Unsaturated Fatty Acids Induce Cytotoxic Aggregate Formation of Amyotrophic Lateral Sclerosis-linked Superoxide Dismutase 1 Mutants*

Formation of misfolded protein aggregates is a remarkable hallmark of various neurodegenerative diseases including Alzheimer disease, Parkinson disease, Huntington disease, prion encephalopathies, and amyotrophic lateral sclerosis (ALS). Superoxide dismutase 1 (SOD1) immunoreactive inclusions have been found in the spinal cord of ALS animal models and patients, implicating the close involvement of SOD1 aggregates in ALS pathogenesis. Here we examined the molecular mechanism of aggregate formation of ALS-related SOD1 mutants in vitro. We found that long-chain unsaturated fatty acids (FAs) promoted aggregate formation of SOD1 mutants in both dose- and time-dependent manners. Metal-deficient SOD1s, wild-type, and mutants were highly oligomerized compared with holo-SOD1s by incubation in the presence of unsaturated FAs. Oligomerization of SOD1 is closely associated with its structural instability. Heat-treated holo-SOD1 mutants were readily oligomerized by the addition of unsaturated FAs, whereas wild-type SOD1 was not. The monounsaturated FA, oleic acid, directly bound to SOD1 and was characterized by a solid-phase FA binding assay using oleate-Sepharose. The FA binding characteristics were closely correlated with the oligomerization propensity of SOD1 proteins, which indicates that FA binding may change SOD1 conformation in a way that favors the formation of aggregates. High molecular mass aggregates of SOD1 induced by FAs have a granular morphology and show significant cytotoxicity. These findings suggest that SOD1 mutants gain FA binding abilities based on their structural instability and form cytotoxic granular aggregates.

Formation of misfolded protein aggregates is a remarkable hallmark of various neurodegenerative diseases including Alzheimer disease, Parkinson disease, Huntington disease, prion encephalopathies, and amyotrophic lateral sclerosis (ALS). Superoxide dismutase 1 (SOD1) immunoreactive inclusions have been found in the spinal cord of ALS animal models and patients, implicating the close involvement of SOD1 aggregates in ALS pathogenesis. Here we examined the molecular mechanism of aggregate formation of ALS-related SOD1 mutants in vitro. We found that long-chain unsaturated fatty acids (FAs) promoted aggregate formation of SOD1 mutants in both dose-and time-dependent manners. Metal-deficient SOD1s, wild-type, and mutants were highly oligomerized compared with holo-SOD1s by incubation in the presence of unsaturated FAs. Oligomerization of SOD1 is closely associated with its structural instability. Heat-treated holo-SOD1 mutants were readily oligomerized by the addition of unsaturated FAs, whereas wild-type SOD1 was not. The monounsaturated FA, oleic acid, directly bound to SOD1 and was characterized by a solid-phase FA binding assay using oleate-Sepharose. The FA binding characteristics were closely correlated with the oligomerization propensity of SOD1 proteins, which indicates that FA binding may change SOD1 conformation in a way that favors the formation of aggregates. High molecular mass aggregates of SOD1 induced by FAs have a granular morphology and show significant cytotoxicity. These findings suggest that SOD1 mutants gain FA binding abilities based on their structural instability and form cytotoxic granular aggregates.
Amyotrophic lateral sclerosis (ALS) 1 is a progressive and fatal neurodegenerative disorder that mainly affects motor neurons in the brain stem and spinal cord. Approximately 10% of ALS patients are familial cases, with autosomal dominant inheritance. More than 90 different mutations in the gene coding for superoxide dismutase 1 (SOD1) have been identified in about 20% of familial ALS (FALS) families (1,2). Although the molecular mechanisms of selective motor neuron degeneration by SOD1 mutants in FALS remain largely unknown, common pathological features of conformational diseases, as evidenced by SOD1 immunoreactive inclusions, are found in the spinal cord of ALS patients and in the SOD1 mutant FALS mouse model (3)(4)(5)(6)(7)(8). The characteristics of FALS resemble those of many other neurodegenerative diseases in which a causative protein undergoes a conformational rearrangement, which endows it with a tendency to aggregate and form deposits within affected tissues.
SOD1 is a 32-kDa homodimeric enzyme that decreases the intracellular concentration of superoxide radicals by catalyzing their dismutation to O 2 and H 2 O 2 . ALS-linked mutations of SOD1 are distributed throughout the primary and tertiary structures, and most mutations appear unrelated to the dismutase activity. Many biochemical and biophysical studies have reported that SOD1 mutants are structurally unstable compared with wild-type forms (10 -13). These observations suggest that the mutations primarily affect the structural stability of SOD1 rather than the enzyme activity.
The half-life of SOD1 mutants is shorter than that of wildtype forms in cultured cells (14). SOD1 mutants form a complex with Hsp70 and CHIP, which promotes degradation of SOD1 through the ubiquitin-proteasome system (15). Hsp70 directly binds metal-deficient wild-type SOD1 as well as SOD1 mutants, suggesting that destabilized SOD1 is targeted by the molecular chaperone system (15,16). These observations imply that structural stability of SOD1 may also be strongly involved in refolding by the chaperone or in degradation of SOD1 by the ubiquitin-proteasome system. On the other hand, aggregates of mutant SOD1 are observed to have aggresome-like morphology when cells are treated with a proteasome inhibitor (14). This aggresome-like morphology resembles pathological inclusions of neurodegenerative diseases in affected tissues. These findings suggest that in disease states, misfolded proteins overwhelm the protein handling systems, including chaperones and proteasomes. Therefore, the formation of cellular inclusions may be required for other factors to act as modulators to promote protein aggregates. In fact, lipid molecules, including unsaturated fatty acids (FAs), phosphatidylserine, and phosphatidylinositol, promote amyloidogenesis of amyloid ␤-peptides, tau (17), and ␣-synuclein (18,19) in vitro. These molecules are biologically significant as mediators for signal-ing and inflammation during disease progression of neurodegeneration.
Here we investigated in vitro SOD1 aggregation affected by FAs to create an aggregation model system for FALS. We demonstrated that unsaturated FAs promote self-assembly and cytotoxic aggregate formation of SOD1. Aggregation by FAs is strongly correlated with structural instability and FA binding activity of SOD1, which may have significant implications in FALS.

EXPERIMENTAL PROCEDURES
Expression, Purification, and Characterization of Recombinant SOD1 Proteins-pcDNA3-SOD1 (20) was digested with EcoRV and XhoI and subcloned into blunted NcoI and XhoI sites of pET-15(b) (Novagen) to construct the expression plasmid. Expression of recombinant SOD1 proteins was induced in BL21(DE3)pLysS by adding 1 mM isopropyl 1-thio-␤-D-galactopyranoside, 0.1 mM CuCl 2 , and 0.1 mM ZnCl 2 until cells were grown to 0.6 absorbance unit at 600 nm, and then bacterial cells were further cultured at 23°C for 6 h. Cells were pelleted and resuspended in TNE buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 0.1 mM EDTA) supplemented with protease inhibitor mixture (Roche Applied Science). Cells were then disrupted by sonication. Insoluble materials were removed by centrifugation at 10,000 ϫ g for 30 min. Supernatant was collected for further purification. Purification was performed according to Hayward et al. (11), with minor modifications. Briefly, ammonium sulfate powder was added to the supernatant to 65% saturation with gentle stirring on ice. The supernatant, after centrifugation at 10,000 ϫ g for 30 min, was directly loaded for phenyl-Sepharose (Amersham Biosciences) column chromatography. The column was thoroughly washed with TNE buffer containing 2 M ammonium sulfate. Proteins were eluted using a linearly decreasing salt gradient. SOD1 activity measurement using a xanthine/xanthine oxidase-based method (21) identified fractions containing SOD1. Activity fractions were desalted by ultrafiltration using a centricon filter (Millipore). SOD1 was re-metallated as described previously (22). The proteins were then loaded onto a Q-Sepharose (Amersham Biosciences) anion exchange column and eluted using a linearly increasing salt gradient toward a buffer containing 200 mM NaCl and 10 mM Tris-HCl, pH 8.0. Fractions containing SOD1 were pooled and concentrated. Homogeneity of SOD1 was Ͼ95%, as verified by SDS-PAGE with Coomassie Brilliant Blue staining. Specific activity of the purified enzymes was assayed and calculated by bovine SOD1 (Cayman) or human SOD1 purified from erythrocytes (Sigma-Aldrich) as standards. Fully metallated SOD1 was delipidated using hydroxyalkoxypropyl dextran type III (Sigma-Aldrich) as described previously (19) before de-metallation. Metal-deficient apo-enzymes were prepared as described previously (23), and loss of enzyme activity was confirmed after de-metallation. The metal content of purified enzymes was estimated as described previously (22).
Oligomerization of SOD1-A stock solution of 25 mM FAs was prepared in 0.01 M NaOH containing 25 M butylated hydroxytoluene. SOD1 proteins were filtered through Microcon YM-100 (100-kDa cutoff) filters (Millipore) to remove high molecular mass SOD1 before oligomerization. FAs were added directly to preincubated SOD1 at 37°C in 50 mM phosphate buffer, pH 7.2, containing 150 mM NaCl and 0.1 mM EDTA and further incubated at the same temperature.

SDS-PAGE and Western
Blotting-For detection of SOD1 oligomers, SDS-PAGE was performed under non-reducing conditions using 12.5% polyacrylamide gels. After oligomerization of SOD1, protein samples were prepared in SDS-PAGE loading buffer (62.5 mM Tris-HCl, pH 6.8, 1% SDS, 5% glycerol, and 0.007% bromphenol blue) in the absence of ␤-mercaptoethanol and then boiled at 95°C for 3 min before loading. Western blotting was performed as described previously (24). Briefly, proteins were transferred to Hybond ECL nitrocellulose membranes (Amersham Biosciences), followed by UV cross-linking, boiling membranes in 2% SDS and 50 mM Tris, pH 7.6, for 10 min, and extensive washing in Tris-buffered saline. For detection of SOD1, rabbit anti-SOD1 antibody (Stressgen) was used.
Glycerol Density Gradient Centrifugation and Densitometric Analysis-A glycerol linear gradient of 10 -40% was prepared in a centrifuge tube. Formation of the SOD1 oligomer was performed as described above. Approximately 200 l of incubated SOD1 was layered onto a glycerol cushion and separated by centrifugation with a SW41Ti rotor (Beckman) at 35,000 rpm for 15 h. In a parallel experiment, protein standards (Amersham Biosciences) were separated simultaneously in order to calibrate fractions. Fractions were subjected to SDS-PAGE under non-reducing conditions, and then Western blotting was performed. Western blot images were analyzed using image analysis software (Scion Image Beta 4.02; Scion Corp.).
Solid-phase Oleic Acid Binding-Sodium salt of oleic acid (Sigma-Aldrich) was coupled to EAH-Sepharose (Amersham Biosciences) by 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide (Pierce) to prepare oleate-Sepharose according to Peters et al. (25). Oleic acid coupling was verified by binding bovine serum albumin and recombinant ␣-synuclein protein. Mock-Sepharose was prepared from EAH-Sepharose by blocking coupling ligand with 1 M acetic acid. For the binding assay, 200 ng of Microcon-filtered protein was incubated with oleate-Sepharose or mock-Sepharose in 400 l of phosphate-buffered saline containing 0.1 mM EDTA at 37°C for 30 min with agitation. Protein bound to Sepharose was settled on a spun column and washed extensively with phosphate-buffered saline. The bound protein was then eluted with 50% ethanol. Eluates were subjected to SDS-PAGE and Western blotting.
Transmission Electron Microscopy-SOD1 proteins (40 M) were incubated at 37°C for 24 h in 50 mM phosphate buffer (pH 7.2) containing 150 mM NaCl and 0.1 mM EDTA supplemented with 100 M arachidonic acid. The samples were absorbed to a glow-charged supporting membrane on 400-mesh grids and fixed by floating on 2.5% glutaraldehyde and 4% paraformaldehyde in 0.1 M phosphate buffer for 5 min. After three washes with distilled water, samples were negatively stained by 2% sodium phosphotungstic acid and dried. Specimens were observed in a LEO 912AB electron microscope (LEO Electron Microscopy), operated at 100 kV.
Toxicity Assay-Cytotoxicity of protein aggregates was measured as described previously (26,27). In brief, neuro2a mouse neuroblastoma cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum and 2 mM glutamine in 5% CO 2 at 37°C. Cells were differentiated in serum-free Dulbecco's modified Eagle's medium with 0.3 mM dibutylyl cAMP before use. Cells were plated at 30,000 cells/well in 96-well plates and differentiated overnight. The medium was removed, and prepared SOD1 aggregates were added in new medium without phenol red. After incubation for 18 h at 37°C, the cells were assayed using an MTS reduction assay kit (Promega). Another plate also treated as described above was stained for 1 min with trypan blue, and stained cells were counted as dead cells.

RESULTS
Unsaturated Fatty Acids Promote Self-assembly of SOD1s-We expressed and homogeneously purified recombinant human SOD1s from the bacterial expression system (Fig. 1A). The purified wild-type and G93A enzymes showed comparable specific activity; however, A4V mutant showed ϳ56% activity compared with that of wild-type enzyme (Fig. 1B). The zinc ion content of the purified enzymes showed almost full occupancy; however, copper ion content of A4V was 54.5% of the wild-type level (Fig. 1C). Specific activity was correlated with copper ion occupancy of purified enzyme, indicating proper metal loading in the active site.
We next examined the effect of long-chain FAs on oligomerization of SOD1 proteins. Wild-type and mutant (A4V and G93A) SOD1 were incubated with various concentrations of arachidonic acid (AA) as described under "Experimental Procedures." After incubation, oligomerized SOD1 was subjected to SDS-PAGE and then detected by Western blotting. Under reducing conditions, mainly bands of ϳ16 and 38 kDa, corresponding to monomer and dimmer sizes of SOD1, respectively, were detected ( Fig. 2A). In contrast, under non-reducing conditions, smeared patterns of Ͼ50 kDa in size were supposed to be SOD1 oligomers ( Fig. 2A). These observations suggest that disulfide bonds maintained SOD1 oligomers. Thus, non-reducing SDS-PAGE was thought to be an efficient method to detect SOD1 oligomers and aggregates. Among the holo-enzymes, wild-type and G93A were not oligomerized; instead, they segregated as monomer and dimer size bands (Fig. 2B, top panel). After incubation with Ͼ100 M AA, holo-A4V showed a faint smear pattern that was seen from 50 kDa to near the stacking gel range beside monomer-and dimer-size bands (Fig. 2B, top  panel). In contrast, all metal-deficient enzymes, regardless of mutations, were oligomerized in the presence of Ͼ30 M AA (Fig. 2B, bottom panel). Apo-enzymes demonstrated higher oligomerization propensity than holo-enzymes depending on AA concentration (Fig. 2B). Thus, AA efficiently promoted oligomerization of SOD1s.
Next, we performed a time-course analysis of SOD1 oligomerization in the presence of AA. Metal-deficient G93A and A4V were oligomerized in a time-dependent manner (Fig. 2C). Maximum oligomerization was reached within 60 min of incubation in the presence of AA (Fig. 2C).
We then examined the effect that various FAs, including stearic acid, oleic acid, linoleic acid, and AA, have on SOD1 oligomerization. Unsaturated FAs, including oleic acid, linoleic acid, and AA, promoted SOD1 oligomerization (Fig. 3). However, saturated FAs and stearic acid had little effect on SOD1 oligomerization (Fig. 3). SOD1 oligomerization induced by FAs required at least monounsaturated FAs. This result may reflect the difference of solubility between unsaturated and saturated FAs in the buffer.
We verified the formation of SOD1 oligomers using a 10 -40% glycerol density gradient centrifugation because presumable artifacts after detection of SOD1 oligomers using nonreducing SDS-PAGE may have remained. After fractionation, we could not observe high molecular mass SOD1 oligomers from the incubated sample in the absence of AA; fractions were Ͻ67 kDa and potentially represented monomer and dimer states (Fig. 4A, top panel). In contrast, we detected high molecular mass oligomers in fractions of Ͼ440 kDa from the incubated sample in the presence of AA (Fig. 4A, bottom panel). Under these conditions, SOD1 with molecular mass of Ͻ67 kDa was dramatically decreased compared with the sample incu- FIG. 1. Characterization of purified recombinant SOD1s. A, purified SOD1s were separated using SDS-PAGE and stained with Coomassie Brilliant Blue. B, dismutase activity of the purified SOD1s was assayed by the xanthine/xanthine oxidase-based method. One unit of the activity is defined as the amount of enzyme needed to exhibit 50% of dismutation of the superoxide radicals. C, metal content of the purified SOD1s was measured using 4-pyridylazoresorcinol assay in 6 M guanidine-HCl. bated in the absence of AA (Fig. 4A, bottom panel). Although oligomers of Ͼ440 kDa were fractionated by the glycerol density gradient centrifugation, these were detected as monomer, dimer, and smeared high molecular mass bands that reached stacking gels under non-reducing SDS-PAGE (Fig. 4A, bottom  panel). This indicates oligomers are partly disrupted during the boiling of the SDS-PAGE loading buffer. We next performed densitometric analysis from Western blotting images to estimate the amount of oligomerized SOD1 (Fig. 4B). The resulting image analysis found that immunoreactivity for oligomers was ϳ80% of the total immunoreactivity.
Structural Instability of SOD1 Is Correlated to Oligomerization Propensity and FA Binding-We showed the FA-induced oligomerization propensity of apo-SOD1s was higher than that of holo-SOD1. This implies that protein stability might be strongly associated with FA-induced oligomerization propensity. Among the holo-enzymes, wild-type and G93A were not oligomerized under our experimental conditions (Fig. 2B, top  panel). To examine the correlation between oligomerization propensity and protein stability of holo-enzymes, holo-SOD1 was heated and then oligomerized by AA. In the absence of AA, only heat-treated A4V was oligomerized (Fig. 5, left panel). In the presence of AA, heat-treated G93A and A4V were highly aggregated, but under the same conditions, wild-type SOD1 was not (Fig. 5, right panel). Oligomerization was observed above 58°C for G93A and above 48°C for A4V (Fig. 5, right  panel). In the previous study, A4V was more unstable than G93A for heat treatment analyzed by differential scanning calorimetry (12). This result suggests that structural instability is strongly correlated with oligomerization propensity induced by FAs.
Although we showed that FAs promoted SOD1 oligomerization, the mechanism is not perfectly understood. Similarly, unsaturated FAs oligomerize ␣-synuclein and tau. In the case of ␣-synuclein and tau, FAs were bound to proteins, which suggested that oligomerization mechanisms underlie the FA binding characteristics of protein. To examine whether SOD1 binds to FAs, we carried out a solid-phase oleic acid binding assay. Among the holo-enzymes, very small amounts of holo-A4V were bound to the oleate-Sepharose column, whereas wildtype and G93A were not (Fig. 6A). All of the apo-enzymes were bound to oleate-Sepharose, regardless of their mutations (Fig.  6A). In contrast, bound proteins were not observed in mock-Sepharose (Fig. 6A). Nearly all of the input amounts of metaldeficient proteins were bound, which was estimated by 50% input. This finding suggests that metal-deficient SOD1 proteins strongly bind to FAs. We next examined whether heattreated holo-enzymes bind to FAs. Apo-enzymes were used as control binding. Heat-treated SOD1 mutant (G93A) at 58°C and 68°C was bound to FAs, whereas wild-type was not (Fig.  6B). The results of the FA binding assay were strongly correlated with the oligomerization propensity of SOD1. These findings suggest that FA binding alters the conformation of SOD1 to form oligomers.
FA-induced SOD1 Aggregates Result in Granular Morphology and Are Cytotoxic-We analyzed the ultrastructure of SOD1 aggregates by electron microscope. SOD1 proteins (ϳ40 M) were incubated in the presence of 100 M AA at 37°C for 24 h. Holo-enzymes were heated at 50°C for 30 min before incubation in the presence of AA. After incubation, granular aggregates were observed in all of apo-enzymes and heattreated SOD1 mutants (Fig. 7A). In contrast, no visible materials were found in wild-type holo-SOD1s, even though they were heat-treated (Fig. 7A). The morphology of the aggregates was round or amorphous large granules composed of clustered small granules (Fig. 7A). We could not observe any visible protein aggregates in the samples incubated without AA, except in apo-A4V, which revealed a fibril structure (data not shown).
We next examined the effect of FA-induced aggregates on cell viability of differentiated neuro2a cells. Aggregates of SOD1s were formed using the same methods as described for observation under an electron microscope. Aliquots incubated in the presence or absence of AA were diluted in the culture medium, which was directly added to differentiated neuro2a cells. After incubation for 18 h, toxicity was assessed with MTS reduction (Fig. 7B) and trypan blue staining (Fig. 7C). The presence of the granular aggregates formed by AA from Apo-SOD1s and heat-treated SOD1 mutants significantly reduced cell viability (Fig. 7, B and C). In contrast, no significant decrease of viability was detected when the cells were exposed either to incubated proteins in the absence of AA or to the buffer solutions used to form the aggregates in the absence of added protein (Fig. 7, B  and C). These findings suggest that FA-induced SOD1 aggregates were highly toxic to the cells. DISCUSSION Numerous neurodegenerative diseases are accompanied by highly insoluble inclusions of protein aggregates within characteristic neuronal populations. In the case of FALS, the protypical Lewy body-like hyaline inclusions, composed largely of granule-coated fibrils of SOD1-insoluble filaments, have been detected in the spinal cord of FALS patients with SOD1 gene mutations (5,28). Although there has been controversy about whether such inclusions are a cause or a consequence of the neuronal degeneration, accumulating evidence suggests that aggregates formed via misfolded proteins, especially soluble oligomeric assemblies, may cause cell injury (29 -31). Moreover, cytotoxicity of protein aggregates may have common features because granular aggregates form non-pathological proteins that can also be toxic (26). These findings suggest the avoidance of protein aggregation may be crucial for therapy of FIG. 6. Solid-phase oleic acid binding assay shows apo-SOD1 or thermally destabilized SOD1 bound to oleate-Sepharose. A, solid-phase binding assay was performed as described under "Experimental Procedures." Approximately 50% input (100 ng of proteins) was electrophoresed to estimate the quantity of FA binding SOD1. B, holo-SOD1s (wild-type and G93A) were thermally destabilized at the indicated temperatures for 30 min and then directly loaded on oleate-Sepharose. Apo-enzymes were used as positive controls for oleic acid binding. In the present study, we demonstrated that unsaturated FAs promoted SOD1 oligomerization at physiological pH. SOD1 oligomers were detected by SDS-PAGE under non-reducing conditions. Although immunoreactivity for SOD1 oligomers was decreased in SDS-PAGE under reducing conditions, SOD1 oligomers were considerably SDS-resistant under non-reducing conditions. Based on this method, we found that apo-SOD1 proteins were highly oligomerized by AA compared with holo-SOD1 proteins in time-dependent and FA concentration-dependent manners (Fig. 2, B and C). Metal-deficient SOD1s may be representative of misfolding intermediates for their oligomeric assemblies because they are oligomerized independent of their mutations. These findings suggest that metal-deficient SOD1 proteins have a high oligomerization propensity, which is consistent with previous studies (9,10,13,32). Moreover, heating of holo-SOD1 mutants increased the tendency to form oligomer complexes, especially in the presence of AA; however, the wild-type holo-SOD1 did not form oligomers, even after heating to 68°C and exposure to AA (Fig. 5). This finding suggests that mutations of SOD1 primarily affect their conformation. Our time-course analysis of oligomerization demonstrates that FAs induced the oligomerization process fairly rapidly. We could detect oligomers within 1 h of incubation in the presence of AA (Fig. 1C). Glycerol density gradient centrifugation analysis showed that oligomer species were roughly estimated to be Ͼ80% of the total SOD1 after a 90-min incubation in the presence of AA (Fig. 4). The conversion efficiency and the speed of oligomer formation may be considered as supportive evidence that these reactions occur in vivo.
Aggregations of misfolded proteins are primarily affected by their mutations, especially in inherited conformational diseases. Mutant proteins in conformational diseases have a common characteristic of easily unfolding in a physiological condition and favoring aggregate formation. Protein aggregation has also been shown to be modulated by several factors, including protein concentration, pH, and interactions with other elements such as lipid molecules. It has been reported that FAs stimulated the polymerization of amyloid ␤-peptides, tau (17,33), and ␣-synuclein (18,19) in vitro. These studies suggest that FAs play a pivotal role as nucleates in the self-assembly of misfolded proteins. Although the precise mechanism of how lipid molecules accelerate protein aggregation has not been elucidated, it has been proposed that lipid-bound proteins change their conformation or anionic surfaces, presenting as micelles or vesicles, which can serve to nucleate aggregate formation (18,34,35). We confirmed that apo-SOD1s or heattreated holo-SOD1 mutants were bound to oleic acid (Fig. 6). The FA binding properties of SOD1s were strongly correlated to their conformational instability. These results are consistent with the notion that misfolding intermediates of SOD1 caused by mutations or metal loss may be facilitated by FAs to form oligomeric structures. Another possible mechanism is protein oxidation by FAs. Oxidation also enhances misfolding and aggregation of SOD1 (32). In particular, FAs can lead to the production of radicals because they are easily peroxidized by auto-oxidation to generate peroxyl radicals. However, we could not inhibit SOD1 oligomerization using even a considerable amount of radical scavenger (data not shown). Moreover, oxidized derivatives of FAs also induced SOD1 oligomerization to a similar extent with fresh FAs (data not shown). This finding suggests that oxidation or oxidative damage of SOD1 does not directly drive SOD1 oligomerization. Rather, it is most likely to be associated with a SOD1-destabilizing event.
Recently, several studies for in vitro aggregation of SOD1 have been published. Aggregation of SOD1 can be induced by metal-catalyzed oxidation (32), trifluoroethanol, or heat treatment (10), which induces oxidative modification or protein destabilization. This indicates that structurally unstable SOD1 has an influence on its aggregate formation in vitro. Crystallographic studies suggest that metal-deficient SOD1 forms an amyloid-like assembly caused by non-native conformational changes and permits dimer interaction (36,37). This amyloidlike structure was represented by prolonged incubation of SOD1 at acidic pH (9). In the present study, ultrastructural analysis showed that the FA-inducing aggregates had round or amorphous morphology with clustered tiny spherical aggregates (Fig. 7A). They resemble pre-fibrillar aggregates of the N-terminal domain of Escherichia coli HypF protein or aggregates of the Src homology 3 domain of cytosolic phosphatidylinositol 3-kinase as reported by Stefani and co-workers (26). They demonstrated that granular aggregates of proteins, even non-pathological proteins, are cytotoxic when applied externally (26). Our data also demonstrate that granular aggregates of SOD1s reveal significant cytotoxicity (Fig. 7, B and C). Although the cytotoxic mechanism of the aggregates is not completely understood, it has been proposed that such pre-fibrillar intermediates may lead to cytotoxicity by permeabilization of the membrane bilayer (38,39).
The present findings may provide considerable pathological implication for FALS. Lipid molecules such as FAs may be positive modulators for misfolded protein aggregations. Most misfolded proteins including SOD1 mutants are rapidly degraded by the ubiquitin-proteasome system. Unsaturated FAs may promote misfolded protein aggregations before they are degraded. In addition, cytotoxic aggregate formation of SOD1 may require FAs because granular aggregates structures were markedly observed in SOD1s incubated with AA. Although it is not clear whether the cytotoxic aggregates of SOD1s are generated intracellularly, we have provided a protein aggregation model system to help understand the pathological significance of FAs as a positive modulator for the aggregate formation in FALS. We believe that our system will contribute to efficient drug screening for inhibitors of SOD1 aggregation.