A Primary Role for Disulfide Formation in the Productive Folding of Prokaryotic Cu,Zn-superoxide Dismutase*

Background: Prokaryotic Cu,Zn-superoxide dismutase (SodC) forms an intramolecular disulfide bond. Results: Disulfide formation is essential for folding and enzymatic activation of SodC. Conclusion: The thiol-disulfide status controls the intracellular stability of SodC. Significance: The oxidizing environment of the periplasm is required for antioxidant activity of SodC. Enzymatic activation of Cu,Zn-superoxide dismutase (SOD1) requires not only binding of a catalytic copper ion but also formation of an intramolecular disulfide bond. Indeed, the disulfide bond is completely conserved among all species possessing SOD1; however, it remains obscure how disulfide formation controls the enzymatic activity of SOD1. Here, we show that disulfide formation is a primary event in the folding process of prokaryotic SOD1 (SodC) localized to the periplasmic space. Escherichia coli SodC was found to attain β-sheet structure upon formation of the disulfide bond, whereas disulfide-reduced SodC assumed little secondary structure even in the presence of copper and zinc ions. Moreover, reduction of the disulfide bond made SodC highly susceptible to proteolytic degradation. We thus propose that the thiol-disulfide status in SodC controls the intracellular stability of this antioxidant enzyme and that the oxidizing environment of the periplasm is required for the enzymatic activation of SodC.

Cu,Zn-superoxide dismutase (SOD1) 2 is an enzyme that converts superoxide anion into oxygen and hydrogen peroxide at a catalytically active site of a bound copper ion (1). In addition to the copper-binding site, a typical form of SOD1 is also equipped with a zinc-binding site and an intramolecular disulfide bond (2). It is, however, important to note that the copperand zinc-binding sites are not always conserved among SOD1 proteins. For instance, a zinc-binding site is missing in SOD1 from Mycobacterium tuberculosis (3), and an SOD1-like protein from Bacillus subtilis does not even bind a catalytic copper ion (4). On the other hand, the intramolecular disulfide bond is completely conserved in all species possessing SOD1, suggesting its crucial role in the structure and function of SOD1.
In eukaryotic SOD1, the overall structural fold is not significantly affected by the thiol-disulfide status (5), but formation of the disulfide bond is essential for enzymatic activity by controlling orientation of the arginine residue that optimally guides superoxide anion to the catalytic copper site (6). Also, an enzymatically active form of eukaryotic SOD1 exists as a non-covalent homodimer, but reduction of the intramolecular disulfide bond facilitates monomerization and eventually misfolding of SOD1, which has been observed as a pathological change in a subset of amyotrophic lateral sclerosis (2). Formation of the intramolecular disulfide bond is, therefore, a critical factor for SOD1 to assume its enzymatically competent conformation.
Notably, the activation process of eukaryotic SOD1, including its metal binding and disulfide formation, is well regulated in the cell. Binding of a zinc ion has been proposed to first occur in SOD1 (2), although the mechanism of zinc ion acquisition remains totally unknown. Next, this zinc-bound form of SOD1 is specifically recognized by a copper chaperone protein, CCS, and a copper ion is then supplied from CCS to SOD1 (7). Simultaneously, a disulfide bond is introduced in SOD1 by CCS, which facilitates the homodimerization of SOD1 and finalizes the enzymatic activation of SOD1 (6). Given that eukaryotic SOD1 is mainly localized in the highly reducing cytoplasm (8), additional care should be taken to retain the disulfide bond. Indeed, x-ray structural analysis of SOD1 has shown that the disulfide bond is buried at the dimer interface (9); therefore, dimerization of SOD1 presumably protects the disulfide bond from its reductive cleavage in the cytoplasm.
In addition to the CCS-dependent activation of SOD1, a CCS-independent pathway(s) has been proposed for the maturation of SOD1 in several eukaryotes (10), and interestingly, prokaryotes do not possess a gene corresponding to CCS. A thiol-disulfide oxidoreductase, DsbA, has been suggested to form the disulfide bond in prokaryotic SOD1 (often called SodC) (11); however, the activation process of SodC remains unknown. Unlike eukaryotic SOD1, SodC is localized in the periplasmic space, where free copper/zinc ions would be available for activation of SodC through a simple diffusion process (12). Furthermore, the periplasmic space provides a more oxidizing environment than the cytoplasm (13), implying no need for metallochaperones to introduce the disulfide bond as well as metal ions and also no need to protect the disulfide bond from reduction.
Indeed, not all SodC proteins form homodimers; notable examples include the monomeric SodC from Escherichia coli (14), Salmonella enterica (15), and B. subtilis (4), in which the disulfide bond is highly solvent-exposed. Although SodC activity has been considered to play important roles in the pathogenicity of bacteria (16), it remains unknown how enzymatic activation of SodC is regulated, and the role of the disulfide bond in the physiological function of SodC is obscure.
In this study we have used E. coli SodC (17) as a model for a monomeric SOD1 of prokaryotes and found that the intramolecular disulfide bond is essential for the folding and intracellular stability of SodC. Unlike eukaryotic SOD1, E. coli SodC without the disulfide bond has little secondary structure, albeit with ability to bind a copper ion. Reductive cleavage of the disulfide bond was, furthermore, found to make SodC highly susceptible to proteolytic degradation in vivo as well as in vitro. Based upon these results, we propose that the intracellular activity of prokaryotic SodC is regulated by its thiol-disulfide status.

EXPERIMENTAL PROCEDURES
Preparation of SodC Proteins-A cDNA of E. coli SodC without the N-terminal periplasmic signal sequence was cloned into a pET-15b plasmid (Novagen) where the thrombin cleavage recognition sequence was replaced with a HRV3C cleavage recognition sequence. The C74A/C169A mutations (the numbering is based on the entire sodC gene including the N-terminal signal peptide) were inserted by the inverse PCR method using a KOD-Plus-Neo DNA polymerase (TOYOBO). All constructs in this study were confirmed by DNA sequencing.
E. coli BL21(DE3) (New England Biolabs) was transformed with the respective plasmids, and expression of SodC proteins was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside at 37°C for 6 h in LB medium with ampicillin. Cells were lysed with a cycle of freeze-thaw, resuspended, and ultrasonicated in PBS with 2% Triton X-100, 5 mM MgSO 4 , and 8 mg/liter DNase I. The soluble supernatant after centrifugation at 20,000 ϫ g for 20 min was filtered through a 0.22-m syringe filter and loaded onto a Profinity IMAC Ni 2ϩ resin (Bio-Rad). The resin was washed with 50 mM Tris, 500 mM NaCl, 10 mM imidazole, pH 8, and then the bound SodC proteins were eluted with 50 mM Tris, 100 mM NaCl, 250 mM imidazole, pH 7. For introduction of the disulfide bond in SodC in vitro, the eluted SodC proteins were further incubated with 100 M CuSO 4 at 4°C for 1 h.
To remove the N-terminal His tag, the SodC proteins were then incubated with 1 unit of HRV3C protease (Novagen) per 100 g of SodC at 4°C for 20 h. Samples were further purified by size exclusion chromatography using a gel filtration column (TSKgel G2000SW, TOSOH) equilibrated with 50 mM Tris, 100 mM NaCl, pH 7.0. After confirming successful removal of the His tag from SodC by SDS-PAGE, the purified tag-free SodC was then demetallated by precipitation with 20% trichloroacetic acid (TCA) followed by washes with cold acetone. The dried protein pellets were redissolved in an appropriate metalfree buffer that was treated with Chelex 100 resin (Bio-Rad). The concentration of SodC was spectroscopically determined from the absorbance at 225 and 215 nm (18). SOD Activity Assays-3.2 g of SodC proteins in 50 mM HEPES, 100 mM NaCl, pH 8.0, were assayed for SOD activity using the SOD assay kit WST (DOJINDO) in a 96-well plate, and the absorbance at 450 nm was measured using a plate reader (Epoch, BioTek). For the metallated forms, an equimolar amount of CuSO 4 /ZnSO 4 was added before the assay.
Cu 2ϩ and Co 2ϩ Ion Titration Experiments-E,E-SodC S-S/noCys was titrated with CuSO 4 or CoCl 2 in 100 mM NaOAc, 100 mM NaCl, pH 5.5, or 10 mM HEPES, 100 mM NaCl, pH 7.4, respectively (19,20). After the addition of CuSO 4 or CoCl 2 , sample solutions were centrifuged at 7,500 rpm for 1 min to remove any insoluble materials and then used for the measurement of visible absorption spectra with a spectrophotometer (UV-1800, Shimadzu). All buffer solutions were treated with Chelex 100 resin to remove metal ion contamination. Furthermore, to avoid contamination of metal ions in samples (Zn 2ϩ ions, in particular), we used a plastic disposable cuvette in place of a glass cell. Spectroscopic Analysis-Far-UV CD spectra were measured with a J-720WI spectropolarimeter (Jasco). Concentration of protein samples was set to 10 M in 20 mM sodium phosphate, 50 mM NaCl, pH 8.0, treated with Chelex 100 resin. For the analysis of metallated SodC proteins, an equimolar amount of CuSO 4 /ZnSO 4 was added to the samples before measurements.

Quantitation of Copper and Zinc Ions by Colorimetric Assay-
Fourier transform infrared spectroscopy (FTIR) spectra were measured by using an IRAffinity-1S spectrophotometer (Shimadzu) attached with an attenuated total reflection accessory (DuraSamplIR II, nine reflections). SodC proteins were demetallated by TCA precipitation and redissolved in a deuterated buffer containing 100 mM Tris and 100 mM NaCl at pD 8.4. The protein concentration was set to 1 mM. The second derivative of the absorption spectrum was obtained using the LabSolutions IR software (Shimadzu). Peaks were assigned based on the work of Byler and Susi (21) and of assignments on bovine SOD1 (22).
Size Exclusion Chromatography-50 l of 10 M SodC was loaded onto a gel filtration column (TSKgel G2000SW, TOSOH), and the absorbance change at 215 nm was monitored. To avoid contamination of divalent metal ions from the HPLC system, 50 M EDTA was included in the running buffer, 50 mM Tris and 100 mM NaCl, pH 7.0. For the analysis of metallated SodC, a 1.2-fold molar excess amount of CuSO 4 and/or ZnSO 4 was added to SodC proteins before loading onto the column. For the estimation of molecular weight, the column was calibrated with bovine serum albumin (66 kDa), chicken egg albumin (45 kDa), carbonic anhydrase (29 kDa), equine heart myoglobin (17.7 kDa), ␣-lactoalbumin (14.2 kDa), and insulin (5.7 kDa).
Western Blotting Analysis of SodC in E. coli-E. coli BW25113 and the sodC/dsbA/degP/prc-knock-out cells from the Keio collection (23) were provided by National BioResource Project (NBRP) (National Institute of Genetics, Mishima, Japan); E. coli were cultured in LB medium to stationary phase at 37°C, and the periplasmic fraction was obtained by a cold osmotic shock method (24) that was slightly modified. Cells were first resuspended in 50 mM Tris, 20% (w/v) sucrose, 2.5 mM EDTA, 2 mM CaCl 2 , pH 8.0, together with 100 mM iodoacetamide for protecting free thiols and then incubated on ice for 10 min. After centrifugation at 20,000 ϫ g for 10 min, the pellets were resuspended in 0.5 mM MgSO 4 with 100 mM iodoacetamide and incubated on ice for 10 min. The supernatant obtained by centrifugation at 20,000 ϫ g for 10 min was collected as the periplasmic fraction. The total amount of proteins in the periplasmic fraction was measured by using a Micro BCA Protein Assay kit (Thermo Scientific) using bovine serum albumin as a standard.
To examine the effects of reductants on endogenous SodC, dithiothreitol (DTT, 0 -2.5 mM) was added to the E. coli BW25113 cultures in the stationary phase. 0.15 g/liter chloramphenicol was also added to inhibit protein synthesis (25). The sodC-knock-out E. coli cells transformed with a pGEX-4T-1 plasmid in which the glutathione S-transferase gene was replaced by SodC with an N-terminal periplasmic signal sequence were also used for these experiments. After the addition of DTT and chloramphenicol, E. coli cells were cultured for 1 h at 37°C, and the periplasmic fraction was then isolated as described above.
Periplasmic proteins were analyzed by non-reducing SDS-PAGE using a 12.5% polyacrylamide gel. After electrophoresis, the gel was immersed in a 1% ␤-mercaptoethanol solution for 15 min to reduce the disulfide bond in SodC (26). Proteins in the gel were then blotted onto a PVDF membrane (0.2 m, GE Healthcare) and incubated with rabbit anti-SodC sera (1:100) that was prepared by immunization of a rabbit with purified recombinant SodC protein (Unitech). Goat anti-rabbit conjugated with HRP (1:3000, Pierce) was used as a secondary antibody, and the blots were developed with ImmunoStar LD (Wako).
Proteolytic Digestion of SodC-0.3 g/liter E,E-SodC S-S/noCys was incubated with 0.3 ϫ 10 Ϫ5 -0.3 ϫ 10 Ϫ1 g/liter Pronase (Calbiochem) at 37°C for 30 min in 50 mM Tris, 100 mM NaCl, pH 7.0. The digestion reaction was quenched by adding a protease inhibitor mixture (cOmplete, EDTA-free, Roche Applied Science) and a Laemmli sample buffer with ␤-mercaptoethanol. The samples were boiled and then loaded onto a 12.5% polyacrylamide gel and analyzed by SDS-PAGE. After electrophoresis, the gel was stained with Coomassie Brilliant Blue to visualize protein bands.

Formation of the Disulfide Bond Is Essential for Enzymatic
Activity of SodC-To test if formation of the intramolecular disulfide bond was essential for enzymatic activity of SodC, recombinant E. coli SodC with the disulfide bond (SodC S-S ) and a disulfide-null mutant protein in which Cys 74 and Cys 169 were replaced with Ala (SodC noCys ) were prepared. The disulfidereduced form of SodC (SodC SH ) was also prepared by incubation of SodC with DTT. As described under "Experimental Procedures," the two thiols of Cys 74 and Cys 169 in SodC were oxidized to form an intramolecular disulfide bond by the addition of CuSO 4 , which was then removed by TCA precipitation. SodC noCys proteins were also treated with TCA precipitation; treatment of SodC proteins with TCA precipitation assures the preparation of demetallated SodC (E,E-SodC S-S and E,E-SodC noCys ). As shown in Fig. 1A, the electrophoretic mobility of SodC S-S analyzed by SDS-PAGE was retarded in the presence of a reductant, DTT, whereas SodC SH and SodC noCys did not change their mobility upon the addition of DTT. These results confirmed successful introduction of the disulfide bond in SodC S-S .
SodC catalyzes the disproportionation of superoxide anion, which can be spectrometrically assayed based upon the formation of formazan by superoxide anion (see "Experimental Procedures"). As shown in Fig. 1B, both E,E-SodC S-S and E,E-SodC noCys showed no activity, consistent with a catalytic role of a bound copper ion. Indeed, significant activity became evident after the addition of a copper ion to SodC S-S ; however, SodC noCys remained inactive even in the presence of a copper ion. Given that the further addition of a zinc ion did not activate SodC noCys , formation of the intramolecular disulfide bond is essential for the enzymatic activity of SodC. A, 10 g of either SodC S-S , SodC SH , or SodC noCys with and without DTT was loaded on a 12.5% polyacrylamide gel for SDS-PAGE analysis. In the absence of DTT, SodC proteins were modified with iodoacetamide before electrophoresis. DTT reduced the disulfide bond in SodC S-S and thereby retarded the electrophoretic mobility of SodC. B, an SOD activity assay was performed by using 3.2 g of SodC proteins in 50 mM HEPES, 100 mM NaCl, pH 8.0. SOD activity was represented as the inhibition efficiency (%) of formazan formation by superoxide anion. Metallated forms were prepared by incubation of E,E-SodC with equimolar CuSO 4 /ZnSO 4 for 30 min before the assay.

SodC Binds a Copper Ion in the Absence of the Disulfide Bond but Remains
Inactive-Given that SodC noCys was inactive even in the presence of copper ions, it is possible that formation of the disulfide bond is essential for binding of the copper ion in SodC. To examine this possibility, the titration of SodC proteins with Cu 2ϩ ions was monitored by visible absorption spectroscopy.
When an equimolar amount of Cu 2ϩ ion was added to E,E-SodC S-S , an absorption peak was observed at 655 nm ( Fig. 2A), consistent with the binding of a copper ion at the copper-binding site (27). A new peak at 830 nm then emerged after the addition of a 2-fold molar excess of Cu 2ϩ ions to SodC S-S ( Fig.  2A), suggesting that a Cu 2ϩ ion was also bound in the zincbinding site (19). Further addition of Cu 2ϩ ions to SodC S-S increased the intensity of an absorption peak at 760 nm ( Fig.   2A), which corresponds to that of free Cu 2ϩ ions in the acetate buffer (data not shown). Based upon these results, SodC S-S preferentially binds a Cu 2ϩ ion at its copper-binding site; therefore, SodC S-S becomes enzymatically active in the presence of copper ions.
Unexpectedly, E,E-SodC noCys was also found to bind a Cu 2ϩ ion in the copper-binding site; a characteristic absorption peak at 650 nm appeared after the addition of an equimolar Cu 2ϩ ion to E,E-SodC noCys (Fig. 2B). It is, therefore, unlikely that the lack of enzymatic activity in SodC noCys (Fig. 1B) is due to an inability to bind a copper ion. Notably, however, further addition of Cu 2ϩ ions to SodC noCys did not produce the absorption peak around 830 nm corresponding to a Cu 2ϩ ion bound at the zincbinding site; instead, an absorption peak at 730 nm due to free copper ions increased in intensity (Fig. 2B). In addition, when . For clarity, the spectra of SodC proteins with 1 and 2 eq mol of CoCl 2 are shown as thick curves. E, 200 M E,E-SodC S-S (thin curve) or E,E-SodC noCys (thick curve) in 100 mM NaOAc, 100 mM NaCl, pH 5.5, was mixed with an equimolar amount of CoCl 2 followed by the addition of an equimolar amount of CuSO 4 . chelator, EDTA (500 M), the dissociation of the copper ion was found to be slightly more rapid from SodC noCys than that from SodC S-S , although the dissociation completed within 5 min in both SodC noCys and SodC S-S (data not shown). This implies roles of the disulfide bond in increasing the affinity, albeit slight, of the bound copper ion in SodC. Taken together, formation of a disulfide bond is not necessary for binding of a copper ion in SodC but appears to be critical to attain a native conformation for enzymatic activity.
A Disulfide Bond Is Required for a Native Conformation around the Zinc-binding Site-It should be noted that, in the absence of the disulfide bond, the zinc-binding site in SodC was not able to bind an additional copper ion; therefore, the disulfide bond may be required for establishing the native conformation of the zinc-binding site. As described previously (19,28), a Co 2ϩ ion has been used as a substitute for a spectroscopically silent Zn 2ϩ ion and was titrated to E,E-SodC S-S and E,E-SodC noCys . The addition of an equimolar amount of Co 2ϩ ions to E,E-SodC S-S resulted in a characteristic absorption peak centered at 563 nm (Fig. 2C), indicating binding of a Co 2ϩ ion at the zinc-binding site (19). Upon further addition of Co 2ϩ ions, an additional absorption peak at 600 nm emerged, consistent with binding of a Co 2ϩ ion at the copper-binding site (Fig. 2C) (29,30). In SodC S-S , therefore, the zinc-binding site is considered to be available for binding of a Zn 2ϩ ion.
In contrast to SodC S-S , however, SodC noCys was unable to bind a Co 2ϩ ion; the addition of up to a 10-fold molar excess of Co 2ϩ ions to E,E-SodC noCys produced an absorption peak at 520 nm (Fig. 2D), which corresponds to that of free Co 2ϩ ions in the HEPES buffer (data not shown). We also tested the binding of Co 2ϩ ion to SodC proteins in the presence of Cu 2ϩ ion. In SodC, Cu 2ϩ binding to the copper site (Fig. 2, A and B) and Co 2ϩ binding to the zinc site (Fig. 2C) produced distinct absorption spectra, and the sum of those spectra was observed when both Cu 2ϩ and Co 2ϩ ions were added to E,E-SodC S-S (Fig.  2E). In contrast, the absorption peaks characteristic to the binding of Co 2ϩ ions at the zinc-binding site were not observed when both Cu 2ϩ and Co 2ϩ ions were added to E,E-SodC noCys (Fig. 2E); instead, only the absorption peak at 650 nm due to the binding of Cu 2ϩ ion to the copper-binding site was confirmed. Accordingly, SodC noCys was considered to have limited ability to bind metal ions at the zinc-binding site even in its copperbound form.
We have also confirmed no binding of a Zn 2ϩ ion in SodC noCys by direct quantitation of Zn 2ϩ ions. E,E-SodC S-S and E,E-SodC noCys were first incubated with 1.2-fold molar excess of Zn 2ϩ ions, and unbound Zn 2ϩ ions were then removed by ultrafiltration. Approximately an equimolar amount of Zn 2ϩ ions (133 Ϯ 16%) remained in SodC S-S , whereas SodC noCys was unable to bind Zn 2ϩ ions (28 Ϯ 10%). These results thus show that the zinc-binding site was not properly formed in SodC without the disulfide bond. Therefore, even though the disulfide bond is not required to bind a copper ion in SodC, the conformation of SodC in the absence of the disulfide bond is not optimized for catalyzing the disproportionation of a superoxide anion.
SodC Folds into a ␤-Sheet Structure upon Formation of the Disulfide Bond-To investigate the effects of the disulfide bond on the folding of SodC, the content of secondary structure in SodC was examined by circular dichroism (CD) spectroscopy. The CD spectrum of E,E-SodC noCys had a large negative peak at 199 nm (Fig. 3A), and its spectral shape is characteristic of a random coil structure (31). E,E-SodC SH , in which both Cys residues (Cys 74 and Cys 169 ) are in the thiol state, also exhibited a similar CD spectrum to that of E,E-SodC noCys with a large negative peak at 199 nm (Fig. 3A), confirming little artificial effects of the Cys-to-Ala mutations on the conformation of SodC lacking the disulfide bond. Given that disulfide-reduced human SOD1 has been shown to fold into a structure rich in ␤-sheets (5), almost no secondary structure in SodC without the disulfide bond was unexpected.
When the disulfide bond was introduced into SodC (E,E-SodC S-S ), the CD spectrum exhibited remarkable changes (Fig.  3A); the intensity of the negative peak at 199 nm largely decreased, and the spectral shape also changed. We thus suppose that these spectral changes highlight important roles of the disulfide bond in the secondary structure formation of SodC.
As shown in Fig. 3B, the addition of either Cu 2ϩ ion, Zn 2ϩ ion, or both to E,E-SodC noCys also slightly decreased the intensity at 199 nm of its CD spectrum, but the spectral shape was not markedly affected. It should also be noted that the addition of metal ions to E,E-SodC S-S further changed the CD spectrum (Fig. 3C); in particular, the spectral shape of SodC S-S was significantly changed in the presence of both copper and zinc ions. Binding of metal ions in SodC S-S is hence supposed to have effects on the structural maturation, but how those metal ions affect the conformation of SodC S-S is difficult to speculate from the spectral changes of SodC upon binding of metal ions.
On the other hand, a relatively large difference in the CD spectral shape between SodC S-S and SodC SH (or SodC noCys ) allows us to elucidate the conformational changes of SodC caused by the disulfide formation, and indeed, the difference spectrum between E,E-SodC S-S and E,E-SodC SH (or E,E-SodC noCys ) is consistent with the formation of a ␤-sheet structure with a characteristic negative peak at 218 nm and a positive peak at 200 nm (Fig. 3D) (31). These results thus support an important role for the disulfide bond in realizing ␤-sheet structures in SodC.
Formation of a ␤-sheet structure in SodC by introduction of the disulfide bond was also confirmed by FTIR. Fig. 4A shows the second derivative FTIR spectrum of E,E-SodC noCys , and peaks were observed at 1,641 and 1,665 cm Ϫ1 , which were assigned to a random coil and a ␤-turn structure, respectively (21). These results are consistent with the random coil nature of E,E-SodC noCys suggested by CD spectroscopy (Fig. 3A). A peak was also observed at 1,617 cm Ϫ1 , but its assignment remains obscure. Notably, the spectrum of E,E-SodC noCys remained unchanged by the addition of both copper and zinc ions but was significantly different from that of SodC with the disulfide bond; the second derivative FTIR spectrum of E,E-SodC S-S detected peaks at 1,634 and 1,677 cm Ϫ1 due to a ␤-sheet structure (Fig. 4A). A peak was also observed at 1,658 cm Ϫ1 (Fig. 4A), which could be assigned to either a random coil or an ␣-helical

Roles of Disulfide Formation in SodC Folding
structure (22). Further addition of both copper and zinc ions to E,E-SodC S-S did not significantly change the spectrum (Fig.  4A), again consistent with a primary role of the disulfide bond in the formation of secondary structures. Moreover, a difference FTIR spectrum between E,E-SodC S-S and E,E-SodC noCys is also shown in Fig. 4B, where the increased intensity of absorption peaks at 1,631 and 1,680 cm Ϫ1 corresponding to ␤-sheet structures was observed upon formation of the disulfide bond.
Based upon these results, a ␤-sheet structure was formed in SodC after introduction of the disulfide bond; therefore, SodC without the disulfide bond was not able to become enzymatically active even in the presence of copper ions.
Formation of a Disulfide Bond Makes the Conformation of SodC More Compact-The random coil nature of SodC without the disulfide bond was also inferred by size exclusion chromatography. As shown in Fig. 5A, E,E-SodC noCys eluted at 7.3

FIGURE 3. CD spectral analysis of secondary structures in SodC.
A-C, 10 M SodC in 20 mM sodium phosphate, 50 mM NaCl, pH 8.0, was used for measurement of a CD spectrum. In panel A, the CD spectra of E,E-SodC noCys , E,E-SodC S-S , and E,E-SodC SH are shown as dotted, thin solid, and thick solid curves, respectively. In panel B, an equimolar amount of either CuSO 4 (thin solid curve), ZnSO 4 (broken curve), or both (thick solid curve) was added to E,E-SodC noCys , and the CD spectra were measured. For comparison, the CD spectrum of E,E-SodC noCys is shown again as a dotted curve. In panel C, CD spectra were measured after an equimolar amount of either CuSO 4 (broken curve), ZnSO 4 (dotted curve), or both (thick solid curve) was added to E,E-SodC S-S . For comparison, the CD spectrum of E,E-SodC S-S is shown again as a thin solid curve. D, a difference CD spectrum between E,E-SodC noCys (dotted curve in A) and E,E-SodC S-S (thin solid curve in A) is shown as a thin curve, whereas a thick curve represents a difference CD spectrum between E,E-SodC SH (thick solid curve in A) and E,E-SodC S-S (thin solid curve in A). ml, and the elution volume was found to significantly increase to 8.2 ml in the presence of the disulfide bond (E,E-SodC S-S ). Given that the elution volume of 7.3 and 8.2 ml corresponds to 45 and 24 kDa, respectively, formation of the disulfide bond may alter the quaternary structure of SodC (16 kDa). Nonetheless, the increase of elution volume also supports a decrease in the hydrodynamic volume of SodC upon formation of the disulfide bond, which is consistent with our idea that SodC S-S assumes a more compact conformation than SodC without the disulfide bond.
In contrast, the addition of either Cu 2ϩ or Zn 2ϩ ion to E,E-SodC noCys did not change the elution profile (Fig. 5B), corroborating our findings that the metal binding was not sufficient for the formation of ␤-sheet structures in the disulfide-null SodC. It is, however, notable that the addition of both Cu 2ϩ and Zn 2ϩ ions to E,E-SodC S-S further increased its elution volume to 8.5 ml, corresponding to 19 kDa, which is close to the size of monomeric SodC (Fig. 5A). The addition of Zn 2ϩ but not Cu 2ϩ ion shifted the elution volume of E,E-SodC S-S to that of Cu,Zn-SodC S-S (data not shown). Accordingly, binding of metal ions (zinc ion, in particular) can affect the conformation of SodC S-S , but we also suppose that disulfide formation plays more significant roles in folding SodC into a compact conformation.
Intracellular Stability of SodC Is Controlled by Disulfide Formation-It is well known that proteins with random coil structures are highly susceptible to proteolytic degradation in cells. We, therefore, suspected that intracellular SodC is quite unstable before it forms the disulfide bond. Indeed, E,E-SodC noCys in vitro was almost completely digested with a nonspecific protease mixture, Pronase, at a Pronase/SodC ratio of 10 Ϫ3 , whereas E,E-SodC S-S remained resistant to proteolytic degradation (Fig. 6A). Also, binding of copper and zinc ions to SodC S-S (Cu,Zn-SodC S-S ) further increased the resistance to Pronase, suggesting roles of metal ions in the structural stabilization of SodC. Even in the presence of metal ions, however, SodC noCys was found to be more susceptible to proteolytic degradation than SodC S-S (Fig. 6A).
In E. coli, SodC is equipped with an N-terminal signal peptide for its transport to the periplasmic space (32), where the signal peptide is cleaved from SodC. To examine the effects of the disulfide bond on the intracellular stability of SodC, sodCknock-out E. coli cells were transformed with plasmids encoding SodC/SodC noCys with an N-terminal signal peptide under the control of a tac promoter. As shown in Fig. 6B, SodC with a disulfide bond, in which the N-terminal signal peptide was cleaved, was detected in a periplasmic fraction extracted from E. coli cells grown to the stationary phase. In contrast, almost no SodC noCys was confirmed in the periplasmic fraction (Fig.  6B) or cytoplasmic fractions (data not shown), supporting the idea that formation of the disulfide bond increases the intracellular stability of SodC proteins. Intracellular protein levels are, however, also affected by factors other than the susceptibility to proteolysis; therefore, we attempted to directly reduce the disulfide bond in SodC under conditions in which protein synthesis was inhibited.
The sodC-knock-out E. coli cells, in which SodC with an N-terminal signal peptide was overexpressed, were first cultured to the stationary phase. DTT, an efficient reductant permeable to the outer membrane, was then added to the culture together with chloramphenicol to halt new protein synthesis. Periplasmic extracts were prepared by cold osmotic shock in the presence of a thiol-specific modifier, iodoacetamide, and the thiol-disulfide status of SodC was investigated by Western blotting. As shown in Fig. 6C, all of overexpressed SodC proteins were found to possess the disulfide bond in the absence of DTT, and the disulfide-reduced form of SodC appeared only after the addition of Ͼ1.0 mM DTT. These results well reproduce a previous study (11) and show that the disulfide bond in existing SodC proteins is reduced by DTT. Furthermore, we have noted that the total amounts of SodC were significantly decreased by the addition of DTT (Fig. 6C). Based upon SDS-PAGE analysis of the periplasmic fraction followed by Coomassie Brilliant Blue staining, we confirmed that total protein levels and composition of proteins in the periplasmic fraction were not drastically affected by the addition of DTT (data not shown); therefore, the disappearance of SodC by the addition of increasing amounts of DTT is not considered to be due to DTTinduced disruption of the cell wall and leakage of periplasmic proteins. In addition, endogenous SodC proteins in E. coli BW25113 cells were found to possess the disulfide bond but disappeared when treated with increasing amounts of DTT (Fig. 7A). Because the expression level of endogenous SodC was not as high compared with that of SodC expressed from plasmid-borne cDNA (Fig. 6C), the remaining disulfide-reduced form of endogenous SodC would be below the detection limit of Western blot (Fig. 7A). Given that protein synthesis is largely suppressed in the stationary phase and was further inhibited by chloramphenicol, these data support the idea that reduction of the disulfide bond facilitates the degradation of existing SodC proteins.
A thiol-disulfide oxidoreductase, DsbA, has been proposed to introduce a disulfide bond in SodC at the periplasmic space of E. coli (11); therefore, we assumed that knock-out of DsbA retards disulfide formation and thereby promotes the degradation of SodC. Unexpectedly, in dsbA-knock-out E. coli cells, all endogenous SodC were found to form the disulfide bond (Fig.  7A), suggesting that a DsbA-independent pathway(s) is available for disulfide formation in SodC. Despite this, the band intensity of SodC in dsbA-knock-out E. coli was found to be significantly lower than that of the parent strain, BW25113 (Fig.  7A). Also, SodC expression was found to be more sensitive to DTT when DsbA was absent (Fig. 7A). Quite notably, throughout the experiments, no bands corresponding to the disulfidereduced form of endogenous SodC were observed, implying the immediate degradation of SodC upon reduction of the disulfide bond.
We suspected that disulfide-reduced SodC could accumulate in the absence of an endogenous protease responsible for degradation of SodC, but we failed to detect disulfide-reduced SodC when a periplasmic protease, DegP or Prc, was deleted from E. coli cells (Fig. 7B). Although further studies will be required to test several other periplasmic proteases for their involvement in the degradation of disulfide-reduced SodC, it is also possible that not one but several proteases are involved. Despite this, disulfide-reduced SodC was considered to assume a random coil-like conformation that was highly susceptible to proteolytic degradation. Therefore, based upon the results here, we propose a mechanism whereby the intracellular stability of SodC is regulated by its thiol-disulfide status.

DISCUSSION
SOD1 is a ubiquitous antioxidant enzyme that has been identified in all aerobic organisms from bacteria to humans, where dysfunction of SOD1 is often detrimental. Indeed, deletion of sod1 gene in yeast (33), flies (34), and mice (35) results in decreased lifespan, and furthermore, dominant mutations in human SOD1 have been identified to cause a familial form of amyotrophic lateral sclerosis (36). Also, in infectious bacteria, periplasmic SodC proteins combat against the respiratory burst of phagocytes, assuring successful survival in the host (16). Therefore, the enzymatic activation of SOD1/SodC controls various physiological processes, and we have found that formation of the disulfide bond plays a primary role in the folding and activation of SodC.
Folding of SodC Is Regulated Primarily by Formation of the Disulfide Bond-As shown in this study, prokaryotic E. coli SodC has little secondary structure in the absence of the disulfide bond. This is in sharp contrast to previous findings on eukaryotic SOD1; for example, a ␤-barrel-like folding pattern of human SOD1 is maintained even in the disulfide-reduced state  (5), whereas its quaternary structure has been shown to be affected by the thiol-disulfide status (6,37). Notably, furthermore, formation of the disulfide bond is absolutely necessary for SodC to bind metal ions at the zinc-binding site (Fig. 2); however, human SOD1 has been shown to tightly bind a cobalt ion at the zinc-binding site even without the disulfide bond (20). Compared with human SOD1, E. coli SodC possesses a seven-residue insertion in the loop region ("S-S subloop") containing Cys 74 , which is tethered with Cys 169 through the disulfide bond (Fig. 8A) (38). In the absence of the disulfide bond, therefore, generation of the zinc-binding site in E. coli SodC may be hampered by significant fluctuations of the relatively long S-S subloop. Formation of a disulfide bond is generally able to constrain structural fluctuations of proteins (39); therefore, foldability of SOD1 proteins presumably becomes increasingly dependent upon disulfide formation by extension of the S-S subloop structure.
Although further investigations are necessary to determine why prokaryotic SodC but not eukaryotic SOD1 is almost unfolded in the absence of the disulfide bond, we suspect that the disulfide-dependent folding of SodC is an important aspect for describing the intracellular localization of SodC (Fig. 8B). Eukaryotic SOD1 is a major cytoplasmic protein and is activated via CCS-dependent/independent pathways in the highly reducing environment of the cytoplasm (7,10). A folding process of eukaryotic SOD1 should, therefore, be tolerant of the reducing environment and would not be significantly depen-dent upon its thiol-disulfide status. Unlike eukaryotic SOD1, however, prokaryotic SodC needs to be transported to the oxidizing periplasmic space after translation in the strongly reducing cytoplasm. The transportation of SodC to the periplasm is guided by its N-terminal signal sequence (32); however, a protein with a strong tendency to fold may experience conformational difficulties for going through a transporter. This is reminiscent of the mitochondrial import of yeast SOD1, in which only the disulfide-reduced apo form of SOD1 is eligible for being transported from cytoplasm to the intermembrane space of mitochondria (40). In the mitochondrial intermembrane space, yeast SOD1 acquires the disulfide bond as well as copper and zinc ions, which appears to prevent SOD1 from returning to the cytoplasm. Accordingly, structural folding of SodC triggered by formation of the disulfide bond would significantly contribute to its partitioning between the periplasm and cytoplasm; in that case, the random coil nature of disulfide-reduced SodC is considered to be advantageous for transportation across cellular compartments (Fig. 8B).
Activation Mechanism of SodC in Prokaryotic Cells-After being transported to the periplasm, SodC needs to acquire copper and zinc ions and also the disulfide bond for enzymatic activation. Almost no free copper/zinc ion has been shown to exist in the cytoplasm, and the periplasm is considered to contain significantly higher copper/zinc concentrations than the cytoplasm (41). Although it is still unclear how much freely available copper/zinc ions exist in the periplasmic space, SodC is overlaid on that of human SOD1 (PDB ID: 1HL5, shown in gray). A seven-residue insertion in the S-S subloop of E. coli SodC is colored red. A copper ion (cyan), a zinc ion (orange), and the disulfide bond (yellow) are also shown. B, our proposed mechanism of SodC activation regulated by disulfide formation. SodC with an N-terminal signal peptide is translated in the cytoplasm and transported to periplasm. In the periplasm, disulfide-reduced SodC remains unfolded and highly susceptible to proteolytic degradation; therefore, the disulfide formation is required to occur as the first step of SodC maturation (a disulfide-first mechanism). Then SodC with the disulfide bond binds copper/zinc ions to become enzymatically active. Alternatively, an as-yet unidentified copper chaperone may activate SodC by simultaneously supplying a disulfide bond as well as a catalytic copper ion.
will acquire its metal cofactors after being transported to the periplasm. In eukaryotic SOD1, the first step of the activation appears to be the binding of a zinc ion in the disulfide-reduced apo state (6). Notably, however, our results show that disulfidereduced SodC is unable to bind a cobalt/zinc ion (Fig. 2). Also, even in the presence of metal ions, the disulfide-null SodC is considered to remain in the extended conformation and exhibited high susceptibility to proteolysis (Fig. 6A). Taken together, formation of the disulfide bond is considered to first occur rapidly in SodC just after translocation to the periplasm, which is then followed by the binding of metal ions; otherwise, SodC would become degraded without being enzymatically activated (the "disulfide-first" mechanism in Fig. 8B).
In many eukaryotic cells the disulfide bond is introduced into SOD1 by the copper chaperone, CCS (6); however, prokaryotes including E. coli do not have CCS. Instead, E. coli is equipped with the disulfide bond formation (Dsb) system in the periplasm that catalyzes the correct introduction of disulfide bonds into periplasmic proteins (42). The disulfide bond in exogenously overexpressed SodC in E. coli has been shown to become more susceptible to reduction with a reducing agent upon deletion of DsbA, a periplasmic protein that serves as a disulfide bond donor to proteins (11). Even in dsbA-knock-out E. coli cells, however, the disulfide bond was introduced efficiently into all endogenous SodC proteins (Fig. 7A), suggesting a DsbA-independent mechanism for disulfide formation in SodC. Furthermore, no accumulation of the disulfide-reduced endogenous SodC was observed in E. coli cells where DsbA, DsbB, DsbC, DsbD, or DsbG was deleted. 3 Nonetheless, there is a notable caveat about no apparent roles of the Dsb system on the disulfide formation in SodC; namely, it may be difficult to detect disulfide-reduced SodC simply because of its high susceptibility to proteolytic degradation in cells. It thus remains an open question whether formation of the disulfide bond in SodC is performed solely by the Dsb system, and further investigations will be required to identify a DsbA-independent pathway(s) in the SodC activation mechanism. Also, a cupric ion itself is able to function as an oxidant to form a disulfide bond. Because the disulfide-null SodC can bind a copper ion (Fig. 2B), the interaction between SodC and a cupric ion might play roles in introducing the disulfide bond into SodC under oxidizing environment of the periplasm.
As proposed in the CCS-dependent activation of eukaryotic SOD1, it is also likely that disulfide formation in SodC occurs simultaneously with the copper supply step by an as-yet unidentified copper chaperone protein in E. coli. Interestingly, in S. enterica, a periplasmic cupro-protein, CueP, has been recently proposed to supply a copper ion to SodC (43), whereas it remains to be tested if CueP is able to form the disulfide bond in SodC. E. coli lacks CueP, but its functional counterpart, CusF, has been shown to bind a cuprous ion in the periplasm and have a copper chaperoning activity to the membrane protein CusB (44). Although all SodC proteins in cusF-knock-out E. coli cells were again found to possess the disulfide bond, 3 the ability of CusF to introduce a disulfide bond as well as a copper ion into SodC is currently under investigation in our group.
In conclusion, we propose here that disulfide formation is a primary event for the folding of prokaryotic SodC. This is consistent with the fact that the bacterial periplasm, where SodC is localized, generally provides an oxidizing environment for stabilizing a disulfide bond. When the disulfide bond is lacking, however, SodC is rapidly degraded and loses its enzymatic activity. Although it remains obscure how much the periplasm can change its redox potential during the life cycle of bacteria, the disulfide bond in SodC could function as a sensor to modulate the enzymatic activity in response to the redox environment of the periplasm. Also, SodC activity is an important countermeasure of pathogenic bacteria against the respiratory burst of macrophages; therefore, reduction of the disulfide bond in SodC would be a promising strategy to mitigate the virulence of such bacteria.