice (either methionine plus formaldehyde, DFM or TFM) for 30 min at 30 °C. Aliquots for each of the reactions were loaded on to an SDS gel in order to visualize any changes in molecular mass of TvMGL1. The positive controls formaldehyde and TFM (when processed by MGL) were selected for the experiment because they are known cross-linking agents [36,37]. Incubation of TvMGL1 with either formaldehyde, TFM or DFM followed by an SDS/PAGE analysis revealed the presence of higher molecular mass proteins than expected for TvMGL1 (Figure 4). Based on the integrated density values, the concentrations for the higher molecular mass proteins (corresponding to ~88 kDa) were 1.99 μM, 199 nM and 225 nM for the samples treated with formaldehyde, TFM or DFM respectively. Importantly, samples containing methionine in place of the potential prodrugs resulted in no observable changes in the molecular mass of TvMGL1.
During the course of the cross-linking experiments, the appearance of a lower molecular mass band (~32 kDa) was observed for the samples treated with DFM and TFM (Figure 4), similar to a cross-linking study for the processing of TFM by EhMGL (MGL from Entamoeba histolytica) . However, further analysis of the cleavage product was not possible, as the protein fragment was either lost or failed to resolve during the subsequent chromatographic purification steps (i.e. size exclusion, ion exchange and C8 chromatographic steps).
To provide further evidence that the thioacylating agents may lead to intermolecular cross-linking of the enzyme, attempts were made to reduce cross-linking by either inhibiting the enzyme or by scavenging any reactive intermediates with a nucleophilic compound included in the reaction buffer. Cysteamine was selected as a suitable candidate for this experiment because it is known to efficiently react with carbonyl groups  and with the free or bound PLP to produce a thiazolidine . Therefore protection against cross-linking might be achieved when cysteamine reacts with the electrophilic thioacylating agent or inactivates the enzyme upon sequestering PLP. Upon visualization of the SDS gel, the addition of cysteamine reduced the extent of intermolecular cross-linking for the samples treated with formaldehyde, TFM and DFM (Figure 4). These results strongly support the notion that the processing of either TFM or DFM by TvMGL1 could produce some protein cross-linking as well, although levels of these cross-linked proteins, under the above experimental conditions, appear relatively low.
Cellular toxicity of the fluorinated methionine analogues in organisms that express TvMGL1
To examine the differences in the cytotoxicity effects of TFM and DFM, the level of inhibition for cell growth, using an E. coli model system that expresses TvMGL1, was measured spectrophotometrically by following an earlier reported procedure . For this experiment, the mgl1 gene from T. vaginalis was cloned into an expression vector and overexpressed in E. coli. The cells transformed with the plasmid containing the mgl1 gene were incubated with various concentrations of the fluorinated analogues, and the level of inhibition was measured relative to a negative control (i.e. cells not treated with the compound). The positive controls were conducted in the presence of either ampicillin and NaF as inhibitors since ampicillin is known to inhibit bacterial growth, and fluoride ion can inhibit certain enzymes such as acid phosphatase . As expected, ampicillin and an excess of NaF inhibited cell growth, whereas no effects were observed upon treatment with methionine (Table 2 and Supplementary Figures S5b and S5a respectively at http://www.BiochemJ.org/bj/438/bj4380513add.htm). When the cells were treated with DFM and TFM, inhibition of cell growth was observed (an EC50 value of ~20 μM for both compounds; Table 2 and Supplementary Figures S5c and S5d respectively).
The negative control cells (lacking the mgl1 gene) were examined in order to correlate the inhibition of cell growth with the activation of the fluorinated methionine analogues by TvMGL1. Cell growth was inhibited when treated with either ampicillin or NaF, but not with DFM, TFM or methionine (Supplementary Figure S6 at http://www.BiochemJ.org/bj/438/bj4380513add.htm). Overall, the fluorinated methionine analogues did not inhibit cell growth for E. coli cells lacking TvMGL1.
To determine if DFM is toxic to organisms that endogenously express MGL, the growth inhibitory effects of DFM were studied on T. vaginalis, an anaerobic pathogen known to contain high levels of MGL activity [8,10]. DFM was found to be active against this pathogen, approximately 20-fold less potent than the clinically approved drug metronidazole (Table 2 and Supplementary Figures S5e and S5f), and 2-fold less potent than the EC50 value obtained in the E. coli model (Table 2). Assuming that the transporter efflux between TFM and DFM in T. vaginalis does not significantly differ from the E. coli model, the growth inhibitory effects should reflect the potency of the compound between the two model organisms.
19F-NMR studies demonstrated that TvMGL1 rapidly processed DFM as the 19F-NMR signals rapidly disappeared in the presence of enzyme (Figure 2d). The proposed intermediates, difluoromethanethiol and thioformyl fluoride were not observed by 19F-NMR, since the species were likely to be too short lived for the observable timeframe of the experiments. Similar results were observed for the processing of TFM by TvMGL1, i.e. no 19F-NMR signals were observed for trifluoromethanethiol and thiocarbonyl difluoride. Thus evidence for the processing of DFM and TFM by TvMGL1 was supported by the additional resonance that coincided with the fluorine nucleus of the fluoride ion, i.e. the final decomposition product.
The notion that the above reactive intermediates are too short lived to be observed is consistent with the thiols existing as thiolates based on the calculated pKa values, and the anionic hyperconjugation and electrostatic effects that favour their decomposition to the thioacylating agents . The observed reactivity of the latter compounds supports the calculated heats of reaction (Supplemental Figure S8), in which nucleophilic attack on the thioacylating agents by water or amines are thermodynamically favourable processes. This would explain the production of fluoride ion as observed by 19F-NMR (Figure 2), and previous reports on the instability of trifluoromethanethiolate and thiocarbonyl difluoride [9,42,43].
The incompatibility of the derivatizing agent MBTH with DFM in the presence of TvMGL1 led to the development of an LDH-coupled assay. For the turnover of methionine by TvMGL1, the kinetic parameters obtained from the LDH-coupled assay were consistent with the values obtained utilizing the MBTH assay . For the processing of TFM, the kcat for TvMGL1 was ~6-fold greater than the previously reported activity of a MGL1/MGL2 mixture from T. vaginalis, and with similar Km values, which were obtained utilizing the MBTH assay method . These differences in catalytic activity were likely to have arisen from the differences in kinetic parameters between the two MGL isoenzymes as reported for the turnover of methionine . Additionally, its catalytic activity was comparable with the EhMGL2 enzyme, and its Km value was similar to EhMGL1, both of which were obtained utilizing the MBTH method . For the processing of DFM by an MGL enzyme, to our knowledge, this is the first report of its kinetic values.
The correlation between increasing fluorination of the compound and increasing rate of reaction (i.e. TFM>DFM>methionine) would suggest that the inductive effect and anionic hyperconjugation of the fluorinated methyl thiolates improve its leaving ability. This is also consistent with the idea that the C–S bond cleavage is a rate-determining step of the enzyme as reported by a 1H-NMR study on MGL from Pseudomonas putida .
The activity of the enzyme did not appear to be compromised as a result of the increase in volume for each fluorine atom in the fluorinated methionine analogues (19% for the trifluoromethanethio group and 12% for the difluoromethanethio group relative to the methanethio group) . Additionally, the enzyme appears to favour the enzyme–substrate complex for DFM over methionine as indicated by the 10-fold difference in Km, despite its bulky substitution. On the other hand, TFM resulted in a similar Km value as methionine, which suggests that the bulkier substitution did not enhance the enzyme–substrate complex. Overall, the processing of DFM by TvMGL1 is catalytically more efficient than TFM and methionine.
The processing of the fluorinated methionine analogues by TvMGL1 probably inactivated the enzyme as a result of covalent modification. This was supported by a similar finding for the processing of TFM by EhMGL , intermolecular cross-linking of TvMGL1 (Figure 4), the thioacylating agent reacting with a restrained secondary amine in the PLP model system (Supplementary Figures S1 and S2) and the thermodynamically favourable process of the thioacylating agent reacting with nucleophiles such as the side chains of lysine residues on the surface of the protein or water (Supplementary Figure S8). It is tempting to extrapolate the results for the amount of cross-linked dimers to explain the apparent decrease in the initial rate for the LDH-coupled assay after 60 s, i.e. 6.6 nM and 7.5 nM of cross-linked homodimeric enzymes for the enzymatic processing of TFM and DFM respectively (Supplemental Figure S3c). However, caution should be exercised when applying quantitative values to other experiments such as the LDH-coupled assay and the cell-inhibition assay, because of differences in protein concentration, quantities of cross-linking agent and potential side reactions. More importantly, the rate of protein cross-linking is expected to be influenced by a variety of factors such as proximity effects, number of collisions, amount of cross-linking agent and competing side reactions (e.g. dimer or monomer form of MGL, concentration of macromolecules and the amount of competing water molecules for the cross-linking agent). Therefore cross-linking is expected to be higher within the crowded cellular cytoplasm such as the cell inhibition assay (an estimated concentration of 300–400 g/l of protein and RNA inside E. coli) . On the other hand, cross-linking is expected to be lower for dilute solutions such as the LDH-coupled assay (4.4 mg/l of TvMGL1 and 0.15 g/l of LDH) and the cross-linking experiment (0.88 g/l of TvMGL1).
In the PLP model system, analysis of the results suggested that the processing of DFM and TFM by MPAL led to the production of thioformyl fluoride and thiocarbonyl difluoride respectively. Therefore it was very probable that these intermediates were generated during the course of the enzymatic reaction, since MPAL is known to follow the same γ-elimination reaction steps as MGL in the turnover of methionine [20,32]. Thus thioformyl fluoride has the potential to thioformylate an amine, whereas thiocarbonyl difluoride has the potential to sequentially thiocarbamoylate two amines.
The model proposed above does not rule out the possibility that thioformyl fluoride may sequentially react with two amines to produce an amidine with the elimination of the sulfur atom as hydrogen sulfide (Supplementary Figure S4, c). Evidence for the latter proposal was supported by the intermolecular cross-linking of TvMGL1 for the enzymatic processing of DFM (Figure 4). The results suggested that upon processing of DFM by TvMGL1, the enzyme (~44 kDa) was cross-linked to an adjacent subunit to form a covalently linked dimer (~88 kDa), which was similar to the cross-linking pattern observed for the two positive controls (formaldehyde and TFM). The enzymatic processing of DFM led to two faint higher molecular mass proteins that corresponded to cross-linked trimers (~132 kDa) and tetramers (~176 kDa), which was consistent with the cross-linking pattern for the positive control (formaldehyde). The intermolecular cross-linking results were unexpected since the release of thioformyl fluoride is expected to react with only a single lysine residue. Overall, the additional finding remains consistent with the idea that basic side chain residues on the surface of the enzyme were susceptible to covalent modification following the processing of the fluorinated methionine analogues by MGL. The different modes of covalent modification described in this paper suggest that there are likely to be a variety of cytotoxic and/or inhibitory events taking place within the organism expressing MGL.
In the E. coli model system, the quantity of the fluorinated methionine analogue required to inhibit 50% of cell growth was 2-fold greater than the values obtained for the intact parasites, T. vaginalis (Table 2) and E. histolytica  in a 24 h and 72 h time period respectively. This may indicate that there are some limitations to using the E. coli model when compared with intact parasite studies, such as the differences in their metabolism (see below). Overall, the model system provides a facile and safer screening method for testing potential compounds before further evaluation on MGL-containing organisms (e.g. T. vaginalis, E. histolytica and P. gingivalis).
For the processing of the fluorinated methionine analogues by MGL, the reaction of the thioacylating agent with cellular macromolecules is probably the mode of cellular toxicity, given the evidence supporting its existence in the present study. In the E. coli model, it is unlikely that cellular toxicity arises from the release of the fluoride ion upon hydrolysis from the thioacylating agent, since a high concentration of NaF is required to inhibit E. coli growth (Table 2), which is consistent with a previous study . In a similar study, Hofsten  found that the release of intracellular fluoride upon induction of β-galactosidase by fluoro β-D-galactoside (which is known to hydrolyse intracellularly to fluoride ion) does not inhibit cell growth in E. coli. In T. vaginialis, the reported release of toxic H2S upon hydrolysis of thiocarbonyl difluoride  probably does not inhibit cell growth; since the anaerobic parasitic protozoa lacks the cellular target cytochrome oxidase , and MGL is implicated in generating H2S from homocysteine for cysteine biosynthesis [37,50]. On the other hand, H2S may have inhibited the growth in E. coli since the organism contains cytochrome oxidase .
In summary, a detailed investigation of the enzymatic processing of fluorinated methionine residues was undertaken, providing insight into the mechanistic aspects of substrate and inhibitor turnover by this key protozoan enzyme. Trapping of the resulting reactive compounds (i.e. thioacylating agents) provides, for the first time, chemical knowledge of the structure and reactivity of these compounds. The toxicity of these compounds were further studied in intact E. coli and T. vaginalis, and provide further impetus to the exploration of these fluorinated structural types to other areas of cellular biochemistry.
Ignace Moya synthesized all of the compounds, and performed the experiments and analysed the data for most of the experiments shown. Gareth Westrop and Graham Coombs performed the experiments and analysed the data for the model pathogen T. vaginalis G3. John Honek performed the calculations for the heats of formation and 19F-NMR theoretical chemical shifts. Ignace Moya and John Honek wrote the paper and designed the project.
This work was supported by the Natural Sciences and Engineering Research Council of Canada (to J.F.H.), an Ontario Graduate Scholarship (to I.A.M.), and the University of Waterloo.
We thank Jan Venne for her assistance with the 19F-NMR experiments, Dr Richard Smith for performing the ESI–MS experiments, and Dr Chris Phenix (Thunder Bay Regional Research Institute, Thunder Bay, Canada) and Dr Yu Luo (University of Saskatchewan, Saskatoon, Canada) for their helpful discussion and critical reading of the paper prior to submission.
Abbreviations: CA, cysteamine; DFM, L-difluoromethionine; ESI–MS, electrospray ionization MS; IPTG, isopropyl β-D-thiogalactoside; KAN, kanamycin; LB, Luria–Bertani; LDH, lactate dehydrogenase; MBTH, 3-methyl-2-benzothiazalone hydrazone hydrochloride; MGL, methionine γ-lyase; EhMGL, MGL from Entamoeba histolytica; MPAL, N-methylpyridoxal chloride; PLP, pyridoxal-5′-phosphate; TFM, L-trifluoromethionine; TvMGL1, MGL1 from Trichomonas vaginalis
- © The Authors Journal compilation © 2011 Biochemical Society