Structural and dynamic aspects related to oligomerization of apo SOD1 and its mutants

Results and Discussion

In the present study we characterized the apo form of WT SOD1 and 2 ALS-related mutants with 2 extreme behaviors in terms of oligomerization rates: T54R oligomerizes with rates slightly slower than WT SOD1, whereas I113T has an oligomerization rate more than twice that of WT SOD1 (14).

In the crystal structures of the metal-free form of WT SOD1 and its mutants T54R and I113T, the asymmetric unit contains 2 biologically relevant dimers (A–B and C–D). Although one has a very well-defined electron density throughout the entire sequence, the other dimer has clear breaks in the electron density in the regions encompassing residues 68–78 (loop IV) and residues 125–140 (loop VII). Overall, the structures are very similar to each other. Only few minor local differences in terms of conformation, buried surface areas, and residues involved in H-bond interactions are detected (Table S1). Peculiar for apo T54R SOD1 is a mutation-related hydrogen bond, which is formed at the dimer interface, i.e., between NH2 of Arg-54 of each monomer and OD1 of Asn-19 of the other monomer in the same dimer. This interaction might lead to a partial stabilization of the dimeric state with respect to the WT protein and might correlate with the slightly slower oligomerization rates for apo T54R compared with the WT protein (14), because oligomerization presumably occurs through monomerization (19). The free Cys-6, in all 3 apo SOD1 structures, is essentially buried because of the constraints imposed by 2 H bonds with Ile-18, whereas the other free cysteine, Cys-111, has a considerably high (75%) solvent-exposed side chain (thiol group).

The 3 structures determined here clearly resemble the already available partially apo (20% zinc in 1 of the 2 dimers) WT SOD1 structure [Protein Data Bank (PDB) ID code 1HL4] (18). Similarly, the same electron density breaks were found in only 1 of the 2 dimers. On the contrary, the fully metalated (Cu₂,Zn₂; holo) structures of both WT SOD1 [PDB ID code 1HL5 (18)] and I113T SOD1 mutant [PDB ID code 1UXL (20)] show a well-defined electron density throughout the entire sequence for each dimer present in the asymmetric unit. The backbone rmsd between the holo and apo form of both WT and I113T SOD1 is ≈0.45 Å. The main structural differences between the holo and apo states are observed in the loops connecting the β-strands, where the electron density is broken in one of the dimers in the structures of the apo state. Similar electron density breaks were also observed for the structure of the apo state of the H46R SOD1 mutant (21), the only, up to now published structure of a completely metal-free SOD1 mutant. On the contrary, the structures of ALS-related SOD1 mutants, in the fully metalated state are very similar to each other, and particularly well ordered throughout the sequence (20–25).

WT SOD1 and its mutants, T54R and I113T, in the apo state, form a continuous, extended arrangement of β-barrels stacked up along a direction (crystallographic b-axis) perpendicular to the dimer interface (Fig. 1A). The tetramer A+B+C+D comprising the asymmetric unit is surrounded by 4 others in the ac plane. The intermolecular contacts between the 2 dimers in the asymmetric unit are between monomers A and D and monomers B and C. The H-bonding connections are different for the 2 pairs (Fig. 1B). A H-bonding network is also present between molecules in adjacent asymmetric units, with 7 strong and 3 weak H bonds between monomers A and D1 (i.e., monomer of an adjacent asymmetric unit) and with 6 strong and 2 weak H bonds between monomers B and C1 (Fig. 1B). The view orthogonal to the crystallographic b-axis shows that the apo WT SOD1 β-barrel forms a zigzag array of filaments (Fig. 1C). This behavior is similar to that already observed for apo WT SOD1 reported by Strange et al. (18) and is common to amyloid-like fibrils (26).

The comparison between the structures of WT SOD1 and ALS-related mutants was unable to shed light on the structural basis of different behaviors in oligomerization rates between WT and pathogenic mutants, although it did point at some conformational disorder as a consequence of the lack of metal ions. Therefore, a characterization of the apo state in solution of WT SOD1 and its mutants is necessary.

The ¹H-¹⁵N HSQC spectra of the apo forms of both WT SOD1 and the 2 pathological mutants i.e., T54R and I113T, have reduced signal dispersion with respect to those of the metalated form (Fig. S1), indicating that some parts of the protein do not have a well-defined conformation. High protein instability and strong tendency to form high molecular weight oligomers (13) prevented us from collecting the triple resonance NMR experiments necessary to achieve a specific resonance assignment at 298 K. A protein sample analysis, aiming at finding the best compromise between protein stability toward oligomerization and line broadening effects of NMR signals, led us to acquire all of the spectra at 0.6 mM protein concentration and 288 K. In these experimental conditions, dimeric apo WT SOD1 is stable for periods long enough to collect some of the triple resonance NMR experiments necessary for sequence-specific resonance assignment even if more than one sample from the same preparation was necessary to acquire all of the experiments.

Through the NMR experiments, 68% of the backbone atoms (N, HN, and Cα) were assigned for apo WT SOD1. The unassigned peaks were all clustered in the central part of the ¹H-¹⁵N HSQC spectrum, thus experiencing severe resonance overlap. They were located mainly in loops connecting the β-strands, in particular in loop IV, which contains most of the metal-binding residues. This spectral pattern was already observed in the apo state of the monomeric form of SOD1 obtained through residue mutations at the subunit–subunit interface (17). In the latter case, mutations of the 2 free cysteines (6 and 111) (as WT SOD1) prevented oligomerization, allowing to reach a much higher protein concentration that, combined with the half molecular weight, led to a more complete assignment, which was used here for comparison purposes.

From the analysis of the assigned chemical shift resonances it appears that the secondary structural elements of the β-sheet of the SOD1 β-barrel that comprises β-strands 1, 2, 3, and 6 exist also in the metal-free state, whereas β-strands 4, 5, 7 and 8 are much shorter. NH–NH long-range NOEs are indeed present within β-strands 1, 2, 3, and 6, whereas they are mostly missing in the other β-sheet, which contains some of the metal ligand residues (Fig. 2). From the analysis of the NOESY spectrum, it also appears that most of the long-range ¹H–¹H NOEs, involving side-chain protons, are missing even within the secondary structural elements.

Combined chemical-shift variations of backbone amide groups between the dimeric apo SOD1 and its fully metallated form (Fig. 3A) indicate significant structural changes in loop VII, similar to that already observed for monomeric apo SOD1 (17). Furthermore, a number of NH groups in this loop show, mainly at low temperature, another set of signals with lower intensity caused by a minor conformation or a group of fast exchanging conformers in slow exchange with the rest of the conformations. Analysis of the chemical-shift variations between the monomeric and dimeric apo states of SOD1 (Fig. 3B) confirms that the absence of metal ions similarly affects the 2 forms, particularly in the electrostatic loop. The only detected differences can be ascribed to the mutation of the 2 free cysteines residues (C6A and C111S) and 3 residues (F50E, G51E, E133Q), which, by inducing protein monomerization, sizably affects the NMR signals of the residues at the dimer interface. The latter 5 mutations are present only in the “artificial” monomeric form.

Also the dynamical properties of apo WT dimeric SOD1 are affected by the absence of the metals (Fig. 4). The ¹⁵N relaxation rates of backbone NHs, measured in the temperature range 288 to 310 K, are consistent with the protein being essentially only in the dimeric state, with its tumbling rate increasing with decreasing temperature. Residues located in the first β-sheet of the β-barrel have R₁ and R₂ relaxation rates and ¹⁵N{¹H}–NOEs values characteristic of a structured protein (0.75 ± 0.04 s⁻¹, 27.6 ± 0.7 s⁻¹, and 0.76 ± 0.03, respectively at 298 K). Similar average values were found for the assigned residues of strands β7 and β8 (Fig. 4), which indeed remain significantly structured, as also confirmed by the presence of a NH–NH long-range NOE between β8 and β1 (Fig. 2).

The relaxation rates and ¹⁵N{¹H}–NOEs, which are homogeneous in the β-strand structures, are instead dramatically altered in the electrostatic loop VII. The latter has, at 298 K, lower than average R₂ values_, higher than average R₁ values, and lower ¹⁵N{¹H}–NOEs, which become even negative at 310 K, indicating that its residues experience motions faster than the overall protein tumbling, i.e., faster than a nanosecond. Internal motions in the subnanosecond time scale were also observed in loop VII of monomeric apo SOD1 (17), confirming that the absence of metal ions similarly affects the dynamical properties of both monomeric and dimeric apo forms. The overall spectral features in solution for loop VII suggest that this region, and the nonassigned residues located in the other loop regions, sample a wide range of conformations that interconvert each other very fast.

¹⁵N relaxation rates of backbone NHs were also measured for apo T54R SOD1 (Fig. S3). The results are overall very similar to the data obtained for apo WT SOD1 (Tables S2–S4), showing that, also for this mutant, the mobility of loop VII is dramatically affected by the lack of metal ions; as also confirmed by the negative values of the ¹⁵N{¹H}–NOEs. The similar mobility behavior of apo WT and T54R mutant is consistent with their similar rates of oligomerization (14).

Unfortunately, the I113T mutant is too unstable, preventing the acquisition of relaxation data at temperature >288 K. At this temperature this mutant has overall the same dynamical properties as WT and T54R SOD1, being essentially structured in β-sheet 1 and β-strand 8, but showing higher backbone dynamics in loop VII where a sizable number of residues have negative ¹⁵N{¹H}–NOEs values even at 288K (Fig. S4 and Table S2). At variance with WT and T54R SOD1 mutant, in I113T some residues located in other loop regions experience extensive dynamics on the subnanosecond time scale. In particular, some residues of loop V (residues 90–94), which connects β5 and β6, show a peculiar drop of the ¹⁵N{¹H}–NOEs and R₂ values and a corresponding increase of R₁ values consistent with fast local motions (Table S2). Interestingly, β5 and β6 are located at the edge of the 2 sheets constituting the SOD1 barrel and are thought to be involved in amyloid fibril formation through intermolecular aggregation with other edge strands in other fALS-related mutants (18).

The observed local flexibility of the SOD1 apo forms with respect to the metalated ones is consistent with their solvent accessibility. Indeed, the overall number of NH protons exchanging fast with the bulk solvent, as measured from H₂O/D₂O exchange processes, is much higher in the apo state than in the metalated one. In the apo forms of both WT and T54R SOD1, after 20 min the samples have been dissolved in D₂O, ≈60 residues were completely exchanged, whereas only ≈30 were exchanged in the metalated one. Nevertheless, hydrogen exchange protection factors of apo WT and T54R SOD1 (Fig. 4D and Fig. S3D) reveal that residues located in the secondary structure elements of the β-sheet constituted by strands β1, β2, β3, and β6 and some residues of β8 still show a significant degree of order, in agreement with the presence of a β-structure hydrogen-bond pattern. In particular, in β6 it was possible to quantify exchange rate values only for the amide protons involved in hydrogen bonds with residues located in β3, whereas the other amide protons exchanged too fast with the solvent. Overall WT and T54R SOD1 have very similar H/D exchange properties.

Particularly relevant for the oligomerization process that occurs for apo WT SOD1 at physiological conditions (13), is the solvent exposure of the free cysteines (Cys-6 and Cys-111). Although Cys-111 is highly solvent exposed in both protein forms, Cys-6 has a dramatically different solvent accessibility (Fig. S5). In the metalated form its NH is essentially buried and protected from the solvent and indeed its NH signal is still present after 5 days in D₂O. On the contrary, in the apo form it almost completely disappears after only 4 h. Also, the NH signals of adjacent residues (residues 4–8, β-strand l) all disappear in 4 h, suggesting a high solvent accessibility of the region around Cys-6. The high solvent accessibility of Cys-6 and β-strand l is observed for apo T54R mutant as well, suggesting, as for apo WT SOD1, that the metalated form is rigid and solvent protected, whereas the lack of metal ions makes this region and the entire protein highly dynamic and more solvent accessible.

All of these features indicate that the apo state of SOD1 in solution is characterized by a distribution of conformations, particularly for β-strands 4 and 5 and loop VII.

Similar behavior was observed in the SOD-like protein from Bacillus subtilis (27), where its NMR properties indicate a conformational mobility for most of the protein, characterized by defined secondary-structure elements and a dynamic tertiary structure, at variance with the X-ray crystal structure of the same protein, which shows a well-ordered tertiary structure.

The overall studies here presented for the apo state of SOD1 and its mutants in solution also explain the behavior of these proteins with respect to crystallization. Crystallization trials were performed at 2 different temperatures on both the apo and metalated forms of WT SOD1 and the 2 mutants (Table S5), i.e., at 288 K, which is the temperature at which the crystals discussed above were obtained, and at 310 K, which is the temperature at which the oligomerization studies were carried out (13, 14). Crystals of the apo state of WT SOD1 and the 2 mutants can be obtained only at 288 K, whereas crystals for the metalated forms of WT SOD1 and the mutants were obtained at both temperatures. This result indicates that temperature has a major influence on the crystallization of the apo state whereas it is almost negligible on the metalated one. This overall behavior is consistent with the dynamic properties and conformational disorder of the apo state. A decrease in temperature slows down the interconversion process among the various conformations and increases the population of the most stable states, which can therefore crystallize.

The local disordered state observed for the metal-free SOD1 may appear to be in contradiction with its cooperative unfolding behavior reported by differential scanning calorimetry measurements (28). This contradiction can be, however, reconciled by the presence of a significant portion of still existing tertiary structure and extensive structural interactions at the dimer interface caused by the preserved dimeric nature.

Concluding Remarks

We have recently shown (13) that oligomerization of apo SOD1 involves oxidation of the 2 free cysteines (Cys6 and Cys111) with the formation of intersubunit disulfide bonds, thus linking a high number of protein molecules in high molecular weight species. Some have suggested that the formation of SOD1 aggregates are the consequence of both covalent disulfide cross-linking and noncovalent interactions (29), whereas others proposed that extensive disulfide cross-linking is not required for the formation of mutant SOD1 aggregates (30). Recent studies showed the importance of nonphysiological intermolecular disulfide bond between cysteines 6 and 111 in mutant SOD1 for high molecular weight aggregate formation for protein ubiquitylation and neurotoxicity, which are all dramatically reduced when these cysteines are substituted (31). In any case, there is a general agreement on the critical role played by cysteines 6 and 111, in the modulation of human SOD1 aggregation (29, 31).

We have shown that only the lack of metal ions makes SOD1 oligomerization possible (13). The reason for the dramatic different behavior of apo and metalated forms of SOD1 is now better understood. Indeed, the solvent exposure of the reduced cysteines changes dramatically from the metallated form to the apo one. Only in the latter state a free cysteine can bind another one of a different monomer to form the soluble oligomer. The crystal structures, on the contrary, are not informative on this respect as they clearly represent only one of the multiple conformations taken in solution by the protein. Consistently, the apo form of both WT and the mutants fail to crystallize at physiological temperature because of the high disorder and internal mobility.

The information obtained from the NMR spectra indicates that in solution apo WT SOD1 samples a range of conformations, which are highly disordered in some parts. Higher temperatures accelerate exchange among these conformations and could populate new ones. This behavior explains why only the disordered, locally unfolded, metal-free state has a dramatic protein flexibility that makes accessible conformations prone to oligomerize, whereas the rigid structure of the metalated protein is unable to do it.

Overall the present extensive structural and dynamical characterization of the apo state of WT SOD1 and some of its mutants showed that the lack of metal ions and the subsequent protein flexibility allows the free cysteines (in particular Cys-6) to become exposed and therefore ready to get oxidized and form the disulfide bonds that give rise to the soluble oligomers.

Materials and Methods

Supplementary Material