Detoxification of superoxide without production of H₂O₂: Antioxidant activity of superoxide reductase complexed with ferrocyanide

Abstract

The superoxide radical O₂^·̅ is a toxic by-product of oxygen metabolism. Two O₂^·̅ detoxifying enzymes have been described so far, superoxide dismutase and superoxide reductase (SOR), both forming H₂O₂ as a reaction product. Recently, the SOR active site, a ferrous iron in a [Fe²⁺ (N-His)₄ (S-Cys)] pentacoordination, was shown to have the ability to form a complex with the organometallic compound ferrocyanide. Here, we have investigated in detail the reactivity of the SOR–ferrocyanide complex with O₂^·̅ by pulse and γ-ray radiolysis, infrared, and UV-visible spectroscopies. The complex reacts very efficiently with O₂^·̅. However, the presence of the ferrocyanide adduct markedly modifies the reaction mechanism of SOR, with the formation of transient intermediates different from those observed for SOR alone. A one-electron redox chemistry appears to be carried out by the ferrocyanide moiety of the complex, whereas the SOR iron site remains in the reduced state. Surprisingly, the toxic H₂O₂ species is no longer the reaction product. Accordingly, in vivoexperiments showed that formation of the SOR–ferrocyanide complex increased the antioxidant capabilities of SOR expressed in an Escherichia coli sodA sodB recA mutant strain. Altogether, these data describe an unprecedented O₂^·̅ detoxification activity, catalyzed by the SOR–ferrocyanide complex, which does not conduct to the production of the toxic H₂O₂ species.

The superoxide radical O₂^·̅, the one-electron reduction product of molecular oxygen, is a toxic by-product of oxygen metabolism (1, 2). Fortunately, cells possess O₂^·̅ detoxifying enzymes, which are essential to their survival in the presence of oxygen. Two different types of O₂^·̅ detoxifying enzymes have been described so far. The first one is the well known metalloenzyme superoxide dismutase (SOD), present in almost all aerobic cells (1, 2). It catalyzes the dismutation of O₂^·̅ into H₂O₂ and O₂:

The second system, only found in prokaryotic cells, is the recently characterized nonheme iron superoxide reductase (SOR), which catalyzes reduction of O₂^·̅ into H₂O₂ (3–6):

Although both systems produce H₂O₂, a toxic substance precursor of HO˙ radicals that needs to be eliminated by catalase or peroxidase enzymes, the elimination of O₂^·̅ by SOD or SOR appears to be an essential cellular process to allow organisms to survive in the presence of O₂ (1, 2, 6).

The SOR active site consists of a mononuclear ferrous ion in an unusual [Fe²⁺ (N-His)₄ (S-Cys)] square pyramidal pentacoordination (7, 8). The free, solvent-exposed, sixth coordination position is the site of O₂^·̅ reduction (5, 6). In the case of the SOR from Desulfoarculus baarsii, the reaction with O₂^·̅ proceeds through two reaction intermediates (9, 10). The first one, presumably a Fe³⁺-peroxo species, is formed by the almost diffusion-limited binding of O₂^·̅ to the ferrous active site (k = 1 × 10⁹ M⁻¹·s⁻¹). This intermediate undergoes two sequential protonation processes, first yielding a second intermediate, possibly a Fe³⁺-hydroperoxo species and then the final reaction products, H₂O₂ and the ferric active site (10).

Recently, spectroscopic (11, 12) and crystallographic studies (8) have shown that the SOR active site binds ferrocyanide [also referred as hexacyanoferrate (II) or K₄Fe(CN)₆] at its sixth coordination position through a cyano bridge between the iron and the ferrocyanide molecule (Fig. 1). The complex was observed for both the reduced and oxidized forms of the SOR iron active site (8). The almost perfect steric and electrostatic complementarities of ferrocyanide with the active site suggested that it could act as a potent inhibitor of SOR activity by preventing O₂^·̅ access to the iron.

Fig. 1.

Crystal structure of the active site of the SOR from D. baarsii in complex with ferrocyanide (8). The coordination of Fe(CN)₆ to the SOR iron site occurs through a cyano bridge. The cyanide moieties are stabilized by hydrogen bonds (red dotted line) and van der Walls interactions (thin black dotted lines).

This SOR–ferrocyanide adduct is a previously undescribed complex between an organometallic compound and the iron site of a metalloenzyme. Here, we have studied in detail the reaction mechanism of this complex with O₂^·̅. We found that the complex reacts efficiently with O₂^·̅, in a different way than SOR alone, and eliminates O₂^·̅ without production of toxic H₂O₂. Accordingly, formation of the SOR–ferrocyanide complex in vivo specifically increases the protective effect of SOR expression in Escherichia coli.

Results

Reaction Mechanism of the SOR–Ferrocyanide Complex Studied by Pulse Radiolysis.

The reaction of D. baarsii SOR with O₂^·̅ in the presence of potassium ferrocyanide was investigated by pulse radiolysis as described in refs. 9 and 10. O₂^·̅ was generated in <1 μs by scavenging radiolytically generated HO˙ free radicals by 100 mM formate in O₂-saturated solution at pH 7.6. Aliquots of concentrated ferrocyanide were added to the solution just before the irradiation. The SOR protein was present in large excess with regard to O₂^·̅ ([SOR], 100 μM; [O₂^·̅], 3 μM], to provide pseudo first-order conditions. The kinetics of the reaction of SOR with O₂^·̅ were investigated every 5–10 nm between 395 and 660 nm, in the presence of 0, 0.25, 0.5, 1, 1.2, 1.5, 2, 5, 7, and 10 molar equivalents of ferrocyanide with respect to SOR.

Fig. 2A shows representative kinetic traces of the reaction in the 0.1-ms time scale, recorded at 580 nm in the presence of 0, 1, and 1.2–10 molar equivalent of ferrocyanide. In the absence of ferrocyanide, as described in refs. 9 and 10, absorbances reached a maximum ≈50 μs after the pulse to form the first reaction intermediate. At low ferrocyanide:SOR ratio, increasing the concentration of ferrocyanide slowed down the reaction and induced a decrease of the maximal absorbance variation (Fig. 2A). No additional change occurred above a ferrocyanide:SOR ratio of 1.2, up to 10 molar equivalents of ferrocyanide, at all of the wavelengths investigated. For a given ferrocyanide concentration, all of the kinetic traces fit a single exponential time-dependent equation, with identical rate constants k_1app. At ferrocyanide:SOR ratio <1.2 (Fig. 2B), k_1app decreased linearly with increasing ferrocyanide concentration. Above a slight molar excess of ferrocyanide, the reaction became ferrocyanide-independent, with a k_1app value of (1.70 ± 0.08) × 10⁴ s⁻¹, a value ≈6 times lower than in the absence of ferrocyanide (Fig. 2B). In the presence of 5 molar equivalents of ferrocyanide with respect to SOR, the k_1app value was found to be proportional to the SOR concentration (Fig. 2B Inset). A K_d value of 0.80 ± 0.07 μM was determined for the binding of ferrocyanide to SOR (Fig. 6, which is published as supporting information on the PNAS web site). These data indicate that as soon as enough ferrocyanide is present in solution to form a stoechiometric complex with SOR (ferrocyanide:SOR ratio ≥1.2), a bimolecular reaction between the SOR–ferrocyanide complex and O₂^·̅ occurs, with a rate constant k₁ value of (1.8 ± 0.1) × 10⁸ M⁻¹s⁻¹.

Fig. 2.

(A) First 300 μs of the reaction of SOR (100 μM) with O₂^·̅ (3 μM), generated by pulse radiolysis, followed at 580 nm, in the presence of 0, 1, and 1.2–10 molar equivalents of ferrocyanide, with respect to SOR, at pH 7.6. The lines were calculated for best fit to an exponential model. (B) Dependence of the observed rate constants k_1app on ferrocyanide concentration. (B Inset) Shown is the dependence of k_1app versus SOR concentration in the presence of 5 molar equivalents of ferrocyanide with respect to SOR.

In the absence of ferrocyanide, as described in refs. 9 and 10, a second reaction intermediate was formed ≈10 ms after the pulse. Fig. 3 shows representative traces at 580 nm, in the presence of 0, 1, and 1.2–10 molar equivalents of ferrocyanide, in the millisecond timescale. Between 0 and 1.2 molar equivalents of ferrocyanide, increasing the concentration of ferrocyanide-induced variation of the amplitude of the traces at the different wavelengths investigated (Fig. 3 and data not shown). No further change occurred above a ferrocyanide:SOR ratio of 1.2, up to 10 molar equivalents of ferrocyanide. For ferrocyanide:SOR ratios between 1.2 and 10, after the first rapid process described by k₁, all of the kinetic traces were single exponential processes, with identical rate constants k_2app = 445 ± 20 s⁻¹ (Fig. 3). In the presence of 5 equivalents of ferrocyanide with respect to SOR, this rate constant did not depend on the SOR concentration (Fig. 3 Inset). These data show that the free ferrocyanide molecules present in the solution do not react with the complex at this stage. At constant ionic strength, in the presence of 5 equivalents of ferrocyanide, k_2app value was found to vary linearly with Tris and formate concentrations (Fig. 7, which is published as supporting information on the PNAS web site), whereas k₁ value did not depend on these concentrations (data not shown). When Tris concentration was varied between 1 and 10 mM, k_2app increased by a factor of 1.6. More significantly, when formate concentration was varied between 10 and 100 mM, k_2app increased by a factor of 3.5 (Fig. 7). These data show that formate, and more moderately Tris, are involved in the reaction process described by k_2app.

Fig. 3.

First 16 ms of the reaction of SOR (100 μM) with O₂^·̅ (3 μM), generated by pulse radiolysis, followed at 580 nm, in the presence of 0, 1, and 1.2 to 10 molar equivalents of ferrocyanide with respect to SOR at pH 7.6. The lines were calculated for best fit to a biexponential model. (Inset) Shown is the dependence of k_2app versus SOR concentration in the presence of 5 molar equivalents of ferrocyanide with respect to SOR.

The data are consistent with the formation of two reaction intermediates. Their spectra were constructed at the delay time of their maximum formation. Fig. 4A shows the spectrum of the first intermediate recorded 300 μs after the pulse for a SOR:ferrocyanide ratio of 5. It differs from that recorded in the absence of ferrocyanide (10). Two peaks are observed at 420 and 580 nm. The 580-nm peak is 2.2 times lower than that of the first reaction intermediate observed without ferrocyanide at the same wavelength. The absorbance band centered at 420 nm is very similar to that of a ferricyanide [Fe³⁺(CN)₆] species.

Fig. 4.

Transient absorption spectra (2-cm path-length cuvette) formed upon reaction of SOR (100 μM) in the presence of 0 and 5 molar equivalents of ferrocyanide with O₂^·̅ (3 μM), generated by pulse radiolysis, at pH 7.6. (A) Reconstituted spectra of the first reaction intermediates. □, absence of ferrocyanide 50 μs after the pulse; ●, presence of 5 molar equivalents of ferrocyanide 300 μs after the pulse. (B) Reconstituted spectra of the second reaction intermediates. □, absence of ferrocyanide 10 ms after the pulse; ■, presence of 5 molar equivalents of ferrocyanide 10 ms after the pulse. Dotted line is the spectrum of a ferricyanide species at 3 μM.

The reconstituted spectrum of the second reaction intermediate at its maximum, 10 ms after the pulse (SOR:ferrocyanide ratio of 5, Fig. 4B), also strongly differs from the corresponding one, recorded without ferrocyanide (10). Actually, it is identical to that of a ferricyanide [Fe³⁺(CN)₆] species, with a single peak at 420 nm and at the same concentration as that of O₂^·̅ that reacted with the complex (3 μM) (Fig. 4B, dotted line).

Control experiments indicated that ferrocyanide alone (500 μM), in the absence of SOR, did not react with a 3 μM concentration of O₂^·̅ (data not shown). Also, ferrocyanide alone (6 μM) or in the presence of SOR (1 μM) did not catalyze dismutation of a 10 μM concentration of O₂^·̅ at pH 7.6 (data not shown).

Final Products of the Reaction.

Continuous γ-ray irradiation experiments, in which O₂^·̅ is at a steady state (13), are well adapted to analyze the final products of the reaction between the SOR–ferrocyanide complex and O₂^·̅ (10). Irradiation of aqueous solutions always leads to H₂O₂ production during the primary steps of water radiolysis (the so-called radiolytic H₂O₂, G ≈ 0.07 μmol·J⁻¹; ref. 13). Here, the dose was equal to 88 Gy, hence the radiolytic H₂O₂ production was equal to 6.1 μM. As illustrated in the Introduction from Eqs. 1 and 2, the measurement of the H₂O₂ production allows distinction between a reduction or a dismutation of O₂^·̅. In the presence of SOR alone, a 55 μM concentration of O₂^·̅ (yield ≈ 0.62 μmol·J⁻¹) was stoechiometrically reduced into H₂O₂ (Table 1) as reported in ref. 10. In the presence of increasing amounts of ferrocyanide, the H₂O₂ production decreased and, surprisingly, in the presence of more than one equivalent of ferrocyanide, only radiolytic H₂O₂ was detected. These data suggest that scavenging of O₂^·̅ by the SOR–ferrocyanide complex does not produce H₂O₂.

Table 1.

Products for the reaction of the SOR–ferrocyanide complex with a 55 μM concentration of O₂^·̅, generated by γ-ray radiolysis (88 Gy at 17.6 Gy/min) at pH 7.6

Ferrocyanide, molar equivalent	[H2O2]final,* μM	[H2O2]corr,† μM	[SOR] oxidized,‡ μM	[Ferricyanide],‡ μM
0	62 ± 4	56	57	0
0.5	44 ± 4	38	nd	nd
1	30 ± 1	24	50	14
2	6.1 ± 0.5	0	47	15
3	4.7 ± 1.2	0	44	17
5	7.2 ± 0.5	1	42	18
10	4.7 ± 2.0	0	nd	nd
SOR + 5 equivalents ferrocyanide + 50 μM H2O2	54 ± 1	48	nd	nd
No SOR, ferrocyanide 200 μM	34 ± 2	28	/	0

The SOR solution (100 μM) contained various molar equivalents of ferrocyanide compared with the SOR iron active site. Each value represents the mean of at least three independent experiments. nd, not determined; /, not relevant.

*Dosage of H₂O₂ with the leuco crystal violet method.

^†Corrected from the radiolytic yield for H₂O₂ = 6.1 μM.

^‡Determined from the spectra of Fig. 8 by using ε_{644 nm} = 1,900 M⁻¹·cm⁻¹ for the oxidized SOR active site and ε_420nm = 1,010 M⁻¹·cm⁻¹ for the ferricyanide.

To verify that the SOR–ferrocyanide solution does not lead to catalase activity, we incubated a solution containing 100 μM SOR and 5 molar equivalents of ferrocyanide with a 50 μM concentration of H₂O₂ for 15 min, room temperature at pH 7.6. No decrease in H₂O₂ concentration was observed. Also, a solution containing 100 μM SOR with 5 molar equivalents of ferrocyanide was allowed to react with 55 μM O₂^·̅ in the presence of 50 μM H₂O₂ (Table 1). At the end of the reaction, 48 μM H₂O₂ were detected in the reaction mixture (corrected from the radiolytic yield for H₂O₂). This result confirms that no catalase activity was present in the solution during the reaction with O₂^·̅. Inhibition of the HRP enzyme used to quantify H₂O₂ was excluded, because it still is possible to detect H₂O₂ added before the reaction or resulting from the radiolysis. Finally, ferrocyanide alone does not trap H₂O₂, because O₂^·̅ produced during the continuous γ-ray irradiation is quantitatively dismutated into H₂O₂ in the presence of a 200 μM concentration of ferrocyanide, and ferrocyanide was not oxidized (Table 1).

The final difference spectra of the SOR–ferrocyanide solutions after reaction with O₂^·̅ (reference is unreacted solutions) allow quantification of oxidized SOR and ferricyanide (Table 1; see Fig. 8, which is published as supporting information on the PNAS web site). Identical spectra were obtained in the presence of catalytic amounts of catalase (data not shown). In the absence of ferrocyanide, as reported in ref. 10, SOR is stoechiometrically oxidized by O₂^·̅ (Table 1). In the presence of increasing equivalents of ferrocyanide, the concentration of oxidized SOR is lowered, whereas that of ferricyanide increases, as determined from the apparition of its absorbance band at 420 nm (Fig. 8 and Table 1). The sum of the oxidized amounts of SOR plus ferricyanide remains constant and equal to the concentration of O₂^·̅ produced during the experiment (Table 1). Treating these irradiated solutions with iridium chloride or ammonium persulfate induces full oxidation of SOR and ferrocyanide, the concentrations of which are identical to those before irradiation (data not shown). These data indicate that neither the active site of SOR nor the Fe(CN)₆ species have been destroyed during the reaction.

FTIR Spectroscopy Analysis of the Reaction.

Fig. 5 shows FTIR difference spectra of the SOR–ferrocyanide solutions (with 2 and 5 ferrocyanide equivalents) after minus before the reaction with O₂^·̅. The experimental conditions were identical to those reported in Table 1. The presence of a negative band at 2,037 cm⁻¹ and a positive band at 2,116 cm⁻¹ indicates that during the reaction with O₂^·̅, there was a net consumption of ferrocyanide together with a net formation of ferricyanide, respectively. The frequency of the IR bands are typical for free species in solution, i.e., not complexed to SOR. Quantification of ferrocyanide and ferricyanide from the 2,037 and 2,116 cm⁻¹ bands indicates that ≈18 and 20 μM concentrations of free ferrocyanide have been consumed and 15–18 μM and 20–23 μM concentrations of free ferricyanide have been formed during the experiments containing 2 and 5 equivalents of ferrocyanide, respectively. The ferrocyanide consumption is nearly identical to that of the ferricyanide production, which fits well with the values reported in Table 1.

Fig. 5.

FTIR difference spectra calculated from the absorption spectra of SOR–ferrocyanide solutions recorded after minus before the reaction with 55 μM O₂^·̅, generated by γ-ray radiolysis. The solution contained a 100 μM concentration of SOR and various concentration of ferrocyanide at pH 7.6. In the upper spectrum, the sample contains 2 molar equivalents of ferrocyanide with respect to SOR. In the lower spectrum, the sample contains 5 molar equivalents of ferrocyanide with respect to SOR. The quantification of the ferrocyanide (2,037 cm⁻¹) and ferricyanide (2,115 cm⁻¹) bands were deduced from FTIR spectra recorded in the same conditions with solutions of known concentration for each species.

It has been reported that the SOR–ferrocyanide complex for the Treponema pallidum and Desulfovibrio vulgaris enzymes exhibits characteristic FTIR bands at 2,095, 2,047, and 2,024 cm⁻¹, assigned to the ν(CN) stretching mode of the bridging, equatorial, and axial cyanide groups of ferrocyanide, respectively (12). The FTIR spectrum of a concentrated preparation of a D. baarsii SOR–ferrocyanide complex shows almost identical features (Fig. 9, which is published as supporting information on the PNAS web site). The FTIR spectrum of the SOR–ferrocyanide complex after irradiation, in the same experimental conditions as those of Table 1 (followed by concentration), shows no significant modifications of the frequency and intensities of the bands associated with the ferrocyanide complexed with SOR (Fig. 9). These data show that the SOR–ferrocyanide complex has not been altered during the reaction with O₂^·̅ and that the ferrocyanide bound to SOR remains in the reduced state after the reaction with O₂^·̅.

Search for Fenton Reaction.

It is well known that H₂O₂ leads to HO˙ radicals in the presence of iron complexes (1, 2). If this very powerful oxidant would form during the reaction of the SOR–ferrocyanide complex with O₂^·̅, it would react instantaneously with the protein or the ferrocyanide present in solution. The electrospray mass spectrum of the irradiated SOR–ferrocyanide complex (5 equivalents of ferrocyanide, same experimental conditions as those of Table 1) shows that the mass of the SOR polypeptide is not modified, with only ions detected at 14,028 ± 2 Da (data not shown). This result demonstrates the absence of oxidative damage on the SOR polypeptide. Together with the UV-visible and FTIR data, it confirms that the SOR–ferrocyanide complex is not altered after the reaction with O₂^·̅.

In Vivo, the SOR–Ferrocyanide Complex Increases Protection Against Oxidative Stress.

The results show that the SOR–ferrocyanide complex scavenges O₂^·̅ without generating H₂O₂. Given the high toxicity of H₂O₂, such a property could result in a significant increase of the antioxidant capability of SOR in vivo, resulting in enhanced protection of the cell against oxidative damage. We tested this hypothesis by analyzing the phenotype of an E. coli sodA sodB recA mutant strain in which SOR was expressed in the presence or in the absence of ferrocyanide in the culture medium (Table 2). The E. coli mutant did not grow in the presence of oxygen because of the combined lack of superoxide dismutase (sodA sodB) and the DNA strand-break repair activity (recA), which results in lethal oxidative damage (14). Previous studies have shown that full overexpression of SOR efficiently compensates for the lack of SOD in E. coli (14). As shown in Table 2, moderate expression of SOR, achieved by induction of the sor gene with low IPTG concentrations, efficiently but not totally compensates the lack of SOD. Expression of an E. coli Mn SOD in the same experimental conditions leads to a similar phenotype (Table 2). When 1 mM ferrocyanide is added to the culture medium, the viability of the SOR expressing bacteria is increased by a factor of three. The difference also is significant with 5 mM ferrocyanide, which is the upper limit of apparent toxicity of ferrocyanide for E. coli cells (data not shown). Addition of ferrocyanide at 1 or 5 mM in the culture of the Mn SOD-expressing bacteria, did not induce any significant effect on the aerobic survival of the cell (Table 2). These data show a specific effect of the presence of ferrocyanide in increasing the antioxidant properties of the SOR-expressing system in vivo.

Table 2.

Effect of ferrocyanide and SOR production on the aerobic survival of a sodA sodB recA E. coli mutant (QC 2375)

[Ferrocyanide], mM	Aerobic survival,* %
	QC 2375 pJF119EH (control vector)	QC 2375 pMJ25 (Sor+)	QC 2375 pCWSOD (Sod+)
0	0.003	21 ± 4†	23 ± 6‡
1	0.003	72 ± 7‡	34 ± 12‡
5	nd	51 ± 10†	17 ± 5†

Anaerobic cultures of QC 2375 transformed with pJF119EH, pMJ25, or pCWSOD were plated on LB medium plus 2 × 10⁶ M IPTG and different concentrations of ferrocyanide under anaerobic and aerobic conditions. Colonies were counted after 24 h of incubation at 37°C. nd, not determined.

*Survival was calculated as the ratio of the number of colonies under aerobic conditions to those under anaerobic conditions. Values are the means of four experiments. 100% corresponds to 1.8–2.3 × 10⁸ colonies.

^†Tiny colonies, became larger after 48 h.

^‡Large colonies became larger after 24 h.

Soluble extracts were prepared from cells overexpressing SOR and grown in the presence of 1 mM ferrocyanide. After oxidation with ammonium persulfate, the UV-visible analyses of the soluble extracts reveal the presence of a 420 nm band, characteristic of ferricyanide. This band is not present in extracts from cells grown in the absence of ferrocyanide (Fig. 10, which is published as supporting information on the PNAS web site). In addition, size exclusion chromatography of the soluble extracts showed that ferrocyanide is associated with the SOR protein (Fig. 10). These data demonstrate that a SOR–Fe(CN)₆ complex is formed within the cells when ferrocyanide is added to the culture medium.

In the presence of NADH or NADPH as electron donors, cells extracts were found to catalyze the reduction of the oxidized forms of the SOR–Fe(CN)₆ complex (Table 3, which is published as supporting information on the PNAS web site). These data show that the soluble extracts contain NAD(P)H-dependent reductases that can regenerate the active form of the complex. The reduction rates of the complex were found similar to that reported for SOR alone (4). These data indicate that the reaction of the SOR–Fe(CN)₆ complex with O₂^·̅ can be catalytic within the cell.

Discussion

The recent crystal structure of SOR from D. baarsii revealed that the organometallic compound ferrocyanide makes a complex with the enzyme active site (8). This structure is the only example of a complex between ferrocyanide and an iron site of a metalloenzyme. The ferrocyanide entirely plugs the SOR active site, with several surface residues involved in hydrogen bonds with the cyanide moieties (Fig. 1). This intriguing structure initially suggested that ferrocyanide could be a potentially strong inhibitor of SOR activity.

Here, we show that, on the contrary, the SOR–ferrocyanide complex reacts very efficiently with O₂^·̅. However, the presence of the ferrocyanide adduct markedly modifies the reaction mechanism. Although the complex, like SOR alone, carries out one-electron reduction chemistry on O₂^·̅, H₂O₂ is no longer the final reaction product. The absence of H₂O₂ production in the reaction of the SOR–ferrocyanide complex with O₂^·̅ is in full agreement with the specific increase of the antioxidant capabilities of SOR expression in an E. coli sodA sodB recA mutant strain when complexed with ferrocyanide.

In Scheme 1, we propose a reaction mechanism of the SOR–ferrocyanide complex with O₂^·̅, in which the reduction process essentially is carried out by the ferrocyanide adduct. This mechanism is supported by the following points.

Scheme 1.

Proposed mechanism of the SOR–ferrocyanide complex with O₂^·̅.

Pulse radiolysis and binding experiments show that as soon as enough ferrocyanide is present in solution to saturate the SOR active site, the SOR–ferrocyanide complex reacts with O₂^·̅ with a bimolecular rate constant of 1.8 × 10⁸ M⁻¹·s⁻¹. This rate constant is ≈6 times lower than that obtained in the absence of ferrocyanide but still represents a very efficient scavenging activity toward O₂^·̅. From this bimolecular reaction, two intermediate species are formed sequentially, with different absorption spectra from those observed in the absence of ferrocyanide (Fig. 4 and Scheme 1). In both intermediates, the presence of an absorption band centered at 420 nm shows that the ferrocyanide (Fe²⁺) moiety of the complex has been oxidized into ferricyanide (Fe³⁺). The additional band centered at 580 nm in the first intermediate (Fig. 4A) does not support an oxidation of the SOR iron active site, because in that case, a broader band centered at 650 nm with a higher intensity would be expected (4, 10). It could be consistent with the presence of a peroxide species, as discussed below. For the second intermediate, quantification of the ferricyanide formation corresponds exactly to the amount of O₂^·̅ that reacted with the complex. No other absorbance bands are observed, suggesting that this intermediate corresponds to a SOR(Fe²⁺)–ferricyanide(Fe³⁺) complex. A one-electron redox chemistry therefore is carried out by the ferrocyanide moiety of the complex, whereas the SOR iron site remains in the reduced state (Scheme 1).

It should be noted that, because ferrocyanide is negatively charged, one would have expected strong electrostatic repulsions with the superoxide anion. However, when complexed to SOR, the electrostatic properties of the complex with the presence of positive charges near the ferrocyanide, Lys-48 for example (Fig. 1), could be one of the features that allow reaction of O₂^·̅ with the ferrocyanide adduct.

FTIR analysis of the final reaction products a few minutes after the reaction with O₂^·̅ show that the protein-bound Fe(CN)₆ is in a reduced state (Fig. 9), and the SOR active site is mostly in the oxidized state (Fig. 8). These data suggest that a redox equilibrium has settled between the various oxidized and reduced species present at the end of the reaction, on a longer time scale than that observed by pulse radiolysis (Scheme 1). This equilibrium most likely corresponds to an intramolecular electron transfer in the second intermediate (Scheme 1), similar to what occurs spontaneously when reduced SOR is mixed with ferricyanide, to form a SOR-Fe³⁺-CN-Fe²⁺(CN)₅ species (4, 11, 12). This complex is stable in solution, according to its K_d value of 0.48 ± 0.05 μM (Fig. 6). We note that the presence of only 0.7 mol of oxidized SOR generated per mol of O₂^·̅ together with the oxidation of 0.3 mol of free ferrocyanide per mol of O₂^·̅ into free ferricyanide indicates that the mixed-valence SOR-Fe³⁺-CN-Fe²⁺(CN)₅ has been reduced partially by free ferrocyanide (Scheme 1). From the data of Table 1, an equilibrium constant value of K ≈ 0.1 can be calculated for this reduction process.

No H₂O₂ is produced at the end of the reaction. The absence of H₂O₂ does not result from any catalase activity of the reacting solution, nor from inhibition of the HRP enzyme used to quantify H₂O₂. Also, we observed no evidence for Fenton-like reactions, which would lead to HO˙ production and destruction of the SOR–ferrocyanide complex. This result is in line with the electronic balance of the SOR ferrocyanide solution at the end of the reaction, which indicates that O₂^·̅ has been reduced with only one electron. To form HO˙ from O₂^·̅, two electrons are required.

Consequently, assuming a one-electron reduction process of O₂^·̅, a peroxide species should be formed, at least transiently, at the level of the first reaction intermediate. Reaction of this transient peroxide species with an organic solute present in the solution, formate or Tris buffer, could explain the absence of H₂O₂ formation at the end of the reaction. This hypothesis is supported by the dependence of the rate constant k₂ on formate concentration (Fig. 7). These data suggest that the peroxide intermediate species reacts with formate to form a performic acid at the end of the reaction, instead of H₂O₂ (Scheme 1). Because in SOR alone, the peroxide intermediate species formed during the catalytic cycle is quantitatively transformed into H₂O₂ and does not react with organic solutes, this reaction with formate appears to be specific to the SOR–ferrocyanide complex.

In a cellular context, such oxidation of an organic component to form an alkyl peroxide, weakly reactive as compared with the toxic H₂O₂ (1, 2), may result in a more efficient antioxidant property for the SOR–ferrocyanide complex than that of SOR alone. In fact, this is exactly what we observed in vivo on SOD-deficient E. coli cells complemented with the sor gene. The E. coli mutant strain sodA sodB recA does not grow at all in the presence of oxygen unless complemented with sor or sod genes (14). With sor expression, the addition of ferrocyanide in the culture medium increased by a factor of three the aerobic survival of the E. coli mutant strain. This effect was specific to the presence of SOR and was not observed with SOD expression. We showed that in these conditions, the SOR–ferrocyanide complex is formed within the cells and that it can act as a catalyst to detoxify O₂^·̅. The soluble cell extracts contain NAD(P)H-dependent reductases that can regenerate its reduced active form. Altogether, these data support the hypothesis that the SOR–ferrocyanide complex is more efficient in O₂^·̅ detoxification than SOR alone. The oxidized cellular components produced by the reaction of the SOR–ferrocyanide complex with O₂^·̅ could be much less damaging species than H₂O₂, possibly because they cannot be involved in the formation of the hydroxyl radicals by the Fenton reaction.

Addition of ferrocyanide as a cofactor to SOR provides a new detoxification mechanism for superoxide that may have a biological relevance. Although synthesis of ferrocyanide in cells has not been investigated to date, the presence of cyanide as a metal ligand has been well characterized in hydrogenases (15). Interestingly, hydrogenases are usually found in the same type of bacteria where SORs are naturally expressed and these cells possess enzymatic systems that can synthesize CN⁻, HypE and HypF (16), and incorporate it into proteins. These data leave open a possible formation of ferrocyanide in these bacteria. One may anticipate that formation of the SOR–ferrocyanide complex could be associated with specific oxidative stress conditions, iron overload for example, which make cells highly and specifically sensitive to H₂O₂.

Materials and Methods

Bacterial Strains and Plasmids.

E. coli strain QC2375 was previously described in ref. 14. Plasmid pMJ25 is a pJF119EH derivative in which the sor gene from D. baarsii is under the control of a tac promotor (14). The plasmid pCW-SODA (C. Weill, personal communication) is a pMJ25 derivative in which the sor gene is replaced by the E. coli sodA structural gene, allowing expression of the Mn SOD under the control of a tac promotor.

Protein Preparation.

The recombinant SOR from D. baarsii was purified as described in ref. 4. The protein was isolated with a reduced active site, not oxidized in the presence of oxygen.

Pulse Radiolysis Experiments.

Free radicals were generated by a 200-ns pulse of 4.5 MeV from a linear electron accelerator located at the Curie Institute (Orsay, France). O₂^·̅ was generated by scavenging radiolytically generated HO˙ free radicals by 100 mM formate in O₂ saturated solution/10 mM Tris·HCl (pH 7.6) (9). The doses per pulse were ≈5 Gy ([O₂^·̅] = 3 μM), calibrated with a thiocyanate dosimeter (9). The reaction was followed spectrophotometrically at 20°C, in a 2-cm path-length cuvette.

γ-Ray Radiolysis Experiments.

γ-Ray irradiations were carried out at 20°C by using a cobalt-60 source at a dose rate of 17.6 Gy·min⁻¹. The SOR solutions were made up in 10 mM Tris·HCl buffer (pH 7.6)/100 mM sodium formate and saturated with pure O₂ at room temperature. Concentrated potassium hexacyanoferrate (II) was added to the solution just before irradiation. The duration of irradiation was 5 min (88 Gy). Immediately after irradiation, UV-visible spectra of the solution were recorded in parallel with that of a nonirradiated solution kept under the same conditions. Hydrogen peroxide production was determined immediately after irradiation by using the leuco crystal violet HRP method as described in ref. 17.

Electrospray Ionization Mass Spectroscopy.

Mass spectra were obtained on a PerkinElmer (Thornhill, ON, Canada) Sciex API III+ triple quadrupole mass spectrometer equipped with a nebulizer-assisted electrospray source operating at atmospheric pressure. Samples were in 10 mM ammonium acetate.

FTIR Spectroscopy.

FTIR absorption spectra were recorded at 4 cm⁻¹ resolution by using a Bruker 66 SX spectrometer equipped with a KBr beam splitter and a nitrogen-cooled MCT-A detector. Each spectrum corresponds to 300 scans. Fifteen microliters of the liquid sample was deposited between two CaF₂ windows separated with a ≈25-μm spacer.

Abbreviations

SOD

superoxide dismutase

SOR

superoxide reductase.