Reaction of the desulfoferrodoxin from Desulfoarculus baarsii with superoxide anion. Evidence for a superoxide reductase activity

Desulfoferrodoxin is a small protein found in sulfate-reducing bacteria that contains two independent mononuclear iron centers, one ferric and one ferrous. Expression of desulfoferrodoxin from Desulfoarculus baarsii has been reported to functionally complement a superoxide dismutase deficient Escherichia coli strain. To elucidate by which mechanism desulfoferrodoxin could substitute for superoxide dismutase in E. coli, we have purified the recombinant protein and studied its reactivity toward O-(2). Desulfoferrodoxin exhibited only a weak superoxide dismutase activity (20 units mg(-1)) that could hardly account for its antioxidant properties. UV-visible and electron paramagnetic resonance spectroscopy studies revealed that the ferrous center of desulfoferrodoxin could specifically and efficiently reduce O-(2), with a rate constant of 6-7 x 10(8) M(-1) s(-1). In addition, we showed that membrane and cytoplasmic E. coli protein extracts, using NADH and NADPH as electron donors, could reduce the O-(2) oxidized form of desulfoferrodoxin. Taken together, these results strongly suggest that desulfoferrodoxin behaves as a superoxide reductase enzyme and thus provide new insights into the biological mechanisms designed for protection from oxidative stresses.

Desulfoferrodoxin is a small protein found in sulfatereducing bacteria that contains two independent mononuclear iron centers, one ferric and one ferrous. Expression of desulfoferrodoxin from Desulfoarculus baarsii has been reported to functionally complement a superoxide dismutase deficient Escherichia coli strain. To elucidate by which mechanism desulfoferrodoxin could substitute for superoxide dismutase in E. coli, we have purified the recombinant protein and studied its reactivity toward O 2 . . Desulfoferrodoxin exhibited only a weak superoxide dismutase activity (20 units mg ؊1 ) that could hardly account for its antioxidant properties. UVvisible and electron paramagnetic resonance spectroscopy studies revealed that the ferrous center of desulfoferrodoxin could specifically and efficiently reduce O 2 . , with a rate constant of 6 -7 ؋ 10 8 M ؊1 s ؊1 . In addition, we showed that membrane and cytoplasmic E. coli protein extracts, using NADH and NADPH as electron donors, could reduce the O 2 . oxidized form of desulfoferrodoxin.
Taken together, these results strongly suggest that desulfoferrodoxin behaves as a superoxide reductase enzyme and thus provide new insights into the biological mechanisms designed for protection from oxidative stresses.
Desulfoferrodoxin (Dfx) 1 is a small, nonsulfur iron protein that has been isolated from several strains of anaerobic sulfatereducing bacteria (1,2). Although no enzymatic activity could be associated to Dfx, the physicochemical properties of its iron centers have been well documented (1)(2)(3). Recently, the threedimensional structure of Dfx from Desulfovibrio desulfuricans has been solved at a resolution of 1.9 Å (4). Dfx is a homodimer with a molecular mass of 2 ϫ 14 kDa. The monomer is organized in two protein domains, each with a specific mononuclear iron center named center I or center II. Center I contains a mononuclear ferric iron coordinated by four cysteines in a distorted rubredoxin-type center. Center II has a ferrous iron with square pyramidal coordination to four nitrogens from histidines as equatorial ligands and one sulfur from a cysteine as the axial ligand (4). The midpoint redox potentials have been reported to be 2-4 mV for center I and 90 -240 mV for center II (2,3). The high redox potential value for center II explains the stability of the ferrous ion in the presence of oxygen.
Initially, the structural dfx gene was cloned and sequenced from Desulfovibrio vulgaris Hildenborough and was named rbo (5). rbo was found upstream of the rubredoxin gene, forming an operon. The encoded 14-kDa protein was tentatively named rubredoxin oxidoreductase (Rbo) because it was likely to function in oxidation-reduction with rubredoxin as a redox partner (5). Independently, a protein isolated from D. desulfuricans and D. vulgaris and named Dfx was found to be encoded by the rbo gene (1,2,6). However, up to now, Dfx did not show any evidence for a rubredoxin oxidoreductase activity, and its physiological role remains unclear. Consequently, the name of the corresponding gene changed from rbo to dfx.
Recently, Pianzzola et al. (7) attempted to clone the sod gene from the sulfate-reducing bacteria Desulfoarculus baarsii by functional complementation of a superoxide dismutase (SOD)deficient mutant of Escherichia coli. They actually found a complementing gene showing high sequence identity with dfx from D. vulgaris. However, although expression of dfx could fully complement the SOD phenotype, no SOD activity could be detected in vivo, raising the question of the functional basis for the successful complementation (7,8). Furthermore, it has been shown that deletion of the dfx gene increased the oxygen sensitivity of D. vulgaris when exposed to transitory aerobic conditions (9). This strongly suggested that in these anaerobic bacteria, Dfx plays a role in the defense against oxidative stress.
In the present work, we have purified the recombinant Dfx from D. baarsii, and we have investigated the mechanism by which Dfx could trap superoxide and thus replace superoxide dismutase in E. coli.
Overproduction and Purification of Dfx-E. coli QC774 pMJ25 cells were grown aerobically at 37°C in M9 minimal medium complemented with 0.4% glucose, 2 g/ml thiamin, 1 mg/ml casamino acid, 2 mM isopropyl-1-thio-␤-D-galactopyranoside, 1 mM FeSO 4 ⅐7 H 2 O, and 200 g/ml ampicillin (12 ϫ 1000 ml in a 2-liter Erlenmeyer flask). Growth was performed overnight until the culture reached an A 600 nm of about * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
2.3. All following operations were carried out at 4°C and pH 7.6. The cells were collected by centrifugation. The cell pellet (50 g, wet weight) was suspended in 150 ml of 0.1 M Tris/HCl and sonicated. After ultracentrifugation at 45,000 rpm during 90 min in a Beckman 50.2 Ti rotor, the supernatant was used as soluble extract for further purification. Soluble extract (160 ml, 800 mg of protein) was treated with streptomycin sulfate (final concentration, 2% w/v) during 1 h of stirring and then centrifuged for 20 min at 14,000 ϫ g. The supernatant was treated with 2-3 mg of pancreatic DNase for 15 min and then precipitated with ammonium sulfate (final concentration, 80% w/v). After centrifugation (20 min at 14,000 ϫ g), the pellet was dissolved in 0.1 M Tris/HCl (10 ml, 80 mg/ml), and the solution was loaded onto an ACA 54 column (360 ml) equilibrated with 20 mM Tris/HCl. Proteins were eluted at a flow rate of 0.4 ml⅐min Ϫ1 with the same buffer. A pink fraction corresponding to the volume of elution of low molecular weight protein was collected in a total volume of 57 ml (2 mg/ml). At this stage, the protein solution had already a distinct visible spectrum that resembled that of Dfx. The ratio A 280 nm /A 503 nm was 7. SDS-PAGE on this fraction exhibited three major protein bands located at about 16, 14, and 10 kDa. Protein fractions of 10 mg were then further chromatographed using a Bio-Rad Biologic system on an anion exchange column, Uno Q (Bio-Rad), equilibrated with 10 mM Tris/HCl. A linear gradient was applied (0 -0.2 M NaCl) in 10 mM Tris/HCl, with a flow rate of 1 ml⅐min Ϫ1 during 30 min.
Analytical Determination-SDS-PAGE gels (15% polyacrylamide) were done according to Laemmli (10). The gels were calibrated with the Amersham Pharmacia Biotech low molecular weight markers. The native molecular mass of the protein was determined with a Superdex 75 gel filtration column (120 ml, Amersham Pharmacia Biotech) equilibrated with 25 mM Tris/HCl, pH 7.6, and 150 mM NaCl using a flow rate of 0.4 ml⅐min Ϫ1 . Bovine serum albumin (66 kDa), ovalbumin (45 kDa), trypsine inhibitor (20.1 kDa), and cytochrome c (12.4 kDa) were used as the markers for molecular mass. The void volume was determined with ferritin (450 kDa). Protein concentration was determined using the Bio-Rad protein assay reagent (11) with bovine serum albumin as a standard. Protein-bound iron was determined by atomic absorption spectroscopy. UV-visible spectra were recorded on a Varian Carry 1 spectrophotometer using a 1-cm-path quartz cuvette. EPR measurements were made on a Bruker EMX 081 spectrometer equipped with an Oxford Instrument continuous flow cryostat.
N-terminal Sequence Analysis-The proteins were separated by SDS-PAGE and then transferred to a ProBlot TM membrane (Applied Biosystem) as described by the manufacturer. NH 2 -terminal amino acid sequence determination was performed using an Applied Biosystems gas phase separator (model 477A) with on-line analysis of the phenylthiohydantoin derivatives.
Mass Spectrometry-Mass spectra were obtained on a Perkin-Elmer Sciex API IIIϩ triple quadrupole mass spectrometer equipped with a nebulizer assisted electrospray source (ionspray) operating at atmospheric pressure.
Assays for SOD Activity-The SOD activities were evaluated using the cytochrome c reduction assay modified from McCord and Fridovich (12). The assay was performed at 25°C in 3 ml of reaction buffer (50 mM potassium phosphate, pH 7.6) containing 22 M cytochrome c, 200 M xanthine, 500 units/ml catalase, and an amount of xanthine oxidase that gives an initial rate of ⌬A 550 nm ϭ 0.025 per min in the absence of SOD activity. Reduction of ferricytochrome c was followed at 550 nm, and rates were linear for at least 4 min. One unit of SOD is defined as the amount of protein that inhibits the rate of the reduction of ferricytochrome by 50%. The SOD activities were also evaluated using another test, the NBT reduction technique, modified from Beauchamp and Fridovich (13). The assay was performed at 25°C in 3 ml of reaction buffer (50 mM Tris/HCl, pH 7.6) containing 22 M NBT, 200 M xanthine, 500 units/ml catalase, and an amount of xanthine oxidase that gives an initial rate of ⌬A 560 nm ϭ 0.0165 per min in the absence of SOD activity. Reduction of NBT was followed at 560 nm during 0.5 min, in order to keep absorbance changes fairly low and avoid precipitation of formazan. One unit of SOD is defined as the amount of protein that inhibits the rate of the reduction of NBT by 50%. Under our experimental conditions, using the same commercial CuZn-SOD preparation, NBT test gave a 4-fold more elevated SOD specific activity compared with that determined with the cytochrome c reduction test.
Preparation of Cell Extracts and Reduction of Center II-Aerobic cultures of E. coli strain QC774 pMJ25 were grown in Luria-Bertani medium at 37°C and were harvested at about 0.3 A 600 nm . All of the following steps were performed at 4°C. Cultures were centrifuged, washed in cold 50 mM potassium phosphate buffer, pH 7.8, and resuspended in about 3% of the original culture volume in the same buffer. The cells were then sonicated and centrifuged for 20 min at 6000 rpm to remove cell debris. The supernatant, which contained both cytosolic and membrane material, was fractionated by a centrifugation at 45,000 rpm (Beckman, 70.1 Ti rotor) for 2 h. The supernatant cytosolic fraction (2.3 mg/ml), was removed and stored at Ϫ80°C. The pelleted membranes vesicles were resuspended in 50 mM potassium phosphate buffer, pH 7.8, plus 100 mM NaCl and recentrifuged. The resulting pellet was resuspended in 0.2% of the original culture volume in 50 mM potassium phosphate buffer, pH 7.8. The membrane fractions were stored at 0°C and used within a week.
Oxidation of Dfx-KO 2 stock solutions were prepared as followed. Me 2 SO was dried over 3-Å molecular sieves. Potassium superoxide was dissolved in anhydrous Me 2 SO under a dry atmosphere of argon. No crown ether was added. The concentration of superoxide was determined by using its absorbance in the UV with an extinction coefficient of 2086 M Ϫ1 cm Ϫ1 at 260 nm. Me 2 SO stock solutions of potassium superoxide (1-2 mM) were prepared immediately before each experiment. The fully oxidized form of Dfx was also obtained by the addition of a 2-fold molar excess of potassium ferricyanide to the purified Dfx. Complete oxidation of center II was verified by UV-visible spectrophotometry. Excess of potassium ferricyanide was eliminated by washing with 50 mM Tris/HCl, pH 7.6, using a Centricon pM30.

RESULTS
Purification of the Recombinant Dfx-In order to avoid any possible contamination by SOD activities from the host strain, the dfx gene was overexpressed in the E. coli QC 774 strain, in which the sodA and sodB genes were insertionally inactivated (7). A two-step purification protocol, with a gel filtration on ACA 54 and anion exchange chromatography on Uno Q (Bio-Rad) as described under "Experimental Procedures," was set up. Samples were analyzed at all steps by SDS-PAGE and UV-visible spectroscopy, because Dfx exhibits characteristic absorption bands responsible for the pink color of the protein (1). During Uno Q chromatography, a pink fraction, eluted at 50 mM NaCl, was collected (fraction A). It contained only two polypeptide bands at 14 and 16 kDa with about equal intensities, as shown by SDS-PAGE analysis. Another pink fraction was eluted at a slightly higher NaCl concentration (fraction B). In addition to the 14-and 16-kDa polypeptides, fraction B contained also a major polypeptide of 10 kDa, as shown by SDS-PAGE.
The 14-and 16-kDa polypeptides had the same PER-LQVYKCE N-terminal sequence, identical to the N-terminal translated sequence of the D. baarsii dfx structural gene, without the N-terminal Met residue (7). Furthermore, they had the same mass, as shown by electrospray mass spectrometry analysis, with only ions detected at 14,028 Ϯ 2 Da, confirming the absence of the N-terminal Met residue. These data suggested that the 14-and 16-kDa bands seen in SDS-PAGE originated from the same Dfx polypeptide. In fact, the proportion of the two polypeptide bands was found to be correlated to the presence of the reducing agent, dithiothreitol or ␤-mercaptoethanol, during electrophoresis (data not shown). From 800 mg of soluble extracts, 23 mg of pure Dfx (fraction A) were obtained.
The 10-kDa protein, present only in fraction B, exhibited the N-terminal ADAQKAADNKKPVN sequence, which is identical to the N-terminal sequence of the mature form of HdeA (14). HdeA is a highly abundant periplasmic protein of E. coli of unknown function (14). Fraction B appeared then not homogeneous and was not further characterized.
Finally, during Uno Q chromatography, a protein peak, eluted from the column at 90 mM NaCl, also contained the 14and 16-kDa proteins, together with other minor contaminants. However, this fraction was colorless and did not display the UV-visible spectrum characteristic of Dfx. It thus probably contained the apo form of Dfx and was thus discarded.
Gel-filtration experiments on Superdex 75 column with the native recombinant Dfx, as described under "Experimental Procedures," gave an apparent molecular mass of 27,000 Da (data not shown), showing that Dfx from D. baarsii is a homodimer.
Iron Content/Absorption Spectra of the Recombinant Dfx-The iron content of Dfx was determined by atomic absorption spectrophotometry. A value of 1.97 Fe/polypeptide chain (14,026 Da) was found, suggesting that both center I and center II were fully metallated.
As shown in Fig. 1, the UV-visible spectrum of the as-isolated Dfx exhibits absorptions at 370 and 503 nm contributed by the ferric iron from center I (1, 2). The ratio A 280 nm /A 503 nm was 4.5. The value of the molar extinction coefficient at 503 nm was determined to be 4,400 M Ϫ1 cm Ϫ1 (center I). This value was similar to the corresponding value reported for D. desulfuricans and D. vulgaris Dfx (1,2). Fig. 1 also shows the spectrum of the protein treated with an excess of potassium ferricyanide. It is characteristic of the gray form of Dfx, with absorptions contributed by the ferric irons from both centers I and II (3). In the inset to Fig. 1, the difference spectrum provides the contribution of the oxidized center II with absorption bands centered at 644 and 330 nm (3). The value of the molar coefficient at 644 nm was found to be 1,900 M Ϫ1 cm Ϫ1 (center II).
EPR Spectroscopy Analysis-The EPR spectrum of the protein, recorded at 4 K, displays resonances at g ϭ 7.7, 5.7, 4.1, and 1.8 (data not shown). It is similar to that reported for the pink form of Dfx from D. desulfuricans (1) and D. vulgaris (2). It is typical for a distorted FeS 4 center (S ϭ 5/2), assigned to center I (1, 2). When the EPR spectrum was recorded at 10 K, the intensity of the g ϭ 7.7 feature decreased, whereas that of the g ϭ 5.7 feature increased (data not shown). This is consistent with the former being derived from the ground state and the latter from an excited state, as previously reported (1). The 4 K EPR spectrum of the oxidized protein presents a signal at g ϭ 4.3, in addition to the features at g ϭ 7.7, 5.7, and 1.8 (data not shown). This spectrum is similar to that reported for the ferric form of Dfx from D. desulfuricans (3) and from D. vulgaris (2), and the signal at g ϭ 4.3 was attributed to oxidized center II (2,3).
Dfx Is Not a Superoxide Dismutase-The ability of Dfx to catalyze the dismutation of O 2 . was assayed from its inhibitory effect on the reduction of cytochrome c by superoxide, generated by the xanthine-xanthine oxidase system (12). In Fig. 2A are shown the traces of the reduction of cytochrome c in the presence of different amounts of Dfx. With up to 11 g of Dfx, almost no inhibition of cytochrome c reduction could be observed (data not shown). However, larger amounts of Dfx resulted in a two-phase kinetics. An initial lag period, corresponding to a complete inhibition of cytochrome c reduction, was observed with these larger amounts ( Fig. 2A). The dura-tion of the lag period was found to be roughly proportional to the amount of Dfx added in the test cuvette ( Fig. 2A, inset). In the second phase of the reaction, formation of reduced cytochrome c appeared linear with time, but with a slope that slightly decreased with increased Dfx concentration ( Fig. 2A). When Dfx was pretreated with the superoxide-generating system (xanthine-xanthine oxidase) and then assayed for inhibition of cytochrome c reduction, the lag phase could not be observed anymore (Fig. 2B). However, increased amounts of the preincubated Dfx inhibited the reduction of cytochrome c. The addition of 100 g of preincubated Dfx resulted in 50% inhibition of cytochrome c reduction (Fig. 2B, inset), a value corresponding to a specific SOD activity for the preincubated Dfx of 20 units mg Ϫ1 . A comparable value of specific SOD activity was found from the linear second phase of the kinetic of Fig. 2A (data not shown).
When a Dfx solution (39 M in 50 mM Tris/HCl, pH 7.6) was pretreated with a 5-fold molar excess of O 2 . (KO 2 dissolved in Me 2 SO), in the presence of 500 units/ml catalase, results comparable to those shown in Fig. 2B were obtained (data not shown). Similar results were obtained using another assay, the socalled NBT reduction assay (13). The superoxide-dependent reduction of NBT by the xanthine-xanthine oxidase system was inhibited by large amounts of Dfx (data not shown). Kinetics of reduction were linear for at least 0.5 min, and no lag time was observed in the presence of Dfx. 9 g of Dfx induced a 50% inhibition of reduction of NBT (data not shown), a value corresponding to a specific SOD activity of 25 units mg Ϫ1 in the cytochrome c assay. Preincubation of a concentrated Dfx solution with the xanthine-xanthine oxidase system (as reported in Fig. 2B) before the assay gave a comparable value of the specific SOD activity (data not shown).
Superoxide Oxidizes Efficiently Dfx Center II-The results from the cytochrome c and NBT assays suggested that Dfx did not exhibit a significant superoxide dismutase activity. However, the observation, in the cytochrome c assay, of a lag period proportional to the amount of added Dfx ( Fig. 2A) suggested that during this period, O 2 . reacted with Dfx rather than with cytochrome c. As shown in Fig. 3A, incubation of Dfx with the O 2 . generating system (xanthine-xanthine oxidase plus catalase) induced an increase of the protein absorbance in the 600 -700 nm range. Difference spectra clearly showed a specific oxidation of center II, with the appearance of the band centered at 644 nm (Fig. 3A, inset). Under these conditions, after 10 min of incubation, the oxidation of center II was complete. Longer incubation time did not further modify the spectrum of the fully oxidized Dfx (data not shown). The same results were obtained in the absence of catalase (data not shown). induced a complete and selective oxidation of center II, as shown by the increase of the band at 644 nm (Fig. 3B, inset). Addition of an equivalent amount of Me 2 SO without KO 2 had no effect on the visible spectrum of Dfx (data not shown In the presence of high amounts of CuZn-SOD (Fig. 4A) or Fe-SOD (Fig. 4B) Fig. 4, A and B, insets, shows a linear plot of the reciprocal of the initial rate of oxidation of center II (v ox ) versus CuZn-and Fe-SOD concentration, respectively, according to Equation 7. Under these conditions, when the initial rate of the oxidation of center II is decreased by 50% due to the competition of SOD for O 2 . , Equations 4 and 5 can be rearranged to give the following equation.
The concentrations of CuZn-and Fe-SOD that decrease by 50% the rate of oxidation of center II were then graphically determined from the Fig. 4, A and B can be now calculated using Equation 8. Values of 6.8 ϫ 10 8 and 6.5 ϫ 10 8 M Ϫ1 s Ϫ1 in the experiments using CuZn-and Fe-SOD, respectively, were obtained.

Center II of Dfx Is Slowly Oxidized by H 2 O 2 -
The experiments presented above have been set up in the presence of catalase in order to eliminate a possible effect of H 2 O 2 , the superoxide reduction and dismutation product. The ability of H 2 O 2 to react with Dfx and to oxidize center II was nevertheless tested spectrophotometrically. When 100 M Dfx, in 50 mM Tris/HCl, pH 7.6, was incubated with 1 mM H 2 O 2 , the UVvisible spectrum exhibited an increase of the absorbance in the 600 -700 nm range, during the first 5 min of reaction (data not shown). Difference spectra clearly showed a complete oxidation of center II, with appearance of the band centered at 644 nm (data not shown). In the presence of 500 units/ml of catalase, no modification of the UV-visible spectrum could be observed (data not shown). In the kinetic of the oxidation of center II (100 M Dfx in 50 mM Tris/HCl, pH 7.6) by 0.8, 1, 1.5, or 2 mM H 2 O 2 was followed spectrophotometrically at 644 nm, at 25°C. In all cases, oxidation of center II was found to follow a pseudofirst order kinetic (data not shown). A value of the second order rate constant of oxidation of center II by H 2 O 2 of 45 M Ϫ1 s Ϫ1 was determined. This is almost negligible compared with the value of the rate constant of oxidation of center II by O 2 . (see above).

Reduction of Center II by Cell Extracts-
The results presented above have shown that the reduced form of center II of D. baarsii Dfx can transfer one electron to O 2 . very efficiently.
In order to provide evidence that such a reaction could be catalytic within the cell, we have examined the capability of E. coli cell extracts to reduce the oxidized form of center II, which then could be involved in a new reaction cycle with O 2 . .
The fully oxidized form of Dfx was incubated anaerobically with catalytic amounts of cytosolic or membrane cell extracts, in the presence of NADH or NADPH as electron donors (Table  I). Time-dependent reduction of Dfx was followed spectrophotometrically, using a diode array spectrophotometer (data not shown). Both cytosolic and membrane cell fractions were found to catalyze electron transfer to the center II of Dfx. Complete reduction of center II was observed in the presence of NADH and membrane fractions or in the presence of NADPH and cytosolic fractions (data not shown). In all cases, no evidence for reduction of center I was observed during the time course of the reduction of center II (data not shown). However, longer incubation of Dfx with NADH and membrane fractions or NADPH and cytosolic fractions led to complete reduction of center I, giving the fully reduced form of Dfx (data not shown).
As illustrated in Table I, the rate of reduction of center II depended both on the electron donor and the cell fraction. NADH and the membrane fraction or NADPH and the cytosolic fraction gave the higher rate of reduction of center II, with specific activity values of 90 and 120 nmol of center II reduced/ min/mg of extract, respectively. On the other hand, NADH and the cytosolic fraction were poorly active with a specific activity value of 20 nmol/min/mg. NADPH and the membrane fraction were also found to reduce center II but with a low specific activity of 20 nmol/min/mg. Taking into account that no NADPHdependent reductase should be associated with the membrane fraction in E. coli, this residual NADPH-dependent reductase activity could originate from cytosolic contaminant proteins.

DISCUSSION
To identify the mechanism by which Dfx from D. baarsii could protect a SOD-deficient E. coli strain from a superoxide stress, we have investigated the reactivity of Dfx with regard to the superoxide anion.
Our results show that Dfx exhibits a very weak SOD activity (20 units/mg), representing about 0.3% of the specific activity of a CuZn-or Fe-SOD, assayed under comparable conditions. While this study was under way, a comparable low SOD activity of Dfx from D. desulfuricans was also reported (20). The finding that Dfx did not efficiently catalyze the dismutation of O 2 . is not surprising, because no sequence similarity was found between Dfx and any class of SOD characterized so far (7). Furthermore, no SOD activity could be detected in extracts of SOD deficient E. coli strain overproducing Dfx (7,8).
Whether this low SOD activity could nevertheless account for the functional complementation of the SOD deficient E. coli strain when Dfx is expressed within the cell is questionable. Recent results from Gort and Imlay (21) showed that E. coli can tolerate only small decreases in SOD content, and E. coli constitutively synthesizes just enough SOD (Fe-SOD) for protection from endogenous O 2 . . One can estimate that under the overexpression conditions used during the complementation experiments (7), Dfx represents no more than 5% of the total soluble proteins, corresponding to less than 1 unit mg Ϫ1 of SOD activity. Such an amount of SOD activity is certainly too low to protect the cell from fractions of E. coli contained NADH or NADPH reductase activities that may fulfill this function. As expected, the membrane reductase(s) were found to be NADH-dependent, and the cytosolic reductase(s) were found to be rather NADPH-specific. The values of the specific activities of reduction of center II (Table I) are in the range of specific activities reported in crude extracts for many enzymatic systems in E. coli, in agreement with a possible in vivo catalytic reduction of center II.
Altogether, these data strongly support a superoxide reductase activity for D. baarsii Dfx, as previously hypothesized by Liochev and Fridovich (8), which could account for the functional complementation of the SOD-deficient mutant of E. coli strain. Efforts to purify the putative E. coli NADH-or NADPHdependent Dfx reductases are currently under way and would allow to determine the global kinetic parameters for the reduction of superoxide catalyzed by Dfx.
On the other hand, no obvious function could be assigned to Dfx center I yet. The hypothesis that center I could act as an electron relay between cellular reductases and center II was attractive but is not clearly supported by our results. E. coli extracts did not seem to reduce efficiently Dfx center I in the presence of NADH or NADPH as electron donors. In addition, the three-dimensional structure of D. desulfuricans Dfx indicates a distance of 20 Å between center I and II (4), which hardly supports possible electronic interactions between the two redox centers. Further investigations are needed to understand the function of center I in Dfx.
Recently, it has been shown that deletion of dfx gene increases the oxygen sensitivity of D. vulgaris when exposed transitory to microaerophilic conditions (9). Because dfx deletion does not affect growth of D. vulgaris under anaerobic conditions, it was proposed that the main physiological function of Dfx is that of an antioxidant protein in Desulfovibrio spp. (9). This is in line with the functional complementation by Dfx of sodA sodB mutant in E. coli (7), and we propose that Dfx could protect cell against oxidative stress by the same mechanism in both E. coli and in sulfate-reducing bacteria. What could then be the advantage for a cell to have a mechanism of elimination of superoxide by reduction rather than by catalyzing its dismutation with a SOD enzyme ? That Dfx is the survivor of an ancestral system of O 2 . elimination could be considered, but we would favor a more specific function of Dfx, taking into account the particular redox status in sulfate-reducing bacteria. Anaerobic bacteria, and in particular sulfate-reducing bacteria, are known to usually be highly sensitive to exposure to air; during this exposure, a whole array of enzymes and proteins None 0 a Reduction was followed anaerobically at 17°C in a cuvette (0.1-ml final volume) containing 110 M fully oxidized Dfx, 50 mM Tris/HCl, pH 7.6, and 600 M NADPH or NADH. The reaction was initiated by adding 5-20 g of cell extract. UV-visible spectra (800 -260 nm) were taken every 0.6 min using a HP 8453 diode array spectrophotometer. Initial velocities of reduction of center II were calculated from the decrease of absorption at 644 nm.
b One unit of activity is defined as the amount of cell extract catalyzing the reduction of 1 nmol of center II per min.
can be totally inactivated (24). Furthermore, sulfate-reducing bacteria are fully crowded with strongly auto-oxidizable redox proteins, such as redox carriers (ferredoxin, cytochromes, rubredoxin, desulforedoxin, and flavodoxin) or enzymes, such as hydrogenases (24). Upon exposure to O 2 , these proteins are prone to release their electrons, thus inducing a strong superoxide stress (25). Such a process is probably less important in aerobic cells, which have evolved by integrating the electron transport proteins into the membrane in order to minimize such auto-oxidation reactions (22). SOD and catalase have been found in a few sulfate-reducing bacteria (25,26) and could well account for the good aerotolerance that has been reported in these species (26,27). However, the presence of Dfx in these bacteria may provide an additional advantage. It is tempting to suggest that Dfx, by shuttling the electrons from the autooxidizable redox proteins to superoxide preferentially, in a single reaction, could eliminate both superoxide and the source of its production. Another advantage would be that oxidation of redox carriers by Dfx stops as soon as the superoxide stress is over, restoring anaerobic function, without further loss of reducing equivalents. Finally, such a reaction allows these anaerobic bacteria to shut off transitory superoxide production from those redox carriers with no need for sophisticated regulatory systems, such as those found in facultative anaerobes.