Size-dependent neurotoxicity of β-amyloid oligomers

Abstract

The link between the size of soluble amyloid β (Aβ) oligomers and their toxicity to rat cerebellar granule cells (CGC) was investigated. Variation in conditions during in vitro oligomerization of Aβ_1-42 resulted in peptide assemblies with different particle size as measured by atomic force microscopy and confirmed by the dynamic light scattering and fluorescence correlation spectroscopy. Small oligomers of Aβ_1-42 with a mean particle z-height of 1-2 nm exhibited propensity to bind to the phospholipid vesicles and they were the most toxic species that induced rapid neuronal necrosis at submicromolar concentrations whereas the bigger aggregates (z-height above 4-5 nm) did not bind vesicles and did not cause detectable neuronal death. Similar neurotoxic pattern was also observed in primary cultures of cortex neurons whereas Aβ_1–42 oligomers, monomers and fibrils were non-toxic to glial cells in CGC cultures or macrophage J774 cells. However, both oligomeric forms of Aβ_1-42 induced reduction of neuronal cell densities in the CGC cultures.

1. Introduction

The central event in pathogenesis of Alzheimer’s disease (AD) is thought to be intracellular and extracellular accumulation of polypeptide compounds of low molecular mass – so called beta amyloid (Aβ). These molecules tend to aggregate and form complexes of varying size: from small soluble oligomers, bigger protofibrils and finally insoluble fibrils. It is commonly assumed that formation of Aβ fibrils and plaque deposits is a crucial event in the pathogenesis of AD [1]. However, there is accumulating evidence that soluble oligomers are the most cytotoxic form of Aβ although it is still unclear which size and morphology of the aggregates exert neurotoxicity. The level of soluble forms of Aβ was shown to strongly correlate with the severity of the disease [2, 3]. In addition, it has been shown that soluble oligomeric Aβ species are responsible for a decrease in the long-term potentiation and disruption of synaptic plasticity in AD affected neurons [4-7]. Electrophysiological studies demonstrated that a trimeric form of Aβ_1-42 inhibits long-term potentiation in rodents most effectively [8]. Furthermore, it has been reported [9] that freshly prepared Aβ_1-42 oligomers rapidly induce endoplasmic reticulum stress which results in activation of caspases and apoptosis of cortical neurons in culture, whereas aged fibrillar preparations of Aβ_1-42 were much less toxic to neurons. Notably, it has been demonstrated that Aβ_1-42 tends to form thermodynamically stable trimeric and tetrameric forms that are not obligatory intermediates in the amyloid fibril formation pathways [2]. On the other hand, spherical oligomeric particles with molecular weights ranging from 90 to 110 kDa (diameters in the range of 3–5 nm) were shown to drastically increase the permealization of lipid bilayers [10] which is possibly related to the observed neurotoxicity. Even though distinct differences in the cytotoxicity of very large fibrillar assemblies and small soluble oligomers are now a well established fact [11], there is some experimental evidence that the neurotoxicity may differ significantly even within amyloid aggregate populations with relatively small molecular weights [12]. The purpose of the present work was to establish and quantify the dependence of neurotoxic effects on the size of synthetic Aβ_1–42 oligomers with molecular weights below ~100 kDa.

2. Materials and Methods

The procedures used in this study are approved by the European Convention for the protection of vertebrate animals used for experimental and other purposes and according to the Republic of Lithuania law on the care, keeping and use of animals.

2.1 Neuronal-glial culture preparation

Mixed neuronal-glial CGCs cultures were prepared from 7-8 day old Wistar rats as described [13]. Cells were grown in vitro for 5–6 days before exposure to Aβ. The cultures contained 1.8 ± 0.5% microglial cells, as assessed by staining with isolectin GS-IB4, and 7.3 ± 4.9% astrocytes, as assessed by cellular morphology.

2.2 Preparation of Aβ monomers, oligomers and fibrils

Synthetic Aβ_1-42 peptide was from Bachem (Switzerland) and American Peptide (California, USA). HiLyte Fluor™ 555 labeled Aβ_1-42 peptide, used as a small fraction (0.2 mol%) in oligomer preparation for FCS experiments, was from Anaspec (California, USA). Oligomers were generated as described [14, 15]. Briefly, soluble oligomers were prepared by dissolving 1 mg of peptide in 400 μl HFIP for 30-60 min at room temperature. Sonication was used in this step. 100 μl of the resulting seedless solution was added to 900 μl H₂O in a siliconized Eppendorf tube. After 10–20 min incubation at room temperature, the samples were centrifuged for 15 min at 12000 rpm, the supernatant was transferred to a new siliconized tube and HFIP was evaporated. Concentration of HFIP in solution was monitored in FTIR spectra by a decrease in intensity of the 1192 cm⁻¹ band due to the asymmetric CF₃ stretching vibration [16] (see Supplemental Material, Figs. S1 and S2). Samples were incubated in open vials for 24 hr at 20°C. This protocol is denoted as protocol I. FTIR spectroscopy showed also that the usage of siliconized centrifuge tubes, essential to produce fractions of oligomeric particles with sizes below 3 nm, resulted in transfer of some of the silicon oil components into the vials, which caused strong absorption bands near 1261 and 2962 cm⁻¹, assigned to the bending and stretching vibrations of Si–CH₃ and CH₃ groups, respectively. Vehicle was prepared in the same way as oligomeric forms but without Aβ_1-42 (thus containing similar amounts of residual HFIF or silicone oil). To generate larger oligomers, typically 4-10 nm in diameter, the supernatant was transferred to a non-siliconized Eppendorf tube after the centrifugation and was gently purged with nitrogen for 7 min. The preparation was then stirred in the same vial at ~500 rpm for 24 hrs using a magnetic Teflon-coated stirring bar. Such a protocol will be further referred to as protocol II. Fibrils were formed by protocol III in which the aqueous peptide solution obtained after evaporation of HFIP was incubated for 7 days at room temperature. Monomers were prepared by dissolving Aß_1-42 in HFIP and, after removal of HFIP by evaporation, resuspending in DMSO at a concentration of 0.5 mM. All samples were centrifuged for 15 min at 12000 rpm and the supernatant was transferred to a new siliconized tube. Solutions of peptides were stored at −20°C.

2.3 Fractionation of Aß_1–42 preparations

Ultrafiltrations of Aß_1–42 preparations were performed using Microcon YM-10, YM-30 and YM-100 filters (Millipore) according to manufacturer’s instructions. Briefly, the solutions of freshly prepared Aß_1–42 oligomers were filtrated through Microcon filter with cut-off at 100kDa for 12 min at 12000g. The resulting retentate was supplemented with 100 μl H₂O and recovered by reverse spinning to obtain the fraction with weight of >100 kDa. The supernatant of the first filtration was centrifuged either through the 30 kDa filter (12 min at 12000g) to obtain as the supernatant the fraction with weights of <30 kDa or through the 10 kDa filter (30 min at 12000g) to obtain as the supernatant the fraction with weights of <10 kDa. The retentate after filtration through 30kDa filter was recovered by reverse spinning (plus 100 μl H₂O) and used as the fraction with weights >30 kDa. Protein concentration in the samples was determined by the Bradford method.

2.4 Characterization of oligomer preparations

To assess the size and morphology of the preparations of Aß_1–42 oligomers and fibrils, atomic force microscopy (AFM; Agilent 5500, Santa Clara, CA) was used in the tapping mode. Model TESP (Veeco, Plainview, NY) (f=257-301 kHz, k= 20-80 N/m) and model PPP-NCL-20 (f=146-236 kHz, k=21-98 N/m) (Nanosensors, Neuchatel, Switzerland) microcantilevers were used in this work. According to the manufacturers, the probe tip diameters were between 16 and 20 nm. 20 μl of a 10 μM Aß_1–42 solution was spotted on freshly cleaved mica (SPI Supplies, West Chester, PA), incubated at room temperature for 10 min and rinsed with deionized water (Millipore Inc.), then blown dry with a nitrogen stream. Images were acquired at scan rates between 0.5 and 1 Hz with the drive amplitude and force kept to a minimum. The particle size was estimated by measuring the profile of the sample within the sample plane. The mean height of amyloid aggregates (“z-height”) was estimated by using the Plane Correction (Flattening) module of the SPIP software and determining step-height histograms.

Dynamic light scattering (DLS) experiments were performed on a Zetasizer Nano ZS (Malvern Instruments Ltd, UK) using a laser source with λ=633 nm and a detector at a scattering angle of θ = 173 degrees. Relaxation time distributions were obtained numerically from the field autocorrelation function g₁(t) by means of a regularized inverse Laplace transform routine. Particle size distributions were extracted from the latter through the Stokes-Einstein equation assuming spherical shapes of the particles.

Fluorescence correlation spectroscopy (FCS) was carried out on a Zeiss LSM 510 Meta equipped with a Confocor 3 detection module. Excitation light λ = 561 nm, provided by a diode-pumped solid state (DPSS), is coupled into a 40× 1.1 NA LD-C-Apochromat water immersion objective and the fluorescence emission is epi-collected through the same objective and detected by an avalanche photodiode mounted behind a 575–615 nm bandpass filter. FCS data were fitted with a 3D diffusion model [17] providing the oligomer diffusion coefficient which converts into oligomer size via the Stokes-Einstein equation (for corrections for non-spherical aggregate shapes, see Supplemental Material) .

CD spectra were collected on a Jasco J-810 spectropolarimeter. Spectra were obtained from 185 to 270 nm with a resolution of 0.2 nm and a bandwidth of 1 nm in a 1-cm path length quartz cell. Spectra were then converted to mean residue ellipticity and data were analyzed by using CDPro.

FTIR measurements were performed on a Perkin-Elmer model Spectrum GX FTIR spectrometer equipped with a DTGS detector. FTIR spectra were recorded in transmission mode. The spectral resolution was set at 2 cm⁻¹ and spectra were acquired by co-adding 50 scans. HFIP solution spectra were recorded in a sealed cell with 25 μm path length equipped with CaF₂ windows. Aβ_1–42 preparations were deposited on CaF₂ substrate from 60 μM solution and were dried in air.

Details of the MALDI-TOF characterization of peptide oligomers are given in the Supplementary Materials.

2.5 Assessment of cell viability

The viability of neuronal cells in the cultures treated for 24 h with Aß_1–42 or vehicle (control) was assessed by propidium iodide (PI, 7 μM) and Hoechst 33342 (4 μg/ml) staining using a fluorescence microscope OLYMPUS IX71S1F-3. PI-negative cells with weak Hoechst-staining were considered to be viable, whereas cells showing nuclear shrinkage or fragmentation and intensive Hoechst staining but still lacking PI staining were classified as chromatin condensed/fragmented (apoptotic). PI-positive cells were classified as necrotic. Neuronal cells were distinguished from glial cells according to their characteristic shape and nuclear morphology. Microglial cells were additionally identified by using Isolectin GS-IB4 conjugated with AlexaFluor488 (7 ng/ml, Invitrogen). Neuronal cells were counted in at least 5 microscopic fields per well (two wells per treatment) and expressed as a percentage of the total number of neuronal cells per field.

The activity of LDH released into the medium (an indicator of necrosis) was measured spectrophotometrically [18] by monitoring decrease in NADH at 340 nm as pyruvate is converted to lactate (200 μl aliquot of medium taken from cell culture was added to 0.1 M Tris-HCl buffer, pH 7.5, containing 0.1 mM NADH and 1 mM Napyruvate). LDH activity in control groups was equated to 100 %.

2.6 Vesicle preparation

Vesicles were prepared by mixing stock lipid solutions in chloroform at ratio: DOPC/DOPS/Chol (63:7:30) in a glass tube. Lipid ratios were chosen to mimic to a first approximation neuron membrane composition [19]. The chloroform was evaporated by a gentle stream of N₂. The residual solvent was removed by vacuum-drying lipid film for 1 hr. The film then was dissolved in pentane and left to dry overnight in the hood. The film was hydrated by adding 2.5 ml of working buffer 100mM NaCl, NaH₂PO₄ (pH 7.4), sonicating for 60 minutes and incubating with occasional vortexing or until the lipid film at the bottom disappears. Then the lipid preparation was extruder (Avanti Polar Lipids, USA) through a 100nm polycarbonate membrane for 21 times. Size distribution of vesicles was determined by dynamic light scattering. The preparation exhibited a single distribution peak centered around 150 nm (see Supplemental Material). Large excess of vesicles was used in the oligomer binding to phospholipid experiments. Fluorescently labeled oligomers were prepared as described above.

2.7 Statistical analysis

Data are expressed as mean ± SE of 3-7 experiments on separate CGC cultures. Statistical comparison between experimental groups was performed using Student’s t-test. The Pearson’s correlation method (r) with 99% confidence interval was used for correlation analysis.

3. Results

We analyzed the effect of various Aβ_1-42 assemblies on the viability of CGCs. As shown in Figure 1A, treatment of cells with Aβ_1-42 oligomers prepared by protocol I caused a gradual decrease of neuronal viability in a concentration dependent manner: 94% viability in controls decreased to 11% at 2 μM Aβ_1-42. In contrast, Aβ_1-42 fibrils and monomers had no effect on neuronal viability even at 2 μM concentrations. Vehicle had no effect on cell viability. Aβ_1-42-induced cell death was mainly necrotic as the percentage of PI-positive cells was significantly elevated with the increase of Aβ_1-42 concentration (Fig. 1B). Measurement of LDH activity in the culture media confirmed this observation: the activity of LDH released from cells was found to increase gradually from 10% in cells treated with 0.5 μM to about 70% after treatment with 2 μM Aβ_1-42 oligomeric preparations (Fig. 1C). The percentage of apoptotic cells showing nuclear shrinkage and chromatin condensation was minor (5-10%) with all concentrations of Aβ_1-42 oligomers (Fig. 1B).

Figure 1. Effect of differently prepared Aβ_1-42 assemblies on neuronal cell viability in CGC cultures.

A – Effect of Aβ_1-42 oligomers (protocol I), fibrils (protocol III) and monomers on neuronal cell viability; B – Concentration dependence of Aβ_1-42 oligomers (protocol I) to induce necrotic (PI-positive) and apoptotic (condensed/fragmented nuclei) cell death of neurons. C – Effect of Aβ_1-42 oligomers (protocol I) on LDH release from cells into medium. LDH activity in the control group was equated to 100 %. * - statistically significant effect of Aβ if compared to control. Means ± standard errors of 5-7 experiments on separate CGC cultures are presented.

Importantly, Aβ_1-42 oligomers affected only neurons whereas glial cells were resistant to monomeric, oligomeric and fibrillar forms of the peptide (data not shown). In addition, we tested the effect of Aβ_1-42 oligomers on the murine macrophage J774 cell line and found that at 1 μM concentration Aβ_1-42 oligomers prepared by protocol I did not cause cell death: viability of J774 cells after 24 h incubation with Aβ_1-42 was 96.5 ± 0.6% compared to 97.0 ± 0.3% in cells treated with vehicle. Neither Aβ_1-42 monomers nor fibrils were toxic to J774 cells (data not shown).

To determine whether the toxicity of Aβ_1-42 peptides is CGC neuron specific, we performed experiments comparing the effects of various Aβ_1-42 preparations on primary culture of cortical neurons. We found that cortical neurons responded to Aβ_1-42 in a similar manner as CGCs: 1 μM Aβ_1-42 oligomers prepared by protocol I caused a decrease of neuronal viability by 40 % compared to viability of control cultures after 24 h incubation, whereas Aβ_1-42 monomers and fibrils had no effect on viability of cortical neurons (see supplemental Fig. 5S). This indicates that neurons isolated from different parts of the brain are similarly sensitive to Aβ_1-42 oligomers and insensitive to fibrillar and monomeric forms of Aβ_1-42 at least during short time (24 h) of exposure.

In contrast to Aβ_1-42 oligomers prepared by protocol I, oligomers prepared by protocol II were not toxic to CGC cultures. We then questioned whether the two protocols may generate Aβ_1-42 oligomeric particles of different sizes and thus influence the cytotoxicity of preparations. To investigate this possibility, we carried out an AFM analysis of various Aβ_1-42 preparations precipitated on freshly cleaved mica. Typical results are shown in Fig. 2A-C. Evidently, not only the morphology (oligomers vs. fibrils) but also a size of soluble oligomers depends on the preparation protocol. Analysis of AFM images shows that oligomers prepared by protocol I exhibited mean z-height values spanning from 1 to 3 nm (Fig. 2A and 2D), in few cases reaching 4-5 nm. In distinction, protocol II yielded considerably larger rounded species with sizes of 5-10 nm. The lateral size of the oligomers in AFM images, however, is significantly larger. For the smallest particles with a mean z-height of 1.41±0.06 nm (33 counts), the average lateral size of the particle image was 39.8±2.7 nm (30 counts). The lateral size correlated with the z-height of the particles linearly as r_xy = 7z + 26 nm (R² = 0.8), where r_xy and z are the lateral size and z-height of a particle in nm, respectively. The coefficients in the equation varied by up to 45% depending on the direction along which the image was recorded. This may reflect the asymmetry of the imaging tip.

Figure 2. AFM results.

Exemplary images of Aβ_1-42 oligomers prepared by the different protocols: A - protocol I, B - protocol II, C - protocol III. The lateral size of the images is 4×4 μm. The horizontal bars in A and B shows the location at which hight profiles of the images were analyzed. D and E are the corresponding cross-sections of the particles obtained via protocols I and II.

To investigate how drying of the peptide sample for AFM investigation affects aggregate sizes we carried out DLS and FCS measurements in solutions containing different forms of Aβ_1-42 oligomers. Both methods confirm that protocol I typically generates particles with the sizes from 1 to 4 nm. Representative particle size distributions obtained with DLS are shown in Fig. 3A. Preparations obtained via protocol I (light gray) and protocol II (dark gray) differ significantly. In this particular data set, protocol I oligomers show an average diameter of ~2.7 nm, while the protocol II particles had average diameter of ~8 nm (assuming spherical particle shape). This result is consistent with the AFM data and suggests that drying for AFM imaging does not significantly alter particle size. FCS data obtained with fluorescently labeled aggregates (~ 0.2 mol% of total peptide labeled) indicated similar particle size distributions, which, in case of protocol I preparations (Fig. 3B) exhibits two maxima: one with the mean diameter at ~3.4 nm and another with smaller intensity at ~8 nm. MALDI-TOF mass spectra (see Supporting Material, Fig. S3) show oligomer species with molecular weights up to ~22.54 kDa in preparations obtained by protocol I, while molecular weights of up to ~45.16 kDa were detected in preparations obtained via protocol II. In all cases, preparations were polydisperse in AFM images (Fig. 2), DLS and FCS results (Fig. 3A and 3B).

Figure 3. Size distribution of differently prepared Aβ_1-42 oligomers.

A - DLS data: light gray: protocol I aggregates, dark gray: protocol II aggregates. The larger size oligomer preparation contained also a small contribution of sizes above 200 nm (not shown). B – FCS results for protocol I and protocol II oligomers with the same coding as in panel (A).

The relationship between the toxicity of Aβ_1-42 oligomers and their size is presented in Fig. 4. At 1 μM concentration, small Aβ_1-42 particles (1-2 nm z-height in AFM) obtained by the protocol I were highly toxic to neurons: after 24 h of incubation just 10-40% of viable neurons were observed. Toxicity of Aβ_1-42 decreased with increase in the size of the particles: 50-85% of viable cells were observed after incubation with medium-sized Aβ_1-42 particles (3-5 nm z-height, protocols I and II). Larger particles (5-9 nm z-height, protocol II) were non-toxic to neurons. This suggests that the neurotoxicity depends on Aβ_1-42 oligomer size (Pearson’s correlation coefficient r = 0.74) rather than on the protocol by which the particles are generated, as small oligomers prepared by protocol I are the main cytotoxic species but larger particles prepared by either protocols I and II show comparably low cytotoxicity.

Figure 4. Toxicity of Aβ_1-42 oligomers depends on aggregate size as determined with the AFM in terms of z-height.

Aβ_1-42 oligomers were prepared by protocol I and II. CGCs were treated with 1 μM Aβ_1-42 oligomers for 24 h. Each point represents the effect of separate preparation of oligomers on viability of separate CGC culture (mean values are presented for each experiment).

In principle, the toxic effect of Aβ oligomers may be influenced by functional NMDA receptors which expression in CGC cultures is delayed and reaches maximum on about 10-12 DIV. To test this, we performed series of experiments investigating the effect of various forms of Aβ_1-42 on mature, 10-14 DIV CGC cultures. As can be seen in Fig. 5, the effects of Aβ_1-42 peptides were similar as on 5-6 DIV CGC: 1-2 nm oligomers of Aβ_1-42 at 1 μM concentration were the most toxic species to mature CGCs decreasing neuronal viability by about 60% compared to control, whereas lager, 4-6 nm oligomers as well as monomers and fibrils were not toxic to mature CGC. Aβ_1-42-oligomers-induced neuronal death in mature cultures was mainly (about 60%) necrotic (Fig. 5).

Figure 5. Effect of Aβ_1-42 peptides on neuronal cell viability in 10-12 DIV CGC cultures.

Mixed neuronal-glial CGCs cultures were prepared from 7-8 day old Wistar rats as described in Material and Methods. Cells were grown in vitro for 10–14 days, then treated with 1 μM Aβ_1-42 peptide for 24 h. The viability of neurons in the culture was measured by propidium iodide and Hoechst 33342 staining. * - statistically significant effect of Aβ peptide, if compared to control. Means ± standard errors of 4 separate experiments are presented.

For a refined analysis seeking to reduce the polydispersity of preparations shown in Fig. 3B, we fractionated aggregate sizes by filtering to obtain oligomers with weight ranges of <10 kDa, 10 kDa to 30 kDa, and >30 kDa (see Materials and Methods). Fig. 6 shows the dependence of neurotoxicity on those fractionated amyloid preparations. Specifically, the exposure of CGCs to the fraction of Aβ_1-42 oligomers collected after filtering through the 30 kDa filter reduced neuronal survival by ~60% at 0.35 μM Aβ_1-42 concentration and to less than 10% at 1 μM concentration. Other fractions were not significantly toxic to CGCs.

Figure 6. Effect of Aβ_1-42 on neuronal viability depends on the molecular mass of the peptide assemblies.

Aß_1–42 oligomer preparations were fractionated through different Microcon filters as described in Materials and Methods. * - statistically significant effect of Aβ if compared to control. Means ± standard errors of 3-6 separate experiments are presented.

Although the cytotoxicity of Aβ_1-42 oligomers of different sizes differs significantly, CD spectra do not show differences in their secondary structure (Table 1). The contributions of the secondary structure elements are undistinguishable in preparations obtained by protocols I and II. In contrast, distinctive differences in the distribution of secondary structure elements are observed in monomeric peptide. The content of α-helixes and β–sheet structures in the monomer preparation is higher than previously reported (53 % and < 1 %, respectively) [20], presumably because of sonication procedure used in our work. FTIR spectra of Aβ_1-42 oligomers deposited on CaF₂ substrate revealed several subtle size-dependent differences in the Amide-I spectral region (Fig.7). In agreement with the CD data, the β–sheet secondary structure motive (1628 cm⁻¹ band) dominates for both small and larger oligomers. However, the difference spectrum shows larger absorbance of the small oligomers at 1616 and 1696 cm⁻¹. The absorbance increase near 1616 cm⁻¹ indicates a shift of the 1628 cm⁻¹ component to lower wavenumbers for small oligomers, possibly associated with a different organization of the β–sheet structure due to stronger hydrogen bonding, as strong intermolecular β–strand interaction has been shown to result in a low-frequency shift [21]. The high frequency component (1696 cm⁻¹) is associated with the same intermolecular β–sheet structure. The broad positive band near 1668 cm⁻¹ may indicate the presence of relatively larger contents of loop, turn and α–helix structures in small oligomers [21].

Table 1. Distribution of secondary structure elements in Aβ_1-42 preparations as determined by the CD spectroscopy.

Peptide Aβ1-42	α-helix	β-sheet	Turn	Unordered
Aβ1-42 monomers	71 %	13%	8%	8%
Aβ1-42 small oligomers, protocol I	4%	43%	21%	32%
Aβ1-42 larger oligomers, protocol II	4%	43%	21%	32%

Figure 7. Aβ_1-42 aggregate size causes FTIR spectral differences in the Amide I spectral region.

A- Small oligomers (protocol I), B - larger oligomers (protocol II), and C difference spectrum. Transmission spectra were obtained from dried samples deposited on CaF₂ substrates from 60 μM solution.

To assess whether different size oligomer preparations bind differently to phospholipid membranes we carried out vesicle binding test using FCS. Fig. 8 shows FCS correlation curves fitted with appropriate 3D diffusion models for two different size oligomers preparations. The experiment sequence was as follows. First, we observed oligomers prepared via protocol I (Fig. 8A) and protocol II (Fig. 8B) diffusing 3D freely in solutions with no vesicles, data shown in red dots. These data were fitted with single component 3D diffusion model shown as red continuous lines, and providing diffusion times of 0.05 ms and 0.20 ms for small and big oligomer preparations, respectively. Second, both oligomer batches were brought in contact with lipid vesicles, and the change in fluorescently labeled oligomers diffusion time was observed, data shown as black dots. A two-component 3D diffusion model performs the appropriate fitting (shown in black continuous lines) for this scenario: a fast diffusing component corresponding to free (unbound) oligomers and a slow diffusing component corresponding to vesicle bound oligomers. By fixing the fast diffusion time values, known from measurements of oligomers diffusing in solution free of vesicles, we fit for the slow diffusion times and the fractions of bound and free oligomers. To emphasize the quality of the fits, the residuals are shown at the top of each panel. In the case of small oligomers (Fig 8A) the shift in diffusion time is obvious. The fraction of bound oligomers (slow component) is (>90%) and the slow diffusion time (~4ms) infers vesicles of sizes ~110 nm, agreeing with DLS vesicle size measurements (see Supplemental material). In contrast, in the case of big oligomers produced by protocol II (Fig. 8B) the diffusion time shift is minimal due to a very small fraction of bound amyloid oligomers.

Figure 8. Different size Aβ_1-42 oligomers bind differently to lipid membranes.

FCS data (dots) fitted (continuous line) with single component 3D diffusion model for fluorescent labeled Aβ_1-42 diffusing freely in bulk (red) and with two component 3D diffusion model for Aβ_1-42 diffusing in the presence of non-fluorescent lipid vesicles (black). Residuals of the fits are shown at the top of each panel. All data are normalized for comparison purposes. A- small Aβ_1-42 oligomers (protocol I), B- big Aβ_1-42 oligomers (protocol II).

An assessment of cell densities (presented in Fig. 9) shows that Aβ_1-42 oligomers also reduce the numbers of neuronal cells in the cultures: the total number of neurons (viable, necrotic and apoptotic) per sample area in the cultures decreased after treatment with 1 μM Aβ_1-42 for 24 h. Small oligomers (1-2 nm, protocol I) as well as larger oligomers (4-6 nm, protocol II) similarly decreased the number of neuronal cells, by 15% and 19%, respectively. While both fractions of oligomers cause a similar reduction of neuron densities in the culture, only the small Aβ_1-42 aggregates (protocol I) caused a significant increase in cell death. In distinction, fibrils had no effect on density of neuronal cells (Fig. 9).

Figure 9. Aβ_1-42 oligomers decrease the number of neurons in CGC cultures.

After treatment of CGCs with 1 μM of Aβ_1-42, the total number of viable, necrotic and apoptotic neurons was quantified in 5-7 randomly chosen microscopic fields in each well (two wells per treatment). Total number of cells counted per treatment varied between 1600-2500. * - statistically significant effect of Aβ_1-42 if compared to control. Means ± standard errors of 5-7 separate experiments are presented.

Presence of glial cells may influence the toxicity pattern of Aβ_1-42, therefore we additionally tested the effects of Aβ oligomers on more pure CGC cultures treated with cytosine arabinoside (araC) for 10-14 days. After treatment with araC the number of all non-neuronal cells in the cultures was decreased to about 2%. Meanwhile, as shown in Fig. 10, the effects of various forms of Aβ_1-42 were in principal similar to that observed in gliarich CGC cultures: small, 2-3 nm oligomers reduced neuronal viability by about 40% compared to control, whereas bigger, 4-6 nm oligomers as well as fibrils and monomers were non-toxic. In distinction from glia-rich CGC, neuronal densities in araC-treated cultures were not affected by Aβ_1-42: there were 215 ± 10 neurons/field in control, 216 ± 32 – in cultures treated with small oligomers, 218 ± 39 – treated with fibrils, and 231 ± 22 – treated with monomers. These data suggest that Aβ_1-42 oligomers can directly affect neurons causing cell death though the role of glial cells can not be completely ruled out as these cells may contribute to Aβ_1-42 neurotoxicity by other indirect mechanism(s).

Figure 10. Effect of Aβ_1-42 peptides on neuronal cell viability in pure neuronal CGC cultures.

CGCs cultures were prepared from 7-8 day old Wistar rats as described in Material and Methods. Cytosine arabinoside (10 μM) was added within 48 h of plating to prevent glial cell proliferation. Cells were grown in vitro for 10–14 days before exposure to Aβ_1-42. The cultures contained 0.5% microglial cells and 2.1% astrocytes. Cultures were treated with 1 μM of Aβ_1-42 peptide for 24 h. The viability of neurons in the culture was measured by propidium iodide and Hoechst 33342 staining. * - statistically significant effect of Aβ peptide, if compared to control. Means ± standard errors of 4 separate experiments are presented.

4. Discussion

The results of this study show that the most toxic Aβ_1-42 oligomeric particles are those that exhibit 1-2 nm z-heights and circular shape in AFM scans. Assuming the mean density of the protein to be ~1.4 g/cm³, such 1-2 nm diameter spheres should have a mass that is about the molecular weight of the Aβ_1-42 monomer. However, CD indicates just a minor amount of α-helix content, and there is no indication of Amide I vibrational bands (Fig. 6), associated with α-helices, which is the major component of the monomeric form of Aβ_1-42 [22]. Therefore, one can rule out the possibility that the toxic preparations of oligomers are dominated by monomers. This raises the question whether the oligomers imaged in Fig. 2A are of spherical shape. The AFM results indicate lateral sizes of several tens of nanometers in the preparations which exhibit z-heighs from 1 to 2 nm. It is well-known that the apparent size of objects in AFM images is enlarged due to the size of the AFM tip [23]. Especially large distortions occur when the object’s size is below the curvature radius of the tip. In our case the lateral size r_xy of the particles with a z-height of ~1.5 nm was ~ 40±3 nm, with r_xy depending linearly on z. This suggests that the particles shown in Fig. 2A have true lateral dimensions from 2-3 nm up to 10-15 nm. Therefore, they very likely exhibit lateral dimensions that exceed their z-height. The shape of the aggregates thus deviates significantly from spherical but may more closely resemble oblate spheroids.

This assumption also makes the AFM results consistent with the DLS and FCS analysis. Both techniques determine the diffusion coefficients of oligomers and interpret these in terms of particle geometry under certain assumptions [17]. If the assumption is that the peptide aggregate is spherical, then the Stokes-Einstein equation predicts the size distributions shown in Fig. 3A and B, where the FCS results yield a distribution with an average oligomer diameter of 3.6 nm (average of 8 measurements on 3 different samples), suggesting a considerably bigger size of the particle than that from the AFM data (Fig. 2). However, if one extends the standard Stokes-Einstein approach to spheroids [24] as described in the Supplemental Material, this discrepancy may be resolved. For example, an oblate spheroid with its major and minor axes of ~5 and 1.7 nm (aspect ratio 1:3), respectively, will exhibit about the same diffusion constant as a spherical particle of 3.6 nm diameter. This means the oblate spheroid imaged by AFM as an object of 1.7 nm z-height will exhibit the diffusion constant which under spherical approximation yields particle size distribution shown in Fig. 3 with the mean located around 3.6 nm. It is noteworthy that in the case of oblate spheroids, the average molecular mass of the aggregates consistent with the FCS results would therefore be more than a factor of 2 larger than that of the ideal spheres. Similarly, the size distribution of protocol I oligomers peaks around 2.7 nm (Fig. 3A) in the DLS results, assuming spherical shape of the diffusing particles, but an oblate spheroid with its major and minor axes of 3.8 nm and 1.9 nm (aspect ratio of 1:2), respectively, would have the same diffusion coefficient. However, the spherical particle would have the molecular weight of ~ 8 kDa, (consistent with a peptide dimer) while the spheroid would correspond to a molecular mass of ~12 kDa (a peptide trimer). Importantly, these alternate interpretations of the aggregate shapes in the diffusion measurements are consistent with the results of the AFM investigations if one assumes the same deviations from spherical shapes. Taken all together, we suggest that the preparations of oligomeric particles that exhibited high toxicity towards CGCs were populated with Aβ_1-42 species of low aggregation numbers roughly between dimers and pentamers, with geometric shapes that could be approximated by oblate spheroids.

FCS data in Fig.8 shows that different size Aβ_1-42 oligomers bind differently to a phospholipid membrane of the vesicles which composition to first approximation mimics lipid content of the neuron membranes. While the small fraction of oligomers prepared via protocol I causes a significant shift of the FCS spectrum towards longer correlation times, the big ones prepared via Protocol II exhibit just a marginal effect. The increase of the correlation time signals about the slow down of the thermal motion of the particles that bear the fluorescent label, i.e., amoyloid oligomers. The slowing down under the condition of our experiment may occur only if the fluorescent particle binds to a fluorescently “invisible” vesicle. This, as it is evident from the data in Figs. 8A and 8B, happens only in oligomer preparations obtained via protocol I, which are dominated by the Aβ_1-42 particles with characteristic z-heights from 1-2 nm. The marginal change of the FCS spectrum seen in Fig. 8B may be related to the presence of the small amount of the low molecular weight oligomer particles in protocol II preparations. Alternatively, the shift may be induced by very weak binding of large molecular weight oligomers which, in such a case, should exhibit significantly lower binding energy compared to the small particles.

Our data demonstrate a striking correlation between the size of Aβ_1-42 oligomers and their toxicity to neuronal cells. Even though, we did not find any significant differences in the secondary structure of small and bigger Aβ_1-42 oligomers the cytotoxic effect of these forms was clearly different. Therefore, the main determinant of cytotoxicity seems to be the size of oligomeric Aβ_1-42 particles. Small oligomeric forms of Aβ_1-42, with a particle z-height of 1-2 nm, similarly as recently observed [25] with N-terminally methionine-substituted amyloid peptide, are the most toxic species and induce rapid necrotic neuronal cell death at low submicromolar concentrations, whereas aggregates above 4-5 nm (oligomers with n > 14) do not cause significant cell death. While the cytotoxicity of small (17-22 kDa) oligomers has been known for a decade [26], the continuous transition from highly toxic to non-toxic species as their size increase, is reported here for the first time. On the other hand, even though larger oligomers (n ≈ 12, MW = 38/48 kDa) were reported [27] to cause full inhibition of the long term potentiation, our study shows negligible neurotoxicity of particles with >30 kDa (Fig. 6). In contrast to oligomers, Aβ_1-42 monomers or large fibrillar aggregates were not toxic to CGCs, at least during short time of treatment and at the concentrations used here. These data are consistent with other findings suggesting that oligomeric forms of Aβ_1-42 can affect neuronal cell cycle events [28], cause impairment of synaptic plasticity and long-term potentiation [7, 29], endoplasmatic reticulum stress [30] and cell death [9, 31]. Dimeric and tetrameric forms of Aβ_1-42 have been recently shown to be particularly toxic to cortical neurons due to their high binding capacity to lipid membranes [31]. The binding experiment in this work was performed ex situ, after the pre-adsorption of the amyloid incubated vesicles onto the hydrophobic surface of the array for the subsequent SELDI-TOF mass-spectroscopy analysis [31]. Our data, in which the amyloid binding to the vesicles was detected in situ, strongly supports the idea that the smallest oligomers may be primarily responsible for neurotoxicity observed in vitro. In addition, the binding propensity of these species to the phospholipid membrane may differentiate them from the non-toxic oligomer forms.

In most of the studies, the cytotoxic effects of Aβ were observed at higher concentrations (5-10 μM). We show here that the smallest, 1-2 nm oligomeric particles, possibly dimers – pentamers, caused substantial cell death at submicromolar, i.e., pathophysiologically relevant concentrations. Natural Aβ_1-42 peptide dimers were recently isolated from AD patients brains and were found to cause impairment of synaptic plasticity at nanomolar concentrations [32]. Even though the data in [32] strongly suggests that the Aβ_1-42 dimers may directly target the receptors involved in the synaptic transmission, it may not be excluded that the perturbation of the receptor function may occur through the alteration of the physical properties of the phospholipid matrix which accommodates the protein “machinery” of synapses.

Importantly, this study demonstrated that small oligomers of Aβ_1-42 were toxic to both cerebellar and cortex neurons, whereas glial cells in CGC cultures or macrophage cell lines were resistant to Aβ_1-42 insult indicating that neuronal cells are particularly sensitive to Aβ_1-42 oligomers.

Another observation in this study was that bigger (4-5 nm) Aβ_1-42 oligomer particles caused disappearance of neurons from the treated cultures even without detectable increase in cell death. This suggests that larger oligomers may also be toxic by a different mechanism than Aβ_1-42 dimers - pentamers. A similar disappearance of neuronal cells has been previously reported by other investigators in mixed neuronal-glial cultures under conditions when microglia were activated, e.g. after treatment of cells with NO, bacterial endotoxins, etc. [13, 33]. The causes for this are unclear at present. One of the likely explanations may be that Aβ_1-42-affected neurons could be phagocytosed by microglial cells present in mixed CGC cultures. The involvement of microglia in this process is partially supported by our finding that there were no alterations in neuronal densities after incubation with Aβ_1-42 oligomers in araC-treated cultures. On the other hand, it has been reported [34] that fibrillar Aβ can stimulate phagocytic activity of monocytes and microglia. If oligomeric forms of Aβ_1-42 have similar effect then it would be possible to speculate that such activated microglial cells may phagocytose Aβ-affected neurons. And if the rate of phagocytosis exceeds the rate by which morphological features of cell death develop this could result in removal of neurons that did not exhibit apparent signs of cell death. However, such hypothesis requires thorough experimental investigations.

Abbreviations

Aβ

beta amyloid

AD

Alzheimer’s disease

AFM

atomic force microscopy

CD

circular dichroism

CGC

cerebellar granule cells

DIV

days in vitro

HFIP

hexafluoroisopropanol

FTIR

Fourier transformed infrared spectroscopy

DLS

dynamic light scattering

FCS

fluorescence correlation spectroscopy

PI

propidium iodide