Pyroglutamate-Modified Amyloid-β(3–42) Shows α-Helical Intermediates before Amyloid Formation

Abstract

Pyroglutamate-modified amyloid-β (pEAβ) has been described as a relevant Aβ species in Alzheimer’s-disease-affected brains, with pEAβ (3–42) as a dominant isoform. Aβ (1–40) and Aβ (1–42) have been well characterized under various solution conditions, including aqueous solutions containing trifluoroethanol (TFE). To characterize structural properties of pEAβ (3–42) possibly underlying its drastically increased aggregation propensity compared to Aβ (1–42), we started our studies in various TFE-water mixtures and found striking differences between the two Aβ species. Soluble pEAβ (3–42) has an increased tendency to form β-sheet-rich structures compared to Aβ (1–42), as indicated by circular dichroism spectroscopy data. Kinetic assays monitored by thioflavin-T show drastically accelerated aggregation leading to large fibrils visualized by electron microscopy of pEAβ (3–42) in contrast to Aβ (1–42). NMR spectroscopy was performed for backbone and side-chain chemical-shift assignments of monomeric pEAβ (3–42) in 40% TFE solution. Although the difference between pEAβ (3–42) and Aβ (1–42) is purely N-terminal, it has a significant impact on the chemical environment of >20% of the total amino acid residues, as revealed by their NMR chemical-shift differences. Freshly dissolved pEAβ (3–42) contains two α-helical regions connected by a flexible linker, whereas the N-terminus remains unstructured. We found that these α-helices act as a transient intermediate to β-sheet and fibril formation of pEAβ (3–42).

Introduction

Extracellular deposits belong to the pathology of Alzheimer’s disease (AD), a neurodegenerative disorder leading to progressive decline of cognitive functions (1). These extracellular plaques consist of insoluble fibrils composed of amyloid-β peptides (Aβ), which are formed from the amyloid precursor protein through cleavage by β- and γ-secretases (2, 3). Posttranslational modifications increase the diversity of Aβ species, resulting in extensive heterogeneity, as well as in N-terminal and C-terminal truncations. Pyroglutamate-modified Aβ (pEAβ) was described as a major isoform among them; up to 20% of the total Aβs start with a pyroglutamate (pE) at the N-terminus (4, 5, 6). pEAβ(3–x) was reported to be present in senile plaques in amounts equivalent to those found for Aβ(1–x) (7) and to play an important role in plaque formation and neurological deficits (7, 8). Aβ(3–x) is generated by cleavage of the first two amino acid (aa) residues (D1 and A2) from Aβ(1–x) or by alternative β-secretase cleavage leading to E3 at the N-terminus (5). The enzyme glutaminyl cyclase catalyzes intraglutamate lactam ring formation from the N-terminal amino group of E3 and its γ-carboxyl group, leading to pEAβ (9, 10). The formation of the N-terminal pE lactam ring leads to increased aggregation propensity (5, 11), and the resulting assemblies have a stronger seeding potential for other Aβ species (12, 13). Moreover, the lack of a free amino group at the N-terminus renders pEAβ more stable to degradation mediated by amino peptidases (14), and the toxicity compared to nontruncated Aβ peptides independent of their C-terminal lengths is dramatically increased (13).

Structural analysis of Aβ species by NMR spectroscopy has proven to be useful to elucidate detailed structural information on Aβ species under various solution conditions (15, 16). Here, we intend to use NMR spectroscopy to structurally characterize pEAβ (3–42) and to compare it with Aβ (1–42). Due to the severe aggregation propensity, the choice of an appropriate solvent is decisive. It has been shown that Aβ forms α-helical structures in solutions containing micelles (17), hexafluoro-2-propanol (18), or 2,2,2-trifluoroethanol (TFE) (16), facilitating structural studies.

The properties of TFE-water mixtures have been claimed to be similar to those of extracellular medium (19), and such a mixture is said to be a membrane-mimicking co-solvent (20). It is widely accepted that TFE prevents tertiary interactions due to lowered solvent polarity and favors secondary-structural elements, and that it can thus stabilize possible Aβ intermediates on pathway to amyloid formation (21, 22).

To obtain high-quality data from soluble pEAβ before aggregation and to compare structural properties of pEAβ (3–42) to Aβ (1–42) on pathway to aggregation, we started our investigation in aqueous solutions containing TFE. The results indicated that pyroglutamate-modified Aβ has a decreased helix propensity along with increased hydrophobicity and faster aggregation kinetics, even in TFE-water mixtures. We have previously shown that pEAβ (3–40) forms β-sheet-containing structures under conditions where the non-modified isoform Aβ (1–40) forms α-helices (23). Here, we extended these studies by investigating monomeric as well as oligomeric pEAβ (3–42) before aggregation.

Materials and Methods

Peptides

Expression and purification of recombinant, natural-isotope-abundant, [U-¹³C,¹⁵N]-pEAβ (3–42) and [U-¹⁵N]-Aβ (1–42) were performed as described recently (24, 25). Recombinant [U-¹³C,¹⁵N]-Aβ (1–42) was manufactured by Isoloid (Düsseldorf, Germany). Aβ (1–42) in natural isotope abundancy (NA) was chemically synthesized and purchased from Bachem (Heidelberg, Germany).

Sample preparation

All preparations were performed in Protein LowBinding tubes (Eppendorf, Hamburg, Germany). The Aβ peptides were pre-dissolved in 1,1,1,3,3,3-hexafluoro-2-propanol (Sigma-Aldrich, Hamburg, Germany) and incubated for 3 days at room temperature to assure the formation of monomers. 1,1,1,3,3,3-hexafluoro-2-propanol was then removed by lyophilization and samples were stored at −20°C.

Circular dichroism spectroscopy

Far-UV circular dichroism (CD) spectra were recorded on a Jasco J-1100 spectropolarimeter from 260 to 190 nm with a 0.5 nm step size, 50 nm/min scan speed, and 1 nm bandwidth. The peptide (25 μM) was dissolved in buffer (30, 40 or 50% TFE in 50 mM potassium phosphate, pH 2.8) and directly measured in 1 mm path-length cuvettes at 20°C. Signal to noise ratio was increased by accumulation of 10 scans per sample. Spectra for the buffer were subtracted for background correction.

Aggregation assay by thioflavin T fluorescence

Aggregation assays were performed in black non-binding 96-well plates (Sigma-Aldrich) at 20°C. Wells were prepared to a total volume of 100 μL and comprised 25 μM of initially monomeric pEAβ (3–42) and 10 μM thioflavin T (ThT) in buffer (30%, 40%, or 50% TFE in 50 mM potassium phosphate (pH 2.8)). Each reaction was performed five times and background corrected. Fluorescence was monitored using a microplate reader (PolarStar Optima, BMG, Offenburg, Germany) with excitation and emission filters at 440 and 492 nm, respectively. Wells were scanned by bottom read every 15 min, with 30 s shaking before use.

Transmission electron microscopy

Monomeric pEAβ (3–42) and Aβ (1–42) were dissolved in aqueous solutions of 40% TFE in 50 mM potassium phosphate (pH 2.8) and incubated for 5 days at 20°C without agitation. Matured fibrils were absorbed on formvar/carbon-coated copper grids (S162, Plano, Wetzlar, Germany) for 5 min. The grids were washed twice with water, and then the samples were negative stained using 1% (w/v) uranyl acetate for 1 min and finally washed twice with water. Images were recorded with a Libra 120 transmission electron microscope (TEM; Zeiss, Oberkochen, Germany) at 120 kV.

NMR spectroscopy

NMR measurements were performed on Agilent 800 MHz or Bruker 600 MHz spectrometers equipped with cryogenically cooled triple-resonance probes and pulsed z-field gradients. Lyophilized [U-¹³C,¹⁵N]-pEAβ (3–42) and [U-¹³C,¹⁵N]-Aβ (1–42) were directly dissolved in 100% TFE-d2-OH and diluted with 50 mM potassium phosphate (pH 2.8) to a final TFE concentration of 30%, 40%, or 50%. Each experiment was performed with freshly prepared Aβ samples in concentrations of 25 μM for two-dimensional or 100 μM for three-dimensional (3D) experiments in Shigemi tubes (Sigma-Aldrich) at 20°C unless otherwise specified. Two-dimensional (¹H-¹⁵N)-heteronuclear single-quantum correlation (HSQC) and (¹H-¹³C)-HSQC correlation spectra were recorded according to the standard Bruker sequences. Backbone ¹HN, ¹⁵N, ¹³C′, ¹³Cα, and side-chain ¹³Cβ, ¹Hα, and ¹Hβ resonance assignments were obtained using BEST-TROSY (BT) optimized pulse sequences (26, 27) or standard Bruker/Varian pulse sequences. Seven triple-resonance 3D correlation spectra were recorded: BT-HNCA+, BT-HNcoCA, BT-HNCO+, BT-HNcaCO, BT-HNcoCACB, HCCH-COSY (28, 29), and standard Varian (¹H-¹H-¹⁵N)-NOESY-HSQC pulse sequences. Proton and carbon shifts were referenced to the methyl groups of an internal standard of 4,4-dimethyl-4-silapentane-1-sulfonic acid, and nitrogen was referenced indirectly according to its gyromagnetic ratio (30). Spectra were processed by NMRPipe (31) and evaluated with CCPNmr Analysis (32). Assignments were deposited at the Biological Magnetic Resonance Bank with corresponding numbers 26678, 26679, and 26680.

Secondary-chemical-shift calculation and secondary-structure determination

Analysis of secondary chemical shifts is based on the publication by Zhang et al. (33) and sequence corrected (34). Secondary-structure determination was performed using the software TALOS-N (35). Structural propensities were calculated with the program SSP (36).

Results

pEAβ (3–42) is more prone to form β-sheets and high-molecular-weight fibrils

CD spectroscopy of pEAβ (3–42) and Aβ (1–42) was performed to analyze the secondary-structural elements of soluble peptides before aggregation. CD spectra of soluble pEAβ (3–42) were recorded in aqueous buffer containing different TFE concentrations. Each sample was dissolved in buffer before use and measured without further incubation. The CD spectra in 50% TFE indicated α-helical conformation, based on the characteristic minima at 208 and 222 nm, the x axis intercept at 201 nm, and the maximum at 195 nm. CD spectra of soluble pEAβ (3–42) at lower TFE concentrations, however, indicated β-sheet-dominated structures (Fig. 1 A). The CD spectrum of pEAβ (3–42) in 40% TFE showed a mixture of α-helical and β-sheet-containing secondary-structure elements. Analysis of the CD spectrum of pEAβ (3–42) in 30% TFE indicated that the conformation was now mainly β-sheet, based on the minimum at 218 nm, the x axis intercept at 212 nm, and the maximum at 200 nm. In contrast, the CD spectrum of freshly dissolved Aβ (1–42) in buffer containing 50% or 40% TFE evidenced α-helix-dominated structures, whereas the spectrum of Aβ (1–42) in 30% TFE indicated a mixed α-helical/β-sheet structure (Fig. 1 B). In general, the CD results showed that soluble non-fibrillar pEAβ (3–42) had an increased tendency to form β-sheet-containing structures when compared to Aβ (1–42) under identical solution conditions.

Figure 1.

CD spectra and aggregation kinetics of pEAβ (3–42) and Aβ (1–42). Far-UV CD spectra were recorded on a Jasco J-1100 spectropolarimeter at 20°C from 260 to 190 nm, accumulated 12 times, and corrected for the buffer. (A) CD spectra of 25 μM pEAβ (3–42) in 50 mM potassium phosphate (pH 2.8) containing 30%5, 40%, or 50% TFE, respectively (black, red, and blue lines). With decreasing TFE concentration the content of α-helices was reduced and β-sheet-rich structures appeared. (B) CD spectra of 25 μM Aβ (1–42) in 50 mM potassium phosphate (pH 2.8) containing 30%, 40%, or 50% TFE (green, purple, and orange lines), respectively. Although the spectrum of pEAβ (3–42) showed β-sheet structural elements in 50% and 40% TFE, Aβ (1–42) is primarily α-helix rich but forms β-sheet structures in 30% TFE comparable to that of pEAβ (3–42) at higher TFE concentrations, i.e., 40%. (C) Aggregation kinetics measured by ThT assay, where 25 μM pEAβ (3–42) was dissolved in 50 mM potassium phosphate (pH 2.8) containing 30%, 40%, or 50% TFE (black, red, and blue lines, respectively) including 10 μM ThT. (D) Kinetics of 25 μM Aβ (1–42) were measured in 50 mM potassium phosphate (pH 2.8) containing 30%, 40%, and 50% TFE (green, purple, and orange lines) including 10 μM ThT (pH 2.8). The fluorescence emission at 492 nm was monitored on a fluorescence spectrometer (excitation at 440 nm), recorded every 15 min for 62 h. The measurement was performed fivefold and averaged, and the buffer was subtracted for background correction. An increase of ThT fluorescence and faster aggregation kinetics of pEAβ (3–42) were observed with decreasing TFE concentration. No fibril formation of Aβ (1–42) could be detected within 62 h. To see this figure in color, go online.

Aβ fibril formation was investigated from a kinetic point of view by ThT assay. ThT is used to monitor and quantify the presence of fibrils by binding to cross-β-sheet-rich structures present in amyloids. Aggregation kinetics of pEAβ (3–42) showed fibril formation and thus the presence of β-sheet structures in aqueous TFE solutions when compared to Aβ (1–42) under the same conditions. Fig. 1 C shows the aggregation kinetics in 30%, 40%, and 50% TFE at 20°C. The ThT fluorescence intensity of pEAβ (3–42) in 50% TFE did not increase significantly within 60 h. The aggregation kinetics in 40% TFE showed a distinct lag phase of ∼11 h followed by a growth phase and the characteristic stationary phase up to 20 h. Fibril formation in 30% TFE occurred faster; the lag phase lasted 6 h and fluorescence reached its maximum after 11 h. Moreover, the total fluorescence intensity of pEAβ (3–42) incubated in 30% TFE was higher than in 40% TFE, indicating a higher percentage of β-sheeted fibrils. In contrast, fibril formation of Aβ (1–42) was not observed under the same conditions, as no increase in fluorescence intensity was observed in either 50%, or 40% TFE, nor in 30% TFE within 60 h (Fig. 1 D). Thus, within 60 h, Aβ (1–42) did not form fibrils under these solvent conditions.

The morphology of matured aggregates in thermodynamic equilibrium was analyzed with negative-stained electron microscopy. For that, freshly dissolved pEAβ (3–42) and Aβ (1–42) were incubated for 5 days in 40% TFE for aggregation and visualized by TEM. As seen in Fig. 2 (upper) pEAβ (3–42) formed long fibrils in micrometer scale with an average width of 9.7 ± 1.5 nm. The fibrils were twisted, exhibited homogeneous morphology and accumulated into clumps. In contrast, Aβ (1–42) showed a completely different conformation (Fig. 2, lower). The structure of insoluble Aβ (1–42) was amorphous, without any characteristic fibril morphology, and branched into networks of nonfibrillary particles.

Figure 2.

TEM images of pEAβ (3–42) and Aβ (1–42). pEAβ (3–42) (upper) and Aβ (1–42) (lower) peptides were dissolved in 50 mM potassium phosphate (pH 2.8) containing 40% TFE and incubated at 20°C for 5 days. Grids were prepared by negative staining. Aggregated pEAβ (3–42) formed large twisted fibrils, whereas Aβ (1–42) formed non-fibrillary branched networks. The average fibril length of pEAβ (3–42) was 9.7 ± 1.5 nm. To see this figure in color, go online.

Characteristic chemical-shift differences of pEAβ (3–42) compared to Aβ (1–42)

(¹H-¹⁵N)-HSQC spectra of freshly dissolved pEAβ (3–42) in different TFE concentrations ranging from 30% to 50% are shown in Fig. 3 A. Resonance assignments were performed using various 3D experiments, as described in Materials and Methods. The spectra of pEAβ (3–42) recorded in buffer containing 50% TFE exhibits all amide crosspeak signals in a single conformation and similar intensities. The results of pEAβ (3–42) in 40% TFE are similar, with minor peak shifts likely due to the reduced TFE concentration. Nonetheless, some peak intensities were decreased, for example, H13, E22, M35, and G37. This effect was dramatically increased in 30% TFE. There, only the N-terminal crosspeaks of amino acid (aa) residues pE3, F4, and R5 were detectable, whereas all other resonances were missing.

Figure 3.

(¹H-¹⁵N)-HSQC of pEAβ (3–42) and Aβ (1–42). (A) pEAβ (3–42) (100 μM) was dissolved in aqueous solution containing 30%, 40%, or 50% TFE in 50 mM potassium phosphate (pH 2.8) (black, red, and blue lines, respectively). Spectra were recorded on a 600 MHz Bruker spectrometer at 20°C. The spectra of pEAβ (3–42) in 50% TFE differs slightly from that in 40% TFE due to the change in TFE concentration. Notably, some peak intensities, such as those of H13, E22, M35, and G37, are drastically decreased. The spectrum of pEAβ (3–42) in 30% TFE revealed that only the crosspeaks of the very N-terminal aa residues pE3, F4, and R5 were left. (B) Overlay of 100 μM pEAβ (3–42) (red) and Aβ (1–42) (purple) in aqueous solution (50 mM potassium phosphate (pH 2.8)) containing 40% TFE. Peaks with prominent changes in chemical shifts are marked with assignments. Spectra were recorded on a 600 MHz Bruker spectrometer at 20°C. Pyroglutamate modification affects the N-terminal crosspeaks toward H14 (assigned in black). Crosspeaks exclusively for Aβ (1–42) were labeled in purple and in red for pEAβ (3–42). (C) Changes in chemical shifts of pEAβ (3–42) compared to those of Aβ (1–42) were obtained from (¹H-¹⁵N)-HSQC of pEAβ (3–42) and Aβ (1–42) in 40% TFE and calculated according to the formula Δδ(¹H,¹⁵N) = ((Δδ1H^N)² + (1/5^∗Δδ¹⁵N)²)^1/2. Chemical-shift changes are plotted as a function of the primary sequence of pEAβ (3–42). Pyroglutamate formation affects the N-terminal aa residues up to G9 with decreasing effect toward the C-terminus. To see this figure in color, go online.

(¹H-¹⁵N)-HSQC spectra of soluble pEAβ (3–42) and Aβ (1–42) recorded in 40% TFE are compared in Fig. 3 B. Some chemical-shift values differ significantly, although the only chemical difference between the pEAβ (3–42) and Aβ (1–42) is the N-terminus. Chemical-shift differences were plotted as a function of the primary sequence of pEAβ (3–42) (Fig. 3 C). Differences in chemical shifts decreased rapidly from the N-terminus toward the C–terminus, affecting significantly almost 22% of the total aa residues.

Chemical-shift-based secondary-structure analysis

Secondary chemical shifts are the deviation of the measured NMR chemical shifts from known random-coil chemical-shift values and are used to locate and determine secondary-structure elements. Secondary chemical shifts for Cα and C′ in pEAβ (3–42) are shown in Fig. 4 A. Values >1.5 ppm indicate segments of α-helical structures, whereas negative scores are indicative of β-strands or extended conformations. C′ and Cα secondary chemical shifts indicated two α-helical regions from Y10 to D23 and A30 to V36, respectively. This is in accordance with the results for Cβ and Hα secondary chemical shifts, where positive values indicate β-sheets and negative values α-helices (Fig. 4 A). There is no indication for β-sheet structures in the observable, and thus soluble, Aβ moiety.

Figure 4.

Structural characterization of pEAβ (3–42). Structural analysis was based on NMR chemical shift data for 100 μM pEAβ (3–42) in buffer (40% TFE in 50 mM potassium phosphate (pH 2.8)) at 20°C. (A) C′, Cα, Cβ, and Hα chemical shifts calculated as the difference between measured and random-coil chemical shifts, described by Zhang et al. (33), and sequence corrected for neighbor effects (34). (B) SSP score. (C) Secondary-structure determination by TALOS-N (35). The probability for an α-helical conformation is plotted for each aa residue. To see this figure in color, go online.

Additionally, secondary-structural propensities were also calculated using the SSP program (36) with Hα, Cα, Cβ, and C′ chemical shifts as input data (Fig. 4 B). Positive score values are indicative of α-helices and negative scores for β-strands or extended conformation.

As a third approach, secondary structure was determined with TALOS-N using all available chemical shifts (HN, Hα, Hβ, N, Cα, Cβ, and C′) (35), and the result is shown in Fig. 4 C. There were no β-sheet structures, but two α-helical regions were identified with a certainty >80%. The results are in agreement with both previously performed methods, indicating strong α-helices in the regions Y10 to D23 and A30 to V36, with decreased helical propensity in between. Both termini showed negative score values.

NMR chemical-shift analysis by three methods did not provide any evidence for β-sheet structures of monomeric pEAβ (3–42) in 40% TFE. Although the secondary structure of soluble Aβ (1–42) determined by CD spectroscopy differed from that of soluble pEAβ (3–42) under identical conditions, NMR spectroscopy showed the same structural elements for both monomeric isoforms. Aβ (1–42) in 40% TFE showed also soluble monomers consisting of two α-helical regions from Y10 to V24 and A30 to V36, as determined from secondary chemical shifts and TALOS-N (Fig. S1).

Interestingly, soluble and monomeric pEAβ (3–42) in 50% TFE showed the same secondary-structure analysis based on secondary chemical shifts and TALOS-N results: two helices connected by a flexible linker, in accordance with CD data indicating the presence of α-helices and the absence of fibril formation (Fig. S2).

Aggregation-dependent crosspeak attenuation in 3D spectra

Crosspeak intensity changes of Aβ (1–42) compared to pEAβ (3–42) in aqueous 40% TFE-containing buffer, as well as between pEAβ (3–42) in 40% and pEAβ (3–42) in 50% TFE-containing buffer, were analyzed. Therefore, the peak heights observed in 3D BT-HNCA+ and BT-HNCO were plotted as a function of the aa sequence (Fig. 5, A and B). The crosspeak intensities gained from HNCA experiments (recovery delay of 0.3 s) on soluble pEAβ (3–42) were decreased by an average of 15% to 20% compared to the corresponding intensities for monomeric Aβ (1–42) under the same conditions. Especially weak crosspeak intensities, for example, E22 or M35, were almost doubled in the spectra of Aβ (1–42) (Fig. 5 A). For pEAβ (3–42) in 40% TFE compared to that in 50% TFE, crosspeak intensities between V12 and E22, as well as between I31 and G37, were reduced by ∼50%. M35, for example, was affected even more strongly (Fig. 5 A). Nonetheless, one has to mention that viscosity differences present in solutions with varying TFE concentrations may have an influence on crosspeak intensities. The N-terminus and residues G25 to A30 were less affected. Changes in crosspeak intensities gained from HNCO (recovery delay of 0.2 s) were nearly identical: the peak height of pEAβ (3–42) was 15–20% decreased compared to Aβ (1–42) under the same conditions. Comparing the peak intensity of pEAβ (3–42) in 40% TFE to the corresponding intensity in 50% TFE indicated significant signal reduction. All aa residues except for the very N-terminal residues pE3 and F4 are reduced by at least half (Fig. 5 B).

Figure 5.

Normalized crosspeak intensities of pEAβ (3–42) compared to Aβ (1–42). Crosspeak intensities of pEAβ (3–42) and Aβ (1–42) obtained from BT-HNCA+ (A) and BT-HNCO (B) spectra. The peak intensity of pEAβ (3–42) in 40% TFE was normalized to pEAβ (3–42) in 50% TFE (red) and to Aβ (1–42) in 40% TFE (black). The normalized crosspeak intensity is plotted against the primary aa sequence of pEAβ (3–42). Stars indicate overlapping crosspeaks and therefore no analysis of peak height. The Cα and C′ intensities of pEAβ (3–42) are decreased by an average of 15–20% compared to those of Aβ (1–42). Cα and C′ intensities of pEAβ (3–42) in 40% TFE were only half those of pEAβ (3–42) in 50% TFE, but this effect was smaller at the N-terminus and residues G25–A30. To see this figure in color, go online.

Decrease in monomer population over time

The crosspeak intensities of soluble pEAβ (3–42) and Aβ (1–42) were plotted against the incubation time and represent the monomer population in non-steady state and before aggregation (Fig. 6 A). The crosspeak heights of (¹H-¹⁵N)-HSQC spectra were averaged over all residues in each peptide and normalized to their initial values at the first measurement. The monomer population of pEAβ (3–42) decreased drastically and more rapidly compared with that of Aβ (1–42). The crosspeak intensities of monomeric pEAβ (3–42) after 50 h were reduced to ∼10% relative to initial values, whereas monomer-derived crosspeak intensities for Aβ (1–42) were still at 50% of initial values.

Figure 6.

Crosspeak intensities over time. (A) The (¹H-¹⁵N)-HSQC crosspeak intensities were averaged over all residues in Aβ (1–42) (black) and pEAβ (3–42) (red) and reported relative to their initial value at the first measurement. The monomer abundance in pEAβ (3–42) decreased more rapidly than that in Aβ (1–42). (B) Crosspeak intensities over time for 100 μM pEAβ (3–42) (40% TFE in 50 mM potassium phosphate (pH 2.8)) at 20°C. A time series of 48 (¹H-¹⁵N)-HSQCs (each 90 min) was recorded on an 800 MHz Agilent spectrometer. The total acquisition time was ∼72 h. The amide crosspeak intensity for selected aa residues is plotted against time. Within 15 h, most of the crosspeaks were near the background, but the N-terminal aa residues pE3, F4, and R5 (dark blue) were more stable. Amide crosspeak intensities for all aa residues are shown in Fig. S3. To see this figure in color, go online.

This result was supported by analyzing the stability of soluble pEAβ (3–42) in 40% TFE by a time-dependent series of several (¹H-¹⁵N)-HSQC spectra over 3 days and evaluating all residue-specific crosspeak intensities individually. The amide crosspeak intensities were plotted against time and residue sequence position (Figs. 6 B and S3). All crosspeak intensities decreased at almost the same rate except for the very N-terminal residues pE3, F4, and R5. The loss of crosspeak intensity might result from conversion of the peptide monomers to NMR-invisible assemblies whose peaks are broadened beyond detection because of their large size or conformational exchange.

To elucidate whether the chemical exchange results from intra- or intermolecular conformational change of the peptide, (¹H-¹⁵N)-HSQC spectra of monomeric [U-¹⁵N]-pEAβ (3–42) were measured in the presence and absence of an excess amount of [NA]-pEAβ (3–42). The 10-fold molar excess of the peptide in NA reduced the peak intensity to ∼20%. The reduction in peak intensity indicates that the observed chemical exchange over time cannot be (exclusively) caused by intramolecular interactions, but is rather due to intermolecular association, which is concentration dependent (Fig. 7). Thus, the loss in signal intensity over time, as well as at higher protein concentrations, is primarily due to kinetic events of aggregation in a non-steady state and to a lesser extent based on the formation of a monomeric subspecies.

Figure 7.

Normalized crosspeak intensities in the presence of a 10-fold molar excess of non-isotopically enriched peptide. (¹H-¹⁵N)-HSQC peak height of 25 μM [U-¹⁵N]-pEAβ (3–42) with an excess of 250 μM [U-¹⁴N]-pEAβ (3–42) was normalized to 25 μM [U-¹⁵N]-pEAβ (3–42) in 40% TFE. The decrease in crosspeak intensity is plotted against the primary aa sequence of pEAβ (3–42). On average, the peak densities are decreased to ∼20% of the original intensities.

Discussion

TFE is broadly known as a co-solvent that artificially stabilizes secondary-structure elements that may be relevant as possible conformational intermediates (21). It is described as disrupting tertiary interactions due to lowered solvent polarity and favors intramolecular hydrogen bonds, which supports secondary-structure formation in proteins and peptides (22). Molecular dynamics simulation by Roccatano et al. (37) confirmed that the secondary-structure-stabilizing effect of TFE is based on the displacement of water and a decreased dielectric constant favoring intramolecular interactions.

β-sheet and amyloid fibril formation

pEAβ is more prone to aggregate and has an increased tendency to form β-sheet structures in aqueous solutions compared to nontruncated Aβ (5, 13). The conversion from α-helices to β-sheets analyzed by lowering the TFE concentration was described in detail for Aβ (1–40) (38). Chen et al. (38) showed that Aβ (1–40) undergoes a three-state transition from an α-helical conformational state to β-sheets even in TFE concentrations ≤25%. Amyloid formation in the presence of TFE was also described for other unstructured peptides. Khan et al. (39) demonstrated that conalbumin shows the fastest aggregation at a TFE concentration of 15%, although TFE stabilizes α-helices in conalbumin at concentrations >50%. For the islet amyloid polypeptide, it was observed that low concentrations of TFE in combination with heparin induce aggregation, as also occurs in the presence of TFE alone when the concentration is at least 15% (40). Formation of amyloids in the presence of TFE was also shown for, e.g., human carbonic anhydrase II (41), the FF domain of URN1 splicing factor (42), transferrin (43), and α-synuclein (44). Here, our data show that the amyloid-inducing effect of TFE can also be observed for pEAβ (3–42). TFE stabilized secondary-structural elements on pathway to amyloid formation and allowed us to characterize pEAβ species before amyloid formation.

Secondary-structure analysis of freshly dissolved pEAβ (3–42), which is not yet in thermodynamic equilibrium, by CD data suggests an increasing content of β-sheets for pEAβ (3–42) in aqueous solutions with decreasing TFE concentrations. The data indicate a high content of α-helices in 50% TFE. Helix content was reduced and β-sheet content increased with lower TFE concentrations. These data were confirmed by analyzing the aggregation kinetics by ThT assay. No increase in fluorescence intensity, and thus no fibril or β-sheet formation, was detected under conditions where the CD spectra of freshly dissolved peptide indicated only α-helices in the non-steady state. pEAβ (3–42) started to form ThT-positive structures at concentrations where the first β-sheets in soluble peptide were observed by CD spectroscopy. Aggregation kinetics, and thus fibril formation, was accelerated in decreased TFE concentrations, indicating that the observed species is not thermodynamically stable but reacts further toward aggregation and is thus on pathway to amyloid formation—at least under the assayed conditions. TFE has been described as stabilizing intramolecular hydrogen bonds, and thus supports the formation of α-helices and β-hairpins (45). Here, TFE at high concentrations seems to stabilize helical pEAβ (3–42) monomers by destabilizing hydrophobic interactions and thus delaying or even preventing aggregation and fibril formation (46). However, lower TFE concentrations obviously cannot prevent pEAβ (3–42) from forming β-strands and fibrils. Thus, the choice of solvent conditions is critical to the investigation of intermediate structures that are thermodynamically necessary for amyloid formation to occur. By using TFE-water mixtures, we have found solution conditions where amyloid formation of pEAβ (3–42) is observed but is dramatically decelerated, allowing structural studies of possible metastable intermediates between unstructured pEAβ (3–42) monomers and fibrils.

CD data for freshly dissolved Aβ (1–42) in the non-steady state showed α-helical conformation in 40% TFE, according to a previously described simulation study (47), and β-sheet structures in 30% TFE. Interestingly, Aβ (1–42) fibril formation could not be detected either in 40% or in 30% TFE, as shown by ThT assay, underlining that aggregation kinetics are prevented. Under conditions where the thermodynamic equilibrium is not reached, helical Aβ (1–42) is more stabilized by TFE than pEAβ (3–42), as pEAβ (3–42) shows helix-to-sheet conversion at lower TFE concentrations. This is consistent with the behavior of the C-terminally shortened isoforms pEAβ (3–40) and Aβ (1–40) (23).

Incubation of pEAβ (3–42) and Aβ (1–42) under identical conditions leads to aggregates that differ in size and morphology as observed by TEM. Aggregates of pEAβ (3–42) in TFE solution conditions are β-sheet rich and thus represent a type of amyloid fibrils, whereas aggregates of Aβ (1–42) formed under the same conditions have an irregular amorphous ultrastructure.

Fibrillization of pEAβ (3–42) is preceded by the formation of an α-helical intermediate

Dependent on the TFE concentration, freshly dissolved pEAβ (3–42) showed β-strand-containing secondary structure, as observed by CD spectroscopy. Aggregation kinetics indicated formation of amyloid fibrils. NMR chemical-shift data analysis allowed us to determine the secondary structure of the pEAβ (3–42) monomer on a residue-specific level before aggregation starts. Interestingly, the NMR-visible species of freshly dissolved pEAβ (3–42) indicated the presence of only α-helical monomers; no β-strand-containing monomeric species could be observed. Analysis of secondary chemical shifts determined two helical regions from Y10 to D23 and A30 to V36 connected with a linker (33, 34, 36). Additionally, lowering the TFE concentration did not change the observed chemical shifts of pEAβ (3–42) monomers toward β-strand structures, but the crosspeak intensities were drastically decreased. The β-sheet-rich structures detected via CD spectroscopy are thus not observable by solution-state NMR, due to exchange processes or to the high molecular weight of the assembly state. However, aggregation kinetics monitored by ThT assay and analysis of fibrils by TEM under the same conditions indicated that at thermodynamic equilibrium, fibrils are present. Thus, the NMR-visible α-helical species, which is detected in non-steady state before aggregation, has to be an intermediate in aggregation kinetics of pEAβ (3–42). If the observed helical monomer is not a direct intermediate but rather the precursor of a new/second monomeric species on pathway to fibril formation, then chemical exchange would occur within a molecule, and thus, the crosspeak intensities would not depend on the overall peptide concentration. To check this, we added a 10-fold molar excess of pEAβ (3–42) in NA to [U-¹⁵N]-pEAβ (3–42). Interestingly, this led to decreased crosspeak intensity, actually a decrease to 20% compared to the initial signal, indicating intermolecular interactions of these species. The loss of signal is not based on intramolecular interactions, which would be indicative of formation of a second monomeric species. Thus, the concentration of α-helical monomers was reduced due to peptide-peptide interactions, resulting in nonobservable assemblies.

Our results show that under the assayed conditions, pEAβ (3–42) undergoes amyloid fibril formation based on the observed α-helical intermediate. Teplow and co-workers (48) reported that helical intermediates could play a central role in aggregation processes. They showed, by the use of a model peptide, that partially folded intermediates containing no observable β-sheet structures can mediate the initial stages of amyloid fibril formation via helix-rich oligomeric intermediates. They also found that 18 different Aβ species, among others pEAβ (3–42), incubated in glycine buffer showed a transient increase in helicity before the appearance of β-sheet structures (49). The same group showed that TFE stabilizes this partially folded helical conformer, which is a possible intermediate in Aβ fibril assembly (46). Helical intermediates of amyloidogenic peptides were also observed by Anderson et al. (44), who showed that α-synuclein forms fibrils that are strongly correlated with the formation of a monomeric, partly helical intermediate. This intermediate exists in equilibrium with the natively disordered state at low TFE concentrations, whereas high TFE concentrations shift the equilibrium to the α-helical conformation (44). These helical intermediates are prone to fibril formation and have been reported for a number of amyloidogenic peptides by Abedini and Raleigh (50). The authors hypothesized that transiently monomeric helical structures associate, leading to a high local concentration of an aggregation-prone sequence that promotes intermolecular β-sheet formation (50, 51). α-Helical structures of Aβ are also found in the presence of GM1 ganglioside micelles (52, 53, 54). Ikeda et al. (54) showed that with an increasing GM1 concentration, Aβ shows an α-helical form that converts to a β-sheet form that can further convert to a seed-prone β-structure. Neither changes in chemical shifts nor intramolecular interactions within aggregation kinetics could be observed via NMR spectroscopy. Since monomers need to pass through a kinetic barrier, and the presence of an unfolded monomeric pEAβ species could be excluded, it seems deductive that the helical monomeric species is on pathway to fibril formation. This helical monomeric species is stabilized in aqueous TFE solution and represents an intermediate state that occurs before aggregation. By using TFE as a co-solvent, the dynamics of this intermediate to amyloid fibril formation are drastically delayed or even prevented.

Nevertheless, it is not necessary that helical structures lead to an increase in the rate of amyloid formation. Stable helical structures can decrease or actually prevent the rate of amyloid formation (50, 51). Fibril formation of both Aβ (1–42) and pEAβ (3–42), however, is inhibited in TFE concentrations ≥50%, most probably due to the destabilization of hydrophobic interactions between those α-helical monomers. Thus, very high TFE concentrations impose a kinetic barrier to the α-to-β transition (45).

The very N-terminus of Aβ affects its fibrillization kinetics

By comparing the (¹H-¹⁵N)-HSQC spectra of pEAβ (3–42) and Aβ (1–42), it can be seen that the chemical difference between these two Aβ species at the very N-terminal end has a significant effect on the chemical shifts of the residues up to G9 and then, with decreasing effects, further toward the C-terminus. Mischo et al. (55) performed a study regarding the effect of pE formation on unstructured model peptides. Comparison of the (¹H-¹⁵N)-HSQC spectra of these peptides starting with the N-terminal E before and after conversion to pE showed that this modification alters the chemical shifts of only the next two aa residues. In contrast, the difference between Aβ (1–42) and pEAβ (3–42) not only affects the neighboring aa residues but alters the conformational state of the unstructured N-terminus, leading to a decreased monomer stability of pEAβ (3–42). Monomeric pEAβ (3–42) NMR spectroscopy data showed a more rapid decrease in monomer abundance compared to Aβ (1–42). Aβ (1–42) is less prone to aggregation due to its stabilized α-helical secondary structure. On the other hand, the conversion of monomeric pEAβ (3–42) α-helices to oligomeric intermediates results in amyloid fibril formation, as the transient α-helices are not stabilized enough and act as precursors to aggregation. The secondary-structure determinations based on chemical shifts are very similar for Aβ (1–42) and pEAβ (3–42), as two helices are indicated for both peptides. However, the chemical shifts of the unstructured N-terminal region up to G9 are drastically different. It was previously reported that the soluble N-terminus is not in the rigid fibril core (56, 57), but plays a central role in enhancing β-sheet formation and stabilizing amyloid fibrils (58, 59, 60), especially at low pH (61). The N-terminal region might support the aggregation-prone sequence stabilizing β-sheet formation as described by Abedini and Raleigh (50, 51). Similarly, it was shown, using Fourier transform infrared (FTIR) spectroscopy, that Aβ forms a transient α-helical structure at the N-terminal region and that the helix stability is modulated by the ionic strength and the N-terminal modification (62).

All these data suggest that the observed α-helical conformation of pEAβ (3–42) is a transient intermediate on pathway to fibril formation. According to this hypothesis, this pathway could be consisting of 1) the conversion of unstructured pEAβ (3–42) to α-helices; 2) the formation of α-helical assemblies, leading to a high local concentration of an aggregation-prone sequence; 3) the conversion of this aggregation-prone sequence to β-strands; and 4) the formation of amyloid fibrils (Fig. 8). This sort of self-assembly based on transient α-helices has been demonstrated also for model peptides (48, 63), islet amyloid polypeptide (64), and α-synuclein (44), and also in vivo for the assembly of silk into its cross-β structure (65).

Figure 8.

Schematic diagram of the transition of an α-helical intermediate to β-sheets. α-helices are depicted as cylinders and β-strands as zigzag lines. Unstructured pEAβ (3–42) monomers form TFE induced α-helices. These helices bundle together and generate a high local concentration of an aggregation-prone sequence that is likely to form β-strands. The formation of β-strands disrupts the α-helices and leads to the formation of β-sheet-rich assemblies.

The formation of Aβ oligomers is kinetically preferred unless this species is unstable at high concentrations. These oligomers then allow the fibril-formation-prone residues to overcome the kinetic barrier to the formation of β-sheet-rich amyloid fibrils (66). Interestingly, Aβ was shown to be in a monomer-oligomer equilibrium in concentrations up to 3 μM (67), which is much higher than the estimated concentration in the cerebrospinal fluid of either normal or diseased brains (68, 69). This indicates that the energy level of nucleation is high and that a significant proportion of Aβ might be monomeric in vivo.

The precise structures of Aβ in vivo are still unknown. Based on the aa sequence it is likely that the C-terminal domain of Aβ adopts a β-strand conformation and the N-terminal domain is in equilibrium between an α-helix and β-strand (70). Thus, modification of the very N-terminus may results in an altered secondary structure favoring aggregation. Since Aβ is partially composed of the amyloid precursor protein transmembrane region, it might adopt largely helical conformation in vivo and transition from α-helices to β-sheets would promote amyloid formation (48). On the other hand, if the peptide is mainly unstructured in solution comparable to in vitro data, then the formation of a partially helical intermediate would be on pathway to amyloid formation (48). There is evidence to the conclusion that pEAβ (3–42) forms transient intermediates with α-helical secondary structure as a precursor to β-sheet and fibril formation. However, as shown in this study, TFE as a co-solvent is useful to support the stabilization of α-helical intermediates which may also play a central role in extracellular in vivo amyloid formation of pEAβ (3–42).

Conclusions

The aggregation propensity and structural properties of pEAβ (3–42) were analyzed under different solvent conditions and compared to those of Aβ (1–42). In addition, pEAβ (3–42) was further studied in secondary-structure-stabilizing solution conditions to characterize possible intermediates on pathway to amyloid fibril formation. The pE-modified peptide forms β-sheet structures, starts to aggregate, and forms fibrils even in 40% TFE, as visualized by CD, ThT assay, and TEM. In contrast, Aβ (1–42) keeps its α-helical conformation under the same solvent conditions. High-resolution NMR spectroscopy of pEAβ (3–42) in a solution state was performed to obtain complete backbone and partial side-chain chemical-shift assignments of pEAβ (3–42) under these conditions. Analysis of the secondary structures of pEAβ (3–42) and Aβ (1–42) in aqueous 40% TFE indicated that both form dominantly α-helical structures in two regions connected by a flexible and disordered linker. However, in pEAβ (3–42), this conformation is only transient and thermodynamically metastable and converts further into fibrils via solution-NMR-invisible β-sheet-rich assemblies.

Author Contributions

M. Schwarten, C.D., L.G., and D.W. conceived and designed the experiments. C.D., M. Schwarten., K.R., M. Stoldt, and T.Z. carried out experiments. C.D., M. Schwarten, J.L., and D.W. analyzed the data. C.D., M. Schwarten., L.G., and D.W. wrote the manuscript.

Editor: Jeff Peng.