Biophysical Characterization of Aβ42 C-terminal Fragments—Inhibitors of Aβ42 Neurotoxicity

Abstract

A key event in Alzheimer’s disease (AD) is age-dependent, brain accumulation of amyloid β-protein (Aβ) leading to Aβ self-association into neurotoxic oligomers. Previously, we showed that Aβ oligomerization and neurotoxicity could be inhibited by C-terminal fragments (CTFs) of Aβ42. Because these CTFs are highly hydrophobic, we asked if they themselves aggregated and if so, what parameters regulated their aggregation. To answer these questions, we investigated the dependence of CTF aqueous solubility, aggregation kinetics and morphology on peptide length and sequence, and the correlation between these characteristics and inhibition of Aβ42-induced toxicity. We found that CTFs up to 8-residues long were soluble at concentrations >100 µM and had a low propensity to aggregate. Longer CTFs were soluble at ~1–80 µM and most, but not all, readily formed β-sheet-rich fibrils. Comparison to Aβ40-derived CTFs showed that the C-terminal dipeptide I41-A42 strongly promoted aggregation. Aggregation propensity correlated with previously reported tendency to form β-hairpin conformation but not with inhibition of Aβ42-induced neurotoxicity. The data enhance our understanding of the physical characteristics that affect CTF activity and advance our ability to design, synthesize, and test future generations of inhibitors.

A key event in Alzheimer’s disease (AD) etiology is assembly of amyloid β-protein (Aβ) into neurotoxic oligomers (1). Aβ oligomers induce severe neuronal injury and likely are the primary neurotoxins acting in AD (2–6). Two predominant forms of Aβ comprising 40 (Aβ40) or 42 (Aβ42) amino acid residues are produced in vivo. The difference between Aβ40 and Aβ42 is the absence or presence of the C-terminal residues I41 and A42, respectively. Though this is a small structural difference, it has strong implications on the biophysical and biological behaviors of the two Aβ alloforms. Aβ42 has been shown to be more neurotoxic (7), form higher-order oligomers, and follow a different oligomerization pathway compared to Aβ40 (8, 9). These observations correlate with structural stabilization of the C-terminus of Aβ42 by the I41-A42 dipeptide (8, 10, 11). Studies using discrete molecular dynamics have suggested that a turn centered at G37–G38 rigidifies the C-terminus in Aβ42, but not Aβ40 (12, 13). In agreement with these results, using replica-exchange molecular dynamics with an all-atom protein model, Yang et al. also found that in Aβ42 the sequences spanning residues 28–37 and 39–42 are connected by a turn, which was stabilized by multiple hydrophobic interactions involving I41 and A42 (14). Multiple solution-state nuclear magnetic resonance (NMR) studies of aqueous Aβ40 and Aβ42 (15–18) support the notion that the C-terminus of Aβ42 is more rigid than that of Aβ40.

In view of the critical role of the C-terminal region of Aβ42 in self-assembly, we hypothesized that peptides derived from this region might disrupt Aβ42 self-assembly and therefore inhibit Aβ42 toxicity. To test this hypothesis, we prepared a series of C-terminal fragments (CTFs) of Aβ42, [Aβ(x–42), x = 28–39] (19) (Table 1) and evaluated their efficacy as inhibitors of Aβ42-induced neurotoxicity (20). All CTFs except Aβ(28–42) showed significant inhibition of Aβ42-induced toxicity. Of the 12 CTFs tested, Aβ(30–42), Aβ(31–42), and Aβ(39–42) were the strongest inhibitors (20). Interestingly, inhibition of toxicity did not correlate with peptide length and had a complex relationship with peptide sequence. These findings raised questions regarding the biophysical properties of CTFs themselves: how the sequence relates to the biophysical properties of each peptide, and what structural/biophysical features contribute to inhibition of Aβ42-induced toxicity. Answering these questions is critical for designing future generations of CTF-based peptidomimetic inhibitors of Aβ assembly and toxicity.

Table 1.

Biological and biophysical characteristics of Aβ42 CTFs and control peptides.

Peptide	Sequence	Cell Viabilitya (%)	Max. solubilityb (µM)	Aggregation Ratec(nm/h)	β –sheet formation T50d(h)	β-hairpine (%)	Coil-turne (%)
Aβ(39–42)	VVIA	89±5***	140±30	-	-
Aβ(38–42)	GVVIA	83±3***	156±33	-	-
Aβ(37–42)	GGVVIA	73±2***	143±27	-	-
Aβ(36–42)	VGGVVIA	80±3***	134±19	-	-
Aβ(35–42)	MVGGVVIA	82±4***	149±33	-	-
Aβ(34–42)	LMVGGVVIA	76±3***	132±29	-	-
Aβ(33–42)	GLMVGGVVIA	81±3***	134±37	1.8±0.2	25±3
Aβ(32–42)	IGLMVGGVVIA	79±2***	54±24	-	-
Aβ(31–42)	IIGLMVGGVVIA	105±5***	62±17	1.0±0.2	36±4	25	29
Aβ(30–42)	AIIGLMVGGVVIA	97±4***	11±3	8.8±3.4	6.6±0.3	42	16
Aβ(29–42)	GAIIGLMVGGVVIA	72±3**	22±9	6.6±3.3	4.1±0.1	43	4
Aβ(28–42)	KGAIIGLMVGGVVIA	-	~1	-	-
Aβ(34–40)	LMVGGVV	66±2	169±13	-	-
Aβ(30–40)	AIIGLMVGGVV	98±7***	196±1	0.3±0.1	-	4	28
Aβ(21–30)	AEDVGSNKGA	63±7	129±5	-	-

^aCell viability (mean±SEM) was calculated from at least three independent experiments (n≥18). Statistical significance was calculated compared with Aβ42 alone using t-test.

^***p<0.001;

^**p<0.01.

^bMaximal average solubility (mean±SEM) was measured by AAA in 4–7 measurements. The cell viability and solubility data of CTFs reported previously were shown here for comparison (20).

^cparticle growth rates (mean±SEM) were calculated from 3 independent measurement;

^dT₅₀ values were determined by sigmoidal regression using the function [ f (x) = y₀ + a / (1+e^x₀−x / b) ] (Fig. 2C);

^ethe percentage of CTF superclusters obtained by IM-MS and MD simulations in water at 300 K (44). “-” means no data available for this peptide. Blank cells mean data not obtained for this peptide.

Here, we characterized the aqueous solubility of the CTFs, their tendency to aggregate into fibrils and form β-sheet, and correlated these characteristics with the previously characterized inhibition of Aβ42-induced toxicity. To gain further insight into the specific interactions that affect CTF properties, in addition to the original Aβ42 CTF series, we included two Aβ40-derived CTFs, Aβ(34–40) and Aβ(30–40), to evaluate the importance of the C-terminal I41-A42 dipeptide at the C-terminus of Aβ42, and the sequence Aβ(21–30), which contains the putative folding nucleus of both Aβ40 and Aβ42 (11). The sequences of addition peptides were shown in Table 1. We present a systematic study of the aggregation properties of these peptides and discuss relationships among sequence, biophysical properties, and inhibition of Aβ42 toxicity.

MATERIALS AND METHODS

Peptide Preparation

Aβ42 was synthesized by solid-phase techniques (21) using 9-fluorenylmethoxycarbonyl chemistry, as described previously (22), purified by high-performance liquid chromatography and analyzed by mass spectrometry and amino acid analysis (AAA). Aβ(30–40), Aβ(34–40), and Aβ(21–30) were prepared using the same method. Aβ42 CTFs were prepared and characterized as described previously (19).

Toxicity Inhibition Assay

Inhibition of Aβ42-induced toxicity was performed as described previously (20). Briefly, PC-12 cells were used 48 h after differentiation. Solutions of Aβ42 and each peptide at a 1:10 concentration ratio, respectively, were incubated with the cells for 15 h. Cell viability was determined by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay using the CellTiter 96^® kit (Promega, Madison, WI). Negative controls included DMSO at the same concentration as in the peptide solutions and media alone. A positive control was 1 µM staurosporine for full kill, which was used to represent a 100% reduction in cell viability, based on which the percentage viability of all of the experimental conditions was calculated. At least three independent experiments with six replicates (n ≥ 18) were carried out, and the results were averaged and presented as mean±SEM.

Solubility

Peptide solutions or suspensions were prepared at 200-µM nominal concentration by dissolution at 2 mM in 60 mM NaOH and dilution with 10 mM sodium phosphate, pH 7.4, to the final nominal concentration. The solution was sonicated for 1 min and then filtered through an Anotop 10 syringe filter with 20-nm pore-size (Whatman, Florham Park, NJ). Parallel recordings of CD and DLS data started immediately following filtration. Four to seven replicates were measured for each peptide. The actual peptide concentrations were determined by AAA and the results were presented as mean±SEM.

Dynamic Light Scattering

Peptide solutions were measured using an in-house-built system with a He–Ne laser model 127 (wavelength 633 nm, power 60 mW, Spectra Physics Lasers, Mountain View, CA). Light scattered at 90° was collected using image-transfer optics and detected by an avalanche photodiode built into a 256-channel PD2000DLS correlator (Precision Detectors, Bellingham, MA). The size distribution of scattering particles was reconstructed from the scattered light correlation function using PrecisionDeconvolve software (Precision Detectors, Bellingham, MA) based on the regularization method by Tikhonov and Arsenin (23). The data are average of three independent experiments.

Circular Dichroism Spectroscopy

Far-UV CD spectra were recorded using a J-810 spectropolarimeter (Jasco, Easton, MD) with a thermostable sample cell (23°C) using cuvettes with 1-mm path length. Twenty measurements were collected between 190–260 nm with 1-s response time, 20-nm/min scan speed, 0.2-nm resolution, and 2-nm bandwidth, and averaged after background subtraction. Measurements were taken every 24 h for 4 d for Aβ(33–42) through Aβ(39–42), and in shorter time intervals for longer CTFs. The data are representative of three independent experiments. All CD data were converted from CD signal (millidegrees) into mean residue molar ellipticity (deg•cm²•dmol⁻¹) using the equation [θ] = θ•10⁻¹• l⁻¹•c⁻¹, where l is the path length in cm and c is the molar concentration. Secondary structure content initially was calculated using the deconvolution programs Selcon3 (24), ContinLL (25), and CDSstr (26) within the CDpro (2004) software package. ContinLL consistently generated fits with the lowest root-mean-square deviations among these three programs and therefore was chosen for deconvolution of all CD spectra.

Electron Microscopy

Eight-µL aliquots of peptide solutions were spotted on glow-discharged, carbon-coated Formvar grids (Electron Microscopy Science, Hatfield, PA). Samples of Aβ(37– 42) through Aβ(39–42) were incubated for 30 min, Aβ(32–42) through Aβ(36–42) for 15–20 min, Aβ(30–40), Aβ(34–40), and Aβ(21–30) for 10 min, and Aβ(28–42) through Aβ(31–42) for 60–90 min. The solutions were wicked gently with filter paper. The samples were then fixed with 5 µL 2.5% glutaraldehyde for 20 min and stained with 5 µL 2% uranyl acetate for 30 min (Aβ(28–42) through Aβ(31–42)) or 1% uranyl acetate for 10 min (Aβ(32–42) through Aβ(39– 42), Aβ(30–40), Aβ(34–40), and Aβ(21–30)). After cautious removal of staining solutions with filter paper, the grids were air-dried. Three to six replicates of each peptide were analyzed using a CX 100 transmission electron microscope (JEOL, Peabody, MA). The diameter and length of each peptide were analyzed using ImageJ (available at http://rsbweb.nih.gov/ij/). Ten separate measurements were averaged and the data reported as mean±SEM.

RESULTS

Inhibition of Aβ42-induced Toxicity

Previously, a cell-viability screen showed that all CTFs of Aβ42, except Aβ(28–42), which was highly toxic itself, inhibited Aβ42-induced toxicity (20). Here, we characterized the two Aβ40 CTFs and Aβ(21–30) by the same viability assay using the MTT assay (27, 28) in differentiated PC-12 cells (29, 30) with 5 µM Aβ42 and 10-fold excess of each peptide. Aβ42 alone caused a robust (≈40%) reduction in cell viability. Aβ(30–40) showed strong inhibitory effect to Aβ42-induced toxicity, similar to Aβ(39–42) and Aβ(30–42), whereas Aβ(34–40) and Aβ(21–30) were inactive. The cell viability of CTFs and control peptides are shown in Table 1.

Peptide Solubility

CTFs are highly hydrophobic peptides (19) and therefore are expected to be poorly soluble and aggregate in aqueous solutions. To estimate peptide solubility we used a simple filtration assay. Lyophilized peptides were dissolved or suspended in 10 mM sodium phosphate at 200-µM nominal concentration (see Materials and Methods), sonicated for 1 min, and filtered through a 20-nm pore-size filter (alumina-based polar membrane) to remove insoluble material. Following this treatment, the actual concentrations were determined by AAA and are shown in Table 1.

CTFs up to 10 amino acids long were found to be soluble between ~100–200 µM. Longer peptides were soluble between ~10–80 µM except for the longest CTF, Aβ(28–42), which was found to have the lowest solubility (~1 µM). Both Aβ40 CTFs had higher solubility in this assay than any of the Aβ42 CTFs. The solubility found for Aβ(21–30) (~130 µM) was surprisingly lower than expected considering the hydrophilic nature of this peptide and previous solution-state NMR studies done at millimolar concentrations in ammonium acetate, pH 6.0, at 10°C (11, 31, 32). We note that an underlying assumption in using the filtration assay was that the degree of non-specific adsorption of the peptides to the filter membrane would be independent of the sequence. This assumption likely is reasonable for the CTFs, which are all highly hydrophobic. However, the unexpectedly low concentration found for Aβ(21–30) following filtration presumably reflects strong adsorption of this highly polar peptide on the polar, alumina-based filter membrane, rather than actual solubility.

Peptide aggregation

The limited solubility of CTFs, particularly those longer than 10 residues, was expected taking into account that they originated from the hydrophobic C-terminus of Aβ42. To elucidate further the behavior of these peptides we asked whether their low solubility was a result of clumping together to form amorphous aggregates or reflected aggregation into amyloid fibrils. Amyloid fibrils are non-crystalline but have a high degree of order, reflected in a cross-β-structure, in which β-strands are arranged perpendicular to the fibril axis (33, 34). To answer this question, we studied time-dependent particle growth, conformational change, and morphology of the peptides using dynamic light scattering (DLS), circular dichroism (CD) spectroscopy, and electron microscopy (EM), respectively.

Ideally, the kinetics of aggregation and conformational transition would have been studied with all peptides at the same concentration. However, this was not feasible because the solubility of the CTFs varied over two orders of magnitude. If all CTFs had been diluted to the concentration of the least soluble peptide (<10 µM), observation of aggregation, conformational transitions, or morphological changes for the more soluble peptides would have necessitated to conduct experiments for weeks or even months. Instead, we studied the peptides at their maximal concentration following filtration through 20-nm pore-size filters as described above. In a limited number of cases, certain peptides were diluted to allow comparing aggregation rates of different peptides at similar concentrations.

To study particle size growth over time, we used DLS, a common method for studying protein aggregation non-invasively (35–37). When evaluating the data obtained here, it is important to keep in mind that because the peptides could not be studied at the same concentration, the interpretation of the data is qualitative and not quantitative. We found that most of the Aβ42 CTFs, including Aβ(32–42) (~55 µM), Aβ(35–42) through Aβ(39–42) (>100 µM), and also Aβ(34–40) (~170 µM) and Aβ(21–30) (~130 µM), did not show appreciable particle growth up to 96 h. In addition, the solubility of Aβ(28–42) was too low to allow reliable measurement and the aggregation of Aβ(34–42) (~130 µM) was slow and the spectra had low signal-to-noise ratio that precluded reliable calculation of aggregation rate. Particle growth was observed for Aβ(x–42) CTFs with x=29, 30, 31, and 33 and for Aβ(30–40). The change in hydrodynamic radius (R_H) of these CTFs over 96 h is shown in Fig. 1A and the average particle growth rate (dR_H/dt) is given in Fig. 1B and Table 1. The general trend among those four Aβ42 CTFs was faster aggregation with longer sequence, but the correlation between length and aggregation kinetics was not linear. Aβ(29–42) and Aβ(30–42) aggregated substantially faster than Aβ(31–42) and Aβ(33–42) even though they were measured at substantially lower concentrations. Aβ(30–40) aggregated more slowly than the four Aβ42 CTFs despite similar length, demonstrating the strong contribution of the I41-A42 dipeptide to promoting aggregation.

FIGURE 1.

Particle growth rate. A) Time course of average R_H was calculated from whole particle size distributions in solutions of Aβ(29–42), Aβ(30–42), Aβ(31–42), Aβ(33–42), or Aβ(30–40) at the concentrations indicated. Each data point represents mean±SEM calculated from the average R_H of eight consecutive DLS measurements during 45–60 min. Aggregation of Aβ(29–42) and Aβ(30–42) was followed until the upper limit of detection was reached. B) Average aggregation rates of Aβ(29–42), Aβ(30–42), Aβ(31–42), Aβ(33–42), and Aβ(30–40). The data represent mean±SEM of 3 independent experiments.

It is noteworthy that immediately following filtration through a 20-nm pore-size filter (t=0), particles of R_H = 40–100 nm were observed for the Aβ42 CTFs shown in Fig. 1A, indicating that some rapid self-association occurred. Nevertheless, the total intensity of scattering was low, suggesting that most of the peptide existed in a disassembled state or as small oligomers. This suggests that relatively few 40–100-nm particles formed rapidly between the end of the filtration and the beginning of DLS measurement (1–2 minutes). Such particles also may have extended rather than globular shapes, which would result in a large measured R_H.

Secondary Structure

We used CD spectroscopy to investigate the correlation between aggregation and β-sheet formation. CD spectra of each peptide were recorded over 96 h. In general, the data observed were in agreement with the DLS results. The initial spectra of all peptides showed a high proportion of unordered structure. The spectra of Aβ(32–42), Aβ(35–42) through Aβ(39–42), Aβ(34–40), and Aβ(21–30) did not change during the experiment. Fig. 2A shows characteristic CD spectra of ~150 µM Aβ(38–42) as an example. Other peptides showed a time-dependent spectral change from a minimum at 197 nm to a maximum at 198 nm with simultaneous development of a minimum at 218 nm indicating transformation from unordered conformation to a β-sheet-rich structure. Fig. 2B shows characteristic CD spectra of ~60 µM Aβ(31–42) as an example. Similar changes in the spectra, albeit at different rates, were observed for Aβ42 CTFs longer than Aβ(35–42) except Aβ(32–42) and for Aβ(30–40). In all cases, an isodichroic point was observed at 212 nm, indicating a one-step transition from unstructured to β-sheet-rich conformation.

FIGURE 2.

Time-dependent conformational change. A) Representative CD spectra of 156 µM Aβ(38–42) recorded in time intervals of 24 h. The spectra showing a minimum at 197 nm are characteristic of a statistical coil and remain unchanged for 4 days. B) Representative CD spectra of 62 µM Aβ(31–42) recorded in time intervals of 24 h. The initial spectrum showing a minimum at 197 nm is characteristic of statistical coil. The development of a maximum at 198 nm and a minimum at 218 nm indicate conformational change to β-sheet-rich structures. C) Representative time course of β-sheet formation calculated as described in Materials and Methods is shown for Aβ(29–42), Aβ(30–42), Aβ(31–42), Aβ(33–42), Aβ(34–42) and Aβ(30–40) at the concentrations indicated.

To quantify the rate of secondary structure transformation, the spectra were deconvoluted using the program ContinLL (25). Reprehensive time-course β-sheet formation were shown in Fig. 2C. The time in which half-maximal β-sheet conformation formed (T₅₀, Table 1) was calculated to facilitate quantitative comparison with solubility and DLS data. Aβ(34–42) and Aβ(30–40) showed a small increase in β-sheet content at ~170 µM and ~195 µM, respectively, during the time of measurement. Aβ(33–42) did not show conformational conversion at ~50 µM, whereas at ~130 µM it converted to β-sheet at a rate similar to that of Aβ(31–42) at ~60 µM. Aβ(29–42) at ~20 µM and Aβ(30–42) at ~10 µM converted to β-sheet within several hours (Fig. 2C). Consistent with the DLS results, Aβ(32–42) appeared to be an outlier. At ~55 µM, Aβ(32– 42) showed no conformational change up to 96 h, suggesting that in contrast to other CTFs, this CTF formed unordered, rather than fibrillar aggregates.

Morphology

To determine the morphology of peptide aggregates, aliquots of each peptide solution were examined by EM directly after dissolution (day 1) and following incubation for 7 days (Fig. 3).

FIGURE 3.

Time-dependent peptide morphology. Peptide solutions used were of the following concentrations: Aβ(28–42), 1±0.7 µM; Aβ(29–42), 14±1 µM; Aβ(30–42), 9±0.5 µM; Aβ(31–42), 22±0.6 µM; Aβ(32–42), 16±0.6 µM; Aβ(33–42), 80.0±0.1 µM; Aβ(34–42), 99±4 µM; Aβ(35–42), 122±1 µM; Aβ(30–40), 191±10 µM. Electron micrographs were recorded immediately after sample preparation (day 1) and one week later (day 7).

Electron micrographs of Aβ(35–42) through Aβ(39–42), Aβ(34–40), and Aβ(21–30), showed non-fibrillar aggregates. The morphology of Aβ(35–42) is shown as an example in Fig. 3. Aβ(34–42) and longer Aβ42 CTFs, except Aβ(32–42), were found to form fibrils, which displayed substantial morphological variability. On day 1, Aβ(34–42) formed long (>500 nm) fibrils with diameter d = 17±1 nm. After 7 days, wide, ribbon-like fibrils with diameter 26±2 nm were observed. Fibrils of Aβ(31–42) and Aβ(33–42) had a twisted, filamentous morphology. The average diameters of Aβ(31–42) and Aβ(33–42) fibrils were 5.0±0.2 nm and 10.2±0.9 nm on day 1, respectively, and they were >500 nm long. The morphology of these fibrils did not change appreciably between day 1 and day 7.

Aβ(28–42), Aβ(29–42), and Aβ(30–42) were examined at concentrations ≤10 µM. On day 1, non-fibrillar aggregates or short threads were observed whereas on day 7, the morphology was characterized by multiple short fibrils (average diameter = 8.8±0.5 nm, average length = 72±6 nm). The appearance of multiple short fibrils is in agreement with formation of multiple nuclei, consistent with fast aggregation and β-sheet formation of these peptides. In contrast, the shorter peptides, Aβ(31–42) through Aβ(34–42), except Aβ(32–42), yielded long fibrils, suggesting that for these peptides the rate of fibril elongation was substantially higher than the rate of nucleation.

Immediately after preparation, thread-like structures were observed for Aβ(32–42), whereas at day 7, the predominant morphology was non-fibrillar aggregates as predicted for this CTF based on the combination of relatively low solubility, slow aggregation, and no observation of β-sheet formation. Aβ(30–40) showed predominantly non-fibrillar aggregates on day 1, whereas on day 7, long (>500 nm) and twisted fibrils, 12±1 nm in diameter were observed.

DISSCUSION

Previously, Aβ42 CTFs were found to inhibit the neurotoxicity inflicted by full-length Aβ42 supporting the idea that peptides derived from the C-terminus of Aβ42 would disrupt the assembly of Aβ42 into toxic oligomers. Here, we expanded our initial study (20) to a systematic biophysical characterization of solubility and aggregation of all the CTFs reported previously, and included two Aβ40 CTFs and the fragment Aβ(21–30) derived from the putative Aβ folding nucleus.

As shown in Table 1, using a filtration method to estimate peptide solubility, we found that Aβ(33–42) and shorter peptides were soluble at >100 µM, whereas longer peptides had substantially lower solubility. The data indicate that CTF solubility relates roughly to peptide length and depends on the particular amino acid sequence of each peptide. The inclusion of the two Aβ40 CTFs in the current study revealed that the presence of the C-terminal dipeptide, I41-A42, confers a strong decrease in solubility (cf. Aβ(30–40) with Aβ(30–42) and Aβ(34–40) with Aβ(34–42), Table 1), supporting the idea that these two residues stabilize aggregation-prone conformations in Aβ(x–42) relative to Aβ(x–40).

Aβ(34–42) has been studied previously by solid-state NMR and found to form fibrils in which the peptide chains were in an anti-parallel arrangement (38). Here, we asked if fibril formation was a common phenomenon to all CTFs or whether the low solubility we observed for CTFs longer than Aβ(35–42) might have reflected amorphous aggregation. Peptide/protein fibrillation depends on a number of factors, including hydropathy (39), secondary structure propensity of each residue (40), the context of each residue within the sequence (41–43), and peptide length. Comparison of solubility, rates of particle growth (DLS) and conformational transition (CD), and morphology (EM) was difficult because not all peptides could be dissolved at the same concentration. One way to overcome this difficulty would have been comparing all the peptides at the maximal concentration of the least soluble CTF. However, this would have resulted in low signal-to-noise ratio in DLS and CD experiments, and likely would have required very long measurement times to observe aggregation of certain CTFs. Instead, we chose to study each CTF near its highest concentration as determined by the filtration assay described above and in particular cases, to compare certain CTFs at higher dilutions.

We found that Aβ(35–42) and shorter CTFs did not aggregate, convert to β-sheet, or form fibrils within the timeframe of measurement at concentration >100 µM. In contrast, Aβ(34–42) and longer CTFs, except Aβ(32–42), aggregated into β-sheet-rich fibrils (Table 1 and Fig. 1– Fig 3).

However, despite this simple division of the CTF series to short (4–8 residues) and long (9–15 residues), analysis of the different datasets showed that the relations among peptide length, solubility, fibrillogenesis tendency, and inhibitory activity were complex. For example, in some cases, peptides with similar lengths behaved similarly but in other cases, differed substantially. Thus, taking into account the differences in concentration, Aβ(29–42) and Aβ(30–42), or Aβ(31–42) and Aβ(33–42) had comparable aggregation rates, whereas the aggregation rates of Aβ(30–42) and Aβ(31–42) differed substantially (Fig. 1B). These data suggest that above the 8-residue cutoff, under which little or no aggregation is detected, aggregation rate and fibril formation depend on the particular amino acid sequence of each peptide and the context of each residue.

To gain a better understanding of how solubility and aggregation tendency correlate with each other and with the biological activity of the peptides, we calculated linear correlations among the different datasets, which, depending on the parameter, ranged from as little as 3 to as many as15 data points for the entire peptide series. This analysis showed that solubility alone was a poor predictor of the aggregation tendency or biological activity of peptides. The r² values calculated for the correlation of peptide solubility with rate of aggregation (dR_H/dt, DLS measurement), half-maximal time of β-sheet formation (T₅₀, CD measurement), or inhibitory activity (IC₅₀, MTT assay) were 0.42, 0.41 (not shown), and 0.04 (Fig. 4A), respectively. In contrast, as might be expected, the rates of aggregation and β-sheet formation were correlated (r²=0.86, Fig. 4B). However, neither of these parameters showed high correlation with inhibition of Aβ42-induced toxicity (IC₅₀ with dR_H/dt, r²=0.45; IC₅₀ with T₅₀, r²=0.30, not shown).

FIGURE 4.

Relationships among biophysical and biological properties. A) Linear regression analysis correlating inhibition of Aβ42-induced toxicity with CTF solubility (r² = 0.04, p = 0.52). B) Linear regression analysis correlating aggregation rates of Aβ(29–42), Aβ(30–42), Aβ(31–42), and Aβ(33–42) with T₅₀ values of β-sheet formation (r² = 0.86, p = 0.07). C) Linear regression analysis correlating solubility of Aβ(29–42), Aβ(30–42), Aβ(31–42), and Aβ(30–40) with propensity for β-hairpin conformation (r² = 0.95, p = 0.03). D) Linear regression analysis correlating aggregation rates of Aβ(29–42), Aβ(30–42), Aβ(31–42), and Aβ(30–40) with propensity for β-hairpin conformation (r² = 0.78, p = 0.11). E) Linear regression analysis correlating T₅₀ values of β-sheet formation of Aβ(29–42), Aβ(30–42), and Aβ(31–42) with propensity for β-hairpin conformation (r² = 0.99, p = 0.01). F) Linear regression analysis correlating inhibition of Aβ42-induced toxicity of Aβ(29–42), Aβ(30–42), Aβ(31–42), and Aβ(30–40) with propensity for coil-turn conformation (r² = 0.85, p = 0.08). The symbols used in panels B-F are Aβ(29–42) – Δ, Aβ(30–42) – ✻, Aβ(31–42) –♦, Aβ(33–42) – ▢, and Aβ(30–40) – ∇.

Next, we asked whether a correlation existed between any of the physical parameters measured here or the inhibition of Aβ42-induced toxicity and the three-dimensional structures of the CTFs. Recently, Wu et al. have used ion-mobility mass-spectrometry (IM-MS) combined with all-atom molecular dynamics (MD) simulations in the gas phase to examine the conformation of Aβ CTFs. In addition, the conformation of several CTFs was further calculated in the presence of explicit water molecules. Two predominant aqueous conformations were found—β-hairpin and “coil-turn” (44). Aβ(29–42) and Aβ(30–42) were reported to have a similar propensity (~40%) to form β-hairpin, whereas shortening the sequence by one N-terminal residue to Aβ(31–42) reduced the β-hairpin propensity to ~25%. Removal of the two C-terminal residues of Aβ(30– 42) resulted in a dramatic decrease of β-hairpin formation propensity from 40% to 4% in Aβ(30–40). In agreement with their lower tendency to form β-hairpin, Aβ(31–42) and Aβ(30–40) showed higher tendency to form a coil-turn structure (29% and 28%, respectively) than Aβ(30–42) (16%) and Aβ(29–42) (4%) (Table 1).

The β-hairpin propensities reported by Wu et al. showed moderate to high correlation with the CTFs’ solubility (r²= 0.95), aggregation rate dR_H/dt (r²= 0.78), and T₅₀ of β-sheet formation (r²=0.99) (Fig. 4C–Fig E). Interestingly, IC₅₀, values of inhibition of Aβ42-induced toxicity showed relatively high correlation with coil-turn conformation propensity of CTFs (r²= 0.85) (Fig. 4F). Taken together, the data suggest that of the longer CTFs studied here, those that tend to adopt a β-hairpin conformation, such as Aβ(29–42) and Aβ(30–42) have low solubility, form abundant fibrillogenesisi nuclei, and aggregate rapidly into short β-sheet-rich fibrils. In contrast, CTFs that have a relatively low tendency to form β-hairpin, such as Aβ(31–42) and Aβ(30–40), have higher solubility, form few nuclei, and aggregate at slower rates into long fibrils. The latter three peptides, Aβ(30–42), Aβ(31–42), and Aβ(30–40), are strong inhibitors of Aβ42-induced toxicity, presumably thanks to interaction with full-length Aβ42 enabled by their coil-turn conformation (Fig. 4F).

The contextual importance of position of each amino acid residue in the sequence is highlighted by comparing Aβ(32–42) with Aβ(30–40). Both these peptides comprise 11 residues and they share >80% sequence identity. Moreover, the two N-terminal residues of Aβ(30–40), A30 and I31, are identical, in reverse order, to the C-terminal I41 and A42 in Aβ(32–42). Nevertheless, Aβ(30–40) formed β-sheet-rich fibrils whereas Aβ(32–42) did not. Surprisingly, despite this behavior, Aβ(30–40) was ~4-times more soluble than Aβ(32–42). In addition, in contrast to the poor inhibitory activity and slight toxicity of Aβ(32–42) (20), Aβ(30–40) was a strong inhibitor of Aβ42 toxicity.

The longest CTF included in our study, Aβ(28–42), stood apart from the rest of the peptides. Aβ(28–42) had the lowest solubility of all CTFs and was the only sequence displaying higher toxicity than Aβ42 itself (20). One explanation for the low solubility observed is the amphipathic nature of the K side chain, which includes a long hydrophobic butylene arm and a positively charged amino group. Though K typically is considered a hydrophilic residue, hydrophobic interactions of the butylene moiety with other side-chains (45) may explain the considerable decrease in solubility of Aβ(28–42) relative to Aβ(29–42).

The high toxicity of Aβ(28–42) may be explained, at least partially, by Coulomb interactions between the positive charge of K28 in Aβ(28–42) and negatively-charged phosphate head groups of membrane phospholipids. Similar to Aβ(28–42), Aβ(25–35) has a net +1 charge and is highly toxic (46). In both Aβ(28–42) and Aβ(25–35), the net positive charge results from the ε-NH₃ ⁺ group of K28, which also has been proposed to mediate the interaction of full-length Aβ with the plasma membrane (47). Interestingly, most of the amino acid substitutions in Aβ that cause familial AD and/or cerebral amyloid angiopathy, namely, those caused by the Dutch (E22Q) (48), Arctic (E22G) (49), Iowa (D23N) (50), Tottori (D7N) (51), and English (H6R) (52) mutations, and the recently discovered deletion of glutamate 22 (E22Δ) (53, 54), cause an increase of one unit in the positive charge of Aβ. The Italian mutation (E22K) (55) causes an increase of two positive charge units. Thus, increased positive charge may be an important factor contributing to the toxic effect of amyloidogenic/hydrophobic peptides, presumably by increasing their tendency to interact with negatively charged membranes.

In summary, our study suggests that although grossly the biophysical properties of Aβ42 CTFs are length-dependent, the correlation between peptide length, peptide conformation, solubility, aggregation tendency, and inhibitory activity are complex. Peptides up to 8-residues long have relatively high solubility and low aggregation propensity. Longer peptides, excluding Aβ(32–42), have low aqueous solubility and readily form β-sheet-rich fibrils. The presence of the C-terminal dipeptide, I41-A42, increases these tendencies substantially. Aggregation rates of CTFs correlate with β-hairpin propensity and β-sheet formation. In contrast, inhibition of Aβ42-induced toxicity shows poor correlation with peptide length, solubility, aggregation rate, β-hairpin propensity in the monomer, or β-sheet formation in the fibril, but correlates with a propensity to form a coil-turn conformation. These results shed light on the parameters that modulate the biophysical properties and the inhibitory activity of the CTFs. The data provide guidelines for design of future generations of CTF-based inhibitors with improved characteristics against Aβ-induced toxicity.

Abbreviations

AAA

amino acid analysis

Aβ

amyloid β-protein

AD

Alzheimer’s disease

CD

circular dichroism spectroscopy

CTF

C-terminal fragment

DLS

dynamic light scattering

EM

electron microscopy

IM-MS

ion-mobility mass-spectrometry

MD

molecular dynamics

MTT

(3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

NMR

nuclear magnetic resonance.