Assessing the Role of the Active-site Cysteine Ligand in the Superoxide Reductase from Desulfoarculus baarsii*

Superoxide reductase is a novel class of non-heme iron proteins that catalyzes the one-electron reduction of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{O}_{2}^{{\bar{.}}}\) \end{document} to H2O2, providing an antioxidant defense in some bacteria. Its active site consists of an unusual non-heme Fe2+ center in a [His4 Cys1] square pyramidal pentacoordination. In this class of enzyme, the cysteine axial ligand has been hypothesized to be an essential feature in the reactivity of the enzyme. Previous Fourier transform infrared spectroscopy studies on the enzyme from Desulfoarculus baarsii revealed that a protonated carboxylate group, proposed to be the side chain of Glu114, is in interaction with the cysteine ligand. In this work, using pulse radiolysis, Fourier transform infrared, and resonance Raman spectroscopies, we have investigated to what extent the presence of this Glu114 carboxylic lateral chain affects the strength of the S—Fe bond and the reaction of the iron active site with superoxide. The E114A mutant shows significantly modified pulse radiolysis kinetics for the protonation process of the first reaction intermediate. Resonance Raman spectroscopy demonstrates that the E114A mutation results in both a strengthening of the S—Fe bond and an increase in the extent of freeze-trapping of a Fe-peroxo species after treatment with H2O2 by a specific strengthening of the Fe—O bond. A fine tuning of the strength of the S—Fe bond by the presence of Glu114 appears to be an essential factor for both the strength of the Fe—O bond and the pKa value of the Fe3+-peroxo intermediate species to form the reaction product H2O2.

Superoxide reductase (SOR) 4 catalyzes the one-electron reduction of O 2 . to H 2 O 2 , providing an antioxidant defense in some anaerobic or microaerophilic bacteria (1)(2)(3)(4)(5): Historically, the only enzymes known to eliminate superoxide were superoxide dismutases, which catalyze the disproportionation of superoxide (6). The recently discovered SOR activity involving the reduction of superoxide has changed our view of how toxic superoxide is eliminated in cells and how this activity protects the cell against oxidative stress and lethal levels of superoxide. SORs are a novel class of non-heme iron proteins that can be classified into one-iron proteins, which possess only the activesite iron center (called Center II) (7)(8)(9)(10)(11)(12), and two-iron proteins (2,13,14), which possess an additional rubredoxin-like Fe 3ϩ -(SCys) 4 , center (called Center I) for which the function and role are not known (9,15). In the reduced state, the SOR active site consists of a non-heme Fe 2ϩ center in an unusual [His 4 Cys 1 ] square pyramidal pentacoordination (7,13,14). The sixth coordination site of the Fe 2ϩ center is vacant and suggests the most obvious site for O 2 . binding. The SOR active site Fe 2ϩ center reacts specifically at nearly diffusion-controlled rates with O 2 .
according to an inner sphere mechanism (16 -21). Although one or two reaction intermediates have been proposed depending on the enzyme studied, it is now generally accepted that these intermediate species are Fe 3ϩ -peroxide species (16 -24). For the Desulfoarculus baarsii enzyme, the reaction mechanism was proposed to involve formation of two intermediates (Scheme 1) (16,19). The first one, a proposed Fe 3ϩ -peroxo species, results from the bimolecular reaction of O 2 . with Fe 2ϩ .
It then undergoes a diffusion limited protonation process to form a second reaction intermediate, a Fe 3ϩ -hydroperoxo species (19). A second protonation process at the level of the Fe 3ϩhydroperoxo species would allow the release of the reaction product H 2 O 2 . Although the donor of this second protonation event is not yet known, it was proposed to be associated with the presence of a water molecule at the SOR active site (25). Finally, the resulting Fe 3ϩ atom of the active site becomes hexacoordinated with a conserved glutamate residue, Glu 47 in D. baarsii (Scheme 1) (7,26). The strictly conserved cysteine axial ligand in the SOR active site has been hypothesized for long to be an essential feature in the reactivity of the enzyme with superoxide (5,8). It was also proposed that, similarly to that reported for cytochrome P450 enzymes, rubredoxins, or ferredoxins (27)(28)(29), H-bonds on the cysteine ligand could finely tune the strength of the Fe™S bond and then, in the case of SOR, might play an important role in the reactivity of SOR with superoxide (30). Two peptide NH groups from the Leu/Ile and His of the well conserved tetrapeptide Cys 116 -Asn 117 -Leu/Ile 118 -His 119 (numbered according to the D. baarsii SOR sequence) are within H-bonding distance of the cysteine ligand sulfur. In the case of the SOR from D. baarsii, for which a crystal structure was determined at 1.15-Å resolution (14), the Ile 118 and His 119 peptide N atoms are at 3.41 and 3.26 Å, respectively, from the cysteine sulfur ligand, suggesting relatively weak H-bonds from these two potential donors (Fig. 1). However, up to now, no experimental data have been reported to investigate the functional role of this cysteine residue. Sitedirected mutagenesis at this position (Cys 3 Ala) in the case of Archaeoglobus fulgidus SOR enzyme was reported (11) to result in an apoprotein, and thus not suitable for any study of the reaction with superoxide.
In the case of the D. baarsii enzyme, redox-induced FTIR difference spectra (exploiting the Fe 2ϩ 3 Fe 3ϩ redox change at the SOR active site) revealed that a protonated carboxylate group, proposed to be the side chain of Glu 114 , undergoes conformational modifications upon oxidation of the active-site iron (26). This highlights some marked interaction between this carboxylic group and the iron active site. Glu 114 is located near the Cys 116 side chain, buried inside the protein in a non-polar environment, compatible with a protonated carboxylate side chain. Its two carboxylic oxygens face directly the sulfur and iron atoms, with distances of about 5.6 -6.5 Å (Fig. 1). Such an orientation suggests some dipolar interactions, which could explain the FTIR difference signals. These interactions could also contribute to the modulation of the strength of the S™Fe bond. Glu 114 is conserved among the two-iron type of SORs, as well as the one-iron SOR of Treponema pallidum. For the T. pallidum enzyme, a similar interaction with the iron active site and the lateral chain of a protonated carboxylate was also observed by FTIR spectroscopy (26). On the other hand, most of the one-iron SORs such as the neelaredoxins and as represented by, for example, Pyrococcus furiosus, do not have this glutamate residue, where instead a serine residue is found (7). The side chain of this serine residue shows no probable H-bonds donated by the OH group, according to the crystal structure of the P. furiosus enzyme (7).
In this work, using pulse radiolysis, FTIR, and resonance Raman spectroscopies, we have investigated to what extent the presence of the Glu 114 carboxylic lateral chain could affect the strength of the S™Fe bond and the reaction of the iron active site with superoxide. We report a characterization of the E114A mutant in D. baarsii showing significantly modified pulse radiolysis kinetics for the protonation process of the first reaction intermediate. RR spectroscopy shows very local changes at the level of the Fe 3ϩ -S(Cys) bond, and demonstrates that the mutant strongly stabilized an iron peroxo species in its active site. This highlights the function of this carboxylic group on the fine tuning of the S™Fe strength and in the role of the cysteine ligand in the reactivity of the enzyme with superoxide.

EXPERIMENTAL PROCEDURES
Materials-For pulse radiolysis experiments, sodium formate and buffers were of the highest quality available (Prolabo Normatom or Merck Suprapure). Oxygen was from ALPHA GAZ. Its purity is higher than 99.99%. Water was purified using an Elga Maxima system (resistivity 18.2 megaohm). K 2 IrCl 6 was from Strem Chemical Inc.
Bacterial Strain and Plasmids-The complementation tests of the Escherichia coli QC2375 mutant strain were carried out as previously described (16). Plasmid pMJ25 is a pJF119EH derivative, in which the sor gene from D. baarsii is under the control of a tac isopropyl 1-thio-␤-D-galactopyranoside (IPTG)-inducible promotor (2). (ii) one oxygen atom of the Glu 114 side chain is located at 2.63 Å from the backbone CϭO of His 49 . This is consistent with a protonated Glu 114 carboxylate side chain forming a H-bond with the backbone CO of His 49 . In black are shown the distances between the two oxygen atoms of the carboxylic side chain of Glu 114 and the sulfur and iron atoms (5.6 -6.5 Å). SCHEME 1. Reaction mechanism of the SOR from D. baarsii with O 2 . (from Ref. 19).
Site-directed Mutagenesis and Protein Purification-Two primers were designed for PCR-based site-directed mutagenesis to create the D. baarsii SOR mutant E114A. Primer 1 (5Ј-CAAGGTCGTGGCCCGCGCATACTGCAACATCCACG-3Ј) and primer 2 (5Ј-CGTGGATGTTGCAGTATGCGCGGG-CCACGACCTTG-3Ј) contained the mutation of interest (underlined). Mutagenesis was carried out on plasmid pMJ25 with the QuikChange site-directed mutagenesis kit from Stratagene. The mutation was verified by DNA sequencing. The resulting plasmid, pCWE114A, was transformed in E. coli DH5␣. Overexpression and purification of the E114A mutant protein was carried out as reported for the wild-type protein (2). The purified mutant protein was found to be as stable as the wild-type protein. The UV-visible absorption spectrum of the purified protein exhibited an A 280 nm /A 503 nm ratio of 4.8 -4.9 and appeared to be homogeneous, as seen by SDS-PAGE analysis (data not shown). Protein concentration was determined using the Bio-Rad protein assay reagent. Full metallation of the two mononuclear iron sites was verified by atomic absorption spectroscopy, with a value of 2.0 iron atoms per polypeptide chain (data not shown). The protein was isolated with an oxidized Center I (⑀ 503 nm ϭ 4,400 M Ϫ1 cm Ϫ1 ) and a fully reduced Center II.
Pulse Radiolysis-Pulse radiolysis measurements were performed as described elsewhere (16,19). Briefly, free radicals were generated by irradiation of O 2 -saturated aqueous protein solutions (100 M), in 10 mM of the different buffers used, 0.1 M sodium formate with 200-ns pulses of 4 MeV electrons at the linear accelerator at the Curie Institute, Orsay, France. Superoxide anion, O 2 . , was generated during the scavenging by formate of the radiolytically produced hydroxyl radical, HO ⅐ , as previously described (16). Doses of about 5 gray per pulse resulted in about 3 M O 2 . . Reactions were followed spectrophotometrically, between 450 and 750 nm, at 20°C in a 2-cm path length cuvette. Kinetic traces were analyzed using a Levenberg-Marquardt algorithm from the Kaleidagraph software package (Synergy Software). ␥-Ray Irradiations-␥-Ray irradiations were carried out at 20°C using a cobalt-60 source, at a dose rate of 5.9 gray min Ϫ1 . The SOR protein to be irradiated, 22 M, was in 10 mM Tris-HCl buffer, pH 7.6, 100 mM sodium formate, 500 units/ml of catalase from Aspergillus niger and saturated with pure O 2 . The duration of irradiation was 6 min (35 gray) to obtain a stoichiometric amount of O 2 . (radiolytic yield 0.62 mol J Ϫ1 ) relative to SOR. Resonance Raman-Resonance Raman spectra were recorded using a modified single-stage spectrometer (Jobin-Yvon) equipped with a liquid nitrogen cooled back-thinned CCD detector as described elsewhere (22,23). Briefly, excitation at 647.1 nm (30 milliwatts) was provided by a Spectra Physics Series 2000 Kr ϩ laser. Protein samples were 3-5 mM and in 100 -200 mM buffer. A holographic notch filter (Kaiser Optical) was used to reject stray light. Spectra were calibrated using the exciting laser line, along with the SO 4 2Ϫ (983 cm Ϫ1 ) and ice (230 cm Ϫ1 ) Raman bands from a frozen aqueous sodium sulfate solution. Spectral resolution was Ͻ3 cm Ϫ1 with entrance slits at 100 m and frequency accuracy was Ϯ1 cm Ϫ1 . Reported spectra were the result of the averaging of 40 single spectra recorded with 30 s of accumulation time. Baseline corrections were performed using GRAMS 32 (Galactic Industries). In all reported spectra, the contributions from ice have been subtracted using the GRAMS 32 software by cancellation of the 230 cm Ϫ1 band. In general for the resonance Raman samples, 1 l of concentrated protein (1-5 mM) was deposited on a glass slide sample holder and then transferred into a cold helium gas circulating optical cryostat (STVP-100, Janis Research) held at 15 K. For the rapidly frozen H 2 O 2 -treated Raman samples, 5 l of concentrated SOR protein (3-6 mM) on ice was manually mixed rapidly with an equal volume of H 2 O 2 solution whose concentration was adjusted to obtain a final SOR protein: H 2 O 2 ratio of about 1:6 eq. 3-5 l of the resulting mixture were promptly transferred to the glass slide sample holder and immediately immersed in liquid nitrogen before transferring to the optical cryostat for Raman measurement. The entire mixing/freezing operation took no more than about 5 s.
Optical absorbance measurements were made using a Varian Cary spectrophotometer, in 1-cm path length cuvettes. Low temperature (4.2 K) X-band EPR spectra were recorded on a Bruker EMX 081 spectrometer equipped with an Oxford Instrument continuous flow cold helium gas cryostat.

Spectroscopic Characterization of the SOR E114A Mutant-
The UV-visible absorption spectrum of the as isolated E114A mutant exhibited the characteristic absorption bands at 370 and 503 nm, arising from the ferric iron of Center I (2); these values are similar to those observed for the wild-type protein (2). When treated with a slight excess of K 2 IrCl 6 , the spectrum of the mutant exhibited an increase in absorbance in the 500 -700 nm range, reflecting oxidation of the Center II iron atom (19). The well defined absorption band corresponding to the oxidized Center II band can be more easily seen when the intense contributions of Center I are subtracted (Fig. 2). The maximum of this absorption band strongly depends on the pH, as was determined for the wild-type SOR protein (19). At pH 7.6, in the E114A mutant protein, the band corresponding to the oxidized Center II is centered at 630 nm, with a molar extinction coefficient value of 1,660 mM Ϫ1 cm Ϫ1 . Upon increasing the pH from 5 to 10.2, the Center II absorption band shifts from 632 to 547 nm (Fig. 2), with an apparent pK a Function of the Cysteine Axial Ligand in SOR Activity value of 8.8 Ϯ 0.1 (Fig. 2, inset). This pK a value of the so-called "alkaline transition" is similar to that found for the wild-type (pK a ϭ 9.0 Ϯ 0.1) (19) and corresponds to the displacement of the glutamate 47 ligand by a hydroxide ion at basic pH values on the sixth coordination position of the active-site ferric iron (25). However, in the case of the E114A mutant, the absorption maxima of the Fe 3ϩ -Glu (632 nm) and the Fe 3ϩ -OH (547 nm) forms are blue shifted from 12 to 13 nm compared with that reported for the wild-type protein (644 and 560 nm, respectively) (19). This might reflect some electrostatic changes at or near Center II in the mutant protein.
The 4 K EPR spectra of as isolated E114A mutant displayed resonances at g ϭ 7.7, 5.7, 4.1, and 1.8, similar to those obtained for the wild-type SOR (2) (data not shown). These resonances are typical of those of a Fe 3ϩ ion in a distorted tetrahedral FeS 4 center and originated from Center I (2). The 4 K EPR spectrum of the E114A mutant oxidized with a slight excess of K 2 IrCl 6 exhibited an additional signal at g ϭ 4.3 (Fig. 3), identical to that observed for wild-type SOR (2). This g ϭ 4.3 signal was attributed to the oxidized Center II, with a rhombic (E/D ϭ 0.33) high-spin (S ϭ 5/2) ferric ion (2,30).
For the wild-type protein, FTIR difference spectra for reduced and oxidized iron Center II showed bands at 1770 and 1760 cm Ϫ1 , respectively (Fig. 4, thin line, and Ref. 26), assigned to the carbonyl stretching mode, (CϭO), of a protonated carboxylate group of an Asp or Glu side chain located in a nonpolar environment (26). The comparison of reduced minus oxidized FTIR difference spectra of the wild-type SOR (Fig. 4, thin line) with that of the E114A mutant (thick line) clearly indicates that the 1770/1760 cm Ϫ1 differential signal is not observed when Glu 114 is replaced with alanine, a residue that does not possess a carboxylic group. Thus, Glu 114 , located in a hydrophobic protein environment at Ϸ6 Å of the iron Center II accounts for the 1770/1760 cm Ϫ1 differential signal (26). Bands at 1200 (oxidized) and 1188 (reduced) cm Ϫ1 (Fig. 4, thin line) can be also assigned to the Glu 114 side chain (C™O) stretching mode.
In addition, Fig. 4 further shows that only modest global protein structural changes are induced by the E114A mutation. Slight spectral changes are observed in the region where pep-   tide (CϭO) and (CN) ϩ (NH) modes contribute, at 1700-1610 and 1580-1520 cm Ϫ1 , respectively. The infrared bands at 1109 (reduced) and 1098 (oxidized) cm Ϫ1 , assigned to the ring mode of histidine ligands of the iron, remain unchanged for the mutant.
Reaction This value is almost identical to that reported for the wild-type protein (20). Furthermore, k 1 was found to be pH independent between 5 and 9.5 (Fig. 5A), as it was reported for the wild-type protein (19).
This first intermediate decayed monoexponentially on the millisecond time scale (0.4 -10 ms) with a k 2 value independent of the protein concentration (data not shown). As reported for the wild-type protein, in the E114A mutant k 2 is directly proportional to H ϩ concentration between pH 5 and 8 (Fig. 5B), with logk 2 ϭ logk 0 Ϫ (0.80 Ϯ 0.1) ϫ pH. A slope close to unity (i.e. 0.8 Ϯ 0.1) indicates that, like the wild-type protein (19), the decay of the first intermediate involves a single protonation process as the rate-limiting step, the proton being directly provided by the bulk aqueous solvent. However, between pH 5 and 8, k 2 values determined for the E114A mutant are 10 times smaller than that of the wild-type protein (Fig. 5B). This is reflected by the k 0 value, the limit of the protonation rate constant when pH values approach 0, calculated to be (1.7 Ϯ 0.6) ϫ 10 8 M Ϫ1 s Ϫ1 for the E114A mutant and (1.4 Ϯ 0.7) ϫ 10 9 M Ϫ1 s Ϫ1 for the wild-type protein (Fig. 5B and Ref. 19). These data show that the E114A mutation results in a decrease of the rate of protonation of the first reaction intermediate.
Above pH 8.4 and up to pH 9.5, in contrast with the wild-type protein, k 2 values for the E114A mutant became pH independent (Fig. 5B). Similar behavior at basic pH values was also described for the wild-type SORs from Desulfovibrio vulgaris (18) and Archaeoglobus fulgidus (20,21) and was proposed to be associated with a protonation event arising from a water molecule instead of H 3 O ϩ at more acidic pH values (20,21). This behavior could be also associated with the presence of a protonated base close to the active site. In the wild-type SOR from D. baarsii, where such a pH-dependent process was observed up to pH 10, this protonated base might be not be effective enough to out-compete the protonation process from the bulk solvent. For the E114A mutant, at basic pH values, the situation could be the opposite, due to the slower protonation process arising from the bulk.
The UV-visible absorption spectrum of the first reaction intermediate for the E114A mutant was reconstructed from the absorbance at different wavelengths obtained after 50 s of the reaction at pH 7.6 (data not shown). This spectrum exhibits a maximum at 610 nm and is identical to the spectrum of the first reaction intermediate reported for the wild-type protein, which was associated with a Fe 3ϩ -peroxo species (19). Such a species results from the fast binding, at the sixth coordination position, of the superoxide anion and its subsequent reduction by the ferrous iron.
The spectrum of the second intermediate of the E114A mutant was reconstructed at pH 7.6, where k 2 exhibits a direct dependence with H ϩ concentration and at pH 9.1, where k 2 was found to be pH independent (Fig. 6). Both reconstructed spectra are similar and comparable with that determined for the second reaction intermediate in the wild-type protein (19), with a relatively narrow absorption band centered at 625 nm. This second intermediate, resulting from the protonation of the Fe 3ϩ -peroxo species was proposed to be a Fe 3ϩ -hydroperoxo species (19).
As previously described (19), with our instrumental device, we are not able to follow the reaction kinetics above 30 ms, a time range in which the last step of the reaction is expected to occur (Scheme 1). This last step would correspond to the dissociation of the second reaction intermediate, the release of H 2 O 2 and formation of the Fe 3ϩ active site species. Such a final reaction step is supported by the appearance at the end of the reaction of a UV-visible absorption spectrum characteristic of the oxidized active site (Fig. 6, dotted line), which clearly differs from that of the second reaction intermediate. This final spectrum was obtained from ␥-ray irradiation experiments, carried out under the same experimental conditions as pulse radiolysis. This spectrum corresponds to a stoichiometric oxidation of the SOR Center II (⑀ 630 nm ϭ 1,660 mM Ϫ1 cm Ϫ1 , at pH 7.6) by O 2 . . Fig. 7 shows the 15 K RR spectra of the wild-type and E114A mutant oxidized with K 2 IrCl 6 , excited using a 647.1 nm laser, in resonance with the S 3 Fe 3ϩ charge transfer band of the active site (22,23). The pre-resonance contributions from Center I (the Fe 3ϩ rubredoxin-like center) have been subtracted, as previously described (23). The active site modes thus enhanced predominantly arise from the Fe 3ϩ -S(Cys) moiety. The 299/316/323 cm Ϫ1 cluster of bands in the D. baarsii wild-type spectrum (Fig. 7a) was attributed to primarily Fe 3ϩ -S(Cys) stretching mode and bending mode contributions (23). Because the 299 cm Ϫ1 band is the most intense, it is most likely predominantly Fe 3ϩ -S(Cys) stretching in character (23). The 742 cm Ϫ1 band was assigned to the S-C ␤ stretching mode of Cys 116 and the weakly enhanced Fe 3ϩ -N(His) stretching mode bands from the Fe 3ϩ -coordinating histidine residues are seen at 216, 234, and 238 cm Ϫ1 (Ref. 23 and Fig. 7). For the K 2 IrCl 6 -oxidized proteins (Fig. 7), mutation of the Glu 114 residue significantly perturbs the active site vibrational modes associated with the cysteine ligand, as compared with wild-type. The 299/316/323 cm Ϫ1 cluster of bands (wild-type) is shifted to 291/313/323 cm Ϫ1 (E114A) indicating a general significant downshift of the Fe™S(Cys) stretching and bending modes for the E114A mutant, corresponding to a significant weakening of the S™Fe 3ϩ bond for this mutant.

Resonance Raman Experiments: Stabilization of a Fe 3ϩ -peroxo Species on the Active Site-
The C ␤ -S stretching frequency is sizably increased from 742 (wild-type) to 755 (E114A) cm Ϫ1 (Fig. 7). These data indicate that the E114A mutation results in an increase in electron density in the S-C ␤ bond probably from a slight geometry change of the bond angle. The 357 cm Ϫ1 band (wild-type), also assigned to a cysteine deformation mode containing significant C-N character (23), is not sizably changed in the mutant E114A spectrum (i.e. 356 cm Ϫ1 ). However, Fig. 7 indicates some small but significant changes in some of the vibrational frequencies associated with the Fe-N modes of the iron-coordinating histidine residues. For example, the 216 cm Ϫ1 frequency in the wildtype spectrum shifts to 214 cm Ϫ1 . All the above observations clearly show that the E114A mutation results in localized structural changes primarily associated with Cys 116 , with a notable decrease of the strength of the S™Fe bond.
The treatment of SOR with a slight excess of H 2 O 2 followed by rapid freezing was previously reported to form a metastable Fe 3ϩ -peroxo species at the SOR active site, in particular when the oxidized active site sixth ligand was removed (i.e. the E47A mutation) (22,23). These studies allowed a detailed resonance Raman characterization of a particular high-spin Fe 3ϩ -peroxo species exhibiting (Fe 3ϩ -O 2 ) and (O™O) stretching frequencies of 433-438 cm Ϫ1 and 850 -852 cm Ϫ1 , respectively (22,23). When the E114A mutant was similarly treated with 6 eq of H 2 O 2 at pH 8.5 and rapidly frozen (less than 5 s), the RR spectrum of the E114A mutant exhibited new bands at 446 and 851 cm Ϫ1 (Fig. 8) that were not present in the spectrum of the K 2 IrCl 6 -oxidized protein (Fig. 7). These band frequencies are consistent with the (Fe™O 2 ) and (O™O) stretching modes, respectively, of a Fe 3ϩ -peroxo species, as previously verified for the wild-type and E47A mutant of the SOR from D. baarsii using 18 O labeling (22,23). The intensities of these two bands are similar to that of the 756 cm Ϫ1 band (C-S stretching mode), a marker for the active site in the Fe 3ϩ state, as reported for the E47A mutant (at 742 cm Ϫ1 ) (Fig. 8). For the wild-type protein, the two RR bands corresponding to the Fe 3ϩ -peroxo species are . generated by ␥-ray irradiation in the presence of a catalytic amount of catalase, or by reaction with a slight excess of K 2 IrCl 6 . FIGURE 7. Low temperature (15 K) resonance Raman spectra of the SOR active site wild-type and E114A mutant from D. baarsii oxidized with 3 eq of K 2 IrCl 6 excited at 647.1 nm. a, E114A mutant; b, wild-type. SOR concentration was 3 mM in 100 mM Tris-HCl buffer, pH 8.5; laser power was 30 milliwatts at the sample; spectral resolution was Ͻ3 cm Ϫ1 . Contribution of the Fe 3ϩ rubredoxin-like center (Center I) and ice were subtracted from each spectrum (23).
very weak compared with the 742 cm Ϫ1 band ( Fig. 8 and Ref. 23), indicating that this species is trapped to a lesser degree. 5 In contrast, for the E114A mutant, the intensities of the peroxo bands indicate that when treated with H 2 O 2 , this mutant trapped and stabilized by rapid freezing a significant population of the Fe 3ϩ -peroxo species at the active site, as was observed for the D. baarsii E47A mutant (22,23). However, in the case of the E114A mutant, the (Fe™O 2 ) stretching mode is observed at a significantly higher frequency compared with the E47A mutant (446 cm Ϫ1 compared with 438 cm Ϫ1 , respectively), whereas the (O™O) stretching mode remains comparable (851 cm Ϫ1 for the E114A and 850 cm Ϫ1 for the E47A). These observations indicate that the Fe™O bond is stronger for the Glu 114 mutant than it is for wild-type and E47A.
The low temperature EPR spectrum of a rapidly frozen solution of E114A mixed with H 2 O 2 in conditions comparable with those of the RR samples corresponding to Fig. 8, exhibits a feature at g ϭ 4.3, similar to that observed for the K 2 IrCl 6 -oxidized protein (Fig. 3). Except those for the iron corresponding to Center I, which remained unchanged, no other signals were observed. This indicates that the peroxo species formed in these conditions is associated with a high spin rhombic ferric ion.
Altogether these results indicate that, like the E47A mutant, the E114A mutation strongly stabilizes a high spin Fe 3ϩ -peroxo species in its active site. However, the Fe™O bond of the peroxo is significantly strengthened in the E114A mutant.
Effects  Fig. S1). An NADPH:superoxide oxidoreductase activity for the wild-type SOR of 9 M NADPH ox / min was determined. The E114A mutation does not modify the NADPH:superoxide oxidoreductase activity of the protein (supplemental data Fig. S1). As shown previously (31), under these conditions, the O 2 . flux is the rate-limiting process for NADPH consumption. Consequently this test does not provide additional information on the effects of the E114A mutation on the subsequent steps after O 2 . binding at the SOR active site. However, it demonstrates that turnovers still occur with this mutant. The ability of the SOR E114A mutant protein to complement E. coli superoxide dismutase deficiency was tested as described in Ref. 16. The E. coli sodA sodB recA mutant cannot grow in the presence of oxygen because of the combined lack of superoxide dismutase activity (sodA sodB) and the DNA strand break repair activity (recA) that results in lethal DNA oxidative damage. As shown in Table 1, at low IPTG concentration, the production of the E114A mutant SOR protein restored only 23% of the aerobic growth of sodA sodB recA mutant, whereas the wildtype SOR protein restored 42%. At higher IPTG concentration, 58% of anaerobically growing E. coli cells producing the E114A SOR were able to grow aerobically compared with 78% for those producing wild-type SOR. These data indicate that production of E114A SOR can compensate for superoxide dismutase deficiency in E. coli, but with significantly less efficiency than the wild-type SOR.

DISCUSSION
The presence of a cysteine axial ligand in the SOR iron active site has long been proposed to be an essential feature in the reactivity of the enzyme with superoxide (5,8). For heme-thiolate proteins like cytochrome P450 and for [Fe™S] cluster proteins (27)(28)(29), it has been proposed that H-bonds on the cysteine ligand would change the strength of the S™Fe bond and therefore modulate the chemistry of these proteins. A similar situation is expected for SOR where such changes in the S™Fe bond should play some role in modulating the reactivity of SOR and c, wild-type. SOR concentration was 3 mM in 100 mM Tris-HCl buffer, pH 8.5; laser power was 30 milliwatts at the sample; spectral resolution was Ͻ3 cm Ϫ1 . Contribution of the Fe 3ϩ rubredoxin-like center (Center I) and the ice were subtracted from each spectrum (23). The 230 cm Ϫ1 band is due to ice contributions.

TABLE 1 Aerobic survival of superoxide dismutase-deficient E. coli cells depending on their production of wild-type or E114A mutant of SOR
Anaerobic cultures of sodA sodB recA E. coli mutant QC 2375 transformed with pJF119EH, pMJ25, or pCWE114A were plated on LB medium plus 5 M or 1 mM IPTG, under anaerobic and aerobic conditions. Colonies were counted and after 24 h incubation at 37°C.

Plasmid
Aerobic survival a a Survival was calculated as the ratio of the number of colonies under aerobic conditions to those under anaerobic conditions. Values are the means of five experiments. 100% corresponds to 1.8 -2.3 ϫ 10 8 colonies. b ND, not determined. with superoxide. Because it is trans to the superoxide binding site, the cysteine ligand should play a critical role in donating electron density to the iron metal center so as to promote Fe 2ϩ 3 superoxide electron transfer, or H 2 O 2 dissociation from highspin Fe 3ϩ , or both. In particular for SOR, in the well conserved tetrapeptide Cys 116 -Asn 117 -Leu/Ile 118 -His 119 , two peptide NH groups from the Leu/Ile 118 and His 119 are within H-bonding distance of the Cys 116 ligand sulfur (Fig. 1) (14). Furthermore, for the D. baarsii enzyme, as previously suggested (26) and confirmed here by FTIR measurements, the carboxylic group of the Glu 114 side chain, which directly faces the S™Fe bond at about 5-6 Å from both atoms, is in interaction with the iron atom in the active site (Fig. 1). Glu 114 is well conserved in the two-iron SORs, whereas it is replaced by a Ser residue in the one-iron type of SORs, such as the P. furiosus enzyme (7).
In this work, we have mutated the Glu 114 residue of the SOR from D. baarsii, into alanine. The data reveal that the presence of the Glu 114 carboxylic group allows a fine tuning of the S™Fe bond strength, which appears to be a determining factor in the rate of formation and release of the reaction product, H 2 O 2 .
The pulse radiolysis experiments presented here showed that the E114A mutation does not change the overall reaction mechanism of SOR with superoxide (Scheme 1) and still reacts very rapidly with superoxide. Although with our pulse radiolysis apparatus we were not able to kinetically characterize the final step of the catalytic cycle (formation of H 2 O 2 and the resting ferric active site), this final step does exist for the E114A mutant (Fig. 6), as it does for wild-type and other mutants we have studied (16,19).
However, we have observed that the mutation specifically decreases by a factor of 10 the rate constant for the protonation process of the Fe 3ϩ -peroxo intermediate species, to form the Fe 3ϩ -hydroperoxo intermediate species. An interpretation for this very specific effect is not obvious at first glance. In fact, because the Glu 114 is located on the opposite side to the intermediate peroxo adducts (Fig. 1) and that, as seen here by FTIR spectroscopy, the E114A mutation does not induce marked global structural changes in the protein, the decrease of the protonation rate of the peroxo species is not expected to result from structural constraints that could appear in the active site. Also, the mutation does not apparently modify the proton source, because as in the wild-type protein, the proton is still provided directly by the bulk solvent (Fig. 5). Consequently, this slower protonation process could be associated with some changes in the intrinsic properties of the Fe 3ϩ -peroxo intermediate species. Interestingly, the E114A mutation removes the dipolar interaction with Cys 116 (Fig. 4) as well as potentially removing an H-bond with the backbone CϭO group of His 49 whose side chain is ligated to the iron atom (Fig. 1). All these factors could affect the local electrostatics around the Cys-Fe moiety as well as slightly altering its bonding geometry, distance, and therefore the bond strength. Several arguments reported here are in agreement with the above proposals.
First, we previously reported that the wild-type protein exhibits an alkaline transition of the S 3 Fe 3ϩ charge transfer UV-visible absorption (644 to 560 nm, ⌬ ϭ 84 nm, pK a ϭ 9.0), which has been attributed to the displacement of the Glu 47 ligand by OH Ϫ at basic pH values (8,24). This results in an increase in the electron density at the Fe 3ϩ site (25). The E114A mutant exhibits a similar alkaline transition with similar pK a and magnitude values (pK a ϭ 8.8 and ⌬ ϭ 85 nm). However, the mutation induces spectral blue-shifts of 12-13 nm for both the Fe 3ϩ -Glu 47 and Fe 3ϩ -OH ligated states (632 to 547 nm, respectively). The significantly different S 3 Fe 3ϩ charge transfer absorption maxima of the E114A indicates localized changes at the active site that affect electron density at the high spin Fe 3ϩ atom (8).
Second, the RR spectra show that the E114A mutation has very specific effects on the cysteine ligand (Fig. 7). When compared with the wild-type and the E47A proteins, the E114A mutation induces a significant weakening of the S™Fe 3ϩ bond (299/316/323 to 291/313/311 cm Ϫ1 ) and a concomitant strengthening of the C-S bond (743 to 756 cm Ϫ1 ) of the cysteine ligand. These localized structural changes are compatible with the observed blue shift (12-13 nm) of the S 3 Fe 3ϩ charge transfer band of the oxidized iron site in the E114A mutant.
From these data, one can conclude that the carboxylic side chain of Glu 114 , which faces directly the S™Fe bond at about 5-6 Å distances of both atoms, allows a fine tuning of the strength of the S™Fe bond. In particular it induces a strengthening of the S™Fe bond.
Such an effect should have consequences on the bonding of the Fe 3ϩ -peroxo intermediates that form during the reaction with superoxide. Up to now, the reaction of SOR with H 2 O 2 is the only way that has been reported to allow a detailed characterization by resonance Raman of a Fe 3ϩ -peroxo species in the active site of SOR (23,24). Although it is not known yet if such a peroxo species is identical to the ones formed during the reaction of the protein with superoxide, the characterization of a Fe 3ϩ -peroxo species here and elsewhere in SORs by resonance Raman are most likely informative with respect to the properties of the intermediates species formed in the reaction with superoxide. In support of that, it has been recently reported that the iron complex SOR mimic [Fe II (cyclam)-PrS] ϩ , when reacted with superoxide, and forms an high-spin iron peroxo intermediate with Fe™O and O™O resonance Raman bands (32) very similar to those observed for the SOR protein treated with H 2 O 2 (23,24).
When the E114A mutant is reacted with a slight excess of hydrogen peroxide, we showed that it stabilizes an iron peroxo species at its active site, which can be freeze-trapped and characterized by resonance Raman spectroscopy. The RR spectrum exhibits intense (Fe™O 2 ) and (O™O) mode bands, at 446 and 851 cm Ϫ1 , respectively, characteristic of the ferric iron peroxo species similarly trapped in other SORs and which has been assigned to a side-on high-spin Fe 3ϩ -peroxo species (22,23). EPR experiments presented here show that the peroxo species in the E114A mutant is also associated with a high-spin ferric iron site.
Such a high-spin iron-peroxo species was readily trapped and characterized in another D. baarsii SOR variant, the E47A mutant protein (22,23). In the wild-type protein, a similar peroxo species, with RR band frequencies identical to those of the E47A mutant, was also observed but was trapped to a much lesser degree. This is consistent with the fact that the wild-type SOR active site accommodates such species only transiently because the Glu 47 residue binds to the sixth coordination position of the Center II ferric site, facilitating H 2 O 2 release by a ligand exchange mechanism (23). Thus, for the E47A mutant, the stabilization of the peroxo species was associated with the presence of the alanine residue incapable of ligating to the ferric iron (23). Accordingly, in the E47A mutant, the Fe™O and O™O stretching modes exhibited identical vibrational frequencies as those of the wild-type protein (Fig. 8 and Refs. 22 and 23). Interestingly, the (Fe™O) band of the iron-peroxo species in the E114A mutant is shifted to higher frequency (446 compared with 438 cm Ϫ1 in the wild-type and E47A proteins), whereas the (O™O) bands exhibit comparable frequencies (850 -851 cm Ϫ1 ). This indicates that the Fe™O bond is stronger in the E114A mutant than in the E47A and wild-type proteins. A stronger Fe™O bond is in good agreement with stabilization of the peroxo species in the E114A mutant protein. Thus, the mechanism by which the E47A and E114A mutants induce a stabilization of the peroxo species appears to be fundamentally different.
For the E114A mutant, strengthening of the Fe™O bond correlates well with the observed weakening of the S™Fe bond that indicates a decreased electron donation from the cysteine ligand to the iron. For a high-spin Fe 3ϩ -peroxo end-on or side-on bonding situation, the in-plane * orbital of the peroxo ligand overlaps with the Fe 3ϩ d xz orbital (Refs. 23 and references therein). Thus any electron density donation to the filled * orbital, from the iron center and originating from the cysteine ligand in the trans position, is expected to weaken the Fe™O 2 bond. The weaker S™Fe 3ϩ bond in the E114A mutant (compared with wild-type) indicates decreased cysteinate electron donating power to the iron and the peroxo * orbital, and therefore a stronger Fe™O 2 bond (compared with wild-type) is expected, and indeed observed. Such trans ligand donation effects have been observed for Fe 3ϩ -alkylperoxo species in SOR model complexes for which decreased Lewis basicity (i.e. decreased electron density donation from the trans ligand) was correlated with increased Fe™O bond strength and stability (33).
The strengthening of the Fe™O bond of the Fe 3ϩ -peroxo species observed in the E114A mutant, with no modification of the strength of the O™O bond is expected to change the acidity of the peroxo species, most likely increasing it. Interestingly, the Brönsted catalysis equation predicts that among a series of bases, the logarithm of the rate constant of protonation of a species is directly proportional to its pK a value. This appears to be in very good agreement with the decrease of the rate of the protonation of the Fe 3ϩ -peroxo species intermediate observed in the reaction of the E114A mutant with superoxide (Fig. 5, decreases of k 2 value, Scheme 1). Thus, the properties of the peroxo species characterized by RR from the reaction of SOR with H 2 O 2 account well for the specific effect of the E114A mutation on the protonation rate of the first reaction intermediate during the reaction with superoxide. Thus, we propose that the decrease of this protonation rate can be directly associated with a decrease of the pK a value of the peroxo intermediate, which as discussed above, originates from the strengthening of the Fe™O (peroxo) bond in that mutant.
Using the Fpr reductase from E. coli as a source of electrons for SOR, we showed that the E114A mutation did not inhibit turnovers with O 2 . . However, in cellular conditions, the complementation of an E. coli sod mutant strain with the D. baarsii sor gene was significantly impaired by the E114A mutation. Altogether with our previous conclusions that the E114A mutant stabilizes a peroxide intermediate in its active site, these data suggest that an efficient release of the peroxide intermediate to form H 2 O 2 is important for the SOR detoxification activity.
In conclusion, these data highlight a functional role for the carboxylic side chain of Glu 114 in the SOR from D. baarsii. This carboxylic group, which is in interaction with the S™Fe bond, directly contributes to a fine tuning of the S™Fe strength. The data obtained with the E114A mutant allow to gauge the importance of the strength of the S™Fe bond in the reaction with superoxide, which appear to directly affect both the strength of the Fe™O bond and the pK a values of the iron-peroxo intermediate that is formed during catalysis. Both factors are essential in the formation of the reaction product H 2 O 2 , which in SOR arises from protonations of the peroxo intermediate and cleavage of the Fe™O bond. One should note that a strengthening of the Fe™O bond in the E114A mutant should also markedly affect the last step of the reaction cycle with superoxide. This final step involves a second protonation process at the level of the Fe 3ϩ -hydroperoxide intermediate together with the cleavage of the Fe™O bond to form and release H 2 O 2 (Scheme 1). Further works are needed to specify this point.
P450 enzymes like SOR form Fe 3ϩ peroxide intermediate species with a trans cysteinate ligand during their catalytic cycles. However, at the difference of the P450 enzymes that favor the heterolytic cleavage of the O™O bond of the peroxide intermediate to form the high valent iron oxo species (34), our previous RR experiments demonstrated that the SOR active site can accommodate unusual Fe 3ϩ -peroxo species with particularly weak Fe™O bond and strong O™O bond, which, in addition to the spin state of the iron (33), clearly favors the Fe™O bond cleavage to form its reaction product H 2 O 2 . Our present work further shows that the strength of the S™Fe bond, which is finely tuned by the cysteine environment, significantly contributes to this unusual weak Fe™O bond in SOR.
Very recently, the stabilization of the peroxide intermediate in the E114A mutant has allowed the determination of a high resolution structure of three slightly different conformations of high-spin iron peroxide intermediates in the SOR active site with end-on coordination (35). It should be noted that previously, based on the (Fe™O)-(O™O) frequency comparisons with high-spin peroxo-iron complexes and D 2 O exchange experiments, a side-on peroxo conformation was proposed for SOR (22,23). It is possible that a side-on species could correspond to an early deprotonated peroxo state rapidly trapped in solution by flash-freezing at 77 K in Refs. 22 and 23 and in this work, but which was not trapped in the crystal under different experimental conditions (35). Further studies are needed to specify this point, but we stress that the arguments we make in this paper apply for both an end-on or side-on Fe 3ϩ -peroxo configurations.
These crystallographic studies highlight a specific protonation mechanism of the Fe™OOH on the proximal oxygen to release H 2 O 2 (35). This provides additional important data that account for the differences in the evolution of the Fe 3ϩ -peroxide species between SOR and P450 enzymes.