Functional, structural, and spectroscopic characterization of a glutathione-ligated [2Fe–2S] cluster in poplar glutaredoxin C1

Results and Discussion

Cloning and Expression of Plant Grxs in E. coli.

To gain insight into the function and structure of plant Grxs, poplar Grxs of subgroup I (Grx C1, C2, C3, and S12) were produced as recombinant proteins in E. coli, purified to homogeneity and compared with the previously characterized Grx C4 (2, 12). Among the various Grx characterized in all kingdoms, GrxC1 is unique to plants having a YCGYC active site, whereas poplar Grx C2 contains a YCPFC active site, Grx C3 and C4 a “classical” YCPYC active site, and Grx S12 a WCSYS active site similar to the one found in human Grx 2 (CSYC) [supporting information (SI) Fig. 5]. Unexpectedly, among the five Grxs of subgroup I, recombinant Grx C1 (12,514 Da) eluted in exclusion size chromatography as two successive peaks, one with a higher oligomerization state and a brown coloration characteristic of the presence of an Fe–S center, and a colorless peak of smaller size. Gel filtration compared with proteins of known molecular mass, indicated molecular masses of 34 and 17 kDa for the brown holoprotein and colorless apoprotein fractions, respectively (data not shown). The data are best interpreted in terms of a dimeric holoprotein (predicted M _r = 25 kDa) and a monomeric apoprotein (predicted M _r = 12.5 kDa). Moreover, NMR studies of the holo- and apo- forms of Grx C1 indicate that these proteins are present as a dimer and monomer, respectively (11).

To ensure that the Fe–S center incorporation in Grx C1 is not caused by the removal of the 10 first amino acids (see Materials and Methods) and determine whether Fe–S center incorporation is unique to poplar Grx C1 or is a property of all plant Grx containing a CGYC active site, we have also produced the full-length Grx C1 and the Grx C1 homolog from Arabidopsis thaliana (At5g63030, AtGrx C1). Both recombinant holoproteins display the same visible absorption characteristics and have not been characterized further (data not shown).

Nature of the Fe–S Center in Grx C1.

The nature of the Fe–S center in holoGrx C1 was initially assessed by using a combination of analytical and spectroscopic techniques. Samples purified to homogeneity contained 1.1 ± 0.1 mol of Fe per mol of protein, a result that is consistent with the presence of one [2Fe–2S]²⁺ center per dimer or one [4Fe–4S]²⁺ center per tetramer. However, the UV-visible absorption and CD spectra are uniquely characteristic of a [2Fe–2S]²⁺ cluster (Fig. 1 A) and are very similar to those recently reported for human Grx2 (10). There are, however, some significant differences compared with other structurally characterized biological [2Fe–2S]²⁺ centers. For example, the absorption spectrum of Grx C1 is dominated in the 400- to 480-nm region by a single band centered at 430 nm, whereas other biological [2Fe–2S]²⁺ centers generally exhibit two resolved bands centered near 420 and 460 nm (12–17). On the basis of the theoretical and experimental ε₂₈₀ values for the apoprotein (9.6 mM⁻¹cm⁻¹), ε₂₈₀ and ε₄₃₀ values for the [2Fe–2S]²⁺ center are estimated to be 2.6 mM⁻¹cm⁻¹ and 5.0 mM⁻¹cm⁻¹, respectively (A ₄₃₀/A ₂₈₀ ratio = 0.41 ± 0.02). These extinction coefficients are indicative of ≈0.5 [2Fe–2S]²⁺ clusters per monomer based on published extinction coefficients for biological [2Fe–2S]²⁺ centers (18). Hence the UV-visible absorption and Fe analytical data are consistent in estimating one [2Fe–2S]²⁺ cluster per Grx C1 dimer for homogeneous samples with A ₄₃₀/A ₂₈₀ ratios = 0.41 ± 0.02.

In agreement with the anticipated diamagnetic S = 0 ground state that results from antiferromagnetic coupling of two high-spin (S = 5/2) Fe(III) centers, purified samples of holoGrx C1 do not exhibit an observable EPR signal. Attempts to reduce the [2Fe–2S]²⁺ cluster under anaerobic conditions to yield a stable [2Fe–2S]⁺ cluster by using excess dithionite, electrochemical reduction in the presence of redox mediators or deazariboflavin-mediated photoreduction have not been successful. Absorption studies indicate a gradual and irreversible bleaching of the visible absorption and parallel EPR studies revealed a very weak S = 1/2 EPR signal, g _‖ = 2.04 and g _⊥ = 1.93, invariably accounting for <0.02 spins per [2Fe–2S] cluster (Fig. 1 C). Hence the cluster appears to be reductively labile, with reductive degradation proceeding via an unstable S = 1/2 [2Fe–2S]⁺ intermediate. The inability to reduce the cluster to a stable [2Fe–2S]⁺ state argues against an electron transfer function for the cluster in Grx C1.

The vibrational properties of the Fe–S cluster in holoGrx C1 were characterized by resonance Raman studies (19). The resonance Raman spectrum of Grx C1 in the Fe–S stretching region (240–450 cm⁻¹), obtained with 457.9-nm excitation, is shown in Fig. 1 D. The Fe–S stretching frequencies are most similar to those of [2Fe–2S]²⁺ clusters with complete cysteinyl ligation (14, 20), and distinct from [2Fe–2S]²⁺ centers with two histidine ligands at a unique Fe site (21) or with a single serine or arginine ligands in place of one of the cysteine ligands (22–25). The Fe–S stretching modes for the [2Fe–2S]²⁺ centers in Grx C1 are readily assigned by direct analogy with the assignments established via isotope shifts and normal mode calculations for representatives of the all-cysteine-ligated [2Fe–2S]²⁺ clusters from each of the major classes of ferredoxins, i.e., plant-type (typified by Spinacia oleracea ferredoxin), hydroxylase- or Isc-type (typified by adrenodoxin) and thioredoxin-type (typified by Clostridium pasteurianum ferredoxin) (SI Table 1) (14, 20). Hence, taken together, the analytical and spectroscopic properties of the holoform of Grx C1 are consistent with a dimeric protein containing one [2Fe–2S]²⁺ center coordinated exclusively by cysteinyl ligands.

Identification of the Fe–S Cluster Ligands by Site-Directed Mutagenesis.

Mutagenesis studies involving each of the three conserved cysteine residues of Grx C1 were used to investigate the cysteine residues involved in [2Fe–2S]²⁺ cluster ligation. No cluster was observed in the Grx C1 C31S variant, indicating that the catalytic cysteine (Cys-31) is likely to be a cluster ligand. However, both the Grx C1 C34S and C88S variants were able to assemble [2Fe–2S]²⁺ clusters with visible absorption and CD characteristics identical to that of the [2Fe–2S]²⁺ center in holoGrx C1 (data not shown) and resonance Raman spectra very similar to those of holoGrx C1 (Fig. 1 D). On the basis of iron-determinations (0.38 ± 0.04 and 0.19 ± 0.04 Fe/monomer for C88S and C34S, respectively) and A ₄₃₀/A ₂₈₀ ratios (0.16 ± 0.02 and 0.06 ± 0.02 for C88S and C34S, respectively), aerobically purified samples of Grx C1 C88S and C34S variants were found to contain ≈0.4 and ≈0.2 [2Fe–2S]²⁺ clusters per dimer, respectively. As discussed below, the substoichiometric cluster content of the C88S and C34S variants appears to be a consequence of enhanced O₂ sensitivity, suggesting that both Cys-34 and Cys-88 may play a role in stabilizing the [2Fe–2S]²⁺ cluster against oxygen degradation. In light of the resonance Raman evidence for a [2Fe–2S]²⁺ cluster with complete cysteinyl ligation, the mutagenesis results were initially very puzzling but were subsequently rationalized by the crystallographic results.

The stability of the Fe–S cluster in wild-type and mutant forms of Grx C1 was monitored aerobically by measuring the absorption at 430 nm in the presence of various reductants and oxidants as a function of time (Fig. 2). The results are quite similar to those observed for human Grx2 (10), i.e., reduced glutathione stabilizes the cluster in Grx C1 and in the Grx C1 C34S and C88S variants, whereas oxidized glutathione slightly destabilizes the cluster. DTT also stabilizes the cluster but to a lesser extent than reduced glutathione. We took advantage of the stabilizing ability of reduced glutathione to purify holoforms of AtGrx C1, which was less stable than poplar Grx C1 and poplar Grx C1 C34S and C88S in the presence of GSH.

The sequence required for assembling a [2Fe–2S] cluster in Grxs was investigated by expressing proteins with altered active sites. The active sites sequences of Grx C2, C3 and C4 were thus mutated into CGYC and the one of Grx C1 into CPYC. UV-visible absorption studies of samples purified in the presence of GSH showed that Grx C1 G32P was no longer able to incorporate a [2Fe–2S]²⁺ cluster, whereas Grx C2 P24G/F25Y, Grx C3 P38G, and Grx C4 P28G, all assembled a [2Fe–2S]²⁺ cluster with absorption properties very similar to those of Grx C1. Grx C4 P28G was unique in being the only Grx protein investigated in this work that was purified exclusively in the Fe–S cluster-bound form. Analytical and quantitative UV-visible absorption/CD studies of Grx C4 P28G indicated 1.1 ± 0.1 [2Fe–2S]²⁺ cluster per dimer and the absorption, CD, and resonance Raman spectra are very similar to Grx C1, indicating analogous electronic, vibrational, and structural properties for the [2Fe–2S]²⁺ centers (Fig. 1 B and D). Taken together, these mutagenesis data indicate that the presence of a glycine rather than proline after the first catalytic cysteine is an essential determinant for the ability of Grxs to assemble a [2Fe–2S]²⁺ cluster.

The importance of the glycine residue after the first catalytic cysteine, coupled with lack of role for the second active-site cysteine in cluster assembly, raised the possibility that the class of monothiol Grxs with CGFS active sites may also be able to incorporate Fe–S clusters. This possibility is particularly intriguing because the yeast Grx5 protein, which has been shown to be involved in Fe–S assembly (7, 8), has a CGFS active site. The Grx C1 Y33F/C34S with a CGFS active site was found to incorporate a Fe–S center with UV-visible absorption properties similar to those of Grx C1. Although the presence of an Fe–S cluster has not been reported to date for any purified samples of yeast Grxs with a CGFS active site, this may be a consequence of instability under aerobic conditions or a requirement for other as yet unidentified amino acids needed for the Fe–S assembly.

The recent discovery of a [2Fe–2S]²⁺ cluster in human Grx2 (10), which has a CSYC active site, prompted investigation of Grx S12, which contains a similar CSYS active site. The aerobically purified WT protein and the S30G mutant, which contains a CGYS active site similar to Grx C1 C34S, did not incorporate a cluster, suggesting that either serine is not able to play a role identical to the glycine residue or that other determinants not present in the Grx S12 sequence, but in human Grx2, are important for Fe–S cluster incorporation. Human Grx2 has two active-site cysteines (Cys-37 and Cys-40) and two extra cysteines (Cys-28 and Cys-113), not conserved in either Grx S12 or in plant sequences but which can form a disulfide bridge (26). Mutagenesis and spectroscopic results lead to the proposal for a subunit bridging cluster ligated by Cys-28 and Cys-113 (10). However, the great similarity in the UV-visible absorption and CD properties of the [2Fe–2S]²⁺ cluster between human Grx2 and poplar Grx C1, combined with the absence of any mutagenesis results for Cys-37 in human Grx2, suggests that this proposal needs to be reevaluated in light of the results presented herein. In fact, a very recent paper by Johansson et al. (26) describing the x-ray crystallographic structure of human Grx2 fully supports the proposal made here, the human Grx2 is actually organized in a very similar way to poplar Grx C1 (26). The Cys-28 and Cys-113 in human Grx2 may only have a structural role that facilitates Fe–S cluster assembly rather than providing cluster ligands. An important feature that may explain the absence of Fe–S cluster incorporation in Grx S12 is the presence of a tryptophan residue before the catalytic cysteine. All of the plant Grx proteins that have thus far been shown to incorporate an Fe–S cluster have, instead, a tyrosine residue. This possibility is supported by recent structural evidence on poplar Grx C4, which was shown to be in equilibrium between monomeric and dimeric states (27). In the dimeric state, the two active-site tyrosine residues (YCPYC) are involved in the formation of a new interface, which delimits a pocket in which some small cofactors like Fe–S clusters could be incorporated. We hypothesize that the presence of a proline residue prevents cluster incorporation.

Structure of the Holoprotein.

The x-ray structure of Grx C1, solved to 2.1-Å resolution, indicates that the holoprotein organizes as a tetramer containing one [2Fe–2S] cluster in the crystalline state. This result is surprising because gel filtration, NMR, analytical and UV-visible absorption studies of cluster-replete holoprotein all indicate a dimeric structure containing one [2Fe–2S] cluster in solution (11). However, as the holoprotein gradually loses its cluster under air, it seems likely that the tetrameric organization detailed below results from cocrystallization of holo and apo forms. In the x-ray structure, the four molecules of Grx in the asymmetric unit exhibit a unique “dimer of dimer structure” shown in Fig. 3 A. Each polypeptide has a typical Trx fold with an averaged pair-wise rms Cα discrepancy of 0.54 Å. One pair of chains A and B and another pair of chains C and D are related by noncrystallographic quasi 2-fold axes (171.1° and 175.4° rotation, respectively), which are roughly parallel to each other (7.3°) and have a distance of ≈10 Å between them. Dimer A/B and dimer C/D are related by third quasi 2-fold symmetry of 176.9° and 179.4° rotations respectively roughly perpendicular to the first two 2-fold axes.

Multiwavelength anomalous diffraction (MAD) analysis revealed the position of Fe atoms at the center of an electron density island, which connects chains A and C. A model of a [2Fe–2S] cluster fits well with the electron density of the metal center. The two Fe atoms of the [2Fe–2S] cluster are linked to the Cys-31 Sγ atom of chains A and C and a compound not assigned to any of the polypeptide side-chains in the electron density map, suggesting the existence of nonprotein ligands. As a consequence, the A–C structure is likely to provide a good representation of the holodimer in solution. The nonprotein ligand was separated from the polypeptide chains by HPLC after an acid treatment of Grx and chemically identified by MALDI-TOF and NMR as GSH, even though the preparation used for crystallography was made in absence of glutathione. In fact, a GSH model fits well into the electron density map, revealing that the two Fe atoms are coordinated by two inorganic sulfur atoms of the cluster, two Cys-31 Sγ atoms, and two Cys Sγ atoms of GSH. In the overall structure, chains A and C, one [2Fe–2S] cluster, and two glutathione molecules are arranged obeying a noncrystallographic 2-fold symmetry, the axis of which passes through the two inorganic sulfur atoms of the [2Fe–2S] cluster. Tyr 30, Tyr 33, and GSH molecules in chains A and C surround the [2Fe–2S] cluster, sequestering it from the solvent. Hence, both tyrosines in the YCGYC active-site sequence of Grx C1 play a role in stabilizing the cluster (see details of the Fe–S cluster environment in Fig. 3 B). The structure is consistent with the mutagenesis and spectroscopic properties of holoGrx C1.

In an earlier work by Feng et al., we have determined the 3D structure of the reduced apo form of Grx C1 by NMR spectroscopy (11). In addition, by comparing the resonances of the holodimeric form to those of the apo form, we have proposed a model in which the iron sulfur center is coordinated by two cysteines and two external glutathione molecules. In general, the crystallographic data presented here match well those of the Feng paper, and, in particular, the organization of the subunit is similar to that described by NMR (11). The crystallographic structure, however, provides near atomic resolution of the iron–sulfur center and indicates that the model was only partly correct. We had earlier hypothesized that the two Cys residues as well as the two GSH ligands were located in cis with respect to the plane created by the two iron atoms. The data presented here reveal that they are actually placed in trans in the real structure.

In our crystal, Grx chains A–D are each bound to one of four GSH molecules through hydrogen bonds, salt bridges, and hydrophobic interactions except that two of the GSH molecules are covalently linked to the [2Fe–2S] cluster. The four GSH molecules share similar conformations and have essentially the same relative position with respect to their associated Grx molecules. Two GSH molecules in complex with chains B and D face the Cys Sγ atoms of the catalytic cysteine residues at a distance of ≈3 Å, without forming covalent bonds. Chains B and D do not directly interact with the [2Fe–2S] cluster but serve to stabilize the tetramer structure.

Among the poplar subgroup I Grxs, only Grx C1 forms an iron–sulfur cluster. Inspection of the structure indicates that the presence of a proline residue adjacent to the catalytic cysteine, as in Grx C2, C3, or C4, would likely interfere with the cluster formation because of steric constraints involving the side chain. In light of the role played by the two tyrosine residues, the presence of a small residue, and especially a glycine, is likely to be essential for Fe–S cluster incorporation.

Oxidant Properties of Grx C1.

In the bacterial periplasm, the formation of disulfide bonds is catalyzed by the soluble enzyme DsbA, which is recycled by the membrane protein DsbB that, in turn, transfers electrons to quinones and the respiratory chain (3). Mutational inactivation of the dsbB gene results in several phenotypes including complete loss of motility (28). Following up on recent studies on a mutated Trx with a CACA active site that incorporates a [2Fe–2S]²⁺ cluster (5), we fused Grx C1 to the TorA leader peptide that directs export across the cytoplasmic membrane via the twin arginine transporter (TAT). The TAT pathway allows the membrane translocation of proteins that have acquired metal cofactors and/or disulfide bonds in the cytoplasm (3, 29). The TorA-Grx C1 fusion was exported into the periplasm and restored motility to a significant extent (Fig. 4 A). To further examine the ability of exported Grx C1 to mediate the formation of disulfide bond formation in the periplasm, we monitored the activity of alkaline phosphatase (PhoA), an enzyme that contains two disulfide bonds that are essential for folding into the active native conformation (30). PhoA activity is completely abolished in dsbB/dsbA-deleted cells (a decrease of >100-fold over the WT parental strain) (Fig. 4 B). However, expression of TorA-Grx C1 restored the level of PhoA activity to that of WT MC1000 strain. Secretion of Grx C1 via the PhoA signal sequence is targeted to the general secretory pathway, (SEC), whereby proteins are exported in an essentially unfolded conformation also restored PhoA activity in E. coli LM108. In contrast to the TAT pathway, the SEC pathway is unable to export proteins with cofactors such as iron sulfur clusters. All of the Grxs tested (poplar Grx C1 or Grx C1 G32P, Grx C4, and E. coli Grx1 and -3) were able to promote disulfide bond formation in AP (Fig. 4 B and data not shown), indicating that the apo form of Grx C1 and probably more generally all dithiol Grxs are able to promote disulfide bond formation in vivo. Concerning the holo Grx C1, the restoration of AP activity could arise from apoprotein generation because of the instability of the cluster. These results were surprising because the cells lack DsbB, and there is no known periplasmic oxidant for the recycling of reduced Grx C1 after it transfers a disulfide bond to a substrate protein. The best explanation is linked to the postulated presence of GSSG in E. coli periplasm (31).

One way to understand this oxidant property is to measure the redox potential of the protein. Both the Grx C1 holo- and apoproteins exhibit redox-active, two-electron processes when titrated by using the thiol-specific reagent monobromobimane (mBBr) to monitor the oxidation state of the proteins. Oxidation–reduction midpoint potential (E_m) values of −160 ± 10 mV and −180 ± 10 mV were measured at pH 7.0 for the holoprotein and apoprotein, respectively. Given the two-electron nature of these redox couples and the fact that mBBr detection was used to monitor the redox process, it seems highly likely that these E_m values, which are slightly more positive than the E_m values previously reported for Grx1 (−233 mV) and Grx3 (−198 mV) of E. coli (32), arise from disulfide/dithiol couples. These more positive values for plant Grxs may be a general feature, because the redox potential of Grx C4 also lies in the same range with an E_m value of −170 mV ± 10 mV. The redox process detected with the holoprotein certainly does not reflect redox cycling of the Fe–S center, which is essentially inert under the conditions of these redox titrations. It rather results from the disulfide/dithiol redox activity of apo form generated after interaction with mBBr during the 2-h period required for protein equilibration in the redox buffer. Indeed, monitoring the Fe–S center spectrum in the presence of mBBr indicates that it is almost completely degraded in this time (data not shown). Assuming this rather oxidant redox potential for Grx C1, we also have measured the reductase activity of apo Grx C1 using classical Grx substrates, hydroxyethyl disulfide and dehydroascorbate. The catalytic parameters are similar to those found with other poplar Grxs (data not shown).

Localization and Physiological Role of Grx C1.

The function of Grx C1 in plants is clearly of considerable interest because of its ability to bind an Fe–S cluster. A redox function seems very unlikely in light of the inability of the [2Fe–2S]²⁺ cluster to undergo reversible redox cycling. To address the function of Grx C1, we first determined the localization of the protein, by cloning in-frame the entire sequence of Grx C1 before the GFP protein, and this construction was used to bombard leaf cells from Nicotiana benthamiana. As shown in SI Fig. 6, the fluorescence associated with the construction is localized to the cytosol.

Previous studies have implicated some Grxs in the maturation of Fe–S clusters. In vitro experiments demonstrated that E. coli Grxs in the presence of GSH facilitate assembly of a [4Fe–4S] cluster in the fumarate nitrate reductase regulatory protein by reducing disulfide bridges formed between cluster-ligating cysteines in the apoprotein (6). Yeast Grx5, which has a CGFS active-site sequence that is likely to be able to assemble an Fe–S cluster based on the results reported above, has been shown to play an important role in mitochondrial Fe–S cluster biogenesis (7, 9). Other prokaryotic and eukaryotic monothiol Grxs with CGFS active-site sequences have recently been shown to be effective functional substitutes for yeast Grx5 (8). Although the specific role of yeast Grx5 in Fe–S cluster biogenesis has yet to be elucidated, ⁵⁵Fe radiolabeling studies of knockout mutants suggest that it is likely to facilitate the transfer of clusters preassembled on the Iscu1p scaffold protein into acceptor proteins (9). The recent discovery that Grx5 is also required for vertebrate heme synthesis (33) is particularly interesting in this regard, because it raises the possibility that cluster-bound Grx5 plays a direct role in regulating heme biosynthesis in mammals by facilitating assembly of a [4Fe–4S] cluster on iron regulatory protein 1 or activating ferrochelatase by insertion of the catalytically essential [2Fe–2S] cluster (34).

In plants, components of the Fe–S clusters assembly machinery have been described in both mitochondria and plastids, but nothing is known concerning the assembly of cytosolic Fe–S cluster proteins (35). In yeast, where GSH was identified as being important for cytosolic Fe–S cluster maturation, at least six different proteins have been implicated as possibly being involved in cluster assembly (36). As Grx C1 coordinates a [2Fe–2S] cluster by using two external glutathione molecules, an intriguing hypothesis would be that Grx C1 could serve as a scaffold for assembly of cytosolic Fe–S clusters or a chaperone for transfer of Fe–S clusters assembled on scaffold proteins to cytosolic acceptor proteins. In vitro experiments to test this hypothesis by using appropriate Fe–S cluster acceptor and donor proteins remain to be completed.

The results presented herein clearly demonstrate that apo Grx C1 is functional in disulfide bond formation. It is currently unclear whether this functionality is maintained in holo Grx C1, but this seems unlikely because the active-site cysteine is a cluster ligand. Moreover, human Grx2, which has a [2Fe–2S] cluster with analogous spectroscopic and stability properties, was shown to be inactive in classical Grx assays. Hence the holo- and apo- forms of Grx C1 can have different functions within the cell, and it is possible that the [2Fe–2S] cluster has dual roles in Fe–S cluster biosynthesis and in sensing the redox status of the cell to activate the disulfide oxidoreductase activity of Grx C1 during conditions of oxidative stress. The latter role was suggested for the [2Fe–2S] cluster in human Grx2, because it appeared to be required only under conditions of oxidative stress (10). Although the presence of a similarly coordinated Fe–S cluster was shown in mammalian cells for Grx 2, there is still no evidence that this cluster is present in plant cells. In vivo studies with plants are planned to address the putative role of the cluster as a sensor of oxidative stress.