Dissociation of Human Copper-Zinc Superoxide Dismutase Dimers Using Chaotrope and Reductant INSIGHTS INTO THE MOLECULAR BASIS FOR DIMER STABILITY*

The dissociation of apo- and metal-bound human cop-per-zinc superoxide dismutase (SOD1) dimers induced by the chaotrope guanidine hydrochloride (GdnHCl) or the reductant Tris(2-carboxyethyl)phosphine (TCEP) has been analyzed using analytical ultracentrifugation. Global fitting of sedimentation equilibrium data under native solution conditions (without GdnHCl or TCEP) demonstrate that both the apo- and metal-bound forms of SOD1 are stable dimers. Sedimentation velocity experiments show that apo-SOD1 dimers dissociate coop-eratively over the range 0.5–1.0 M GdnHCl. In contrast, metal-bound SOD1 dimers possess a more compact shape and dissociate at significantly higher GdnHCl concentrations (2.0–3.0 M ). Reduction of the intrasubunit disulfide bond within each SOD1 subunit by 5–10 m M TCEP promotes dissociation of apo-SOD1 dimers, whereas the metal-bound enzyme remains a stable dimer under these conditions. The Cys-57

H 2 O 2 ) through redox cycling of its catalytic copper ion (5,6). The x-ray crystal structures of SOD1 from higher organisms are all very similar and reveal a homodimeric protein in which each monomeric subunit folds into an eight-stranded Greek key ␤-barrel, binds one copper and one zinc ion, and contains an intrasubunit disulfide bond (7)(8)(9)(10)(11).
The evolution of SOD1 proteins has been the subject of much discussion (12)(13)(14)(15)(16). The prokaryotic and eukaryotic enzymes demonstrate significant sequence similarity, and both are characterized by the Greek key ␤-barrel fold. Many bacterial SOD1 proteins, such as that from Photobacterium leiognathi, are also dimeric, although they often possess distinct dimer interfaces and have incorporated different strategies for the electrostatic recognition of the negatively charged superoxide substrate (12,14). In contrast, the SOD1 protein from Escherichia coli is monomeric (13,17) and possesses catalytic activity and heat stability comparable to the dimeric eukaryotic SOD1 proteins (17,18). The existence of both monomeric and dimeric Greek key ␤-barrel SOD1 proteins led to the suggestion that the prokaryotic and eukaryotic enzymes diverged from one ancient monomeric ancestor and then subsequently converged to their distinct dimeric enzymes (12). In terms of those SOD1 proteins that are dimers, the dimeric architecture appears necessary to maintain catalytic efficiency through the proper positioning of charged residues that guide the negatively charged superoxide substrate into the active site channel (12,15,16,19). This concept is supported by the fact that an engineered monomeric human SOD1 protein created by site-directed mutagenesis of interface residues possesses disorder in the regions responsible for electrostatic guidance resulting in a substantial decrease in catalytic activity (20). The dimeric architecture of eukaryotic SOD1 proteins could also arise from the need for these enzymes to acquire reactive copper from the copper chaperone for SOD1 (21). The copper chaperone for SOD1 contains an SOD1-like domain (domain 2) that facilitates recognition of its target, presumably through the formation of a copper chaperone-SOD1 complex (22,23).
The biochemistry and cell biology of the dimeric eukaryotic SOD1 proteins have been studied for over three decades. Recently, there has been intense renewed interest in the physical properties of the human enzyme because of its linkage to amyotrophic lateral sclerosis (ALS, motor neuron disease, Lou Gehrig's disease) (24,25), a fatal neurodegenerative disorder. A subset of inherited ALS (familial ALS, FALS) cases are caused by dominantly inherited mutations in SOD1. Accumulating evidence strongly suggests that the pathogenic mutations lead to aggregation of SOD1 proteins and that this propensity for selfassociation is somehow toxic to motor neurons (reviewed in Refs. 26 and 27). Indeed, high molecular weight, insoluble protein complexes containing FALS SOD1 are observed in spinal cord neurons of ALS patients, and in transgenic mice expressing these mutant proteins (28 -30). Nevertheless, the detailed composition of the FALS SOD1-containing aggregates formed in vivo is currently unknown, and it is possible that the aggregating species is a monomeric form of pathogenic SOD1, particularly because many of the FALS mutations fall directly at the dimer interface, at the intrasubunit disulfide bond, and in regions that affect metal binding. Thus, defining the molecular mechanisms and determinants that control dissociation of wild type and pathogenic human SOD1 dimers is of keen interest.
A wide range of studies have demonstrated qualitatively that both metal binding and the presence of an intact intrasubunit disulfide bond are contributors to the stability to the SOD1 homodimer (31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48). However, a rigorous quantitative analysis of the interrelationships between the molecular determinants of SOD1 dimer stability has yet to be described. In this report we demonstrate using analytical ultracentrifugation that both the apo-and metal-bound forms of SOD1 fall into the class of stable protein dimers that do not dissociate by mass action at the very low concentrations used in a sedimentation equilibrium experiment. Sedimentation velocity experiments in the presence of chaotrope or reductant allow characterization of the additional stability imparted by metal binding and the intrasubunit disulfide bond. The solution results have been combined with crystallographic data to provide a molecular explanation for the existence of different SOD1 macromolecular shapes and multiple SOD1 dimeric species with different stabilities.

EXPERIMENTAL PROCEDURES
Materials-Monobasic and dibasic potassium phosphate, acetonitrile (Optima grade), formic acid, sodium chloride, glacial acetic acid, ultrapure guanidine hydrochloride, sodium hydroxide, yeast extract, peptone, dextrose (glucose), and sodium acetate were obtained from Fischer Scientific. Tris(2-carboxyethyl)phosphine hydrochloride (TCEP-HCl) was purchased from Hampton Research. Ammonium sulfate and Tris base were purchased from U.S. Biologicals. EDTA was acquired from Mallinckrodt. Primers were purchased from Invitrogen. Agarose was obtained from Sigma. Pfu DNA polymerase and deoxynucleotides came from Stratagene. Glass beads were purchased from Biospec. All solutions unless otherwise noted in the text were prepared using de-ionized water passed through a Millipore ultrapurification system.
Protein Production, Metal Ion Removal, and Analysis of Metal Content-Purification methods are described in complete detail in Ref. 49. Site-directed mutagenesis of human wild type SOD1 was performed similarly to previously published methods (50,51), and the entire coding region of wild type and C57S SOD1 constructs was sequenced to verify the correct sequence. SOD1 proteins were expressed in the EGy118 strain of Saccharomyces cerevisiae, which lacks the endogenous yeast sod1 gene (52). SOD1 proteins expressed in this system are N-acetylated and have high metal content. After growth on selective medium, cells were cultured in ϳ30 liters of YPD medium (1% yeast extract, 2% peptone, and 2% dextrose) at 30°C for 36 h and harvested by centrifugation. The cells were resuspended in three volumes of chilled lysis buffer (200 mM Tris buffer, pH 8.0, 0.1 mM EDTA, 50 mM NaCl), and lysed using a blender with 0.5-mm glass beads. Because SOD1 loses its metal ions at low pH in the presence of chelators (6,53), the lysate was maintained at pH 7.0 during the lysis procedure by adjusting with 1 M sodium hydroxide when necessary. The lysate was centrifuged at 8000 ϫ g for 1 h at 4°C, and the supernatant was slowly brought to 60% ammonium sulfate saturation while stirring on ice. The mixture was stirred on ice for an additional 30 min and then centrifuged at 8000 ϫ g for 1 h at 4°C.
The supernatant (ϳ800 ml) was filtered through glass wool and applied to ϳ200 ml of fast flow phenyl-Sepharose (Amersham Biosciences) packed in a low pressure sealed column (Amersham Biosciences C26/40 column) pre-equilibrated with Buffer A (2.0 M ammonium sulfate, 150 mM NaCl, and 50 mM potassium phosphate, pH 7.0) using a peristaltic pump at a flow rate of ϳ8 ml/min. After loading, the column was washed with at least two column volumes of Buffer A and eluted with a step gradient of decreasing concentrations of ammonium sulfate (0.1 M every 45 ml, ϳ8 ml/min flow rate). Fractions were analyzed by SDS-PAGE, and those containing SOD1 were combined and concentrated. The protein was dialyzed into 2.25 mM potassium phosphate buffer, pH 7.0, containing 160 mM NaCl and subjected to gel filtration on a Sephadex G-75 superfine (Sigma) column pre-equilibrated with the same buffer. If required, some fractions were exchanged into 2.25 mM potassium phosphate buffer, pH 7.0, and subjected to an additional chromatographic step using DEAE-Sephadex A-50 (Sigma) or DEAE-cellulose (U.S. Biologicals) pre-equilibrated with 10 mM potassium phosphate buffer, pH 7.0. The proteins were eluted from these columns using a step gradient with increasing concentrations of potassium phosphate. The purity and correct molecular weight of the SOD1 proteins were verified by SDS-PAGE and electrospray ionization mass spectrometry using a PerkinElmer Life Sciences Sciex API III triple quadrupole mass spectrometer (Thornhill, Canada). Positive ion mass spectra were collected for all samples after dilution to ϳ20 pmol/l in 50:50:1 water/acetonitrile/formic acid (v/v). Data analysis was performed using BioMultiView version 1.3.1 (PerkinElmer Life Sciences, Sciex, Canada).
"Metal-bound" SOD1 was protein that was not subjected to the dialysis procedure described below and was used in these centrifugation experiments as purified. Apo-SOD1 proteins were generated by dialysis at low pH in the presence of EDTA using methods similar to those previously published (6). Spectra/Por 1 dialysis tubing with a molecular mass cutoff range of 6000 -8000 Da (Spectrum Laboratories Inc.) was rinsed with distilled water before adding the metal-bound protein. The protein solution was dialyzed five times at 4°C against 10 mM EDTA and 100 mM sodium acetate, pH 3.8, followed by dialysis three times against 100 mM NaCl and 100 mM sodium acetate, pH 3.8, to remove SOD1-bound EDTA (54). The protein was then dialyzed five times in 100 mM sodium acetate, pH 5.5, to remove NaCl. All dialysis buffers were filter sterilized using a 1000-ml Stericup Filter (Millipore), and those that did not contain EDTA were passed through a column containing ϳ150 ml of Chelex 100 resin (Bio-Rad) to remove trace metals. All glassware and equipment that came into contact with the apoprotein solutions were treated with a solution containing EDTA and rinsed with metal-free water.
Inductively coupled plasma, atomic emission mass spectrometry was performed at the Elemental Analysis Facility (UCLA, Department of Chemistry and Biochemistry) using a Thermo Jarrel Ash (Thermo Electron) IRIS 1000 inductively coupled plasma, atomic emission instrument to quantify metal ions that were present and to confirm that apo-SOD1 was metal-free. Three readings were taken per wavelength (three wavelengths per element) and averaged. Metal ion concentrations were converted from parts per billion to molarity and compared with protein concentration to calculate equivalents of metal ions bound per SOD1 dimer. Metal-bound wild type SOD1 in these studies contained, on average, ϳ2.9 equivalents of zinc and ϳ0.6 equivalents of copper per dimer, whereas the C57S mutant contained ϳ2.53 equivalents of zinc and ϳ0.13 equivalents of copper per dimer. All apo-SOD1 proteins contained Ͻ0.05 equivalents of both copper and zinc.
Analytical Ultracentrifugation-All sedimentation velocity and sedimentation equilibrium experiments were performed in a Beckman Optima XL-I at the Center for Analytical Ultracentrifugation of Macromolecular Assemblies (The University of Texas Health Science Center at San Antonio, Department of Biochemistry). Buffer density and viscosity corrections were made according to data published by Laue et al. (55) as implemented in UltraScan version 6.2 software (56). The partial specific volume of the SOD1 holoenzyme estimated from the protein sequence according to the method by Cohn and Edsall (57) was 0.72485 cm 3 /g. All solutions were made metal-free using a batch preparation with Chelex 100 resin and filtered using a Steriflip Filter system (Millipore Corp.). TCEP was dissolved in metal-free buffer and adjusted to pH 5.5 using sodium hydroxide. All SOD1 samples were prepared at the same time before each run by diluting a ϳ10 mg/ml (ϳ300 M) protein solution to the specified concentrations.
Sedimentation velocity runs were performed at 20°C using an eighthole AN-50 rotor and double-sector charcoal/Epon filled centerpieces. SOD1 samples with an initial absorbance at 280 nm of 0.8 (ϳ2.4 mg/ml; ϳ74 M) were centrifuged at 50,000 rpm. Scans were collected using absorbance optics and analyzed by the method of van Holde and Weischet (58) as implemented in UltraScan (56). This analysis yields the integral distribution of diffusion-corrected sedimentation coefficients (59,60). SOD1 monomer and dimer sedimentation coefficients were predicted with the HydroPro software (61) based on the crystal structure of human SOD1 (pdb code 1HL5 (62)).
Sedimentation equilibrium experiments were performed at 20°C using a four-hole AN-60 rotor and double-sector charcoal/Epon-filled centerpieces. Three 120-l samples of protein with an initial absorbance at 280 nm of 0.3, 0.4, and 0.5 were sedimented to equilibrium at 24,000, 27,500 and 31,000 rpm. Scans were collected with absorbance optics at 230 and 280 nm. The radial step size was 0.001 cm, and each c versus r datapoint was the average of 20 independent measurements. Points exceeding 0.9 A were excluded from the fits. Data spanning the concentration range of 2-100 M SOD1 were examined by global fitting and Monte Carlo analyses using UltraScan. Global fitting such as that employed here enhances the confidence in each fitted parameter value (63). Equilibrium data were fit to multiple models. The most appropriate model was chosen based on visual inspection of the residual run patterns, and on the best statistics. 95% confidence intervals were determined by Monte Carlo analysis.

Analytical Ultracentrifugation of the Apo-and Metal-bound
Forms of Native Human SOD1-In the present study, the structure and stability of the apo-and metal-bound forms of human SOD1 have been characterized rigorously in the analytical ultracentrifuge using both sedimentation velocity and equilibrium approaches. Sedimentation velocity runs initially were performed under native conditions (100 mM acetate buffer, pH 5.5) without chaotrope or reductant. Velocity scans were analyzed by the method of van Holde and Weischet (58). This analysis method effectively removes the contributions of diffusion to boundary spreading to yield the integral distribution of s 20,w of all species in the sample, G(s). Consequently, a G(s) plot of boundary fraction versus s 20,w will be vertical if the sample is homogeneous, and will have a positive slope if the sample is heterogeneous (59,66). The G(s) plots in Fig. 1 demonstrate that under regular buffer conditions apo-SOD1 (filled circles) sedimented as a homogeneous ϳ2.6 S species, while metal-bound SOD1 (open circles) sedimented as a single ϳ3.1 S species. Prediction of the dimer s 20,w using the crystal structure coordinates (pdb code 1HL5 (62)) and hydrodynamic bead modeling (61) yielded a value of 2.9 S, strongly suggesting that the apo-and metal-bound SOD1 proteins were both stable dimers.
To characterize directly the quaternary structure of native apo-and metal-bound human SOD1, we performed sedimentation equilibrium experiments and analyzed the data by global fitting approaches. Twelve individual datasets resulting from multiple speeds and multiple loading concentrations (Figs. 2 and 3) were globally fit to three models (single component, monomer-dimer equilibrium, two ideal non-interacting components) using the UltraScan program (56). The goodness of fit was judged using a combination of the fit statistics (Figs. 2 and 3, legends) and the randomness of the residuals (Figs. 2 and 3,  upper plots).
The sedimentation equilibrium data for the apo-and metal-bound SOD1 samples in both cases were best fit by a noninteracting, two-component model. The major SOD1 component in each sample had a molecular mass of 32.5 kDa, equivalent to an SOD1 homodimer (31.6 kDa calculated). The mass of the minor component was 24 -28 kDa. PAGE confirmed that the minor, lower molecular weight species was present in the sedimentation equilibrium samples (data not shown). This minor species likely represents a proteolytic product that appears during the longer time course of the sedimentation equilibrium experiment, because this species was not detected in the sedimentation velocity runs or by PAGE run on the same sedimentation velocity samples immediately after the centrifugation experiment.
The sedimentation equilibrium results demonstrate that both the apo-and metal-bound forms of SOD1 are stable dimers that do not fractionally dissociate into monomers under even the most dilute concentrations used in the sedimentation equilibrium experiments (2 M). This is indicative of a dimerization K d of ϳ10 Ϫ8 M or lower, even in the absence of bound metals. An additional important conclusion from these results is that metal binding results in a more compact SOD1 dimer, i.e. the difference in s 20,w between the 2.6 S apo-dimer and 3.1 S metal-bound SOD1 dimer is due to a difference in shape rather than mass.
Guanidine Hydrochloride-induced Dissociation of SOD1 Dimers-Because both the apo-and metal-bound forms of human SOD1 behave as stable dimers in sedimentation velocity and equilibrium experiments under native conditions, the chaotrope guanidine hydrochloride (GdnHCl) was used to induce subunit dissociation (33,34,36,(42)(43)(44)67). Fig. 4A shows the effects of increasing concentrations of GdnHCl on the dissociation of the 2.6 S apo-SOD1 dimer, as assayed by sedimentation velocity. Because GdnHCl is a denaturant at high concentrations, care was taken to mix solutions well to minimize local concentration effects. The native apo-dimer dissociated into a 1.4 S monomeric species over the range of 0.25-1.25 M GdnHCl. Bead modeling predicts a value of 1.8 S for the monomeric subunit in the crystallographic dimer. The G(s) plots in 0.5-1.0 M range were heterogeneous and had shapes characteristic of a self-associating system (59). This would be expected if the dimer was sufficiently destabilized in 0. The G(s) plots of metal-bound SOD1 in GdnHCl are shown in Fig. 4B. The plots were similar to those for the apo-protein (Fig.  4A), with two important differences. First, dissociation of the native dimer occurred between 2.0 -3.0 M GdnHCl. Second, the metal-bound SOD1 sample sedimented as a homogeneous 1.8 S species in 3.0 -5.0 M GdnHCl, and as a 1.4 S species in concentrations greater than or equal to 5.0 M (Fig. 4B, inset).
The weight-average s 20,w of apo-and metal-bound SOD1 were plotted against GdnHCl concentration (Fig. 4C) to illustrate more clearly the effects of metal binding on SOD1 macromolecular shape and dimer stability. The large difference in shape between the native apo-and metal-bound SOD1 dimers is clearly apparent (Fig. 4C, region I). Both with and without bound metals, SOD1 dimer dissociation appeared to be cooperative with increasing GdnHCl. However, more GdnHCl (Ͼϳ1 M) was required to induce equivalent states of dissociation of the metal-bound SOD1 dimers compared with the apo-dimer (Fig. 4C, region II). Finally, GdnHCl-dependent dissociation of the metal-bound SOD1 dimer initially produced a 1.8 S monomer that was converted to a 1.4 S species at higher GdnHCl concentrations. In contrast, dissociation of the apo-SOD1 dimer produced only the 1.4 S species (Fig. 4C, region III).
Dissociation of the SOD1 Dimer Induced by Intrasubunit Disulfide Bond Reduction-To probe the contribution of the intrasubunit disulfide bond formed between Cys-57 and Cys-146, we incubated apo-and metal-bound human SOD1 dimers with increasing concentrations of the optically neutral reductant, TCEP (Tris(2-carboxyethyl)phosphine). In these experiments the native apo-SOD1 dimer sedimented as a homogeneous ϳ2.7 S species, and began to dissociate in 5 mM TCEP (Fig.  5). At 5-10 mM TCEP, the shapes of the G(s) plots were heterogeneous and characteristic of a self-associating system (59), whereas at 40 mM TCEP nearly all the apo-SOD1 sedimented as a ϳ1.7 S monomer. Curves of the percent disulfide reduction (determined by mass spectrometry) plotted against TCEP concentration closely parallel the sedimentation velocity data (data not shown). In contrast to the apo-protein, the metalbound SOD1 dimer remained intact in 100 mM TCEP (data not shown). Thus, as with chaotrope, the one or more structural changes due to metal binding also stabilize the dimer against reductant-induced dissociation.
The involvement of the intrasubunit disulfide bond was examined further by characterizing the hydrodynamic properties of C57S SOD1 mutant, which is incapable of forming the disulfide bond. Fig. 6 shows the G(s) plots of the apo-and metalbound forms of C57S SOD1 under buffer conditions where chaotrope and reductant were not present. The metal-bound C57S sample sedimented as a homogeneous dimer of ϳ2.8 S. In contrast, the apo-C57S protein sedimented as a homogeneous monomer of ϳ1.8 S. These data unequivocally demonstrate that metal binding is able to stabilize the SOD1 dimer independently of intrasubunit disulfide bond formation.

Identification of Five Distinct SOD1 Species-A combined
G(s) plot summarizing the multiple SOD1 species identified in our studies is shown in Fig. 7. SOD1 dimers span the range of ϳ2.6 -3.1 S and have different macromolecular shapes depending on their metal content and the presence of the intrasubunit disulfide bond. The ϳ3.1 S metal-bound SOD1 dimer structure contains an intact disulfide bond and well ordered zinc and electrostatic loop elements (see below) and is the most compact. The ϳ2.8 S species in Fig. 7 is the metal-bound form of C57S.
We propose that the ϳ10% decrease in S value relative to the metal-bound wild type protein is due to increased flexibility of the disulfide loop that results from the loss of the Cys-57 to Cys-146 disulfide bond. It is also possible that this difference could be due to the fact that C57S contains slightly less metal ions than does the wild type protein (see "Experimental Procedures"). The apo-dimer, which contains an intact disulfide bond but no metals, sediments at only ϳ2.6 S. This is a large ϳ20% decrease in S value relative to the metal-bound wild type enzyme. The difference in sedimentation behavior of apo-SOD1 and metal-bound SOD1 is likely due to a general "tightening" of the overall shape of the molecule caused by the ordering of the electrostatic and zinc loop elements upon metal binding (see Fig. 8, A and B). In support of this, recent crystallographic analyses reveal that the electrostatic and zinc loops become disordered when metals ions are lost (62,68). From these data it is evident that the absence of metals generates greater frictional drag on the protein then does absence of the disulfide bond in the metal-bound protein. Two types of monomers were observed, those that sediment near 1.8 S and those that sediment near 1.4 S. The ϳ1.8 S group includes metal-bound SOD1 in 3.0 -5.0 M GdnHCl, apo-SOD1 monomers dissociated by TCEP, and apo-C57S SOD1 monomers in buffer alone. The 1.8 S monomer has the same S value as that predicted from the crystal structure of metalbound SOD1, and presumably corresponds to the folded structure similar to that seen in the crystal structure. Indeed, x-ray and NMR studies have shown that both metal-free and metalbound engineered monomeric SOD1 (F50E/G51E) proteins possess a three-dimensional fold very similar to that observed in the dimeric form of the enzyme (15, 69, 70). Because metal binding maintains the SOD1 monomer in the 1.8 S conformation in 3.0 -5.0 M GdnHCl, these concentrations of GdnHCl are presumed to cause only disruption across the dimer interface in the metal-bound protein, leading to dissociation into folded, metal-bound monomers (see "flattened" region III for the met- al-bound protein in Fig. 4C). Higher concentrations of GdnHCl lead to formation of the same monomeric 1.4 S species observed upon dissociation of the apo-protein. Hence, we conclude that the 1.8 to 1.4 S transition results from unfolding of the monomer in the absence of metals, and that GdnHCl at Ն5 M leads to metal ion loss.
There is no chaotrope or reducing agent present in the C57S sample. We conclude that the similar sedimentation behavior of the apo-wild type protein in the presence of small amounts of TCEP and the apo-C57S protein is due to their structural similarity, because both are incapable of forming the intrasubunit disulfide bond. TCEP is an excellent disulfide reducing agent that has advantages over other reducing agents such as dithiothreitol (71)(72)(73). For example, it is optically neutral at 280 nm, can reduce disulfide bonds in most proteins, and is stable over a wide range of pH values, including those where dithiothreitol is no longer effective. Intriguingly, the protein concentration in these velocity experiments is ϳ74 M, and 40 mM TCEP does not result in complete reduction of the disulfide bond. A recent study on the yeast SOD1 enzyme strongly suggests that, once formed, the intrasubunit disulfide bond is kinetically stable, even in the presence of excess reductants (74).
Structural Basis for SOD1 Dimer Stability-X-ray crystallography reveals the overall three-dimensional architecture of dimeric human SOD1 as shown in Fig. 8A. The robust SOD1 dimer interface buries ϳ640 Å 2 of solvent-accessible surface area per polypeptide and includes numerous main-chain to main-chain hydrogen bonds, apolar interactions, and watermediated hydrogen bonds. The dimer interface is formed by reciprocal contacts between the C-terminal ␤-strand (strand 8, purple in Fig. 8A) and residues from the disulfide loop (also purple in Fig. 8A) from each monomer. Reduction of the disulfide bond (or its abrogation via the C57S mutation) is envisioned to result in increased mobility of both the disulfide loop and ␤-strand 8 in each protomer, which in turn is predicted to weaken their reciprocal interactions across the dimer interface. An important conclusion from our studies is that the disulfideintact, human SOD1 dimer is unusually stable even in the absence of bound metal ions.
As shown in Fig. 8B, copper binding has three main structural effects: 1) it stabilizes the ␤-barrel and the C-terminal ␤-strand of each monomer by anchoring copper ligands His-46, His-48, and His-120; 2) it helps to anchor the disulfide loop against the ␤-barrel through His-48 and His-63; and 3) it ties the electro- FIG. 8. a, the structural basis for the stability of the human SOD1 dimer (pdb code 1HL5 (62)). The relationship between the two subunits is indicated. Each subunit of the protein binds one copper ion (blue), binds one zinc ion (green), contains one disulfide bond (yellow), and possesses an eight-stranded Greek key ␤-barrel (gray) topology. Protruding from each ␤-barrel are two prominent loop elements termed the "electrostatic" (red) and "zinc" (green) loops. The electrostatic loop (loop VII, residues 121-143) forms part of the active site channel and contains charged amino acids that help steer the negatively charged superoxide anion toward the catalytic copper, enhancing the rate of the disproportionation reaction (76 static loop to the disulfide loop and ␤-barrel through hydrogen bonds between the non-liganding nitrogen of His-48 and the carbonyl oxygen of Gly-61 and the non-liganding nitrogen of His-120 and the carbonyl oxygen of Gly-141, respectively. Fig. 8B also shows that the link between the electrostatic and the disulfide loops is strengthened through the side chain of Arg-143, that forms two hydrogen bonds with the carbonyl oxygen of Gly-61 and one with the carbonyl oxygen of Cys-57. His-63 binds copper and zinc simultaneously in the Cu(II) form of the enzyme, linking the two metal binding sites to the disulfide loop. Zinc binding to His-71, His-80, and Asp-83 strongly influences the conformation of the zinc loop, which in turn interacts intimately with the electrostatic loop. Asp-124 of the electrostatic loop directly links the copper and zinc binding sites by forming hydrogen bonds simultaneously with copper ligand His-46 and zinc ligand His-71. Thus, metal binding plays an important structural role by organizing and stabilizing the above-mentioned hydrogen bonding networks stabilizing the monomer, which in turn stabilizes the SOD1 homodimer.
Implications for SOD1-linked Familial Amyotrophic Lateral Sclerosis-Since the link between mutations in copper-zinc superoxide dismutase (SOD1) and familial ALS (FALS) was first described ϳ10 years ago, laboratories worldwide have sought to understand how the mutations render the SOD1 protein toxic to motor neurons. Evidence is accumulating that this toxic property comes from the ability of the mutant SOD1 proteins to assemble into higher order structures, either soluble oligomers or insoluble aggregates, that somehow interfere with the neuronal cellular machinery. A model was proposed recently for pathogenic SOD1 aggregation where a metal-deficient, partially misfolded dimer acts as the building block for the formation of amyloid-like and helical filamentous arrays (68). Because the detailed composition of the FALS-containing aggregates formed in vivo is currently unknown, it is possible that the aggregating species is the monomeric form of pathogenic SOD1. This concept is supported by recent studies that suggest that oxidized, monomeric pathogenic human SOD1 is on the pathway to aggregation (75). The potential importance of FALS mutations to the dissociation behavior of SOD1 is shown in Fig. 8C. Pathogenic mutations that fall at the dimer interface and are predicted to directly affect dimer stability are shown as black spheres. Two such metal-bound pathogenic mutants, A4V and I113T, have been shown by low angle x-ray scattering experiments to possess an altered, more tenuous selfassociation in solution relative to the wild type protein (50).
Pathogenic mutations that fall in the "metal-binding region" are shown as yellow spheres. Members of this class are known to be metal-deficient to varying degrees (compared with the wild type enzyme) when the proteins are isolated from their expression system and are therefore inherently more prone to monomerization (this study and Refs. 27, 46, and 68). Finally, some pathogenic SOD1 mutants are more prone to reduction or loss of the intrasubunit disulfide bond. Hayward and colleagues (48) demonstrated using partially denaturing gels that the disulfide bond in many FALS mutants appears more susceptible to reduction, and the mutants are more susceptible to monomerization than the wild type protein. Indeed, the C146R pathogenic mutant of human SOD1 is predicted to behave similarly to C57S, because both mutations abrogate the intrasubunit disulfide bond. It therefore will be intriguing to compare the dissociation behavior of these pathogenic proteins to that of the wild type enzyme in the analytical ultracentrifuge where the molecules can be studied in solution free from interaction with polyacrylamide gels or gel filtration resins.