Research article

The crystal structure of human GLRX5: iron–sulfur cluster co-ordination, tetrameric assembly and monomer activity

Catrine Johansson, Annette K. Roos, Sergio J. Montano, Rajib Sengupta, Panagis Filippakopoulos, Kunde Guo, Frank von Delft, Arne Holmgren, Udo Oppermann, Kathryn L. Kavanagh


Human GLRX5 (glutaredoxin 5) is an evolutionarily conserved thiol–disulfide oxidoreductase that has a direct role in the maintenance of normal cytosolic and mitochondrial iron homoeostasis, and its expression affects haem biosynthesis and erythropoiesis. We have crystallized the human GLRX5 bound to two [2Fe–2S] clusters and four GSH molecules. The crystal structure revealed a tetrameric organization with the [2Fe–2S] clusters buried in the interior and shielded from the solvent by the conserved β1-α2 loop, Phe69 and the GSH molecules. Each [2Fe–2S] cluster is ligated by the N-terminal activesite cysteine (Cys67) thiols contributed by two protomers and two cysteine thiols from two GSH. The two subunits co-ordinating the cluster are in a more extended conformation compared with iron–sulfur-bound human GLRX2, and the intersubunit interactions are more extensive and involve conserved residues among monothiol GLRXs. Gel-filtration chromatography and analytical ultracentrifugation support a tetrameric organization of holo-GLRX5, whereas the apoprotein is monomeric. MS analyses revealed glutathionylation of the cysteine residues in the absence of the [2Fe–2S] cluster, which would protect them from further oxidation and possibly facilitate cluster transfer/acceptance. Apo-GLRX5 reduced glutathione mixed disulfides with a rate 100 times lower than did GLRX2 and was active as a glutathione-dependent electron donor for mammalian ribonucleotide reductase.

  • analytical ultracentrifugation
  • crystallization
  • glutaredoxin
  • glutathione
  • iron–sulfur
  • ribonucleotide reductase


GLRXs (glutaredoxins) are small evolutionarily conserved thiol–disulfide oxidoreductases that are involved in the maintenance of cellular redox homoeostasis and in cellular redox signalling. GLRXs have also been implicated in the maintenance of cytosolic and mitochondrial iron homoeostasis, and proteins from different organisms have been shown to co-ordinate [2Fe–2S] clusters [16].

Structurally, GLRXs belong to the thioredoxin-fold superfamily of proteins, but are distinguished from the thioredoxins by their specificity for GSH. Depending on the number of cysteine residues in the active-site sequence they can be divided into monothiol (CGFS) and dithiol (CP/SYC) GLRXs. Dithiol GLRXs are efficient catalysts of protein disulfide reactions, especially in reduction of mixed disulfides with glutathione. They have been shown to protect cells from oxidative stress and apoptosis, to regulate transcription factors and have, through these processes, been implicated in several disease-related conditions [7]. Monothiol GLRXs show a higher degree of sequence identity compared with the dithiol GLRXs and can in eukaryotes be further categorized into single-domain and multi-domain proteins. Some monothiol GLRXs have a five amino acid insertion in the loop preceding the active site, as well as a WP motif with unknown function. They contain several conserved amino acids known to be involved in GSH binding, but have low or no measurable activity in assays using established model substrates for GLRXs [810]. Alternatively, monothiol GLRXs have been implicated in iron regulation [9,1114] and iron–sulfur cluster biogenesis [15]. Hence, monothiol GLRXs are structurally similar to, but biochemically different from dithiol GLRXs.

To date four GLRXs have been identified in the human genome: two dithiol GLRXs, GLRX1 and GLRX2, a multi-domain monothiol, GLRX3 [16], and the mitochondrial single-domain monothiol, GLRX5. In addition, there are two TXNRDs [TXN (thioredoxin) reductases] with an N-terminal GLRX domain (TXNRD3 and TXNRD1) [17]. Human GLRX5 is well conserved among eukaryotes, and initial studies in yeast revealed early an important role in mitochondrial iron–sulfur cluster biogenesis [11]. Phylogenetic profiling predicts that GLRX5 is part of the Isc iron–sulfur assembly machinery [18], and it is suggested to be required for transfer and insertion of clusters into acceptor proteins after the clusters have been assembled on the IscU scaffold protein [19]. In addition, yeast two-hybrid experiments demonstrate that GLRX5 interacts with the Isa1 scaffold protein [18]. Wingert et al. [15] showed that deficiency of the GLRX5 orthologue in zebrafish (known as Shiraz) affects haem biosynthesis through a cytosolic IRP (iron regulatory protein) [15]. In humans, a silent mutation in the GLRX5 gene that results in decreased levels of GLRX5 protein, results in sideroblastic-like microcytic anaemia characterized by mitochondrial iron overload and impaired haem synthesis [20].

In the present paper we describe the crystal structure of the [2Fe–2S]-bound human GLRX5, and present a biochemical characterization of the protein. Our results reveal a tetrameric organization with pronounced differences in the lability or accessibility of the [2Fe–2S] cluster compared with human GLRX2, and show that the apoprotein possesses low, but significant, GSH–disulfide oxidoreductase activity.


Expression and purification of recombinant human GLRX2 and GLRX5

Recombinant human GLRX2 was expressed and purified as described previously [6]. A template plasmid encoding full-length human GLRX5 was obtained from the Mammalian Gene Collection (cDNA clone IMAGE:6066312). A construct comprising residues 35–148, lacking the N-terminal mitochondrial signal, was cloned into pNIC28-Bsa4 (GenBank® accession number EF198106) with a TEV (tobacco etch virus) cleavable N-terminal His6 tag. For expression of native GLRX5, the plasmid was transformed into a phage-resistant derivative of Escherichia coli strain BL21(DE3) carrying the pRARE2 plasmid for rare codon expression. The cells were grown at 37 °C in Terrific Broth supplemented with 50 μg/ml kanamycin and 34 μg/ml chloramphenicol, until the culture reached a D600 of 1.5. The temperature was decreased to 18 °C and protein expression was induced with 0.1 mM IPTG (isopropyl β-D-thiogalactopyranoside) overnight. The cells were collected by centrifugation (5000 g for 15 min) and frozen at −80 °C. For expression of SeMet (selenomethionine)-labelled GLRX5, the expression plasmid was transformed into E.coli strain B834(DE3) and cultures were grown at 25 °C in LB (Luria–Bertani) medium supplemented with 50 μg/ml kanamycin. The cells were grown until they reached a D600 of 1.0, were washed three times with sterile water and resuspended in M9 minimal medium (Molecular Dimensions). The culture was used to inoculate M9 minimal medium supplemented with 40 mg/l SeMet, 50 μg/ml kanamycin and 100 μM FeCl3, and grown at 37 °C until they reached a D600 of 0.8. The temperature was decreased to 18 °C and protein expression was induced with 0.5 mM IPTG overnight. The cells were harvested by centrifugation (5000 g for 15 min) and stored at −80 °C. Cell pellets were resuspended in 50 mM Hepes (pH 7.5), 500 mM NaCl, 20 mM imidazole, 5% glycerol and protease inhibitors (Complete, Sigma) and lysed by sonication (20 kHz with 2 s on and 3 s off for 10min). Cell debris and nucleic acids were removed by addition of 0.15% polyethyleneimine followed by centrifugation for 45 min at 40000 g. The proteins were purified by nickel-affinity chromatography (5 ml Ni-Sepharose FF, GE Healthcare) using a stepwise gradient of imidazole, and fractions were analysed by SDS/PAGE. The protein was concentrated and the buffer was exchanged to 50 mM Hepes (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 10 mM GSH, 3 mM DTT (dithiothreitol) and 0.5 mM TCEP [tris-(2-carboxyethyl)phosphine] using an Amicon centrifugal filtration unit [Millipore, 3kDa MWCO (molecular mass cut-off)]. The histidine tag was removed by incubating GLRX5 with TEV protease (150 μg of TEV/10 mg of GLRX5) for 16 h at 4 °C. The protein was re-purified by applying to a nickel-affinity column (500 μl of Ni-Sepharose FF, GE Healthcare) and the flow-through was collected. The TEV-cleaved protein was concentrated using an Amicon 3K-centrifugal filtration unit and used for crystallization and further characterization.

Apo-GLRX5 was obtained by addition of 5 mM EDTA followed by gel-filtration chromatography on a HiLoad 16/60 Superdex 200 column (GE Healthcare) equilibrated with a buffer composed of 10 mM Hepes (pH 7.5), 500 mM NaCl, 5% (v/v) glycerol, 5 mM GSH and 0.5 mM TCEP.


The concentration of GLRX5 was determined by measuring the absorbance at 280 nm using a Labtech Nanodrop 1000 spectrophotometer and a predicted molar absorption coefficient of 11460 M−1·cm−1. The stability of the iron–sulfur clusters was monitored aerobically at 20 °C for 240 min by measuring the absorbance at 320 and 420 nm using a Polar Star Omega plate reader (BMG Labtech) under various conditions using 2–5 mM GSH, GSSG, H2O2 or EDTA.

Molecular mass analysis

The molecular mass of GLRX5 was verified by electrospray-ionization MS (Agilient 1100 series LC/MSD).

Analytical gel-filtration chromatography

Apo- and holo-GLRX5 were applied to an analytical 10/300 Superdex 200 column (GE healthcare) equilibrated with 10 mM Hepes (pH 7.5), 300 mM NaCl, 5 mM GSH and 0.5 mM TCEP at 4 °C. The column was calibrated with low-molecular-mass standards comprising BSA (66 kDa) and bovine carbonic anhydrase (29 kDa).

Analytical ultracentrifugation

Sedimentation-velocity experiments were carried out on a Beckman Optima XL-I analytical ultracentrifuge (Beckman Instruments) equipped with an AnTi-50 rotor and cells with double sector centrepieces. Protein samples were studied at a concentration of 50 μM in 10 mM Hepes (pH 7.5) and 300 mM NaCl at 10 °C, employing a rotor speed of 45000 rev./min. Radial absorbance scans were collected using absorbance optics at a wavelength of 280 nm in continuous scan mode, at 2 min intervals with a radial step size of 0.003 cm. Aliquots (300 μl) were loaded into the sample chambers of double channel, 12 mm centrepieces and 310 μl of buffer was used in the reference channels. Data were analysed using the SEDFIT [21] software package whereby differential sedimentation coefficient distribution, c(s), were obtained by direct boundary modelling to Lamm Equation solutions. Sedimentation coefficients, s, were obtained by integration of individual peaks in the calculated c(s) values, after fitting of the frictional ratio, in order to allow these distributions to be corrected for the effects of diffusion. The software package SEDNTERP [22] (version 1.08) was used in order to convert the obtained sedimentation coefficient values into the equivalent values in water at 20 °C, s020,w, taking into account the solvent density (1.01313 g/ml), viscosity (1.567×10−2 P) and partial specific volume (0.7255 or 0.7316 ml/g for the dimer or monomer).

Enzymatic characterization of GLRX5

To measure GLRX activity, Di-Eosin-GSSG [23] was used to glutathionylate BSA (S.J. Montano, J. Lu, J and A. Holmgren, unpublished work). The Eosin-GSH-BSA was used as a substrate, and release of eosin-GSH, which is highly fluorescent with a 545 nm emission, was measured. A black 96-well plate was used in a PerkinElmer Victor3 multi-label counter containing a final well volume of 100 μl with 1 μM apo-GLRX5 or 10 nM GLRX2, 0.25 mM NADPH and 45 nM yeast GR (glutathione reductase) in 0.1 M potassium phosphate buffer (pH 7.5), 1 mM EDTA and 20 μM of Eosin-GSH-BSA. The reaction was started by addition of 25, 50, 100, 200, 400 or 800 μM GSH, followed by recording the fluorescence emission at 545 nm after excitation at 520 nm. Controls where no GSH was added in the reactions were included in the experiment. In a separate experiment, concentrations of 0, 0.5, 1, 2 and 4 μM GLRX5 or 0.625–20 nM GLRX2 were incubated with 0.5 mM GSH, 0.25 mM NADPH and 45 nM yeast GR in 0.1 M potassium phosphate buffer (pH 7.5) and 1 mM EDTA. The reaction was started by adding 20 μM Eosin-GSH-BSA and the fluorescence was recorded. Controls in the absence of GLRX2/5 were included. In the final experiment, 60 nM rat recombinant full-length TXNRD1 [24] replaced GR and was incubated with up to 4 μM GLRX5 to test whether this enzyme used GLRX5 as a substrate. The fluorescence increase per min at 545 nm was calculated within a linear range of the reaction curve to determine the catalytic activity of GLRX5 and GLRX2.

RNR (ribonucleotide reductase) activity with GSH was used to test the activity of GLRX5. Proteins R1, R2 and p53R2 of RNR were expressed and purified as described previously [25,26]. Determination of RNR activities were carried out as described previously [25]. Briefly, RNR enzyme was reconstituted by mixing recombinant R1 and R2 or p53R2 proteins. Activity was assayed following the conversion of [3H]CDP into [3H]dCDP. R1 and R2 or p53R2 proteins were pre-incubated with 2 mM ATP and 10 mM MgCl2 at 37 °C for 15 min and the reaction was initiated by adding reaction mixture containing 40 mM Tris/HCl buffer (pH 7.6), 2 mM ATP, 10 mM MgCl2, 200 mM KCl, 20 μM FeCl3 and 0.5 mM [3H]CDP (~10000 c.p.m./nmol) in a final volume of 50 μl. When a GLRX system was used together with RNR, the samples contained 2 μM apo-GLRX5, 0.15 μM GR, 1 mM NADPH and 10 mM GSH. Incubation was carried out at 37 °C for 60 min. The reaction was terminated by the addition of 1 M HClO4, which also hydrolyses dCDP to dCMP. The amount of [3H]dCMP radioactivity was quantified by liquid-scintillation counting after ion-exchange chromatography on Dowex-50 columns. The activity was calculated as nmol of dCDP produced (measured as dCMP) per time of incubation, and OriginPro 8.1 software used for data analysis.


SeMet-substituted crystals were grown by vapour diffusion employing the sitting drop method, using 75 nl of protein (37 mg/ml) and 75 nl of well solution containing 70% (v/v) MPD (2-methyl-2,4-pentanediol) and 0.1 M Hepes (pH 7.5). Native crystals were grown using the same technique in a drop of 150 nl of protein (93 mg/ml) and 150 nl of well solution containing 50% (v/v) PEG [poly(ethylene glycol)] 300, 0.2 M MgCl2, 0.1 M cacodylate (pH 6.5) and 0.01 M spermine tetrahydrochloride. Both types of crystals were grown aerobically at 20 °C, were shaped like thin rods and were visibly brown. The crystals were flash-cooled in liquid nitrogen straight from the drop without additional cryo-protection.

X-ray data collection and refinement

Both the SeMet and the high-resolution X-ray data were collected at 100 K on beamline X10SA at the Swiss Light Source. The SeMet crystal diffracted to 2.7 Å (1 Å=0.1 nm) and was collected at the wavelength λ=0.9794 Å. After processing in Mosflm [27] the data were analysed in XPREP (Bruker AXS) and four SeMet sites were found using SHELXD [28] with an anomalous signal to 3.1 Å. Initial phases were calculated with SHELXE and the space group pinpointed to P43212. The phases were improved in SHARP [29], and Buccaneer [30] was used to build fragments of secondary structural elements into the solvent-flattened maps. Positions for four molecules of the closest homologue, GLRX C1 from Populus tremula x tremuloides (PDB code 2E7P), were found by manual molecular replacement in Coot [31] using the Buccaneer-built fragments as a guide. After rigid body refinement in REFMAC5 [32] and phase improvement using Parrot [33], the maps enabled a first rough rebuild in Coot. The resulting model was used for molecular replacement into the higher-resolution native data with Phaser [34]. The native data were collected at λ=0.979Å in five wedges along the thin, rod-like native crystal that showed highly anisotropic diffraction. In each segment, 20 degrees of data survived radiation damage. The frames were integrated with Mosflm and data were merged and scaled using Scala from CCP4 suite to a final resolution of 2.4 Å [27]. Before refinement commenced, 5% of the data were set aside for the calculation of Rfree. The maps from the Phaser solution were automatically traced by ARP/wARP [35], and the model was improved by manual rebuilding in Coot cycled with restrained refinement in REFMAC5 including TLS (translation/liberation/screw) groups. Refinement included all data to 2.4 Å, although, due to the anisotropy of the data, the final resolution of the structure is considered to be 2.6 Å. The final model and structure factors have been deposited in the PDB with the accession code 2WUL. Statistics are presented in Supplementary Table S1 (at

Electrostatic surface potential

The PDB2PQR server [36] was used to convert the protein files into PQR format and charges were assigned using the PARSE force field. The APBS (Adaptive Poisson–Boltzmann Solver) plugin for PyMOL (DeLano Scientific; was used to map the electrostatic potential (±5 kT/e) on to the molecular surface of the protein.


Purification and analysis by UV–visible spectroscopy

On the basis of previous experience with human GLRX2, holo-GLRX5 was purified aerobically in the presence of reduced GSH and TCEP. Holo-GLRX5 had a visible brown colour and absorbed light of 320, 415, 420 and 457 nm wavelength, characteristic of a [2Fe–2S] centre [1], whereas the apoprotein had no absorption in this region (Supplementary Figure S1 at The stability of the clusters in holo-GLRX5 and holo-GLRX2 was followed at 420 nm after addition of GSH, EDTA, GSSG or H2O2 (Figure 1). GLRX5 precipitated immediately when 2 mM GSSG was added to the holoenzyme (results not shown) and 75 min after addition of 5 mM H2O2. This is in contrast with holo-GLRX2 where both GSSG and H2O2 slowly destabilizes the cluster, but does not precipitate the protein. Addition of EDTA affected cluster dissociation faster in holo-GLRX5 compared with holo GLRX2, whereas GSH seemed to stabilize the cluster in both proteins. These results reveal differences in cluster stability and interaction with redox compounds.

Figure 1 Stability of the iron–sulfur cluster in GLRX5 in comparison with GLRX2

The stability of the [2Fe–2S] cluster in GLRX2 (A) and GLRX5 (B) was monitored aerobically by measuring the absorbance at 420 nm at 20 °C for 240 min using a Polar Star Omega plate reader after addition of 2 mM of either GSH, H2O2 or EDTA, and the absorbance corrected to the protein concentration used. Grey line, untreated protein; black line, GSH; broken lines, H2O2; open circles, EDTA.

Analysis of the oligomeric state

In gel-filtration chromatography experiments, holo- and apo-GLRX5 eluted as two peaks corresponding to apparent molecular masses of 52 kDa (tetramer) and 13 kDa (monomer) (Figure 2A). This is in contrast with holo GLRX2, which is a dimer in solution. The tetrameric quaternary state of holo GLRX5 was further supported by sedimentation velocity analytical ultracentrifugation which confirmed that holo-GLRX5 is a tetramer in solution, whereas apo-GLRX5 is monomeric (Figure 2B). Hence, tetramerization of GLRX5 is coupled to iron–sulfur cluster binding.

Figure 2 The oligomeric state of apo- and holo-GLRX5 analysed by gel-filtration chromatography and by sedimentation velocity analytical ultracentrifugation

(A) Apo- (broken line) and holo- (solid line) GLRX5 were applied on to a Superdex 200 column equilibrated with 10 mM Hepes (pH 7.5), 300 mM NaCl, 5 mM GSH and 0.5 mM TCEP at 4 °C. The column was calibrated with low-molecular-mass standards (grey filled area) consisting of BSA (66 kDa) and bovine carbonic anhydrase (29 kDa). (B) Sedimentation velocity plot of the differential sedimentation coefficient distribution, c(s), against the apparent sedimentation coefficient corrected to water at 20 °C, s20,w of GLRX5, together with the differential molecular mass distribution, c(M), against the system mass, M. Experiments were conducted with a protein concentration of 13 μM in 10 mM Hepes (pH 7.5) and 300 mM NaCl at 4 °C. The protein sediments as a monomer (broken line) in the absence of an [2Fe–2S] cluster, or as a mixture of monomer and tetramer (solid line) in the presence of an [2Fe–2S] cluster.

Molecular mass analysis

The molecular mass of native and SeMet-labelled holo-GLRX5 was determined to be 12852 and 12946 Da respectively, in agreement with the predicted mass of the expressed constructs. Analysis of apo-GLRX5 identified three peaks of 13464, 13158 and12851 Da (Figure 3). The masses of the first two peaks are consistent with the addition of two and one GSH molecule respectively, whereas the third peak potentially could correspond to an intramolecular disulfide. Tryptic digestion of the apoprotein followed by ion-trap MS analysis identified only glutathionylation of Cys67, and no other cysteine-containing fragment corresponding to glutathionylation of Cys122 or an intramolecular disulfide was detected (results not shown).

Figure 3 Molecular mass of apo GLRX5

The molecular mass of apo GLRX5 was verified by electrospray mass ionization-time-of-flight MS.

Crystal structure of [2Fe–2S]-bound GLRX5

The crystallographic asymmetric unit contains 436 protein residues in four chains (A41–149, B41–149, C40–148 and D41–149), four GSH molecules, two inorganic [2Fe–2S] clusters, four chloride ions, one molecule of PEG and 93 water molecules. In addition, the electron-density map reveals an unknown entity in close proximity to Phe69 and GSH in each molecule that was left unmodelled. More than 99% of residues are in favoured regions and all are in allowed regions of a Ramachandran plot calculated by MolProbity [37].

The [2Fe–2S] clusters are each co-ordinated by two protein chains and two GSH molecules where the iron atoms are tetrahedrally co-ordinated by active-site Cys67, a cysteine contributed by GSH and two inorganic sulfurs. In this way there are two [2Fe–2S] co-ordinated dimers whereby ~1050 Å2 of the surface area of the two protein chains are buried upon dimer formation. Analysis of the crystal packing suggests a tetrameric form (Figure 4A) with an additional 840 Å2 of surface area buried upon tetramer formation. The cluster is completely buried from the exterior, and solvent accessibility is only possible via the central aperture. The GSH molecules and the position of the Phe69 side-chain further reduced the solvent accessibility of the cluster (Figure 4C). Each protein chain comprises a central four-stranded mixed β-sheet flanked by five α-helices, typical of the thioredoxin superfamily of proteins. Cys67 is located at the N-terminus of helix α2 and the loop between β strand β1 and helix α2 (β1-α2 loop, residues 59–67) forms a β-hairpin which is folded back on to the protomer and packs against α2 and β2.

Figure 4 Structure of human GLRX5

(A) Tetramer coloured by protein chain with the β1–α2 loop shown in crimson. The [2Fe–2S] cluster is represented by orange and yellow spheres (for iron and sulfur respectively), whereas the GSH is shown as sticks with carbon coloured white, nitrogen coloured blue and oxygen coloured red. Phe69 which shields the cluster from solvent is also shown in stick representation with carbon coloured by protein chain. (B) Interactions between GSH and the protein chains where main-chain contacts are depicted by black triangles and hydrogen bonds are shown by broken lines. (C) Stereoview of [2Fe–2S] cluster co-ordination and intersubunit interactions. The view is looking at the tetramer edge as shown by the arrow and icon in (A). [2Fe-2S] cluster co-ordination and intersubunit interactions are shown by black and yellow dotted lines respectively. Features associated with protein chain B are marked by an asterisk.

Lys59, which is conserved in both monothiol and dithiol GLRXs, anchors the beginning of the β1-α2 loop, whereby the backbone NH forms a hydrogen bond with the hydroxy group of Ser70 (2.8–3.0 Å, in the four protein chains), whereas the backbone carbonyl forms a hydrogen bond to the backbone NH in Gln66 (3.0 Å). The Lys59 side-chain has two ionic interactions with the GSH that is closely associated with the protein chain: one with the glycine carboxylate (2.5–2.7 Å) and another with the cysteine thiolate (3.0–3.4 Å). NZ of Lys59 also forms a hydrogen bond with the backbone carbonyl of Gln66 (2.6–2.9 Å) from the associated protein molecule that co-ordinates the cluster. The β1–α2 loop is further stabilized by hydrogen bonds between the backbone NH of Thr61 and the backbone carbonyl in Gln64, and between the hydroxy of Thr61 and the backbone NH of Gln64 (3.0 Å). Finally, Cys67 anchors the C-terminus of the β1-α2 loop through its role in co-ordinating the [2Fe–2S] cluster. The charged side-chains of Glu63 and Gln64 in the β1–α2 loop, as well as Asp92 and Asp93 in the β2–α3 loop, stretch out away from the protein.

The only other cysteine residue present in the protein sequence, Cys122, is located in the beginning of helix α4. Cys122 is separated from Cys67 by 7.6–8.4 Å (sulfur-to-sulfur distance) and the side-chain of Phe69 intercalates between the two cysteine side-chains.

The interface between the two molecules that co-ordinate the cluster is not completely symmetrical: Arg97 interacts across the interface by forming intermolecular hydrogen bonds with the main-chain carbonyl of Pro65 (Arg97A-Pro65B, Arg97C-Pro65D) or the Gln66 side-chain (Arg97B-Gln66A, Arg97D-Gln66C). Two residues, Trp106 and Pro107, located on the α3–β3 loop stabilize the tetramer by packing against the α4 helix in the neighbouring subunit (A-D and B-C).

Each of the GSH molecules that co-ordinate the [2Fe–2S] cluster makes extensive contacts with an individual protein chain (Figure 4B). The glutamate carboxylate accepts hydrogen bonds from the backbone nitrogens of Cys122 and Asp123, and the glutamate amine participates in a salt bridge with the side-chain of Asp123. The GSH cysteine has two backbone interactions with the main-chain of Ile109, and the glycine carboxylate has ionic interactions with the side chains of Lys59, Arg97 and Lys101.

Analysis of the surface charge

Analysis of the sequence of GLRX5 reveals that it is an acidic protein with a predicted isoelectric point of 4.5. This is in contrast with GLRX2 which has a predicted isoelectric point of 9.2. Mapping the electrostatic potential on to the surface of the proteins, as shown in Figure 5, illustrates the difference in the surface properties of the two proteins. The results suggest that GLRX5 and GLRX2 probably have different interaction partners, or possibly even interact with each other.

Figure 5 Electrostatic potential (±5 kT/e) mapped on to the molecular surface

(A) GLRX5 tetramer. (B) GLRX2 dimer.

Enzymatic activity

Since one of the most obvious roles for GLRXs in iron–sulfur cluster assembly or transfer is to reduce disulfides or mixed disulfides with GSH on iron–sulfur acceptor proteins, we measured whether GLRX5 could reduce mixed disulfides with GSH. When GSH was used as an electron donor, micromolar concentrations of GLRX5 catalysed the reduction of glutathionylated BSA (Figure 6A). This can be compared with human GLRX2, which under the same conditions reduced glutathionylated BSA at nanomolar concentrations (Figure 6B). These results show that the relative catalytic activity of GLRX5 is approx. 500-fold lower than GLRX2. No activity was detected when TXNRD1 was used as an electron donor instead of GSH. Thus, unlike GLRX2 [38], GLRX5 is not a direct substrate of TXNRD1 (results not shown). To study the reduction of the GLRX–glutathione intermediate, potentially formed in the second step in the reaction above [39,40], 20 μM Eosin-GSH-BSA was used with either 1 μM GLRX5 or 10 nM GLRX2, and the GSH concentration varied (Figures 6C and 6D). These results revealed that the GSH–GLRX5 mixed disulfide is reduced with a rate at least 100-fold lower than that of GSH–GLRX2 (Table 1).

Figure 6 Characterization of GLRX5 and comparison with GLRX2 as GSH–disulfide oxidoreductases

In a 100 μl reaction volume, 0.25 mM NADPH, 0.5 mM GSH and 45 nM yeast GR were incubated with 0–4 μM GLRX5 (A) or 0–20 nM GLRX2 (B). Subsequently, GSH concentrations of 25–800 μM were added to 1 μM GLRX5 (C) or 10 nM GLRX2 (D) in the presence of 0.25 mM NADPH and 45 nM yeast GR. Reactions were started by the addition of 20 μM Eosin-GSH-BSA and the relative fluorescence was recorded at 545 nm. Hanes–Woolf plots are shown as insets and the OriginPro 8.1 software was used for data analysis.

View this table:
Table 1 Kinetic constants for GSH-dependent reduction of mixed disulfides

To test whether GLRX5 was acting as an electron donor for RNR, experiments were performed with different concentrations of GSH in combination with human GLRX5 and mouse R1–R2 complex. With 5 mM GSH, no RNR activity was recorded with the GLRX5 system, whereas 10 and 20 mM GSH showed significant activity for RNR catalysis (Figure 7A). RNR activity was also recorded for the R1–R2 complex with various concentrations of human GLRX5 (0.1–5 μM) in the presence of 10 mM GSH, 1 mM NADPH and 0.15 μM GR (Figure 7B). A Km value of 0.4 μM was obtained for GLRX5 with the R1–R2 complex. The same experiment was performed with the R1–p53R2 complex and it displayed a typical apparent Km value of 0.9 μM for GLRX5 (Figure 7B).

Figure 7 Effect of GLRX5 and GSH on mouse RNR complexes in reduction of CDP to dCDP

(A) Samples with 200 μg/ml R1 and 40 μg/ml R2 were assayed with various concentrations of GSH. The reaction was started by adding reaction mixture supplemented with 2 μM GLRX5, 1 mM NADPH and 0.15 μM yeast GR. (B) RNR activity of 200 μg/ml R1 with 100 μg/ml R2 (●) or 100 μg/ml p53R2 (Δ) were measured in the presence of increasing amounts of GLRX5 with 1 mM NADPH, 10 mM GSH and 0.15 μM yeast GR. The results represent two independent experiments with duplicate samples. The kinetic data was determined from a double- reciprocal Lineweaver–Burk plot using the OriginPro 8.1 software.


After analysis of the GLRX5 crystal structure revealed a tetrameric organization, studies to determine the quaternary structure in solution were initiated. Analytical ultracentrifugation and size-exclusion chromatography support that holo-GLRX5 is tetrameric, whereas the metal-free protein is monomeric. Thus far, only the monothiol GLRX6 from Saccharomyces cerevisiae is reported to be a tetramer in solution when co-ordinating an iron–sulfur cluster [10]. A recently described structure of GLRX6 is of the iron–sulfur-free GSH-bound monomeric form, whereas the tetrameric form apparently requires both iron–sulfur and an N-terminal domain [41]. Analysis of the crystal packing of the E. coli GLRX4 (a homologue of human GLRX5), using PISA (Protein interfaces, surfaces and assemblies service) at the European Bioinformatics Institute also supports a tetrameric quaternary structure for GLRX4. Although the [2Fe–2S]-bound dithiol poplar GLRX C1 (GRXC1) crystallizes with a tetramer in the asymmetric unit, only two chains co-ordinate a [2Fe–2S] cluster, whereas the other two chains are apoproteins. Consequently it was shown that the tetrameric form of GRXC1 is a result of crystal packing, since the protein is dimeric in solution. Thus the apparent oligomeric structure for the [2Fe–2S]-bound monothiol GLRXs (GLRX4 and GLRX5) is different than the dimeric structure observed for the [2Fe–2S]-bound dithiol GLRXs (GLRX2 and GRXC1).

Compared with the GLRX2 dimer, the two protein chains that co-ordinate the [2Fe–2S] cluster in GLRX5 are in a more extended conformation. The two GLRX5 protomers bind the cluster in a head-to-head fashion in which the two β-sheets are essentially co-planar. However, in GLRX2 the protomers that co-ordinate the [2Fe–2S] cluster are organized such that the β-sheets form a ~120 ° angle and one is rotated by approx. 90 °. The extended arrangement seen in GLRX5 was also observed for E. coli GLRX4 [42] and could be necessary to support tetramerization, since the bent conformation observed for the GLRX2 dimer would sterically interfere with this type of tetramer oligomerization.

The intersubunit interactions in GLRX5 are more extensive than those observed in human GLRX2, and are found between Lys59, Pro65, Gln66 and Arg97. Interestingly Lys59, Pro65 and Arg97 are conserved in monothiol GLRXs, and Lys59, Pro65 and Gln66 are located on the extended β1–α2 loop that in the tetrameric structure shields the [2Fe–2S] cluster. The position of the β1-α2 loop is similar in the E. coli GLRX4 [2Fe–2S]-bound structure [42], whereas in the solution structure of monomeric E. coli GLRX4 [43], the loop protrudes away from the protein. Structural alterations of this loop have previously been suggested to be involved in cluster transfer [42], since it affects the position of Cys67 and contains several residues that both interact with GSH and are involved in protomer interactions. In holo GLRX5, the clusters are buried in the tetrameric structure and largely inaccessible. However, if the proteins implicated in cluster transfer were to interact with the β1–α2 loop and potentially also the β2–α3 loop in protomer A and B or C and D, they would gain close proximity to the cluster. In the present structure, the conformation of the loops are stabilized by several hydrogen bonds and charged residues in both the β1–α2 loop (Glu63 and Gln64) and the β2–α3 loop (Asp92 and Asp93), which is conserved among most monothiol GLRXs, reach out into the solution. Thus it would be interesting to perform mutational studies of these residues once the interaction partners of GLRX5 are known.

Besides the cysteine residue in the CGFS active-site motif, GLRX5 possess a second semi-conserved cysteine residue, Cys122, that, in yeast, has been shown to form an intramolecular disulfide with Cys67 [44,45]. We have not been able to find conclusive evidence of such a disulfide, which would require structural rearrangements, including unwinding of the first turn of helix α2. This would not be unprecedented, as a similar type of disulfide and helix unwinding has been observed upon oxidation of both human GLRX1 and human TXN1 [46,47]. Thus the proved sensitivity of apo- and holo-GLRX5 to GSSG or H2O2 might be due to destabilizing structural rearrangements, possibly by formation of such an intramolecular disulfide. The higher reactivity towards GSSG compared with H2O2 is potentially due to the glutathione-binding site or that GSH in the cluster might exchange with the GSH in the buffer, similarly to holo-GLRX2 [48].

MS analysis revealed that GLRX5 in vitro can be covalently modified by one and two GSH molecules if the [2Fe–2S] cluster is not present. Although we were only able to identify a GSH-mixed disulfide with Cys67, the most plausible explanation is that both Cys67 and Cys122 are modified by GSH. In the present structure of GLRX5, Cys67 is located ~3.8 Å from the GSH cysteine thiol and could readily form a mixed disulfide without significant rearrangement if the [2Fe–2S] cluster was not present. GSH bound non-covalently to S. cerevisiae GLRX6 [41] adopts a similar binding mode as it does when covalently attached in GSH–GLRX mixed disulfides [6], and in human GLRX2 the non-covalently bound GSH also makes similar interactions with the protein, as it does when co-ordinating the [2Fe–2S] cluster [6]. Thus glutathionylation of GLRX5 would protect the cysteine thiol groups from further oxidation and potentially also facilitate transfer or acceptance of the clusters.

Both glutathionylated BSA and RNR revealed that apo-GLRX5 has low, but significant, deglutathionylation activity. In the reduction of glutathionylated BSA, the second half reaction is a 100-fold lower than that measured for GLRX2, and this result is similar to most monothiol GLRXs [9,44,49]. Reduction of the GLRX–GSH mixed disulfide by GSH might be inhibited by the five amino-acid insertion on the β1–α2 loop, or could be disfavoured if the GLRX–GSH mixed disulfide serves as a scaffold for iron–sulfur clusters [50]. GLRX5 also reduced the R1–R2 complex with a Km which is comparable with previous studies with TXN1 and GLRX1 [25]. However, the specific activity and apparent Vmax with RNR with R2 or p53R2 were found to be much lower for GLRX5 compared with GLRX1 or GLRX2 [25]. Thus the activity of both of the RNR complexes with GLRX5 reconfirms the glutathionylation mechanism for GLRX catalysis in contrast with the dithiol mechanism for TXN. Although the results clearly show that GLRX5 is active, its in vivo role in ribonucleotide reduction may be limited.


Catrine Johansson designed the research together with Udo Oppermann, and performed the purification, crystallization, spectroscopy and molecular mass analysis. Kunde Guo helped with the purification and crystallization, and Panagis Filippakopoulos performed the analytical ultracentrifugation. Annette Roos solved the structure with help from Frank von Delft. Sergio Montano and Rajib Sengupta performed the enzyme kinetics under the supervision of Arne Holmgren. Catrine Johansson, Annette Roos, Panagis Filippakopoulos, Rajib Sengupta, Arne Holmgren, Udo Oppermann and Kathryn Kavanagh contributed to data analysis, preparation of Figures and final revision of the manuscript, which was primarily edited by Catrine Johansson.


This work was supported by the Swedish Research Council [grant number 3529]; the Swedish Cancer Society [grant number 961]; and The K.A Wallenberg Foundation [grant number 2006–0192]. R.S. was supported by a Wenner-Gren postdoctoral fellowship. The Structural Genomics Consortium is a registered charity (number 1097737) funded by the Canadian Institutes for Health Research, the Canadian Foundation for Innovation, Genome Canada through the Ontario Genomics Institute, GlaxoSmithKline, Karolinska Institutet, the Knut and Alice Wallenberg Foundation, the Ontario Innovation Trust, the Ontario Ministry for Research and Innovation, Merck and Co, the Novartis Research Foundation, the Swedish Agency for Innovation Systems, the Swedish Foundation for Strategic Research and the Wellcome Trust. The study was supported by the Oxford NIHR Biomedical Research Unit.

Abbreviations: GLRX, glutaredoxin; GR, glutathione reductase; IPTG, isopropyl β-D-thiogalactopyranoside; PEG, poly(ethylene glycol); RNR, ribonucleotide reductase; SeMet, selenomethionine; TCEP, tris-(2-carboxyethyl)phosphine; TEV, tobacco etch virus; TXN, thioredoxin; TXNRD, TXN reductase


View Abstract