β‐amyloid model core peptides: Effects of hydrophobes and disulfides

Abstract

The mechanism by which a disordered peptide nucleates and forms amyloid is incompletely understood. A central domain of β‐amyloid (Aβ21–30) has been proposed to have intrinsic structural propensities that guide the limited formation of structure in the process of fibrillization. In order to test this hypothesis, we examine several internal fragments of Aβ, and variants of these either cyclized or with an N‐terminal Cys. While Aβ21–30 and variants were always monomeric and unstructured (circular dichroism (CD) and nuclear magnetic resonance spectroscopy (NMRS)), we found that the addition of flanking hydrophobic residues in Aβ16–34 led to formation of typical amyloid fibrils. NMR showed no long‐range nuclear overhauser effect (nOes) in Aβ21–30, Aβ16–34, or their variants, however. Serial ¹H‐¹⁵N‐heteronuclear single quantum coherence spectroscopy, ¹H‐¹H nuclear overhauser effect spectroscopy, and ¹H‐¹H total correlational spectroscopy spectra were used to follow aggregation of Aβ16–34 and Cys‐Aβ16–34 at a site‐specific level. The addition of an N‐terminal Cys residue (in Cys‐Aβ16–34) increased the rate of fibrillization which was attributable to disulfide bond formation. We propose a scheme comparing the aggregation pathways for Aβ16–34 and Cys‐Aβ16–34, according to which Cys‐Aβ16–34 dimerizes, which accelerates fibril formation. In this context, cysteine residues form a focal point that guides fibrillization, a role which, in native peptides, can be assumed by heterogeneous nucleators of aggregation.

Abbreviations

β‐ME

β‐mercaptoethanol

Aβ

β‐amyloid

ESI

electrospray ionization

HSQC

heteronuclear single quantum coherence spectroscopy

MALDI‐TOF

matrix‐assisted laser desorption ionization‐time of flight

NOESY

nuclear overhauser effect spectroscopy

ROESY

rotating‐frame overhauser effect spectroscopy

TCEP

tris(2‐carboxyethyl) phosphine

TFA

trifluoroacetic acid

ThT

thioflavin T

TIS

triisopropylsilane

TOCSY

total correlational spectroscopy

1. INTRODUCTION

Formation of amyloid fibrils generally includes nucleation, with a transition from largely unstructured monomers to aggregates rich in β‐sheets, usually parallel and in‐register. In addition, peptides such as Aβ adopt multiple distinct forms or polymorphs, distinguishable by electron microscopy, solid‐state NMR, and other techniques.1, 2, 3, 4, 5, 6, 7, 8, 9 In this article, we compare two hypotheses about how fibrillizing peptides, such as the Aβ peptides, acquire whatever limited structure they possess. According to a first hypothesis, fibrillization begins with a structured core domain that initiates and guides the limited folding. A second hypothesis is that there are few if any specific interactions; rather, Aβ could first become partially aligned through interactions of only modest specificity, for example, by binding a metal ion, or through interactions between hydrophobic side chains in a micelle‐like aggregate. According to this second hypothesis, there would be no structured core domain.

We compare these hypotheses in the studies reported, below. We report on several model peptides, which are internal fragments of Aβ peptides or variants thereof. Peptides studied include Aβ21–30, Aβ16–34, without or with an N‐terminal Cys extension, and a cyclic version of Aβ21–30. Although these internal fragments are not physiological, they allow us to address questions that we would not be able to address with full‐length Aβ peptides: in initial studies of Aβ1–40 with a Cys residue appended to the N‐terminus, we observed that this peptide formed disulfide bonds very rapidly and was then insoluble even at low peptide concentrations and in fairly harsh solvents (e.g., dimethyl sulfoxide (DMSO)/water/trifluoroacetic acid [TFA] mixtures). This preluded NMR studies. Other investigators had studied Cys‐containing Aβ peptides, but these studies also did not include NMR. For example, scanning Cysteine mutagenesis was used to probe solvent accessibility within fibrils,10, 11 and other Cys‐containing mutant forms of Aβ12, 13 were developed to examine cytotoxicity or small oligomers. The advantage of using the shorter peptides is that they allowed us to ask three questions: (a) Are short internal fragments, such as Aβ21–30, structured, as has been proposed?14, 15, 16, 17, 18, 19, 20, 21, 22 (b) Does the addition of Cys, or cyclization, enhance weak interactions between side chains elsewhere in the peptide, and thereby foster aggregation—that is, by an entropic effect? (c) Does addition of flanking hydrophobic residues convert the internal section of Aβ into an aggregating (fibrillizing) peptide?

We will show, below, that Aβ21–30 does not form fibrils, but rather, remain monomeric at all concentrations tested. Attempts to stabilize transient structures through addition of an N‐terminal Cys residue (Cys‐Aβ21–30) or cyclization (cyclo‐Aβ21–30) also led to peptides that always remained unstructured (CD and NMR) and monomeric, and did not form fibrils. On the other hand, extending this domain at both ends to include stretches of hydrophobic amino acids leads to a peptide that aggregates into fibrils. Furthermore, we will also show that addition of the N‐terminal Cys to Aβ16–34 enhances the rate of this fibril formation.

2. RESULTS

2.1. Rationale for the use of model Aβ peptides

We hypothesized that there is no core domain that organizes the limited folding of Aβ into β‐sheet‐rich amyloid fibrils. To test this hypothesis, we synthesized model peptides to make the following comparisons:

1. We examined peptides from the central region of Aβ peptides, residues 21–30, which has been proposed as an autonomously folding domain.14, 15, 16, 17, 18, 19, 20, 21, 22 In order to stabilize any weak interactions that might exist in this region, we also synthesized a series of peptide variants of Aβ21–30, including Cys‐containing and a cyclic peptide, cyclo‐Aβ21–30. Cys‐Aβ21–30 was monomeric and dimeric only (size exclusion chromatography [SEC], Figure S1a), soluble, and unstructured by CD spectroscopy and was not further studied.

2. We compared the hydrophilic Aβ21–30 peptide to longer peptides containing this domain, but extending it at each end by adding several hydrophobic residues, that is, Aβ16–34. We hypothesized that aggregation depends on the presence of multiple weak interactions through the hydrophobic effect; thus, Aβ16–34 would form fibrils while Aβ21–30 would not.

3. We also compared Aβ16–34 with Cys‐ Aβ16–34, that is, the same peptide as in (2), but containing an additional N‐terminal Cys residue. We hypothesized that this Cys residue would act as a “focal point” for interactions between Aβ molecules and would thereby accelerate aggregation by allowing for dimerization, and bypassing a rate‐limiting step in aggregation.

For (3), we initially synthesized variants of Aβ1–40 and Aβ13–38 containing a single Cys residue appended to the N‐terminus. These peptides, however, in contrast to either unmodified Aβ1–40 or Aβ13–38, rapidly formed disulfide bonds and became insoluble in aqueous buffer near neutral pH and even at pH 3 or 10. They were insoluble, furthermore, even in the presence with of organic solvents (e.g., DMSO), denaturants (e.g., urea), or detergents. For this reason, we turned to shorter congeners of Aβ1–40 that maintained the basic domain structure of the full molecule. Aβ16–34 retains the amino acids in the two β‐sheets of most Aβ fibrils, and the intervening bend region. Initial experiments (discussed below) showed that this peptide also formed fibrils that exhibit thioflavin T (ThT) fluorescence, and had a fibrillar appearance in transmission electron microscopy (TEM) similar to fibrils formed by Aβ1–40. We also found that appending an N‐terminal Cys to Aβ16–34 resulted in a manageable peptide that could be purified, and remained sufficiently soluble in neutral buffers for study by NMR.

Figure 1 and Table S1 show the peptides used in these studies.

Figure 1.

Sequences of Aβ1–40 and internal fragments of Aβ used in these studies. Aβ21–30 and Aβ16–34 were used to make variants with either an N‐terminal Cys residue, or cyclized variants

2.2. Aβ21–30 and variants

Hypothetically, residues 21–30 of Aβ might form an autonomously folding domain that could guide the acquisition of structure, albeit limited, that accompanies aggregation into fibrils in full‐length Aβ peptides, and possibly in the shorter peptides discussed below, such as Aβ16–34. To investigate this possibility, we synthesized Aβ21–30, Cys‐Aβ21–30, and cyclo‐Aβ21–30. All of these peptides were highly soluble (up to ~50 mM), and required no disaggregation procedures, for example, initial dissolution in DMSO, hexafluoroispropanol (HFIP), or 1 mM NaOH. Despite the lack of a disaggregation procedure, Aβ21–30 was monomeric by SEC (Figure S1a) and sedimentation equilibrium analytical ultracentrifugation (M = 963, calculated molecular weight = 947.0 g mol⁻¹, Figure S1b); SEC of Cys‐Aβ21–30 showed monomer and dimer only. CD spectroscopy of this peptide up to 500 μM indicated no structure (Figure 2a), and the mean residue ellipticity did not change at the concentrations tested, which also indicates the absence of aggregation. CD spectra of cyclo‐Aβ21–30 up to 500 μM were essentially identical and indicated no structure (Figure 2b). Aβ21–30, Cys‐Aβ21–30, and cyclo‐Aβ21–30 did not form fibrils. Analysis of ¹H,¹H‐total correlational spectroscopy (TOCSY), ‐rotating‐frame overhauser effect spectroscopy (ROESY), and‐nuclear overhauser effect spectroscopy (NOESY) (Figure 2c,d) ¹H,¹⁵N‐heteronuclear single quantum coherence spectroscopy (HSQC) and ¹H,¹³C‐HSQC for Aβ21–30 (Figure S2, Table S2 a,b), Cys‐Aβ21–30 (not shown), and cyclo‐Aβ21–30 (Figure S3, Table S3 a,b) indicated no long‐range connectivities (for some small peptides and other molecules, NOESY signal can approach zero or become negative. Thus, it was possible that this peptide would have little or no NOESY signal, but this clearly is not the case). The chemical shifts obtained from these spectra were analyzed by TALOS+ as described in Section 4.11. In all cases, no secondary (or higher order) structure was detected; rather, all peptides were predicted to be coils, by both sequence and NMR chemical shift data. These experiments did not include a disaggregation procedure; hence, while trace amounts of organic solvent (DMSO or HFIP) might, in theory, abrogate or eliminate weak structures under other circumstances, this clearly was not the case under our conditions.

Figure 2.

Circular dichroic and NMR spectroscopy of Aβ21–30 and cyclo‐Aβ21–30. (a) Circular dichroic spectra of Aβ21–30 at concentrations of 200–500 μM, as indicated in the legend. (b) Circular dichroic spectrum of cyclo‐Aβ21–30 at 500 μM. In (a) and (b), the spectra are plotted as Mean Residue Ellipticities, which do not vary with peptide concentration, consistent with monomeric status. (c) Superimposed ¹H,¹H‐TOCSY (blue) and ¹H,¹H‐ROESY (red) spectra of Aβ21–30. We also acquired ¹H,¹H‐NOESY spectra (not shown), which were essentially the same as the ¹H,¹H‐ROESY spectra. (d) Superimposed ¹H,¹H‐TOCSY (blue) and ¹H,¹HNOESY (red) spectra of cyclo‐Aβ21–30. NOESY, nuclear overhauser effect spectroscopy; ROESY, rotating‐frame overhauser effect spectroscopy; TOCSY, total correlational spectroscopy

The NMR finding around which the proposal of a folded structure for this region rested was a connectivity in the ¹H,¹H‐ROESY spectrum of Glu22 Hα to Ala30 HN.14 A subsequent paper by the same group assigned the peak to two overlapping peaks, Glu22 Hα to Ala30 HN and Lys28 Hα to Ala30 HN.15 Using a higher field instrument (900 rather than 500 MHz), however, another group20 showed that there were no long range rotating‐frame overhauser effect (ROEs), and only weak medium range ROEs; furthermore, the peak in question was an i,i + 2 contact between Lys28 HR and Ala30 HN. Like the latter group, we also observed no long range ROEs or NOEs, albeit under somewhat different solvent conditions. Nevertheless, under our solvent conditions, we conclude that these peptides are monomeric and unstructured.

2.3. Initial characterization of Aβ16–34 and Cys‐Aβ16–34

Initial characterization of these peptides included ThT fluorescence as a marker of fibril formation, electron microscopy of fibrils, SEC, and sedimentation equilibrium analytical ultracentrifugation.

As expected for fibril‐forming peptides, both Aβ16–34 and Cys‐Aβ16–34 displayed typical ThT fluorescence after ~1–3 days (Figure 3a). Fluorescence was greater for Cys‐Aβ16–34 than for Aβ16–34 under the same fibrillization conditions. Adding 0.1% (vol:vol) β‐mercaptoethanol (β‐ME) at the start of the fibrillization reaction greatly diminished ThT fluorescence of Cys‐Aβ16–34. Transmission electron microscopy of both peptides indicated the presence of typical amyloid fibrils (Figure 3b,c).

Figure 3.

Initial characterization of Aβ16–34 and Cys‐Aβ16–34. (a) Thioflavin T fluorescence of Aβ16–34 (green), Cys‐Aβ16–34 (blue), Cys‐Aβ16–34 in the presence of β‐ME (red). Points represent mean ± SE for three replicate samples. Note that the two y‐axes have different scales. (b, c) Transmission electron micrographs of fibrils of (b) Aβ16–34 and (c) Cys‐Aβ16–34. Peptides were disaggregated as described in Methods and allowed to fibrillize at 37°C under quiescent conditions. Magnification is ×15,000. (d, e) Size exclusion chromatography of Aβ16–34 and Cys‐Aβ16–34, respectively, at various times, as indicated in figure. (f) Analytical ultracentrifugation (sedimentation equilibrium) of Aβ16–34. (g) Residuals for the fit of data shown in panel (f). β‐ME, β‐mercaptoethanol

Both peptides were mainly or entirely monomeric when first dissolved. This was shown by SEC (Figure 3d,e) and analytical ultracentrifugation (sedimentation equilibrium, Figure 3f,g). In the case of Aβ16–34, the size of the monomer peak gradually diminished over 3 days. No oligomer peak was observed in SEC. Analytical sedimentation equilibrium ultracentrifugation confirmed the predominance of monomers: M = 2,584, calculated molecular weight = 1,978.3 g mol⁻¹. In SEC of Cys‐Aβ16–34, there was a more rapid decline in the monomer peak than for Aβ16–34. In addition, peaks with elution positions of dimers and oligomers were observed.

2.4. Formation of disulfide bonds by Cys‐Aβ16–34

Initial observations from SEC and ThT fluorescence indicated that Cys‐Aβ16–34 formed fibrils more rapidly than Aβ16–34. Furthermore, these experiments also indicated that the final solubility of Aβ16–34 was greater than that of Cys‐Aβ16–34. As a preliminary to NMR experiments discussed below, we measured the rate of disulfide bond formation as a function of temperature. The rate of disulfide bond formation (measured using the Ellman reagent) increased with temperature (Figure 4), as expected. Kinetics were analyzed as a second‐order reaction, that is,

$$A=\frac{\left(A_0-A_{\mathrm{eq}}\right)}{\left(1+k{A}_{0}t\right)}+A_{\mathrm{eq}}$$ (1)

where A = absorbance at 412 nm (from Ellman reagent), A ₀ = A _412nm at t = 0, V _eq = A _412nm extrapolated to infinite time, t = time, and k = rate constant. The rate constants fit the Arrhenius equation (Figure 4, inset). As discussed below, the second‐order rate equation is an approximation, albeit a fairly accurate one, for the kinetics scheme used below in our analysis of NMR data.

Figure 4.

Concentration of thiols, assayed using the Ellman reagent, of Cys‐Aβ16–34, treated as in serial TOCSY experiments, at various temperatures (from 5°C to 25°C). Initial values are normalized to unity. Lines are nonlinear least squares fit to second‐order reaction equation. Inset shows an Arrhenius plot of the second‐order rate constants; least square fit yielded $\ln \left(k\right)=\ln \left(A\right)-\frac{E_{\mathrm{a}}}{R}\left(\frac{1}{T}\right)=16.33-5.5E03\left(\frac{1}{T}\right)$, where k = rate constant, A = pre‐exponential factor, E _a = activation energy, T = absolute temperature, and R = 8.314 × 10⁻³ kJ mol⁻¹ K⁻¹; fit of straight line, r = 0.98. TOCSY, total correlational spectroscopy

2.5. NMR spectroscopy of Aβ16–34 and Cys‐Aβ16–34

Assignments of peaks were made by standard methods; peak heights and volumes were measured using NMRView software. In addition to ¹H,¹H‐TOCSY and ‐NOESY spectra, spectra acquired included ¹³C‐ and ¹⁵N‐HSQC of Aβ16–34 and Cys‐Aβ16–34 (Figures S4–S7). Chemical shifts are listed in Tables S4 and S5 for Aβ16–34 and Cys‐Aβ16–34, respectively.

No long‐range nOes were observed for Aβ16–34 and Cys‐Aβ16–34, consistent with CD spectra showing no stable secondary structure (not shown). No peak broadening accompanied loss of peak volume and height in any of the time courses, that is, the volume/height ratio showed no change over time. Although soluble oligomeric intermediates are observed for most or all fibril‐forming peptides, and indeed were observed by other methods in the case of Cys‐Aβ16–34 (Figure 3e; see also Figure S10, discussed below), no new NMR peaks were observed over the course of these experiments. The lack of such peaks could be due to low concentrations of oligomeric species, and/or peak broadening for larger oligomers, and/or similarity in chemical shifts between monomers and dimers or oligomers.

In order to compare the fibrillization of Aβ16–34 and Cys‐Aβ16–34 at a site‐specific as well as a global level, we acquired serial ¹H,¹H‐TOCSY spectra, and in some cases, ¹H‐NOESY spectra for these peptides. As indicated above (SEC, ThT fluorescence kinetics), Cys‐Aβ16–34 appeared to form fibrils more rapidly than Aβ16–34. This was confirmed in the experiments to be described below.

We determined conditions that would allow a direct comparison of Aβ16–34 and Cys‐Aβ16–34. Initial experiments showed that at appropriate peptide concentrations for observing both fibrillization and NMR spectra (≥100 μM), the optimal temperature for following serial (time‐dependent) TOCSY spectra was 15°C. The read‐out of serial TOCSY experiments was decay of peak intensities (volumes and heights) as the fibrillization reaction proceeded. These initial experiments showed that at 5°C, Cys‐Aβ16–34 precipitated within ~3 days, but the spectra of Aβ16–34 barely changed at all even over longer time periods, indicating that fibril formation either did not occur or was extremely slow. On the other end of the spectrum, at temperatures ≥20°C, both peptides fibrillized, but many of the peaks could no longer be observed. An additional experimental limitation at 25°C was that Cys‐Aβ16–34 precipitated within ~12 hr, limiting the number of time points and thus the accuracy of the kinetics parameters that could be obtained (TOCSY experiments took 45–90 min, depending on conditions). Therefore, we settled on 15°C as a temperature at which both peptides could be compared.

Time courses were obtained for three separate samples of each peptide. Because Cys‐Aβ16–34 forms disulfide bonds while Aβ16–34 cannot, different schemes were used to analyze the kinetics of peak decay for these two peptides, as discussed in the next sections.

2.6. Kinetics of peak decay for Aβ16–34

For Aβ16–34, ¹H,¹H‐TOCSY peak volumes (and heights) decayed over the course of 2–3 days, and this decay was analyzed as a pseudo‐first‐order approach to equilibrium, that is,

$$V=\left(V_0-V_{\mathrm{eq}}\right)e^{-\mathit{kt}}+V_{\mathrm{eq}}$$ (2)

where V = peak volume, V ₀ = peak volume at t = 0, V _eq = peak volume extrapolated to infinite time, t = time, and k = rate constant. That Aβ peptide fibrillization approached an equilibrium was in agreement with many previous studies, especially by those of Wetzel and associates.23, 24 The average rate for sites within Aβ16–34 (Figures 5 and S8) was 0.016 ± 0.004 hr⁻¹ (mean ± SD). ¹H spectra also were acquired after each of the serial TOCSY experiments, and the peaks in the amide region (7.0–9.0 ppm) were integrated to obtain area under the peaks. These values similarly decayed over the time course of the experiment (Figure S9). Analyzed using the same equation, the pseudo‐first‐order rate constant was 0.026 ± 0.0006 hr⁻¹ (mean ± SD), in reasonable agreement with the mean of rate constants obtained from the serial TOCSY spectra. The mean value for V _eq (Equation 2) was 0.810 ± 0.0510. For all peaks volumes, V _eq was higher for Aβ16–34 than for Cys‐Aβ16–34 (kinetics analyzed below).

Figure 5.

Rate constants obtained from analysis of serial TOCSY spectra of Aβ16–34. Data were analyzed as a pseudo‐first‐order approach to equilibrium, as described in text. Data represent means from three experiments on replicate samples; error bars are SDs for the rate constant parameter. TOCSY, total correlational spectroscopy

2.7. Kinetics of peak decay for Cys‐Aβ16–34

In contrast to Aβ16–34, attempting to fit data on Cys‐Aβ16–34 showed clear disagreement with the pseudo‐first‐order kinetic scheme. We hypothesized the following kinetic scheme:

Whereas reactions of both Aβ16–34 and Cys‐Aβ16–34 represented an approach to equilibrium, dimerization of Cys‐Aβ16–34 is followed by very rapid precipitation of the peptide, with little dimer remaining in solution, while monomer precipitates at a rate similar to that of Aβ16–34. This is borne out by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS‐PAGE) of peptide remaining in solution at the end of the kinetics experiment (Figure S10), which shows a higher proportion of monomeric peptide remaining in solution than in the pellet, which was mostly dimeric. Thus,

$$2A\stackrel{k_2}{\to }{A}_2$$ (3a)

$$A+A_n\stackrel{k_1}{\to }{A}_{n+1}$$ (3b)

where A and A ₂ are monomeric and dimeric peptide, A _n and A _n+1 fibrils or large oligomers not observable in TOCSY experiments, and k ₁ and k ₂ are rate constants. Thus, the rate equation is given by:

$$\frac{\mathit{dA}}{\mathit{dt}}=-k_1\left[A_n\right]\left[A\right]-k_2{\left[A\right]}^2$$ (4)

that is, parallel pseudo‐first‐ and second‐order reactions. Since the concentrations of oligomers and fibrils are low compared with monomer and do not undergo large changes with time, the first term on the right side of the above equation can be written as a pseudo‐first‐order reaction, that is,

$$\frac{\mathit{dA}}{\mathit{dt}}=-k^{\prime }\left[A\right]-k_2{\left[A\right]}^2$$ (5)

where k′ is the pseudo‐first‐order rate constant. Integrating this differential equation yields:

$$A=\frac{\left(A_0-A_{\mathrm{eq}}\right)k^{\prime }}{\left(\left(A_0-A_{\mathrm{eq}}\right)k_2+k^{\prime }\right)e^{k^{\prime }t}-\left(A_0-A_{\mathrm{eq}}\right)k_2}+A_{\mathrm{eq}}$$ (6)

Or, put into terms of peak volumes:

$$V=\frac{\left(V_0-V_{\mathrm{eq}}\right)k^{\prime }}{\left(\left(V_0-V_{\mathrm{eq}}\right)k_2+k^{\prime }\right)e^{k^{\prime }t}-\left(V_0-V_{\mathrm{eq}}\right)k_2}+V_{\mathrm{eq}}$$ (7)

The solution of this differential equation is given in Supporting Information. From this analysis (Figures 6a,b and S11), the pseudo‐first‐ and second‐order rate constants averaged over all the analyzed TOCSY cross peaks were 0.0022 ± 0.0064 hr⁻¹ and 0.10 ± 0.01 volume units (VU)⁻¹ hr⁻¹, respectively (mean ± SD). Additional statistical analyses of these fits include calculation of the Akaike information coefficient25, 26, 27 (see Supporting Information) and likelihood functions28, 29, 30 (Figure S12).

Figure 6.

Rate constants obtained from analysis of serial TOCSY spectra of Cys‐Aβ16–34. Data were analyzed as a combined second‐order and pseudo‐first‐order approach to equilibrium, as described in text. Data represent means from three experiments on replicate samples; error bars are SDs for the rate constant parameter. (a) Pseudo‐first‐order rate constants. (b) Second‐order rate constants. TOCSY, total correlational spectroscopy

Although the units of these second‐order rate constants cannot be compared directly to those obtained from disulfide bond formation (Figure 4), the first half‐lives were quite similar: ~10 hr at 15°C in both cases. An additional point is that the range of second‐order rate constants was fairly narrow, from 0.08 to 0.13 VU⁻¹ hr⁻¹. This point is also reflected in the fact that the SD for the rate constants was ~10% of the mean value, that is, 0.10 ± 0.01 VU⁻¹ hr⁻¹. Thus, in addition to accelerating the fibrillization of Cys‐Aβ16–34 compared with Aβ16–34, the addition of the Cys residue has an ordering effect on aggregation. The mean value for V _eq for Cys‐Aβ16–34 was 0.24 ± 0.049 (mean ± SD), compared with 0.81 ± 0.051 for Aβ16–34, confirming that Cys‐Aβ16–34 was less soluble than Aβ16–34.

Finally, the above serial TOCSY experiment was repeated in the presence of 100 mM tris(2‐carboxyethyl) phosphine (TCEP). This concentration of TCEP retards but does not eliminate disulfide bond formation. In general, the second‐order rate constants were ~72% those observed in the absence of TCEP (Figure S13).

3. DISCUSSION

In this article, we compared internal fragments of Aβ, all of which contain the central segment, Aβ21–30, a hydrophilic, flexible region which, in Aβ fibrils and some oligomers, forms a “bend” between the two parallel, in‐register β‐sheets.31, 32, 33, 34 In the longer peptides, Aβ16–34 and Cys‐Aβ16–34, as in Aβ1–40 and Aβ1–42, this central domain is flanked by stretches of hydrophobic amino acids. We compared Aβ16–34 to Cys‐Aβ16–34 and found that the Cys residue increased the rate of fibril formation, and had other effects on kinetics (discussed further, below). Both peptides formed typical amyloid fibrils, as shown by ThT fluorescence and transmission electron microscopy. Both peptides were unstructured in solution by CD spectroscopy, and in ¹H,¹H‐NOESY (and ¹H,¹H‐ROESY) experiments, no long‐range nOes were observed in any of the peptides examined.

The results above demonstrate the following three points:

1. A putative structured core domain (Aβ21–30) is monomeric and highly soluble, and is unstructured by CD and NMR spectroscopy. Although in theory this region could serve as a “focal point” to organize aggregation, we saw no evidence for structure in Aβ21–30, nor in congeners of Aβ21–30 in which we attempted to stabilize marginally stable structures in Aβ21–30 by the addition of a Cys residue or cyclization.

2. Aβ16–34 and Cys‐ Aβ16–34, in which the above peptide is extended to include hydrophobic residues at each end, was able to aggregate and form typical amyloid fibrils.

3. The addition of a Cys residue to the N‐terminus of Aβ16–34 yields a peptide, Cys‐Aβ16–34, that fibrillizes more rapidly, and has a lower final solubility that Aβ16–34 itself. The kinetics of fibrillization was followed in serial NMR experiments. The more rapid fibrillization and lower final solubility were apparent in serial ¹H,¹H‐TOCSY experiments, where the final peak volumes (V _eq in Equations 2, 7) were uniformly higher for Aβ16–34 than for Cys‐Aβ16–34. No peak broadening was observed. The disulfide bond formed by Cys‐Aβ16–34 appears to provide a “focal point” for rapid fibrillization of the peptide. Aβ16–34, lacking such a focal point, forms fibrils more slowly, and retains greater final solubility.

In Cys‐Aβ16–34, the N‐terminal Cys allowed formation of stable dimers, and once formed, these dimers fibrillized very rapidly. The disulfide bonded Cys residues, then, seemed to act as a “focal point” which organized the aggregation process. Once the disulfide was formed, the remainder of the peptide appears to “zip up” and form an insoluble fibril. Although there are no Cys residues in full‐length Aβ peptides, of course, metal ions35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45 or lipid surfaces,46, 47 among other heterogeneous nucleators could subsume the role of the Cys residue in Cys‐Aβ16–34 and act as a “focal point” for aggregation of full length Aβ peptides.

The formation of amyloids begins with nucleation, at which point the peptide begins to acquire β‐sheet structure, developing in a previously unstructured peptide. The intermediates and final products in the process of fibril growth (extension) would be too large to observe by solution NMR. Thus, we infer that our observations bear upon the early stages of amyloid formation, including nucleation.

We propose, then, that the nucleation of Aβ aggregation could consist of two theoretically distinct processes, illustrated in the scheme of Figure 7. Metal ions, for example, or other types of heterogeneous nucleators could lead to peptide‐peptide association; this would, in turn, bring hydrophobic side chains close to one another and thereby favor a more stable complex made through the hydrophobic effect. An additional factor is the FF motif in full‐length Aβ peptides and in Aβ16–34 and Cys‐Aβ16–34, but not Aβ21–30: this motif has been shown to form 3₁₀ helices which can act as an oligomerizing precursor to amyloid fibrils, especially in a membrane environment.48, 49, 50, 51

Figure 7.

Schematic showing how the Cys residue in Cys‐Aβ16–34 might accelerate and order fibrillization of this peptide compared with that of Aβ16–34. (a) Aβ16–34, containing no Cys or metal‐binding sites, is unstructured in solution, and self‐associates only through weak and nonspecific interactions, for example, between hydrophobic side chains. (b) Cys‐Aβ16–34 is also unstructured in solution, but can form disulfide bonds. (c) Formation of disulfide bonds favors interactions between side chains and backbone moieties in the linked chain, and thus can serve as a nidus to accelerate and order formation of fibrils. (d) As additional molecules bind to this growing aggregate, with continued disulfide bond formation, the aggregate grows and precipitates

As expected, NMR peak decay of Aβ16–34 and Cys‐Aβ16–34 followed different kinetic schemes. The kinetics of peak decay for Aβ16–34 was consistent with a pseudo‐first‐order rate equation. In contrast, the kinetics of peak decay for Cys‐Aβ16–34 was dominated by the formation of disulfide bonds, and hence by a second‐order reaction. Addition of TCEP to Cys‐Aβ16–34 reduced the level of disulfide bond formation and slowed the rate of decay of peak volumes.

One of the fundamental features of amyloids is their structural polymorphism. At the same time, seeding solutions of an amyloidogenic peptide with fibril seeds tends to produce replicate fibrils,1, 4, 52, 53, 54, 55 indicating that much or most of this polymorphism arises during the nucleation phase, not during fibril growth/extension. Furthermore, the degree of polymorphism is not necessarily the same for all amyloidogenic peptides, so there can be greater or lesser degrees of polymorphism for various amyloids. The kinetics of fibrillization of Aβ16–34 and Cys‐Aβ16–34 suggest that in the latter peptide, there is a nidus at which aggregation starts, and this nidus not only accelerates but also orders the aggregation process. This could lower the degree of polymorphism of the latter peptide. For Aβ16–34, which lacks both the disulfide and the His residues that bind divalent metals ions in Aβ1–40, the eventual formation of fibrils appears to depend mainly on what remains in this peptide out of all the residues in Aβ1–40, that is, the hydrophobic (mainly aliphatic) amino acids. The hydrophobic amino acids are sufficient to lead to fibril formation of Aβ16–34, since Aβ21–30 and cyclo‐Aβ21–30, lacking these residues, do not fibrillize at all. Fibrillization of Aβ16–34, however, is a slow and apparently random process that is delayed by the lack of a clearly defined starting point, which the Cys residue in Cys‐Aβ16–34 or the metal ions in Aβ1–40 provides.

4. EXPERIMENTAL PROCEDURES

4.1. Synthesis of linear peptides

Peptides synthesized for these studies are depicted in Figure 1 and listed in Table S1. Peptides were synthesized using an Applied Biosystems 433A synthesizer (Foster City, CA) as previously described.56 Briefly, peptides were synthesized with an Fmoc‐Wang resin on a 0.25 mM scale with FastMoc chemistry, using N‐hydroxybenzotriazole and 2‐(1H‐benzotriazol‐1‐yl)‐1,1,3,3,‐tetramethyluronium hexafluorophosphate as coupling reagents (CSBio). Coupling times were extended by ~3.3 times normal, and residues V18, V24, G25, S26, and N27 were doubly coupled. After synthesis, peptides were cleaved from the resin in 95% TFA, 2.5% water, and 2.5% triisopropylsilane (TIS) for Aβ21–30 and Aβ16–34; and in 94% TFA, 2.5% H₂O, 2.5% TIS, and 1.0% ethanedithiol for Cys‐Aβ21–30, Cys‐ Aβ16–34, and Aβ13–38 (all vol:vol). After 2 hr of cleavage, peptides were purified by reverse phase high performance liquid chromatography (HPLC), using a Preparative Zorbax C18 column, on a Hewlett‐Packard 1050 HPLC, with a gradient between solvents A (0.1% [vol:vol] TFA in water) and B (0.1% TFA [vol:vol] in acetonitrile), the exact gradient depending on the peptide. Peptides were checked for purity (>95%) by analytical reverse phase (C18) HPLC and matrix‐assisted laser desorption ionization‐time of flight (MALDI‐TOF) or electrospray ionization liquid chromatography (ESI LC)‐mass spectrometry, as previously described.56 All peptides were stored at −20°C or −80°C.

4.2. Synthesis of cyclic peptides

To make cyclic peptides, peptides were made with an N‐terminal Cys residue, incorporated in place of one of the Ala residues (as indicated below), and a C‐terminal thioester, so that cyclization could be carried out by a native chemical ligation reaction.57, 58, 59 Thus, we synthesized NH–CEDVGSNKGA–SR, which is Aβ21–30, with Cys substituted for Ala21, and with a C‐terminal thioester (R = leucyl mercaptopropionic acid). Peptide chains were extended using optimized manual Boc Solid Phase Peptide Synthesis.57, 58, 59 Peptides were then simultaneously deprotected and cleaved from the resin by treatment with anhydrous hydrogen fluoride (HF) containing p‐cresol (9:1, vol:vol) for 1 hr at 0°C. After evaporation of the HF under reduced pressure, crude products were precipitated and triturated with chilled diethyl ether. The peptide products were then dissolved in 50% aqueous acetonitrile and purified, or lyophilized for storage. Cyclization was carried out by native chemical ligation, essentially as described elsewhere.60, 61 Briefly, peptides at ~2 mg/ml were dissolved in 100 mM sodium phosphate, pH 7.20 and 10 mM TCEP. To monitor the course of the reaction, small aliquots of the reaction mixture were examined by analytical reverse phase‐high performance liquid chromatography (RP‐HPLC). Cyclization of the deprotected linear peptides was very efficient, taking less than 1 hr. After forming the cyclic peptide by native chemical ligation, Cys residues were converted to Ala by reductive desulfuration, essentially as described elsewhere.62 Briefly, Raney nickel was prepared by adding 25 mg NaBH₄ to 150 mg Ni(OAc)₂(H₂O)₂ dissolved in 2 ml of deionized H₂O. After 5 min, the slurry was filtered through a medium sintered glass funnel. 0.5 mg of peptide was dissolved in 2 ml of 35 mM TCEP, 6 M urea and 200 mM sodium phosphate, pH 6.80. The Raney nickel slurry was added to the peptide solution. The reaction was monitored by HPLC using a C18 column maintained at 60°C, and using a 20:80 to 60:40 acetonitrile:H₂O (both 0.1% TFA, vol:vol) gradient. The peaks were checked using MALDI‐TOF, or more recently, by LC‐ESI mass spectrometry. Purification of cyclic peptides was essentially as described above. An example of this reaction is shown in Figure S14.

4.3. Electron microscopy

Peptides were dissolved directly into 10 mM phosphate buffer, pH 7.20, 0.02% (wt:vol) NaN₃ at room temperature to a final peptide concentration of 500 μM. A disaggregation procedure was not performed, because the goal of these experiments was to determine whether the peptides were capable of forming oligomers and fibrils at all. Solutions were allowed to aggregate at 37°C for 48 hr. Then, solutions were centrifuged at 16,000g for 3 min. Most of the supernatant was discarded, and the remaining solution and fibrillar material was vortexed for 3 s. The slurries were then applied to glow‐discharged, 400‐mesh carbon‐coated support films. All samples were stained with 0.1% (wt:vol) uranyl acetate. Excess solution was wicked off with paper tissue (Kimwipes), and the grids were dried at room temperature. Micrographs were taken on an FEI Tecnai F30st‐STEM microscope at magnifications of ×15,000, ×49,000, and ×98,000. The CCD camera multiplied the magnification by a factor of ×1.4.

4.4. Circular dichroic spectroscopy

Samples were freshly prepared at peptide concentrations of 100–500 μM in 10 mM phosphate buffer, pH 7.20, 0.02% (wt:vol) NaN₃. In general, freshly dissolved samples were used, because at higher concentrations (≥300 μM) of Aβ16–34 and Aβ13–38, fibrils formed to some extent after an incubation of ~4–6 hr. Samples were centrifuged at 16,000g for 3 min prior to measuring the spectra, and only the supernatant was examined. CD spectra were recorded using an Aviv (Lakewood, NJ) model 202 spectropolarimeter with a temperature‐controlled cuvette holder. Three hundred microliters of each sample was added to a 0.1 cm path‐length cell (Starna), and three scans were obtained from 280 to 180 nm at intervals of 1 nm with a 2 s averaging time. The temperature of the cuvette holder was 25°C. The bandwidth was 1 nm. Data were subsequently processed by zeroing each spectrum individually by subtracting the solvent baseline, and averaging the three spectra. Ellipticities were converted to mean residue ellipticities (deg cm² dmol⁻¹).

4.5. Size exclusion chromatography

Fresh solutions of Aβ21–30, Aβ16–34, and Cys‐Aβ16–34 at concentrations from 100 to 500 μM were prepared in 10 mM phosphate buffer, pH 7.20, 0.02% (wt:vol) NaN₃ immediately prior to chromatography. To calibrate the column, molecular weight standard peptide solutions (100 μl, ~1 mg/ml) and glycine (100 μl, ~1 mg/ml) were injected onto a Superdex Peptide 10/300 column (GE Healthcare) attached to an Agilent model 1100 HPLC. Chromatography of Aβ21–30, Aβ16–34, and Cys‐Aβ16–34 was performed using a mobile phase of 10 mM sodium phosphate, pH 7.40 (also containing 0.02% NaN₃, wt:vol); flow rate was 0.5 ml/min. Column temperature was room temperature (generally ~22°C), and the effluent was monitored at by UV absorbance at 220 nm. In some experiments, as indicated in Section 2, 0.1% (vol:vol) β‐ME was included in the solvent.

4.6. Analytical ultracentrifugation

Fresh samples of Aβ21–30 and Aβ16–34 were prepared in 10 mM sodium phosphate, pH 7.20, 0.02% (wt:vol) NaN₃ to a concentration of 250 μM. A buffer blank was also prepared. In order to avoid contamination by dust or other large particles, samples centrifuged twice at 16,000g for 3 min, followed by filtration through a 0.22 μm syringe filter (Corning; Corning, NY) into a separate, autoclaved Eppendorf tube. The samples were then centrifuged at the Biophysical Core Facility at the University of Chicago, at 36,000 rpm for 72 hr, at 20°C. UV scans were measured every 2 hr at 220 and 235 nm. Equilibrium was attained in this time, as shown by the lack of change in the absorbance gradient over the previous 24 hr. Apparent molecular weights were obtained from the equation:

$$\ln \left(A\right)=\frac{M\left(1-\overline{\mathit{v\rho }}\right){\omega }^2{r}^2}{2\mathit{RT}}+C$$

where A = absorbance, M = weight‐averaged molecular weight, $\overline{v}$ = partial specific volume, ρ = solvent density (1.02 g/cm³), ω = angular velocity (rad/s), r = distance from the center of rotation (cm), R = gas constant = 8.3 × 107 erg K⁻¹ mol⁻¹, T = 293 K, and C = integration constant. Partial specific volume was determined from amino acid composition using the program SEDNTERP (for discussion, see Reference 63). For Aβ21–30 and Aβ16–34, these values were 0.698 and 0.770, respectively. In each case, three separately made samples were analyzed, and the results were aggregated to calculate M.

4.7. Peptide aggregation by sedimentation assay with SEC

Aggregation of peptides into fibrils was measured using a sedimentation assays (for example, see Reference 24). For a typical aggregation assay, 5 mg of lyophilized, HPLC purified Aβ16–34 or Cys‐Aβ16–34 was dissolved in neat DMSO to a concentration of ~1.5 mM, and divided into two replicate solutions. Peptide concentrations were calculated from absorbance at 257 nm (extinction coefficient = 400 M⁻¹ cm⁻¹, due to the two Phe residues), and in the case of Cys‐containing peptides, using the Ellman assay (see below). To start aggregation, 50 mM sodium phosphate, pH 7.40, was added to give the desired peptide concentration, typically 100 μM. Peptide was incubated at 4°C. Because the second replicate solution could not be analyzed simultaneously with the first, the second replicate solution was initially stored at 4°C, and pH~2–3 for 60 min (the length of time needed to analyze the first sample chromatographically) before adding buffer to raise the pH. The two samples, although treated slightly differently, gave essentially identical results. Immediately after adding buffer, and at various times thereafter, 100 μl aliquots were removed, centrifuged for 10 min at 15,000g; the top 50 μl was removed and injected onto a Superdex Peptide 10/300 column, as described above. The peak areas corresponding to monomeric, dimeric, and oligomeric peptide were recorded as a function of time. As described in Section 2, Aβ16–34 showed only a monomer peak, whereas some samples of Cys‐Aβ16–34 showed monomer, dimer, and oligomer peaks.

4.8. ThT fluorescence

In some instances, ThT fluorescence of precipitates was measured using standard assays,64, 65 in order to document that the precipitates were amyloids. In some ThT fluorescence experiments on Cys‐Aβ16–34, indicated in Section 2, 0.1% (vol:vol) β‐ME was included in the solvent. Peptide concentrations, solvents, and other conditions were as described in the previous section. As previously described56 (see Figure 3a), for peptides showing a lag phase in aggregation, kinetics were analyzed using a stretched exponential equation; for peptides aggregating without a lag phase, the equation of a first‐order approach to equilibrium was used.

4.9. Dissolving peptides for NMR experiments

In general, the goal in dissolving the various Aβ peptides was to obtain as complete disaggregation as possible. Accordingly, the following procedure was followed in most cases: peptide was dissolved initially in neat HFIP, and then lyophilized (the shorter Aβ peptides [Aβ21–30 and cyclo‐Aβ21–30] required no such disaggregation procedure and the same results were obtained with or without such a procedure). The lyophilized powder was then dissolved in neat DMSO (for NMR, D₆‐DMSO). In kinetics experiments, described below, aggregation was initiated by diluting this solution with 0.05 M sodium phosphate, pH 7.40 so that the final peptide concentration was 100 μM and DMSO concentration was 2% (vol:vol). Because of residual TFA from peptide purifications, the final pH was 7.1 as measured by microelectrode, and in some cases, by the position of 0.5 mM Tris and/or (4‐(2‐hydroxyethyl)‐1‐piperazineethanesulfonic acid) (HEPES) added to the sample as a pH indicator.66 Any undissolved material (generally not visible) was removed by centrifugation (16,000g). Initially, an approximate peptide concentration was determined initially by A ₂₅₇ using ε = 1,400 M⁻¹ cm⁻¹. Actual final peptide concentration was determined by Bradford assay calibrated by amino acid analysis of at least triplicate samples. For some experiments, for example, obtaining natural abundance ¹H,¹⁵N‐ and ¹H,¹³C‐HSQC spectra, peptide concentration was considerably higher, up to 3 mM for Aβ21–30 and its variants, and ~300 μM for Aβ16–34 or Cys‐Aβ16–34. ¹H,¹H‐TOCSY, ‐ROESY,67, 68 and ‐NOESY, and ¹H,¹⁵N‐ and ¹H,¹³C‐HSQC spectra were acquired at the NMR Facility of the University of Chicago, using either a 500 or 600 MHz Bruker Avance spectrometer (with Topspin 3), at temperatures from 5°C to 35°C, as indicated in Section 2. NMR assignments were obtained at 5–25°C in 5°C increments. Spectra were processed using either TopSpin or NMRPipe,69 and further analyzed using Sparky, NMRView (from One Moon Scientific, Inc.), and CARA.70 The spectra were referenced using sodium trimethylsilylpropanesulfonate (DSS)

4.10. Serial measurements of 2D homonuclear correlation spectra

Serial ¹H,¹H‐TOCSY, and in some cases, serial ¹H, ¹H‐NOESY or ¹H,¹H‐ROESY spectra were obtained for two reasons.

First, serial spectra were obtained to observe decay of peaks, associated with aggregation, at a site‐specific level. In these experiments, lyophilized peptides were dissolved as described above. Serial TOCSY spectra were obtained every ~1.5 hr for 1–5 days (the length of the experiment depended on the rate of fibrillization); serial NOESY spectra required ~3 hr per spectrum. Periodically, 1D ¹H spectra were obtained between serial TOCSY or NOESY spectra. Equations for analyzing these kinetics data are given in Section 2. The units of peak areas and volumes are referred to simply as Area or VU, and are not otherwise defined. In general, all kinetics experiments were performed on three replicate samples of peptide solution, and the results presented represent mean values obtained for the three replicates.

Second, as will be described in Section 2, no long range nOes or rOes were observed in the spectra of any of the peptides examined, including Aβ21–30, Cys‐Aβ21–30, cyclo‐Aβ21–30, Aβ16–34, and Cys‐Aβ16–34. On the possibility that such nOes or rOes appeared in spectra only transiently, however, we assessed changes in these spectra over time, by acquiring serial spectra (¹H,¹H‐TOCSY and ‐NOESY). As described in Section 2, no new peaks developed over the course of these experiments.

4.11. TALOS+ torsional angle estimates and PyMol plotting

Chemical shift data for ¹H, ¹⁵N, and ¹³C atoms were obtained from NMR peak assignments for each backbone residue and corresponding side chains for Aβ21–30, cyclo‐Aβ21–30, Aβ16–34, and Cys‐Aβ16–34. These chemical shift data were then uploaded to the TALOS+ website71, 72 (http://spin.niddk.nih.gov/bax/nmrserver/talos/) for data processing, to obtain information on backbone residue torsional angles and secondary structure probabilities, with S2 values indicating strength of prediction.73

4.12. Measurement of disulfide bond formation by thiol assays

For the Cys‐containing peptide, Cys‐Aβ16–34, the rates of disulfide bonds formation were measured under the same conditions that were used for measurements of serial NMR spectra. After making solutions of Cys‐Aβ16–34 as described above, the solution was divided among three glass vials, each containing 200 μl. To initiate formation of disulfide bonds, 20 μl of 100 mM sodium phosphate, pH 7.40, was added to each tube. Thiols were assayed using the Ellman reagent at t = 0, and again at various intervals.74, 75, 76 For the Ellman test, a stock solution of 40 mg of Ellman reagent (5,5′‐dithiobis‐(2‐nitrobenzoic acid)) was made in 10 ml of sodium phosphate, pH 8.00, and stored at 4°C. For each assay, 33 μl of this stock solution was mixed with 20 μl of sample and 922 μl of 0.1 M sodium phosphate, pH 8.50. This mixture was allowed to react at room temperature for 15 min, and then absorbance at 412 nm was measured; a buffer control was subtracted from this value at each time point. Concentration of free thiol was calculated using an extinction coefficient (for the thiolate of Ellman reagent) of 14,150 M⁻¹ cm⁻¹.

4.13. SDS‐PAGE of Cys‐Aβ16–34 after serial TOCSY spectra

To assess the proportion of monomeric and disulfide‐linked dimeric Cys‐Aβ16–34 in pellet and supernatant after serial TOCSY experiments, the peptide was first centrifuged at 16,000g and divided into supernatant and pellet, and each fraction was analyzed by Tris‐tricine SDS‐PAGE77 of Cys‐Aβ16–34. To obtain a dimer standard for SDS‐PAGE, Cys‐Aβ16–34 was oxidized using H₂O₂ by the method of Sidova et al.78 Experimental details and the results are presented in Supporting Information.

Supporting information

Supporting Figure 1A Size exclusion chromatography of Aβ21‐30 and Cys‐Aβ21‐30.

Supporting Figure 1B: Analytical ultracentrifugation (sedimentation equilibrium) of Aβ21‐30.

Supporting Figure 1C: CD spectra of Aβ21‐30.

Supporting Figure 2A: ¹H,¹³C‐HSQC spectrum of Aβ21‐30.

Supporting Figure 2B: ¹H,¹⁵N‐HSQC spectrum of Aβ21‐30.

Supporting Figure 3A: ¹H,¹³C‐HSQC spectrum of cyclo‐Aβ21‐30.

Supporting Figure 3B: ¹H,¹⁵N‐HSQC spectrum of cyclo‐Aβ21‐30.

Supporting Figure 4: ¹H,¹H‐TOCSY and ‐NOESY spectra of Aβ16‐34.

Supporting Figure 5: ¹H,¹H‐TOCSY and ‐NOESY spectra of Cys‐Aβ16‐34

Supporting Figure 6: ¹³C‐ and ¹⁵N‐HSQC of Aβ16‐34.

Supporting Figure 7: ¹³C‐ and ¹⁵N‐HSQC of Cys‐Aβ16‐34.

Supporting Figure 8: individual plots for decay of peak volumes in serial ¹H,¹H‐TOCSY spectra of Aβ16‐34.

Supporting Figure 9: Decay of amide region integrated areas of ¹H‐1D NMR spectra for Aβ16‐34.

Supporting Figure 10: SDS‐PAGE of Cys‐Aβ16‐34 after kinetics experiment.

Supporting Figure 11: Individual plots for decay of peak volumes in serial ¹H,¹H‐TOCSY spectra of Cys‐Aβ16‐34.

Supporting Figure 12: Discussion of Akaike Information Coefficient and Likelihood functions; plot of Likelihood function for A21.HN‐A21.HA (cross peak between HN and Hα of A21).

Supporting Figures 13: Rates for decay of peak volumes in serial ¹H‐TOCSY spectra of Cys‐Aβ16‐34 in the presence of TCEP.

Supporting Figure 14: example of cyclization chemistry.

Supporting Table 1: peptides used in these studies.

Supporting Table 2 A and B: chemical shift data for Aβ21‐30.

Supporting Table 3 A and B: chemical shift data for cyclo‐Aβ21‐30.

Supporting Table 4 A and B: chemical shift data for Aβ16‐34.

Supporting Table 5 A and B: chemical shift data for Cys‐Aβ16‐34

Solution for differential Equation 5.

Hawk LML, Pittman JM, Moore PC, et al. β‐amyloid model core peptides: Effects of hydrophobes and disulfides. Protein Science. 2020;29:527–541. 10.1002/pro.3778