Trifluoroethanol Partially Unfolds G93A SOD1 Leading to Protein Aggregation: A Study by Native Mass Spectrometry and FPOP Protein Footprinting

Abstract

Misfolding of Cu, Zn superoxide dismutase (SOD1) variants may lead to protein aggregation and ultimately amyotrophic lateral sclerosis (ALS). The mechanism and protein conformational changes during this process are complex and remain unclear. To study SOD1 variants aggregation at the molecular level and in solution, we chemically induced aggregation of a mutant variant (G93A SOD1) with trifluoroethanol (TFE) and used both native mass spectrometry (MS) to analyze the intact protein and Fast Photochemical Oxidation of Proteins (FPOP) to characterize the structural changes induced by TFE. We found partially unfolded G93A SOD1 monomers prior to oligomerization and identified regions of the N-terminus, C-terminus, and strands β5, β6 accountable for the partial unfolding. We propose that exposure of hydrophobic interfaces of these unstructured regions serves as a precursor to aggregation. Our results provide a possible mechanism and molecular basis for ALS-linked SOD1 misfolding and aggregation.

Graphic Abstract

Introduction

Amyotrophic lateral sclerosis (ALS), commonly known as Lou Gehrig’s disease in the United States, is a fatal neurodegenerative disease that specifically targets the motor neurons in the spinal cord, brain stem, and cortex.¹ It is the most common motor neuron disease with a lifetime risk of approximately 1 in 2000.² Currently there is no cure or effective treatment.³ Approximately 10% of ALS cases are genetically inherited (fALS); among these, approximately 25% are associated with the Cu, Zn superoxide dismutase (SOD1) gene on chromosome 21 where over 160 site mutations are currently known to cause ALS.⁴

SOD1 is an antioxidant enzyme found in the nucleus, peroxisomes, mitochondrial intermembrane space of eukaryotic cells, and in the periplasmic space of bacteria ⁵. It catalyzes the dismutation of superoxide anion into H₂O₂ and O₂.⁶ The human enzyme is a 32-kDa homodimer, containing one copper and one zinc in the 153-residue subunit. The X-ray crystal structures of SOD1 from many species were solved, mostly in the fully metallized state, showing that the structures are highly conserved (Figure 1A).⁷ Each SOD1 monomer is built upon a β-barrel motif comprising eight anti-parallel β strands, and two large and functionally important loops (i.e., an electrostatic loop and a zinc-binding loop (Figure 1B)).

Figure 1.

(A) Crystal structure of human SOD1 (PDB 1N18), a native dimer. Shown is one monomeric subunit, which contains two major function loops, an electrostatic loop and a zinc-binding loop, and one β-barrel formed by eight β-sheet strands. Zinc ion is denoted as grey sphere, and copper ion as blue sphere. (B) Secondary structure representation of SOD1, showing the eight anti-parallel β strands as large arrows with numbers representing the starting and ending residue index of each strand. The β strands that show increased FPOP footprinting upon TFE treatment are colored green.

In addition to the normal catalytic function of SOD1, there is evidence for a gain-of-function toxicity of SOD1 upon mutation.^{8, 9} Proteinaceous inclusions rich in mutant SOD1 are in tissues of ALS-SOD1 transgenic mice, ALS patients, and in cell-culture models.^{10, 11} The fibrils or insoluble inclusions in ALS, however, are unlikely to be toxic because they arise in a relatively late stage of the disease.¹² The ALS-inducing SOD1 mutants manifest a common shift of the folding equilibrium towards denatured and partially unfolded apo monomers, either disrupting the dimer interface, perturbing the monomer structure, or compromising the formation of mature (e.g., holo, disulfide-bonded) dimers.^13-16 Extensive studies showed that the immature forms of SOD1, obtainable through intramolecular disulfide reduction,¹⁷ metal-ion removal,^{18, 19} heating,^{20, 21} chemical induction^{21, 22} readily form aggregates under physiological conditions, possibly owing to a decrease in the native SOD1 stability and/or an increase in the unfolding rates. Considerable evidence supports that the precursor for SOD1 aggregation is an immature monomeric species. Stathopulos et al.²¹ found that, with trifluoroethanol (TFE) or heat-induced aggregation, mutant SOD1 forms ALS-like aggregates in vitro by populating unfolded or partially unfolded states. Matthews et al.^{18, 23, 24} also showed that the unfolded form of monomeric SOD1 is likely the source of aggregation in motor neurons.

The mechanism and molecular basis of aggregation, however, are still unclear. Eisenberg and colleagues²⁵ found that two heptapeptides derived from SOD1, one from the loop between β6 and β7, the other from β8, are capable of forming cross-beta structures that could serve as precursors to large scale aggregates. Furukawa et al.²⁶ identified a protease-resistant core structure within various fALS-inducing mutant SOD1 aggregates and found three regions (amino acids 1–30, 90–120, and 135–153) that are likely to form a scaffold core for the aggregates. Matthews and co-workers²⁷ recently reported that the C-terminus of SOD1 (loop VII-β8 regions) is involved in a non-native interaction with a disordered N-terminus, a nonnative intramolecular structure that may provide a nucleation site for ALS aggregations. In addition, Kumar and co-workers,²⁸ using molecular dynamics simulations, showed that SOD1 undergoes partial unfolding of β-strands 4, 5, and 6 before TFE-induced aggregation.

Here we present a study of the G93A SOD1 conformational changes and aggregation induced by TFE, which has been commonly used to induce protein aggregation owing to its capability to promote conformational changes in proteins. The effect of TFE on protein destabilization was extensively explored^28-32 to show that moderate amounts of TFE convert the proteins into native-like, aggregation-prone species.^{31, 33} In particular, the fALS SOD1 mutants readily form aggregates in the presence of TFE.^{21, 34} Using far-UV CD, Kumar and co-workers showed that TFE at 15–30% by volume can induce partially unfolded β-sheet-rich extended conformations in A4V SOD1 in a non-monotonic concentration dependent manner;³⁵ the partially unfolded A4V SOD1 subsequently formed aggregates.^{28, 36} The morphology of TFE-induced aggregates greatly resemble those in vivo ALS aggregates.²¹

We selected G93A because its folding and aggregation processes were extensively studied,^{16, 18, 22, 23, 36} and yet the interfaces for aggregation are still not known. Furthermore, the G93A ALS mouse model has both high-expressing and low-expressing versions with different disease severities,³⁷ and antibodies raised against G93A and WT SOD1 exhibit different reactivities for various SOD1 forms,³⁸ emphasizing the need to understand more clearly the structures of oligomers linked to disease. We chose 18% TFE (v/v) for our studies because both the WT and ALS-linked variants of SOD1 unfold and aggregate at this concentration, facilitating future comparisons between mutant and WT. We observed for the G93A SOD1 variant that 18% TFE induces several local unfolding events that lead to aggregation. Meiering and coworkers²¹ previously observed a correlation between the minimum percent of TFE needed to aggregate various SOD1 variants and global stability as measured by thermal unfolding. Additionally, the ALS-linked variants have faster unfolding rates, suggesting a correlation between accessing partially or fully unfolded states and the propensity to aggregate. Importantly, the apo-G93A variant was shown by electron microscopy to assemble from amorphous spheres to elongated fibrils in the presence of TFE. To obtain further insights into the earliest unfolding and oligomerization steps, we chose to focus on structural studies by mass spectrometry rather than kinetics studies of unfolding, which are complicated by the possibility of multiple unfolding and aggregation rates, the inherent challenges of reproducibility when studying aggregation kinetics, and the intrinsic “stochasticity” of SOD1 oligomer formation.^{39, 40}

In our study, we adopted both top-down native mass spectrometry (MS) to explore the formation of oligomers and FPOP footprinting to uncover the conformational changes of G93A SOD1. The use of FPOP to elucidate the surface solvent accessibility of SOD1 proteins was previously demonstrated by Whitelegge and co-workers⁴¹, wherein differential FPOP modifications for metal-free and zinc-bound forms were compared. Herein, we show that these MS-based approaches can detect and localize the partial unfolding, and we propose that the exposure of hydrophobic interfaces of partially unfolded structure can promote the cascade of ALS SOD1 aggregation.

Experimental Section

Materials and Methods

All experiments were performed using recombinant, metal-free apo-SOD1 protein expressed and purified from BL21-Gold(DE3) PLysS E. coli cells (Strategene, Inc., Cedar Creek, TX, USA), as previously described ²⁴. The two free cysteines, Cys6 and Cys111, were mutated to Ala and Ser, respectively, to avoid intermolecular disulfide scrambling; the glycine-to-alanine substitution (G93A) was introduced into this pseudo WT background¹⁸. L-Histidine, L-methionine, 30% hydrogen peroxide, leucine enkephalin acetate hydrate, catalase, trifluoroethanol (TFE), urea, dithiothreitol (DTT), iodoacetamide (IAM), ammonium acetate, formic acid (FA), trifluoroacetic acid (TFA), HPLC-grade solvents, and phosphate buffer saline (PBS) were purchased from Sigma-Aldrich (St. Louis, MO). Sequencing-grade trypsin was purchased from Promega (Madison, WI).

Native Mass Spectrometry of the Intact Protein

A freshly prepared G93A SOD1 protein stock solution (40 ¼M protein, in 10 mM potassium phosphate, pH 7.2), stored at 4 °C prior to use, was buffer exchanged to a 200 mM ammonium acetate, pH 7.0 solution for mass spectrometry (MS) studies by using a Vivaspin® 500 system (Sartorius, Goettingen, Germany). After buffer exchange, the resulting solution was split into two aliquots; 18% (v/v) TFE was added to one aliquot and equal amount of ammonium acetate to the other. The final concentration of G93A was 5 ¼M. Samples were equilibrated for a 1 h incubation at room temperature before conducting the native MS measurements at many time points over a period of 2 weeks. To perform native ESI analysis, 5 ¼L of the protein sample was loaded in a Borosilicate emitter with extra coating (Thermo Fisher Scientific, Waltham, MA); the emitter was interfaced with a Thermo Exactive Plus EMR mass spectrometer (Thermo Fisher Scientific, Waltham, MA). The capillary voltage was 1.0–1.3 kV, and the source temperature was 37 °C. The mass spectrometer was operated in the full-scan mode with a mass resolving power of 70,000 at m/z = 200.

FPOP Footprinting

FPOP footprinting with the incorporation of a reporter peptide was performed as previously described,^{42, 43} and the details are in Supporting Information. A stock solution of G93A SOD1 was treated to give one aliquot treated with 18% (v/v) TFE and another with an equal amount of PBS buffer. Both TFE-treated and untreated samples were submitted to a 1 h incubation at room temperature before FPOP footprinting. To perform FPOP, 50 ¼L of solution containing 5 ¼M G93A SOD1, 8 ¼M leucine enkephalin (reporter peptide), 20 mM hydrogen peroxide, and histidine as radical scavenger, was loaded into a 100 ¼L syringe (Hamilton, Reno, NV), and the solution was advanced by using a syringe pump (Harvard Apparatus, Holliston, MA) into a flow capillary for laser irradiation to give the free radicals. To adjust the labeling times (radical dosage) and to measure the time-dependent (dose) response of TFE-untreated G93A SOD1, various histidine concentrations (0, 0.2, 2.0, 40 mM) were used in FPOP.

Mass Measurement of Intact Protein

The protein-level FPOP footprinting was determined by using a MaXis 4G Q-TOF mass spectrometer (Bruker, Bremen, Germany) coupled to a custom-built, online sample-handling device. In brief, 10 ¼L of footprinted G93A SOD1 (~50 pmol) was loaded onto a C8 reversed-phase trapping column (ZORBAX Eclipse XDB, 2 × 15 mm; Agilent Technologies, Santa Clara, CA) for three-minute desalting with 200 ¼L/min water containing 0.1% (by volume) TFA. The desalted protein was then eluted from the trapping column with a 10-min linear gradient from 2% to 95% solvent B (acetonitrile with 0.1% FA) at 200 ¼L/min.

Bottom-up Analysis of Protein Footprints

All samples were submitted to acetone precipitation for clean-up, and then resuspended in 8 M urea, treated with DTT (10 mM) and IAM (20 mM) sequentially, and diluted with 10 mM PBS buffer (pH 7.5) prior to trypsin digestion. DTT and IAM were used to break disulfide bonds and cap the free cysteines, respectively. Samples were then incubated with 20:1 protein: enzyme (by weight) for 12 h (37 °C, 350 rpm). The digestion was quenched by adding 1% (by volume) formic acid. After digestion, 20 ¼L digests from each sample were loaded in autosampler vial for bottom-up analysis. The LC-MS/MS setup for bottom-up analysis is detailed in Supporting Information.

Data Analysis

The *.raw MS files were searched directly by using Byonic (Protein Metrics Inc., San Carlos, CA) against a custom-built database containing the sequence of G93A SOD1 ⁴⁴. All known HO^• side-chain reaction products were added to the variable modification database for searching for HO^•-modified samples (Supporting Information) ⁴⁵. The alkylation of sample with IAM, which adds a carbamidomethyl group (MW = 57.0214 Da) to Cys-containing peptides, was added as a fixed modification for searching. The searching tolerance window was 15 ppm for precursor ions, and 50 ppm for product ions. Residues Lys and Arg were selected as fully specific cleavage sites with tryptic proteolysis. See Supporting Information for more details of data analysis and quantitation.

Results and Discussion

Native MS Reveals Conformational Changes.

Native ESI (native MS) preserves substantially the native state of proteins when spraying them from aqueous solution containing a volatile salt (e.g., ammonium acetate) at low temperature and neutral pH ^{46, 47}. The process introduces to the gas phase protein ions carrying fewer charges compared to introduction by regular or denaturing ESI because fewer basic residues are exposed in the folded form. Similar to many of other SOD1 mutant variants, G93A SOD1 is natively a dimer ⁴⁸. The native mass spectrum of G93A SOD1 (Figure 2A) shows primarily three peaks, corresponding to the differently charged dimeric G93A species (i.e., +10, +11, and +12 charge states). The mass of G93A dimer was determined by MS to be 31,536.6 Da (theoretical mass is 31,536.5 Da). We could not detect any monomer or higher oligomers of G93A, consistent with the known dimeric nature of G93A SOD1 in its native form.

Figure 2.

Native mass spectra of G93A SOD1: (A) without addition of TFE, G93A SOD1 is predominantly dimer introduced to the mass spectrometer as +10, +11, and +12 dimeric species; (B) under the same spray conditions and with the addition of TFE, the peaks corresponding to dimer were replaced by those corresponding to monomer. The larger charge state distribution of the monomer compared to dimer indicates an increase of solvent-accessible area. The two distributions of charge states (denoted using orange and purple colors, respectively) suggest two conformations of the monomer.

G93A SOD1 changes its conformation, however, upon treatment with TFE. With 1 h incubation, we found an increase of the monomeric species. The expanded charge-state distribution centered at low m/z (higher charge states) suggests a large increase of protein surface area by going from native dimer to monomer ensembles, exposing more residues that are solvent inaccessible in the native state (Figure 2B). The measured mass of monomeric G93A species was 15,770.0 Da by MS (theoretical mass is 15768.3 Da). Furthermore, the two distributions of charge states suggest two structurally different monomeric species, one is more unfolded (carrying more charges; e.g., +15, +14, +13, etc.) and the other somewhat more native (less charged; e.g., +7, +6, +5). The dimer interface of G93A SOD1 is presumably disrupted during the TFE incubation, giving non-native, possibly partially unfolded monomers.

Native MS provides the ability to assess the oligomeric state in the presence of TFE as a function of time, and indeed the native mass spectrum after an extended TFE incubation time (2 days) shows mixed G93A SOD1 species at various oligomeric states, with the 12-mer being the primary species; other species including monomer, dimer, and tetramer are also present (Figure 3A). Interestingly, at 4 days of incubation, the 12-mer and monomer are the only two major species (Figure 3B). The centroid of the charge states for G93A SOD1 12-mer shifts to lower m/z, suggesting a more compact conformation of the 12-mer formed after 4 days of TFE incubation compared to that at 2 days. No native mass spectra could be obtained after 4 days of incubation with TFE, owing to formation of large-scale insoluble aggregates that clogged the ESI emitter. As a comparison, we show that the G93A SOD1 without TFE contains mainly tetramer, dimer, hexamer, and minor levels of 12-mer after 2 weeks of incubation (Figure 3C). No monomers were seen at any time points for G93A SOD1 without TFE.

Figure 3.

Native ESI spectra of G93A SOD1. (A) G93A SOD1 treated with TFE after 2 day of incubation shows mixed species at various oligomeric states, including monomer, dimer, tetramer, and 12-mer; (B) after 4 days of incubation of TFE-treated G93A SOD1, 12-mer and monomer become the predominant species. The centroid of the charge states shifts to lower m/z, likely indicating a more compact conformation of the 12-mer after 4 days incubation, compared to after 2 days incubation; (C) without TFE treatment, the G93A SOD1 is seen as mainly tetramer, dimer, 6-mer, and low level of 12-mer, even after 2-week incubation. This indicates a much slower oligomerization rate for G93A SOD1 without TFE than with TFE treatment.

Given that the results indicate that the monomer is the likely precursor that triggers the downstream oligomerization cascade^{21, 25, 27}, understanding the G93A SOD1 structural changes from native dimer to monomeric species may shed light upon the molecular basis of the aggregation mechanism. In the following study, we focus on the elucidation of the unfolding events and localization of the regions associated with the unfolding of G93A SOD1 induced by TFE. To do this, we performed FPOP footprinting and bottom-up analysis on the native and G93A SOD1 after 1 h TFE incubation.

Footprinting as Function of Radical Lifetime.

To insure that the labeling reactions do not induce conformational changes during the labeling reaction^{49, 50}, we examined the correlation between the extent of FPOP modifications and the lifetime (dosage) of the hydroxyl radicals. The lifetime of the radicals can be controlled by varying the concentration of an amino acid radical scavenger^{45, 51}. We submitted the G93A SOD1 samples to FPOP under different concentrations of histidine (0, 0.2, 2.0, 40 mM) followed by MS analysis of the intact protein and quantified the modification extent of G93A SOD1. The measured mass (15,768.0 Da) is in agreement with the theoretical mass of monomeric G93A SOD1 (15,768.3 Da) that has Cys6 and Cys111 substituted by alanine and serine, respectively. The measured mass verified that our SOD1 protein does not have an acetylated N-terminus, different from the experimental system adopted in the prior FPOP study ⁴¹.

To achieve an accurate time-yield dependence (dosage dependence), we also incorporated Leu-enkephalin in the solution as a reporter peptide whose modification extent is an approximate measure of the labeling time, as described previously⁵². We found that the fraction of unmodified of G93A SOD1 decreased as the labeling time increased at lower histidine concentrations (Figure 4A-D). The linear correlation with the extent of labeling for the reporter peptide (Figure 4E) is consistent with the absence of conformational changes in G93A SOD1 during the time course of labeling. The results also demonstrate the value of adjusting the labeling time by using different concentrations of histidine scavenger to obtain a data trend.

Figure 4.

(A-D) Mass spectra of intact, FPOP-labeled G93A SOD1 (+15 charge state) at different labeling times (radical dosages) as adjusted by using different histidine (scavenger) concentrations. Mass spectra show decreased extent of modification on G93A SOD1 as the labeling time decreases; (E) Linear response curve for the fraction of unlabeled G93A SOD1 vs. the fractional modification of the reporter peptide, as obtained by varying the radical lifetime (dosage) using different histidine concentrations. Coordinates (intersection of the dotted lines) show that the fraction unmodified of G93A SOD1 is 55% when the FPOP labeling time (dosage) corresponds to a 25% yield of oxidized reporter peptide.

Normalization of Labeling Time.

Because TFE can be oxidatively modified by hydroxyl radicals⁵³, the presence of TFE also affects the lifetime/yield of the hydroxyl radicals. This can lead to biased FPOP comparisons between TFE-treated and untreated G93A SOD1 samples⁵⁰. We found that the oxidation yield of the reporter peptide in the presence of 18% (v/v) TFE is 25% with 0.2 mM histidine as radical scavenger. Under the same experimental conditions but in the absence of TFE, however, the oxidation yield of reporter is 38% (Figure 5A). Such differential yield of reporter peptide indicates that the labeling time has been shortened in the presence of TFE. Interestingly, despite the shorter labeling time in the presence of TFE, G93A SOD1 exhibited increased modification at the global level of the protein (Figure 4A), indicating that the protein surface solvent accessibility of G93A SOD1 increases as a result of the TFE incubation. A straightforward, unbiased comparison of G93A SOD1 FPOP yields was confounded at other differential labeling times, as shown in Figure 5.

Figure 5.

Comparison of FPOP yields of G93A SOD1 samples submitted to a 1-h incubation with and without addition of 18% (v/v) TFE. The FPOP labeling experiments were conducted with 0.2 mM histidine as the radical scavenger. (A) Because the presence of TFE shortens the radical lifetime, different yields on reporter peptide were obtained, confounding the comparison of G93A SOD1 FPOP yields. (B) Normalization of the labeling time to that corresponds to 25% reporter modification extent, using the response curve, enables an unbiased comparison of G93A SOD1 FPOP results.

To normalize the FPOP labeling time so that the oxidative modification extents of G93A SOD1 in the two environments (with and without TFE) can be compared, we measured the hydroxyl radical modifications of G93A SOD1 in both environments normalized to the same labeling yield of reporter (i.e., 25% Leu-enkephalin modification). From the response curve (Figure 4E), we determined that, in the absence of TFE, the G93A SOD1 fraction modified is 45% when normalized to the labeling time that corresponds to 25% reporter modification extent. Whereas in the presence of TFE, the G93A SOD1 increases to 71% at the same labeling time (Figure 5B). The significant increase in modification clearly indicates that the conformation of G93A SOD1 becomes more open upon treatment with TFE. This result is in good agreement with the native MS findings showing that the native dimer of G93A SOD1 undergoes unfolding and produces a monomer in the presence of TFE, prior to forming large oligomers (Figure 2 and Figure 3).

FPOP Shows Differences Between Native Dimer and Monomer.

To determine the regions that become surface accessible as the G93A SOD1 native dimer becomes monomer, we submitted the footprinted samples to tryptic digestion and compared the oxidative modification extents normalized to a labeling time where the reporter undergoes a 25% yield in both environments (with and without TFE). By plotting the yield of reporter peptide at each histidine concentration versus the corresponding fraction modified of G93A SOD1 tryptic peptides on the y-axis, we can determine the regional time (dosage)-dependent FPOP modifications of G93A SOD1 (Supporting Information, Figure S1), affording a kinetic output similar to that of HDX curves for peptides^{43, 54}; the relationship between the yield of reporter peptide and time was described previously⁵².

A comparison of yields between the G93A SOD1 dimer and the monomers as a function of location in the protein reveal that several regions undergo local unfolding, including the N-terminus (peptides 1–9, 4–23, 10–23), C-terminus (peptide 144–153), and the β-strands 5 and 6 (peptides 80–91, 92–115). Given the bimodal distribution of peaks representing the monomer, as observed in native MS (Figure 2B), we expect the monomer is conformationally heterogeneous comprised of species exhibiting different levels of unfolding. Thus, these results are a composite due to sampling of different monomer conformations. Importantly, the FPOP results also pinpoint the regions that are not changing (e.g., regions represented by peptides 24–36, 37–69, 116–128, 123–136). These peptides indicate that these regions are not involved in unfolding and also serve as effective negative controls.

FPOP Confirms Disrupted Dimer Interface and Site of Unfolding.

For both the N- and C-terminus of G93A SOD1, the FPOP yields are greater for the monomer than the dimer (Figure 6A). This is consistent with the notion that the native dimeric SOD1 protein uses both its N- and C-termini as dimer interfaces (Figure 6B)⁵⁵. Upon treatment with TFE, those regions become more solvent accessible, a specific observation that is consistent with the results from native MS (Figure 2B), whereby the predominant form of G93A SOD1 under 1-h TFE incubation is a monomer. The results are also in accord with those of Hough and co-workers¹⁴ who showed, with solution X-ray scattering, that there is an increased tendency for the SOD1 dimer to open at the dimer interface upon heat-induced destabilization. A SOD1 toxicity pathway may include a decrease of dimer interface stability and an increase in the population of poorly packed monomeric SOD1 species¹⁶.

Figure 6.

(A) Comparison of normalized FPOP yields of TFE-treated and untreated G93A SOD1 at peptide level. T-test P values were used to indicate statistical significance of the differences (*** for P ≤ 0.001, ** for P ≤ 0.01, * for P ≤ 0.05, and ns for P > 0.05). Regions of G93A SOD1 including the N-terminus (peptide 1-23), C-terminus (peptide 144-153), and peptide 80-115 show significant increase of solvent accessibility going from native dimer to monomer ensembles. The error bars correspond to the± SE of triplicate measurements; (B) Crystal structure of dimeric G93A SOD1 (PDB 3GZP) shows that the N- and C-terminus constitute the dimer interface of G93A SOD1 (N-terminus in light blue, C-terminus in orange); (C) Peptides 80-91 and 92-115 span the entire β-strand 5 and 6, and the adjoining loops (colored in purple).

Besides the dimer interface, a second region located distal from the dimer interface, represented by peptides 80–91 and 92–115 (Figure 6A), also shows statistically significant increases of FPOP yields for the monomer. The X-ray crystal structure of G93A SOD1 indicates this region is comprised of β-strands 5 and 6, part of the closely-packed β-barrel (Figure 1B and Figure 6C). FPOP outcomes corroborate this local compact structure because both peptides 80–91 and 92–115 give lower FPOP yields for the native dimer (Figure 6A), indicating the higher solvent accessibility of the native β-barrel. The footprinting yields for this region, however, increase significantly for the partially unfolded monomeric species. The overall oxidation yields for peptide 80–91, which spans the entire β-strand 5 and part of the adjoining loop (Figure 6C and Figure 7A), increase from 2% to 6%. The yields for peptide 92–115 comprising β-strand 6 (amino acids 95–101) and the full adjoining loop between strands 6 and 7 (amino acids 102–115, Figure 1B) increased significantly from 6% to 32%, indicating a dramatic increase in solvent accessibility and pronounced local unfolding. Examination of the product-ion spectra in the MS/MS analysis reveals hydroxyl-radical modifications of +16 on His80, Val81, and Lys91 (Figure S2). Those differentially modified species are chromatographically separated as shown by their different retention times in the EIC (Figure 7B), allowing quantification at the residue levels (Supporting Information Figure S2) and demonstrating that footprinting can probe local unfolding of proteins⁵⁶. The product-ion spectra also show that Val97 and His110 are two of the modified residues of peptide 92–115, although there are others modified at lower yields (data not shown).

Figure 7.

(A) G93A SOD1 crystal structure with zoom-in view of peptide 80-91 (colored in blue) and residues His80, Val81, and Lys91 (colored in yellow); (B) Extracted ion chromatograms (XICs) of unmodified and +16 modified peptide 80-91. The differently modified isomers (modifications on H80, V81, and K91) are well separated chromatographically because they possess different hydrophobicities than the unmodified peptide.

Structural Results Have Implications for ALS-Associated Aggregation.

The footprinting shows that the TFE-induced G93A SOD1 partial unfolding causes rupture of dimer interface (N-terminal and C-terminal regions, covering β-strand 1, 2, and 8) and local unfolding of β-strand 5 and 6 (Figure 1B and Figure 6B,C). This partial unfolding contributes to the disruption and leads to exposure of structurally promiscuous interfaces that are natively buried, increasing the propensity for self-association linked to the protein aggregation that causes ALS, as hypothesized by others^{14, 16, 18, 19, 26, 57}. Supporting this hypothesis are the results by Nordlund and Oliveberg⁵⁷ who showed, from folding kinetic analysis and φ-value analysis⁵⁸ that, the N-terminal strands β1–3 are the seeds for SOD1 aggregation and that the regions of SOD1 most susceptible to unfolding under physiological conditions are strands β5, β6, and β8. These latter strands serve as protective caps that prevent the hydrophobic edges of N-terminal strands to become exposed. In several ALS-provoking mutant variants, however, this protection mechanism is completely missing owing to early unfolding of these protective caps. Contemporaneously, Ding and Dokholyan¹⁹ showed by all-atom DMD simulations that, for SOD1 monomers, strands β5, β6, and β8 undergo the earliest unfolding, followed by the exposure of sticky edge strands of N-terminus.

Several interesting studies point to the ALS disease relevance of our findings. The TFE-induced aggregation of a series of ALS-linked variants was probed by Sypro Orange fluorescence, demonstrating that hydrophobic exposure was a common property that correlated with aggregation.⁵⁹ Our studies identify precisely where this hydrophobic exposure occurs for G93A. Strikingly, Bertolotti and coworkers³⁴ also demonstrated that these isolated TFE-induced small aggregates could induce intracellular aggregation in a prion-like mechanism. In another compelling study by Bosco and coworkers,⁶⁰ the C4F6 antibody, which is raised against the G93A protein,⁶¹ was shown to bind a misfolded conformation involving β5 and β6. The authors hypothesize that loss of metal binding exposes the normally C4F6 hidden epitope that is recognized in ALS mice and ALS patient samples. In a study by Dokholyan and coworkers,⁶² a trimeric nonnative species that is toxic to neuronal-like cells also exhibits a C4F6 binding site on its surface. Very large aggregates, by contrast, were shown to be neuroprotective.⁶³ Thus, stabilizing the metal-binding loops to limit exposure of the C4F6 epitope can have a potential therapeutic advantage.

In summary, native MS of the intact proteins and MS-based protein footprinting (FPOP) allow an incisive look at the aggregation of TFE-destabilized G93A SOD1. The use of native MS tracks the oligomeric status of G93A SOD1 upon destabilization and shows dissociation of the native dimer to partially unfolded monomeric ensembles prior to formation of oligomers and large-scale aggregates. The FPOP results confirm the partial unfolding of G93A SOD1 and localize the unfolding regions to be the N- and C-termini (β-strand 1, 2, and 8) and strands β5, β6. Proteins with these accessible surfaces may serve as the primary scaffold for the aggregation cascade.

The addition of TFE induces local unfolding of G93A SOD1 to expose hydrophobic regions that elicit SOD1 aggregation. Inhibition of such exposure may lead to prevention of aggregation of SOD1 and slow the fALS progression. This bioanalytical approach may be effective for other aggregating proteins.