Systematically perturbed folding patterns of amyotrophic lateral sclerosis (ALS)-associated SOD1 mutants

m, Niklas Bokna s, Peter M. Andersen, and Mikael Oliveberg Mikael J. Lindberg, Roberth Bystro
Departments of Biochemistry and Clinical Neuroscience, Ume University, S-901 87 Ume, Sweden Edited by Alan R. Fersht, University of Cambridge, Cambridge, United Kingdom, and approved May 18, 2005 (received for review March 9, 2005)

Amyotrophic lateral sclerosis is a neurodegenerative syndrome associated with 114 mutations in the gene encoding the cytosolic homodimeric enzyme CuZn superoxide dismutase (SOD). In this article, we report that amyotrophic lateral sclerosis-associated SOD mutations with distinctly different disease progression can be rationalized in terms of their folding patterns. The mutations are found to perturb the protein in multiple ways; they destabilize the precursor monomers (class 1), weaken the dimer interface (class 2), or both at the same time (class 1 2). A shared feature of the mutational perturbations is a shift of the folding equilibrium toward poorly structured SOD monomers. We observed a link, coupled to the altered folding patterns, between protein stability, net charge, and survival time for the patients carrying the mutations.
neurodegenerative disease protein stability

monomer and interface stability; restoring only one of these components may not be sufficient to suppress the disease. Materials All experiments were done under oxidizing conditions, with the intramolecular disulphide linkage between C57 and C146 kept intact. To avoid artifacts from nonnative disulphide crosslinks between the remaining free cysteines C6 and C111, we used the double mutant C6AC111A (SODpwt) as pseudo-WT for further mutant analysis (24). The substitutions C6AC111A cause only a slight decrease of protein stability but show no influence on the SOD folding behavior (25).
Mutant Selection. For most of the 114 known SOD1 mutations, little

familial form of the fatal neurodegenerative disease amyotrophic lateral sclerosis (ALS) has been linked to a large set of mutations in the ubiquitous enzyme CuZn superoxide dismutase (SOD) (1, 2). Although the mechanism by which mutant SOD causes neural death remains elusive (3), several lines of evidence suggest that ALS is a protein-folding disease (413), in analogy with other neurodegenerative disorders like Parkinsons disease, Lund Huntingtons disease, and CreutzfeldtJakobs disease (14). The cytotoxic component of these diseases seems to be ordered oligomeric states, or aggregates of unfolded or misfolded proteins that escape the cellular quality-control and housekeeping mechanisms (15). More detailed elucidation of the molecular events underlying cytotoxic gain of function has been hampered by generally complex folding and aggregation behaviors of the disease-provoking proteins (15); it is not yet clear from which conformational species the cytotoxic pathway originates or whether the toxicity mechanism is common to all protein-folding disorders or disease-specific. In the case of SOD, thermodynamic analysis suggests that the WT protein is a so called three-state dimer (13, 16, 17) in which the individual monomers adopt their correct folded structures also in the absence of the intermolecular interface (18) and the stabilizing Zn and Cu ions (16, 19). Recently, several findings have implicated such immature monomeric SOD species as precursors in the ALS mechanism (1113, 20, 21). To shed more light on the origin of neurotoxicity in ALS, we have compared the folding behavior of 15 ALS-associated SOD mutants with survival times ranging from 1 to 17 years after the onset of the first symptom (Fig. 1). In support of the notion that ALS is a protein-folding disorder, the folding patterns of the ALS-associated mutants show different and characteristic features that can be linked to their respective disease-progression rates. Common to all mutations is a shift of the folding equilibrium toward the denatured monomer. However, the magnitudes and the mechanistic origins of this shift are found to vary; the ALS-associated mutations reduce the stability of the monomer, the strength of the dimer interface, or both. The observation implies that therapeutic strategies (22) to prevent the noxious effects of mutant SOD at a primary protein level (23) must restore at the same time perturbations of both the
9754 9759 PNAS July 12, 2005 vol. 102 no. 28

or no clinical data have been published. The 15 ALS-associated mutants included in this study are the missense mutations from which reliable and sufficient clinical data are available (see supporting information, which is published on the PNAS web site). The patient survival times from the onset of first symptom to death or tracheotomy are given in Table 2.
Protein Preparation. The mutants were constructed on a background of C6AC111A and the monomeric variant F50EG51E C6AC111A, and they were coexpressed with copper chaperone (yCCS) as described in ref. 26. Apo protein was prepared as described in ref. 25, and the standard buffer was 10 mM Mes (pH 6.3) with 10 mM EDTA to maintain the proteins metal free in all steps of the analysis. Gel Filtration. Dimerization was analyzed by gel chromatography on

a Sephadex S-75 column (Amersham Pharmacia), as described in ref. 25. The concentration of eluted protein was 20 M monomer.
Measurements. Equlibrium titrations were done on a J-810 CD

spectrometer (Jasco) or a *-180 instrument (Applied Photophysics, Surrey, U.K.), and the kinetics were measured on an SX-17MV stopped-flow flourimeter (Applied Photophysics). The temperature was 25C, and the measurements were done at a protein concentration of 4 M monomer.
Data Analysis. Monomeric apoSOD was assumed to display a minimal two-state transition between the denatured state (D) and the fully folded monomer (M), KD-M [D][M] kukf, where kf and ku are the refolding and unfolding rate constants, respectively. The plots of log kf and log ku vs. [urea] were fitted by the following:

log k obs log k f k u) log (10logkf

m f[urea]

10 logku

H 2O

m u[urea]

),

[1]

This paper was submitted directly (Track II) to the PNAS ofce. Abbreviations: ALS, amyotrophic lateral sclerosis; SOD, superoxide dismutase.
To

whom correspondence should be addressed. E-mail: [email protected].

2005 by The National Academy of Sciences of the USA

www.pnas.orgcgidoi10.1073pnas.0501957102

pwt construct apoSOD50/51/6/111 (apoSODmono ) that displays classical two-state behavior (25).

kf D| L ; M ku

Scheme 1.

Fig. 1. The structure of homodimeric SOD (PDB ID code 1SPD), showing the positions of the 15 ALS-associated mutations analyzed in this study.
logK where the slopes mu mf m D-M are used as measures of solvent exposure in the equilibrium unfolding process and the activation process of unfolding and refolding, respectively (27). The stability of the apoSOD monomer was derived from kinetic data according to the following: H 2O H 2O k f ) GD-M RTlnK D-M RTln k u H 2O G m D-M [urea], G D-M

Similar folding transitions are observed for the majority of other single-domain proteins (30) and are characterized by V-shaped chevron plots (Fig. 2) and sigmoidal unfolding transitions upon titration with urea (27). Consistently, folding of the pwt homodimeric protein (apoSODdimer ) has been found to proceed through docking of preformed monomers according to the three-state mechanism (25). kf ka 2D L | ; 2M | L ; M2 ku kd Scheme 2.

[2]

and the stability changes upon point mutation were calculated from the following:
mut pwt mut G D-M G D-M GD-M H 2O H 2O 2.3RT(log k f log k u ), pwt G D-M mut G D-M

[3]

where and are the stabilities of the pseudo-WT H 2O H 2O and mutant proteins, respectively, and log k f and log k u are the mutant-induced changes of the rate constants. The corresponding stability loss upon mutation of the dimeric mut mut mut apoSOD ( G 2D-M G D-M G 2M-M ) was calculated 2 2 from the following:
H 2O H 2O mut G2D-M 2.3 RT log k f log k u 2

log k alog k d),

[4]

Aggregation Assay. The aggregation propensity of the apoSOD

Class 1 ALS-Associated Mutants: Selective Destabilization of the Free Monomer. The monomeric state of the ALS-associated mutant
D90A D90A (apoSODmono ) is characterized by a pronounced increase of the unfolding rate constant, whereas the effect on the refolding kinetics is much smaller (Fig. 2). The accompanying shift of the transition midpoint toward lower urea concentration yields a destabilization of the monomeric species corresponding to D90A pwt D90A GD-M GD-M GD-M 0.67 (Eq. 3, Table 1). From the small change of log kf, it can be further concluded that the mutation D90A has only subtle effect on the transition-state but destabilizes mainly the fully structured monomer (25, 27). In terms of the folding process, this result indicates that the mutational perturbation sets in mainly after the rate-limiting step, on the downhill side of the folding barrier (31). Despite the marked effect on the folded monomer, D90A shows no effect on the dimer interface; the D90A chevron plot of apoSODdimer yields a dimer dissociation rate pwt constant (kd) identical to that of apoSODdimer (Fig. 2). Such selective destabilization of the SOD monomer is referred to as class 1 folding. Of the 15 ALS-associated mutants characterized in this study, 4 display class 1 behavior (see supporting information).

mutants was determined according to the protocol of Stathopulos et al. (28), in which aggregation of freshly prepared apo protein was induced by thermal scans and measured by light scattering at 450 nm. The aggregation temperature (Tagg) was defined as the temperature at which the light intensity of the scattered light reaches 0.5 V corresponding to approximately half the amplitude of the aggregation transition. Protein concentration was 2 M, and the pH was adjusted to 6.0 with 20 mM Mes. The temperature gradient in the thermal scans was 0.7C per minute, and the values of Tagg given in Table 2 are the averages of at least two independent measurements. Results
Folding of apoSOD: A Unique Fingerprint for Docking of Preformed Monomers. In contrast to many other homodimeric proteins (29),

the structure of the SOD monomers can form independently of the dimer interface (18, 25). This structural integrity of the SOD monomer is nicely demonstrated by the folding of the monomeric
Lindberg et al.

PNAS July 12, 2005 vol. 102 no. 28 9755

BIOCHEMISTRY

where log kd is the dimer dissociation rate constant and the effect on the associative event log ka was assumed to be zero. Data analysis was performed by using KALEIDAGRAPH software (Synergy Software, Reading, PA).

pwt The experimental hallmark of apoSODdimer is a chevron plot in which the refolding reaction proceeds over the transition state for monomer formation () and the unfolding limb displays a characteristic downward kink at 5 M urea. The kink is due to a Hammond shift of the rate-limiting step to a later transition state () describing dimer dissociation (kd) (see the free-energy profiles in Fig. 2). The nature of the early and late transition states can be independently confirmed by varying the protein concentration; elevated protein concentration decreases only log ku below 5 M urea, but it has no appreciable effect on log kf or the unfolding reaction above 5 M urea (25). Accordingly, the folding process of homodimeric apoSOD displays a unique kinetic fingerprint where all of the microscopic rate constants shown in Scheme 2 can be obtained directly from the chevron data in Fig. 2. From such data, pwt the overall stability of the homodimeric apoSODdimer was recently pwt estimated to G2D-M2 15.7 kcalmol at 0 M urea, and the corresponding stability of the dimer equilibrium was recently pwt approximately 12 kcalmol (Kd 1.5 nM) estimated to G2M-M 2 (25). Together with a first-order dissociation rate constant of log kd 2.8 at 0 M urea, the results yields a second-order rate constant (ka) of 2 106 M1s1 for association of the homodimer, approaching that expected for a purely diffusive process (25). A simplifying feature of a diffusion limited assembly of the SOD homodimer is that ka can be assumed to be independent of mutation (i.e., mutant-induced changes of G2M-M2 can be determined from changes in kd alone).

Fig. 2. The folding kinetic of SOD. (Upper) Chevron plots for pseudo-WT SOD (black) and representative mutants (red) exemplifying the class 1, 2, and 1 2 pwt pwt mutant mutant behavior associated with neurotoxic function in ALS. Units are given in s1. apoSODdimer (F), apoSODmono (), apoSODdimer (), and apoSODmono (). The ts pwt are from Schemes 1 (Eq. 1) and 2 (25, 51), and the dotted line shows specically the urea dependence of log kd for apoSODdimer . The small offset in log kf observed for all monomeric proteins is due to the dimer-splitting substitutions F50EG51E (25). (Lower) Schematic folding free-energy proles at 0 M urea showing the energetic perturbations induced by the mutants. The class 1 mutant D90A shows a selective destabilization of the monomer without affecting the dimer interface (log kd 0). In contrast, the class 2 mutant L144F has no effect on the monomer stability but fails to dimerize. The class 1 2 mutants G41S and I104F affect both the monomer and interface stability. Chevron plots for the other mutants in Tables 1 and 2 are given in the supporting information.

Class 2 ALS-Associated Mutants: Selective Destabilization of the Dimer Interface. In contrast to D90A, the ALS-associated mutation L144F

shows no impact on the monomer stability. Although the limbs of the L144F chevron plot are biased toward higher values, there is no shift of the transition midpoint (Fig. 2). This slight decrease of the folding activation barrier accompanying the mutation L144F is in good agreement with earlier reports on unspecific hydrophobic stabilization of the transition-state ensemble upon increasing the number of solvent-exposed Phe (32). Nevertheless, the mutation L144F fails to dimerize: the chevron plot of the L144F species lacking the interface-splitting substitutions F50EG51E yields unfolding rate constants that are identical to those of the monomeric protein (Fig. 2). The small change in log kf reflects simply the intrinsic energetic contribution of the dimer-splitting mutations F50EG51E (25). The weakened dimer interface of L144F is further indicated by chromatography, in which the elution peak is

pwt (i.e., the proshifted toward the monomeric control apoSODmono tein is partly dissociated under the chromatographic conditions; Fig. 3). In this study, L144F represents the only case of clean class-2 behavior, suggesting that this ALS-provoking mutation is a rare group.

Class 1 2 ALS-Associated Mutants: Destabilization of both Monomer and Dimer Interface. The third, and dominant, class of mutants

displays a combination of class 1 and 2 behavior by affecting both the monomer and interface stability. An intriguing example is provided by the twin mutations G41S and G41D. Despite the close structural appearance of these substitutions, G41S and G41D are characterized by markedly different disease phenoclass with mean survival times of 1 and 17 years, respectively. Both mutations lead to a radical destabilization of the SOD monomer, revealed by greatly increased values of log ku and transition midpoints near 0 M

Table 1. Kinetic parameters for monomeric apoSOD

G, kcalmol
3.03 0.11 1.41 0.05 0.12 0.05 0.10 0.12 0.35 0.05 3.84 0.29 1.16 0.04 1.10 0.04 2.37 0.03 0.60 0.05 2.12 0.02 2.31 0.01 1.25 0.03 1.78 0.04 2.80 0.04 0.50 0.10

H2 O log KD-M

MP, M
1.59 0.06 0.74 0.04 0 0.08 0.07 0.04 0.08 1.89 0.04 0.63 0.04 1.06 0.03 1.16 0.02 0.12 0.04 0.56 0.05 1.14 0.02 0.64 0.03 0.92 0.04 1.56 0.03 0.34 0.04

1.36 0.04 0.97 0.03 1.36 0.04 1.82 0.09 1.36 0.04 1.34 0.01 1.62 0.02 1.63 0.03 1.54 0.01 1.36 0.04 1.51 0.01 1.20 0.01 1.70 0.01 1.54 0.01 1.28 0.01 1.77 0.06

3.59 0.07 2.01 0.03 1.27 0.03 1.89 0.03 1.62 0.02 4.16 0.21 2.47 0.02 2.44 0.03 3.28 0.02 1.80 0.03 3.07 0.02 2.90 0.01 2.62 0.02 2.85 0.03 3.34 0.03 2.14 0.03

1.19 0.09 2.14 0.09 2.16 0.12 1.97 0.09 0.59 0.23 1.38 0.09 1.42 0.09 0.49 0.08 1.79 0.09 0.67 0.08 0.53 0.08 1.31 0.08 0.92 0.09 0.17 0.09 1.86 0.11

from kinetic data (Eq. 1), where mf for all mutants was locked to the WT value of 0.95; this value also is obtained by global t. Calculated from the minima of the chevron plot through tting of Eq. 1 in the linear regime up to 5 M urea. Approximated with the WT value because exact measurements were hampered by poorly dened refolding limbs of the mutant protein. The approximation is justied by the kinetic amplitude data showing that the midpoints for these mutations are close to 0 M urea.
9756 www.pnas.orgcgidoi10.1073pnas.0501957102

Lindberg et al.

survival times for L106V and I104F are 2.3 and 13 years, respectively (Table 2). Data for the other class 1 2 mutants are shown in Tables 1 and 2 and in the supporting information.
Aggregation. The aggregation of the apoSOD dimer shows a marked dependence on pH: e.g., the apparent midpoint (Tagg) for G41D varies from 68C at thermally induced aggregation of apoSODdimer pH 6.3 to 32C at pH 5.4 (data not shown). To obtain a suitable span of the light-scattering profiles we adjusted accordingly the pH to 6.0, yielding aggregation temperatures between 32C and 54C for the mutant proteins (Table 2). As shown in Supplementary materials, the thermal transitions of the SOD mutants are nearly sigmoidal. Even so, the aggregation process is not reversed upon lowering the temperatures after completed scans. In comparison with other systems, the reproducibility of the aggregation data are surprisingly good and the variation in Tagg is typically 2C (Table 2). Preliminary screens by atomic force microscopy indicate that the thermally induced aggregates are small disk-like structures rather than large fibrils (data not shown).

Fig. 3. Size-exclusion chromatograms showing the dimer and monomer pwt pwt L106V composition of apoSODdimer (black), apoSODmono (blue), apoSODdimer (green), L144F and apoSODdimer (red). Dimers elute at 10.5 ml, and folded monomers elute at 12 ml. Consistent with the kinetic data, L144F displays an intermediate elution position indicating that this mutant describes a dynamic equilibrium between the monomeric and homodimeric state.

G41S urea (Fig. 2 and Table 1). As a consequence, apoSODmono and, G41D especially, apoSODmono are only partly able to fold on their own under physiological conditions but sample unfolded conformations a substantial part of their time. The different monomer stabilities of G41S and G41D are to some extent balanced by the accompanying effects on the dimer interface; the chevron plot of G41S is characterized by a substantially weakened interface apoSODdimer pwt (kd is 10 times higher than for apoSODdimer ), whereas G41D apoSODdimer approaches class 1 behavior. Thus, in terms of dimer stability, the two mutants turn out rather similar (Table 2). An important chemical difference between the mutants is, nevertheless, that G41D introduces an extra negative charge. Similar folding patterns are observed for the mutants L106V and I104F that reveal substantially destabilized monomers and dimer interfaces that are L106V weakened to different extent. The unfolding limb of apoSODdimer shows only a small deviation from that of the monomeric species I104F displays log kd (see supporting information), whereas apoSODdimer values close to those of the pseudo-WT dimer (Fig. 2). In this case, however, the substitutions do not alter the proteins net charge and the extent of mutational perturbation on the protein stability is correlated with the associated disease-progression rates: the mean

Discussion
Folding of SOD Reveals a Unique Kinetic Fingerprint Characteristic for a Three-State Dimer. Folding of the apoSOD monomer shows all of

the characteristics of a two-state transition (i.e., the folding process takes place in one step over an extensive free-energy barrier; ref. 27). Such cooperative folding transitions is a shared feature of most small proteins (30) and suggest that the underlying folding freeenergy landscapes of natural proteins are highly optimized (33, 34). The biological advantage of optimized folding cooperativity could be that it reduces the chance of erroneous protein aggregation by preventing the accumulation of partly structured intermediates and excessive conformational breathing of the native state (34, 35). Interestingly, the cooperative formation the apoSOD monomer is also the rate-limiting step for folding of the apoSOD homodimer under physiological conditions, yielding a three state folding mechanism according to Scheme 2. Such independent formation of the monomers before dimerization is also predicted from the relatively small ratio between interfacial and intramonomer contacts in the SOD molecule (29). Other homodimeric proteins with more extensive interfaces form obligatory dimers where folding and association occur simultaneously in one step (29, 36). An important

Table 2. The kinetic parameters for the SOD homodimer

G, kcalmol
4.31 0.12 3.47 0.33 4.47 0.16 4.05 0.22 0.48 0.31 2.65 0.12 0.90 0.11 0.65 0.11 2.98 0.14 2.22 0.12 1.24 0.11 3.62 0.11 2.48 0.12 1.89 0.12 4.56 0.15

Class
12 12 12 12 1 12 1 1 12 1 12 12 12 2 12

MP, M
2.64 0.08 1.19 0.06 0.60 0.30 0.92 0.20 0.47 0.03 2.32 0.04 1.30 0.10 1.39 0.11 2.11 0.13 0.98 0.10 1.11 0.04 1.95 0.14 0.87 0.14 1.66 0.11 1.88 0.05 0.91 0.02

Tagg, C
55.2 0.9 52.3 0.3 53.7 0.1 32.8 0.5 34.5 0.9 45.5 0.4 35.0 2.6 40.0 1.5 35.5 1.4 46.7 1.0 42.8 0.7 43.3 1.2 48.7 0.9 52.2 1.7

Survival time, years

No. of patients (n)

pWT A4V G41D G41S H43R H46R L84V G85R D90A G93A E100G I104F L106V I113T L144F V148G
The

1.03 0.05 0.34 0.05 0.98 0.24 1.55 0.06 1.21 0.15 1.05 0.01 1.15 0.06 1.18 0.06 1.28 0.03 1.39 0.09 1.25 0.02 0.93 0.05 1.09 0.14 1.32 0.06 0.99 0.02 1.34 0.03

2.13 0.15 1.72 1.10 1.12 1.89 1.56 2.14 2.14 1.73 1.92 1.75 0.78 1.23 0.81 0.64

value of the unfolding rate constant at 5.8 M urea. from the minima of the chevron plot through tting of Eq. 1 in the linear regime near the transition region. Average value from two or three measurements. The disease survival time is the time from onset of rst symptom to death or tracheotomy. See supporting information for references on clinical data. For the most frequent mutation D90A, we chose to use survival data for heterozygous patients since genetic studies suggest that the D90A-homozygous patients also carries a disease-modifying gene (50).
Estimated

Lindberg et al.

PNAS July 12, 2005 vol. 102 no. 28 9757

benefit of this relatively simple folding behavior of the apoSOD homodimer is that it allows detailed examination of mutational perturbations from kinetic data alone.
Structural Rationale for the Class 1, 2, and 1 2 Behavior. The class 1 mutation D90A has no detectable effect on the dimer interface and only slightly destabilizes the folded monomer through selectively increased values of the unfolding rate constant, ku (Fig. 2). The behavior is in full accordance with predictions from the x-ray structure (37, 38); the D90 side chain is localized at the protein surface distant from the dimer interface but forms a delocalized set of hydrogen bonds to the backbone amide groups of D92, G93 and V94 across the hairpin between strands 5 and 6. Truncation of these hydrogen bonds is expected to selectively destabilize the isolated monomer without affecting the dimer interface significantly. The selective effect on the dimer interface observed for the class 2 mutation L144F can be accounted for by the slight structural differences observed from comparison of the monomeric and dimeric crystal structures (37, 38). In the monomer, L144 protrudes freely into the solvent, whereas in the dimeric state, the side chain is forced into close contact with its neighboring residues through alterations of the backbone. As a consequence of this structural adjustment, insertion of the larger Phe becomes sterically obstructive, leading to a destabilization of the dimer upon mutation L144F. The strong long-range coupling between the geometry at position 144 and the dimer interface could be mediated by the flanking C57C146 disulphide link: the integrity of this SOS bond is critical for the dimeric structure, and its cleavage has been demonstrated to promote monomerization (25, 39). Consistently, the folding patterns accompanying the class 1 2 mutations G41D and G41S can be assigned to conformational strain of the polypeptide backbone. In the WT protein, the backbone adopt anomalous torsion angles ( and ) at position 41 that are only suitable for Gly. Hence, mutation to the less bendable residues D or S leads to backbone distortions that can propagate over considerable distances in the monomer structure and also affect the strength of the dimer interface. In the case of I104F and L106V, the introduction of oversized side chains at completely buried positions destabilizes the folded structure through steric clashes. The larger impact on kd observed for L106V is most likely due to its close proximity to the interface region, where the obstructive side chain more directly interfere with the geometry required for strong dimer assembly. In support of this explanation, the neighboring replacements A4V and I113T have been observed to markedly distort the crystal packing of the dimer interface (12). The impact of the other mutations listed in Tables 1 and 2 can be rationalized along similar lines. Together, it is thus possible to link in a straightforward manner changes in folding patterns with changes in protein structure for the ALS-associated SOD mutants. Of particular interest is the structural adjustment upon dimerization, explaining how seemingly benign local perturbations can still affect interactions that are distant in the tertiary structure. Common Denominator for the ALS-Associated Mutants: Shift of the Folding Equilibrium Toward Denatured Monomers. The shared fea-

Fig. 4. Schematic representation of the three-state folding reaction of SOD, including the different effects of the 15 ALS-associated mutations characterized in this study. D, unfolded protein; M, folded monomer; M2, homodimer. Common for the class 1, 2, and 1 2 mutations is a shift of the folding equilibrium toward unfolded and partly structure monomers, indicating that the starting material for the toxic gain of function is to be found within this broad ensemble of poorly dened species.

especially those in the denatured ensemble. The overall small effects on the refolding kinetics, show further that the mutational perturbations have only limited impact on the early folding events but start to develop on the downhill side of the folding barrier, just before the formation of the fully developed monomer (27). Perturbation of these late folding events contributes to make the mutant structures poorly wrapped up and increases its propensity to unfold locally. That is, besides shifting the equilibrium toward the denatured ensemble, the mutants more easily sample the partly structured states decorating the downhill side of the folding barrier (34). The starting material for noxious gain of function could then be either species in the denatured ensemble or partly structured states on the monomer folding pathway. In both instances, this includes misfolded species that normally reside as high-energy states (Fig. 4). In combination with reductive cleavage of the C57OC146 disulphide bond (40), the severely destabilized monomers of several class 1 2 mutants will not be able to adopt their folded structures without support from the dimer interface: the ALS-associated SOD mutants have been found to exhibit an increased susceptibility to disulphide reduction (40), contributing to further decrease their unfolding midpoints by 1 M urea (M.J.L. and M.O., unpublished data; see midpoints in Table 1). Accordingly, these monomers are likely to sample an extensive repertoire of denatured and potentially promiscuous states in vivo, providing a plausible link to their enhanced cytotoxicity. Rather than provoking misfolding or changing the folding trajectory, the diseaseassociated mutations of SOD simply increase the cellular load of denatured protein through energetic perturbations of the folding free-energy profile. The results seem thus to have implications also

tures of the ALS-linked mutations in Tables 1 and 2 hold information about from which point in the folding process the toxic pathway starts. The most pronounced feature is a common shift of the folding equilibria toward the ensemble of unfolded and partly structured species preceding the fully folded apo monomer (Fig. 4). For class 1 mutations, the selective destabilization of the folded monomer increases the occupancy of denatured protein through decreased values of GD-M, whereas for class 2 mutations the equally selective shift of the dimer equilibria (G2M-M2) increases the population of both denatured and folded monomers. Analogously, for class 1 2 mutations, the combined effect on GD-M and G2D-M2 increases the occupancy of all monomeric species and
9758 www.pnas.orgcgidoi10.1073pnas.0501957102

Fig. 5. Plots of protein stability changes (G) vs. average survival time after diagnosis. G for ALS-associated mutations that do not alter the net charge of SOD shows a high correlation with survival time (R 0.91). Increasing the net charge of the protein causes a shift toward longer survival time, whereas decreasing the charge has the opposite effect. This behavior is expected for a disease mechanism associated with protein aggregation (43).

Lindberg et al.

for other neurodegenerative conditions where the target proteins are yet intractable to detailed experimental analysis.
Link Among Protein Stability, Net Charge, and Disease Progression.

We have indicated from simplistic equilibrium analysis that the extent of stability loss of the ALS-associated SOD mutations is related to the severity of the disease (26). In this study, we confirm and lend further detail to this finding. For mutations that do not alter the net charge, we observe a clear relation between protein stability (Eq. 4) and mean survival time after the onset of the first symptom: the lower the stability, the shorter the survival time (Fig. 5). Upon examination of the data, the immediate impression is that the relation is not linear but hints a threshold at G 1.5 kcalmol. Below this threshold, the survival times drastically drop to a constant level of 4 years. However, such detailed interpretations of the data are currently subject to considerable uncertainty because they rely partly on features within the statistical errors of the patient survival times (Table 2). Even so, it is clear that severely destabilized mutants are linked to short survival times (R 0.91). The result has the signature of a disease-provoking event related to protein aggregation (4143). In support of such a disease model, the stability pattern shown in Fig. 5 is broken systematically by the mutations that change also the net charge. With a pI 4.85.1 (44), SOD holds a net negative charge under physiological conditions, and incrementing this repulsive charge through G41D seems to slow down the disease progression (Fig. 5). Vice versa, the mutations that decrease the magnitude of the net negative charge are linked to survival times that are shorter than expected from protein stability alone. The effect is particularly clear for the moderately destabilized mutations G85R and D90A, whereas the more severely destabilized E100G inevitably converges with the other low stability mutations (Fig. 5). Independent evidence for a coupling between net charge and pathogenic propensity is provided from the genetics; of all known ALS-associated SOD mutants, 30 reduce the magnitude of the net negative charge, whereas only 8 produce an increase (45). Thus, familial ALS seems to respond to a combination of low protein stability and reduced magnitude of the net negative charge, which are characteristics that have been found to be critical also in other neurodegenerative disorders (43, 46). Although, the effect of
1. Rosen, D. R., Siddique, T., Patterson, D., Figlewicz, D. A., Sapp, P., Hentati, A., Donaldson, D., Goto, J., ORegan, J. P., Deng, H. X., et al. (1993) Nature 362, 5962. 2. McCord, J. M. & Fridovich, I. (1969) J. Biol. Chem. 244, 60496055. 3. Brown, R. H., Jr., (1998) Nat. Med. 4, 13621364. 4. Shinder, G. A., Lacourse, M. C., Minotti, S. & Durham, H. D. (2001) J. Biol. Chem. 276, 1279112796. 5. Okado-Matsumoto, A. & Fridovich, I. (2002) Proc. Natl. Acad. Sci. USA 99, 90109014. 6. Durham, H. D., Roy, J., Dong, L. & Figlewicz, D. A. (1997) J. Neuropathol. Exp. Neurol. 56, 523530. 7. Bruijn, L. I., Houseweart, M. K., Kato, S., Anderson, K. L., Anderson, S. D., Ohama, E., Reaume, A. G., Scott, R. W. & Cleveland, D. W. (1998) Science 281, 18511854. 8. Johnston, J. A., Dalton, M. J., Gurney, M. E. & Kopito, R. R. (2000) Proc. Natl. Acad. Sci. USA 97, 1257112576. 9. Kunst, C. B., Mezey, E., Brownstein, M. J. & Patterson, D. (1997) Nat. Genet. 15, 9194. 10. DiDonato, M., Craig, L., Huff, M. E., Thayer, M. M., Cardoso, R. M., Kassmann, C. J., Lo, T. P., Bruns, C. K., Powers, E. T., Kelly, J. W., et al. (2003) J. Mol. Biol. 332, 601615. 11. Rakhit, R., Crow, J. P., Lepock, J. R., Kondejewski, L. H., Cashman, N. R. & Chakrabartty, A. (2004) J. Biol. Chem. 279, 1549915504. 12. Hough, M. A., Grossmann, J. G., Antonyuk, S. V., Strange, R. W., Doucette, P. A., Rodriguez, J. A., Whitson, L. J., Hart, P. J., Hayward, L. J., Valentine, J. S. & Hasnain, S. S. (2004) Proc. Natl. Acad. Sci. USA 101, 59765981. 13. Furukawa, Y., Torres, A. S. & OHalloran, T. V. (2004) EMBO J. 23, 28722881. 14. Stefani, M. & Dobson, C. M. (2003) J. Mol. Med. 81, 678699. 15. Dobson, C. M. (2003) Nature 426, 884890. 16. Stroppolo, M. E., Malvezzi-Campeggi, F., Mei, G., Rosato, N. & Desideri, A. (2000) Arch. Biochem. Biophys. 377, 215218. 17. Malvezzi-Campeggi, F., Stroppolo, M. E., Mei, G., Rosato, N. & Desideri, A. (1999) Arch. Biochem. Biophys. 370, 201207. 18. Banci, L., Benedetto, M., Bertini, I., Del Conte, R., Piccioli, M. & Viezzoli, M. S. (1998) Biochemistry 37, 1178011791. 19. Strange, R. W., Antonyuk, S., Hough, M. A., Doucette, P. A., Rodriguez, J. A., Hart, P. J., Hayward, L. J., Valentine, J. S. & Hasnain, S. S. (2003) J. Mol. Biol. 328, 877891. 20. Arnesano, F., Banci, L., Bertini, I., Martinelli, M., Furukawa, Y. & OHalloran, T. V. (2004) J. Biol. Chem. 279, 4799848003. 21. Lynch, S. M., Boswell, S. A. & Colon, W. (2004) Biochemistry 43, 1652516531. 22. Cohen, F. E. & Kelly, J. W. (2003) Nature 426, 905909. 23. Ray, S. S., Nowak, R. J., Brown, R. H., Jr., & Lansbury, P. T., Jr. (2005) Proc. Natl. Acad. Sci. USA 102, 36393644. 24. Lepock, J. R., Frey, H. E. & Hallewell, R. A. (1990) J. Biol. Chem. 265, 2161221618.

protein charge is effectively screened at long distances in the aqueous interior of the cell, the electrostatic component of even single charge alterations could be considerable at close range. More detailed elucidation of these factors is currently limited by the statistics of the clinical data; survival times are so far available for only half of the reported ALS-associated SOD mutations, and only 12 of these survival times are based on data from 10 patients.
The Role of Protein Aggregation. As a rule, protein aggregation is

favored by mutations that bring the net charge of the protein closer to neutrality (46). Following the notion that ALS is a proteinaggregation disease, it is interesting to note that the two charge alterations G41D and D90A that are outliers in terms of protein stability fall on their predicted positions in a plot of survival time vs. aggregation temperature. That is, the mutant with the highest net charge yields the lowest aggregation propensity (Table 2 and supporting information). However, on the whole, the data show no relation between survival time and in vitro aggregation. A major reason is that the mutations A4V and V148G yield aggregation temperatures that are much higher than anticipated from their thermodynamic and chemical properties alone. One possibility is that these highly destabilized mutants aggregate by alternative pathways involving an early decomposition of the starting material into small precursor aggregates (47, 48). Such local decomposition does not need to show up in the light-scattering experiments and could contribute to underestimate the intrinsic stickiness of the free monomers. Another explanation is that homoaggregation as measured under oxidizing conditions in the test tube poorly reflects the in vivo situation where noxious gain of function could arise from disulphide reduced species or from more complex (mixed) aggregation processes involving both other proteins (5) and membrane lipids (49). Common to all such associative mechanisms is nevertheless that they are expected to depend on net charge and protein stability (50).
We thank Stefan Marklund for stimulating discussions and Katarina Wallgren for technical assistance. This work was supported by the Swedish Research Council, Go ran Gustafssons Stiftelse, and Bertil Ha llstens Research Foundation (Hja rnfonden, Uppsala).
25. Lindberg, M. J., Normark, J., Holmgren, A. & Oliveberg, M. (2004) Proc. Natl. Acad. Sci. USA 101, 1589315898. 26. Lindberg, M. J., Tibell, L. & Oliveberg, M. (2002) Proc. Natl. Acad. Sci. USA 99, 1660716612. 27. Fersht, A. R. (1999) Structure and Mechanism in Protein Science: A Guide to Enzyme Catalysis and Protein Folding (Freeman, New York). 28. Stathopulos, P. B., Rumfeldt, J. A., Scholz, G. A., Irani, R. A., Frey, H. E., Hallewell, R. A., Lepock, J. R. & Meiering, E. M. (2003) Proc. Natl. Acad. Sci. USA 100, 70217026. 29. Levy, Y., Wolynes, P. G. & Onuchic, J. N. (2004) Proc. Natl. Acad. Sci. USA 101, 511516. 30. Jackson, S. E. (1998) Fold Des. 3, R81R91. 31. Hedberg, L. & Oliveberg, M. (2004) Proc. Natl. Acad. Sci. USA 101, 76067611. 32. Viguera, A. R., Vega, C. & Serrano, L. (2002) Proc. Natl. Acad. Sci. USA 99, 53495354. 33. Onuchic, J. N. & Wolynes, P. G. (2004) Curr. Opin. Struct. Biol. 14, 7075. 34. Lindberg, M., Tangrot, J. & Oliveberg, M. (2002) Nat. Struct. Biol. 9, 818822. 35. Jacob, M., Schindler, T., Balbach, J. & Schmid, F. X. (1997) Proc. Natl. Acad. Sci. USA 94, 56225627. 36. Shakhnovich, E. I. (1999) Nat. Struct. Biol. 6, 99102. 37. Ferraroni, M., Rypniewski, W., Wilson, K. S., Viezzoli, M. S., Banci, L., Bertini, I. & Mangani, S. (1999) J. Mol. Biol. 288, 413426. 38. Deng, H. X., Hentati, A., Tainer, J. A., Iqbal, Z., Cayabyab, A., Hung, W. Y., Getzoff, E. D., Hu, P., Herzfeldt, B., Roos, R. P., et al. (1993) Science 261, 10471051. 39. Doucette, P. A., Whitson, L. J., Cao, X., Schirf, V., Demeler, B., Valentine, J. S., Hansen, J. C. & Hart, P. J. (2004) J. Biol. Chem. 279, 5455854566. 40. Tiwari, A. & Hayward, L. J. (2003) J. Biol. Chem. 278, 59845992. 41. Otzen, D. E., Kristensen, O. & Oliveberg, M. (2000) Proc. Natl. Acad. Sci. USA 97, 99079912. 42. Thirumalai, D., Klimov, D. K. & Dima, R. I. (2003) Curr. Opin. Struct. Biol. 13, 146159. 43. Chiti, F., Stefani, M., Taddei, N., Ramponi, G. & Dobson, C. M. (2003) Nature 424, 805808. 44. Wenisch, E., Vorauer, K., Jungbauer, A., Katinger, H. & Righetti, P. G. (1994) Electrophoresis 15, 647653. 45. Andersen, P. M., Sims, K. B., Xin, W. W., Kiely, R., ONeill, G., Ravits, J., Pioro, E., Harati, Y., Brower, R. D., Levine, J. S., et al. (2003) Amyotroph. Lateral Scler. Other Motor Neuron Disord. 4, 6273. 46. Chiti, F., Calamai, M., Taddei, N., Stefani, M., Ramponi, G. & Dobson, C. M. (2002) Proc. Natl. Acad. Sci. USA 99, 1641916426, Suppl. 4. 47. Otzen, D. E. & Oliveberg, M. (2004) Protein Sci. 13, 14171421. 48. Vaiana, S. M., Palma-Vittorelli, M. B. & Palma, M. U. (2003) Proteins 51, 147153. 49. Sparr, E., Engel, M. F., Sakharov, D. V., Sprong, M., Jacobs, J., de Kruijff, B., Hoppener, J. W. & Killian, J. A. (2004) FEBS Lett. 577, 117120. 50. Parton, M. J., Broom, W., Andersen, P. M., Al-Chalabi, A., Nigel Leigh, P., Powell, J. F. & Shaw, C. E. (2002) Hum. Mutat. 20, 473. 51. Sanchez, I. E. & Kiefhaber, T. (2003) J. Mol. Biol. 325, 367376.

Lindberg et al.

PNAS July 12, 2005 vol. 102 no. 28 9759

BIOCHEMISTRY