To survive hostile conditions, the bacterial pathogen Mycobacterium tuberculosis produces millimolar concentrations of mycothiol as a redox buffer against oxidative stress. The reductases that couple the reducing power of mycothiol to redox active proteins in the cell are not known. We report a novel mycothiol-dependent reductase (mycoredoxin-1) with a CGYC catalytic motif. With mycoredoxin-1 and mycothiol deletion strains of Mycobacterium smegmatis, we show that mycoredoxin-1 and mycothiol are involved in the protection against oxidative stress. Mycoredoxin-1 acts as an oxidoreductase exclusively linked to the mycothiol electron transfer pathway and it can reduce S-mycothiolated mixed disulphides. Moreover, we solved the solution structures of oxidized and reduced mycoredoxin-1, revealing a thioredoxin fold with a putative mycothiol-binding site. With HSQC snapshots during electron transport, we visualize the reduction of oxidized mycoredoxin-1 as a function of time and find that mycoredoxin-1 gets S-mycothiolated on its N-terminal nucleophilic cysteine. Mycoredoxin-1 has a redox potential of −218 mV and hydrogen bonding with neighbouring residues lowers the pKa of its N-terminal nucleophilic cysteine. Determination of the oxidized and reduced structures of mycoredoxin-1, better understanding of mycothiol-dependent reactions in general, will likely give new insights in how M. tuberculosis survives oxidative stress in human macrophages.
Every year 9.4 million people are infected with Mycobacterium tuberculosis, the causative agent of tuberculosis, a disease that results in 1.7 million deaths worldwide (Goldberg et al., 2012). Multi-drug resistance and co-infection with HIV challenge the global efforts to stop the disease and necessitate the search for new anti-TB drugs and drug targets. An appealing cellular system to target is the redox homeostasis network of M. tuberculosis. During the course of the disease, M. tuberculosis is encapsulated by alveolar macrophages to form granulomas. In these necrotic regions the pathogen is exposed to high concentrations of reactive oxygen intermediates (ROI) and reactive nitrogen intermediates (RNI) produced by the macrophages (Nathan and Shiloh, 2000; Ehrt and Schnappinger, 2009).
To survive hostile conditions, M. tuberculosis continuously monitors the variation in the intracellular redox state (Trivedi et al., 2012). Two component proteins, DosS and DosT, sense changes in oxygen, nitric oxide and carbon monoxide levels, while WhiB3 and anti-sigma factor RsrA monitor changes in intracellular redox state. With these and other unidentified redox sensors, M. tuberculosis orchestrates its metabolic pathways to survive in nutrient-deficient, acidic, oxidative, nitrosative, and hypoxic environments inside granulomas. M. tuberculosis employs ROI and RNI protecting and scavenging enzymes like catalase (KatG), superoxide dismutase (SodA and SodC), alkylhydroperoxide reductase (AhpC), and the one-cysteine peroxiredoxin AhpE to neutralize the redox stress generated by the host immune system (Manca et al., 1999; Piddington et al., 2001; Bryk et al., 2002; Li et al., 2005).
The major low-molecular-weight thiol in Mycobacteria that maintains the intracellular thiol-disulphide homeostasis during bursts of oxidative stress is mycothiol (MSH). MSH plays an important role in the defence against several external stresses including oxidative stress, alkylating agents and antibiotics (Rawat et al., 2002; Buchmeier et al., 2006). MSH comprises a functional cysteine residue and the saccharides inositol and N-glucosamine (Newton et al., 1995), and is present in millimolar concentrations (Newton et al., 1996). Functionally, MSH can be considered as the equivalent of glutathione (GSH). The MSH pool in the cell is maintained in the reduced form by the flavoprotein mycothiol disulphide reductase (Mtr) (Patel and Blanchard, 1999). Mtr reduces mycothiol disulphide (MSSM), sometimes referred to as mycothione, to MSH at the expense of NADPH. Recently, we uncovered a new thiol-disulphide reductase that shuttles electrons between MSH and arsenate reductase, in Corynebacterium glutamicum, mycoredoxin-1 (Mrx1) (Ordóñez et al., 2009). Mrx1 is a small 10 kDa protein with a glutaredoxin-like sequence. The glutaredoxins (Grxs) are already known for a long time (Holmgren, 1976) as GSH-dependent disulfide oxidoreductases that catalyse a variety of thiol-disulfide exchange reactions, including protein glutathionylation and deglutathionylation (Lillig et al., 2008; Mieyal et al., 2008; Gallogly et al., 2009). Grxs are small, heat-stable proteins with a typical Trx-fold structure (Eklund et al., 1984; Martin, 1995; Collet and Messens, 2010). During catalysis, the thiolate of the active site nucleophilic cysteine of Grx attacks the disulfide bond of a S-glutathiolated protein, releasing the protein thiol in the reduced form while becoming itself glutathiolated. Then, GSH attacks the glutathiolated Grx, releasing reduced Grx and GSSG.
Here, we report the structure and the function of Mycobacterial Mrx1. Mrx1 is exclusively linked to the mycothiol electron transfer pathway and it can reduce S-mycothiolated mixed disulphides. With this study on Mrx1, we have identified the partner oxidoreductase of MSH and we bridge the gap between protein disulphides and MSH.
The Δmrx1 and mshA− strains of Mycobacterium smegmatis are more sensitive to oxidative stress
To assess the physiological role of Mrx1 we determined the phenotype of a mrx1 and a mshA knockout in Mycobacterium smegmatis. M. smegmatis is a fast growing non-pathogenic Mycobacteria often used as a model to study M. tuberculosis (Shiloh and DiGiuseppe Champion, 2010). MshA is the first enzyme in the biosynthesis pathway of MSH that encodes for a glycosyl transferase, and which seems to be essential for the growth of M. tuberculosis in the absence of catalase (Buchmeier and Fahey, 2006). Mrx1 and MshA are highly conserved between M. smegmatis and M. tuberculosis (72% and 69% sequence identity respectively). For all the tested oxidizing and H2O2-inducing agents [hydrogen peroxide, menadione, cumene hydroperoxide (CuOOH), diamide, CuSO4, iodoacetamide (IAM), sodium hypochlorite and 2,4-dinitrochlorobenzene (DNCB)], hydrogen peroxide, Cu2+ and IAM stress give a significantly larger zone of clearance for Δmrx1 than for wild type (Fig. 1 and Table S1). On the other hand, the mshA− strain of M. smegmatis (Rawat et al., 2002) is much more sensitive to oxidative stress than the Δmrx1 strain, and this for all chemical reagents tested, indicating the importance of MSH for survival under oxidative stress conditions. In conclusion, MSH seems to be important for protection during oxidative stress, while the role of Mrx1 in maintaining proteins in a reduced state seems to be rather limited. We assume that Mrx1 might become important for the demycothiolation of mycothiol-protected sulphurs after a period of stress.
Oxidized Mrx1 is specifically reduced with electrons from the MSH/Mtr/NADPH pathway
First, we identified the electron transfer pathway coupled to Mrx1. To do so, we reconstructed the MSH/Mtr/NADPH electron transfer pathway by incubating MSH with Mtr and NADPH. Oxidized Mrx1 with a single disulphide between its active site cysteines was added as substrate for this pathway (Fig. 2A). By monitoring the decrease in the absorption at 340 nm due to NADPH consumption, we demonstrate that electrons are transferred via the MSH/Mtr/NADPH pathway to oxidized Mrx1. After the reaction Mrx1 was analysed on a Vydac C4 reversed phase column (RPC). The 280 nm RPC trace line shows a shift from oxidized to the reduced Mrx1 after the reaction, proving the reduction of oxidized Mrx1 at the expense of NADPH by the MSH/Mtr/NADPH pathway. We also checked the stoichiometry of the reaction by varying the concentration of MSH. The minimum concentration of MSH needed for the reduction of 50 μM oxidized Mrx1 is 100 μM MSH (Fig. S1). Decreasing the concentration to 50 μM MSH drops the initial velocity to 16% of the initial velocity, while doubling the concentrations (200 μM) does not increase the initial velocity significantly (116%). The electron transfer reaction needs 2 molecules of MSH to reduce the disulphide bond of oxidized Mrx1. Although not present in M. tuberculosis, we also tested the reduction of oxidized Mrx1 by GSH (Fig. 2B and Fig. S1). We could show that oxidized Mrx1 is slowly reduced by GSH in a coupled enzyme assay with yeast glutathione disulphide reductase (GR), GSH and NADPH. After 20 min reaction with 200 μM GSH in the GR/NADPH assay 75% of the oxidized Mrx1 is reduced, while 100 μM MSH in the Mtr/NADPH assay reduces all the oxidized Mrx1. However, comparing the initial velocities of the reduction of Mrx1 by MSH in the Mtr/NADPH assay with the reduction by GSH in the GR/NADPH assay shows that Mrx1 is kinetically much more specific for MSH. Further, we tested the possibility that Mrx1 receives electrons from the thioredoxin reductase (TrxR)/NADPH electron transfer pathway (Fig. 2C). No electron transfer was observed, and after the reaction, Mrx1 was still in its oxidized form. So, oxidized Mrx1 seems to be specifically reduced by the MSH/Mtr/NADPH pathway in M. tuberculosis.
Mrx1 reduces MSH mixed disulphide bonds using electrons from the MSH/Mtr/NADPH pathway
As Mrx1 and MSH seem to play a role in the defence against hydrogen peroxide (Fig. 1), we then investigated the possible role of Mrx1 as hydrogen peroxide scavenger. Hydrogen peroxide was added to the MSH/Mtr/NADPH pathway in the presence and absence of Mrx1 (Fig. S2). The addition of Mrx1 does not result in an additional decrease of the absorption at 340 nm. With the Ferrous Oxidation of Xylenol orange (FOX) assay (Wolff, 1994), we tested the residual hydrogen peroxide concentration after the reaction. The MSH/Mtr/NADPH pathway consumed all hydrogen peroxide and the addition of Mrx1 had no extra effect (Fig. S2). MSH scavenges the hydrogen peroxide in the absence of Mrx1, and the formed MSSM drives the electron transfer by consuming NADPH. However, the hydrogen peroxide scavenging role of MSH is only at a low rate of 10−2 M−1 s−1 (using 500 μM MSH), which is several orders of magnitudes lower than the reaction rate of the known peroxidases of M. tuberculosis (104 to 108 M−1 s−1) (Jackett et al., 1978; Manca et al., 1999; Piddington et al., 2001; Bryk et al., 2002). A direct hydrogen peroxide detoxification role of MSH is therefore highly unlikely in vivo.
Finally, we tested whether Mrx1 can reduce a mixed disulphide with MSH, which is formed during oxidative stress to prevent irreversible cysteine oxidation to sulphonic acid (Roos and Messens, 2011). In a mycothiol-coupled hydroxyethyl disulphide (HED) assay (Holmgren, 1979) (Fig. 3A), we showed that Mrx1 reduces the mixed disulphide between MSH and HED. Furthermore, Mrx1 cannot reduce the mixed disulphide between GSH and HED (Fig 3B). Thus, Mrx1 seems to be a specific reductase that exclusively recognizes and reduces mixed disulphides with MSH.
The saccharide moieties of mycothiol inhibit electron transfer to oxidized Mrx1
To further explore the specificity of Mrx1 for MSH, we performed inhibition experiments using the saccharides moieties of MSH: myo-inositol and N-acetyl-D-glucosamine (GlcNAc). Changes in the initial velocity of the reduction of 50 μM oxidized Mrx1 by 100 μM MSH, 0.5 μM Mtr and 500 μM NADPH were evaluated in the presence of varying concentrations of saccharides (Fig. 4). To check the specificity of inhibition, we used glucose. In the presence of 100 μM myo-inositol, the initial velocity drops with a factor of two, while 100 μM glucose gives only 13% inhibition. Increasing the concentration to 500 μM myo-inositol results in residual background activity only, while with 500 μM glucose still 45% activity remains. GlcNAc is a stronger inhibitor than myo-inositol since addition of 100 μM GlcNAc already gives background residual activity, and increasing the concentration to 500 μM gives similar residual activity. A concentration of 100 μM of GlcNAc seems to compete directly with the 100 μM of MSH for the binding on oxidized Mrx1. Combining 100 μM myo-inositol with 100 μM GlcNAc gives again background activity. These inhibition data clearly indicate the presence of a binding pocket for the saccharide moieties of MSH on the surface of Mrx1.
Mrx1 has a thioredoxin structural fold with a putative MSH binding site
We solved the solution structures of oxidized and reduced Mrx1. The 1H, 15N chemical shifts for all residues were assigned in oxidized Mrx1, except for residues 15 and 16 which were not detected in the 1H,15N correlated spectra (Fig. 5A), likely due to conformational exchange in the active site region. The 1H, 13C and 15N chemical shifts of reduced and oxidized Mrx1 have been deposited in the BioMagResBank (http://www.bmrb.wisc.edu/) under accession numbers 18322 and 18325 respectively. Talos+ dihedral and NOE-derived distance restrains were used to calculate the structures of reduced and oxidized Mrx1. The coordinates were deposited to the Protein Databank under accession numbers 2LQO and 2LQQ respectively. The ensembles of 20 lowest energy structures of both forms show very good Ramachandran statistics while fulfilling the experimental data (Table 1), reflecting the high quality of the 3D structures. The ensemble of oxidized Mrx1 is slightly less well defined reflected by the higher backbone root mean square deviation (rmsd) over the ensemble, likely a result of the lower number of restraints obtained for oxidized Mrx1.
Table 1. Structural statistics for the 20 NMR structures of reduced and oxidized Mrx1
Reduced PDB: 2LQO
Oxidized PDB: 2LQQ
All residues except those in the flexible N- and C-terminus were used in the RMSD analysis. Ramachandran statistics were obtained for all residues from PROCHECK analysis.
Distance restraints total
Short range (i − j = 0)
Medium Range (1 ≤ |i − j| ≤ 4)
Long range (|i − j| ≥ 5)
NOE violations > 0.5 Å
0 ± 0
1.1 ± 0.8
Dihedral violations > 5°
0.1 ± 0
0.5 ± 0.9
RMSD from average (Å)
Backbone N, CA, C′, O
0.43 ± 0.06
0.60 ± 0.08
0.91 ± 0.07
1.12 ± 0.08
Most favoured regions (%)
Additional allowed regions (%)
Generously allowed regions (%)
Disallowed regions (%)
Mrx1 has a thioredoxin-fold consisting a four-stranded antiparallel β-sheet surrounded by three α-helices (Fig. 5D). This is the common fold found in several oxidoreductases including but not limited to thioredoxin (Trx), glutaredoxin (Grx), NrdH-redoxin and tryparedoxin (Holmgren et al., 1975; Martin et al., 1993; Alphey et al., 1999; Xia et al., 2001). In all these proteins and in Mrx1, the redox active CXXC motif is located at the N-terminus of the first α-helix (α1) and a conserved proline (Pro57) opposite to the active site is present in cis conformation (Fig. 5C). In both the oxidized and the reduced forms of the protein, the N-terminal cysteine of the CXXC motif (Cys14) is solvent exposed whereas the C-terminal cysteine (Cys17) is more buried.
The major structural differences between the reduced and oxidized forms are located around the C14G15Y16C17 active site motif (Fig. 5B and D). In the oxidized form, the 2 cysteines are disulphide bonded resulting in a rearrangement of the surrounding residues to allow the formation of the disulphide bond. Procheck analysis shows 4 β-strands formed between the residues 6–10 (β1), 32–35 (β2), 58–61 (β3) and 66–68 (β4) in the reduced form, and between residues 6–9 (β1), 31–34 (β2), 58–61 (β3) and 66–68 (β4) in the oxidized form. The 3 α-helices consist of residues 15–26 (α1), 40–48 (α2) and 72–83 (α3) in the reduced form, and of residues 16–26 (α1), 40–49 (α2) and 72–83 (α3) in the oxidized form (Figs 5D and 6A). In 9 out of the 20 models of oxidized Mrx1, a short helical turn (η1) appears at the N-terminus of helix α2 (Figs 5D and 6A). Comparison with Grx and Trx shows that a similar helical turn is present in the oxidized and glutathionylated form of yeast Grx1 (PDB codes: 3C1R and 3C1S), and in both oxidized and reduced yeast Grx2 (PDB codes: 3CTG and 3CTF). In E. coli Grx1 (PDB code: 1EGO), Grx3 (PDB code: 1FOV) and M. tuberculosis TrxC (PDB code: 2I1U) this short helix is absent.
The linker between strand β4 and helix α3 (Thr68 till Ala72) determines the position of helix α3 relative to strand β4. This five amino acid long linker is conserved in 90 putative Mrx1 sequences (blast search within the group of Actinomycetes using M. tuberculosis Mrx1 as search entry) and contains a highly conserved patch (T68N69P70) at the surface of Mrx1 (Fig. 6A and B). This patch is connected to a conserved groove formed by Val56, Pro57 and Thr58 connecting the solvent exposed patch with Cys14. The groove is only exposed in half the models (compare the structures of the two NMR models in Fig. 6B). In the other half of the models, the flexible side-chain of Tyr16 covers the groove.
In Mrx1, this groove might be involved in the interaction with the amide bond between Cys and N-glucosamine of MSH. So far, attempts to dock MSH into the groove of Mrx1 yielded unconvincing results. Mrx1 may however bind MSH via an induced fit mechanism as observed for the binding of GSH to human Grx1 (Jensen et al., 2011). Mrx1 lacks the binding pocket for GSH as found in Grx (Bushweller et al., 1992), and there is no hydrophobic pocket present as in Trx and NrdH-redoxin (Stehr and Lindqvist, 2004) for the binding of TrxR (Lennon et al., 2000; Stehr et al., 2001).
Mrx1 gets S-mycothiolated on its nucleophilic cysteine
The reaction mechanism of the reduction of Mrx1 by MSH was examined by recording heteronuclear single quantum correlation (HSQC)-spectra during the course of reaction. We added 15N-labelled oxidized Mrx1 to the MSH/Mtr/NADPH electron transfer mixture and we followed the reaction in the NMR tube (Fig. 7A). HSQC-spectra of Mrx1 were recorded at different time points during the course of the reduction reaction. To slow down the reaction and to enable the recording of consecutive HSQC spectra in function of time, the experiment was carried out at room temperature at pH 6.5. We selected the cross-peaks of Thr11, Asn69 and Gly84 (Fig. 7A). Initially (t = 0), in oxidized Mrx1, we see a single cross-peak for all three residues (Fig. 7A). After 20 min, additional cross-peaks appeared for Thr11 and Asn69. Both residues are located in the vicinity of the active site (Fig. 6B). For Gly84 at the C-terminus of Mrx1 only a single cross-peak was observed. We took the cross-peak of Thr11 to follow the changes in function of time (Fig. 7A). Comparing the spectra recorded for the oxidized and reduced form (Fig. 5A), we could conclude that one of the new cross-peaks of Thr11 is for the reduced form (blue). The other intermediate cross-peak (red) was never observed before and disappeared during the course of the reaction (Figs 7A and S3).
If we plot the intensities of the cross-peaks of reduced and oxidized Thr11 and the newly formed intermediate in function of time (Fig. 7B), we see that the intensity of intermediate cross-peak initially increases, but gradually decreases as more of the reduced form is formed. We hypothesized that this newly identified intermediate cross-peak corresponds to Thr11 in a population where Mrx1 forms a mixed disulphide with MSH. To verify this hypothesis, we titrated twenty molar excess of MSH into a sample of 15N labelled oxidized Mrx1 (Figs 7C and S3). The intermediate cross-peak appeared at exactly the same position as after 20 min in the reaction with MSH, Mtr and NADPH, confirming our hypothesis.
We next used mass spectrometry to prove that this intermediate is a mixed disulphide between MSH and Mrx1. First, we identified the formation of a mixed disulphide between MSH and the thiol of a synthetic peptide FLRTSCG. The peptide (50 μM) was incubated at 37°C for 15 min with an equimolar amount or a 10-fold excess of MSSM. The ESI-MS spectrum of the reaction products showed the presence of a dominant peak at m/z 783.4 (z = 1) corresponding to the unreacted peptide together with a peak at m/z 1267.4 (z = 1), corresponding to a mass shift of 484 Da that corresponds to the addition of MSH to the peptide (Fig. S4). The peptide was only 8% and 42% S-mycothiolated depending on the molar ratios of peptide : MSSM (1:1 and 1:10). Upon collision-induced dissociation fragmentation in the gas phase the m/z 1267.4 precursor ion showed two major neutral losses of 180 Da and 452 Da corresponding to the loss of inositol or AcDehydroalanine-GlcN-Ins respectively (Fig. S4). Similarly, loss of the malate group on the structurally similar bacillithiol has previously been reported (Chi et al., 2011). Subsequent fragmentation of the two major daughter ions in MS3 confirmed the localization of MSH on the cysteine residue of the peptide.
To trap Mrx1 as a mixed disulphide in complex with MSH, we incubated the Mrx1 C17A mutant for 5 min with MSSM at room temperature. The reaction was stopped with an excess of N-ethylmaleimide (NEM) to block all remaining free thiols. After digestion with Endo Lys-C, the peptides were analysed by LC-MS/MS to identify the mixed disulphide modification. The average sequence coverage was 92% and the N-terminal peptide that contains the nucleophilic cysteine (MVTAALTIYTTSWCGYALRLK) was found with and without the N-terminal methionine at percentages of 55% and 45%, respectively, as estimated by extracted ion chromatograms (XIC) and peaks area integration. In control conditions without MSSM, the Cys was found completely modified by NEM. In the presence of MSSM, the mass of the N-terminal peptide was found to be 484 Da higher, consistent with the covalent attachment of MSH. A precursor ion of m/z 1423.7 (z = 2) corresponding to a mixed disulphide between the N-terminal peptide and MSH was further analysed in MS/MS (Fig. 7D). As for the synthetic peptide, the fragmentation spectra were dominated by two major neutral losses of 180 Da and 452 Da, which confirms the presence of MSH. Some fragmentation also took place at the peptide bonds to produce b- and y- series of ions, which could unambiguously identify the Cys as S-mycothiolated. Quantification of the extent of modification was done by peak area integration of the extracted ion chromatograms (XIC) for each molecular species and found to be equal to 17% and 52% depending on the molar ratio between the protein and MSSM (1:1 or 1:10).
Altogether, the mass spectrometric and HSQC analyses indicate the formation of a mixed disulphide between MSH and Mrx1 as a reaction intermediate during the reduction of oxidized Mrx1 by MSH (Fig. 7E). MSH nucleophilically attacks the disulphide of oxidized Mrx1 forming a mixed disulphide between MSH and Mrx1. A second molecule of MSH reduces the mixed disulphide releasing reduced Mrx1 and MSSM.
Oxidized and reduced Mrx1 are energetically equivalent and the redox potential is −218 mV
Having shown that Mrx1 reduces disulphide bonds (Fig. 2), we wanted to understand the thermodynamic driving force for this reaction. Like for the reductases Trx and Grx, the reaction is a bimolecular nucleophilic substitution reaction (SN2) in which the difference in stability between the oxidized and reduced state may be correlated with the energetics of the reduction reaction (Collet and Messens, 2010). We measured the thermal unfolding of oxidized and reduced Mrx1 using circular dichroism (Fig. 8). The unfolding curves of both oxidized and reduced Mrx1 are similar and we obtained an identical midpoint at ∼ 80°C. For chemical unfolding, we needed to increase the temperature to 50°C to obtain fully unfolded Mrx1 at 8 M urea (Fig. 8). Urea-induced unfolding at 50°C resulted in a midpoint of unfolding at 4.5 M urea. Both thermal and urea-induced unfolding are irreversible. Mrx1 is highly stable under physiological conditions. Oxidized and reduced Mrx1 are energetically equivalent, which was also observed for surface exposed oxidase DsbA of S. aureus (Heras et al., 2008).
Another way to predict whether a reaction is thermodynamically favourable is by measuring the redox potential, which is proportional to the Gibbs-free energy of a reaction. We measured the redox potential of Mrx1 in equilibrium with GSH. From the plot of the reduced fraction of Mrx1, obtained from elution peak integration on reverse phase chromatography, versus the redox potential of the GSH buffer solution, we obtained a midpoint potential of −218 mV (Fig. 8C). The redox potential of Mrx1 is in the same range as the redox potentials reported for glutaredoxins: E. coli Grx1 has a redox potential of −233 mV, and the redox potential of E. coli Grx3 is −198 mV (Aslund et al., 1997). Energetically, it will be favourable for Mrx1 to reduce disulphide bonds with a redox potential higher than −218 mV.
The pKa of Cys14 is fine-tuned by hydrogen bonding with Cys17
Among many other factors (Jensen et al., 2008), the rates of the thiol-disulphide exchange reactions are strongly influenced by the pKa of the nucleophilic cysteine involved in the reaction (Shaked et al., 1980). Due to the local electrostatic environment of nearby residues of the nucleophilic cysteine in the CXXC motif of oxidoreductases (Bulaj et al., 1998; Hansen et al., 2005), the pKa of the N-terminal cysteine is relatively low compared with the pKa value of 8.6 for a cysteine (Kallis and Holmgren, 1980; Dyson et al., 1991; Kortemme and Creighton, 1995; Thurlkill et al., 2006; Lillig et al., 2008). This allows the N-terminal cysteine to perform a nucleophilic attack on the substrate disulphide (Lillig et al., 2008). We determined the pKa of the two cysteines in Mrx1 by recording the absorption at 240 nm during a pH titration (Roos et al., 2012). For the N-terminal cysteine (Cys14), we measured a pKa of 6.8, whereas the pKa of the C-terminal cysteine (Cys17) was 11.1 (Fig. 9A). The structural environment of Cys14 shows hydrogen bonds of the sulphur atom of Cys14 with several neighbouring residues. Cys14 hydrogen bonds to the backbone amide of Gly15 and the backbone amide of Tyr16 in 17 of the 20 lowest energy structures of reduced Mrx1. Additionally, in 10 structures, a hydrogen bond is formed between Cys14 and the backbone amide of Cys17. These hydrogen bonds stabilize Cys14 as a thiolate, lowering its pKa as seen for Trx and Grx (Bulaj et al., 1998; Foloppe et al., 2001; Foloppe and Nilsson, 2004; Hansen et al., 2005). Remarkably, the pKa of Cys14 is almost two units higher than the pKa of the N-terminal cysteine of Grxs, which is typically lower than 5.
To better understand the structure, dynamics and electrostatics of the active site of Mrx1, including the stabilization of its thiolate, four extensive molecular dynamics (MD) simulations of the protein in explicit solvent were performed (225 nanoseconds each) with Cys14 treated as a thiolate (Foloppe et al., 2001; Foloppe and Nilsson, 2004). Such simulations provide further insights into the organization of the CGYC active site motif (Foloppe and Nilsson, 2004). The starting NMR conformers presented various conformations and interactions for their active site-reduced cysteines (Table S2). The simulations provided a more detailed and precise picture of the structural dynamics of the Mrx1 active site, relevant to the chemistry of Mrx1, such as the pKa of Cys14.
The four simulations of Mrx1 converged on a clearly dominant trans orientation for Cys14 (Fig. S5 and Table S3), despite different starting conformations (Table S2). This trans conformation positions the Cys14 thiolate sulphur (Cys14_Sγ) such that it frequently hydrogen-bonds the backbone amide groups of Tyr16 and Cys17 (Fig. 9 and Table S3). Using the correlation between the pKa and the charge on the sulphur determined by Natural population analysis (NPA), we calculated the pKa of Cys14 to be in the range of 6.1 to 7.6 when Cys17 is in trans conformation, depending on the snapshot. However, Cys14_Sγ can also be hydrogen-bonded by the thiol of Cys17, depending on the conformation of Cys17. Cys17 populated both the trans and gauche− (g−) conformations during the simulations (Table S3 and Fig. S5), with several transitions between these conformers. This is important, since the conformation of the C-terminal cysteine (Cys17) controls whether it can hydrogen bond the thiolate of the N-terminal cysteine (Cys14). Such a hydrogen bond can only be formed when the C-terminal cysteine is g−. In this conformation, the calculated pKa of Cys14 is in the range of 5.1 to 5.6. To test the role of Cys17 on the pKa of Cys14, we tried to measure the pKa of the Mrx1 C17A mutant. Unfortunately, this mutant was not stable over a broad enough pH range. Instead, starting from the model with Cys17 in g− (Cys14–Cys17 hydrogen bond) we in silico mutated Cys17 to Ala. With this model, we could calculate the pKa of Cys14. We found an increase in the pKa of Cys14 of about one unit, which confirms the importance of the hydrogen bond between both active site cysteines.
For comparison, in MD simulations of Grx, the C-terminal cysteine overwhelmingly populated the g− conformer, contributing to lower the thiolate pKa further (Foloppe et al., 2001; Foloppe and Nilsson, 2004; 2007). Interestingly, in Mrx1 Cys17 populated both g− and trans corresponding to the lower calculated pKa values (5.1–5.6) and the higher Cys14 pKa values (6.1–7.6) respectively. Thus, the stabilization of the C-terminal Cys17 trans conformation in Mrx1 compared with Grx can rationalize why the measured Cys14 pKa in Mrx1 is higher than its counterpart in Grx. This suggests that the dynamics of C-terminal cysteine can provide a mechanism to fine-tune the pKa of Cys14, and thus the redox properties of Mrx1.
The mechanisms by which M. tuberculosis detoxifies ROI and RNI are of particular interest because this knowledge will help us to understand both pathogenesis and persistence, which in turn leads to latent infection. M. tuberculosis engages a diverse variety of defence systems against oxidative stress (Manca et al., 1999; Piddington et al., 2001; Bryk et al., 2002; Li et al., 2005). In all these detoxification mechanisms electrons are needed. These electrons are provided by the general electron fuel of the cell originating from NADPH, and are transferred to oxidized substrates and protein disulphides. We determined an electron transfer pathway involving Mrx1 (Fig. 10). Mrx1 belongs to a new class of Grx-like proteins exclusively coupled to MSH. The reduction mechanism of oxidized Mrx1 proceeds through the formation of a mixed disulphide between Mrx1 and MSH (Figs 7 and 10). A second MSH molecule breaks the mixed disulphide releasing reduced Mrx1 and MSSM. A similar reduction mechanism is reported for GSH to reduce dithiol Grx (Fernandes and Holmgren, 2004).
Mrx1 has only 34% sequence similarity with E. coli Grx1, and is characterized by a CGYC active site sequence motif. Mrx1 has a typical Trx-fold that consists a four stranded β-sheet surrounded by three α-helices. The active site CGYC motif is located at the N-terminus of the α1 helix. Opposite to the active site motif, Pro57 is present in cis conformation. This cis-proline is conserved among all proteins with a Trx-fold and plays a role in maintaining the conformation of the active site (Gleason, 1992; Collet and Messens, 2010).
Interestingly, the C-terminus of Mrx1 structurally resembles Trx more closely than Grx. A long linker between strand β4 and helix α3 is present in Mrx1, orienting helix α3 almost antiparallel to strand β4. In Trx and NrdH (Stehr and Lindqvist, 2004), this conformation allows the binding of TrxR (Lennon et al., 2000; Stehr et al., 2001). Grx are characterized by a short linker placing helix α3 perpendicular to strand β4 causing steric hindrance for TrxR binding. Despite this structural similarity with Trx, oxidized Mrx1 is not recognized and reduced by TrxR. A hydrophobic pocket close to the CXXC motif needed for the binding of TrxR is not present in Mrx1 (Stehr et al., 2001).
The long linker (T68 till A72) between strand β4 and helix α3 forms a conserved patch connected to Cys14 via a conserved groove (T55V56P57T58) harbouring cis-Pro57 (Fig. 6). In E. coli Grx3, the sequence motif T51V52P53 located in a similar groove interacts with the peptide bond between Cys and γ-Glu of GSH (Nordstrand et al., 1999). We speculate that this groove contains the MSH binding site.
Grx are known to reduce a wide range of target proteins involved in several crucial cellular functions including the protection against oxidative stress (Rouhier et al., 2005; Li et al., 2007). Therefore, it is highly likely that Mrx1 with a redox potential (-218 mV) in the same range as Grx (Aslund et al., 1997) will also be involved in many cellular reduction pathways. We previously showed that Mrx1 of C. glutamicum provides the electrons to reduce an arsenate-mycothiol adduct (Ordóñez et al., 2009). In M. tuberculosis, no MSH targets of Mrx1 have been identified so far, but we demonstrate here reduction of mycothiol-coupled HED. In addition, we show that Mrx-1 does not reduce mixed disulphides with GSH. Further specificity is also observed for the reduction of oxidized Mrx1, which kinetically is in favour of receiving electrons from MSH.
The first step of the disulphide reduction mechanism of Grx is a nucleophilic attack of the N-terminal cysteine of Grx on its target disulphide (Fernandes and Holmgren, 2004). Therefore, the N-terminal cysteine of Grx is predominately present as thiolate under physiological conditions. Using a combined approach of experimentally determined pKas and MD followed by NPA charge analysis, we could show that in Mrx1 the pKa of Cys14 depends on the conformation of Cys17. The gauche− conformation of Cys17 decreases the pKa of Cys14 via a thiol-thiolate hydrogen-bond, while the trans conformation of Cys17 disrupts this hydrogen bond and results in a higher pKa for Cys14. In the solution structure of reduced Mrx1, Cys17 significantly populates the trans conformation, resulting in calculated pKa values of Cys14 between 6.1 and 7.6. Experimentally, a pKa of 6.8 was obtained for Cys14 supporting the computational calculation of the pKa. MD simulations show that Cys17 can also adopt a gauche− conformation and form a hydrogen bond with Cys14. This extra hydrogen bond lowers the pKa of Cys14 to 5.1–5.6. Thus, the conformation of Cys17 is an important factor in regulating the pKa of Cys14. This low pKa allows Cys14 to function as a nucleophile in the first step of the reaction mechanism in the same way as the N-terminal cysteine of Grx.
The mrx1 deletion mutant of M. smegmatis shows increased, but limited sensitivity to hydrogen peroxide, Cu2+ and iodoacetamide (IAM). The rather limited sensitivity might be due to compensatory upregulation of defensive enzymes in the mrx1 deletion mutant. Alternatively, the function of Mrx1 might become especially important in the period after oxidative stress to reactivate mycothiol protected proteins.
The pathway of MSH biosynthesis has been established, and the enzymes involved include a glycosyltransferase (MshA), a phosphatase (MshA2), a deacetylase (MshB), an ATP-dependent ligase (MshC) and an acetyltransferase (MshD) (Newton et al., 2008). In the absence of MSH (mshA− strain) (Fig. 1), the sensitivity to oxidative stress inducing compounds is much stronger since the protection by MSH is not possible. It has been shown that mshA deletion mutants require catalase to grow and are resistant to the pro-drug ethionamide (Vilchèze et al., 2008). The stronger sensitivity of a mshA− mutant strain indicates that peroxidases or peroxiredoxins might need MSH to protect Mycobacterium spp. against ROI and RNI (Newton et al., 1999). We should note that all mutants lacking MSH in both M. smegmatis and M. tuberculosis grow slower than the wild type when starting form a frozen stock (Newton et al., 2008). Our Δmrx1 strain in M. smegmatis does not display this lag in growth as observed for the MSH-deficient strains.
Similar to the Δmrx1 strain, increased sensitivity to oxidative stress has been reported for deletion mutants of the grx genes in E. coli and yeast (Collinson et al., 2002; Fernandes and Holmgren, 2004). The yeast Grxs are directly linked to hydrogen peroxide by their peroxidase activity (Collinson et al., 2002). Mrx1 does not directly eliminate hydrogen peroxide from the cell, but may be linked to the oxidative stress response in an indirect way by keeping essential protein cysteines in the reduced state. We showed that Mrx1 specifically reduces MSH mixed disulphides. Protein S-mycothiolation might be a reversible post-translational thiol-modification to protect vulnerable cysteines from overoxidation or to control redox sensing regulators in response to oxidative stress in Mycobacteria, such as reported for S-glutathionylations in eukaryotes and S-bacillithiolations in Bacillus subtilis (Rokutan et al., 1991; Chai et al., 1994; Lee et al., 2007; Chi et al., 2011). Mrx1 might then rescue the protected cysteines by reducing the mixed disulphide (Fig. 10), like observed for Grx (Seres et al., 1996). Grx are also known to reduce a vast variety of proteins linked to oxidative stress, including but not limited to, the transcription factor OxyR, which activates the expression of several antioxidant defence genes in response to increased hydrogen peroxide concentrations (Zheng et al., 1998). M. tuberculosis contains no OxyR but has a related thiol-based LysR-type regulator OxyS that controls the catalase gene and has redox-sensitive Cys residues that could require Mrx1 for reduction (Li and He, 2012).
In Streptomyces coelicolor, another actinomycete, a sigma/antisigma pair SigR/RsrA controls the response to thiol-oxidative stress (Kim et al., 2012). RsrA is a zinc-binding, redox-sensitive anti-sigma factor that sequesters SigR under reducing conditions (Antelmann and Helmann, 2011). Under disulphide stress conditions, which are physiologically different from peroxide stress, an intramolecular disulphide-bond is formed in RsrA, resulting in the release of Zn(II) and active SigR. Free SigR induces in a negative feedback loop several thiol-disulphide reducing systems to restore the thiol redox balance. The Mrx1/MSH system might also be involved in this negative feedback, which makes RsrA a possible candidate substrate for the Mrx1/MSH reducing system.
In conclusion, Mrx1 is a novel member of the Trx-family, which uses the MSH electron transfer pathway to reduce disulphides in M. tuberculosis (Fig. 10), helping this pathogen to survive oxidative stress. Our findings open new avenues in our general understanding of the oxidative stress defence mechanisms of Mycobacteria.
The mrx1 and mshA− knockout strains in Mycobacterium smegmatis
MSMEG_1947 (mrx1) and flanking DNA was amplified from M. smegmatis mc2155 genomic DNA using primers 5′-GATGTCGGCCTGCGTCTTCT-3′ and 5′-AGGTGCTCATCGCCAACGGT-3′ and cloned into vector pcr2.1 (Invitrogen). The hygromycin resistance cassette was excised from pucHyg using Sma1 and inserted into the Pml1 site within MSMEG_1947 to obtain the suicide vector. The suicide vector was electroporated into competent M. smegmatis cells and plated on Middlebrook 7H10 medium with hygromycin (75 μg ml−1). All colonies that arose on this medium were replica plated on Middlebrook 7H10 medium with kanamycin. Putative knockout mutants did not grow on kanamycin but grew on hygromycin. Genomic DNA was extracted from these mutants followed by PCR amplification to confirm disruption. One mutant, ΔMrx1, was investigated further. The mshA− transposon mutant of Mycobacterium smegmatis has been described (Newton et al., 2003).
Disk diffusion assay
Disk diffusion assays of wild type, mshA− and Δmrx1 were performed as described (Rawat et al., 2002). The oxidants hydrogen peroxide (0.5 μmol), menadione (10 pmol), cumene hydroperoxide (0.2 μmol), diamide (0.2 μmol), sodium hypochlorite (5 μmol), 2,4-dinitrochlorobenzene (10 nmol), the metal Cu2+ (10 nmol CuSO4) and the alkylating agent iodoacetamide (0.1 μmol) were spotted on a 5 mm diameter sterile paper disk and placed on a Middelbrook 7H9 agar plates streaked with wild type, mshA− or Δmrx1. The plates were incubated at 37°C for 3 days. The zone of bacterial clearance around each paper disk was measured and the results of the different knockouts were analysed using one-way anova test (comparing all knockouts) or the student t-test (comparing one knockout to one other).
Purification of mycothiol
MSH was purified as described (Ordóñez et al., 2009).
The TrxR, GR and Mtr electron transfer assays
Mrx1 was oxidized using a 10-molar excess of diamide for 30 min at room temperature and purified on a superdex75HR column equilibrated with 50 mM HEPES/NaOH pH 8.0, 150 mM NaCl to obtain a pure sample. An aliquot of the 80% ammonium sulphate precipitation of Mtr or TrxR was pelleted by centrifugation and resuspended in 50 mM HEPES pH 8.0. Ammonium sulphate was removed by dialysis against 50 mM HEPES pH 8.0.
The assay mixtures of Mtr was prepared by diluting all components, excepts oxidized Mrx1 in the 50 mM HEPES/NaOH pH 8.0 buffer to a final concentration of 0.5 μM Mtr, 500 μM MSH and 500 μM NADPH. The TrxR assay mixture consists of 0.5 μM TrxR and 500 μM NADPH. The GR assay mixture consists of 0.5 μM GR, 500 μM GSH and 500 μM NADPH. All concentrations were calculated taking into account the subsequent addition of oxidized Mrx1. The mixtures were incubated for 20 min at 37°C in a 96-well plate. The assay was started by adding 50 μM of oxidized Mrx1 in each well. The absorption at 340 nm was monitored. For each reaction mixture a control well without oxidized Mrx1 was added.
To assess the specificity of oxidized Mrx1 for the GSH and MSH, the GR and Mtr electron transfer assay was reconstructed as described and several concentrations of GSH (200, 100 μM) and MSH (200, 100, 50 μM) were compared.
The Mrx1 peroxidase activity, FOX and HED assay
A mixture of 0.5 μM Mrx1, 500 μM MSH, 3 μM Mtr and 500 μM NADPH was prepared. In the control mixture, Mrx1 was absent. The mixture was pre-incubated for 20 min at 37°C before the different concentrations of H2O2 were added. The consumption of NADPH was monitored at 340 nm. To estimate the reaction rate, 10 μM and 5 μM H2O2 were used as substrate. Under these conditions (substrate concentration << KM), the reaction rate is proportional to the initial velocity divided by the substrate (H2O2) and the initial enzyme (MSH) concentration. The assay was performed in duplicate.
The Ferrous Oxidation of Xylenol orange (FOX) assay is performed as described (Wolff, 1994).
The HED assay for GSH (Holmgren, 1979) was modified for MSH. The mixed disulphide between MSH or GSH and 2-HED was formed by incubating 700 μM HED with 1 mM MSH or GSH, respectively, at 30°C for 3 min. This substrate (250 μM final concentration) was added to the electron transfer mixture with a final concentration 0.5 μM Mtr, 250 μM NADPH and 2 μM Mrx1. As control, Mrx1 was omitted of the mixture. The assay was performed at 25°C and the absorption monitored at 340 nm.
The competition assay with N-acetyl glucosamine and myo-inositol
The Mtr electron transfer assay was prepared as described above. A reaction mixture with a final concentration (taking in account the addition of the competitor and Mrx1) of 0.5 μM Mtr, 100 μM MSH and 500 μM NADPH was prepared. N-acetyl glucosamine (100 and 500 μM) and myo-inositol (100 and 500 μM) were added as competitor of MSH. Glucose (100 and 500 μM) was used as a negative control. The mixture was incubated for 10 min at 37°C. Oxidized Mrx1 (final concentration of 500 μM) was added, and the consumption of NADPH was recorded for 20 min at 340 nm. The assay was performed in duplicate.
The redox potential, stability of Mrx1 and the pKa of the active site cysteines
The redox potential and the stability was determined as described (Roos et al., 2007). The pKa of the nucleophilic cysteine was determined spectrophotometrically (Roos et al., 2007). To measure the pKa of the C-terminal cysteine the poly-buffer was changed to a buffer consisting of 10 mM Tris, 10 mM sodium phosphate, 10 mM sodium borate and 10 mM CAPS pH 13.0.
The detection of the mixed disulphide between Mrx1 and MSH
Oxidized 15N labelled Mrx1 was prepared as described in a buffer solution of 50 mM sodium phosphate pH 6.5, 150 mM NaCl. Mtr was dialysed to the same buffer. MSH, Mtr and NADPH were pre-mixed and oxidized 15N Mrx1 was added to obtain final concentrations of 265 μM Mrx1, 998 μM MSH, 2.4 μM Mtr and 1 mM NADPH. 15N-HSQC spectra were recorded every 30 min.
In a second experiment MSH was titrated stepwise into a 300 μM oxidized 15N 13C Mrx1 sample till a 20 molar excess of MSH was reached. 15N-HSQC spectra were recorded.
Expression of single labelled 15N and 13C, and double labelled 15N, 13C Mrx1
A preculture in 2× TY medium supplemented with 50 μg ml−1 kanamycin and 1% glucose was grown overnight at 37°C. One litre of medium enriched with 1 g l−115N ammonium chloride and 4 g l−113C glucose was inoculated with 20 ml preculture and grown to a cell density of 0.7. The cells were induced with 1 mM IPTG and grown overnight at 25°C. Double and single labelled Mrx1 were purified as described.
Nuclear magnetic resonance assignment and structure determination
Double labelled (15N and 13C) Mrx1 was prepared for NMR experiments at 1 mM in 20 mM sodium phosphate buffer at pH 6.6, 9% D2O, in the presence or absence of 2 mM DTT for the reduced and oxidized form respectively. All NMR spectra were acquired at 25°C on a Varian NMR Direct-Drive Systems 600 or 800 MHz spectrometers. 2D 15N-HSQC and 13C-HSQC, 3D 15N- and 13C-NOESY-HSQC (mixing time of 125 ms) and triple resonance spectra CBCA(CO)NH, HNCACB, HNCO, HBHA(CO)NH, C(CO)NH and HCCH-TOCSY (Sattler et al., 1999) and aromatic (HB)CB(CGCD)HD and (HB)CB(CGCDCE)HE (Yamazaki et al., 1993) were recorded on the 13C,15N-labelled samples. All NMR data were processed using NMRPipe (Delaglio et al., 1995) and analysed by NMRView (Johnson, 2004) and CCPNMR (Vranken et al., 2005). Backbone and side-chain assignments were obtained using NMRView as described previously (Zorzini et al., 2011) and verified from 3D 15N- and 13C-NOESY-HSQC spectra using CCPNMR.
NOE cross-peaks for reduced and oxidized Mrx1 were extracted from the corresponding 3D 15N- and 13C-NOESY-HSQC and 2D NOESY spectra. These NOEs and dihedral restraints from Talos+ (Shen et al., 2009) were used as input for the structure calculations using CYANA version 2.1. NOEs were assigned using the automated NOE assignment procedure of CYANA (Guntert et al., 1997; Herrmann et al., 2002) followed by a final structure refinement in explicit solvent using the RECOORD protocol (Nederveen et al., 2005), which runs under CNS (Brunger et al., 1998). The 20 lowest-energy structures of each set were used for the final analysis.
J.M. and N.V.N. are group leaders of the VIB. D.V. is Collaborateur logistique at FNRS-FRS. We would like to thank Almudena Villadangos for purifying mycothiol and for the cloning of Mtr and Mrx1; Katherine Barretto and K.P. Papavinasasundaram for constructing of the M. smegmatis ΔMrx1 strain and Aleksandra Dziewulska for experimental assistance.
This work was made possible thanks to financial support from the following institutions (i) agentschap voor innovatie door Wetenschap en technologie (IWT), (ii) Vlaams Instituut voor Biotechnologie (VIB), (iii) the HOA-project of the Vrije Universiteit Brussel (VUB), (iv) National Institutes of Health grant GM061223 (MR) and (v) the Swedish Research Council (LN). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. KVL thanks the IWT for a PhD fellowship. L.B. and G.R. thank the FWO for postdoctoral fellowship.