Pulsed hydrogen–deuterium exchange mass spectrometry probes conformational changes in amyloid beta (Aβ) peptide aggregation

Abstract

Probing the conformational changes of amyloid beta (Aβ) peptide aggregation is challenging owing to the vast heterogeneity of the resulting soluble aggregates. To investigate the formation of these aggregates in solution, we designed an MS-based biophysical approach and applied it to the formation of soluble aggregates of the Aβ₄₂ peptide, the proposed causative agent in Alzheimer’s disease. The approach incorporates pulsed hydrogen–deuterium exchange coupled with MS analysis. The combined approach provides evidence for a self-catalyzed aggregation with a lag phase, as observed previously by fluorescence methods. Unlike those approaches, pulsed hydrogen–deuterium exchange does not require modified Aβ₄₂ (e.g., labeling with a fluorophore). Furthermore, the approach reveals that the center region of Aβ₄₂ is first to aggregate, followed by the C and N termini. We also found that the lag phase in the aggregation of soluble species is affected by temperature and Cu²⁺ ions. This MS approach has sufficient structural resolution to allow interrogation of Aβ aggregation in physiologically relevant environments. This platform should be generally useful for investigating the aggregation of other amyloid-forming proteins and neurotoxic soluble peptide aggregates.

Protein aggregation is one of the immediate causes of Alzheimer’s, Parkinson, and Huntington diseases, motivating biophysical studies of the responsible proteins. More than 20 small proteins undergo amyloidosis in humans. In Alzheimer’s disease (AD), the aggregation of the 40- or 42-aa-long amyloid beta (Aβ) peptide, generally called Aβ₄₀ or Aβ₄₂, respectively, is proposed to be involved in the onset of the disease (1, 2). Aβ₄₂ is more amyloidogenic and more neurotoxic than Aβ₄₀. Although the amyloid-cascade hypothesis suggests that the Aβ-containing amyloid plaques are responsible for neurodegeneration (3–7), other studies suggest that soluble aggregates of Aβ₄₂ are more neurotoxic than the amyloid plaques (8–13).

The amyloid plaques in AD-affected brains contain high levels of copper, zinc, and iron (14–20). Among these, Cu has drawn the most attention because the Aβ precursor protein is likely a Cu-chaperone protein (21). Several studies of Cu²⁺–Aβ₄₀ interactions show that Cu²⁺ can promote Aβ₄₀ aggregation (14, 18, 19).

The structure of Aβ₄₂ and its aggregates, although studied extensively, remains of high interest. Studies of amyloid fibrils invoke X-ray crystallography (22–24), EM (19, 25, 26), and thioflavin T fluorescence (19, 27), revealing the polypeptide’s global behavior, whereas NMR studies provide residue-level information for the fibrils (28–30). Nevertheless, we know little about soluble Aβ aggregates owing to their intrinsically high heterogeneity.

MS should offer an opportunity for investigating soluble aggregates of Aβ₄₂. Thus far, there are no MS-based, time-dependent studies of the formation of soluble aggregates. Moreover, there are no other biophysical studies of Aβ₄₂ aggregation at the peptide (regional) level. MS, however, was used for analyzing the aggregated Aβ fibrils (31–33) and, with ion mobility (34–36), for soluble Aβ aggregates. Hydrogen–deuterium exchange (HDX) (37–42), even with top-down sequencing, can afford residue-level information (43, 44) and provide insight on Aβ₄₂ fibril core structure (31, 32) and its recycling (33, 45).

In light of the dearth of aggregation studies at the peptide level, we have used herein pulsed HDX to study the aggregation of the Aβ₄₀ and Aβ₄₂ peptides. Our platform is suitable for confirming the effect of temperature, agitation, and presence of Cu²⁺ ions on Aβ aggregation. Pulsed HDX, developed by Englander and coworkers (46), can be used for detecting protein folding intermediates (47) and membrane protein behavior (48). Others have also applied pulsed HDX to analyze fibril (33) and oligomer (43, 44) Aβ structures. To explain the experimental data, we used a self-catalyzed aggregation model, and its success suggests utility for other amyloid-forming proteins.

Materials and Methods

Protein Preparation.

A synthetic, wild-type human form of Aβ₄₀ (Keck Biotechnology Resource Laboratory) and a recombinant, wild-type human form of Aβ₄₂ (rPeptide) were used. Sample preparation followed a reported procedure (49) as detailed in Supporting Information. Aβ films were thawed and dissolved in dry DMSO at 1 mM before aggregation. Aβ aggregation studies were initiated by diluting 1:19 (vol/vol) monomeric Aβ (either Aβ₄₀ or Aβ₄₂) into PBS buffer (pH 7.4) at 25 °C. Aggregation was allowed to occur over various incubation times ranging from 1 min to 48 h and under various experimental conditions as described in Supporting Information. Native gel electrophoresis and Western blotting were conducted as previously described (19).

Pulsed HDX.

HDX experiments were carried out by mixing the incubated protein sample with D₂O buffer in a 1:1 (vol/vol) ratio (pulsed labeling for 1 min at 0 °C). HDX was then quenched by mixing the exchanging solution with 30 μL 3 M urea and 1% TFA to give a pH of 2.5. Pepsin digestion of Aβ and data analysis and processing (50) were used as described in Supporting Information. Importantly, the HDX time was fixed for each experiment, so the term “time” hereafter means aggregation time unless stated otherwise.

Finke–Watzky Modeling.

The experimentally determined time-dependent data (mass shift vs. incubation time) were characterized phenomenologically by following the recommendation of Finke and coworkers (51) to use the simplest model consistent with the data. A bootstrap resampling method (52–54) was then used to evaluate the uncertainty of t_1/2 (Supporting Information gives details).

Results and Discussion

Our goal is to implement a general MS approach to monitor Aβ aggregation under different incubation conditions. Both HDX and aggregation are time-dependent phenomena. To minimize aggregation during HDX and thereby separate aggregation and HDX, we used pulsed HDX whereby we allowed Aβ to incubate for various times and then submitted the mixture to a rapid pulsed HDX (1 min) followed by the usual quench procedure. Before MS analysis, we centrifuged the sample, split it into two equal parts representing the top and bottom parts, respectively, and analyzed the latter aliquot that is enriched in Aβ oligomers and aggregates.

Comparison of Aβ₄₂ and Aβ₄₀ Aggregation by Native Gel and Western Blotting.

After various aggregation times, we analyzed the Aβ₄₀ and Aβ₄₂ samples by native gel and Western blotting to permit an accurate comparison of the aggregated species (Fig. S1). We found that the top half of the solution after centrifugation showed no detectable aggregates but only Aβ monomer, indicating that most of the monomer is homogeneously distributed, whereas larger soluble species move to the bottom half. We used the Aβ₄₀ peptide as a control, which showed only the presence of Aβ₄₀ monomer in all of the “bottom-half” samples (Fig. S1, a–c), in agreement with previous work (55). Aβ₄₂, however, showed a quite different behavior. After only 1 min of incubation, formation of low-molecular-weight (LMW) Aβ₄₂ oligomers (i.e., mostly trimer and tetramer) was observed. With longer incubation time, high-molecular-weight (HMW) soluble Aβ₄₂ oligomers in the 50- to 110-kDa range were observed. The amyloid fibrils, which are too large to enter the gel, were observed at the top of gel lanes. All these results are consistent with previously published work showing that Aβ₄₂ has a greater tendency to form aggregates than Aβ₄₀ (34, 55).

Pulsed HDX Analysis of Soluble Aβ Species.

To conduct a more detailed characterization of various regions of Aβ, we applied pepsin proteolysis following the pulsed HDX treatment described above. Because most of the soluble Aβ species were on the bottom-half fraction upon centrifugation, we performed all analyses on the bottom-half solution. Seven peptides were observed: the N-terminus region (1–19 and 4–19), middle region (20–33, 20–34, and 20–35), and C-terminus region (35–40/42 and 36–40/42). We used three of these peptides (i.e., 1–19, 20–35, and 36–40/42) to analyze the HDX results because these fragments provide full Aβ coverage.

The pulsed HDX approach used to probe the Aβ aggregation states as a function of incubation time is unlike most HDX protocols, in which the protein is incubated in D₂O buffer for various times, quenched, and analyzed. Here the HDX time is a short (1 min) “pulse” after various incubation times. This pulse time is sufficient to allow the maximum exchange for this largely intrinsically disordered protein yet is negligible compared with the incubation times. We view this pulsed HDX as a “recording” tool to monitor aggregation without competing with it, allowing us to separate experimentally the aggregation from HDX.

Initially, we followed the aggregation of Aβ₄₀ at 25 °C by pulsed HDX as a control experiment (Fig. 1). All three peptic peptides monitored show a constant HDX extent as a function of incubation time. This indicates that no detectable conformational changes or self-association occurred from 1 min to 48 h at 25 °C. The apparent protection is likely due to two factors: (i) The exchange time is sufficiently short for the pulse (1 min) that HDX is not complete and (ii) there is significant back-exchange for the highly disordered Aβ₄₀ (Supporting Information). This extent of HDX is consistent not only with our native gel and Western blotting experiments, but also with the common perception that Aβ₄₀ forms fewer aggregates than Aβ₄₂ (55, 56).

Fig. 1.

Pulsed HDX results for three peptides from Aβ₄₀. Aggregation was done at 25 °C and in the absence of Cu²⁺. Peptic peptides 1–19 (A), 20–35 (B), and 36–40 (C) are represented by triangles, circles, and diamonds, respectively. Lines were fit by using “linear fit” in OriginPro 8.5.

By contrast, HDX of Aβ₄₂ showed an appreciable increase in protection by a modified sigmoidal behavior (Fig. 2, solid line; raw HDX data shown in Figs. S2–S5). The first stage is rapid and exponential-like, showing a rapid increase in protection. A first plateau follows wherein no significant change occurs in protection. A second increase in protection appears as a sigmoidal upward break in the curve, followed by a second plateau, indicating that the system reaches equilibrium among all of the species. The HMW species correspond to Aβ₄₂ soluble aggregates that concentrate upon centrifugation to the bottom-half solution. Importantly, these results are consistent with native gel and Western blotting that show formation of HMW Aβ₄₂ soluble species on the same time scale as the pulsed HDX experiments (Fig. 2, Right, lane 1).

Fig. 2.

Pulsed HDX (Left) and Western blotting (Right) results for three peptic peptides from Aβ₄₂. Aggregation was done at 25 °C and in the absence (solid triangles, circles, and diamonds) or presence of Cu²⁺ (hollow triangles, circles, and diamonds). Peptic peptides 1–19 (A), 20–35 (B), and 36–42 (C) are represented by triangles, circles, and diamonds, respectively. Dashed lines are fits using “linear fit” in OriginPro 8.5. Solid lines are fits using a modified F–W model in Mathcad. Native gel and Western blotting of the aggregated Aβ₄₂ samples (incubated at 25 °C for 48 h), in the absence (lane 1) or presence of Cu²⁺ (lane 2).

We attribute the first rapid increase in the protection level to formation of small Aβ₄₂ oligomers (dimer, trimer, and tetramer, etc.), in line with previous work (34, 57). The plateau that follows indicates no additional significant hydrogen-bond formation or conformational change. During this stage, larger oligomers are forming with little increase in protection as Aβ₄₂ molecules add to the large soluble aggregates. After some time, the oligomers do reorganize, in an autocatalytic fashion, to a structure showing higher protection. The upward-breaking sigmoidal curve represents this reorganization and is consistent with the behavior of other amyloid-like proteins (i.e., they undergo nucleation, growth, and stabilization) (51). Whereas the rate of nucleation is slow, giving nearly constant HDX, the growth rate is much faster, giving the sigmoidal behavior of HDX. Similar sigmoidal behavior was reported recently for Aβ₄₂ aggregation followed by a fluorescent tetramethylrhodamine-labeled Aβ₄₂ derivative (57). Our approach has an advantage in that it can be used directly without the need for adding amino acids or fluorophores to the protein.

Application of Finke–Watzky Modeling and Statistical Evaluation by a Bootstrap Strategy.

Amyloid-like proteins usually undergo a self-catalyzed aggregation process, which involves one nucleation step and one growth step. Although there are many models to describe a slow, continuous nucleation, we chose the simplest one to fit the time-dependent aggregation of Aβ₄₂ (Fig. 2), following the recommendation of Finke and coworkers (51). We modified the Finke–Watzky (F–W) model by adding a “dimerization” process (Eq. 1). The model describes a reaction that proceeds from a monomer Aβ in state A to a “dimer” in state B (Eqs. 1–3). Admittedly, state B, consisting of a heterogeneous mix of small oligomers, is more complicated than can be addressed in this modified F–W model. Nevertheless, considering it as dimer, we subsequently allowed its transformation to state C by nucleation of large, soluble oligomers, heterogeneous in structure and number. The model includes a forward rate constant k₁ of oligomerization, a rate constant k₋₁ for the reverse reaction, a forward nucleation rate constant k₂ from B to C, and a forward self-catalysis rate constant k₃ involving B and C. A useful parameter that can allow for a direct comparison of different aggregation curves is t_1/2, which is the time point at which the number of Aβ molecules in state C equals the remaining Aβ molecules.

In fitting the data to the F–W model, we sought a single parameter to characterize the outcome. The t_1/2 value is likely that parameter, and it could be obtained by averaging the t_1/2 values from three determinations. However, to obtain a better measure and use the full statistical value of the data, we evaluated the statistics by using a bootstrap resampling strategy. This gives a more reliable t_1/2 along with its precision, permitting the use of a simple t test for establishing differences. By applying the bootstrap strategy (Supporting Information) to the F–W model (52–54) (Fig. 3 and Fig. S6), we retrieved not only t_1/2 values, but also rate-constant information and precision. Although there are many families of k₋₁, k₁, k₂, and k₃ values that fit the experimental data, they are highly correlated (i.e., an increase in k₂ and a commensurate decrease in k₃ gives a good fit to the model). There is, however, a nearly constant t_1/2 for all of the fits for one peptide. More importantly, the t_1/2 values corresponding to the three peptides are different. The middle region (i.e., 20–35) has the shortest t_1/2 (1,070 ± 30 min), whereas C-terminus (i.e., 36–42) and N-terminus (i.e., 1–19) regions have longer t_1/2 values (1,230 ± 30 min and 1,420 ± 20 min, respectively; Fig. S7). That these differences among the three peptides are statistically significant is supported by the t test at ≥99.99% confidence level, made possible by using the bootstrap strategy. Although it may be surprising that the outcome treats the three peptides as separate entities, we note that proteins are complex species and give, for example, different HDX rate constants for different regions. These results now establish that the middle region (i.e., residues 20–35) is the first to report entering an organized state. This is consistent with most molecular dynamics simulations that point to the middle region as either a folding nucleus or a dimerization interface (58–60). The hinge region identified in the solid-state NMR results for Aβ₄₂ fibrils is from S26 to I31 (28), which lies in the middle of the region we identified here. This smaller region may also be important in initiating aggregation, as suggested by aggregation studies of myriad mutation forms. For example, Flemish (A21G) (61), Arctic (E22G) (62), Dutch (E22Q) (61, 63), Iowa (D23N) (64), and others (65) exhibit different aggregation behavior compared with the wild-type Aβ₄₂, supporting that the middle region of Aβ is important in the aggregation process.

Fig. 3.

Flowchart for the modified F–W modeling and bootstrap strategy.

The C terminus (residues 36–42) aggregates second, confirming that it plays a smaller but still important role in aggregation (49). All of the residues in this region (i.e., VGGVVIA) are hydrophobic, serving as an interface for Aβ₄₂ aggregation. Besides, the only differences between Aβ₄₀ and Aβ₄₂ are the last two residues of this region (i.e., IA) of Aβ₄₂.The strong differences in aggregation between Aβ₄₀ and Aβ₄₂ hint at the importance of the hydrophobic interaction within this region. As a result, we propose that without the last two residues the hydrophobic interaction would be insufficient to “pull” small aggregates together and form larger aggregates.

The N terminus (residues 1–19) aggregates last, consistent with its hydrophilic nature. Unlike other approaches showing that the N terminus is flexible (28), our data show some ordering in this region, and as expected, it is the last involved in the aggregation process. The observation that mice never develop AD, however, may be consistent with the importance of the N terminus, because the three mutations between mouse vs. human Aβ represent amino acid substitutions in this region (i.e., R5G, Y10F, and H13R). Nevertheless, the N terminus is involved in aggregation, and this might be the cause for the structural difference between soluble Aβ aggregates and insoluble Aβ fibrils. In addition, the structural difference may be relevant to the different levels of neurotoxicity exhibited by soluble Aβ aggregates vs. insoluble fibrils. Thus, these pulsed HDX studies should have an important impact in the design of drugs that can alter the aggregation of Aβ species and ultimately their neurotoxicity.

Pulsed HDX Analysis of Aβ₄₂ Fibrils.

Even though the liquid chromatography (LC)–MS/MS identification showed full coverage for the peptides, there was some undigested Aβ₄₂ observed as monomer mass in the above experiments, and its corresponding relative peak intensity increased with increasing incubation time. Our online pepsin digestion has been tested in many other projects (66, 67), and the digestion efficiency approaches 100% for proteins, most of which are larger than Aβ₄₂. The Aβ₄₂ peptide showed the lowest digestion efficiency, suggesting the presence of amyloid fibrils or other HMW aggregates that are slow to proteolyze. The protection levels of these undigested species (Fig. S8A) are higher than that of Aβ₄₂ species that can undergo pepsin proteolysis. Given that we observed no significant peaks corresponding to the monomeric mass of Aβ₄₀ in its aggregation experiments, the result for Aβ₄₂ is further evidence that the amyloid fibrils are the source of peaks representing undigested Aβ₄₂.

To study directly the amyloid fibrils, we applied pulsed HDX to the preformed Aβ₄₂ fibrils, which represent the final stage of Aβ aggregation. Digestion of amyloid fibrils is difficult (31, 68) and usually requires the use of “strong” chemicals including hexafluoroisopropanol or DMSO to dissociate all aggregates into Aβ₄₂ monomer (49). This time-consuming and MS-unfriendly digestion protocol is not suitable for an HDX platform. We tested the effectiveness of the HDX quenching procedure (3 M urea with 1% TFA) by using transmission EM and found the protocol was able to dissociate to some extent the amyloid fibrils, allowing us to measure the HDX of the fibrils (Fig. S9).

Upon pulsed HDX of the preformed fibrils we observed both the peptides formed by pepsin digestion (i.e., 1–19) and the undigested Aβ₄₂. The peptides and the full Aβ₄₂ are similarly protected (89 ± 1% for the undigested Aβ₄₂ and 85 ± 1%, 84 ± 1%, and 91 ± 2% for 1–19, 20–35 and 36–42, respectively), consistent with the peptides’ being proteolytic fragments from the amyloid fibrils. This conclusion is consistent with a study that shows the recycling mechanism of Aβ₄₂ fibrils with the Aβ₄₂ monomer (45). The monomer thus formed carries information of the “imprinted” fibrils because quenching HDX preserves information before MS analysis. These monomers digest well with pepsin owing to a lack of the highly packed structure characteristic of the amyloid fibrils. Thus, the peptide fragments are useful “fibril reporters” because they contain the same protection as their precursor fibrils. These amyloid fibrils, which become ultimately the main component in the incubation, are considerably more difficult to interrogate because they have a highly compact structure. Nevertheless, the highly organic mobile phase we used is sufficient to denature some of the fibrils and release detectable amounts of Aβ₄₂ monomer (31). The protection level of undigested Aβ₄₂ seen in the previously mentioned experiments (Fig. S8A) is similar to that of the preformed fibrils, both digested and undigested, indicating that the full Aβ₄₂ peptide originates from the Aβ₄₂ fibrils whereas the peptic peptides arise from soluble species.

To rule out a significant contribution of undigested Aβ₄₂ to our time-dependent studies described earlier, we measured directly the digestion of amyloid fibrils by examining the relative peak-area ratios in extracted ion chromatograms (EICs) of the peptide fragments and the undigested Aβ₄₂. We used the N-terminal region (i.e., 1–19) as a measure of fibril digestion, and the ratio of its EIC peak area compared with that of the undigested Aβ₄₂ is 0.078, indicating that only a small fraction of fibrils undergo digestion. Thus, the contribution of fibril digestion to the peptide-fragment signals is likely small because the peptide 1–19 signal area in the EIC was considerably greater than the EIC area corresponding to the undigested protein (discussed below). Because HDX reports an ensemble average, the protection levels seen in the digested peptides are a measure of soluble Aβ₄₂ aggregates, rather than of the insoluble Aβ species.

Pulsed HDX Analysis of Factors Affecting Aβ₄₂ Aggregation.

One motivation for the HDX pulsed platform is to develop a tool to investigate the many factors that affect Aβ₄₂ oligomerization and aggregation (and of other amyloid-forming proteins). Such a platform should rapidly provide insight on those factors that affect Aβ₄₂ aggregation. Shifts in or losses of the sigmoidal behavior are a measure of the effect. One factor of high interest to us is that of Cu²⁺, which has been previously implicated in Aβ₄₂ aggregation (14, 18, 19). Thus, we added Cu²⁺ to the PBS buffer before mixing with Aβ₄₂ such that the final stoichiometry of Cu²⁺ and Aβ₄₂ is 1:1 (molar ratio). The pulsed HDX results of Aβ₄₂ incubation in the absence of Cu²⁺ provide the reference point for this experiment. The protection level of Aβ₄₂ incubated with Cu²⁺ is remarkably linear and unchanged for the three peptic peptides (i.e., 1–19, 20–35, and 36–42; Fig. 2, dashed lines; raw data shown in Fig. S3), in contrast to their sigmoidal behavior in the absence of Cu²⁺ (Fig. 2, solid lines; raw data shown in Fig. S2). The plateau stage of protection in the presence of Cu²⁺ is similar to that represented by the first plateau for 1–19 in the absence of Cu²⁺ (55%). The protection is lower than that of both 20–35 and 36–42 (56% vs. 60% and 60% vs. 75%, respectively).

We hypothesize that the differences of protection in the presence and absence of Cu²⁺ is that Cu²⁺ stabilizes different forms of Aβ aggregates by interacting with them. The coordination sites for Cu²⁺ are likely on the N terminus (i.e., H6, H13, and H14) (14, 17), and coordination to these sites would lead to more protection of that region of Aβ₄₂. By contrast, the hydrophobic interactions between the C termini of Aβ peptides may be diminished to compensate for the Cu²⁺ coordination. In addition, native gel and Western blotting of the samples corresponding to the largest difference on the pulsed HDX curve show much less HMW soluble aggregates (50–110 kDa) and more LMW aggregates (i.e., mainly trimer and tetramer) in the presence vs. absence of Cu²⁺ (Fig. 2, Right). This result reinforces our previous report showing that Cu²⁺ stabilizes the soluble Aβ₄₂ aggregates and inhibits formation of amyloid fibrils (14, 19).

We also tested the ability of our pulsed HDX approach under rapidly aggregating conditions such as higher temperature and with agitation. We chose a higher and more physiologically relevant temperature (37 °C) and agitated the solution (at 150 rpm, Fig. 4, solid line; raw data shown in Fig. S4). Compared with the results at 25 °C in the absence of Cu²⁺ and without agitation, the aggregation behavior speeded up considerably. The lag phase disappears, indicating that the transformation from LMW aggregates to HMW aggregates is expedited. Next, we repeated the experiment at higher temperature and subjected the solution to agitation, as well as added Cu²⁺ to the solution. According to the results just discussed, applying higher temperature and agitation should decrease the lag phase of Aβ₄₂ aggregation, whereas Cu²⁺ should slow down the aggregation. Indeed, the results (Fig. 4, dashed line; raw data shown in Fig. S5) indicate that the Aβ₄₂ aggregation is faster than that at 25 °C, yet a lag phase is observed in the presence of Cu²⁺. This agrees with the hypothesis that higher temperature and agitation accelerate the aggregation. However, the results showed slower kinetics than that for Aβ₄₂ aggregation at 37 °C, 150 rpm, in the absence of Cu²⁺, providing evidence that Cu²⁺ slows down the Aβ₄₂ aggregation, even at 37 °C, yet without completely preventing it.

Fig. 4.

Pulsed HDX results for three peptides from Aβ₄₂. Aggregation was done at 37 °C, with agitation at 150 rpm and in the absence (solid triangles, circles, and diamonds) or presence of Cu²⁺ (hollow triangles, circles, and diamonds). Peptic peptides 1–19 (A), 20–35 (B), and 36–42 (C) are represented by triangles, circles, and diamonds, respectively. Solid lines are fits using a simple F–W model and dashed lines are fits using a modified F–W model in Mathcad.

As we observed in experiments of Aβ₄₂ incubated at 25 °C, the peaks corresponding to the Aβ₄₂ monomer (undigested species) were observed in all experiments mentioned in this section. The extent of protection of the corresponding undigested species is higher than observed for peptide fragments in the same experiments, and similar to those of both digested and undigested preformed fibrils (Fig. S8 B–D).

The ratios of peak intensities corresponding to Aβ₄₂ that can be digested relative to that resistant to digestion now range from 17 to 1.1. After longer incubation times, more and more Aβ₄₂ becomes resistant to digestion. These ratios are 210 to 14 times higher than the ratio calculated from preformed fibrils described above (i.e., 0.078), indicating that only a very small amount of the fibrils was digested and thus is not expected to interfere to a large extent with the protection levels calculated from these experiments. Importantly, this result indicates that our platform reports only on the soluble Aβ species and not insoluble aggregates.

Conclusion

A new implementation of the pulsed HDX technique interfaced with LC/MS provides an opportunity to examine the details of aggregation of the Aβ peptide at various stages. By using this method, we successfully decoupled the aggregation from the HDX process. Importantly, we extracted kinetic information on the Aβ₄₂ aggregation at 25 °C, indicating that the middle region of the Aβ₄₂ peptide (i.e., 20–35) was the “seeding” region in aggregation, followed by the C-terminus hydrophobic region (i.e., 36–42) and then the N-terminus hydrophilic region (i.e., 1–19). Finally, we showed that this approach allowed us to examine directly the factors that affect the oligomerization of Aβ₄₂. For example, at 37 °C and with agitation, Aβ₄₂ aggregated faster than at 25 °C. To the contrary, the presence of Cu²⁺ slowed down the Aβ₄₂ aggregation, presumably by complexing the polypeptide in the N-terminal region and stabilizing the soluble Aβ₄₂ species.

We envision this approach as a general tool to support future efforts to measure rates of Aβ₄₂ oligomerization and aggregation as a function of various parameters (e.g., concentration, presence of different ligands or proteins, and pH). We also see the need for more effort in using MS site-specific ion activation (i.e., electron-transfer dissociation) to obtain information at the amino acid level and probe in more detail the aggregation interface and the Cu²⁺ binding sites. Alternative footprinting approaches including fast photochemical oxidation may also offer complementary views to understand amyloid formation. Moreover, the method described herein is applicable to the study of various experimental conditions on the oligomerization and aggregation of other amyloid-forming proteins involved in neurodegenerative diseases.