Conformational Disorder of the Most Immature Cu, Zn-Superoxide Dismutase Leading to Amyotrophic Lateral Sclerosis*

Misfolding of Cu,Zn-superoxide dismutase (SOD1) is a pathological change in the familial form of amyotrophic lateral sclerosis caused by mutations in the SOD1 gene. SOD1 is an enzyme that matures through the binding of copper and zinc ions and the formation of an intramolecular disulfide bond. Pathogenic mutations are proposed to retard the post-translational maturation, decrease the structural stability, and hence trigger the misfolding of SOD1 proteins. Despite this, a misfolded and potentially pathogenic conformation of immature SOD1 remains obscure. Here, we show significant and distinct conformational changes of apoSOD1 that occur only upon reduction of the intramolecular disulfide bond in solution. In particular, loop regions in SOD1 lose their restraint and become significantly disordered upon dissociation of metal ions and reduction of the disulfide bond. Such drastic changes in the solution structure of SOD1 may trigger misfolding and fibrillar aggregation observed as pathological changes in the familial form of amyotrophic lateral sclerosis.

Mutations in Cu,Zn-superoxide dismutase (SOD1) 2 are linked to familial forms of amyotrophic lateral sclerosis (fALS) (1). A major pathological change observed in SOD1-related fALS is the abnormal accumulation of misfolded mutant SOD1 proteins in affected motor neurons (2). Actually, many in vivo as well as in vitro studies have supported that pathogenic muta-tions facilitate the misfolding of SOD1 proteins (3); however, the molecular mechanism triggering the misfolding of SOD1 remains controversial.
SOD1 is known as one of the most stable proteins to the extent that its melting temperature (T m ) is Ͼ90°C (4); therefore, a misfolding event appears quite unlikely for SOD1. Nonetheless, SOD1 was found to have acquired such high stability through several post-translational processes including copper and zinc binding and disulfide bond formation (Fig. 1A). Actually, disulfide-reduced apoSOD1 exhibits significantly decreased stability (T m ϳ 42°C) and is susceptible to unfolding/ misfolding at physiological temperatures (5,6). Intracellular deregulation of metal binding and/or disulfide formation will, hence, be a key event triggering the misfolding of SOD1.
Notably, many pathogenic mutations are found to disturb the post-translational control of SOD1 maturation (7,8) and thereby increase intracellular fractions of the apo- (9) and/or disulfide-reduced state (10). Only when both metal ions and disulfide bond are absent, SOD1 forms fibrillar aggregates (11). Given that SOD1 fibrillation is a pathological hallmark in SOD1-related fALS patients (12) as well as model mice (13), the most immature form of SOD1 will provide a clue to understand the molecular pathomechanism of this devastating disease. In a number of previous studies, the roles of metal binding and disulfide formation in the misfolding of SOD1 have been suggested by a variety of experimental methods (6, 8, 11, 14 -19); however, conformational information on SOD1 in solution lacking both metal ions and the disulfide bond, which is the only state accessible to fibrillar aggregates (11), is still limited. Also, a crystal structure of apoSOD1 lacking the disulfide bond has implied little impacts of metal binding and disulfide formation on its overall folding pattern albeit with increased disorder at loop regions (19). This is probably because the crystallization process sorts the folded conformation out of many other conformations adopted by immature SOD1 in solution. Indeed, significant changes in chemical shifts were concentrated on loop regions (loops IV and VII; Fig. 1A) between the solid state and solution NMR spectra of apoSOD1 with a disulfide bond in crystalline and solution state, respectively (20). Nonetheless, the degree of disorder in the loop regions of immature SOD1 is still ambiguous. To understand the molecular mechanism of SOD1 misfolding in SOD1-related fALS, conformational features of apoSOD1 lacking a disulfide bond, which is prone to fibrillation, therefore need to be clarified in more detail.
Here, we have investigated conformational features of the most immature SOD1 lacking both metal ions and the disulfide bond by using spectroscopic and scattering methods. Although metallated SOD1 with a disulfide bond exists as a homodimer, apoSOD1 without the disulfide bond has been shown to favor a monomeric state and is considered to adopt a protomer conformation (21). In this study we nonetheless found that a conformation of disulfide-null apoSOD1 in solution was significantly distinct from that of the SOD1 protomer; in particular, loop regions usually connected via bound metal ions and a disulfide bond (loops IV and VII; Fig. 1A) lost their restraint and were significantly disordered upon both dissociation of metal ions and reduction of the disulfide bond. Such distinct conformational disorder of SOD1 realized only in its most immature state will be discussed in relation to the pathological changes observed in SOD1-related fALS cases.

Experimental Procedures
Preparation of SOD1 Proteins-Escherichia coli SHuffle TM (New England BioLabs) was transformed with a pET15b plasmid containing cDNA of human SOD1 with an N-terminal His 6 tag, and the expression of His-tagged SOD1 proteins was induced by culturing the transformed cells in the presence of 0.1 mM isopropyl ␤-D-1-thiogalactopyranoside at 20°C for 16 h. His-tagged SOD1 proteins in the lysates were purified with a HisTrap HP (1 ml, GE Healthcare) and dialyzed against 50 mM NaOAc, 100 mM NaCl, 50 mM EDTA, pH 4.0, at 4°C for 16 h so as to remove the metal ions bound to SOD1. The resultant apoSOD1 was further dialyzed against 100 mM Na-P i , 100 mM NaCl/5 mM EDTA, pH 7.0 (NNE buffer), and then incubated with thrombin at 37°C for 3 h. Thrombin was removed from the sample solution using HiTrap benzamidine (1 ml, GE Healthcare), and the purified apoSOD1 was obtained using gel filtration column chromatography (Cosmosil 5Diol-300-II, Nacalai Tesque) with NNE buffer as the running buffer. For preparation of the Zn 2ϩ -bound form of SOD1, the NNE buffer of apoSOD1 samples was first exchanged to a Chelex-treated buffer without EDTA, and then an equimolar amount of ZnSO 4 was added to apoSOD1.
Concentrations of copper and zinc ions in the samples were determined by using Metallo Assay Copper LS (Metallogenics) and Metallo Assay Zinc LS (Metallogenics), respectively. Almost no contamination of copper and zinc ions (Ͻ1 mol % of total SOD1 concentrations) was confirmed in apoSOD1 samples examined here. The thiol-disulfide status of SOD1 was further confirmed by SDS-PAGE; after protection with a thiolspecific modifier, iodoacetamide, SOD1 samples were resolved by non-reducing SDS-PAGE. Electrophoretic mobility of SOD1 has been known to increase in the presence of the conserved intramolecular disulfide bond (22), and we thereby confirmed that our SOD1 S-S samples had the intramolecular disulfide bond (data not shown).
Size-exclusion Chromatography with an Online Multi-angle Light Scattering (SEC-MALS)-60 M (ϳ1 g/liter) SOD1 was loaded on a gel filtration column (TSKgel G2000SW, TOSOH) fitted to an HPLC system (Shimadzu), and the absorbance change at 280 nm of the elution was monitored. For the analysis of apoSOD1, 5 mM EDTA was included in the running buffer (100 mM Na-P i , 100 mM NaCl, pH 7.0) to prevent the adventitious loading of metal ions to SOD1 during gel filtration. A molecular size of the protein eluted from the column was determined by multi-angle light scattering using miniDAWN TREOS (WYATT Technology) connected online to the HPLC system.
Aggregation Assay-The aggregation of SOD1 proteins was monitored by the increase of solution turbidity. 20 M E,E-SOD1 noCys and E,E-SOD1(57/146) S-S in NNE buffer were set in a 96-well plate and agitated with a POM ball (3/32 inch, SAN-PLATEC) at 1200 rpm at 37°C. For the examination of Zn 2ϩbound SOD1, 5 mM EDTA in NNE buffer was replaced with 0.2 mM ZnSO 4 . Solution turbidity was monitored by absorbance change at 350 nm using a plate reader (Epoch, BioTek).
Spectroscopic Methods-For measurement of circular dichroism spectra, SOD1 samples (10 and 400 M for far-UV and near-UV regions, respectively) were prepared in a Chelextreated buffer (20 mM Na-P i , 50 mM NaCl, pH 8.0), and a J-720WI spectropolarimeter (Jasco) was used. Fluorescence spectra of 5 M SOD1 in NNE buffer were measured using fluorescence spectrophotometry (F-4500, Hitachi) with excitation at 282 nm. Fourier transform infrared (FTIR) spectra were measured by using an IRAffinity-1S spectrophotometer (Shimadzu) attached with an attenuated total reflection accessory (DuraSamplIR II, nine reflections). SOD1 proteins (ϳ1 mM) were demetallated by precipitation with trichloroacetic acid and redissolved in a deuterated buffer of 100 mM Tris, 100 mM NaCl at pD 7.0.
Nuclear Magnetic Resonance-15 N-Labeled SOD1 proteins were prepared by culturing E. coli SHuffle TM harboring the plasmid for expression of His-tagged SOD1, and the protein expression was induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside in M9 minimal media containing 15 NH 4 Cl, 3 M ZnSO 4 , and 30 M CuSO 4 at 20°C for 43 h. His tag-free demetallated SOD1 proteins were purified as described above. Using Zeba TM Spin Desalting Columns (ThermoFisher), the NNE buffer in purified SOD1 samples was exchanged to Chelextreated buffer containing 100 mM Na-P i and 100 mM NaCl at pH 7.4 (NN buffer).
NMR experiments were done at 283 K using Bruker AVANCE III 600 MHz spectrometer and Agilent UNITY Inova 600 MHz equipped with a cold probe. The spectra were processed using NMRPipe (23), and the data analysis was performed with Olivia 1.16.9. 3 15 N-Labeled E,E-SOD1(57/ 146) S-S and E,E-SOD1 noCys in NN buffer (ϳ3 mM) were diluted 10ϫ with NN buffer containing 10% D 2 O, and the 1 H, 15 N heteronuclear single quantum correlation (HSQC) spectra were then measured as a reference. For hydrogen-deuterium exchange experiments, the 15 N-labeled proteins (ϳ3 mM) were diluted 10ϫ with a D 2 O buffer containing 100 mM Na-P i and 100 mM NaCl at pD 7.4. The 1 H, 15 N HSQC spectra were collected 5 min after dilution, and the spectral changes were mon-itored every 4 h. The NMR signals were analyzed based on a previously published assignment table (25)(26)(27). Using the resonance intensities before H/D exchange as a reference, relative intensities were calculated.
Small-angle X-ray Scattering (SAXS)-SAXS data were obtained using NANO-Viewer (Rigaku) equipped with a high brightness x-ray generator, RA-Micro7 (Rigaku), and a hybrid pixel array detector, PILATUS 200K (DECTRIS). SAXS measurements were performed at 10°C using a series of samples serially diluted from 6.58 to 2.53 g/liter (SOD1 noCys ) or 8.64 to 3.11 g/liter (SOD1(57/146)) in 50 mM Tris, 100 mM NaCl, 5 mM EDTA at pH 7.0. To remove any oligomeric/aggregated species, the E,E-SOD1 noCys samples were first loaded on the gel filtration column (G2000SW, TOSOH) equilibrated with 50 mM Tris, 100 mM NaCl, 5 mM EDTA at pH 7.0, and protein fractions of a symmetrical and sharp elution peak was collected and immediately used for SAXS measurements. A successive series of scattering images (3 min ϫ 10 frames) was recorded, and then only the images free from radiation damages (10 frames in all experiments) were used to obtain scattering curves. The circular averaged data were normalized by exposure time (3 ϫ 10 min) and protein concentration, thereby the scattering curves, I(Q), were obtained, where Q ϭ 4sin()/, 2 is the scattering angle, and is the wavelength of the x-ray (1.5418 Å). I(Q) at very low angle was fitted with the Guinier approximation using the equation, where I(0) and R g are the forward scattering intensity (Q ϭ 0) and the radius of gyration, respectively. Experimental SAXS data were manipulated using PRIMUS (28) and GNOM (29). The theoretical R g value of the crystal structure of SOD1 (PDB ID 2C9V) was calculated using CRYSOL (30).
Shape Reconstructions-Low-resolution shapes of E,E-SOD1 noCys were restored from SAXS data using GASBOR (31), scored with DAMAVER (32), and visualized with the SITUS package (33). The rigid-body and ensemble models of E,E-SOD1 noCys were refined by using BUNCH (34) and EOM (35), respectively. During both refinements the ␤-barrel scaffold (residues 5-52 and 84 -124) of E,E-SOD1 noCys was treated as a rigid body harboring three flexible loops in its N terminus (residues 1-4), a region including loop IV (residues 53-83), and a region from loop VII to the C terminus (residues 125-157). Coordinates of the rigid body were taken from an x-ray crystal structure of SOD1 (PDB ID 2C9V). The flexible loops were first assumed as dummy residues, and the conformations were then refined against the experimental SAXS data while keeping the rigid body structure intact. In the ensemble modeling, 10,000 hypothetical rigid-body models were randomly generated, from which a plausible ensemble was suggested by refining populations and compositions of the hypothetical structures.

Results and Discussion
SOD1 has four Cys residues in total, among which Cys-57 and Cys-146 form an intramolecular disulfide bond (Fig. 1A). The other two Cys residues (Cys-6 and Cys-111) do not form disulfide bonds in native SOD1 but are susceptible to aberrant oxidation (36). To avoid oxidative modifications during experiments, Cys-6 and -111 of SOD1 in this study were mutated to Ser (SOD1(57/146)); the presence of a Cys-57-Cys-146 disulfide bond in SOD1(57/146) was confirmed electrophoretically (SOD1(57/146) S-S ) (data not shown). Also, SOD1 in which all four Cys residues were mutated to Ser (SOD1 noCys ) was used as the model of SOD1 lacking the disulfide bond.

SOD1 Becomes a Monomer with Decreased Contents of Secondary Structures upon Losing Both Metal Ions and the Disul-
fide Bond-To understand conformational features of the most immature SOD1, a quaternary structure of apoSOD1 noCys (E,E-SOD1 noCys ) was examined by SEC-MALS. Native SOD1 is a homodimeric protein (Fig. 1A), but the monomer-dimer equilibrium has been known to be regulated by the binding of metal ions and the formation of the disulfide bond (21,37). Indeed, we have successfully reproduced our previous findings (21); E,E-SOD1 noCys was monomeric with a molecular mass of 16.0 kDa and that either the introduction of a disulfide bond (E,E-SOD1(57/146) S-S ) or the binding of a Zn 2ϩ ion (E,Zn-SOD1 noCys ) was sufficient to render SOD1 dimeric with a 32.0-kDa molecular mass (Fig. 1, B and C). Previous studies have shown that a protomer of the SOD1 dimer can be prepared with F50E/G51E double mutations (38); indeed, our SEC-MALS analysis also showed that both apo-and Zn 2ϩ -bound forms of SOD1(F50E/G51E) with a disulfide bond were monomeric with a 16.0-kDa molecular mass (Fig. 1, B and C). Quite notably, however, the elution volume of SOD1(F50E/G51E) S-S (8.87 ml) was significantly smaller than that of E,E-SOD1 noCys (9.39 ml) even though both SOD1 species were monomeric (Fig. 1B). It is thus expected that monomeric SOD1 in the most immature state adopts a conformation that is significantly altered from that of a protomer of native SOD1 dimer.
To further confirm the roles of metal binding and disulfide formation in the conformation of SOD1, we have performed secondary structural analysis on the distinct metallation/thioldisulfide status of SOD1 using circular dichroism (CD) spec-  Fig. 2A). In contrast, the binding of a zinc ion and the formation of a disulfide bond were found to play significant roles in attaining the secondary structures ( Fig. 2A); upon dissociation of a zinc ion, the peak around 230 nm became vague, and a negative CD signal around 200 nm increased its intensity. In E,E-SOD1 noCys (red open circles, Fig. 2A) in particular, a negative CD signal at 208 nm that was evident in other SOD1 species examined in this study was lost, and an additional signal was observed around 200 nm, implying the presence of random coils (40).
We also measured the CD spectra in the near-UV region and thereby attempted to characterize the effects of disulfide formation and Zn 2ϩ binding on the tertiary structure of SOD1. A near-UV CD spectrum reflects the environments around aromatic side chains (Trp, Phe, Tyr) and disulfide bonds (41). As shown in Fig. 2B, significant differences in spectral shape were observed between SOD1(57/146) S-S and SOD1 noCys , which probably reflects the contribution of the disulfide bond to CD signals in this region (240 -290 nm) (41). On the other hand, the effects of Zn 2ϩ ion on the spectrum were minimal in SOD1(57/ 146) S-S and SOD1 noCys . Although interpretation of near-UV CD spectra has not been well established (41), the results would support no drastic changes of SOD1 in the regions near Trp and Phe residues (no Tyr residue in SOD1) upon losing a Zn 2ϩ ion.
The presence of random coils in E,E-SOD1 noCys was further confirmed with FTIR spectroscopy. As shown in Fig. 2C (red open circles), the second derivative of an IR spectrum of E,E-SOD1 noCys showed an absorption peak at 1642 cm Ϫ1 , which is characteristic to random coils (42). Upon the addition of an equimolar Zn 2ϩ ion to E,E-SOD1 noCys , an absorption peak at 1630 cm Ϫ1 that corresponds to ␤-sheet structures (42) emerged, whereas random coil structures indicated by an absorption peak at 1643 cm Ϫ1 still remained (Fig. 2C, red-filled  circles). In contrast, both apo-and Zn 2ϩ -bound SOD1 with its disulfide bond exhibited an absorption peak at 1630 cm Ϫ1 but not 1642 cm Ϫ1 , suggesting the formation of ␤-sheet structures (Fig. 2C, black-filled and open circles). Although SOD1 has already been shown to become fibrillated only in the apo-and disulfide-reduced state (11), we have confirmed again in this study that E,E-SOD1 noCys but not E,E-SOD1(57/146) S-S is fibrillogenic and also that the addition of excess Zn 2ϩ ions suppresses the fibrillation of SOD1 proteins (Fig. 2D). Accordingly, a distinct conformational disorder of apoSOD1 that is realized only in the absence of the disulfide bond may have the potential to trigger the fibrillation of SOD1.
The ␤-Barrel-like Scaffold of SOD1 Is Not Significantly Affected in the Most Immature State-Whereas increased fractions of random coils were evident in apoSOD1 upon losing the disulfide bond, it is notable that the absorption peak at 1673 cm Ϫ1 , corresponding to ␤-sheet structures, were still observed in E,E-SOD1 noCys (Fig. 2C, red open circles). The structure of SOD1 consists of a ␤-barrel-like fold with two major loop regions (loop IV and loop VII; also see Fig. 1A). As suggested by previous studies (14,19,(25)(26)(27), upon dissociation of metal ions and/or reduction of the disulfide bond, the two major loops appear to increasingly fluctuate with a mostly retained structure of the ␤-barrel-like core region.
To probe the effects of the disulfide reduction on the protein folding of our apoSOD1 samples, we took advantage of the fact that SOD1 has a single Trp residue in the ␤3 strand of the ␤-barrel-like scaffold and examined its fluorescence properties in E,E-SOD1(57/146) S-S and E,E-SOD1 noCys . In particular, the wavelength of maximum fluorescence emission of Trp has been known to be an indicator of changes in the environment surrounding Trp residues (43). Actually, the fluorescence peak was red-shifted (ϳ6 nm) when E,E-SOD1 noCys and E,E-SOD1(57/ 146) S-S were unfolded in the presence of guanidine hydrochloride (data not shown); nonetheless, the spectra were almost completely overlapped between E,E-SOD1(57/146) S-S and E,E-SOD1 noCys (data not shown). These results hence further sup-port that the ␤-barrel structure in apoSOD1 is not largely affected upon reduction of the disulfide bond.
Actually, previous NMR studies on several immature states of SOD1 have shown that the ␤-barrel structure is well maintained and that the flexible loop regions (loop IV and VII) experience extensive mobility upon removal of metal ions and/or reduction of the disulfide bond (14,(25)(26)(27). To confirm the distinct behaviors of the ␤-barrel structure and the loop regions, we performed a hydrogen-deuterium (H/D) exchange analysis on the 1 H, 15 N HSQC spectra of our E,E-SOD1 noCys and E,E-SOD1(57/146) S-S proteins at 283 K. After dilution of the protein sample with deuterated buffer, intensities of the resonances from amide protons decreased due to the H/D exchange. Fig. 3 represents time-dependent and heterogeneous changes of the resonance intensities in E,E-SOD1 noCys and E,E-SOD1(57/146) S-S over the entire proteins. In this study, 20% of resonance intensities relative to those before H/D exchange were set as a threshold; namely, the resonances of the residues colored red almost disappeared with Ͻ20% of the original intensities before dilution, whereas significant intensities (Ͼ20%) of resonances were maintained in the residues colored  Table 1). C, the pair-distance distribution function, P(r), from the scattering profile is shown: E,E-SOD1 noCys (filled circles) and E,E-SOD1(57/146) S-S (open circles). D max values obtained from P(r) are summarized in Table 1. blue. In both E,E-SOD1 noCys and E,E-SOD1(57/146) S-S , most of the residues with relatively slow H/D exchange (colored blue) are found in ␤-strands, consistent with a robust ␤-barrel scaffold of SOD1 (Fig. 3). Moreover, all of the observed resonances from the residues in loop IV and VII already disappeared in E,E-SOD1 noCys at 4 h after H/D exchange (Fig. 3A), but notably, resonances from some of the residues in loop IV and VII of E,E-SOD1(57/146) S-S (Gly-61, Arg-69, His-71, Asp-76, Glu-78 in loop IV; Thr-135 and Ser-142 in loop VII) retained significant intensities even after 36 h of exchange (Fig. 3B). These data are thus consistent with the flexible nature of those loop regions and also imply that the loop regions in E,E-SOD1 would become more disordered upon losing the disulfide bond. To get more insight into the effects of disulfide reduction on the conformation of demetallated SOD1 species, we further evaluated their molecular size and shape by utilizing SAXS. The Most Immature SOD1 Has a Significantly Distinct Conformation from That of Dimeric SOD1 in Solution- Fig. 4A shows scattering curves of E,E-SOD1(57/146) S-S and E,E-SOD1 noCys at infinite dilution. From the Guinier plot of the scattering curve, forward scattering intensity, I(0), was first determined from the intercept of the linear fit (Fig. 4B) and then used for estimation of a relative molecular mass (Table 1). Based upon the estimates, E,E-SOD1(57/146) S-S and E,E-SOD1 noCys were found to be a dimer and a monomer in solution, respectively. These estimated masses are also consistent with the ones empirically calculated from the Porod volume (V p , Table 1), further strengthening the fact that apoSOD1 became a monomer upon reduction of the disulfide bond (Fig.  1, B and C).
Furthermore, in the Guinier plot (Fig. 4B) the slope of the linear fit contained information on the radius of gyration, R g , of a molecule. As summarized in Table 1, the R g value of E,E-SOD1(57/146) S-S was estimated from the plot as 21.5 Å, which is comparable with the one calculated based upon the crystal structure of matured human SOD1 (PDB ID 2C9V, 20.9 Å) (30). In contrast, E,E-SOD1 noCys showed an R g of 18.4 Å, which was significantly larger than the one calculated from the protomer in the crystal structure of SOD1 dimer (15.5 Å) (30). A less compact structure in E,E-SOD1 noCys was further supported by the pair-distance distribution function, P(r); E,E-SOD1 noCys and E,E-SOD1(57/146) S-S had a maximum dimension, D max , of 58 and 68 Å, respectively (Fig. 4C, Table 1). The observed D max of E,E-SOD1(57/146) S-S was comparable with the one calcu-FIGURE 5. Significant disorder of loops IV and VII describes a distinct conformation of the most immature SOD1. A, an overall shape of E,E-SOD1 noCys (surface model) that was restored from the scattering curve is visualized using the SITUS package, onto which a monomer unit of the SOD1 crystal structure (ribbon model) is superimposed (normalized spatial discrepancy ϭ 0.9203). Cys-57 and -146, which are involved in the formation of an intramolecular disulfide bond, are shown in yellow. Regions allowed for conformational changes during refinements are colored in cyan (loop IV) and pink (from loop VII to the C terminus). B, an experimentally observed scattering curve of E,E-SOD1 noCys (open circles) is compared with the one calculated from a monomer unit of the SOD1 crystal structure (blue, ϭ 7.29). A theoretical curve of the most representative model of E,E-SOD1 noCys with peeled loops IV and VII is shown in red ( ϭ 1.98), which is superimposed in panel C onto the low resolution envelope of E,E-SOD1 noCys (normalized spatial discrepancy ϭ 0.8867). The details and statistics of the modeling are summarized in Table 2.
lated from the crystal structure of matured SOD1 dimer (70.7 Å), but again, E,E-SOD1 noCys in solution had a D max that was larger than of a protomer of SOD1 in its crystal structure (47.7 Å). Collectively, these results clearly indicate that monomeric E,E-SOD1 noCys adopts a relatively extended conformation compared with a monomer unit of the dimer.
Actually, a low resolution model of E,E-SOD1 noCys , which was restored from the scattering curve using GASBOR (31), could not be superimposed well to a monomer unit of the native SOD1 dimer and was more ellipsoidal (Fig. 5A, Table  2). More precisely speaking, our experimental SAXS curve of E,E-SOD1 noCys was significantly deviated from the theoretical one calculated using a protomer of the SOD1 crystal structure ( ϭ 7.29; Fig. 5B). These observations may be reconciled with our results to point to the conformational disorder of loops IV and VII upon losing the disulfide bond in apoSOD1. To confirm this, we refined the conformations of loops IV and VII in the SOD1 crystal structure against the experimental SAXS curve of E,E-SOD1 noCys while keeping the ␤-barrel scaffold intact. As shown in Fig. 5, B and C, a rigid body model of disulfide-null apoSOD1, in which loops IV and VII were "peeled" off from the ␤-barrel scaffold, gave a scattering curve quite similar to the experimentally observed one ( ϭ 1.98).
Nevertheless, loops IV and VII in the rigid body model are highly extended (Fig. 5C) and thus seem to be inappropriate for the description of a single distinct conformer. Rather, taking into account that previous studies showed significant fluctuations at loops IV and VII in immature forms of SOD1 (14,(25)(26)(27), E,E-SOD1 noCys is considered to adopt multiple conformations with highly mobile loops IV and VII. Indeed, compared with the rigid body model, the experimental SAXS curve of E,E-SOD1 noCys was a better fit to the theoretical scattering curve from an ensemble of conformations carrying the flexed loops IV and VII ( ϭ 1.14) (Fig. 6A, Table 2). Most of the simulated conformations exhibited R g of 16 -19 Å and D max of 40 -70 Å, whereas minor conformations also populated at R g 21-27 Å and D max 70 -100 Å (Fig. 6, B and C). It is, therefore, possible that the most immature apoSOD1 without the disulfide bond is monomeric with significant fluctuations in loops IV and VII (Fig. 6D), resulting in a conformation distinct from that of a protomer of the native SOD1 dimer.
Pathological Significance of the Most Immature SOD1-Misfolding/aggregation of SOD1 proteins is a major pathological change in SOD1-related fALS patients (44). It has been suggested that demetallation and/or disulfide reduction are involved in misfolding of SOD1 for the formation of insoluble aggregates (3). Actually, only the most immature state of SOD1 (modeled by E,E-SOD1 noCys in this study) was accessible to the formation of fibrillar aggregates (Fig. 2D) (11). In other words, E,E-SOD1 noCys can be regarded as a precursor for fibrillation, and we have found here that its structural conformation is unique and distinct from that of the dimeric SOD1 in the metallated and/or disulfide-bonded states. In particular, severe disorder at loops IV and VII is considered to be responsible for such a unique conformation of fibrillation-prone E,E-SOD1 noCys (Figs. 5 and 6).
Given that loop IV contains all three zinc ligands (His-71, His-80, Asp-83) and the copper-zinc bridging ligand (His-63) (Fig. 1A), demetallation is expected to cause increased disorder around the loop. Indeed, it has been suggested that in the 1 H, 15 N HSQC spectra of apoSOD1 SH/S-S , the amino acid residues at loops IV and VII exhibit significant changes in their chemical shifts upon binding of a Zn 2ϩ ion (25,26). Many of the amino acid residues in loop IV have not been identified both in the crystal and solution structures of metal-deficient SOD1 SH/ S-S due to severe thermal fluctuations (14,19,45). Moreover, the disulfide bond tethers loop IV (Cys-57) to the ␤-strand after loop VII (Cys-146) (Fig. 1A); therefore, lost of the disulfide bond is also a significant contribution to the increased fluctuation of the loops. Actually, the absence of the disulfide bond facilitated the H/D exchange of amide protons at loops IV and VII of apoSOD1 (Fig. 3). Given that either demetallation or disulfide reduction alone is not sufficient for triggering fibrillation of SOD1 proteins (Fig. 2D) (11), the increased disorder of loops IV and VII, which are realized only in the most immature state of SOD1, could facilitate fibrillation.
As described above, loop IV includes both the disulfide bonding Cys residue (Cys-57) and the Zn 2ϩ binding ligands (Fig. 1A); therefore, fluctuations in loop IV caused by the removal of metal ions would affect the conformation around the disulfide bond. Indeed, lesser amounts of dithiothreitol were required to reduce the disulfide bond in E,E-SOD1 S-S compared with its Zn 2ϩ -bound form, E,Zn-SOD1 S-S (Fig. 7A), suggesting that the disulfide bond becomes increasingly accessible upon losing the metal ions. Also, the chemical modification of Cys-57 and -146 in the disulfide-reduced state (SOD1 SH ) was found to proceed more efficiently in the apo form than in its Zn 2ϩ -bound form (Fig. 7B); certain amounts of the unmodified state remained in E,Zn-SOD1 SH , but the modification was almost completed in E,E-SOD1 SH . This is also supported by the NMR-derived solution structures of apo-and Zn 2ϩ -bound SOD1 with a disulfide bond (14,46) in which Cys-57 and Cys-146 are found to become more exposed to the solvent upon dissociation of a Zn 2ϩ ion (47) (Fig. 7C). Collectively, therefore, these data show that the disulfide bond as well as its constituents, Cys-57 and -146, are more exposed to the solvent upon losing metal ions. SOD1 is localized mostly in the cytoplasm, where the redox environment is highly reducing; therefore, exposure of the disulfide bond to the solvent would increase the chance of being attacked by endogenous reductants such as glutathione. Once the disulfide bond is reduced, apoSOD1 became a monomer with an ellipsoidal conformation in which loops IV and VII are possibly wide-open (Figs. 5 and 6).
Although both concentration and specific activities of SOD1 were decreased in erythrocytes from affected SOD1related fALS patients (48), it is notable that enzymatic activity of SOD1 is almost fully retained in transgenic mice expressing human SOD1 with pathogenic mutations (49). In physiological conditions, therefore, mutant SOD1 could exist initially as a matured state: i.e. a copper, zinc-bound state with a disulfide bond. Given that affinity for Zn 2ϩ ion has been shown to decrease in mutant SOD1 proteins in vitro (7), mutant SOD1 in the matured state would nevertheless gradually lose its bound metal ions in the cytoplasm with high chelating capacity for metal ions. In motor neurons with a meter-long axon in particular, more than a year will be necessary for SOD1 to be anterogradely transported to the nerve termini (50, 51); therefore, it may be difficult for FIGURE 6. An experimental SAXS curve of E,E-SOD1 noCys was reasonably fit with an ensemble of conformations. From 10,000 hypothetical rigid-body models with the randomly flexed loops IV and VII, 10 independent ensembles of conformations were reconstructed to fit the experimental SAXS curve of E,E-SOD1 noCys (Fig. 4A). A, a theoretical curve obtained from a representative ensemble of conformations with the flexed loops IV and VII is shown in red and well matched with the observed scattering curve (open squares, ϭ 1.14). A scattering curve calculated from a monomer unit of the SOD1 crystal structure is again shown for comparison (blue, ϭ 7.29). A distribution of R g (B) and D max of conformations (C) in each of those 10 ensembles is shown. D, representative conformations in a population with (left) 17.9 Å and (right) 24.2 Å of R g are shown. Regions allowed for conformational changes during refinements are colored cyan (loop IV) and pink (from loop VII to the C terminus). mutant SOD1 to survive in a matured form during the axonal transport. Losing metal ions from the matured state of mutant SOD1 could gradually occur in a year-long scale and contribute to the accumulation of the most immature SOD1 proteins (Fig. 7D).
Actually, the amount of SOD1-positive inclusions formed in model mice appear to be negatively correlated with the affinity of mutant SOD1 for metal ions. For example, human SOD1 with mutation abrogating (H46R) or significantly reducing (G85R) metal binding have been found to accumulate significant amounts of inclusions, whereas formation of inclusions was quite limited in the model mice expressing mutant SOD1 retaining the metal binding ability (e.g. G37R, G93A) (24,52). These pathological observations are thus consistent with our proposed mechanism where failure of metal acquisition in SOD1 increased the amounts of the disulfide-reduced state in the cytoplasm (Fig. 7D), and a unique conformation of the most immature SOD1 will be a promising target for controlling the pathogenicity of mutant SOD1 in ALS.