Amyotrophic Lateral Sclerosis Mutations Have the Greatest Destabilizing Effect on the Apo- and Reduced Form of SOD1, Leading to Unfolding and Oxidative Aggregation*

Mutant forms of Cu,Zn-superoxide dismutase (SOD1) that cause familial amyotrophic lateral sclerosis (ALS) exhibit toxicity that promotes the death of motor neurons. Proposals for the toxic properties typically involve aberrant catalytic activities or protein aggregation. The striking thermodynamic stability of mature forms of the ALS mutant SOD1 (Tm >70 °C) is not typical of protein aggregation models that involve unfolding. Over 44 states of the polypeptide are possible, depending upon metal occupancy, disulfide status, and oligomeric state; however, it is not clear which forms might be responsible for toxicity. Recently the intramolecular disulfide has been shown to be required for SOD1 activity, leading us to examine these states of several disease-causing SOD1 mutants. We find that ALS mutations have the greatest effect on the most immature form of SOD1, destabilizing the metal-free and disulfide-reduced polypeptide to the point that it is unfolded at physiological temperatures (Tm < 37 °C). We also find that immature states of ALS mutant (but not wild type) proteins readily form oligomers at physiological concentrations. Furthermore, these oligomers are more susceptible to mild oxidative stress, which promotes incorrect disulfide cross-links between conserved cysteines and drives aggregation. Thus it is the earliest disulfide-reduced polypeptides in the SOD1 assembly pathway that are most destabilized with respect to unfolding and oxidative aggregation by ALS-causing mutations.

Mutant forms of Cu,Zn-superoxide dismutase (SOD1) that cause familial amyotrophic lateral sclerosis (ALS) exhibit toxicity that promotes the death of motor neurons. Proposals for the toxic properties typically involve aberrant catalytic activities or protein aggregation. The striking thermodynamic stability of mature forms of the ALS mutant SOD1 (T m >70°C) is not typical of protein aggregation models that involve unfolding. Over 44 states of the polypeptide are possible, depending upon metal occupancy, disulfide status, and oligomeric state; however, it is not clear which forms might be responsible for toxicity. Recently the intramolecular disulfide has been shown to be required for SOD1 activity, leading us to examine these states of several disease-causing SOD1 mutants. We find that ALS mutations have the greatest effect on the most immature form of SOD1, destabilizing the metal-free and disulfide-reduced polypeptide to the point that it is unfolded at physiological temperatures (T m < 37°C). We also find that immature states of ALS mutant (but not wild type) proteins readily form oligomers at physiological concentrations. Furthermore, these oligomers are more susceptible to mild oxidative stress, which promotes incorrect disulfide cross-links between conserved cysteines and drives aggregation. Thus it is the earliest disulfide-reduced polypeptides in the SOD1 assembly pathway that are most destabilized with respect to unfolding and oxidative aggregation by ALS-causing mutations.
Amyotrophic lateral sclerosis (ALS) 1 is a neurodegenerative disease caused by selective degeneration of motor neurons in the brain and spinal cord (1). About 10% of total ALS cases are of familial origin, and of these, ϳ20% are caused by point mutations in the abundant enzyme Cu,Zn-superoxide dis-mutase (SOD1) (2). SOD1 is a dimer of identical subunits, each of which contains a copper and a zinc ion, and catalyzes conversion of superoxide anion to oxygen and hydrogen peroxide at a dinuclear copper-zinc site (3). Studies of SOD1 knock-out mice reveal no motor neuron disease (4), indicating new toxic functions of ALS mutations. Among the proposed gain of functions is aberrant redox chemistry arising from modification of active copper and zinc sites in ALS mutants. For example, enhanced peroxidase activity (5), protein nitration (6), and superoxide production (7) have been proposed. The requirement of copper ion for toxicity has, however, been put into question, because ALS symptoms are observed in mice expressing the SOD1 mutant in which all of the copper ligands are mutated (8).
Another hypothesis for a toxic function acquired by the mutant SOD1 proteins is an increased propensity for cytoplasmic aggregation, which is one of the pathological features common to familial ALS (9). ALS mutations have been considered to induce protein misfolding and/or destabilization; in fact, ALS mutations provoke a decrease of 1-6°C in the melting temperature of SOD1 (10), although wild type (WT) SOD1 remains active after treatment at 80°C (11). Reduced zinc affinity in ALS mutants has been proposed to facilitate destabilization of mature SOD1 (12), and conversely zinc binding has been known to increase the protein stability (11). Many studies have focused on how mutations predispose the active mature protein to adventitious reactions; however, studies of the polypeptide in the earlier stages are now emerging (13)(14)(15) as the physiological pathways controlling activation of the enzyme are uncovered. Although it remains unclear how SOD1 acquires zinc ions, the copper chaperone for SOD1 (CCS) typically controls copper acquisition (16 -18) and disulfide formation in apo-SOD1 (19). Given that ALS mutants are more susceptible to disulfide reduction compared with the WT protein (20), adventitious reductions have been proposed as an important step in the disease. It is not known, however, if altered regulation of the thiol/disulfide status is associated with ALS mutations. In fact, many of the early physicochemical studies were conducted on mixtures of the SOD1 proteins in various degrees of posttranslational modifications. In particular, the status of the essential intra-subunit disulfide and the protein quaternary structure is important but frequently unreported.
In this study, we focus on how the stability of both WT and ALS mutant proteins is influenced by post-translational modifications. Given that four post-translational events precede SOD1 activation, i.e. copper and zinc binding, disulfide formation, and dimerization, it is clear that SOD1 can adopt 44 canonical microstates, where metal binding at incorrect sites are not taken into account. It was not clear which states of the protein are most destabilized by disease-causing mutations. By examining representative subgroups of these states, we can now establish a hierarchy for the effects of zinc binding and disulfide formation on stability and aggregation of WT SOD1 and three ALS mutants (A4V, G85R, and G93A). The results indicate that the unmodified polypeptide is least stable and show that ALS mutants in this state are unfolded and oligomerized at physiological temperature. Disulfide formation and zinc binding reduce oligomer formation by increasing the thermal stability of SOD1. The disulfide-reduced forms, furthermore, are highly susceptible to disulfide-linked multimerization upon oxidative stress, indicating that regulation of intra-or inter-molecular disulfide formation can determine whether SOD1 aggregates or adopts the mature state. In contrast to many models that focus on altered attributes of the mature SOD1 enzyme, these results now focus attention on aberrant properties of the most immature polypeptide and on the role that post-translational modifications may play in SOD1 aggregation.

EXPERIMENTAL PROCEDURES
Preparation of the Apo and Reduced Human SOD1 Proteins-Proteins were overexpressed in Escherichia coli BL21(DE3) and purified as described previously (19,21). For preparation of the disulfide-reduced protein, 750 l of as-isolated human SOD1 (hSOD1) (ϳ3.0 g/liter) was anaerobically treated with 200 mM dithiothreitol (DTT) to reduce disulfide bonds. After anaerobic incubation at 37°C for an hour, solutions were acidified with 0.4% trifluoroacetic acid to reduce the reactivity of thiol groups and to promote the dissociation of metal ions from SOD1. Following incubation in 15% CH 3 CN, 10% CH 3 OH at room temperature for an hour, the proteins were purified by reverse-phase high pressure liquid chromatography through a C4 Vydac 214TP54 column as described (19). The isolated apo-SOD1 has less than 0.1 and 10% for copper and zinc ions, respectively, based on the inductively coupled plasma atomic emission spectroscopic analysis. Reduction of the disulfide bond was confirmed by the 4-acetamido-4Ј-maleimidylstilbene-2,2Јdisulfonic acid (AMS) modification (see below). The resulting apo and reduced SOD1 proteins can be activated by adding copper-bound CCS (data not shown) (19,22), indicating that the protein tolerates relatively harsh treatment with acid and organic solvents.
The disulfide form of the protein was prepared by oxidizing ϳ200 M E,E-hSOD1 SH with 1 mM K 3 [Fe(CN) 6 ] in the presence of 200 M ZnSO 4 and 20% glycerol. After 1 h of aerobic incubation, the solution was acidified and purified by high pressure liquid chromatography as described above. Protein concentrations were determined with the absorption at 280 nm using 5500 M Ϫ1 cm Ϫ1 per monomer as the extinction coefficient (23). The Bradford assay with IgG as a standard was also used to routinely determine the protein concentration.
Thiol Modifications-To analyze the thiol-disulfide status in hSOD1, modification with either AMS (Molecular Probes, Inc.) or iodoacetamide (IA) was employed (19). Protein was precipitated with 20% trichloroacetic acid, washed with acetone, and dried under vacuum. Protein pellets were brought into the anaerobic chamber and redissolved in the modification buffer (50 mM HEPES, pH 7.2, 1 mM EDTA, 1 mM bathocuproine disulfonate, 2.5% SDS) containing either 25 mM AMS or 100 mM IA. After anaerobic incubation at 37°C for an hour, the modified proteins were separated by SDS-PAGE.
Electrophoresis-Proteins were separated on 12.5% polyacrylamide Tris-HCl resolving gel with a 5.4% polyacrylamide stacking gel. Samples were dissolved in Laemmli buffer with/without ␤-mercaptoethanol (␤-ME) and boiled at 95°C for 2 min before loading. The gels were run at constant voltage of 150 V and stained by Coomassie Brilliant Blue. For the Western blot, separated proteins in the SDS-polyacrylamide gel were blotted to a polyvinylidene fluoride membrane (Bio-Rad) and blocked for an hour with the ECL blocking agent (Amersham Biosciences). The membrane was further incubated with the anti-superoxide dismutase (Cu/Zn), human erythrocyte (sheep) (1:1000 dilution, Calbiochem) as the primary antibody, and with rabbit anti-sheep IgG (H ϩ L) (1:1000 dilution, Bio-Rad) as the secondary antibody. Blots were developed with the ECL Plus Western blotting detection system (Amersham Biosciences).
Differential Scanning Calorimetry-Melting temperatures were obtained by using Microcal VP-DSC using a scan speed, 90°C/h, under ϳ31 pounds/square inch of pressure, with the filter set to 5 s. Base-line curves were derived using Origin software (Microcal). About 500 l of 1-2 g/liter SOD1 proteins was used for measurement. Chelex-treated 50 mM K-P i , pH 7.8, was used as a buffer, and 1 mM tris-(2-carboxyethyl)phosphine (TCEP, Molecular Probes, Inc.) was included when the disulfide-reduced proteins were examined. Glycerol-containing buffer was also treated with Chelex before its use for measurements. Solutions were degassed for 25 min using a vacuum pump with stirring before measurements. The experiments were repeated two or three times to estimate the experimental error of T m values. After measurements, samples were submitted to AMS modification followed by the SDS-PAGE analysis, revealing that no air oxidation of the thiol groups occurred in the course of the experiment.
Gel Filtration Chromatography-200 l of 30 M hSOD1 was loaded on Superose 12 HR 10/30 (Amersham Biosciences). For examination of the E,E-hSOD1 states, the column was pre-equilibrated with degassed buffer (50 mM K-P i , 0.1 mM EDTA, pH 7.8) at 4°C, and the flow rate was 1.0 ml/min. In the case of the E,Zn-hSOD1 states, 30 M ZnSO 4 was added to the buffer in place of 0.1 mM EDTA. To prevent possible air oxidation of the Cys groups, 1 mM DTT was added to the eluant in experiments using reduced SOD1 proteins. Absorbance changes were monitored at 215 nm. The column was calibrated using 0.25 g/liter immunoglobulin G, bovine serum albumin, ovalbumin, carbonic anhydrase, horse skeletal myoglobin, E. coli thioredoxin, and aprotinin.

RESULTS
Human SOD1 (hSOD1) has four cysteine residues; Cys 6 , Cys 57 , Cys 111 , and Cys 146 , among which Cys 57 and Cys 146 are conserved and form the intra-molecular disulfide bond (24). In this study, X,Y-hSOD1 SH/S-S signifies protein derivatives in which metal ions have been substituted into the copper-binding site (X) or zinc-binding site (Y). E means empty at the metalbinding site, and the superscripts, SH or S-S, indicate the thiol/disulfide status of the conserved Cys residues. For instance, E,Zn-hSOD1 SH denotes disulfide-reduced protein with a Zn 2ϩ ion at the zinc-binding site but no metal ion at the copper-binding site. To reveal the effects of post-translational modifications on protein stability, we first measured the melting temperatures of E,E-and E,Zn-hSOD1 SH/S-S .
Post-translational Modifications Can Increase SOD1 Thermal Stability-Whereas accurate enthalpy values for SOD1 thermal unfolding are not readily obtained from the differential scanning calorimetry (DSC) thermograms (25), comparisons of the relative melting temperature, T m , reveal relationships among the thermal stabilities of isolated states of SOD1. The unmodified WT protein, E,E-hSOD1 SH , melts between 30 and 50°C and exhibits an endothermic peak at 42.9°C (Fig.  1A). Given that the holoenzyme (Cu,Zn-hSOD1 S-S ) is one of the most stable proteins known in mesophilic organisms (11), it is interesting that the T m of the SOD1 polypeptide in the absence of any post-translational modifications is just above the typical physiological temperature in mammals, ϳ37°C. Disulfide formation and zinc binding stabilize the SOD1 polypeptide, increasing the T m to 49.8 and 58.4°C, respectively. The T m value is further increased to 74.6°C upon formation of the disulfide in E,Zn-hSOD1 (Fig. 1, A and B). Fig. 1C graphically shows increase of T m , ⌬T m , after post-translational modifications relative to the unmodified E,E-hSOD1 SH state. Disulfide formation in E,E-hSOD1 SH increases T m by as much as 6.9°C, whereas a 15.5°C increase is observed upon binding of the Zn 2ϩ ion. Zinc binding therefore has a larger contribution to SOD1 stability than disulfide formation. These increases of T m are comparable with those upon disulfide formation or zinc binding in other proteins (26,27). It is also notable that the ⌬T m of 31.7°C upon both zinc binding and disulfide formation is significantly larger than ⌬T m values for the individual modifications (Fig. 1C). Zinc binding and disulfide formation can thus increase the thermal stability of SOD1.
Given that one of the disulfide-bonding cysteine residues, Cys 57 , is positioned in loop IV, reduction of the loop flexibility upon disulfide formation would result in the SOD1 stabiliza-tion ( Fig. 1, D and E). Furthermore, not only Cys 57 but also all ligands for the Zn 2ϩ ion (His 63 , His 71 , His 80 , and Asp 83 ) are involved in loop IV. Zinc binding would hence be able to introduce structural restraints around loop IV more significantly than disulfide formation, which is consistent with a larger stabilizing effect due to zinc binding (Fig. 1C). Disulfide formation and zinc binding may stabilize SOD1 through reduction of the loop IV flexibility; in fact, E,E-hSOD1 SH is on the threshold of unfolding at physiological temperature even without any disease-associated mutations (Fig. 1A). This is interesting given that protein aggregates containing WT hSOD1 have been found in a subset of sporadic forms of ALS disease where the molecular defects remain unknown (28).
During DSC measurements, formation of non-native disulfide bonds such as the one between Cys 6 and Cys 111 may also occur (25). We thus further tested for stabilizing effects of the post-translational modifications on hSOD1 by using the following Cys mutants: C6S/C111S and C6S/C57S/C111S/C146S (C 4 S). The C 4 S mutant has no Cys residues and can serve as a control for the disulfide-reduced form, whereas the C6S/C111S mutant, which has no free thiol groups when the conserved disulfide bond (Cys 57 -Cys 146 ) is intact, can be compared with the disulfide state of WT protein. The apo-form of the C 4 S mutant and the reduced C6S/C111S mutant exhibit T m values of 37.2 and 37.8°C, respectively (Table I). As with WT protein, zinc addition and disulfide formation in the Cys mutants lead to protein stabilization by increasing T m (Table I). We conclude from these results that the observed stabilization of WT protein upon oxidation arises from formation of the conserved disulfide and not from adventitious disulfides involving Cys 6 or Cys 111 . Another important observation in Table I is that mutation of these two Cys residues slightly stabilizes the modified proteins but destabilizes the unmodified polypeptides; the unmodified polypeptides of both Cys mutants represent a destabilization between 5.1 and 5.7°C from that of WT (⌬T m mut in Table I). The alterations in T m observed upon the Cys mutations may be relevant in the ALS mechanism, because disease-associated mutations have been reported for Cys 6 and Cys 146 (www.alsod. org). In order to examine how disease-associated mutations alter the protein stability, we next measured T m of two common and widely studied ALS mutants, A4V and G93A.
Unmodified ALS Mutant SOD1 Polypeptides Melt at Temperatures Below Physiological Threshold-Previous studies have shown that T m values of ALS mutants are 3-15°C lower compared with that of WT protein, although in these cases the thiol/disulfide status was not determined, and mixtures of states are sometimes apparent in the thermograms (10,29). Most unexpectedly, however, we could not characterize an endothermic transition on the E,E-hSOD1 SH samples of A4V and G93A mutants in 50 mM K-P i , 1 mM TCEP, pH 7.8, between 5 and 100°C. A simple explanation is that this state of the ALS mutants is not fully folded or does not have a distinct structure that is detectable with calorimetry (see below); however, the negative results do not support a strong conclusion. To test this further, we repeated the experiment in the presence of 20% glycerol, which generally increases protein stability and melting temperature by lowering water activity (30). Under these conditions, clear melting temperatures for the E,E-hSOD1 SH states of ALS mutants were obtained, 35.2°C (A4V) and 31.2°C (G93A). These T m values for the mutant E,E-hSOD1 SH states are significantly lower (by 10.6°C in A4V and 14.6°C in G93A) than that of WT protein in the same glycerol-containing buffer (⌬T m mut in Table II). These results show that ALS mutants are completely unfolded in the E,E-hSOD1 SH state at physiological temperature and concentration, even when protein-stabilizing agents are present.
As with WT protein (Table I), disulfide formation and zinc binding in E,E-hSOD1 SH of ALS mutants raise T m values and significantly increase protein stability (Fig. 2, A and B, and Table II). T m values of WT and ALS mutants before and after post-translational modifications in the presence of 20% glycerol are schematically compared in Fig. 2C. Unlike the WT protein, however, disulfide formation alone may be inefficient at stabilizing the A4V and G93A mutant proteins at physiological temperature, especially in the absence of 20% glycerol, T m ϭ 37.4°C (A4V) and 40.3°C (G93A) ( Table II). This is consistent with the previous study showing that apo forms of SOD1 of A4V and G93A are partially unfolded at 23°C (31). In that case, the disulfide status is not reported, but, judging from the preparation method, apo-SOD1 would have the disulfide bond. On the other hand, zinc binding in ALS mutants leads to a T m high enough to be completely folded at physiological temperature, and the lowering of the T m due to ALS mutations is alleviated by zinc binding (Fig. 2C and Table II). To highlight the mutational effects on the protein stability, differences in T m between WT and ALS mutants, ⌬T m mut , are plotted against each state (Fig. 2, D and E). In both ALS mutants, the most significant ⌬T m mut is observed for the E,E-hSOD1 SH state, and post-translational modifications, especially zinc binding, rescue the protein from its destabilization due to the mutations. Because unfolded or incompletely folded proteins will be generally more solvent-exposed, they are prone to inappropriate interactions with proteins (32,33). That ALS mutations drop the T m values of the E,E-hSOD1 SH state below physiological temperature supports the argument that these forms contribute to aggregation. These are the first reports of T m values for the most immature form of SOD1. The new results for more mature forms are consistent with the few published DSC studies on SOD1 and, in some cases, facilitate assignment of T m values reported in the literature for mixtures (Table S1). To probe the SOD1 aggregation states, we characterized each modified state by using the gel filtration chromatography.
Unmodified Form of ALS Mutants Is Oligomerized-Gel filtration studies conducted at 4°C show that E,E-hSOD1 SH of WT is a monomer at 30 M (Fig. 3A), a concentration within the physiological range of 10 -100 M (34,35). Consistent with our previous findings (36), either disulfide formation, zinc addition, or both leads to dimerization (Fig. 3, A and B). No aggregated forms are observed, which is also consistent with the calorimetry results showing that the WT polypeptide melts at 42.9°C (Table II).
Unlike the distinct monomeric and dimeric states observed for WT protein, most of E,E-hSOD1 SH (A4V) elutes from a gel filtration column with an apparent molecular weight corresponding to 50 -60 kDa, and its broad elution profile extends to the region corresponding to the monomeric species (Fig. 3C). Without any modifications, therefore, the A4V mutant is present in oligomerized forms at 4°C. Even in the presence of 0.2 M NaCl, which would reduce possible electrostatic interactions in the oligomer, the oligomeric fractions are still observed (data not shown). Oligomerization may be related to our inability to observe an endothermic peak in the DSC thermogram of E,E-   2. The most immature form of ALS mutant proteins is thermally unfolded at physiological temperature. DSC profiles for the reduced (2.00 g/liter) (A) and the disulfide form (1.73 g/liter) (B) of G93A in 50 mM K-P i , pH 7.8. Reduced sample contains 1 mM TCEP to avoid thiol oxidations during measurements. The endothermic peak for the E,E form is shown as a dotted line, and the solid line is for the zinc-bound form. E,E-hSOD1 SH (G93A) was measured in the presence of 20% glycerol. C, schematic representation of T m for various SOD1 forms in the presence of 20% glycerol (see also Table II). Difference of T m , ⌬T m mut , between ALS mutants and WT in the same state is plotted. D, A4V-WT; E, G93A-WT. hSOD1 SH (A4V) in the absence of glycerol. In the presence of 20% glycerol, which allows detection of T m (Table II), the oligomeric fractions in E,E-hSOD1 SH (A4V) were decreased and shifted to the distinct monomeric and dimeric states (supplemental Fig. S1). Even in the absence of glycerol, the extent of the E,E-hSOD1 SH (A4V) oligomerization is significantly reduced by either zinc binding, disulfide formation, or both (Fig.  3, C and D). Instead, these modifications of the A4V mutant protein favor an equilibrium mixture between the monomeric and dimeric states with correspondingly broad elution profiles. The A4V mutation, therefore, develops the propensity for oligomerization in the E,E-hSOD1 SH state most likely due to the accessibility of unfolded states at physiological temperatures.
The G93A mutant also exhibits a propensity for oligomerization in its unmodified polypeptide state. E,E-hSOD1 SH of G93A elutes at 13.8 ml, which corresponds to a significantly higher apparent molecular weight (ϳ55 kDa) than that of the SOD1 dimer (Fig. 3E). Addition of 20% glycerol can decrease the amount of the oligomeric forms, resulting in distinct monomerdimer mixtures (supplemental Fig. S1). As with the A4V protein, addition of zinc ion, disulfide formation, or both reduces the G93A oligomeric fractions, and the population of states shifts significantly to the dimer form (Fig. 3, E and F). Because little, if any, oligomer in the E,E-hSOD1 SH state is observed for WT protein, oligomerization of the most immature form of SOD1 is a conspicuous feature acquired by ALS mutations. Our results also reveal that the extent of the oligomerization is significantly diminished upon post-translational modifications that stabilize the SOD1 protein.
Mild Oxidants Multimerize E,E-hSOD1 SH via Formation of Inter-molecular Disulfides between Conserved Cysteines-The major intracellular reductant, GSH, is present in the cell with its oxidized species, GSSG, and the sum of their cytosolic concentrations is in the millimolar range for most cells. A mild oxidative stress can therefore be reconstituted by decreasing the GSH/GSSG ratio (37). The GSH/GSSG ratio of 100:1 is typical for the cytosol, whereas a GSH/GSSG ratio of 0.5 may be achieved in oxidative cellular compartments such as the endoplasmic reticulum (37). When 3 M E,E-hSOD1 SH is anaerobically incubated in the presence of GSH and GSSG, SOD1 bands in the high molecular region become intense with a decrease in the GSH/GSSG ratio from 100 to 0.5, regardless of WT or ALS mutants (Fig. 4, upper panel). To avoid thiol oxidations within the gel, all samples are treated with the thiol-specific modifier, iodoacetamide (IA), before sample loading on SDS-polyacrylamide gel. When samples are treated with the reductant ␤-ME before sample loading on a gel, these high molecular weight bands disappear (Fig. 4, lower panel), suggesting that these SOD1 multimers are stabilized through formation of nonspecific inter-molecular disulfide bonds. The dimer band can still be seen, but its fraction is significantly reduced after addition of ␤-ME. These disulfide-linked multimers as well as the oligomer seen in E,E-hSOD1 SH forms of ALS mutants do not bind Congo Red (data not shown), suggesting that the SOD1 oligomers/disulfide-linked multimers are distinct from amyloid-like states (29). The disulfide-linked multimers are insoluble in the absence of detergents such as SDS and Triton X-100 (data not shown), consistent with significant aggregation beyond that shown in Fig. 3. Glutathionylation of SOD1 can also be observed in the mass spectrum (supplemental Fig. S2), and these results strongly suggest that E,E-hSOD1 SH can readily form multiple disulfide bonds under mild oxidizing conditions.
Oxidative modification of the His and Tyr side chains of SOD1 proteins has been proposed, and such reactions may lead to covalent cross-links between SOD1s (38, 39); however, this was not the case in our current assays. The C 4 S hSOD1 mutant shows no multimer bands at any of the GSH/GSSG ratios with or without ␤-ME treatment (Fig. 5), consistent with the idea that these SOD1 polypeptides are linked via inter-molecular disulfide bonds. To test whether the conserved Cys residues, Cys 57 and Cys 146 , are more important in the oxidative multimerization than the other cysteines, we examined two additional Cys mutants, i.e. C6S/C111S and C57S/C146S. As seen in Fig.  5, the E,E-form of the reduced C6S/C111S mutant exhibits increasingly high molecular weight bands upon decreasing the GSH/GSSG ratio, indicating that disulfide-linked multimerization readily occurs through Cys 57 and Cys 146 . In contrast, we do not find SOD1 multimers after parallel treatment of the C57S/ C146S mutant (Fig. 5). These results suggest that formation of disulfide-linked multimers need not involve the nonconserved Cys residues, Cys 6 and Cys 111 . Instead, the conserved Cys residues, Cys 57 and Cys 146 , play an important role in the E,E-hSOD1 SH multimerization upon oxidative stress. Given that zinc ion significantly increases the stability of SOD1 (Fig. 2C), we next examined the possibility that zinc binding to E,E-hSOD1 SH has a protective role in the SOD1 disulfide-linked multimerization.
Zinc Binding Decreases the Extent of Disulfide-linked Multimerization in WT but Not in ALS Mutants-As seen in Fig. 6, formation of intermolecular disulfides is suppressed in E,Zn-hSOD1 SH (WT) relative to the zinc-free form (Fig. 4). Even under the most oxidizing conditions (0.5 of GSH/GSSG), a majority of WT protein migrate as a monomer. This result indicates that, under mild oxidative stress, E,Zn-hSOD1 SH (WT) preferen-tially forms intra-molecular as opposed to inter-molecular disulfide bonds. In ALS mutants, however, multimerization of E,Zn-hSOD1 SH is readily seen upon mild oxidation (Fig. 6). Zinc binding to the protein is thus an important event that increases stability of both mutant and WT SOD1; yet despite the zinc-induced increase in thermal stability, ALS mutants are significantly more susceptible than WT to forming multimers that are stabilized by incorrect disulfide formation.

DISCUSSION
Many of the disease-causing mutations in hSOD1 can give rise to significant amounts of active enzyme; however, other ALS mutations are completely inactive (1). A hallmark of ALS mutants is a dominant "gain of function" that is particularly toxic to motor neurons (9). Whereas cellular targets of the adventitious "function" are not yet clear, two general categories of models for the toxic physicochemical properties of the mature ALS mutants are discussed (1) as follows: (a) deleterious biochemical activities related to the metal cofactors, or (b) the propensity to form toxic aggregates. Here we address two issues of SOD1 stability and folding common to aggregate models. First, we identify the most unstable form of the mutant relative to the WT protein, and we show that it is also the form most prone to oligomerization. Second, we identify an attribute of SOD1 mutants that facilitates oxidative aggregation. The results here suggest that a toxic function acquired by ALS mutations may not be manifested as an attribute of the activated enzyme but of a much earlier stage in the maturation process.
Oligomer Formation in the Most Immature State of ALS Mutants-Most previous studies of hSOD1 stability employed mixtures of the holoprotein with protein in intermediate stages of maturation. In particular, the disulfide status has rarely been established; thus, little is known about the effects of ALS mutations on the disulfide-reduced protein. Biochemical methods for isolating and characterizing the distinct disulfide states for Saccharomyces cerevisiae SOD1 provide new tools to address these issues (19). As seen in Fig. 2C, ALS mutations in human SOD1 have a pronounced destabilizing effect on one form in particular, E,E-hSOD1 SH . Furthermore, we find that this state of ALS mutants is prone to forming oligomers at physiological concentrations where WT exists as a stable monomer (Fig. 3). Oligomerization of ALS mutants in the E,E-hSOD1 SH state is diminished in the presence of chemical chaperones such as glycerol and by the post-translational modifications. Taken together, the results show for the first time that the reduced apo-form of ALS

FIG. 4. Mild oxidation of E,E-hSOD1 SH leads to formation of disulfide-linked SOD1 multimers.
3 M E,E-hSOD1 SH was anaerobically incubated for an hour with 50 mM HEPES, pH 7.2, 0.1 mM EDTA, and 0.1 mM bathocuproine disulfonate in the presence of GSH and GSSG, the ratio of which is indicated in the figure. Total glutathione concentration was kept as 10 mM. After precipitation with 20% trichloroacetic acid, proteins were modified with 100 mM IA to protect thiol groups from aberrant oxidation during electrophoresis. 300 ng of proteins was loaded on 12.5% SDS-polyacrylamide gel without (upper panel) or with (lower panel) the reductant ␤-ME. After electrophoresis, the protein bands were detected with the Western blotting analysis. mutants is mostly unfolded and oligomerized under physiological conditions.
Disulfide-reduced Forms of SOD1 Are Susceptible to Oxidative Aggregation by Mild Physiological Oxidants-We have also identified another attribute of the ALS mutant that becomes apparent upon oxidative stress, namely the propensity to form detergent-soluble multimers cross-linked by intermolecular disulfides. SOD1 has an increased susceptibility to oxidative modification during import into the mitochondria (40) or in transport along the meter-long axon in the neuronal cell (41). Exposure of the most immature form of the mutant polypeptides to a mild oxidant, GSSG, leads to multimers containing intermolecular disulfides (Fig. 4). Most interestingly, this is not the only state susceptible to incorrect disulfide formation; ALS mutants in the zinc-loaded and disulfide-reduced state are much more susceptible to mild oxidative aggregation than their WT counterpart (Fig. 6). Thus, more than one of the intermediate states in the maturation pathway of ALS mutants are susceptible to oxidative aggregation. Disulfide bonds have long been recognized as a major factor for stabilizing folded states (42), but recently they have also been found to play a role in the toxicity of some diseaserelated proteins such as prion (43) and transthyretin (44).
Immature Reduced Protein: a Branch Point between Oligomerization and Maturation-Taken together, the results suggest that the toxic gain of function acquired by the ALS mutant proteins may not be an attribute of the activated enzyme but of much earlier stages in the post-translational pathway. A model for aggregation of the SOD1 polypeptide is shown in Fig. 7, where the unfolded E,E-hSOD1 SH form sits at the branch point between productive folding and aggregation pathways. The unmodified WT polypeptide could readily partition between these pathways even under cellular conditions that include extreme fluxes in SOD1 transcription/translation, reducing environment of the cytosol, or metal-cofactor limitation. Under most conditions, in contrast, this state of ALS mutant polypeptides is unfolded and most likely in equilibrium with oligomers in the cell. The ALS mutations clearly lead to a thermodynamic destabilization, but they may also diminish the folding rate, which is known to be affected in ␤-sheet-containing proteins (45). As proposed in the other human diseases (46,47), we speculate that the mixture of the various states of mutant SOD1 can be accommodated by the cellular quality control systems, including heat shock, copper chaperone proteins, and degradation processes, but only up to a point. Preexisting pools of ALS mutants in the holo-state, i.e. a mutant protein that has undergone post-translational modifications via the CCS-dependent or independent pathways (48), may be reduced to give more immature states under certain cellular stress, thus increasing the steady state concentrations of aggregation-prone forms. Once aggregate formation begins, other disulfide-reduced intermediate states in the maturation pathway such as E,Zn-hSOD1 SH may also be accumulated via adventitious disulfide formation.
Mutations in approximately two-thirds of the residues of SOD1 can cause ALS. These diverse mutations, however, do not cluster; they are distributed throughout the SOD1 structure. Our model wherein toxicity arises from conformational destabilization and unfolding of the unmodified proteins is compatible with this diverse set of mutations as long as the disease-causing mutations change the folding kinetics or thermodynamics of the immature protein. This type of misfolding/ oxidation mechanism is consistent with the previous in vivo studies that deletion of CCS does not modify the onset and progression of the ALS disease in the hSOD1-mouse model (49).
Impact of Oxidative Aggregation on Potential Targets of the Toxic ALS Mutants-The involvement of specific intracellular target(s) in the toxicity of the ALS mutant SOD1 proteins has not yet been established. Leading candidates are axonal transport systems (41), the proteasome (50,51), and the mitochondria (52). It is plausible that SOD1 aggregation retards the slow axonal transport of proteins, promoting motor axon degeneration (53). Unfolding/aggregation of the SOD1 mutants may also facilitate their ubiquitination, through which the proteolysis of the mutants and other proteins will overload proteasomes and inhibit their function (50,51).
The mitochondrial electron transport chain is a predominant source of reactive intracellular oxidants (54), and mitochondrial abnormalities such as vacuolation have been reported as an early pathological feature in several murine lines expressing ALS mutants (52). SOD1 has been shown to be imported into the intermembrane space of mitochondria (55,56), and only the most immature form, E,E-hSOD1 SH , is taken up into intermembrane space (57). Therefore, in the relatively oxidizing environment of the mitochondrial intermembrane space, unstable and oligomeric forms of the reduced ALS mutants would be subjected to adventitious oxidations that can form the disulfide-linked multimers. During mitochondrial import, ALS mutants in the oligomeric E,E-hSOD1 SH state may accumulate at the surface of the mitochondria or inhibit transport of other essential mitochondrial proteins, leading to damage and apoptotic cell death. This mechanism is consistent with recent findings that the ALS mutant is covalently associated with the cytoplasmic surface of the spinal cord mitochondria (40). Given that the spinal cord mitochondria can intrinsically produce higher levels of reactive oxygen species (58), ALS mutants near the spinal cord mitochondria have an increased chance of oxidative insult, which will facilitate the disulfide-linked multimerization and possibly the covalent attachment to crucial mitochondrial components such as an anti-apoptotic Bcl-2 family FIG. 6. Addition of zinc protects WT but not ALS mutant SOD1 from oxidative multimerization. Mild oxidation of disulfide-reduced apo-hSOD1 in the presence of ZnSO 4 was examined by Western blot. 3 M E,E-hSOD1 SH in 50 mM HEPES, pH 7.2, was anaerobically incubated with an equimolar amount of ZnSO 4 for an hour at 37°C, to which GSH and GSSG were then added. After incubation for an hour at 37°C, the proteins were precipitated with 20% trichloroacetic acid, modified with 25 mM AMS, and then loaded on 12.5% SDS-polyacrylamide gel without ␤-ME.
protein (59). These modifications would further lead to mitochondrial injury, release cell death mediators, and contribute to motor neuron cell death (60).
Therapeutic Implications-Our results show that maturation can provide significant protection against aggregation of SOD1 (Fig. 7). If the in vitro observations showing that zinc binding and disulfide formation deter oligomer formation and oxidative aggregation are borne out by further tests in vivo, there are significant therapeutic implications. Overexpression of proteins like CCS that are involved in SOD1 maturation, protein folding (e.g. HSP70), general disulfide formation (protein-disulfide isomerases), or expression of protein degradation machinery (proteasomes) should diminish the extent of aggregation. This approach of stimulating maturation is supported by other biochemical studies showing that an engineered disulfide bond between subunits that force SOD1 dimerization can reduce aggregation (61).
In addition, dietary modification that increases copper and zinc delivery to motor neurons, gene therapy, and/or drug-like molecules that lower SOD1 polypeptide levels or that facilitate the post-translational modifications of SOD1 could also be productive approaches to treating ALS. There are, however, two important caveats about the therapeutic applications concerning this (Fig. 7) and other aggregation models. First, if other models for the molecular origin of the disease, such as adventitious catalytic activity of the mature copper-loaded mutant enzyme, are instead borne out, the opposite therapeutic approach will be warranted. Second, it may be possible to rescue some, but not all, of the disease-causing mutations by stimulating the maturation pathway. Among the more than 100 ALS-associated hSOD1 mutants, seven cannot form intrinsically the essential intramolecular disulfide bond but are nonetheless capable of forming the intermolecular cross-links involved in oxidative aggregation. One of the conserved Cys residues, Cys 146 , is missing in the following seven mutants: Val 118 ins AAAAC (stop at 150), Leu 126 del TT (stop at 131), Leu 126 -STOP, Gly 127 ins TGGG (stop at 133), Glu 132 ins TT (stop at 133), Gly 141 -STOP, and Cys 146 3 Arg. It has been reported that minute quantities of SOD1 aggregates can cause the disease in the mice expressing the truncated mutant, Gly 127 ins TGGG (stop at 133) (62). Decreased affinity of Zn 2ϩ ion in ALS mutants has also been reported (12). The findings here thus suggest that the combined post-translational modifications of zinc binding and disulfide formation may have a protective role in preventing oligomerization and oxidative aggregation of SOD1 from some but not all mutations. In vivo tests of our current aggregation models are now under way, including characterization of the status of the metallation and thiol/disulfide in the SOD1 protein.