|Home | About | Journals | Submit | Contact Us | Français|
The NADPH-dependent homodimeric flavoenzyme thioredoxin reductase (TrxR) provides reducing equivalents to thioredoxin, a key regulator of various cellular redox processes. Crystal structures of photo-inactivated thioredoxin reductase (TrxR) from the Gram-positive bacterium Lactococcus lactis have been determined. These structures reveal novel molecular features that provide further insight into the mechanisms behind the sensitivity of this enzyme toward visible light. We propose that a pocket on the si-face of the isoalloxazine ring accommodates oxygen that reacts with photo-excited FAD generating superoxide and a flavin radical that oxidize the isoalloxazine ring C7α methyl group and a nearby tyrosine residue. This tyrosine and key residues surrounding the oxygen pocket are conserved in enzymes from related bacteria, including pathogens such as Staphylococcus aureus. Photo-sensitivity may thus be a widespread feature among bacterial TrxR with the described characteristics, which affords applications in clinical photo-therapy of drug-resistant bacteria.
The protein disulfide reductase thioredoxin (Trx) maintains thiol groups in a reduced form through thiol-disulfide exchange reactions involving a redox-active WCXXC motif. Trx also provides reducing equivalents to enzymes like ribonucleotide reductase, peroxiredoxins and methionine sulfoxide reductase1. The recycling of Trx is catalysed by the homodimeric flavoprotein thioredoxin reductase (TrxR) that utilizes electrons from NADPH which are shuttled through a tightly bound FAD co-enzyme and a redox active dithiol motif. The TrxR monomers in bacteria, fungi and plants are about 35kDa and composed of an NADPH- and an FAD-binding domain. By contrast, the monomers of TrxR from mammals, dipteran insects and protozoan parasites are larger (ca 50–55kDa) and harbor an interface domain and a C-terminal flexible extension with an additional redox-active motif. In mammalian enzymes this motif contains a selenocysteine residue, while in e.g. Drosophila melanogaster this is replaced by a cysteine apparently without altering the catalytic efficiency2. Based on pioneering work on the three-dimensional structure and biophysical characteristics of Escherichia coli TrxR (EcTrxR) it was proposed that this bacterial enzyme undergoes a major molecular re-arrangement from the so-called Flavin Oxidizing (FO) conformation to the Flavin Reducing (FR) conformation in order to complete a catalytic cycle3,4,5. The FR conformation structure of EcTrxR in complex with Trx indeed confirmed a 66° rotation of the NADPH domain relative to the FAD domain, which relocates NADPH for hydride transfer to FAD and positions the TrxR CXXC motif in proximity of Trx6. Mammalian TrxR does not undergo such a conformational change during catalysis and the reducing equivalents delivered on the re-face of FAD are instead directly transferred to a redox-active CVNVGC motif positioned on the si-face7,8. Subsequently, the electrons are transferred to the C-terminal redox center, in turn reducing the CXXC motif in Trx. These mechanistic differences make TrxR an attractive target for antimicrobial drugs and a range of inhibitors have been explored for this purpose9,10.
The physiological importance of Trx and TrxR are highly dependent on the lifestyle of the organisms as well as access to alternative redox control systems and antioxidant enzymes. For example in E. coli and other glutathione-producing bacteria, Trx is complemented by glutaredoxin as an electron donor to ribonucleotide reductase, whereas glutathione-negative bacteria such as Bacillus subtilis and Staphylococcus aureus lack this pathway11,12. The function of Trx as an electron donor to peroxiredoxins is particularly important in catalase-negative bacteria that rely on thiol-dependent peroxidases to remove damaging peroxides. Many lactic acid bacteria (LAB) lack both glutathione and catalase, suggesting that Trx plays a particularly central role in protection against oxidative stress in these bacteria. Lactococcus lactis, an industrially important LAB model bacterium, contains a TrxR that recycles two Trx (TrxA and TrxD) as well as NrdH, a glutaredoxin-like protein providing electrons to ribonucleotide reductase class Ib13. Additionally, it was recently discovered that reducing equivalents for ribonucleotide reduction can be provided in a TrxR-independent manner through flavoredoxin14. This may explain why knock-out mutants lacking either TrxR or Trx are viable even under oxidative stress conditions15. While investigating the biochemical properties of L. lactis TrxR (LlTrxR) we discovered that the enzyme is susceptible to photo-inactivation by visible light in an oxygen-dependent manner16. This inactivation coincided with a shift in the absorbance spectrum of the tightly bound FAD co-enzyme and oxidation of an isoalloxazine methyl group. To characterize the consequences of this photo-oxidation process in greater detail the three-dimensional structure of inactivated LlTrxR is here solved by X-ray crystallography. We propose possible mechanisms that account for the observed oxidative damage of the enzyme.
We have previously shown that recombinant TrxR from L. lactis is inactivated by visible light in vitro, while the corresponding enzyme from E. coli is more resistant to photo-inactivation under these conditions16. In order to evaluate the sensitivity of native TrxR in an environment resembling the conditions in vivo, cell extracts of mid-late exponential phase cultures of L. lactis and E. coli K-12 were subjected to irradiation over a period of 12h and TrxR activity was measured at different time points by use of a coupled assay with Trx, applying DTNB as the final electron acceptor. A marked decrease in TrxR activity was observed in the L. lactis cell extracts over the course of irradiation, while the activity in the E. coli cell extract was essentially unchanged (Fig. 1). After 12h a ~65% drop in activity was observed in L. lactis cell extract. This apparent slower rate of inactivation of native LlTrxR compared to the recombinant enzyme may be due to light-quenching by endogenous chromophores in the cell extracts.
The overall structure of LlTrxR is similar to other low molecular weight (LMW) TrxRs3,17 with a dimeric organization in which each monomer is composed of NADPH and FAD binding domains (Fig. 2). The secondary structure of each subunit consists of 11 α-helices and 19 β-sheets. The FAD-domain consists of residues 1–112 and 243–308 and the NADPH-domain consists of residues 116–239. The LlTrxR structures were obtained mainly in the FR conformation (Fig. 2a,c), which was unexpected since all available structures of LMW TrxR from other organisms are in the FO conformation, except for the engineered complex between EcTrxR and EcTrx6. LlTrxR crystallized in the FO conformation (Fig. 2b,d) only under reducing conditions in the presence of DTT (Table 1). The propensity of LlTrxR to crystallize in the FR conformation may indicate that the enzyme is stabilized in this conformation. Superposition of LlTrxR Cα-atoms with the structure of EcTrxR3,6, gives RMSD of 0.426Å and 1.137Å for the enzymes in FR (Fig. 2a,c) and FO conformation (Fig. 2b,d), respectively. The relative orientation of the structures thus matches perfectly with the E. coli enzyme and consolidates the catalytic model proposed by Waksman et al. (1994) where the NADPH domain rotates 66° relative to the FAD domain in order to switch between the FO and FR conformations in the catalytic cycle.
No apparent change in the overall structure of LlTrxR in the FR conformation was observed upon exposure to visible light. Close examination of the isoalloxazine ring however reveals increased electron densities around the C7α methyl group over the course of irradiation (Fig. 3a–f), even though the FAD co-enzyme in general has well-defined electron densities with B-factors typically in the range from 20–24Å2. This observation is in accordance with our previous mass spectrometry data demonstrating formation of an aldehyde group in FAD extracted from light-inactivated LlTrxR, which was concluded to be confined to the C7α methyl group based on the spectrophotometric features16. Moreover, an increase in electron density at the meta-position (ortho to the hydroxyl group) of Tyr237, located 4.4Å from C7α is observed as a function of light-exposure (Fig. 3a–f). This density most likely represents the formation of 3,4-dihydroxyphenylalanine (DOPA). Mass spectrometric analysis of proteolytic digests of photo-inactivated LlTxR indeed confirms a mass shift of 16Da confined to Tyr237 (Supplementary Figure 1). Tyr237 is only close to the isoalloxazine ring when LlTrxR is in the FR conformation (Fig. 3a), suggesting that photo-oxidation reactions are primarily taking place while the enzyme is in this conformation. In the FO conformation, Tyr237 is distant from the flavin-binding site and another tyrosine (Tyr133) is in proximity to the isoalloxazine ring (4.9Å from C7α), but no indications of oxidation of Tyr133 were observed in the irradiated structures of LlTrxR. Remarkably, Tyr237 appears to be conserved among most TrxR from Gram-positive bacteria in the phylum Firmicutes, but is replaced mainly by Ala or Phe in other organisms (Supplementary Figure 2).
The observed oxygen- and triplet isoalloxazine-dependent photo-inactivation of LlTrxR16 suggests that molecular oxygen is positioned in the vicinity of the FAD and interacts with the photo-excited co-enzyme. A closer look at the isoalloxazine ring indeed reveals a pocket that potentially may accommodate molecular oxygen due to its hydrophobic character. Met43 positioned on the si-face bends away from the isoalloxazine towards Pro15 and the sidechains of Met18 and Met67 (Fig. 4a) and as a consequence a pocket on the si-face is formed. In the space between the β-carbon of Met43 and the isoalloxazine ring (7.2–7.4Å distance to N10 and N5) a well-defined electron density appears, assigned to a water molecule perfectly centered above the central ring of the isoalloxazine. This water molecule is present in all structures listed in Table 1. The water is in hydrogen bond distance from the 2′-ribityl hydroxyl group (2.8Å) and the alcohol group of Thr46 (2.8Å), and the distance from the water to the N5 and N10 atom of the isoalloxazine is 3.5–3.6Å (Fig. 4a). A closer look at the electrostatic environment on the si-face reveals several hydrophobic residues, most noteworthy are Ile286, Ile49, Leu63 and the β- and γ-carbons of Met43 that together with the π-electron cloud of the conjugated isoalloxazine ring system constitute an overall hydrophobic environment facilitating accommodation of O2 (Fig. 4c). In the presence of O2 at this location, it can be speculated that the Thr46 sidechain can flip so the methyl group points towards the O2, while the 2′-ribityl hydroxyl group can form a hydrogen bond with the 4′-ribityl hydroxyl group. These changes will further increase the hydrophobicity of the si-face, thus increasing the O2 affinity. The re-face of the isoalloxazine ring in the FO conformation is occupied by the Cys134-Cys137 disulfide while in the FR conformation it is surrounded by hydrophilic residues (Asp154, Glu158, Ser155 and Gln285). These features further underlines that the si-face open space is the most likely site of O2 accommodation and thus oxidant generation. The oxygen pocket described here is conserved in Staphylococcus aureus TrxR (SaTrxR), the most closely related structure-determined enzyme (53% sequence identity).
Even though several residues on the si-face of the isoxoalloxazine ring are conserved among TrxR (Fig. 4e) some subtle yet crucial differences in the local environment are observed. Met43LlTrxR appears to be conserved among bacteria in the phylum Firmicutes (Supplementary Figure 2), but in EcTrxR this residue is substituted with Leu44EcTrxR that is centered at the si-face of the isoalloxazine with a 3.8–4.0Å distance from the δ-carbon to the N10 and N5 atoms (Fig. 4b). In addition, Ser64LlTrxR is replaced by the more bulky Met66EcTrxR that hinders Leu44EcTrxR from bending towards Pro15. Leu44EcTrxR thus blocks the pocket on the si-face of the isoalloxazine and restricts access of O2 (Fig. 4d). Noticeably, all available bacterial TrxR structures except those from L. lactis and S. aureus display Leu/Ile in similar positions as Leu44 in EcTrxR (Fig. 4e).
Based on the structural information provided here we propose mechanisms that may account for the observed photo-inactivation of LlTrxR with the involvement of photo-excited FAD and O2 (Fig. 5). The initial step involves light absorption by the isoalloxazine ring and subsequent electron transfer from the ring to O2 with concomitant generation of superoxide radicals (O2−.) and an isoalloxazine radical-cation. Subsequent rapid, spontaneous dismutation of O2−.would give H2O218, as observed in other flavin photo-oxidation reactions19. The isoalloxazine radical-cation may then undergo one of two processes: electron transfer with the nearby Tyr237 to regenerate the isoalloxazine and generate a Tyr237 radical-cation (which would be expected to rapidly deprotonate to give a Tyr phenoxyl radical due to the very low pKa of these species), or deprotonation at one of the ring methyl groups. Such reactions are supported by literature data on the intramolecular photo-oxidation of target substrates by flavins subject to blue light, by electron transfer pathways19.
Tyr phenoxyl radicals undergo rapid reaction with O2−./HOO. to give hydroperoxides, with these being formed at either C3/C5 or C1 due to the high spin density at these sites20,21. Such peroxides are unstable and undergo ready decay to the corresponding catechols, potentially accounting for the formation of 3,4-dihydroxyphenyalanine (DOPA) from Tyr237. An alternative route for decay of the isoalloxazine radical-cation involves deprotonation at a methyl group (a well-established decay pathway of radical-cations22,23) to give a primary alkyl radical (–CH2.). In the presence of O2 this would lead to formation of a peroxyl radical, while in the presence of O2−.a hydroperoxide would be formed. Subsequent decay of the peroxyl radical via the Russell mechanism24, or decomposition of the hydroperoxide25, would yield an aldehyde.
An alternative route to oxidation of the C7α methyl group may be direct hydrogen atom abstraction from the methyl group by the Tyr237 phenoxyl radical (i.e. Tyr237-O.+flavin-CH3→Tyr237-OH+flavin-CH2.→formyl group) but such a pathway would be energetically less favorable due to the relative bond strengths of the C-H and O-H bonds. An alternative process involving formation of singlet oxygen (1O2) by the excited state isoalloxazine could also be envisaged, with this then undergoing decay to H2O2 or formation of an endoperoxide from the Tyr237 residue26. Subsequent decay of the unstable endoperoxide, potentially involving radical formation26, could account for the formation of DOPA from Tyr237 and oxidation of the C7α methyl group on the isoalloxazine ring to the formyl group.
The structure of LlTrxR also reveals an oxygen pocket at the si-face of the isoalloxazine ring, which is absent in EcTrxR. EcTrxR has previously been shown be highly photo-reactive and forms high yields of FAD semiquinones upon irradiation under anaerobic reaction conditions27. In light of the data obtained in the present study, it may therefore be suggested that the difference in photo-sensitivity between LlTrxR and EcTrxR is related to the accessibility of O2 in the active site, rather than the light-absorbing properties of the isoalloxazine ring per se. Besides its proposed role in photo-sensitivity the oxygen pocket might also have implications for the ability of these enzymes to reduce O2. We have reported that LlTrxR displays a ~10-fold increased rate of O2 reduction in the presence of NADPH as compared to EcTrxR16. LlTrxR thus shows a similar tendency to reduce O2 as the TrxR homologue AhpF, a dedicated reductase of the peroxide scavenger AhpC in eubacteria28,29. Interestingly, X-ray crystal structures of AhpF from S. typhimurium and E. coli (PDB 1HYU and 1FL2, respectively), also display a pocket on the si-face of the isoalloxazine and a tyrosine (Tyr344) located in the same position relative to the C7α as Tyr237LlTrxR. It can thus be hypothesized that AhpF also may be sensitive to photo-oxidation. However, it is important to emphasize that a direct comparison between reduction and photo-activation of O2 cannot be made since the molecular mechanisms and kinetic properties of these processes are different.
The novel molecular features of LlTrxR reported here provide insight into the photo-inactivation mechanism involving a protein-bound FAD acting as a photosensitizer that is highly relevant in a wider context. The conservation of Tyr237 and Met43 among TrxR from related Gram-positive bacteria suggests that photo-sensitivity may be widespread among these types of organisms. This opens up interesting perspectives in terms of clinical light treatment, e.g. blue light therapy of pathogenic and multi-drug resistant bacteria such as S. aureus, Streptococcus pyogenes and Bacillus anthracis, which all have TrxR predicted to harbour the described oxygen pocket.
Lactococcus lactis subsp. cremoris (strain MG1363) was grown in SA medium30 containing 1% (w/v) glucose and 2mg/L lipoic acid (GSAL medium). E. coli K-12 (strain MG1655) was grown in LB-medium. One liter synchronized cell culture (in biological triplicate) at exponential phase (OD600 0.6–0.7) was placed on ice for 30min and then centrifuged (4°C). The cell pellet was resuspended in 0.9% NaCl and distributed in ten 2mL screw-cap microcentrifuge tubes, centrifuged, and the pellets were stored at −20°C until use. Proteins were extracted by adding 600μL 100mM potassium phosphate pH 7.5, 1mM EDTA and 400μL glass beads ≤106 μm (Sigma) to each tube and performing bead beating on a FastPrep FP120 homogenizer (Qbiogene), setting 6 for 30s, followed by 1min rest on ice. Bead beating was repeated 6 times followed by centrifugation for 15min at 13,000g. Benzonase (2μL) was added to pooled supernatants (~4.5mL) and incubated for 1h at 21°C. The protein concentration was determined (Bio-Rad Protein Assay, Life Science Research) using BSA as a standard and the samples were aliquoted and stored at −20°C. Cell extracts were thawed, diluted 1:1 with 100mM potassium phosphate pH 7.5, 1mM EDTA and transferred to original 1.5mL Eppendorf tubes with 6×3mm magnets, and placed on a stirrer in a cold room at 4°C. The samples were irradiated with a 1225 Lumen (21W) lamp at 15cm distance as previously described16. Control samples wrapped in foil are referred to as “No Light”. After 0, 1.5, 3, 6, 9 and 12h cell extracts were removed and stored at −20°C for subsequent activity measurements. TrxR activity was assayed as 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB) reduction monitored at 412nm in 150μL format with 100mM potassium phosphate, pH 7.5, 2mM EDTA, 0.20mM DTNB and 0.20mM NADPH. For L. lactis 30μL of undiluted cell extract was used in the assay while the E. coli cell extracts were diluted 3 times in 100mM potassium phosphate, pH 7.5, 1mM EDTA before being added to the assay mixture. The activity of the cell extracts was measured directly or after addition of their respective recombinant thioredoxins (5μM final concentration), LlTrxA and EcTrx113. The addition of recombinant thioredoxin in the assay boosted the signal ~4 and ~2.3 times for L. lactis and E. coli, respectively. The initial activities were determined in at least technical triplicates for each biological replicate.
Production and purification of recombinant His-tagged L. lactis and E. coli TrxR and Trx and subsequent light inactivation of TrxR were performed essentially as described previously16. For crystallization trials 50μM LlTrxR was irradiated for 30, 60, 120, 180, and 240min (with remaining relative initial activities of ~73, 27, 26, 15 and 10%, respectively) and subsequently loaded on a HiLoad Superdex 75 gel filtration column (prep grade, 16/60; GE Healthcare) using 10mM HEPES, 200mM NaCl, 2mM Na-EDTA, pH 7.0 as eluent. Fractions that appeared homogeneous according to the chromatogram and SDS-PAGE were pooled, dialyzed (6−8kDa MWCO, Spectra/Por 1) against 10mM HEPES, 2mM Na-EDTA, pH7.0, and concentrated (Amicon Ultra centrifugal filter unit, 10kDa MWCO) to 8–10mg/mL as assessed from absorbance (456=11,300M−1cm−1)31.
Crystals obtained in FR conformation: LlTrxR (2μL, 8–10mg/mL) was mixed with 2μL reservoir buffer and co-crystallized with 1μL 25mM NADP+ (Sigma-Aldrich) and equilibrated as hanging-drops over a 500μL reservoir containing 35% PEG 1500 (Fluka), 400mM Li2SO4, 20mM HEPES of pH varying in the range 6.0–8.5. Bright yellow octahedron crystals (100μm) appeared after about 5 days at 19°C and grew to about 200μm in length by 14 days. Crystals of light-exposed (30, 60, 120, 180 and 240min) and light-protected LlTrxR were harvested and flash-cooled directly from the drop without additional cryoprotectants.
Crystals obtained in FO conformation: LlTrxR (2μL, 8–10mg/mL) was mixed with 2μL reservoir buffer, and 2μL 60mM DTT and equilibrated as hanging-drops over a 500μL reservoir containing 20% PEG 4000 (Fluka), 400mM Li2SO4. Only non-irradiated LlTrxR crystallized under these reducing conditions. Crystals appeared after ~14 days at 19°C. Crystals were transferred to a cryoprotectant solution consisting of reservoir buffer with 10% ethylene glycol before storing in liquid nitrogen.
X-ray diffraction data were collected at 100K at I911–3, MAXII, Sweden and at ID30A-1, ESRF, France, with wavelengths ranging from 0.97–1.00Å. Diffraction data were indexed, integrated, scaled and merged with XDS and XSCALE32 or MOSFLM33. Molecular replacement was performed with the program MOLREP34 from the CCP4 suite35 using the structure of SaTrxR (PDB 4GCM) as the initial search model for the FO conformation, while EcTrxR (PDB 1F6M) was used as template for the FR conformation. Restrained refinement was carried out using Refmac536 and manual inspection was done in Coot37. Local non-crystallographic symmetry restraints were imposed when appropriate. LlTrxR obtained in FR conformation crystallized in space group P412121 with 1 monomer per asymmetric unit, while FO conformations crystallized in P21 with 4 monomers per asymmetric unit. After refinement, the Ramachandran plots of the structures had 94.4–97.3%% residues in favored regions, 2.5–5.0% residues in allowed regions and 0–0.66% residues outliers. Details on the data collection statistics are provided in Table 1. All RMSD values for structural alignments of the Cα-atoms were calculated using the program Superpose in CCP438. Figures displaying structural data were made with PyMOL (http://www.pymol.org/).
LlTrxR (20μg; irradiated or light-protected) in 8M urea, 50mM NH4CO3 (45μL) was added to 2.4μL 100mM DTT and incubated (40min, 21°C), followed by addition of 2.5μL 200mM iodoacetamide and incubation in the dark (40min, 21°C). After 1:10 dilution with 50mM NH4CO3, endoproteinase GluC (Roche) was added to the samples in a 1:40 enzyme:protein ratio (w/w) and incubated overnight at 29°C. Trifluoroacetic acid was added to the digested samples to a final concentration of 0.5% and subjected to StageTip purification with empore C18 disks (3M) as previously described39. Samples (1μg) were analysed on a Q-Exactive Orbitrap (Thermo Fisher Scientific) coupled to an EASY-nLC 1000 liquid chromatograph (Thermo Fischer Scientific). Peptides were loaded onto a custom-made nanoLC column (15cm, C18, 100Å, 1.9μm particle size, 75μm ID) packed into a Picofrit emitter (New Objectives) and eluted using a 70min gradient at a flow rate of 250nL min−1. Resolution of 70,000, automatic gain control (AGC) value of 3106, maximum injection time (IT) of 20ms and scan range of 300 to 1500m/z were used for MS scans. MS/MS spectra were acquired in data-dependent mode (Top 10 method) with resolution of 17,500, AGC value of 1106, and maximum fragmentation accumulation time of 60ms.
Accession codes: Coordinates and structure factors have been deposited in the Protein Data Bank under accession codes: 5MH4 (FR conformation, no light); 5MIP (FR conformation, 30min light); 5MIQ (FR conformation, 60min light); 5MIR (FR conformation, 120min light); 5MIS (FR conformation, 180min light); 5MIT (FR conformation, 240min light); 5MJK (FO conformation, no light).
How to cite this article: Skjoldager, N. et al. The Structure of Lactococcus lactis thioredoxin reductase reveals molecular features of photo-oxidative damage. Sci. Rep. 7, 46282; doi: 10.1038/srep46282 (2017).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The Q-Exactive Orbitrap mass spectrometer was granted by the Danish Council for Independent Research | Natural Sciences (grant nr. 11-106246). Technical University of Denmark is acknowledged for a PhD stipend (NS) and The Danish Research Foundation, Nordic Bioscience A/S and Technical University of Denmark are acknowledged for a joint PhD scholarship (MR). MJD is supported by a grant from the Novo Nordisk Foundation (grant no: NNF13OC0004294). The MAXIV and ESRF synchrotrons is acknowledged for beam-time and support during data-collection. The research presented has received funding from BioStruct-X (EC FP7 grant agreement N°283570) and DANSCATT (the Danish agency and for Science, Technology and Innovation).
The authors declare no competing financial interests.
Author Contributions P.Ha., P.Hä., and B.S. conceived and directed the project; N.S., O.B., P.Ha., P.Hä. designed experiments; N.S. produced proteins and performed biochemical experiments. N.S., M.B.B. and P.Ha. performed X-ray crystallographic experiments and analysed the data; M.R. and P.Hä. performed mass spectrometry and analysed the data; N.S., M.R., O.B., M.J.D., B.S., P.Ha., P.Hä. discussed results and wrote the paper.