Exposure of Hydrophobic Surfaces Initiates Aggregation of Diverse ALS-Causing Superoxide Dismutase-1 Mutants

doi:10.1016/j.jmb.2010.04.019 J. Mol. Biol. (2010) 399, 512–525

Available online at www.sciencedirect.com

Exposure of Hydrophobic Surfaces InitiatesAggregation of Diverse ALS-Causing SuperoxideDismutase-1 Mutants

Christian Münch and Anne Bertolotti⁎

MRC Laboratory of MolecularBiology, Hills Road, CambridgeCB2 0QH, UK

Received 19 November 2009;received in revised form12 April 2010;accepted 13 April 2010Available online24 April 2010

0022-2836/$ - see front matter © 2010 E

*Corresponding author. E-mail [email protected].

The copper-zinc superoxide dismutase-1 (SOD1) is a highly structuredprotein and, a priori, one of the least likely proteins to be involved in amisfolding disease. However, more than 140, mostly missense, mutations inthe SOD1 gene cause aggregation of the affected protein in familial forms ofamyotrophic lateral sclerosis (ALS). The remarkable diversity of the effectsof these mutations on SOD1 properties has suggested that they promoteaggregation by a variety of mechanisms. Experimental assessment ofsurface hydrophobicity using a sensitive fluorescent-based assay, revealedthat diverse ALS-causing mutations provoke SOD1 aggregation byincreasing their propensity to expose hydrophobic surfaces. These findingscould not be anticipated from analysis of the amino acid sequence. Ourresults uncover the biochemical nature of the misfolded aggregation-proneintermediate and reconcile the seemingly diverse effects of ALS-causingmutations into a unifying mechanism. Furthermore, the method wedescribe here will be useful for investigating and interfering withaggregation of various proteins and thereby provide insight into themolecular mechanisms underlying many neurodegenerative diseases.

© 2010 Elsevier Ltd. All rights reserved.

Keywords: superoxide dismutase; amyotrophic lateral sclerosis; aggrega-tion; neurodegeneration
Edited by S. Radford
Introduction

Aggregation of proteins of unrelated sequence isa major hallmark of neurodegenerative diseases.Amyotrophic lateral sclerosis (ALS) is the mostcommon motor neuron disease. It has an adultonset and is rapidly progressive. The majority ofALS cases are sporadic but a large group ofdominantly inherited forms of the disease arecaused by mutations in the gene encoding theabundantly expressed cytosolic superoxide dismu-tase-1 (SOD1).1,2 How SOD1 mutants provokemotor neuron degeneration remains to be eluci-dated, but it is now well established that SOD1mutations cause familial ALS (fALS) by a gain oftoxic properties and not by a loss of enzymaticactivity. Mice lacking SOD1 do not develop motorneuron disease.3 In contrast, transgenic miceexpressing the ALS-causing SOD1 mutants, inaddition to their endogenous SOD1, develop

lsevier Ltd. All rights reserve

ress:

symptoms reminiscent of the human disease.1,4

Similar to other neurodegenerative diseases, amajor hallmark of ALS is the presence of protein-aceous inclusions in affected neurons. In fALSpatients with SOD1 mutations, aggregated SOD1is the major component of the motor neuroninclusions5,6 and ALS-causing SOD1 mutants alsoform inclusions in transgenic mice.1,7 Thus, ALS-causing mutations confer aggregation propensityas well as some toxic properties to SOD1.Understanding how SOD1 mutations cause aggre-gation of the protein is of major importance, as it isone of the earliest events in the ALS pathogenesis.SOD1 is an abundant and ubiquitously expressed

32 kDa homodimeric enzyme that catalyzes thedismutation of superoxide radicals. Each monomerfolds as an eight-strandedGreek key β-barrel,2 bindsone copper and one zinc ion, and contains a disulfidebond. The native protein is extremely stable andthus, a priori, one of the least likely proteins to beinvolved in a neurodegenerative disease. However,more than 140, mostly missense mutations in SOD1cause fALS.2 SOD1 mutations are scattered throughthe entire sequence of the protein and have verydiverse effects on the properties of the protein. SOD1

d.

http://dx.doi.org/10.1016/j.jmb.2010.04.019

mailto:[email protected]

513Aggregation of ALS-Causing SOD1 Mutants

mutants partition in two groups, on the basis of theirmetal-binding properties and biological activities.One group is characterized by mutations in themetal-binding region (MBR) that perturb metalbinding. The other fALS mutants generally retainwild-type metal content and biological activity andare referred to as wild-type like mutants (WTL).2

How such diverse ALS-causing SOD1 mutations allcause aggregation of the protein is a great puzzle.Most mutations in SOD1 only marginally reduce thestability of the native protein8 and a variety ofdifferentmechanisms have been proposed to explainthe increased aggregation propensity of the diverseSOD1 mutants. Aggregation can arise as a conse-quence of perturbation of the structural integrity dueto loss of metal binding,8–12 reduction of the intra-monomer disulfide bond13–16, oxidation,17 destabi-lization of the dimer18,19 or alteration of post-translational modifications.20,21 Some biochemicalproperties of the mutant proteins such as reductionof the repulsive charge of the protein,22 aberranthydrophobicity of the apo-protein,23 perturbedfolding24,25 or increased unfolding rates24,26 couldaccount for the increased aggregation of SOD1mutants, but it has not been established whether

Fig. 1. Diverse properties of fALS-causing SOD1 mutants. (1SPD).53 Mutants analyzed in this study are highlighted in orafeatures of the mutants analyzed in this study. (b) CalculatedCalculations were performed with PROTEIN CALCULA(Δhydrophobicity) resulting from the mutations. Hydrophydrophobicity values taken form.54

any of these alterations directly cause aggregation.Because of the diversity of the alterations provokedby SOD1 mutations, specific rules have beenproposed to govern aggregation of distinct subsetsof ALS-causing mutants.Some structural alteration of the native fold

ought to be required to elicit SOD1 misfolding.However, because SOD1 is such a stable protein,aggregation is likely to be initiated by local andpossibly subtle unfolding of the native state, ratherthan global unfolding. Therefore, our ability todetect early conformational changes on the mis-folding pathway depends on the availability of asensitive assay. While using the conformation-sensitive dye Sypro Orange, which fluoresces in ahydrophobic environment,27 as a sensitive methodto detect conformational transition, we haveuncovered the biochemical nature of the aggrega-tion-prone conformer. We found that ALS-causingmutations provoke aggregation by increasing thepropensity of diverse SOD1 mutants to exposehydrophobic surfaces, a common feature thatcould not be anticipated from bioinformaticanalysis of the biochemical alterations provokedby the mutations.28

a) Schematic representation of SOD1 structure (PDB entrynge, copper in green and zinc in magenta. Table listing keychange of charge (Δcharge) resulting from the mutations.TOR v3.3. (c) Calculated change of hydrophobicityhobicity values were calculated according to,42 with

514 Aggregation of ALS-Causing SOD1 Mutants

Results

Increased Sypro Orange fluorescence is acommon feature of ALS-causing SOD1 mutants

We selected a set of structurally diverse SOD1mutants, including mutants that account for most ofthe fALS cases with SOD1 mutations.29 The selectedmutants display the broad variety of biophysicaland biochemical properties that characterize fALS-causing SOD1 mutations and different clinicalcharacteristics (Fig. 1a). This comprises six WTLand six MBR mutants selected because 1) they allcause fALS, 2) their positions are scattered through-out the sequence of the protein, 3) they have beenpreviously characterized and exhibit different enzy-matic, physicochemical and structural properties, 4)the disease duration varies greatly for patients withthese different SOD1 mutations (Fig. 1a). Theselected mutants include the A4V mutant, which isboth the most common and the most severe fALS-causing SOD1mutant and retains WTL properties;30

the common H46R mutant with compromisedcopper binding and disease duration of about 17years;31 and the N139K and S134N mutants, which

Fig. 2. Increased surface hydrophobicity upon thermalmutants. (a) Sypro Orange fluorescence during thermal unfoldData are means of 3 independent experiments. s.d. are presemeasured in (a and b), as the mean Sypro Orange-derived fluSOD1WT, using either as-purified or EDTA-treated proteins.

were previously described as nearly indistinguish-able from the wild-type protein.2,10 A bioinformaticanalysis has previously revealed that SOD1 muta-tions have an overall tendency to decrease the netnegative charge of SOD1.22 We also find this trendin the representative set of SOD1mutants selected inthis study, since seven out of the twelve mutantsselected exhibited a decrease in their repulsivecharge (Fig. 1b). In addition, we examined thechange of hydrophobicity introduced by the muta-tions. Four mutations dramatically decreased andtwo increased the hydrophobicity, while the othermutations only caused subtle changes (Fig. 1c).These analyses confirmed that the selected SOD1mutants cover the broad range of propertiescharacteristic of SOD1 mutants.Aiming to set up an assay to detect the conforma-

tional rearrangements on themisfolding pathway,weanalyzed the behaviour of SOD1 mutants in thepresence of a conformation sensitive dye. Proteinswere expressed in insect cells and purified asdescribed9 (and Supplementary Fig. 1). We carriedout thermal denaturation of SOD1 proteins in thepresence of Sypro Orange (Fig. 2a, b and Supplemen-tary Fig. 2). At room temperature, SOD1G37R andSOD1G85R already exhibited a significant fluorescence

denaturation is a generic feature of ALS-causing SOD1ing of as-purified SOD1 without or (b) with 20 mM EDTA.nted in Supplementary Fig. 2. (c) Hydrophobicity valuesorescence maxima (Fmax) or normalized to the Fmax of


in the presence of the conformation sensitive dye, incontrast to all other mutants and wild-type protein(Fig. 2a). This confirmed that under native conditions,SOD1WT and most mutants did not expose hydro-phobic surfaces as expected for tightly foldedproteins. Heat denaturation of all mutants otherthan SOD1G37R and SOD1G85R, in the presence ofSypro Orange, resulted in a marked increase influorescence. Notably, the assay used here revealeddifferences between SOD1WT and the N139K andS134N mutants that have escaped previous analyses.The fluorescence maxima of the as-purified SOD1mutants ranged from about 2-fold, to more than a 14-fold increase relative to the wild-type protein (Fig. 2aand c). We defined relative surface hydrophobicityvalues as the increase in fluorescencemaxima relativeto wild-type SOD1 (Fig. 2c). This revealed that thestructurally diverse SOD1 mutants exposed morehydrophobicity than the wild-type protein.SOD1 variants expressed and purified in similar

conditions as used here were found to exhibitvarious metal contents9 and previous studieshave revealed that metal deficiency increaseshydrophobicity.23,32 We next set up to determinethe metallation status of most of the SOD1 variantsused in this study, in order to determine whether thedifferences of hydrophobicity of the diverse SOD1variants reflected differences in metal content. Inagreement with a previous study, all the purifiedSOD1 variants analyzed had substochiometric cop-per content and the MBR mutants H46R, S134N andD125H were both copper- and zinc-deficient (Sup-plementary Fig. 3a and9). In the other mutantsanalyzed, the zinc site was fully occupied (Supple-mentary Fig. 3a). We failed to detect a correlationbetween low metal content and high amount ofexposed hydrophobic surfaces (Supplementary Fig.3b). However, it has long been suspected that metaldeficiency could be at the origin of the pathogenicityof SOD1 mutants.8,9 We next exposed the purifiedSOD1 variants to 20 mM EDTA and monitored theeffects of such treatment. As expected, the thermaldenaturation profiles, in the presence of SyproOrange, of the metal deficient mutants H46R,S134N and D125H, were very similar with orwithout 20 mM EDTA (Fig. 2a-c). The same wasobserved for theMBRmutant C146R. In contrast, thefluorescence maxima of the other mutants werehigherwhen heat denatured in the presence of EDTA(Fig. 2a-c). Thus, EDTA treatment significantlyreduced the metal content of the SOD1 variantsand this reduction further enhanced exposure ofhydrophobic surfaces of destabilized SOD1mutants.As observed for the as-purified proteins, mostEDTA-treated SOD1 mutants also exhibited fluores-cence maxima that were higher than that of the wild-

Fig. 3. Destabilization of as-purified SOD1A4V by EDTAaggregation. (a) Assessment of surface hydrophobicity byexperiments (n=3) are shown. (b) NuPAGE analysis of SOD1 bof the SOD1 variants on native PAGE (1.2 μg) or denaturing Nuprior electrophoresis. (d, e) Aggregation of SOD1A4V monitor

type protein (Fig. 2b and c). However, EDTAtreatment attenuated the differences in surfacehydrophobicity between the wild-type protein andthe mutants (Fig. 2a-c). This is likely due to thefinding that demetallation increased exposure ofhydrophobic surfaces on SOD1WT (Fig. 2a-c).Environmentally sensitive fluorescent dyes have

been used previously to monitor thermalunfolding.34 The transition midpoint of the dena-turation curves indicates the unfolding temperature(Tm). The apparent Tm values of the EDTA treatedproteins were extracted from the Sypro Orangeanalyses (Supplementary Fig. 3c). We found that theapparent Tm values of our EDTA-treated mutantswere nearly identical to the ones published for thesame mutants in the apo-state (Supplementary Fig.3c and10,33). This reveals that our EDTA-treatedproteins are similar to apo-proteins prepared inother studies. As expected, EDTA treatment dra-matically reduced the Tm of the biologicallymetallated proteins but had a minor effect on themetal deficient proteins H46R, S134N and D125H.Interestingly, except for SOD1N139K, all as-purifiedSOD1 mutants had a lower apparent Tm than thewild-type protein (Supplementary Fig. 3c). We thenanalyzed whether the extent of exposure of hydro-phobicity directly correlated with Tm. While ther-mal unfolding of most SOD1 mutants is required toexpose hydrophobic surfaces, the amount of ex-posed hydrophobic surfaces did not correlatedirectly with the apparent Tm of the differentmutants (Supplementary Fig. 3d). This revealedthat the knowledge of Tm can't predict the amountof surface hydrophobicity exposed by each mutant.Together, these results reveal that exposure of

hydrophobic surfaces is a common feature ofdiverse SOD1 mutants with different metal contentand stability, when subjected to denaturation.

Increased surface hydrophobicity correlateswith aggregation of SOD1A4V

Having found an unprecedented common featurefor these diverse SOD1 mutants we then askedwhether increased exposure of hydrophobic surfacescaused aggregation. We used conditions previouslyestablished to elicit mutant SOD1 aggregation toaddress this question, such as acidic pH in thepresence or absence of a chelator. In buffer alone, orin the presence of as-purified SOD1WT or SOD1A4V,Sypro Orange fluorescence was very low, indicatingthat hydrophobic residues are buried in bothSOD1WT and SOD1A4V, as expected for nativeproteins (Fig. 3a). At physiological temperature,Sypro Orange fluorescence increased dramaticallyonly in the presence of EDTA-treated SOD1A4V,

or low pH exposes hydrophobic surfaces and triggersSypro Orange fluorescence. Means and s.d. of replicateinding to hydrophobic beads. (c) Electrophoretic mobilitiesPAGE (0.75 μg). Proteins were treated as indicated 30 mined by DLS. d.nm: Diameter in nm.

Fig. 3 (legend on previous page)



indicating that EDTA-treated SOD1A4V exhibitedgreater surface hydrophobicity than the SOD1WT

exposed to the same condition, or the untreatedproteins (Fig. 3a), as previously observed.13

Proteins were next exposed to low pH, a conditionthat elicits SOD1A4V aggregation, with or withoutEDTA.35 At pH 3.9, both EDTA-treated and as-purified SOD1A4V became strongly fluorescent inthe presence of Sypro Orange, unlike SOD1WT

exposed to the same conditions (Fig. 3a), revealingthat EDTA or low pH exposed hydrophobic surfaceson SOD1A4V but not on the wild-type protein. Toconfirm these findings, SOD1WT and SOD1A4V werenext incubated on Phenyl Sepharose. After extensivewashing, a fraction of EDTA-treated and as-purifiedSOD1A4V exposed to low pH was selectivelyretained on the hydrophobic resin (Fig. 3b). Theseresults confirm that the fluorescence-based assay isa highly sensitive method for detecting surfacehydrophobicity.To assess whether exposure to acidic pH in the

presence of EDTA efficiently demetallated theproteins, SOD1WT and SOD1A4V were analyzedon native PAGE, since this method was shown toresolve differentially metallated species.9 Themobility of as-purified SOD1A4V was faster thanas-purified SOD1WT (Fig. 3c), in good agreementwith the finding that this mutant had reducedcopper content, compared to the wild-type protein(Supplementary Fig. 3a). At pH 6.3, EDTA-treatedSOD1A4V exhibited the same mobility as theuntreated mutant on native PAGE (Fig. 3c), whileSypro Orange fluorescence increased in the pres-ence of EDTA-treated SOD1A4V, but not theuntreated mutant (Fig. 3a). This reveals thatSypro Orange fluorescence detected changes thatwere not detectable by native PAGE, suggestingthat the Sypro Orange detects conformationalchanges that precede metal loss. At acidic pH,EDTA treatment increased mobility of bothSOD1WT and SOD1A4V on native PAGE, indicatingthat such treatments efficiently depleted metal andproduced apo-proteins (Fig. 3c). As previouslyobserved,36 we also find that demetallation desta-bilized SOD1 dimers (Fig. 3d and e).We next carried out analyses of SOD1 aggrega-

tion, using as-purified and apo-proteins at a 10 μMconcentration.33 Aggregation was monitored bydynamic light scattering (DLS), which revealedthat after 3 days at pH 3.9, both apo- and as-purifiedSOD1A4V were aggregated (Fig. 3d). As previouslyobserved for wild-type SOD1WT, we found that lowpH destabilized both SOD1WT and SOD1A4V

dimers,37 but only apo-SOD1A4V aggregated.Under the same conditions, apo-SOD1A4V dramat-ically increased Sypro Orange fluorescence (Fig. 3a).In good agreement with the high fluorescence ofdemetallated SOD1A4V at low pH, compared to theas-purified protein (Fig. 3a), we observed that apo-SOD1A4V formed larger particles than as-purifiedSOD1A4V (Fig. 3d). After 7 days, EDTA-treatedSOD1A4V also aggregated at pH 6.3 (Fig. 3e). Incontrast, the wild-type protein remained soluble in

each condition tested (Fig. 3d, e, Supplementary Fig.4 and data not shown). All together, these resultsshow that increased exposure of hydrophobicsurfaces correlates with aggregation of SOD1A4V.

TFE exposes hydrophobic surfaces andprovokes aggregation of SOD1H46R

We next examined whether similar propertiesunderlie aggregation of a mutant unrelated to theWTL mutant SOD1A4V. As-purified SOD1WT andthe MBR mutant SOD1H46R were exposed toincreasing trifluoroethanol (TFE) concentrations, acondition previously shown to trigger aggregationof SOD1 mutants,33 as well as other globularproteins.38 TFE increased Sypro Orange fluores-cence (data not shown). However, when thefluorescence of the dye in buffer was subtracted,we found that the fluorescence of SOD1H46R, but notSOD1WT, increased dramatically in Sypro Orange,with increasing TFE concentrations (Fig. 4a). Thisresult indicates that in the presence of TFE,SOD1H46R but not SOD1WT exposed hydrophobicsurfaces, as confirmed by the specific retention ofSOD1H46R, in the presence of TFE, on hydrophobicresin (Fig. 4b). In parallel, aggregation was moni-tored by DLS and revealed that TFE concentrationshigher than 15% triggered aggregation of SOD1H46R,but not SOD1WT (Fig. 4c).To confirm that the large particles observed upon

incubation of SOD1H46R with TFE were aggregates,SOD1WT and SOD1H46R, in non-denaturing bufferor in the presence of increasing concentrations ofTFE, were filtered through a cellulose acetatemembrane and revealed with SOD1 antibodies, asdescribed.39,40 While unaggregated SOD1 filteredthrough the membrane, aggregated SOD1H46R wasretained when the mutant was exposed to 20% TFE(Fig. 4d), indicating that the large particles observedby DSL were SOD1H46R aggregates (Fig. 4c). Equalamounts of proteins were loaded on the filterretardation assay and on a denaturing NuPAGEgel (Fig. 4d, lower panel). However, the amount ofSOD1H46R in 20% TFE that was resolved on thedenaturing gel was consistently lower than that inabsence of TFE. This indicated that aggregates werenot resolved on the gel, suggesting that they mightbe resistant to boiling in reducing buffer. We noticedthat SOD1H46R required lower TFE concentrations totrigger Sypro Orange fluorescence than to aggregate(Fig. 4a, c and d). This indicates that exposure ofhydrophobic surfaces precedes aggregation.We next examined the morphology of mutant

SOD1 aggregates to determine whether they wereamorphous or ordered. Electron microscopyrevealed that SOD1H46R aggregates exhibited bothfibrillar and granular components, reminiscent ofthe granule-coated fibrils found in affected neuronsof ALS patients (Fig. 4e and Ref. 41). Together, theseresults demonstrate that aggregation of two distinctSOD1 mutants, elicited under different conditions,directly correlates with exposure of hydrophobicsurfaces.

Fig. 4. TFE treatment exposes hydrophobic surfaces on as-purified SOD1H46R and provokes fibrillogenesis. (a)Assessment of SOD1H46R surface hydrophobicity by Sypro Orange fluorescence in the presence of increasing TFEconcentrations. (b) NuPAGE analysis of SOD1 binding to hydrophobic beads. (c) Aggregation of SOD1H46R monitored byDLS. (d) Aggregation of SOD1H46R revealed by filter retardation assay and NuPAGE analysis of aliquots of the samplesused in the filter retardation assay. (e) Electron micrographs of negatively stained SOD1H46R fibrils formed in 20% TFEacquired at magnifications of 7100 (upper panel) and 52000 (lower panel). Means and s.d. of 3 independent experimentsare shown in (a and c).


Exposure of hydrophobic surfaces precedesaggregation

Aggregation of SOD1H46R occurred within minutesafter TFE addition,while a longer timewas required todetect aggregation of SOD1A4V at low pH (Figs. 3, 4and data not shown). We took advantage of the slowaggregation kinetics of both as-purified and apo-

SOD1A4V at acidic pH to analyze the aggregationprocess in greater detail. As-purified and apo-SOD1A4V were adjusted to low pH (day 0) andaliquots taken every day to analyze both SyproOrange-derived fluorescence and aggregation. Imme-diately after exposure to low pH, both as-purified andapo-SOD1A4V increased the fluorescence of theconformation-sensitive dye (Fig. 5a, day 0). The


increase in fluorescence preceded the onset of aggre-gation (Fig. 5a and b, day 0). Fluorescence of SyproOrange then reached a maximum after 1 day andgradually decreased after 2 days (Fig. 5a). Concomi-tantly, aggregates gradually grew (Fig. 5b andSupplementary Fig. 5a). Note that the soluble proteinswere barely detectable after 2 days (SupplementaryFig. 5a). Together, these analyses reveal that low pH

exposes hydrophobic regions on the surface ofSOD1A4V and such regions are buried in aggregates.To determine whether these findings were re-

stricted to the experimental conditions used toinduce mutant SOD1 aggregation or were inherentto the aggregation mechanism, we used hightemperature, since this condition was previouslyfound to trigger SOD1A4V aggregation.33 Similar towhat we observed upon destabilization of SOD1A4V

at low pH, heating SOD1A4V to 50 °C increased thefluorescence of Sypro Orange (Fig. 5c). The fluores-cence of the conformation-sensitive dye decreasedover time, while aggregates grew (Fig. 5c and d).Note that no soluble proteins were detectable after1 day (Supplementary Fig. 5c). In contrast, heat didnot provoke fluorescence of SOD1WT or aggregation(Fig. 5c and d). Like the ALS deposits, 3 days-oldaggregates were amyloid-like since they increasedthe fluorescence of Thioflavin T (ThT), an amyloidspecific probe (Fig. 5e and f). In contrast, SOD1WT

didn't increase ThT fluorescence (data not shown).The ThT increase for SOD1 aggregates was modest,in agreement with previous studies.33,35 Congo Redbinding experiments also suggested that SOD1A4V

aggregates, in contrast to the wild-type protein,contain an amyloid component, since they produceda small but reproducible shift in the absorbancespectra of Congo Red (Supplementary Fig. 5b, d), aspreviously reported.33,35 Taken together, theseresults indicate that denaturation of SOD1 mutantsby diverse treatments exposes hydrophobic surfacesand that such regions engage non-native interac-tions to form amyloid-like aggregates. Thus, or-dered assembly of SOD1 aggregates is likely to bedriven by hydrophobic interactions. These experi-ments also reveal that Sypro Orange-derived fluo-rescence is a marker of one of the earliestconformational rearrangement in the aggregationpathway, rather than amyloids per se.

Increased propensity of diverse ALS-causingSOD1 mutants to expose hydrophobic surfacescauses aggregation

In all conditions examined using two specificSOD1 mutants, we observed a correlation betweenexposure of hydrophobic surfaces and aggregation.To determine whether increased exposure of

Fig. 5. Aggregation of as-purified SOD1 mutantsfollows exposure of hydrophobic surfaces. (a) Time courseof measurements of exposed hydrophobicity monitoredby Sypro Orange fluorescence and (b) aggregation ofSOD1A4V at acidic pH monitored by DLS. Note thatdemetallation first provoked a decrease in the size ofparticles measured, indicating that metal loss provokedmonomerization of the protein. (c) Time course ofmeasurements of exposed hydrophobicity and (d) aggre-gation of apo-SOD1A4V exposed to 50 °C. (e, f) aggregationof SOD1A4V monitored by ThT after 3 days of incubationat acidic pH (e) or 50 °C (f). Data are means and s.d. valuesof replicate experiments (n=3).


hydrophobic surfaces correlates with aggregationof diverse SOD1 mutants, we monitored aggre-gation and exposure of hydrophobic surfaces ofthe representative set of 12 mutants. As-purifiedSOD1WT and mutant proteins were exposed to 20%TFE, a condition that was found to elicit aggrega-tion of the diverse SOD1 mutants, but not the wild-type protein. This treatment converted the solubleMBR mutants into aggregates in 20 minutes andwas also efficient to provoke aggregation of theWTL mutants (Supplementary Fig. 6a). Exposure ofsurface hydrophobicity was monitored by Sypro

Orange fluorescence and aggregation monitoredboth by DLS and ThT fluorescence (Fig. 6a and b).ThT fluorescence indicated that aggregates have anamyloid component. Analysis of Congo Red bind-ing confirmed this (Supplementary Fig. 6b). Wefound exposure of hydrophobicity and orderedassembly of SOD1 aggregates, monitored by thesetwo independent methods, were strongly correlat-ed. The strength of the correlation between exposureof hydrophobic surfaces and aggregation is attestedby the very high linear correlation coefficient (Fig.6a, R=0.95 and b, R=0.84). This revealed that

Fig. 6. Aggregation of as-puri-fied ALS-causing SOD1 mutants iscaused by their increased propensi-ty to expose hydrophobic surfaces.(a, b) Correlation between surfacehydrophobicity measured by SyproOrange fluorescence and aggrega-tion measured by DLS (a) and ThTfluorescence (b). Data are meansand s.e.m. from 3 independentexperiments. Measurements weremade 20 minutes after TFE treat-ment. The high linear correlationcoefficient R, denotes the strengthof the correlation. Note that theMBR mutants are more susceptibleto TFE-induced aggregation thanthe WTL mutants. (c) Aggregationof SOD1G85R and SOD1G37R at roomtemperature, revealed by DLS.


exposure of hydrophobicity is a common feature thatprovokes aggregation of diverse SOD1 mutants.To further demonstrate that the exposure of

hydrophobic surfaces predicts aggregation, wefocused on SOD1G37R and SOD1G85R. These twomutants produced high Sypro Orange fluorescenceat room temperature (Fig. 2a), suggesting that theymight aggregate at room temperature. DLS analysesconfirmed this hypothesis (Fig. 6c).

Discussion

A common feature of diverse neurodegenerativediseases is the deposition of aggregated proteins inaffected neurons. Mutations of natively unstruc-tured proteins associated with familial forms ofthese diseases cause aggregation of the proteins byaltering simple intrinsic physicochemical propertiessuch as hydrophobicity, secondary structure andcharge.42,43 However, the rules governing aggrega-tion propensity of highly structured proteins, suchas SOD1, remained unknown. The results presentedhere reveal that increased surface hydrophobicity isa generic feature of structurally diverse ALS-causingSOD1 mutants that could not be anticipated bycalculating the changes in hydrophobicity caused bythe mutation or by using algorithms designed topredict protein aggregation.28 This common prop-erty is intrinsic to fALS mutants but not necessarilyconstitutive. The destabilizing mutations G85R andG37R constitutively expose hydrophobic surfaces.For the other SOD1 mutations, destabilization of theproteins exposes otherwise buried hydrophobicsurfaces and causes aggregation. Previous studieshave shown that SOD1 mutations increase unfold-ing rates24,26,33 and we found here that suchunfolded intermediates expose aggregation-pronesurfaces. While it is expected that partly unfoldedproteins expose some hydrophobic surfaces, andsuch hydrophobic regions exist in SOD1,44 it couldnot be anticipated that diverse ALS-causing SOD1point mutations, causing a variety of alterations onthe properties of the proteins, all increase thepropensity to expose hydrophobic surfaces andthat this common feature is the molecular determi-nant underlying the formation of the orderedaggregates. Furthermore, the effects of the muta-tions analyzed here on the predicted hydrophobicityof the protein are completely random: Some muta-tions increase and some decrease hydrophobicity.It has been previously reported that decreased

stability and metal deficiency are important factorsthat influence SOD1 aggregation.8,9 We did notobserve a direct correlation between metal contentand exposure of hydrophobicity upon thermaldenaturation of the as-purified proteins (Supple-mentary Fig. 3). However, treatment with a chelatorfurther increased exposure of hydrophobic surfacesof the SOD1 variants that where loaded with metal.This indicates that demetallation increases exposureof hydrophobicity and thereby increases aggrega-tion propensity. The MBR mutants were also more

susceptible to TFE-induced aggregation than theWTL (Fig. 6), suggesting that these mutants aremore vulnerable to destabilization. Similar to whatwe observed with metal contents, we found nodirect correlation between Tm and exposed hydro-phobicity (Supplementary Fig. 3). Yet, destabiliza-tion of the natively stable mutants is required toexpose the otherwise buried hydrophobic surfaces.These observations are consistent with the followinginterpretation: metal deficiency, stability, suscepti-bility to unfolding are likely to be interconnected.Destabilization of metallated proteins is likely toprovoke metal loss; conversely metal deficiencydestabilizes the proteins. We found here thatdestabilizing diverse SOD1 mutants by varioustreatments systematically increased exposure ofhydrophobic surfaces on the mutant but not thewild-type protein.Taking advantage of the slow aggregation kinetics

of SOD1 mutants at low pH and high temperature,we found that exposure of hydrophobic surfacesprecedes aggregation (Fig. 5). This shows thataggregation of diverse pathogenic SOD1 mutants isdriven by intermolecular hydrophobic interactionseither between constitutively hydrophobic mutantsor aggregation intermediates exposing hydrophobicsurfaces. Our results are consistent with the struc-tures of SOD1 amyloid-like filaments that haverevealed that these amyloids are formed after localunfolding of the zinc and electrostatic loops thatcreates a hydrophobic interface between twoproteins.45 In two previous studies, a subset ofSOD1 mutants from mice tissues have been capturedon hydrophobic resins.23,46 Our findings reveal thatexposure of hydrophobic surfaces is a generic featureof diverse SOD1 mutants and this property governsaggregation propensity. The increased surface hy-drophobicity of SOD1 mutants may also account forseveral of their properties: Their perturbed folding,both in vitro and in cells,24,25 their selective associa-tion with the cytoplasmic face of mitochondria,47

their recognition by chaperones48 as well as theirdecreased half-life,49 since hydrophobic surfaces,once recognized by chaperones may target theprotein to the degradation machinery.50Whether fALS with SOD1 mutations is a typical

amyloidosis is not clear. Mutant SOD1 inclusions inhuman tissues are composed of 15- 25 nm granules-coated fibrils41 but they are not revealed byamyloid-specific dyes. However, inclusions areThioflavin-S-positive in mice expressing SOD1mutants.39 In vitro, we found that SOD1 aggregatesincreased ThT fluorescence, shifted Congo Redabsorbance towards 541 nm and that they exhibitedboth fibrillar and granular components (Fig. 4e).Previous studies have also reported that SOD1aggregates contain an amyloid component.11,15,33,35

Since minute amounts of misfolded SOD1 causeALS,51 one hypothesis that can reconcile theseseemingly conflicting observations is as follows:SOD1 aggregates contain an amyloid componentbut its low abundance precludes its detection byamyloid dyes in human tissues.


Aggregation-prone hydrophobic surfaces are onlytransiently exposed on the aggregation intermediateand later buried inside the aggregate. These findingsmay provide a molecular basis for the hypothesis ofthe toxic aggregation intermediate. Exposure ofhydrophobic surfaces on the aggregation intermedi-ate may be detrimental to the cell, while the maturefibrillar aggregate, burying these surfaces, is pre-dicted to be less harmful. In aggregation reactionswith slow kinetics, we found that Sypro Orangefluorescence precedes aggregation and ThT fluores-cence (Fig. 5) indicating that Sypro Orange is amarker of an early species in the aggregationpathway rather than another amyloid marker.Sypro Orange reveals a soluble, yet aggregation-prone conformer that builds up aggregates. Whenaggregation is elicited with TFE, Sypro Orangeremains high in the presence of TFE-treated SOD1mutants after the onset of aggregation (data notshown). Since aggregation is observed rapidly afterTFE addition, it is likely that TFE is a more potentdestabilizing agent than pH or temperature. Thissuggests that under harsh conditions, some hydro-phobic surfaces remained exposed on the aggregates,in contrast to the aggregation elicited in milderconditions such as low pH or high temperature (Fig.5). Whether the aggregation pathways are differentunder different aggregation conditions is unclear.Regardless of the aggregation pathway, we foundhere that destabilization of SOD1mutants, by severalmethods, exposes hydrophobic surfaces on themutant proteins but not the wild-type. This revealsthat, the initiation of aggregation is conserved. Wepropose that, any condition, yet to be identified, thatwill destabilize the protein in motor neurons, willexpose hydrophobic surfaces and trigger the orderedassembly of fALS-causing SOD1 mutants.This study identifies the common nature of the

early aggregation-prone conformer of diverse ALS-causing SOD1 mutants and thereby provides essen-tial information to elucidate the mechanism bywhich it provokes motor neuron death and tointerfere with this process.The method used here is broadly applicable,

quantitative and high-throughput. The fluorescent-based assay can be used to screen for compoundsthat prevent the formation of the aggregation-proneconformer, one of the earliest events in the diseaseprocess. In addition, this method will be useful forinvestigating the aggregation propensity of a widevariety of proteins and thereby provide insight intothe molecular mechanisms underlying many neuro-degenerative diseases.

Materials and methods

Protein purification

cDNAs encoding human SOD1mutants were generatedby polymerase chain reactions. Wild-type and mutantcDNAs were cloned in pFastBac1 (Invitrogen) andexpressed in Sf9 cells in the presence of 150 μM CuCl2

and ZnCl2. SOD1 proteins were purified as described.9

Assays were performed in the commonly used 10 mMMES buffer pH 6.38, with or without 20 mM EDTA togenerate apo-SOD1. In Figs. 3 and 5 (a and b), 50 mMMESbuffer pH 6.3 or sodium acetate buffer pH 3.9, were used.Protein concentrations were determined bymeasuring ODat 280 nm and by quantitative amino acid analyses.

Fluorescence measurements

Equal amounts of protein (3 μg in Figs. 2, 3, 5 and 6; 1 μgin Fig. 4) were mixed with an excess of Sypro Orange(10×concentration of S5692 Sigma-Aldrich) in 20 μl in a 96well plate and fluorescence was measured in the 7900HTFast Real-Time PCR System from Applied Biosystems andexpressed as arbitrary units (a.u.). Thermal denaturationswere conducted with a heating rate of 1 °C/1.5 min. Notethat Sypro Orange fluorescence measurements of as-purified SOD1 proteins were very similar in MES bufferpH 6.3 and in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl and10% glycerol (data not shown). Well to well variation wasassessed and s.d. was quantified to be less than 10% in thedifferent wells of a 96 well plate (data not shown).

Hydrophobic binding

Equal amounts of protein (10 μg) were incubated on20 μl Phenyl Sepharose 6 Fast Flow (high sub, GEHealthcare) in 200 μl of the indicated buffer overnight atroom temperature and washed in binding buffer. Total,bound and unbound (supernatant) proteins were resolvedon NuPAGE 4-12% Bis-Tris gels (Invitrogen) and stainedwith Coomassie Brilliant Blue G-250.

Electrophoresis

For native PAGE, samples were equilibrated inLaemmli loading buffer lacking SDS and reducing agents,analyzed on 12% Tris-HCl gels and run in Laemmli bufferwithout SDS. NuPAGE 4-12% Bis-Tris gels (Invitrogen)were run in MES buffer and stained with CoomassieBrilliant Blue G-250.

Aggregation analyses by DLS, ThT fluorescence,electron microscopy and filter retardation assays

Wild-type or mutant SOD1 (10 μM) were incubated attemperatures between 25 to 50° C, in the indicatedbuffers, in the presence of 0 to 20% of 2-2,2-trifluor-oethanol where indicated (Fluka). After 4 to 5 hours(day 0) up to 7 days, the sizes of the particles weremeasured using the Zetasizer Nano S (Malvern) andpresented either as the distribution of particle size or theaverage size (Z average). To monitor amyloid content,reactions were diluted into ThT solutions with a finalconcentration of 10 μM ThT and 30 mM glycine, pH 8.5.Fluorescence emission between 480 nm and 500 nm wasmeasured immediately after dilution, using an excitationwavelength of 446 nm in a Tecan Safire II. Electronmicrographs were acquired by using either a PhilipsEM208 or a Tecnai T12 transmission electron micro-scope. A sample of 2.5 μl of a 10 μM protein solutionwas negatively stained with equal volume of a saturatedsolution of uranyl acetate. Filter retardation assays wereperformed as described in52 and revealed with SOD-100E antibody (Stressgen).


Note that Sypro Orange is not present during anyaggregation reactions.

Acknowledgements

We are very grateful to Nigel Unwin for his helpwith preparing samples for electron microscopy andcomments on the manuscript and to Dr Jason Dayfor the ICP-MS analyses. We thank Benjamin Dehayfor cloning wild-type and G85R human SOD1, LoriPassmore and Graham Fraser for help with electronmicroscopes, David Owen for quantitative aminoacid analyses, Sarath C. Janga and Madan Babu forhelp with data processing and Michel Goedert fordiscussions and comments on the manuscript. TheMedical Research Council and the EMBO YoungInvestigator Programme supported this research.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2010.04.019

References

1. Bruijn, L. I., Miller, T. M. & Cleveland, D. W. (2004).Unraveling the mechanisms involved in motorneuron degeneration in ALS. Annu. Rev. Neurosci.27, 723–749.

2. Valentine, J. S., Doucette, P. A. & Zittin Potter, S.(2005). Copper-zinc superoxide dismutase and amyo-trophic lateral sclerosis. Annu. Rev. Biochem. 74,563–593.

3. Reaume, A. G., Elliott, J. L., Hoffman, E. K., Kowall,N. W., Ferrante, R. J., Siwek, D. F. et al. (1996). Motorneurons in Cu/Zn superoxide dismutase-deficientmice develop normally but exhibit enhanced celldeath after axonal injury. Nat. Genet. 13, 43–47.

4. Gurney, M. E., Pu, H., Chiu, A. Y., Dal Canto, M. C.,Polchow, C. Y., Alexander, D. D. et al. (1994). Motorneuron degeneration in mice that express a humanCu,Zn superoxide dismutase mutation. Science, 264,1772–1775.

5. Shibata, N., Hirano, A., Kobayashi, M., Siddique, T.,Deng, H. X., Hung, W. Y. et al. (1996). Intensesuperoxide dismutase-1 immunoreactivity in intra-cytoplasmic hyaline inclusions of familial amyo-trophic lateral sclerosis with posterior columninvolvement. J. Neuropathol. Exp. Neurol. 55, 481–490.

6. Kato, S., Nakashima, K., Horiuchi, S., Nagai, R.,Cleveland, D. W., Liu, J. et al. (2001). Formation ofadvanced glycation end-product-modified superoxidedismutase-1 (SOD1) is one of the mechanismsresponsible for inclusions common to familial amyo-trophic lateral sclerosis patients with SOD1 genemutation, and transgenic mice expressing humanSOD1 gene mutation. Neuropathology, 21, 67–81.

7. Bruijn, L. I., Houseweart, M. K., Kato, S., Anderson,K. L., Anderson, S. D., Ohama, E. et al. (1998).Aggregation and motor neuron toxicity of an ALS-

linked SOD1 mutant independent from wild-typeSOD1. Science, 281, 1851–1854.

8. Lindberg, M. J., Tibell, L. & Oliveberg, M. (2002).Common denominator of Cu/Zn superoxide dismu-tase mutants associated with amyotrophic lateralsclerosis: decreased stability of the apo state. Proc.Natl Acad. Sci. USA, 99, 16607–16612.

9. Hayward, L. J., Rodriguez, J. A., Kim, J. W., Tiwari, A.,Goto, J. J., Cabelli, D. E. et al. (2002). Decreasedmetallation and activity in subsets of mutant superox-ide dismutases associated with familial amyotrophiclateral sclerosis. J. Biol. Chem. 277, 15923–15931.

10. Rodriguez, J. A., Shaw, B. F., Durazo, A., Sohn, S. H.,Doucette, P. A., Nersissian, A. M. et al. (2005).Destabilization of apoprotein is insufficient to explainCu,Zn-superoxide dismutase-linked ALS pathogene-sis. Proc. Natl Acad. Sci. USA, 102, 10516–10521.

11. Banci, L., Bertini, I., Durazo, A., Girotto, S., Gralla,E. B., Martinelli, M. et al. (2007). Metal-free super-oxide dismutase forms soluble oligomers underphysiological conditions: a possible general mecha-nism for familial ALS. Proc. Natl Acad. Sci. USA, 104,11263–11267.

12. Durazo, A., Shaw, B. F., Chattopadhyay, M., Faull,K. F., Nersissian, A. M., Valentine, J. S. & Whitelegge,J. P. (2009). Metal-free superoxide dismutase-1and three different amyotrophic lateral sclerosisvariants share a similar partially unfolded beta-barrelat physiological temperature. J. Biol. Chem. 284,34382–34389.

13. Tiwari, A. & Hayward, L. J. (2003). Familial amyo-trophic lateral sclerosis mutants of copper/zincsuperoxide dismutase are susceptible to disulfidereduction. J. Biol. Chem. 278, 5984–5992.

14. Lindberg, M. J., Normark, J., Holmgren, A. &Oliveberg, M. (2004). Folding of human superoxidedismutase: disulfide reduction prevents dimerizationand produces marginally stable monomers. Proc. NatlAcad. Sci. USA, 101, 15893–15898.

15. Chattopadhyay, M., Durazo, A., Sohn, S. H., Strong,C. D., Gralla, E. B., Whitelegge, J. P. & Valentine, J. S.(2008). Initiation and elongation in fibrillation of ALS-linked superoxide dismutase. Proc. Natl Acad. Sci.USA, 105, 18663–18668.

16. Karch, C. M., Prudencio, M., Winkler, D. D., Hart, P. J.& Borchelt, D. R. (2009). Role of mutant SOD1disulfide oxidation and aggregation in the pathogen-esis of familial ALS. Proc. Natl Acad. Sci. USA, 106,7774–7779.

17. Rakhit, R., Cunningham, P., Furtos-Matei, A., Dahan,S., Qi, X. F., Crow, J. P. et al. (2002). Oxidation-inducedmisfolding and aggregation of superoxide dismutaseand its implications for amyotrophic lateral sclerosis.J. Biol. Chem. 277, 47551–47556.

18. Hough, M. A., Grossmann, J. G., Antonyuk, S. V.,Strange, R. W., Doucette, P. A., Rodriguez, J. A. et al.(2004). Dimer destabilization in superoxide dismutasemay result in disease-causing properties: structures ofmotor neuron disease mutants. Proc. Natl Acad. Sci.USA, 101, 5976–5981.

19. Wang, J., Caruano-Yzermans, A., Rodriguez, A.,Scheurmann, J. P., Slunt, H. H., Cao, X. et al. (2007).Disease-associated mutations at copper ligand histi-dine residues of superoxide dismutase 1 diminishthe binding of copper and compromise dimerstability. J. Biol. Chem. 282, 345–352.

20. Furukawa, Y., Kaneko, K., Yamanaka, K., O'Hal-loran, T. V. & Nukina, N. (2008). Complete loss ofpost-translational modifications triggers fibrillar




aggregation of SOD1 in the familial form ofamyotrophic lateral sclerosis. J. Biol. Chem. 283,24167–24176.

21. Wilcox, K. C., Zhou, L., Jordon, J. K., Huang, Y., Yu,Y., Redler, R. L. et al. (2009). Modifications ofsuperoxide dismutase (SOD1) in human erythrocytes:a possible role in amyotrophic lateral sclerosis. J. Biol.Chem. 284, 13940–13947.

22. Sandelin, E., Nordlund, A., Andersen, P. M., Mark-lund, S. S. & Oliveberg, M. (2007). Amyotrophiclateral sclerosis-associated copper/zinc superoxidedismutase mutations preferentially reduce the repul-sive charge of the proteins. J. Biol. Chem. 282,21230–21236.

23. Tiwari, A., Xu, Z. & Hayward, L. J. (2005). Aberrantlyincreased hydrophobicity shared by mutants of Cu,Zn-superoxide dismutase in familial amyotrophiclateral sclerosis. J. Biol. Chem. 280, 29771–29779.

24. Lindberg, M. J., Bystrom, R., Boknas, N., Andersen,P. M. & Oliveberg, M. (2005). Systematically per-turbed folding patterns of amyotrophic lateral sclero-sis (ALS)-associated SOD1 mutants. Proc. Natl Acad.Sci. USA, 102, 9754–9759.

25. Bruns, C. K. & Kopito, R. R. (2007). Impaired post-translational folding of familial ALS-linked Cu, Znsuperoxide dismutase mutants. EMBO J. 26, 855–866.

26. Rumfeldt, J. A., Lepock, J. R. & Meiering, E. M. (2009).Unfolding and folding kinetics of amyotrophic lateralsclerosis-associated mutant Cu,Zn superoxide dismu-tases. J. Mol. Biol. 385, 278–298.

27. Hawe, A., Sutter, M. & Jiskoot, W. (2008). Extrinsicfluorescent dyes as tools for protein characterization.Pharm. Res. 25, 1487–1499.

28. Nordlund, A. & Oliveberg, M. (2008). SOD1-associat-ed ALS: a promising system for elucidating the originof protein-misfolding disease. HFSP J. 2, 354–364.

29. Wang, Q., Johnson, J. L., Agar, N. Y. & Agar, J. N.(2008). Protein aggregation and protein instabilitygovern familial amyotrophic lateral sclerosis patientsurvival. PLoS Biol. 6, e170.

30. Ratovitski, T., Corson, L. B., Strain, J., Wong, P.,Cleveland, D. W., Culotta, V. C. & Borchelt, D. R.(1999). Variation in the biochemical/biophysicalproperties of mutant superoxide dismutase 1 enzymesand the rate of disease progression in familialamyotrophic lateral sclerosis kindreds. Hum. Mol.Genet. 8, 1451–1460.

31. Ohi, T., Nabeshima, K., Kato, S., Yazawa, S. &Takechi, S. (2004). Familial amyotrophic lateralsclerosis with His46Arg mutation in Cu/Zn superox-ide dismutase presenting characteristic clinical fea-tures and Lewy body-like hyaline inclusions. J. Neurol.Sci. 225, 19–25.

32. Tiwari, A., Liba, A., Sohn, S. H., Seetharaman, S. V.,Bilsel, O., Matthews, C. R. et al. (2009). Metaldeficiency increases aberrant hydrophobicity of mu-tant superoxide dismutases that cause amyotrophiclateral sclerosis. J. Biol. Chem. 58, 2565–2573.

33. Stathopulos, P. B., Rumfeldt, J. A., Scholz, G. A., Irani,R. A., Frey, H. E., Hallewell, R. A. et al. (2003). Cu/Znsuperoxide dismutase mutants associated with amyo-trophic lateral sclerosis show enhanced formation ofaggregates in vitro. Proc. Natl Acad. Sci. USA, 100,7021–7026.

34. Lo, M. C., Aulabaugh, A., Jin, G., Cowling, R., Bard, J.,Malamas, M. & Ellestad, G. (2004). Evaluation offluorescence-based thermal shift assays for hit identi-fication in drug discovery. Anal. Biochem. 332,153–159.

35. DiDonato, M., Craig, L., Huff, M. E., Thayer, M. M.,Cardoso, R. M., Kassmann, C. J. et al. (2003). ALSmutants of human superoxide dismutase form fibrousaggregates via framework destabilization. J. Mol. Biol.332, 601–615.

36. Arnesano, F., Banci, L., Bertini, I., Martinelli, M.,Furukawa, Y. & O'Halloran, T. V. (2004). Theunusually stable quaternary structure of human Cu,Zn-superoxide dismutase 1 is controlled by bothmetal occupancy and disulfide status. J. Biol. Chem.279, 47998–48003.

37. Khare, S. D., Caplow, M. & Dokholyan, N. V. (2004).The rate and equilibrium constants for a multistepreaction sequence for the aggregation of superoxidedismutase in amyotrophic lateral sclerosis. Proc. NatlAcad. Sci. USA, 101, 15094–15099.

38. Chiti, F., Webster, P., Taddei, N., Clark, A., Stefani,M., Ramponi, G. & Dobson, C. M. (1999). Designingconditions for in vitro formation of amyloid proto-filaments and fibrils. Proc. Natl Acad. Sci. USA, 96,3590–3594.

39. Wang, J., Xu, G., Gonzales, V., Coonfield, M.,Fromholt, D., Copeland, N. G. et al. (2002). Fibrillarinclusions and motor neuron degeneration in trans-genic mice expressing superoxide dismutase 1 with adisrupted copper-binding site. Neurobiol. Dis. 10,128–138.

40. Rousseau, E., Dehay, B., Ben-Haiem, L., Trottier, Y.,Morange, M. & Bertolotti, A. (2004). Targetingexpression of expanded polyglutamine proteins tothe endoplasmic reticulum or mitochondria preventstheir aggregation. Proc. Natl Acad. Sci. USA, 101,9648–9653.

41. Kato, S., Hayashi, H., Nakashima, K., Nanba, E., Kato,M., Hirano, A. et al. (1997). Pathological characteriza-tion of astrocytic hyaline inclusions in familialamyotrophic lateral sclerosis. Am. J. Pathol. 151,611–620.

42. Chiti, F., Stefani, M., Taddei, N., Ramponi, G. &Dobson, C. M. (2003). Rationalization of the effects ofmutations on peptide and protein aggregation rates.Nature, 424, 805–808.

43. Zibaee, S., Jakes, R., Fraser, G., Serpell, L. C.,Crowther, R. A. & Goedert, M. (2007). Sequencedeterminants for amyloid fibrillogenesis of humanalpha-synuclein. J. Mol. Biol. 374, 454–464.

44. Khare, S. D., Wilcox, K. C., Gong, P. & Dokholyan, N.V. (2005). Sequence and structural determinants of Cu,Zn superoxide dismutase aggregation. Proteins, 61,617–632.

45. Elam, J. S., Taylor, A. B., Strange, R., Antonyuk, S.,Doucette, P. A., Rodriguez, J. A. et al. (2003). Amyloid-like filaments and water-filled nanotubes formed bySOD1 mutant proteins linked to familial ALS. Nat.Struct. Biol. 10, 461–467.

46. Zetterstrom, P., Stewart, H. G., Bergemalm, D.,Jonsson, P. A., Graffmo, K. S., Andersen, P. M. et al.(2007). Soluble misfolded subfractions of mutantsuperoxide dismutase-1s are enriched in spinal cordsthroughout life in murine ALS models. Proc. NatlAcad. Sci. USA, 104, 14157–14162.

47. Vande Velde, C., Miller, T. M., Cashman, N. R. &Cleveland, D. W. (2008). Selective association ofmisfolded ALS-linked mutant SOD1 with the cyto-plasmic face of mitochondria. Proc. Natl Acad. Sci.USA, 105, 4022–4027.

48. Wang, J., Farr, G. W., Zeiss, C. J., Rodriguez-Gil, D. J.,Wilson, J. H., Furtak, K. et al. (2009). Progressiveaggregation despite chaperone associations of a


mutant SOD1-YFP in transgenic mice that developALS. Proc. Natl Acad. Sci. USA, 106, 1392–1397.

49. Borchelt, D. R., Guarnieri, M., Wong, P. C., Lee, M. K.,Slunt, H. S., Xu, Z. S. et al. (1995). Superoxidedismutase 1 subunits withmutations linked to familialamyotrophic lateral sclerosis do not affect wild-typesubunit function. J. Biol. Chem. 270, 3234–3238.

50. Arndt, V., Rogon, C. & Hohfeld, J. (2007). To be, or notto be–molecular chaperones in protein degradation.Cell Mol. Life Sci. 64, 2525–2541.

51. Jonsson, P. A., Ernhill, K., Andersen, P. M., Berge-malm, D., Brannstrom, T., Gredal, O. et al. (2004).Minute quantities of misfolded mutant superoxide

dismutase-1 cause amyotrophic lateral sclerosis. Brain,127, 73–88.

52. Rousseau, E., Kojima, R., Hoffner, G., Djian, P. &Bertolotti, A. (2009). Misfolding of proteins with apolyglutamine expansion is facilitated by proteasomalchaperones. J. Biol. Chem. 284, 1917–1929.

53. Deng, H. X., Hentati, A., Tainer, J. A., Iqbal, Z.,Cayabyab, A., Hung, W. Y. et al. (1993). Amyotrophiclateral sclerosis and structural defects in Cu,Znsuperoxide dismutase. Science, 261, 1047–1051.

54. Roseman, M. A. (1988). Hydrophobicity of the peptideC = O.H-N hydrogen-bonded group. J. Mol. Biol. 201,621–623.

Documents

Exposure of Hydrophobic Surfaces Initiates Aggregation of Diverse ALS-Causing Superoxide Dismutase-1 Mutants