Crystal structure of the antioxidant enzyme glutathione reductase inactivated by peroxynitrite.

As part of our studies on the nitric oxide-related pathology of cerebral malaria, we show that the antioxidative enzyme glutathione reductase (GR) is inactivated by peroxynitrite, with GR from the malarial parasite Plasmodium falciparum being more sensitive than human GR. The crystal structure of modified human GR at 1.9-A resolution provides the first picture of protein inactivation by peroxynitrite and reveals that this is due to the exclusive nitration of 2 Tyr residues (residues 106 and 114) at the glutathione disulfide-binding site. The selective nitration explains the impairment of binding the peptide substrate and thus the nearly 1000-fold decrease in catalytic efficiency (k(cat)/K(m)) of glutathione reductase observed at physiologic pH. By oxidizing the catalytic dithiol to a disulfide, peroxynitrite itself can act as a substrate of unmodified and bisnitrated P. falciparum glutathione reductase.

Nitric oxide (NO) is a pluripotent molecule that is involved in both cytoprotective and cytotoxic processes (1)(2)(3)(4). In the latter case, the reaction of NO with metabolically generated reactive oxygen species exacerbates damage due to the formation of more potent agents such as peroxynitrite and nitrosoperoxycarbonate (5)(6)(7).
Peroxynitrite/peroxynitrous acid (ONOO Ϫ /ONOOH; pK a ϭ 6.8) is produced by activated macrophages and other cells via the reaction of NO with superoxide (O 2 . ) at diffusion-controlled rates, and it easily crosses biological membranes (7)(8)(9). In vivo, peroxynitrite is estimated to have a half-life of Ͻ20 ms, primarily due to its reaction with CO 2 and with target molecules, but also to the proton-or methionine-catalyzed isomerization to nitrate (10,11). Nitrosoperoxycarbonate (ONOOCO 2 Ϫ ), formed from ONOO Ϫ and CO 2 , has an even shorter half-life. (6). Depending inter alia on the distance between the source of peroxynitrite and the target structure, the presence of CO 2 can lead to potentiation or reduction of ONOO Ϫ reactivity (10,12).
Peroxynitrite and its derivatives are powerful modifiers of proteins; the effects include nitrosation of cysteine; oxidation of cysteine, tryptophan, and methionine; and nitration of tyrosine (2,4,8). For an amazing variety of human diseases, high levels of nitrotyrosine are found in extra-or intracellular protein pools, which strongly implicates peroxynitrite as a pathophysiological agent (4). In this context, antioxidant enzymes (proteins that deal with the detoxification of the superoxide radical (O 2 . ) and its metabolic derivatives such as H 2 O 2 and glutathione disulfide) have attracted special attention (13)(14)(15)(16)(17)(18). With cerebral malaria, protein nitration in the brain is a typical post-mortem finding (19), which suggests that NO and its derivatives contribute to the pathogenesis and clinical outcome of the disease (3, 19 -22). For studying the NO-and peroxynitrite-related molecular pathology of cerebral malaria, we have chosen the antioxidative enzyme human glutathione reductase (hGR) 1 and its counterpart from the malarial parasite Plasmodium falciparum (PfGR). The intracellular flavoenzyme hGR (EC 1.6.4.2) recycles oxidized glutathione (GSSG) to maintain high levels of reduced glutathione (GSH) by catalyzing the reaction GSSG ϩ NADPH ϩ H ϩ º 2 GSH ϩ NADP ϩ . Due to the fundamental role of GSH in scavenging and removal of deleterious reactive oxygen species, GR plays a crucial part in the antioxidative defense mechanisms of the cell (3,(22)(23)(24). GR is a possible target of NO carrier molecules by virtue of its inhibition in vitro by S-nitrosoglutathione (see Ref. 3), diglutathionyldinitrosoiron complex (see Ref. 3), and peroxynitrite (25,26). Inhibition by S-nitrosoglutathione and diglutathionyldinitroso-iron complex was shown to occur via oxidation of the catalytically essential Cys 63 to a sulfenic acid (R-SOH) and a sulfinic acid (R-SOOH), respectively (3). With regard to inhibition by peroxynitrite, hGR has not yet been studied. Bovine GR, however, was shown to be nitrated at 2 tyrosine residues/protein subunit, a modification that could be prevented by the substrate GSSG (26). It was hypothesized that the 2 residues were the equivalents of Tyr 106 and Tyr 114 in the structurally known hGR.
Here, we report the first picture of enzyme inactivation by peroxynitrite through a crystallographic analysis of modified hGR at 1.9-Å resolution. Furthermore, we show that PfGR (but not hGR) is an effective catalyst of ONOOH-dependent oxidation of NADPH.

EXPERIMENTAL PROCEDURES
Materials-Peroxynitrite was synthesized from sodium nitrite and hydrogen peroxide using established protocols (27,28) and was stored at Ϫ80°C. Recombinant hGR, recombinant hGR with a Y114L mutation, and recombinant PfGR were prepared as described (29). Enzyme kinetic studies were carried out at 25°C in 47 mM potassium phosphate, 200 mM KCl, and 1 mM EDTA (pH 6.9) or in 100 mM potassium phosphate and 0.1 mM DPTA (pH 7.4).
Exposure of Cells and Enzymes to Peroxynitrite-Peroxynitrite was diluted with ice-cold 0.1% KOH to stock solutions ranging from 100 M to 25 mM (⑀ 302 nm ϭ 1.67 mM Ϫ1 cm Ϫ1 ), and ONOO Ϫ additions were made under continuous and vigorous stirring either by microbolus injection or, for steady-state experiments, by infusion with a micropump at a rate of 1 ml/min (8). Because of protein instability, the samples could not be vortexed as recommended in Ref. Absorption Spectroscopy-Flavin spectra of GR at 10 -20 M subunit concentrations in 100 mM potassium phosphate and 0.1 mM DPTA (pH 7.4) were recorded at 25°C. The ⑀ values used for hGR (28) and PfGR (30) were ⑀ 463 ϭ 11.5 and ⑀ 540 ϭ 0.3 mM Ϫ1 cm Ϫ1 for the oxidized enzyme (GR-S 2 ), ⑀ 540 ϭ 4.4 mM Ϫ1 cm Ϫ1 for the complex of reduced enzyme with NADPH (GR-(SH) 2 ⅐NADPH), and ⑀ 423 ϭ 4.2 mM Ϫ1 cm Ϫ1 for nitro-Tyr in GR. The latter value was pH-insensitive between pH 7.4 and 8.4 (see Ref. 26 and inset in Fig. 2).
Peroxynitrite as a Substrate-These assays were carried out at 25°C in 100 mM potassium phosphate and 0.1 mM DPTA titrated with 2 M KOH to pH 8.5. The assay mixture contained 0.5 M GR, varying initial concentrations of peroxynitrite (50 -1000 M), and 100 M NADPH. Because the reaction product (probably nitrite) absorbs at 340 nm, NADPH oxidation could not be used to monitor reaction progress. Instead, the decrease in the peroxynitrite concentration was followed, the ⑀-value for this compound being 1.67 mM Ϫ1 cm Ϫ1 at 302 nm (8,27,28). Alternatively, the reaction was quenched after 20, 40, 60, or 80 s by addition of 1 mM dithiothreitol; subsequently, the NADP ϩ formed in the enzymatic reaction was measured using the glucose-6-phosphate dehydrogenase reaction (30). Bisnitro-GR was found to have the same peroxynitrite reduction activity as the unmodified enzyme, implying that the nitration of GR needs not to be accounted for in the analysis of the kinetics.
Crystallization of Peroxynitrite-treated hGR-We produced inhibited enzyme by repeatedly treating hGR (1 nmol/ml subunit) with 100 M peroxynitrite (initial concentration) in 1-min intervals. When the desired degree of inactivation (75 or 98%, respectively) was reached, the samples were concentrated. Absorption spectroscopy at 423 nm showed a total of 1.6 (2.0) nitrotyrosines/hGR subunit for the 75% (98%) inhibited enzyme. For crystallization, hanging drops of 10 l containing 10 mg/ml protein, 120 mM ammonium sulfate, and 100 mM potassium phosphate (pH 7.5) were equilibrated against reservoirs containing 720 mM ammonium sulfate and 100 mM potassium phosphate (pH 8.0) at 30 or 37°C. The 98% inhibited enzyme gave small crystals of poor quality, which were not pursued further. However, the 75% inhibited enzyme yielded crystals of 0.5 ϫ 0.4 ϫ 0.3 mm 3 within 5 days. These crystals were isomorphous to those of unmodified hGR (31). For storage and handling, they were transferred to a stabilizing solution containing 1 M ammonium sulfate and 100 mM potassium phosphate (pH 8.0). The crystals belong to space group C2, with unit cell parameters a ϭ 119.75, b ϭ 63.65, and c ϭ 84.71 Å and ␤ ϭ 58.42°.
Data Collection and Refinement-X-ray data were collected on R-axis IV image plate detectors (Molecular Structure Corp.) with a Rigaku RUH3R rotating anode (CuK ␣ ) operating at 50 kV and 100 mA with a 0.3-mm collimator. X-ray data were processed (see Table II) with the programs Denzo and Scalepack (32) and NovelR (33). The refined structure of the unmodified enzyme (Ref. 34 and Protein Data Bank code 3GRS) was used as the starting model after the coordinates had been altered to comply with our assignment of the space group to C2. Electron density map calculations and crystallographic refinements were carried out with X-PLOR Version 3.851 using data between 20-and 1.9-Å resolution (F Ͼ 1) corrected for bulk solvent contributions and diffraction anisotropy (35). Bond length and angle parameters for a 3-nitrotyrosine side chain were derived from the crystal structure of 4-hydroxy-3-nitrophenylalaninium nitrate (36).

Inspection of electron density maps with coefficients
showed that the 2 active-site tyrosine residues, Tyr 106 and Tyr 114 , were modified. There was no evidence for the modification of any other residue.
In the case of Tyr 106 , we found clear evidence for a nitro group at the ortho-position (3 or C-⑀1) of the phenol group (see Fig. 3a). However, a more careful examination of the 2F o Ϫ F c electron density map suggested a mixture of wild-type Tyr 106 and nitrated Tyr 106 . This was further confirmed when preliminary positional and individual B-factor refinements with nitrated Tyr 106 at full occupancy produced an average B-factor of ϳ30 Å 2 for phenol ring atoms and of ϳ50 Å 2 for atoms of the 3-nitro group. To estimate the nitrated Tyr 106 /wild-type Tyr 106 ratio, test refinements consisting of 30 cycles of conventional positional and 30 cycles of restrained B-factor refinements were carried out with nitrated Tyr 106 at occupancies of 0.5, 0.6, 0.7, and 0.8, whereas the native conformation of Tyr 106 was maintained as an alternate conformation at a complementary occupancy. The goal of this approach was (i) to find those combinations of occupancies that would produce near wild-type B-values for the wild-type conformation (ϳ20 Å 2 ), and uniform B-values for all atoms in all other conformations and (ii) to minimize peaks in F o Ϫ F c electron density maps. This approach led to the choice of 0.7 nitrated Tyr 106 residues and 0.3 wild-type Tyr 106 residues.
In the case of nitrated Tyr 114 , the 2F o Ϫ F c electron density was ill formed, suggesting multiple conformations. Indeed, both the native conformation and two other conformations with favorable torsion angles could be delineated. Furthermore, positive peaks in F o Ϫ F c electron density maps near the C-⑀1 positions of the phenol rings of the two new conformations suggested that they represented 3-nitrotyrosine side chains. The occupancy of the side chain at position 114 was distributed among the three observed conformations (one for wild-type Tyr 114 and two for nitrated Tyr 114 ), and test refinements were carried out in a manner analogous to the procedure described for Tyr 106 . A model with occupancies of 0.2 for native Tyr 114 and of 0.4 for each of the nitrotyrosine conformers best met the above criteria and was accepted as the optimum.
Attempts to determine the structure of a complex of bisnitro-GR with GSSG were made by incubating crystals in stabilizing solution of pH 8.0 containing either 10 or 20 mM GSSG for 48 h. Examination of F o Ϫ F c and 2F o Ϫ F c electron density maps at 2.0-Å resolution yielded no indication for bound GSSG. We estimate that the limit of sensitivity of the difference Fourier is an occupancy of ϳ20%.

Inactivation of Glutathione Reductase by Tyrosine Nitration-
The peroxynitrite inhibition of hGR and PfGR ( Fig. 1 and Table I) is qualitatively similar to that seen for bovine GR (26), with enzyme inactivation being proportional to nitrotyrosine formation ( Figs. 1 and 2). However, as judged from their catalytic competence (k cat /K m ), hGR and PfGR are Ͼ300-fold more strongly inhibited than bovine GR at the same degree of nitration of 2 Tyr residues/subunit (Table I).
Structure determination revealed that both Tyr 106 and Tyr 114 are modified (see below). To distinguish their roles, we studied the previously characterized enzymatically active mutant Y114L (29). When exposed to peroxynitrite, the mutant lost Ͻ2% of its activity; and as judged from the absorbance change at 423 nm, Ͻ0.1 nitro group was incorporated per enzyme subunit. This unexpected result implies that the modification of Tyr 106 depends on the presence of Tyr 114 or, more likely, of nitro-Tyr 114 .
In the presence of 27 mM bicarbonate (which corresponds to 1.2 mM CO 2 ) at pH 7.4, the degree of inhibition by peroxynitrite was found to be 10-fold less both for hGR and PfGR. This suggests that peroxynitrite itself and not nitrosoperoxycarbonate was the modifying species (6,12). An alternative interpretation is that stirring of the solution was too slow relative to net ONOO Ϫ decomposition by CO 2 (2,6); this possibility will be studied using a stopped-flow apparatus for rapid mixing.
Peroxynitrite as a Substrate-In the oxidized form, GR is characterized by an active-site disulfide (GR-S 2 ); and in the NADPH-reduced form, it is distinguished by an active-site dithiol (GR-(SH) 2 or GR-(SH) 2 ⅐NADPH complex). GR-(SH) 2 ⅐NADPH can be clearly distinguished from GR-S 2 by its absorbance at 540 nm (Fig. 2). In the physiologic reaction, the dithiol of GR-(SH) 2 is reoxidized by dithiol-disulfide exchange with GSSG (29,30,39). Because peroxynitrite is known to oxidize thiols to disulfides (38), we studied whether peroxynitrite can replace GSSG as a substrate according to Equation 1.
As an initial test, GR-S 2 was incubated with NADPH in phosphate buffer (pH 7.4), which expectedly led to an absorbance band at 540 nm (Fig. 2). Upon addition of peroxynitrite, this absorbance disappeared, indicating that the active-site dithiol was reoxidized to the disulfide. The titration cycle (NADPH followed by peroxynitrite and NADPH again) could be repeated Ͼ10 times, although the spectrum of GR-S 2 revealed that the enzyme had been nitrated during these cycles. This implies that both bisnitro-GR and unmodified GR catalyze the reduction of peroxynitrite by NADPH according to Equation 1. To estimate the reaction rates, we conducted steady-state assays at pH 8.5; at this pH, the spontaneous decay of peroxynitrite is slow enough to allow accurate measurements. The k cat values for hGR and PfGR were found to be 10 and 126 min Ϫ1 , respectively, with the apparent K m values of peroxynitrite being ϳ70 M for both proteins.

Structure of hGR Modified by Peroxynitrite-Recombinant
hGR inactivated by 75% with peroxynitrite (bisnitro-GR) was crystallized and structurally characterized at 1.9-Å resolution ( Table II). The electron density maps revealed new density for Tyr 106 and Tyr 114 in the GSSG-binding site of hGR and no unusual electron density features near any other residue. Specifically, there were no indications of dityrosine formation or sulfur oxidation. The selective nitration is consistent with the proposition that protein tyrosine nitration depends on the microenvironment of the targeted tyrosine residues (40). For both modified residues, we interpreted the observed electron density as a mixture of nitrated and wild-type tyrosines, and we were able to estimate the relative amount of each chemical species using systematic structural refinements that tested the plausibility of various models (see "Experimental Procedures"). Position 106 can be described well in terms of a ratio of 0.7:0.3 of nitrated Tyr 106 versus wild-type Tyr 106 , with the overall conformation of nitrated Tyr 106 being very similar to that of the native conformation of Tyr 106 (Fig. 3, a and b). The nitro group points away from the core of the protein structure toward Arg 109 , which provides for a favorable electrostatic interaction. Furthermore, nitrated Tyr 106 maintains its packing against Ala 409Ј and Met 406Ј of the other subunit at the dimer interface, and there is no indication that the dimer interface is perturbed in this region or elsewhere.
In the case of position 114, we were able to delineate two distinct conformations of nitrated Tyr 114 that differ from the native conformation of Tyr 114 , along with some residual unmodified Tyr 114 (Fig. 3c). One of the two orientations of nitrated Tyr 114 directs the nitro group into the bulk solvent region, whereas the other directs this group to the interior of the catalytic site, where it makes hydrogen bonds with active-site water molecules. A satisfactory model for this residue is a fractional ratio of 0.4:0.4:0.2 for nitrated Tyr 114a / nitrated Tyr 114b /wild-type Tyr 114 . Taken together, these crystallographic estimates add up to 1.5 nitrotyrosines, in good agreement with the 1.6 nitrotyrosines/hGR subunit derived from spectroscopic analysis (see "Experimental Procedures" and Fig. 2). The initial peroxynitrite concentrations given here can be correlated with time of exposure to much lower steady-state concentrations (28). The number of nitrotyrosyl residues/hGR subunit (OE) was determined on the basis of the absorbance increase at 423 nm shown in Fig. 2. The presence of 100 M NADPH had no influence on the rate and extent of tyrosine modification. 300 M GSSG, however, protected the enzyme from nitrating inactivation; only 0.3 NO 2 groups were incorporated per subunit. Impairment of Peptide Substrate Binding-The structural data suggest that the major functional impairment in bisnitro-hGR is due to decreased affinity for GSSG as opposed to hindering an electron transfer step in catalysis. Evidence for this is that the K m of GSSG at pH 7.4 is Ͼ20-fold increased over that for the unmodified enzyme (Table I). In addition, soaks of bisnitro-GR crystals in 20 mM GSSG failed to produce binding, whereas fixation of the substrate at high occupancy occurs in crystals of unmodified hGR (23,34). The presence of nitrated forms of tyrosine residues can affect the productive binding of GSSG by imposing steric and, as emphasized here, electrostatic effects. Based on the pK a of 7.0 -8.0 reported for 3-nitrotyrosine residues in proteins (2, 41, 42), nitrated Tyr 106 and nitrated Tyr 114 will be largely in the negatively charged nitrophenolate form at the physiologic pH of 7.4. Charge repulsion between the modified enzyme and its peptide substrate would also explain the change in the K m of GSSG with increasing pH: the K m rises from 820 M at pH 6.9 to Ͼ5000 M at pH 8.0, whereas the K m for the unmodified enzyme is pH-independent in this range.

What Agent Is Responsible for Tyrosine Nitration?-Regard-
ing the mechanism of hGR modification by peroxynitrite, our results imply that it involves a negatively charged or neutral nitrating species because (i) there are no negatively charged residues near the nitrated tyrosines; (ii) the GSSG site of hGR discriminates against positively charged ligands (43); and (iii) the rate of tyrosine nitration in GR is greater than for tyrosine peptides, 2 which indicates a rate-promoting factor, such as local surface charge, in the protein (37,40). Consequently, negatively charged nitrating species such as ONOO Ϫ itself and its reactive adducts are plausible nitrating agents of hGR in vivo, as is the neutral nitrogen dioxide radical that results from the homolytic cleavage of peroxynitrite (44). Such a mechanism contrasts with the proposition that protein tyrosine nitration is caused by a positively charged nitronium (NO 2 ϩ )-like ion, as has been proposed for transition metal ion-catalyzed reactions of peroxynitrite (37,40,45,46). It is possible, however, that Tyr 106 is nitrated by a positively charged species if the nitration of Tyr 114 occurs first, thus creating a local negative charge and changing the electrostatics of the active-site region.
Such introduction of a negative charge at the peroxynitritemodified residue may represent a relevant mechanism also in the pathology of cell signaling (47,48). In signaling cascades, irreversible nitration can affect the same tyrosine residues which are reversibly phosphorylated (45,(47)(48)(49)(50). The two modifications have in common that the generated nascent negative charge has an impact on the conformation as well as on the binding behavior of the affected protein.
Cystine Enzymes as Peroxynitrite Reductases?-PfGR represents an enzyme whose active-site dithiol catalyzes peroxynitrite reduction, probably to nitrite, with a k cat /K m ratio of Ͼ3 ϫ 2 M. Scheiwein, unpublished data.
c R free ϭ R cryst for 5% of reflections against which the model was not refined (35). d Root mean square deviation.
FIG. 3. Crystallographic evidence for the nitration of Tyr 106 and Tyr 114 . a, edge-on stereo view of nitrated Tyr 106 (red) and wild-type Tyr 106 (blue) in the final refined model of bisnitro-GR at 1.9-Å resolution superposed to a 1.9-Å resolution and 2F o Ϫ F c electron density contoured at 1.2. Residues of the other subunit are shown in green. b, stereo view of the face of nitrated Tyr 106 (NIY 106), indicating the interaction of the nitro group with the guanidino group of Arg 109 . c, stereo view of nitrated Tyr 114 /wild-type Tyr 114 showing the three modeled conformations. The two conformations for nitrated Tyr 114 have their nitrophenol groups flat in the plane of the paper (red), whereas the ring of Tyr 114 in wild-type hGR is rotated ϳ60°from this plane (blue). Shown is a detail of the final refined model of bisnitro-GR at 1.9-Å resolution superposed to a 1.9-Å resolution and 2F o Ϫ F c electron density contoured at 1. d, location of Tyr 106 and Tyr 114 in the GSSG-binding site of hGR (Ref. 31 and Protein Data Bank code 1GRA). GSSG (blue) and the 2 tyrosine residues (red) are shown along with other activesite residues such as the catalytic Cys 58 and Cys 63 (violet). 10 4 M Ϫ1 s Ϫ1 (Equation 1). This enzyme activity reflects the chemical fact that peroxynitrite can oxide dithiols to disulfide, but not to higher oxidation states of sulfur (38). As a consequence, peroxynitrite reduction activity is probably an intrinsic property of all enzymes that oscillate between a dithiol and a disulfide form during the catalytic cycle. The list of such enzymes includes thioredoxin reductase and thioredoxins, lipoamide dehydrogenase, trypanothione reductase, as well as sulfhydryl oxidase and asparagusate dehydrogenase (24). The comparison of PfGR with the 12-fold less active hGR already indicates that the catalytic efficiency of peroxynitrite reduction might vary greatly among the listed cystine enzymes.
Pathophysiological and Medicinal Implications-Peroxynitrite is implicated in the pathophysiology of cerebral malaria (19 -21) and other central nervous system disorders (16 -18, 51-54). The effects of peroxynitrite as well as of peroxides are aggravated when the intracellular level of glutathione is decreased (54 -56). In a rat model simulating early events in the pathogenesis of Parkinson's disease, Barker et al. (25) observed that 60% depletion of GSH results in a decrease in GR activity. The peroxynitrite inhibition of GR in that report agrees with our results with the human enzyme. These findings are consistent with the notion that GSH is a scavenger of peroxynitrite. A lowered GSH level leads to increased inactivating nitration of GR. As a consequence, GSSG reduction is slowed down, and the competing reaction (GSSG excretion from the cell) establishes a vicious circle leading to lower GSH levels and lower GR activity (56).
The hypothesis that the nitration of hGR and PfGR is involved in the pathogenesis of cerebral malaria can now be addressed by analyzing post-mortem tissues of cerebral malaria victims with respect to content and distribution of the modified enzyme species. By analogy to the studies on ␣-synuclein (57), a further step would be the analytical use of antibodies that specifically recognize the bisnitrated forms of host and parasite enzymes.
The findings reported here also suggest strategies for developing irreversible chimeric PfGR inhibitors as lead compounds for new antimalarial drugs (58). PfGR is inhibited specifically by methylene blue, which apparently binds to the intersubunit cavity of the enzyme (30). As this drug-binding cavity is connected with the two active sites by intramolecular tunnels, work is in progress to construct a tripartite chimeric compound consisting of methylene blue, a chemical linker that snugly fits the chemistry of the tunnel, and a peroxynitrite donor (59). Methylene blue and the linker guarantee the specificity for the parasite enzyme target, and the peroxynitrite donor is expected to irreversibly modify the active sites.