Fibrillation precursor of superoxide dismutase 1 revealed by gradual tuning of the protein-folding equilibrium

Abstract

Although superoxide dismutase 1 (SOD1) stands out as a relatively soluble protein in vitro, it can be made to fibrillate by mechanical agitation. The mechanism of this fibrillation process is yet poorly understood, but attains considerable interest due to SOD1’s involvement in the neurodegenerative disease amyotrophic lateral sclerosis (ALS). In this study, we map out the apoSOD1 fibrillation process from how it competes with the global folding events at increasing concentrations of urea: We determine how the fibrillation lag time (τ_lag) and maximum growth rate (ν_max) depend on gradual titration of the folding equilibrium, from the native to the unfolded state. The results show that the agitation-induced fibrillation of apoSOD1 uses globally unfolded precursors and relies on fragmentation-assisted growth. Mutational screening and fibrillation m-values (∂ log τ_lag/∂[urea] and ∂ log ν_max/∂[urea]) indicate moreover that the fibrillation pathway proceeds via a diffusely bound transient complex that responds to the global physiochemical properties of the SOD1 sequence. Fibrillation of apoSOD1, as it bifurcates from the denatured ensemble, seems thus mechanistically analogous to that of disordered peptides, save the competing folding transition to the native state. Finally, we examine by comparison with in vivo data to what extent this mode of fibrillation, originating from selective amplification of mechanically brittle aggregates by sample agitation, captures the mechanism of pathological SOD1 aggregation in ALS.

Intracellular aggregation of the homodimeric enzyme superoxide dismutase (SOD1) is a key pathological hallmark of amyotrophic lateral sclerosis (ALS). Despite this coupling with protein-aggregation disease, SOD1 stands out as a relatively soluble protein in vitro: Mutations that are destabilized to the extent that they fail to fold can be left for weeks in physiological buffer without aggregating. Studies of SOD1 aggregation need then to involve certain stimuli that prompts the process to occur (1). One such stimulus is mechanical agitation. Like many other proteins, SOD1 can be driven to form amyloid-like fibrils by shaking or stirring the sample (2–4). Although these experiments may at first seem artificial, they have come to play a key role in our understanding of the aggregation mechanism: Protein aggregation is a high-dimensional process and, as such, its predicted outcome relies on knowledge about all competing pathways (5). In the case of SOD1, this high dimensionality is underlined by the observation that the structures of the fibrillar aggregates are polymorphic and vary with mutation (6). Competing aggregation pathways are also indicated at macroscopic level where the morphology of the SOD1 aggregates can be changed from filamentous to branched protofibrillar or amorphous by replacing the mechanical agitation with incubation in organic solvent (7), addition of lipids (8), or allowing oxidative cross-linking of the protein’s solvent accessible cysteines (9). Non-amyloid aggregates can also be promoted without agitation in physiological buffer by simply increasing the protein concentration to 95–950 μM and extending the incubation times (10). More detailed insight into the SOD1 aggregation mechanism has so far been precluded by difficulty to identify at structural level the monomeric precursor material: Native SOD1 is a metallated dimer with a complex maturation and degradation pathway, including a plethora of intermediate states (11, 12). As observed for β₂-microglobulin (13) and TTR (14); however, it is often assumed that the precursor for SOD1 aggregation is a ruptured apo species, which either exposes sticky material through local unfolding of the central β-barrel (15, 16) cross-links oxidatively via C6/C111 (9), polymerizes via the long active site loops (17), or promotes aggregation by a mixture of mechanisms (10). There is also alternative evidence, based on the congruent effects of ALS-associated mutations on protein stability, which implicate the globally unfolded state as the disease precursor in vivo (18). To distinguish between these possibilities, we examine here the agitation-induced aggregation of apoSOD1 with the toolbox from protein folding: We use linear free-energy relations and urea titration to determine how the fibrillation lag time (τ_lag) and maximum growth rate (ν_max) correlates with the occupancy of states in the unimolecular folding equilibrium. Occupancies of states are determined from chevron kinetics and NMR, and the experiments were done at pH 6.3 and 37 °C to match as far as possible physiological conditions. The results show that the fibrillation process of SOD1 relies on globally unfolded monomers (D), and linear correlations between log τ_lag, log ν_max and log [D] indicate that the fibril growth is assisted by mechanical fragmentation and holds no information about primary nucleation events. Systematic E-scanning of the SOD1 primary sequence suggests moreover that the globally unfolded monomers associate first diffusely to the fibrillar ends, in analogy with the general two-step association pathway for protein–protein binding.

Results and Discussion

Proteins.

Three commonly implicated driving forces for SOD1 aggregation are oxidative cross-linking by the protein’s cysteines (9), assembly via the structurally flexible active site loops (19, 20), and exposure of sticky sequence regions by rupture of the protein’s central β-barrel (6, 21). To determine the effects of these factors on agitated fibrillation in vitro, we examined three different SOD1 variants (Fig. 1): The apoSOD1 monomer containing the native disulphide bond C57–C146 but with the free cysteines C6 and C111 substituted for A to assure folding reversibility (apoSOD1^C57–C146) (22), the reduced apoSOD1 monomer lacking the disulphide bond (apoSOD1^SH SH), and a variant in which the active site loops IV and VII have been removed by protein engineering (apoSOD1^ΔIV ΔVII) (23). Upon incubation at 25 μM under agitation by a magnetic stirrer at pH 6.3 and 37 °C, all three proteins aggregate into fibrils of indistinguishable morphology (Fig. 1). Accordingly, we conclude that the C57–C146 disulphide bond and loops IV/VII are not required for fibrillation, and that the fibrillar core is composed from segments of the β-barrel, which is the only common denominator of the three structures. The results are in good agreement with mass-spectroscopy data from Furukawa and co-workers, which identify a fibrillar core of β-strand segments, coated by loops IV and VII protruding loosely into the solvent (6).

Fig. 1.

Crystal structures and fibril morphologies of the apoSOD1 variants examined. The apoSOD1 monomer with intact C57–C146 crosslink (apoSOD1^C57–C146), the corresponding protein in TCEP with reduced C57 and C146 moieties (apoSOD1^SH SH), and the variant with loops IV (green) and VII (blue) removed by protein engineering (apoSOD1^ΔIV ΔVII). Structures constructed from PDB entry 2XJK. The EM micrographs of the respective protein fibrils, produced by agitation in pure pH 6.3 buffer at 37 °C. Scale bars are 200 nm.

Tuning of the Folding Equilibrium.

The question is then to what extent the SOD1 barrel needs to rupture—or unfold—to become fibrillation competent. To find out, we gradually destabilized the SOD1 structure by urea titration while measuring the response on the aggregation kinetics. Determination of the concentrations of native (N) and unfolded species (D) at each urea concentration was by standard two-state analysis (12, 21, 22, 24, 25).

Scheme 1 where D and N interconvert via a single transition state (‡), k_f, and k_u are the refolding- and unfolding-rate constants, respectively, and k_down is the kinetic pre-factor (26). The equilibrium constant for folding is here:

[1]

and

[2]

where , , and are the values at 0 M urea, and m_D–N = m_u - m_f = ∂ log K_D–N/∂[urea] = ∂ log k_u/∂[urea]-∂ log k_f/∂[urea] define the urea dependence. Consistent with the linear relations in Eqs. 1 and 2, apoSOD1^C57–C146, apoSOD1^SH SH and apoSOD1^ΔIV ΔVII yield all characteristically V-shaped chevron plots under the conditions of the fibrillation assays at 37 °C (Fig. 2 and Table 1). We investigate here how these linear changes of log K_D-N correlate with changes of the fibrillation kinetics; i.e., we vary the occupancy of putative fibrillation-prone material by urea titration (Fig. 2 and Table 1). A benefit of this approach is that it allows larger variation of protein concentration than conventional aggregation assays. With a fixed concentration of 25 μM apoSOD1^ΔIV ΔVII, which is the most stable construct in this study, urea titration permits screening of [N] = 25 μM to 100 pM and [D] = 35 nM to 25 μM. The less-stable variants apoSOD1^C57–C146 and apoSOD1^SH SH yield correspondingly spans of [D] = 0.4 μM to 25 μM and [D] = 8.5 μM to 25 μM, respectively. Thus, to optimize the dynamic range of the fibrillation assays, we choose apoSOD1^ΔIV ΔVII for further experiments as described next.

Fig. 2.

Folding and fibrillation data. (A) Chevron plots of apoSOD1^C57–C146, apoSOD1^SH SH, and apoSOD1^ΔIV ΔVII showing fits of log k_f and log k_u [Eq. 8]. The curvatures in the unfolding limbs at high [urea] are due to transition-state shifts or native-state fraying and were excluded from the fits. The dotted line indicates the urea concentration where D reaches full occupancy. Units are in s^-1. (B) Concentration of D in the fibrillation assays as calculated from the chevron data of apoSOD1^ΔIV ΔVII. Total protein concentration is 25 μM, and units are in M. (C) Fibrillation lag times (τ_lag) of apoSOD1^ΔIV ΔVII vs. [urea] (open circles), derived from fibrillation time courses as described in SI Text, Analysis of Fibrillation Time Courses. The intrinsic [urea] dependence of log τ_lag was measured at full occupancy of the globally unfolded state (D) above 4 M urea. Subtraction of this [urea] dependence yields the direct relation between log τ_lag and log [D] (closed circles). Under these conditions, ∂ log τ_lag/∂[urea] = m_τlag = 0.13 ± 0.06. Units are in s. (D) Corresponding data and normalisation procedure for the elongation rate constants (ν_max) of apoSOD1^ΔIV ΔVII. ∂ log ν_max/∂[urea] = m_ν max = -0.17 ± 0.05. Units are in s^-1.

Table 1.

Kinetic parameters and protein stabilities. Rate constants are in units of s^-1

	apoSOD1C57–C146	apoSOD1SH SH	apoSOD1ΔIVΔVII
*	−0.89 ± 0.03	−1.26 ± 0.08	0.13 ± 0.03
mf* (M-1)	−0.92 ± 0.03	−0.92 ± 0.03 ∥	−0.82 ± 0.02
*	−2.68 ± 0.01	−1.54 ± 0.02	−2.72 ± 0.04
mu* (M-1)	0.44 ± 0.01	0.36 ± 0.01	0.22 ± 0.01
MP † (M)	1.32 ± 0.01	0.22 ± 0.07	2.74 ± 0.05
mD-N ‡ (M-1)	1.36 ± 0.03	1.28 ± 0.07	1.04 ± 0.02
§	−1.79 ± 0.04	−0.28 ± 0.08	−2.85 ± 0.05
¶  (kcal/mol)	2.54 ± 0.07	0.40 ± 0.05	4.04 ± 0.07
MP ** (M)	n.d.	n.d.	2.67 ± 0.04/2.75 ± 0.04
mD-N ** (M-1)	n.d.	n.d.	0.89 ± 0.07/0.98 ± 0.07
MP †† (M)	1.53 ± 0.06	‡‡	2.88 ± 0.06
mD-N †† (M-1)	1.18 ± 0.23	‡‡	0.93 ± 0.09

^*Derived from chevron data according to Eq. 8.

^†Transition midpoint derived from the intersection between log k_f and logk_u.

^‡m_D-N = m_u - m_f [Eq. 2].

^§Calculated from Eq. 2.

^¶.

^∥Locked to m_f of apoSOD1^C57–C146.

^**Derived from NMR data in SI Text, NMR and Equilibrium Unfolding Analysis.

^††Derived from equilibrium unfolding data in SI Text, NMR and Equilibrium Unfolding Analysis.

^‡‡Lack of folded-state baseline prevents reliable estimate.

Verification of Two-State Behavior by Heteronuclear Single Quantum Correlation (HSQC) NMR.

As an independent control of the two-state folding behaviour of apoSOD1^ΔIV ΔVII, we monitored at equilibrium the urea unfolding with HSQC NMR at 37 °C. The results show that the unfolding of apoSOD1^ΔIV ΔVII is a clean first-order transition: The HSQC spectra obtained in the transition region are in all details linear combinations of the native and denatured cross peaks, with no trace of intermediate species (SI Text, NMR and Equilibrium Unfolding Analysis and Fig. S1). Moreover, plots of the levels of N and D vs. [urea] yield coinciding midpoints at MP ≈ 2.7 M urea and equilibrium m_D-N values that match within experimental errors the kinetic m_D-N from chevron data [Eqs. 1 and 2] (SI Text, NMR and Equilibrium Unfolding Analysis, Figs. S2 and S3, and Table 1). The locally unfolded and weakly populated intermediate (< 1%) revealed previously by relaxation-dispersion NMR (16) and hydrogen deuterium-exchange analysis (23), seems thus to remain at high-energy throughout the urea-unfolding transition at 37 °C.

Agitated SOD1 Fibrillation.

Like several other disease-associated proteins and peptides (5), the fibrillation process of apoSOD1 displays a characteristic lag phase followed by a sigmoidal increase of Thioflavin T (ThT) fluorescence (4) (SI Text, Analysis of Fibrillation Time Courses and Fig. S4). Determination of the fibrillation lag times (τ_lag) and maximum growth rates (ν_max) of apoSOD1^ΔIV ΔVII was by standard procedures (SI Text, Analysis of Fibrillation Time Courses), and the resulting plots of τ_lag and ν_max vs. [urea] are shown in Fig. 2. Data shows that the fibrillation kinetics is, on the whole, accelerated by urea addition and that the apoSOD1^ΔIV ΔVII construct maintains its basic fibrillation behaviour also under severely denaturing conditions at 8 M urea (Fig. 2). We find also that all of the SOD1 monomers are trapped in fibrillar form when the reaction has come to completion (SI Text, Analysis of Fibrillation Time Courses) and that the morphologies of the fibrils produced at 8 M urea are indistinguishable from those observed in pure buffer (SI Text, Analysis of Fibrillation Time Courses and Fig. S5). Taken together, this shows that agitated fibrillation is a remarkably robust process, which even persists under solvent conditions that strongly bias the folding equilibrium toward D (Scheme 1). At 8 M urea, the value of K_D–N = [D]/[N] is 0.3·10⁶ (Table 1), corresponding to nearly full occupancy of 25 μM D and merely pM levels of N. Similarly, low levels are expected for the transient intermediate visible by NMR in the absence of urea (16, 23), as this is close to N on the folding coordinate. On this basis, we conclude that the precursor for apoSOD1 fibrillation is not the native protein or the transient intermediate, but the globally unfolded state. Analogous fibrillation from globally unfolded material has been observed for the WW domain (27) and β₂-microglobulin at low pH (28). An attractive, reductionist feature of this interpretation is that it renders apoSOD1 the same fibrillation mechanism as disordered peptides. It also explains why the apoSOD1 fibrils are polymorphic and so readily swap barrel segments in their spines upon point mutation (6): Globally unfolded precursors allow larger malleability than structurally restricted intermediates.

Accounting for the Urea-Dependence of the Fibrillation Kinetics.

From the urea titrations of apoSOD1^ΔIV ΔVII in Fig. 2, it is evident that ν_max reaches its highest values near, or just after, the transition midpoint, and then slowly decreases with increasing [urea]. A matching, opposite change is described by τ_lag. Similar trends are seen for apoSOD1^C57–C146 and apoSOD1^SH SH, albeit that their low-transition midpoints make data less amenable to detailed analysis (SI Text, Normalization of τ_lag and ν_max to log K_D-N and Fig. S6). One explanation for the ν_max and τ_lag maxima would be that the protein populates an aggregation-prone intermediate in the unfolding transition region, which then gradually disappears at higher denaturant concentrations (29). Since there is no trace of intermediates in the NMR analysis, however, it is more likely that the maxima of ν_max result from a trade-off between (i) increased occupancy of D and (ii) intrinsic suppression of fibrillation by the added urea itself. Such suppression by urea is expected for all processes involving burial of solvent-exposed surface area, and there is no reason to believe that folding and fibrillation should be different in this respect. To compensate for the urea dependence, linear functions were fitted the log ν_max and log τ_lag values above [urea] = 4 M where D has full occupancy (Fig. 2). These baselines were subsequently subtracted from the data to produce graphs of log ν_max and log τ_lag vs. [urea] that are linked directly to the occupancy of species in the folding equilibrium (Fig. 2).

Evidence for Fragmentation-Assisted Fibrillation.

The urea-compensated values of log ν_max and log τ_lag show overall linear dependencies on log[D] (Fig. 3), consistent with observations from conventional fibrillation assays where protein concentration is varied by mass (10, 30, 31). A striking feature of the plots in Fig. 3 is that the dependence on protein concentration is very weak with slopes of ∂ log ν_max/∂ log[D] = 0.56 ± 0.03 and ∂ log τ_lag/∂ log[D] = -0.47 ± 0.03. Similarly, low values of ∂ log τ_lag/∂ log[D] are observed for other proteins and peptides with globally unfolded precursors (30). These weak protein-concentration dependences are formally inconsistent with a primary nucleation mechanism where the formation of primary nuclei from soluble monomers is slower than their elongation (32): In the case of apoSOD1^ΔIV ΔVII, it would suggest primary nuclei with an aggregation number of n = 0.5 monomers. Rather, the behavior indicates the involvement of secondary fibrillation pathways like heterogeneous nucleation on fibril surfaces and fragmentation (32). Of particular interest is the role of fragmentation, since the apoSOD1 fibrillation depends critically on mechanical agitation. To examine this further, we use the analytical expressions for ν_max and τ_lag presented by Knowles et al. (30), which, in this case, translate to:

[3]

[4]

where C₊ ∝ k_n[D]^n-1, where k_n is the rate constant for forming a primary nucleus size n (30), and κ is the rate of fibril multiplication according to:

[5]

where k₊ and k_- are the rate constants for elongation and fragmentation, respectively. Notably, ν_max is entirely controlled by κ and independent of the primary nucleation events when fragmentation is operative (k_- > 0). Under such conditions, ∂ log ν_max/∂ log[D] converges to 0.5 (). This is in remarkable agreement with the values of ∂ log ν_max/∂ log[D] = 0.56 ± 0.03 measured in this study (Fig. 3). Correspondingly, we observe that ∂ log τ_lag/∂ log[D] = -0.47 ± 0.03, and ∂ log ν_max/∂ log τ_lag = -1.11 ± 0.02 (Fig. 3). Similarly, weak concentration dependencies of ν_max and τ_lag are found by Radford and co-workers in numeric simulations of secondary fragmentation processes (31). Taken together, these observations suggest that the fibrillation process of apoSOD1^ΔIV ΔVII is dominated by mechanical fragmentation, and that the resulting fibrillar fragments elongate by incorporating globally unfolded monomers. The lag phase contains thus no information about the primary nucleation events, as is often assumed in protein-aggregation kinetics, but is the sole consequence of an exponential increase of growing ends. It remains to establish what gives rise to the first fragmentable species (33). Is it ordering by specific or colloidal association of unfolded species, and to what extent does this primary nucleation rely on external support from contaminant seeds or cuvette/gas interfaces?

Fig. 3.

The fibrillation lag time (log τ_lag) and elongation rate constant (log ν_max) show linear dependencies on the logarithmized concentration of globally unfolded apoSOD1^ΔIV ΔVII (log[D]). The matching slopes of ∂ log τ_lag/∂ log[D] = -0.47 ± 0.03 and ∂ log ν_max/∂ log[D] = 0.56 ± 0.03 suggest a fragmentation-assisted mechanism according to Eqs. 3 and 5. Plots of log τ_lag vs. log ν_max, and τ_lag vs. ν_max (Inset). Data from urea titration (closed circles) and mutagenesis (open circles).

Urea Dependence and Fibrillation m-values.

To shed further light on the apoSOD1^ΔIV ΔVII fibrillation process, we used the urea dependencies of ν_max and τ_lag, which reflect the changes in solvent accessible surface area of the underlying microscopic reaction steps. From the denatured baselines in Fig. 2, we obtain ∂ log ν_max/∂[urea] = m_ν max = -0.17 ± 0.05 and ∂ log τ_lag/∂[urea] = m_τlag = 0.13 ± 0.06 (Fig. 2). The matching absolute values of m_νmax and m_τlag lend further support to the existence of a common scaling factor κ, as inferred in Eqs. 4 and 5. Importantly, this scaling factor allows us to examine the microscopic rate constants k₊ and k_- within the same conceptual framework as the folding kinetics (Fig. 4). The results from such examination are described in SI Text, Fibrillation m-values and suggest that 30–50% of the SOD1^ΔIV ΔVII sequence (i.e., 2 to 4 β-strands) buries in the fibril structure, depending on the position of the elongation barrier (‡^elongation) along m_-/+ (Fig. 4). Interestingly, this estimate is in good agreement with previous data from mass spectroscopy, which indicate fibrillar segments of 3 to 5 β-strands (6). To this end, the relatively low values of m_ν max and m_τlag show that the persistent growth of SOD1 fibrils at high [urea] does not need to imply exceptional stability of the product, but could also stem from low sensitivity to denaturant.

Fig. 4.

Schematic illustration of the bifurcated folding and fibrillation free-energy profiles of apoSOD1^ΔIV ΔVII. The reaction coordinate is solvent-accessible surface area, measured by the m-values in Table 1 and SI Text, Fibrillation m-values. The profiles indicate the case where the folding and fibrillation transition states (‡^folding and ‡^elongation) are placed symmetrically. For simplicity, the stabilities of the fibrillar state (A), the unfolded monomer (D) and the folded protein (N) are normalized, and barrier heights are not to scale. I* is the high-energy intermediate observed by NMR at 0 M urea (16, 23).

Mutational Analysis by E-Scanning.

Association of proteins is generally described as a multistep process that starts by random collision, leading first to a loosely bound transient complex that subsequently rearranges into the close-packed, final complex (34). In treatment of association data, the latter step is often assumed to be irreversible, which in this study yields: Scheme 2 where D is the globally unfolded monomer, A_n, a fibril length n, and A′*A_n, the transient complex. The rate constant k_c describes either folding-after-binding (35), binding by fly-casting (36), or conformational steering by hard-body rearrangement (34), depending on the structural nature of the transient complex A′*A_n. The overall rate constant of fibril elongation then becomes:

[6]

which reduces to k₊ ≈ k_D when k_c≫k_-D, and to k₊ ≈ (k_D/k_-D)k_c when k_c ≪ k_-D (34). Mutational studies of association of rigid proteins have shown that the transient complex, as a rule, is dominated by long-range electrostatic contacts with close-range hydrophobic contacts playing a minor role (34). Accordingly, the effect of mutation on the association kinetics is found to be largest for substitution of charged side chains, whereas truncation of hydrophobic side chains show little effect (34). Under the assumption that similar principles apply to association of disordered precursors to fibrils (Scheme 2), we examined in this study the fibrillation kinetics by E-scanning: the SOD1^ΔIV ΔVII sequence was screened by insertion of negatively charged glutamic-acid moieties (Fig. 5 and SI Text, Mutational Analysis of the Fibrillation Kinetics by E-Scanning, Fig. S7 and Table S1). The measurements were done under fully denaturing conditions at 5 M urea to probe specifically the globally unfolded state (Fig. 2). The results show, somewhat unexpectedly, that the response to mutation is not limited to a few fibrillation hot spots but is relatively uniform across the apoSOD1^ΔIV ΔVII sequence (Fig. 5 and Table 2). Moreover, the effect on τ_lag displays an apparent correlation with global charge that extrapolates linearly to the values of apoSOD1^SH SH and apoSOD1^C57–C146 (Fig. 6). The latter correlation resembles data from other protein-protein interactions (34) and indicates that the reaction involves a loosely bound transient complex, as depicted in Scheme 2. Data implicates also that the decrease of τ_lag upon loop removal is fully accounted for by loss of loop charges: The associated effect of loop-entropy reduction seems subordinate. This weak influence of coil entropy is also evident from the similar τ_lag values of globally unfolded apoSOD1^SH SH and apoSOD1^C57–C146 that differ by the cross-link between C57 and C146 (Table 2). Looking at the individual positions, however, there is one notable outlier in the data set: The mutation V5E/V7E in β1 has no discernible effect on the fibrillation kinetics, despite introducing two new negative charges (Fig. 5 and Table 2). It is also clear that several of the other mutations show variations in τ_lag that go beyond their global charge. Additional factors seem to be at play. To examine possible contributions from sequence hydrophobicity, we selected the apoSOD1^ΔIV ΔVII mutations with a common net charge of -5 (Table 2) and plotted these against local change in Kyte-Dolittle index (Fig. 6). Although data show some scatter, the results point at a basic correlation between decreased hydrophobicity and increased τ_lag; the β1 mutation V5E/V7E is, again, deviant because of its lacking impact on τ_lag. A seemingly analogous influence of loosely packed hydrophobic side chains has been observed in the recognition between the unfolded S peptide and ribonuclease S (37). Finally, we note that none of the apoSOD1^ΔIV ΔVII mutations lead to an acceleration of the fibrillation kinetics as expected from a selective increase of the fragmentation rate constant k_- [Eq. 5]. This indicates that the mutations affect mainly k₊ in the κ term, consistent with the two-step association mechanism in Scheme 2. One explanation for this bias would be that k_- is mainly controlled by close-range contacts in the final fibrillar product, and these are not probed by the current E mutations: Insertion of E side chains in this study yields binding ϕ-values of 1, as has been observed previously for charge mutations outside the binding sites of the TEM1-BLIP, barnase-barstar, and Ral-Ral complexes (34). Substitution of surface charges seems to yield similar effect on the ground state and transition state for breaking up the fibrillar core.

Fig. 5.

Examination of the apoSOD1^ΔIV ΔVII fibrillation kinetics by E-scan under fully denaturing conditions at 5 M urea. (A) Histograms of mutant lag times (τ_lag) fitted by non-normalized gamma distributions (Table 2). Pseudo-wild type (black) and mutants (red and green). (B) Sequence positions of E mutations. (C) Structural positions of E mutations (orange). Positively and negatively charged side chains are shown in blue and red, respectively.

Table 2.

Parameters for fibrillation kinetics at 5 M urea

	τlag* (103 s)	δτlag† (103 s)	νmax* (10-3 s-1)	δνmax† (10-3 s-1)	Charge (e)	Hydrophobicity‡
apoSOD1ΔIVΔVII	3.57	1.30	0.98	0.49	−3	9.87
β1, V5E/V7E	3.93	1.50	1.40	0.45	−5	−2.83
β2, Q15E/N19E	7.82	1.74	0.57	0.22	−5	9.87
β3, S34E	8.39	1.66	1.15	0.86	−4	7.17
β4, L42E/H46E	11.42	4.35	0.74	0.34	−5	2.27
β5, N56E/T58E	9.50	3.65	0.55	0.24	−5	7.07
β6, V64E/S68E	9.22	2.99	0.46	0.19	−5	−0.63
β7, T86E/V88E	11.93	4.15	0.44	0.08	−5	−0.73
β8, L101E/V105E	11.22	5.71	0.47	0.42	−5	−5.41
β2, Q15E/I17E/N19E	20.86	6.91	0.19	0.08	−6	1.87
β4, L42E/V47E	13.70	6.10	0.44	0.16	−5	−5.23
apoSOD1C57–C146	40.30 §	8.27 §	0.10§	0.01§	−8	−65.33
apoSOD1SH SH	29.14	5.64	0.10	0.03	−8	−65.33

^*Derived from ThT data according to SI Text, Analysis of Fibrillation Time Courses).

^†Standard deviation from 7 to 20 measurements.

^‡Calculated from the integrated Kyte-Doolittle plot with a running average of 5 residues (SI Text, Mutational Analysis of the Fibrillation Kinetics by E-Scanning).

^§Estimated by interpolation of data at 4 and 6 M urea (SI Text, Analysis of Fibrillation Time Courses).

Fig. 6.

Correlations between fibrillation lag times (τ_lag) and the parameters net charge and hydrophobicity change. (A) τ_lag shows an overall linear dependence on net charge, consistent with the general behaviour of protein-protein interactions, according to Scheme 2 (34). ApoSOD1^ΔIV ΔVII (-3), E-scan mutants (-4 to -6), and apoSOD1^C57–C146/apoSOD1^SH SH (-8). (B) Plot of τ_lag for the apoSOD1^ΔIV ΔVII mutants with normalized net charge of -5 against change in local hydrophobicity (Table 2). The outlier is the β1 mutation V5E/V7E, which is unique by having no effect on τ_lag. The R-value of 0.85 is without the β1 measurement.

Summary and Conclusions.

The diffusely distributed effect of mutations along the apoSOD1^ΔIV ΔVII sequence can be interpreted in two ways. One possibility is that all probed segments, save β1, are part of the fibrillar structure. Although this interpretation may, at first, seem inconsistent with the size estimate from the relatively low fibrillation m-values (Fig. 2), it can be accommodated with data if the fibrillar products are heterogeneous. The monitored reaction would then represent the average of several pathways, each of which involve different subsets of smaller sequence portions, consistent with the previously seen polymorphic nature of the SOD1 fibrils (6). In light of the global dependence on net change and sequence hydrophobicity (Fig. 6), an alternative scenario would be that the unfolded protein is first captured nonspecifically in a loosely bound transient complex and is thereafter guided into its specific, tightly bound state by the fibrillar surface (Scheme 2). A key feature of this two-step process is that the initial and final complexes need not to rely on the same interactions. In the case of apoSOD1^ΔIV ΔVII, the diffusely bound coil would rearrange into a close-packed segment involving less than half of the sequence material (Fig. 4). Interestingly, such two-step binding would explain why protein solubility is generally found to depend on both global (38) and position-specific (39, 40) physiochemical properties: The diffuse association of the transient complex is to some extent global, whereas the close-packed architecture of the filamentous product relies on specific sequence motifs. Even so, we are not yet in position to identify the sequence regions that compose the fibrillar core of apoSOD1^ΔIV ΔVII: The global correlations with net charge and hydrophobicity indicate that our mutational analysis targets the transient complex alone. The question is then to what extent these in vitro studies reflect the pathologic aggregation of SOD1 in ALS? Of particular concern is the effect of mechanical fragmentation, which biases the process by selecting for aggregates that are structurally brittle: Agitation of β₂-microglobulin is seen to promote straight amyloid-like fibrils under conditions where resting samples would otherwise form curly aggregates (28). Further fuel to this concern is provided by the pathological SOD1 inclusions being not clearly fibrillar but resembling more the Lewy bodies observed in prion diseases (41). It has been hypothesized, however, that fragmentation represents a key component of the pathological fibrillation process, since this is expected to happen naturally, due to the ever present mechanical stress in living tissue (42). Such fragmentation of larger aggregates into smaller segments would not only amplify the precipitation but could also facilitate cell-to-cell transmission and hastening the disease progression (42). Sample agitation would then reproduce innate in vivo fluctuations, rather than biasing the experiments in nonphysiological direction. In this context, it is interesting that the congruent effects of ALS-associated SOD1 mutations point at the globally unfolded state as disease precursor. Either by directly shifting the folding equilibrium of the apo protein toward D (18) or indirectly by decreasing the affinity for the stabilizing Cu and Zn ions (12). Equally notable, bioinformatic analysis has shown that ALS-associated SOD1 mutations preferentially decrease the proteins net negative charge (43), in full correspondence with data in Fig. 6. Taken together, these consistencies indicate that that simplified studies of SOD1 fibrillation in vitro could, after all, capture the essence of the pathological aggregation mechanism: an inherent competition between folding and aggregation that bifurcates from the globally unfolded state.

Materials

Mutagenesis, Expression, Purification, and Experimental Conditions.

Mutagenesis, expression, and purification were as in (22, 23), and all experiments were performed at 37 °C in 10 mM Bis-Tris at pH 6.3, unless otherwise stated.

Electron Microscopy.

Carbon-coated 200-mesh copper grids were applied on top of 20 μL droplets of fibril preparation and blotted for 5 min. Staining was with 1% (weight/volume) uranyl acetate. Images of 18,000× magnification were recorded in a Tecnai G² Spirit BioTWIN microscope with a tungsten filament operating at 80 kV.

Folding Kinetics.

Final protein concentration was 4 μM and the urea grade was ultrapure (MP Biomedicals, Inc.). Kinetic measurements were on a PiStar-180 stopped-flow apparatus (Applied Photophysics, Leatherhead, UK) or by manual mixing on a Varian Cary Eclipse spectrometer (Santa Clara, CA), with excitation at 280 nm and emission collected with a 305 nm long-pass filter or with monochromator at 360 nm. Reduction was by addition of 2.0 mM Tris (2-carboxyethyl) phosphine (TCEP) to all buffers and overnight incubation. Chevron data was fitted with

[8]

where k_obs is the observed rate constant, and the parameters as defined in Eqs. 1 and 2. Data analysis was with KaleidaGraph (Synergy Software, Reading, PA).

Fibrillation Kinetics.

Fibrillation was measured at 37 °C in standard 3.5 ml quarts cuvettes in a Varian Cary Eclipse spectrometer (Santa Clara, CA) with Teflon-coated magnetic stirrers at 1,800 rpm. Sample volume was 2 mL and contained 25 μM protein and 20 μM ThT in 10 mM Bis-Tris at pH 6.3 and varying [urea]. Reduction was by 0.5 mM TCEP. ThT excitation was at 444 nm and emission at 485 nm and measured at every 300 s of incubation. Data treatment was by standard procedures as described in SI Text, Analysis of Fibrillation Time Courses.