Bicarbonate-dependent Peroxidase Activity of Human Cu,Zn-Superoxide Dismutase Induces Covalent Aggregation of Protein

This study addresses the mechanism of covalent aggregation of human Cu,Zn-superoxide dismutase (hSOD1WT) induced by bicarbonate (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document})-mediated peroxidase activity. Higher molecular weight species (apparent dimers and trimers) of hSOD1WT were formed from incubation mixtures containing hSOD1WT, H2O2, and \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}. \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}-dependent peroxidase activity and covalent aggregation of hSOD1WT were mimicked by UV photolysis of hSOD1-WT in the presence of a [Co(NH3)5CO3]+ complex that generates the carbonate radical anion (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{CO}_{3}^{{\bar{{\cdot}}}}\) \end{document}). Human SOD1WT has but one aromatic residue, a tryptophan residue (Trp-32) on the surface of the protein. Substitution of Trp-32 with phenylalanine produced a mutant (hSOD1W32F) that exhibits \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}-dependent peroxidase activity similar to wild-type enzyme. However, unlike hSOD1WT, incubations containing hSOD1W32F,H2O2, and \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}did not result in covalent aggregation of SOD1. These findings indicate that Trp-32 is crucial for \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{CO}_{3}^{{\bar{{\cdot}}}}\) \end{document}-induced covalent aggregation of hSOD1WT. Spin-trapping results revealed the formation of the Trp-32 radical from hSOD1WT, but not from hSOD1W32F. Spin traps also inhibited the covalent aggregation of hSOD1WT. Fluorescence experiments revealed that Trp-32 was further oxidized by \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{CO}_{3}^{{\bar{{\cdot}}}}\) \end{document}, forming kynurenine-type products in the presence of oxygen. Molecular oxygen was needed for \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}/H2O2-dependent aggregation of hSOD1WT, implicating a role for a Trp-32-dependent peroxidative reaction in the covalent aggregation of hSOD1WT. Taken together, these results indicate that Trp-32 oxidation is crucial for covalent aggregation of hSOD1. Implications of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{HCO}_{3}^{-}\) \end{document}-dependent SOD1 peroxidase activity in amyotrophic lateral sclerosis disease are discussed.

This study addresses the mechanism of covalent aggregation of human Cu,Zn-superoxide dismutase (hSOD1 WT ) induced by bicarbonate (HCO 3 ؊ )-mediated peroxidase activity. Higher molecular weight species (apparent dimers and trimers) of hSOD1 WT were formed from incubation mixtures containing hSOD1 WT , H 2 O 2 , and HCO 3 ؊ . HCO 3 ؊ -dependent peroxidase activity and covalent aggregation of hSOD1 WT were mimicked by UV photolysis of hSOD1-WT in the presence of a [Co(NH 3 ) 5 CO 3 ] ؉ complex that generates the carbonate radical anion (CO 3 . ). Human SOD1 WT has but one aromatic residue, a tryptophan residue (Trp-32) on the surface of the protein. Substitution of Trp-32 with phenylalanine produced a mutant (hSOD1 W32F ) that exhibits HCO 3 ؊ -dependent peroxidase activity similar to wildtype enzyme. However, unlike hSOD1 WT , incubations containing hSOD1 W32F , H 2 O 2 , and HCO 3 ؊ did not result in covalent aggregation of SOD1. These findings indicate that Trp-32 is crucial for CO 3 . -induced covalent aggregation of hSOD1 WT . Spin-trapping results revealed the formation of the Trp-32 radical from hSOD1 WT , but not from hSOD1 W32F . Spin traps also inhibited the covalent aggregation of hSOD1 WT . Fluorescence experiments revealed that Trp-32 was further oxidized by CO 3 . , forming kynurenine-type products in the presence of oxygen. Molecular oxygen was needed for HCO 3 ؊ /H 2 O 2 -dependent aggregation of hSOD1 WT , implicating a role for a Trp-32-dependent peroxidative reaction in the covalent aggregation of hSOD1 WT . Taken together, these results indicate that Trp-32 oxidation is crucial for covalent aggregation of hSOD1. Implications of HCO 3 ؊ -dependent SOD1 peroxidase activity in amyotrophic lateral sclerosis disease are discussed.
In a pair of publications, Hodgson and Fridovich demonstrated that bovine Cu,Zn-superoxide dismutase (Cu,Zn-SOD or SOD1) 1 exhibits a nonspecific peroxidase activity (1,2). They provided evidence for formation of a potent oxidant in the presence of H 2 O 2 that was ascribed to a copper-bound hydroxyl radical, at the active site of bovine SOD1. Although the rate constant for the reaction between SOD1 and H 2 O 2 is very low (3), it was shown that the bovine SOD1 peroxidase activity could oxidize a variety of small molecular weight anionic ligands (azide, nitrite, formate, etc.) that are accessible to the active site (1,2,4,5).
However, a perplexing aspect of this peroxidase activity was that even larger molecules (e.g. 2,2Ј-azino-bis-[3-ethylbenzothiazoline]-6-sulfonic acid, urate) that are not accessible to the active site of SOD1 were also oxidized (6). It was later discovered that the bicarbonate anion (HCO 3 Ϫ ) that is present in the buffer was responsible for Cu,Zn-SOD peroxidase-dependent oxidations of substrates in the bulk solution (7,8). It was proposed that the copper-bound oxidant (Cu 2ϩ -OH or Cu III O) could oxidize the HCO 3 Ϫ anion (a physiologically relevant molecule) to the carbonate anion radical (CO 3 . ), a potent oxidant that diffuses out of the active site and causes substrate oxidation (7,8). The fact that CO 3 . , and not hydroxyl radical, is the primary oxidant produced by SOD1-mediated peroxidase activity is physiologically significant, because CO 3 . has a much longer half-life (than hydroxyl radical) and can, therefore, diffuse away from the active site and oxidatively modify critical cellular targets. Liochev et al. (7) suggested that the nitrone spin trap, 5,5Јdimethyl-1-pyrroline-N-oxide (DMPO), is oxidized and hydrolyzed by CO 3 . . Zhang et al. (6) provided the experimental proof for this hypothesis. Using electron spin resonance (ESR) spin trapping in the presence of oxygen-17-labeled water, the investigators showed that the oxygen atom in the DMPO-OH adduct is derived totally from water. Zhang et al. (6) also reported that HCO 3 Ϫ -dependent peroxidase activity can be measured using a variety of methods including fluorescence and optical spectroscopy. More recently, Hink et al. (9) demonstrated that an extracellular SOD (SOD3 or ecSOD) enhanced the hydroxylation of a cyclic nitrone spin trap in the presence of H 2 O 2 and HCO 3 Ϫ . In this work, we report that human Cu,Zn-superoxide dismutase (hSOD1 WT ) also exhibits a HCO 3 Ϫ -dependent peroxi-* This work was supported by National Institutes of Health Grants RR01008, NS40494, and NS40819 (to J. P. C.) and the ALS Association. 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.
ʈ  3 . induces oxidation of a tryptophan residue located on the surface of hSOD1 WT that leads to intersubunit covalent bond formation and subsequent aggregation of the protein. Implications of HCO 3 Ϫ -dependent hSOD1 WT peroxidase activity in the covalent aggregation of SOD1 associated with amyotrophic lateral sclerosis (ALS) are discussed.

EXPERIMENTAL PROCEDURES
Bovine SOD1 was obtained from Roche Diagnostics. Pentammine carbonato complex of Co(III) was synthesized according to the published procedure (10). Briefly, 30 g of Co(NO 3 ) 2 ⅐6H 2 O in 15 ml of water was added to 45 g of ammonium carbonate dissolved in 45 ml of water, followed by the addition of 75 ml of concentrated ammonia. Air was bubbled through the solution for 24 h. The resulting solution was cooled in an ice bath, and the solid product was recrystallized by dissolving it in 55 ml of water at 90°C and then slowly cooling the solution in an ice bath. Pure crystals were isolated and used in the experiments. Chemicals including glycine, SDS, ammonium persulfate, Coomassie G250 stain, and 40% acrylamide/bis were obtained from Bio-Rad. Sodium bicarbonate, hydrogen peroxide, tryptophan, tyrosine, histidine, phenylalanine, and trizma base were purchased from Sigma. DMPO was also obtained from Sigma, and the colored impurity was removed by treatment with activated charcoal (11). POBN, ␣-phenyl-tert-butyl-N-nitrone, and DBNBS were obtained from Sigma and used as received. EMPO was synthesized according to a previous report (12). Azulenyl nitrone was a gift from Dr. David Baker (Florida International University, Miami, FL). DEPMPO was obtained from Dr. Paul Tordo (Universite de Provence, Marseille, France).
Expression and Purification of SOD1-The wild-type human and W32F mutant Cu/Zn-SOD (Cu 2ϩ , 97%; Zn 2ϩ , 108%) were expressed and purified as described previously (13). The SOD1 mutations were created by a two-cycle, polymerase chain reaction-based mutagenesis protocol described by Zhao et al. (14) using mutagenic primers to introduce the desired mutants and subsequently cloned into a pET-3d expression system (Novagen, Madison, WI). Detailed procedures have been described elsewhere (13). SOD1 activity was determined by the cytochrome c method (15). The metallation of the purified SOD1 W32F mutant was determined both by colorimetric PAR assay (13) and ESR spectroscopy. 2 Aggregation of SOD1-In general, SOD1 (1 mg/ml) was incubated with H 2 O 2 (1 mM) in a phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM), sodium bicarbonate (25 mM), and other relevant chemical reagents for 45 min. The reaction mixture was then mixed with an equal volume of Laemmli sample buffer containing 5% ␤-mercaptoethanol and boiled for 6 min. SOD1 (7.5 g) was loaded onto a 10% SDS-PAGE, and electrophoresis was performed at 4°C. The protein bands on the SDS-PAGE were visualized by a Coomassie G250 stain and documented using a Multimage Light Cabinet (Alpha Innotech Corp.). The pattern of SOD aggregation was not altered by increasing the amount of sample buffer, boiling time, or ␤-mercaptoethanol.
Bovine or human SOD1 (1 mg/ml) was mixed with the pentammine carbonato complex of Co(III) (4 mM) in a phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM) and other reagents as specified under "Results." The mixture was then irradiated with UV light using an EiMac VIX 300 UV 300 X xenon arc source. The aggregation of SOD1 induced by UV photolysis of the cobalt complex was also assessed by 10% SDS-PAGE.
ESR Spin-trapping Analysis-A typical reaction mixture (ϳ100 l) for ESR experiments consisted of SOD1 (1 mg/ml), spin traps (25 mM), H 2 O 2 (1 mM), and bicarbonate (25 mM) in a phosphate buffer (100 mM) containing DTPA (0.1 mM). ESR spectra were recorded at room temperature on a Varian E-109 spectrometer operating at 9.5 GHz with a 100-kHz field modulation and equipped with a TE102 cavity.  Spectrometer conditions were as follows: modulation amplitude, 1 G value time constant, 0.064 s; scan time, 2 min; and microwave power, 10 milliwatts. Spectral simulations were performed using the WINSIM program (NIEHS, National Institutes of Health, Research Triangle Park, NC).
Fluorescence Measurements-Fluorescence experiments were performed on a Shimadzu RF-5301 PC spectrofluorometer (Shimadzu Scientific Instruments Inc.). Spectra were obtained at the indicated excitation and emission wavelengths using 3-5-mm and 10 -20-mm slit widths, respectively. 3 .

Radical Anion and Not Hydroxyl Radical-Previously,
we reported that HCO 3 Ϫ dramatically enhances hydroxylation and oxidation of the spin trap DMPO to the DMPO-hydroxyl adduct, DMPO-OH (6). Using oxygen-17-labeled water, we showed that the oxygen atom in the DMPO-OH adduct formed in the SOD1 WT /H 2 O 2 /HCO 3 Ϫ /DMPO system is derived from water and not from H 2 O 2 (6). We have now extended this ESR analysis to include other cyclic nitrone traps (EMPO and DEPMPO) and open-chain nitrones (e.g. POBN).
The addition of H 2 O 2 (0.1-1.0 mM) to an incubation mixture containing SOD1 (bovine or recombinant human Cu,Zn-SOD; 31.7 M), spin traps (25 mM of DMPO, EMPO, DEPMPO, or POBN), HCO 3 Ϫ (25 mM), and DTPA (0.1 mM) in a phosphate buffer (100 mM, pH 7.4) yielded the respective ESR spectrum of the corresponding hydroxylated adduct (Fig. 1A). The computer-simulated ESR spectra (shown with dotted lines in Fig. 1A) were obtained using the ESR parameters (Table I). The addition of Me 2 SO, a frequently used hydroxyl radical scavenger, had no effect on the ESR signal intensity of the hydroxylated adducts (Fig. 1A). In contrast, when Me 2 SO was added to a mixture containing Fe 2ϩ and hydrogen peroxide (the Fenton system), the ESR spectrum of the hydroxylated adduct was replaced by a methyl radical adduct (Fig. 1B). From these results, it can be concluded categorically that free hydroxyl radicals are not generated in the SOD1/H 2 O 2 /HCO 3 Ϫ system and that ESR can be used to monitor the SOD1/H 2 O 2 /HCO 3 Ϫdependent peroxidase activity. These results indicate that Me 2 SO can be used to differentiate between CO 3 . and ⅐ OH formation. Other scavengers such as ethanol, azide, and formate react with both the hydroxyl radical and the carbonate radical anion and therefore cannot be used to distinguish formation of these species by ESR (Table II). In contrast to the cyclic nitrone traps, which form persistent hyroxylated adducts, the hydroxylated adduct of the openchain nitrone, POBN, decomposed to form a secondary radical adduct. Fig. 2 shows the ESR spectra obtained from mixtures containing human SOD1, POBN, H 2 O 2 , DTPA, and different concentrations of HCO 3 Ϫ anion in a phosphate buffer. The spectral intensity increased with increasing HCO 3 Ϫ concentrations. The ESR spectra consisted of two adducts corresponding to POBN-OH and the N-tert-butyl hydronitroxide, MNP-H. We propose that POBN-OH decomposed to the aldehyde and MNPhydroxylamine, which was further oxidized by CO 3 . to the MNP-hydronitroxide. A similar type of radical chemistry has previously been reported for the hydroxyl adduct of ␣-phenyltert-butyl-N-nitrone (16). As reported earlier, the proposed mechanism of hydroxylation of nitrones includes a nucleophilic addition of water to either the nitrone-carbonate radical adduct or to the radical cation intermediate (7). Independent evidence for the intermediacy of CO 3 . was obtained from photolysis studies using the pentamine carbonato complex of cobalt(III) (7). UV photolysis of this cobalt complex has been shown to release CO 3 . (17).
Irradiation of the pentammine carbonato complex of Co(III) in the presence of nitrone traps yielded ESR spectra of radical adducts that were similar to those obtained from the SOD1/ H 2 O 2 /HCO 3 Ϫ system (not shown). Direct ESR detection of the CO 3 . radical formed from mixing concentrated solutions of ONOO Ϫ and HCO 3 Ϫ was achieved using a rapid mixing technique, as reported earlier (18). We FIG. 6. Spin-trapping of the SODderived radical formed from bicarbonate-dependent peroxidase activity. A (top), hSOD1 WT (3 mg/ml) was mixed with H 2 O 2 (1 mM), HCO 3 Ϫ (25 mM), and spin trap DBNBS (10 mM) in a phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM). The mixture was transferred to a capillary tube (100 l), and the ESR spectrum was recorded immediately. Middle, same as top but in the absence of bicarbonate. Bottom, same as top but using bovine SOD1. B (top), the sample from A was ultrafiltrated (10-kDa cutoff), and the ESR spectrum was recorded. Middle, ultrafiltrated sample was treated with Pronase (2 mg/ml) for 1 h, and the ESR spectrum was recorded. Bottom, the central line of the nitroxide adduct (middle) was expanded using a modulation amplitude of 0. reproduced this result in our laboratory and observed that the half-life of this species is extremely short (ϳ6 ms) at physiological pH values (not shown). However, using this rapid mixing technique, we could not detect the carbonate radical anion proposed to be generated in SOD1/H 2 O 2 /HCO 3 Ϫ system (not shown). This is most likely due to a much slower rate of generation of CO 3 . in this system (i.e. rapid decomposition of CO 3 . combined with a slow rate of formation precluded its accumulation to the levels needed for ESR detection Ϫ (0 -25 mM) caused a concentration-dependent increase in the formation of a hSOD1 WT dimer (42-43 kDa) (Fig. 3A). Fig. 3B shows the time-dependent formation of the hSOD1 WT dimer from incubations containing hSOD1 WT , H 2 O 2 , and 25 mM HCO 3 Ϫ anion in a phosphate buffer. The existence of an SOD1 dimer under reducing SDS-PAGE conditions indicates that the subunits are covalently cross-linked via a nondisulfide type linkage. Formation of covalently linked higher molecular weight species (equivalent to dimers of 18-kDa SOD1 subunits) was initially noticeable after 15 min, and with prolonged incubation (60 -120 min), fragmentation occurred. Fig. 3, C and D, depicts covalent cross-linking of hSOD1 WT as a function of increasing H 2 O 2 and hSOD1-WT concentrations.
We then showed that UV photolysis of a cobalt complex that generates authentic CO 3 . radicals in the presence of hSOD1 WT caused its cross-linking (Fig. 3E). There was no apparent dimer formation cross-linking in the dark from incubations containing hSOD1 WT and the cobalt complex (Fig. 3E). These results provide additional evidence for CO 3 . -induced aggregation and covalent cross-linking of hSOD1 WT .

Inhibition of hSOD1 Aggregation by Nitrone Spin Traps-
The addition of the nitrone spin traps DMPO (25 mM), ␣-phenyl-tert-butyl-N-nitrone (25 mM), and azulenyl nitrone (4 mM) to an incubation containing hSOD WT , H 2 O 2 , DTPA, and HCO 3 Ϫ in a phosphate buffer blocked apparent dimer formation, as shown in Fig. 4A. Fig. 4B shows the densitometric analysis of the dimer band obtained in the presence of spin traps. Nitrone spin traps also inhibited UV/cobalt complex-induced covalent dimerization of hSOD1 WT (Fig. 4C). Me 2 SO, a well known trap for hydroxyl radicals, had no effect on hSOD1 WT dimer formation (Fig. 4D). These findings are consistent with the notion that trapping of CO 3 . by nitrones inhibits hSOD1 WT covalent dimerization.

The Effect of Amino Acids on Aggregation of SOD1 in a Human SOD WT /H 2 O 2 /HCO 3
Ϫ System-As shown earlier, evidence for enhanced formation of the hSOD1 WT covalent dimer was noticeable in incubations containing hSOD1 WT , H 2 O 2 , and HCO 3 Ϫ (Fig. 5A, lane p). Inclusion of glycine or phenylalanine had no effect on hSOD1 WT covalent dimer formation induced in the hSOD1 WT /H 2 O 2 /HCO 3 Ϫ system. Both tryptophan and tyrosine totally inhibited covalent dimer formation (Fig. 5A). Similar results were obtained during cobalt complex-photosensitized hSOD1 WT covalent aggregation (Fig. 5C). These results suggest that tryptophan and tyrosine scavenged the oxidant (i.e. CO 3 . ) responsible for hSOD1 WT covalent aggregation. We used DMPO to measure the HCO 3 Ϫ -dependent hSOD1 WT peroxidase activity. The DMPO-OH signal (Fig. 5B) was markedly inhibited in the presence of tryptophan and tyrosine (1 mM). At these concentrations, this signal intensity was not affected by phenylalanine and glycine. These data are consistent with the rapid scavenging of the carbonate anion radical by tryptophan and tyrosine (19). From these results, we conclude that either a tryptophanyl or tyrosyl residue present in hSOD1 WT is the proximal site of interaction with CO 3 . generated during HCO 3 Ϫdependent hSOD1 WT peroxidase activity.
Spin Trapping of the Bicarbonate-mediated Protein Radical Formed in the Human SOD1/H 2 O 2 System-An ESR spectrum that is characteristic of a strongly immobilized nitroxide adduct was detected from incubations containing hSOD1 WT , H 2 O 2 , HCO 3 Ϫ , and DBNBS trap (Fig. 6A, top). When the HCO 3 Ϫ anion was excluded, no spectrum was obtained (Fig. 6A, middle). Under the same experimental conditions, bovine SOD1 did not form a similar nitroxide adduct (Fig. 6A, bottom). Following ultrafiltration, treatment of the nitroxide adduct (Fig.  6B, top) with the Pronase enzyme that cleaved the high molecular weight nitroxide into a low molecular weight nitroxide yielded an isotropic three-line ESR spectrum (Fig. 6B, middle) with a hyperfine coupling constant of 13.6 G. Upon expansion of the center line of the ESR spectrum (Fig. 6B, middle), superhyperfine couplings were resolved (Fig. 6B, bottom). This spectrum was simulated (dotted line in Fig. 6B, bottom) using the following parameters (␣ N , 0.37 G; ␣ H1 , 0.13 G; ␣ H2 , 0.95 G; ␣ H3 , 0.58 G; ␣ Hm (2), 0.92 G). In the presence of tryptophan, the incubation mixture containing bovine SOD1, H 2 O 2 , HCO 3 Ϫ , and DBNBS yielded an intense isotropic three-line ESR spectrum (␣ N ϭ 13.6 G) (Fig. 6C, top). Without HCO 3 Ϫ , no ESR spectrum was obtained (Fig. 6C, middle). Upon expansion of the center line of the ESR spectrum (Fig. 6C, top) using a lower modulation amplitude, superhyperfine couplings could be resolved  1 mM). The solution was transferred to a capillary tube (100 l), and the ESR spectrum was recorded. B, same as above except that hSOD1 W32F was added to the incubation mixture. C, hSOD1 WT (1 mg/ml) was mixed with H 2 O 2 (1 mM), HCO 3 Ϫ (25 mM), and DMPO (25 mM) in a phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM), and the ESR spectrum was recorded using a capillary tube (100 l). D, same as C except in the presence of hSOD1 W32F . E, hSOD1 WT or hSOD1 W32F (1 mg/ml) was mixed with H 2 O 2 (1 mM) and HCO 3 Ϫ (25 mM) in an aerobic phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM). After incubating for 45 min, the aggregation of human SOD1 was measured by SDS-PAGE. (Fig. 6C, bottom). This spectrum was simulated using contributions from a nitrogen atom, three nonequivalent protons, and the two meta-protons present in DBNBS, as reported previously (20) (dotted line in Fig. 6C, bottom) ( Table I). The structure of the adduct was assigned to the DBNBS-tryptophanyl adduct (DBNBS-Trp). Based on these results, the immobilized nitroxide spectrum (Fig. 6A, top) is attributed to trapping of a radical formed from the tryptophan residue in hSOD1 WT . Concomitant with the trapping of the radical formed from the tryptophan residue, DBNBS markedly diminished HCO 3 Ϫ / hSOD1 WT peroxidase-dependent covalent dimer formation from hSOD1 WT (Fig. 6D).
The Effect of Substituting Phenylalanine for Tryptophan on Bicarbonate-mediated Hydroxylation and Oxidation Reactions in the hSOD1 W32F /H 2 O 2 System-The wild-type human SOD1 has a single tryptophan residue (Trp-32) located on the surface of the protein, as shown in the ribbon diagram model (Fig. 7A). To assess the role of tryptophan in hSOD1 WT /H 2 O 2 /HCO 3 Ϫmediated oxidation and aggregation reactions, a site-directed mutant of hSOD1, containing Phe in place of Trp-32, was produced (Fig. 7B). The fluorescence spectra of hSOD1 WT and hSOD W32F confirmed the loss of tryptophan fluorescence upon substitution with phenylalanine (Fig. 7, dotted line). The copper ESR spectra of hSOD1 WT and hSOD1 W32F were identical, demonstrating no change at the active copper site (not shown).
The next step was to compare the ESR spectra of the DBNBS adducts obtained from incubation mixtures containing either hSOD1 WT or hSOD1 W32F , DBNBS trap, H 2 O 2 , and HCO 3 Ϫ in a phosphate buffer (100 mM, pH 7.4) containing 0.1 mM DTPA. An intense ESR spectrum of a protein-derived radical was detected from the hSOD1 WT (Fig. 8A). No immobilized ESR spectrum was detected from the hSOD1 W32F -HCO 3 Ϫ -mediated reaction (Fig. 8C). The HCO 3 Ϫ -dependent peroxidase activity of hSOD1 WT and hSOD1 W32F was monitored by DMPO hydroxylation to DMPO-OH (Fig. 8, B and D). The lack of Trp-32 in hSOD1 W32F actually enhanced hSOD1 W32F /H 2 O 2 /HCO 3 Ϫ -dependent DMPO-OH formation (Fig. 8D), as compared with hSOD1 WT /H 2 O 2 /HCO 3 Ϫ /DMPO (Fig. 8B). This difference may be explained in terms of the reduced availability of the CO 3 . radical (due to increased stoichiometric scavenging of CO 3 . by Trp-32) for DMPO hydroxylation by hSOD1 WT , H 2 O 2 , and HCO 3 Ϫ . This was further verified by using free tryptophan as a substrate for SOD1/bicarbonate-mediated peroxidase activity. The spin trap DBNBS was used to trap the tryptophan-derived carbon-centered radical. The ESR spectra of DBNBS adducts obtained from incubations containing hSOD1 WT or hSOD W32F , DBNBS, H 2 O 2 , and HCO 3 Ϫ in a phosphate buffer containing DTPA are shown in Fig. 8, A and C. In the presence of 50 M  1 mM), and UV-visible spectra were recorded every 5 min. B (left), incubation conditions are the same as in A (left) except that the decrease in the fluorescence spectra was recorded every 5 min. Middle, the same as A (right) except that the fluorescence changes were recorded. Right, incubation conditions are the same as in A (right) except that the emission spectra of products were obtained using the excitation wavelengths as indicated. free tryptophan, an ESR spectrum due to the DBNBS-Trp adduct was detected from hSOD1 W32F /H 2 O 2 /HCO 3 Ϫ /DBNBS system (not shown), and at 1 mM free tryptophan, the intensity of the spectrum was higher (not shown). At a lower concentration of free Trp, spectra from both DBNBS-hSOD1 WT and DB-NBS-Trp were detected; however, at 1 mM free Trp, the ESR spectrum was solely due to the DBNBS-Trp adduct and was almost the same as that observed from hSOD1 W32F (not shown). These results unequivocally demonstrate that hSOD1 WT and hSOD1 W32F exhibit the same extent of bicarbonate-mediated peroxidase activity, and the lack of ESR spectrum from hSOD1 W32F suggests that CO 3 . reacts with the tryptophan residue to form a tryptophan-derived carbon-centered radical that was trapped by DBNBS. It was then of interest to find out whether substituting Trp-32 with Phe-32 has any effect on the aggregation and covalent cross-linking of the protein. We compared the covalent aggregations of hSOD1 WT and hSOD1 W32F in the presence of H 2 O 2 and HCO 3 Ϫ . Fig. 8E shows that covalent dimerization of hSOD1 W32F did not occur under the same experimental conditions (Fig. 8A) that resulted in covalent dimerization of hSOD1 WT . Clearly, the Trp residue at position 32 plays a crucial role in the oxidation, aggregation, and covalent cross-linking of hSOD1 WT caused by HCO 3 Ϫ -mediated peroxidase activity. The Effect of Molecular Oxygen on the Fluorescence Spectra of Tryptophan-derived Oxidation Products-The UV-visible and fluorescence spectral changes in the oxidation products of tryptophan in a bovine SOD1/H 2 O 2 /HCO 3 Ϫ system are shown in Fig.  9. Bicarbonate accelerated the rate of oxidation of tryptophan in the presence of SOD1 and H 2 O 2 (Fig. 9, A (middle) and B (left)). Similar spectral changes due to oxidation of Trp-32 residue were observed in the hSOD1 WT /H 2 O 2 /HCO 3 Ϫ system (Fig. 9, A (right) and B (middle)). Fluorescence spectra (Fig. 9B, right) revealed at least three different products that are formed during HCO 3 Ϫ /H 2 O 2 -dependent oxidation of Trp-32 in hSOD1 WT . The fluorescent intensities of two products (Fig. 10,  A and B) were monitored under aerobic and anerobic conditions in the hSOD1 WT /H 2 O 2 /HCO 3 Ϫ system. Product formation was considerably enhanced in air. Bicarbonate/H 2 O 2 -dependent dimerization of hSOD1 WT was enhanced in air and diminished in N 2 (Fig. 10C). These results indicate that molecular oxygen enhances HCO 3 Ϫ /H 2 O 2 -dependent aggregation and covalent cross-linking of hSOD1 WT via formation of the tryptophanyl carbon-centered and peroxyl radicals (see Scheme 1). DISCUSSION This study revealed that HCO 3 Ϫ augments human SOD1 WT peroxidase activity through formation of CO 3 . , a potent oneelectron oxidant, at the active site of the enzyme. Indeed, secondary formation of CO 3 . provides biological relevance to the peroxidase activity via the formation of a strong oxidant that can diffuse away from the active site, unlike the hydroxyl radical. Because of its selective reactivity, CO 3 . diffuses out of the active site and oxidizes numerous substrates in the bulk solution. The rapid reaction between CO 3 . and the tryptophan residue (Trp-32) located on the surface of the enzyme leads to a covalent aggregation that is nondisulfide in nature (i.e. dithiothreitol-resistant) and, therefore, more irreversible under biological conditions. The ESR spin trapping provided the spectroscopic proof for the formation of a tryptophan-derived radical. The site-directed mutation of Trp-32 by Phe-32 completely prevented HCO 3 Ϫ /hSOD1 WT peroxidase-induced aggregation and covalent cross-linking, demonstrating that Trp-32 oxidation is responsible for the covalent aggregation of hSOD1 WT .
In Vivo Peroxidase Activity of SOD1: The Role of CO 3 . Radical-The peroxidase activity of SOD1 has previously been in-correctly attributed to formation of the hydroxyl radical (21)(22)(23). In the absence of HCO 3 Ϫ , SOD1 WT peroxidase and superoxide dismutase activities were markedly inhibited. In the presence of HCO 3 Ϫ , the SOD1 superoxide dismutase still remained inhibited (not shown), whereas the SOD1 peroxidase activity was considerably increased. Due to its abundance in Incubations were performed in air or under N 2 . The fluorescence intensity was measured after 30 min using the excitation and emission wavelengths (320 and 420 nm). B, same as A except that the fluorescence intensity was measured using excitation and emission wavelengths 365 and 460 nm, respectively. C, after incubating for 45 min, samples obtained from incubations performed under conditions shown in A were analyzed by SDS-PAGE.

cells and tissues, HCO 3
Ϫ -induced oxidative and nitrative reactions are biologically relevant. Recently, evidence for CO 3 . -dependent enhanced peroxidase activity was shown in an in vivo yeast model expressing hSOD1 (24). The spin trap POBN was used to trap the hydroxyethyl radical derived from the reaction between CO 3 . and ethanol generated in situ in yeast cultures (24). The first in vivo spin trapping evidence for increased generation of oxygen-centered radicals was provided in a mouse model (25). The spin trap azulenyl nitrone was administered to ALS SOD1 transgenic mice (SOD1 G93A ), and the corresponding oxidative metabolite azulenyl aldehyde was detected (25). The proposed mechanism involved trapping of a peroxyl radical, followed by a decomposition of the nitroneperoxyl adduct (25,26). Based on the present findings obtained with an analogous open-chain nitrone, POBN, it is conceivable that the azulenyl aldehyde is formed from oxidation and hydrolysis of azulenyl nitrone by CO 3 . (6). Thus, it is also conceivable that the G93A mutant of hSOD1, which causes ALS in humans and motor neuron disease in transgenic mice, may be producing such oxidants in vivo. Comparison between CO 3 . and ⅐ OH Reactivities-Using a steady-state ESR spin trapping measurements, it is difficult to differentiate between ⅐ OH and CO 3 . based on their reaction with conventional hydroxyl radical scavengers (e.g. azide, formate, ethanol) (Fig. 1). Both ⅐ OH and CO 3 . react with these scavengers rapidly to generate the corresponding scavenger-derived radicals, which are then trapped by the nitrone spin traps, yielding a characteristic ESR spectrum of the resulting spin adduct. Thus, in order to differentiate between CO 3 . and ⅐ OH radical formation, it is necessary to monitor the initial rates of spin adduct formation as a function of trap concentrations and cal-culate the rate constants for ⅐ OH and CO 3 . reaction with the relevant traps (27). However, Me 2 SO appears to be an exception. As shown in Fig. 1 (28). This was attributed to the lack of reaction between copperbound hydroxyl radical and Me 2 SO.
In the present work, we obtained independent evidence for CO 3 . -mediated oxidation reactions by UV-photolysis of the pentammine carbonato complex of Co(III) as shown below.
Irradiation of this complex in the presence of DMPO and azide, formate, etc. yielded ESR spectra of adducts that were similar to those obtained from a SOD1/H 2 O 2 /HCO 3 Ϫ system (6). The spin trapping results also revealed that, in the presence of free tyrosine and tryptophan, the DMPO-OH signal intensity obtained in the SOD1/H 2 O 2 /HCO 3 Ϫ system and from cobalt complex/UV light was markedly inhibited in the presence of tyrosine or tryptophan but not in systems containing glycine or alanine. These results are consistent with the observed rate SCHEME 1. Proposed mechanism for covalent aggregation of human SOD1 in the presence of hydrogen peroxide and bicarbonate. REACTION 1 constants reported for CO 3 . and ⅐ OH (Table II). Loss of tryptophan fluorescence in the hSOD1 WT /H 2 O 2 /HCO 3 Ϫ system has been attributed to CO 3 . -mediated oxidation of Trp-32 residue.
Recently, Yamakura et al. (29) reported that peroxynitrite and bicarbonate treatment induced a loss in Trp-32 fluorescence in the hSOD1 WT . The investigators suggested that the oxidative modification of Trp-32 residue in hSOD1 WT enzyme is caused by the peroxynitrite-carbon dioxide adduct. However, it is well known that the peroxynitrite-carbon dioxide adduct or nitrosoperoxycarbonate decomposes to CO 3 . and ⅐ NO 2 radicals at physiological pH (30,31). Thus, the observed modification of Trp-32 reported in that study (29) could be explained on the basis of the reaction between CO 3 . and Trp-32 in hSOD1 WT .
Tryptophan Radical-mediated Covalent Aggregation of Human SOD1: Dependence on Molecular Oxygen-Based on the previous work by Gunther et al. (20) and the present spin trapping results using bovine SOD1 and hSOD1 W32F , we conclude that Trp-32 is the site of oxidation and that the radical adduct is formed from trapping of a carbon-centered radical at the C-3 of the indole ring. As reported previously (20) and reiterated in this study, the ESR simulation of the high resolution spectrum of hSOD1 WT -DBNBS adduct (Table I) strongly supports the structural assignment. The primary carbon-centered radical associated with Trp-32 most likely reacts with molecular oxygen to form the peroxyl radical, as shown in Scheme 1. As reported by Gunther and co-workers (32), the oxygen evolution was detected in incubations containing bovine SOD1 WT , HCO 3 Ϫ , and H 2 O 2 (Fig. 11A, trace 1). As indicated earlier, the bovine SOD1 lacks the tryptophan residue. In the presence of added tryptophan, the oxygen evolution was completely suppressed due to oxygen consumption by the tryptophan-derived radicals. In fact, there was a slight increase in oxygen consumption (Fig. 11A, trace 4). In the presence of DBNBS spin trap, tryptophan-induced oxygen consumption was inhibited (Fig. 11A, trace 2). Similar results were obtained with hSOD WT (which contains a tryptophan residue) in the absence of added tryptophan (Fig. 11B, trace 3). In the presence of the DBNBS spin trap, oxygen evolution was observed from incubations containing hSOD1 WT , HCO 3 Ϫ , and H 2 O 2 (Fig. 11B,  trace 1). This is attributed to trapping of the Trp-32-derived carbon-centered radical inhibiting radical-mediated oxygen consumption. This result further supports the possibility that the DBNBS-hSOD1 WT adduct is formed from trapping the Trp-32 radical by DBNBS and not from a non-radical-mediated "ene" reaction between DBNBS and the indole moiety in Trp-32 (33). Attempts to directly detect and characterize the peroxyl radical at low temperatures, as had been demonstrated in the metmyoglobin/H 2 O 2 system, were not possible in the hSOD1 WT /H 2 O 2 /HCO 3 Ϫ system due to interference from copper ESR absorption. Increased oxygen evolution was observed from incubations containing hSOD1 W32F (mutant lacking the tryptophan residue), HCO 3 Ϫ , and H 2 O 2 (Fig. 11B, trace 2). Results from this study reveal that radical-mediated oxidation of Trp-32 in the presence of molecular oxygen is responsible for H 2 O 2 /HCO 3 Ϫ -dependent covalent aggregation of hSOD1 WT . Anaerobic conditions prevented the covalent dimerization of hSOD1 WT induced by HCO 3 Ϫ and H 2 O 2 (Fig. 10). The oxidative cleavage of the indole ring of tryptophan has been shown to yield multiple products, including hydroxyindole, Nformyl-kynurenine, and kynurenine (34 -36). These products can cause multiple cross-linking reaction pathways, leading to protein aggregation. We have proposed a tentative reaction pathway involving CO 3 . -dependent Trp-32 oxidation, leading to oligomerization of hSOD WT via oxidative derivatives of Trp-32 (Scheme 1). Fig. 9 (left) provides some evidence for formation of oxidative degradation products during HCO 3 Ϫ /H 2 O 2 -dependent SOD1 peroxidase activity. Fluorescence emission spectrum excited at 320 nm yielded a product that had an emission maximum around 420 nm. This peak may be assigned to the Nformyl-kynurenine (excitation, 310 -330 nm; emission, 410 -430 nm). This spectrum also suggests that dimerization does not occur via ditryptophan formation, since ditryptophan exhibits a distinctly different fluorescence (excitation, 320 nm; emission, double peaks at 370 -380 nm). Fig. 9B (right) also shows that excitation at 365 and 410 nm yielded characteristic emission at 450 and 520 nm, respectively. These spectra have previously been assigned to kynurenine (excitation, 365 nm; emission, 460 nm) and to cross-linked products formed from the reaction between kynurenine or 3-hydroxykynurenine and lysine or a histidine amino group (37,38). Results from the preliminary matrix-assisted laser desorption ionization-timeof-flight experiments are consistent with the notion that Trp-32 is oxidized by hSOD1 WT /HCO 3 Ϫ /H 2 O 2 to the kynurenine-type oxidative products (39) (not shown).
Implications of Covalent Aggregation of SOD1 in Familial ALS-Results from this study may increase our understanding of the fundamental mechanism(s) by which some forms of SOD1 mutants induce ALS (40,41). Although the mechanism of pathogenesis in ALS and the role of ALS SOD1 mutants remains controversial and poorly understood, several reports suggest that familial ALS, a dominantly inherited form of ALS, is linked to a toxic gain of function in mutant SOD1 (i.e. increased peroxidase activity, peroxynitrite-mediated nitration, or enhanced aggregation of SOD1) (42)(43)(44). It is now FIG. 11. The effect of tryptophan and DBNBS on SOD1-induced oxygen release. A, bovine SOD1 (2 mg/ml) was incubated with H 2 O 2 (0.5 mM) with or without HCO 3 Ϫ (25 mM) in phosphate buffer (100 mM, pH 7.4) containing DTPA (0.1 mM). Where indicated, incubations contained DBNBS (25 mM) or added tryptophan (5 mM). B, same as A, except that the incubation mixture contained hSOD1 WT or hSOD1 W32F . Oxygen uptake/release experiments were performed using a World Precision Oxygen electrode at room temperature. The electrode was calibrated using air-saturated water (240 M oxygen was equated to 100%). becoming more evident that nitration and oxidation reactions mediated by peroxynitrite are modulated by CO 2 , a ubiquitous cellular component (see Ref. 44 and references therein). The peroxynitrite-CO 2 intermediate decomposes into CO 3 . and ⅐ NO 2 radicals and causes oxidative modifications in amino acids (e.g. tryptophan and tyrosine). Thus, enhanced covalent aggregation of hSOD1 could be mediated by peroxynitrite/CO 2 chemistry or by CO 3 . -dependent peroxidase activity. ALS-associated SOD1 mutants were shown to exhibit increased peroxidase activity (42). It is reasonable to suggest that HCO 3 Ϫ could exacerbate the intrinsic difference between wild-type and familial ALS mutant SOD1 with respect to peroxidase activity. Although protein aggregation is frequently associated with several neurodegenerative diseases (41,46), it is not known whether they are a cause or consequence of toxic oxidative stress in ALS. Recent reports suggest that HCO 3 Ϫ -dependent peroxidase activity is greatly elevated in the spinal cord extracts of SOD1 G93A mice (6). Published data also indicate that high molecular weight complexes of mutant SOD1 (apparent dimers and trimers) are increased in the spinal cord extracts of mutant mice (43,45). It is not known whether these aggregates were formed from noncovalent or covalent aggregation (47,48). In the present study, we present evidence for covalent aggregation of hSOD1 WT that is mediated by a novel oxidative pathway involving the tryptophan-derived oxidation products.