Abstract
The amyloid-β hypothesis of Alzheimer's Disease (AD) focuses on accumulation of amyloid-β peptide (Aβ) as the main culprit for the myriad physiological changes seen during development and progression of AD including desynchronization of neuronal action potentials, consequent development of aberrant brain rhythms relevant for cognition, and final emergence of cognitive deficits.
The aim of this study was to elucidate the cellular and synaptic mechanisms underlying the Aβ-induced degradation of gamma oscillations in AD, to identify aggregation state(s) of Aβ that mediate the peptides neurotoxicity, and to test ways to prevent the neurotoxic Aβ effect.
We show that Aβ1-42 in physiological concentrations acutely degrades mouse hippocampal gamma oscillations in a concentration- and time-dependent manner. The underlying cause is an Aβ-induced desynchronization of action potential generation in pyramidal cells and a shift of the excitatory/inhibitory equilibrium in the hippocampal network. Using purified preparations containing different aggregation states of Aβ, as well as a designed ligand and a BRICHOS chaperone domain, we provide evidence that the severity of Aβ neurotoxicity increases with increasing concentration of fibrillar over monomeric Aβ forms, and that Aβ-induced degradation of gamma oscillations and excitatory/inhibitory equilibrium is prevented by compounds that interfere with Aβ aggregation.
Our study provides correlative evidence for a link between Aβ-induced effects on synaptic currents and AD-relevant neuronal network oscillations, identifies the responsible aggregation state of Aβ and proofs that strategies preventing peptide aggregation are able to prevent the deleterious action of Aβ on the excitatory/inhibitory equilibrium and on the gamma rhythm.
- Alzheimer's disease
- amyloid-β peptide
- BRICHOS domain
- gamma oscillations
- hippocampus
- neuronal synchronization
Introduction
Rhythmic electrical activity or oscillations are a crucial component of brain function (Sederberg et al., 2007). Oscillations in the gamma-frequency band (30–80 Hz) bind neurons into a common temporal matrix, enabling precise temporal coding, spike timing-dependent plasticity, and formation of neuronal assemblies (Singer, 1993). Gamma oscillations are dependent on precise timing of action potential generation and the finely balanced interplay between excitatory and inhibitory neurotransmission in the neuronal network. Gamma oscillations have been suggested to underlie higher cognitive functions, such as sensory perception, attention, and memory (Singer, 1993), and are known to be significantly degraded in Alzheimer Disease (AD) patients, who suffer from deficiencies in their cognitive faculties (Ribary et al., 1991; Uhlhaas and Singer, 2006; Jelic and Kowalski, 2009). It has been shown that increased Aβ levels in AD mouse models result in disrupted timing of evoked action potentials (APs; Stern et al., 2004). However, the exact cellular mechanisms underlying the AD-related degradation of oscillatory activity are yet to be elucidated.
The amyloid cascade hypothesis of AD pathogenesis focuses on toxic amyloid β-peptide (Aβ) aggregates as the main culprit for the deleterious effect on brain function observed in AD and Aβ has been suggested to be responsible for early cognitive decline in AD (Mucke et al., 2000; Selkoe, 2002; Walsh and Selkoe, 2007). Overall, many features of AD pathology can be observed in mutant mice overexpressing amyloid precursor protein (APP) and displaying increased levels of Aβ (Games et al., 1995; Mucke et al., 2000; Götz et al., 2004). Recent studies point to impaired gamma oscillations paralleled by an epileptic phenotype in an AD mouse model, a pathology suggested to be based on a deficit in inhibitory activity (Verret et al., 2012). Many neuronal AD pathologies were also replicated in wild-type rodent brain slices with acute application of Aβ, including synaptic LTP impairment (Shemer et al., 2006; Shankar et al., 2008), as well as neuronal hyperexcitability and resulting network epileptiform activity (Palop et al., 2007; Minkeviciene et al., 2009; Zilberter et al., 2013). Other recent studies also show that Aβ is capable of degrading neuronal gamma oscillations, potentially through similar mechanisms (Nerelius et al., 2009; Pena-Ortega et al., 2012; for review, see Pena-Ortega, 2013).
According to the amyloid cascade hypothesis, processing of APP produces Aβ, aggregation, and fibril formation of which give rise to synaptotoxicity (Mucke et al., 2000) followed by network dysfunction, and eventually dementia (Hardy and Selkoe, 2002; Palop et al., 2006; Palop and Mucke, 2010; Mucke and Selkoe, 2012). Many studies indicate that prefibrillar intermediates, including small soluble oligomers, are most potent in causing neuronal dysfunction, suggesting that oligomeric intermediates in the aggregation process are more toxic than fibrils (Hardy and Selkoe, 2002; Shankar et al., 2008). However, Aβ fibrils are apparently toxic in cell cultures (Chimon et al., 2007; Qiang et al., 2012). Recently, it was shown that secondary nucleation into oligomeric species is mediated by Aβ monomers that bind to the surface of fibrils, and that this is a major source of toxic species (Cohen et al., 2013). Fibrillar Aβ may be toxic as such, but may also cause toxicity by nucleating formation of oligomers, provided that monomeric Aβ is present.
In this study we investigate the relationship between Aβ-induced degradation of gamma oscillations and the concentration of fibrillar over monomeric Aβ forms, elucidate the cellular mechanisms responsible for the toxic Aβ effect and demonstrate the effectiveness of compounds that interfere with Aβ aggregation in the prevention of Aβ toxicity.
Materials and Methods
Animals.
Experiments were performed in accordance with the ethical permit granted by Norra Stockholms Djurförsöksetiska Nämnd to AF (N45/13). C57BL/6 mice of either sex (postnatal days 14–23, supplied from Charles River Laboratories) were used in all experiments. The animals were deeply anesthetized using isofluorane before being killed by decapitation.
Tissue preparation.
The brain was dissected out and placed in ice-cold artificial CSF (ACSF) modified for dissection. This solution contained (in mm); 80 NaCl, 24 NaHCO3, 25 glucose, 1.25 NaH2PO4, 1 ascorbic acid, 3 Na pyruvate, 2.5 KCl, 4 MgCl2, 0.5 CaCl2, 75 sucrose. Horizontal sections (350 μm thick) of the ventral hippocampi of both hemispheres were prepared with a Leica VT1200S vibratome (Leica Microsystems). Immediately after slicing sections were transferred to a submerged incubation chamber containing standard ACSF (in mm): 124 NaCl, 30 NaHCO3, 10 glucose, 1.25 NaH2PO4, 3.5 KCl, 1.5 MgCl2, 1.5 CaCl2. The chamber was held at 34°C for at least 20 min after dissection. It was subsequently allowed to cool to ambient room temperature (∼22°C) for a minimum of 40 min. Peptides were added to the incubation solution 15 min (unless otherwise stated) before transferring slices to the interface-style recording chamber for extracellular recordings or added to the perfusate for patch-clamp recordings in a submerged recording chamber. While incubating, slices were continuously supplied with carbogen gas (5% CO2, 95% O2) bubbled into the ACSF.
Electrophysiology.
Recordings were performed in hippocampal area CA3 with borosilicate glass microelectrodes, pulled to a resistance of 3–5 MΩ. Local field potentials (LFPs) were recorded in an interface-type chamber (perfusion rate 4.5 ml per min) or a submerged-type chamber (for concomitant LFP/patch-clamp recordings) at 34°C using microelectrodes filled with ACSF placed in stratum pyramidale (Fisahn et al., 1998). LFP gamma oscillations were elicited by applying kainic acid (100 nm) to the extracellular bath (Fisahn et al., 2004). The oscillations were allowed to stabilize for 20 min before any recordings were performed.
The interface chamber recording solution contained the following (in mm): 124 NaCl, 30 NaHCO3, 10 glucose, 1.25 NaH2PO4, 3.5 KCl, 1.5 MgCl2, 1.5 CaCl2. The submerged chamber (patch-clamp) recording solution contained the following (in mm): 126 NaCl, 2.5 KCl, 26 NaHCO3, 1.25 NaH2PO4, 1.5 MgCl2, 1.5 CaCl2. Action potentials were recorded from pyramidal cells in area CA3 in cell-attached mode with ACSF-containing patch electrodes. The intracellular solution used for patch-clamp recordings of postsynaptic currents contained the following (in mm): CsMeSO4 140, HEPES 10, MgCl2 2, EGTA 0.6, ATPNa 2, GTPNa 0.3, set to pH 7.2–7.3 with CsOH, osmolarity 270–280 mOsm.
Interface chamber LFP recordings were performed with a 4-channel amplifier/signal conditioner M102 amplifier (Electronics laboratory, Faculty of Mathematics and Natural Sciences, University of Cologne, Cologne, Germany). Submerged chamber LFP recordings and patch-clamp recordings were performed using a Multiclamp 700B (Molecular Devices). The signals were sampled at 10 kHz, conditioned using a Hum Bug 50 Hz noise eliminator (LFP signals only; Quest Scientific), software low-pass filtered at 1 kHz, digitized and stored using a Digidata 1322A and Clampex 9.6 software (Molecular Devices).
Analysis.
Power spectral density plots (from 60-s-long LFP recordings) were calculated in averaged Fourier-segments of 8192 points using Axograph X (Kagi). Oscillation power was calculated by integrating the power spectral density between 20 and 80 Hz. Patch-clamp recordings and concomitant LFP recordings were analyzed using IGOR Pro (WaveMetrics). Data were binned and analyzed over 1 min periods. For the total quantification, the total values were averaged over a 5 min period after 15 min of Aβ application compared with the average of a 5 min initial control period. All data presented normalized was normalized to the average of the first 5 min of control recording before Aβ application.
EPSCs and IPSCs were recorded at −70 mV (EPSCs) and 0 mV (IPSCs) and charge transfer, event amplitude, event frequency, and inter-event-interval (IEI) were analyzed using IGOR Pro with the normalized results representing average values taken over 1 min periods. Paired-pulse ratio experiments were recorded in CA3 pyramidal cells with synaptic potentials evoked by mossy fiber pathway stimulation using a concentric bipolar electrode (dual pulse at 50% maximum response, 100 ms pulse separation, 10 s intersweep interval; values are averages of 10 sweeps). Paired-pulse ratio was defined as the slope of the first EPSP divided by the slope of the second EPSP.
Data are reported as means ± SEM in the text, and as median and upper/lower quartile in the figure box plots. For statistical analysis the Mann–Whitney U test (two-tailed) was used for unpaired data (extracellular) and Student's t test for paired data (patch-clamp). Mann–Whitney U test results are given as the values of the statistic U, the sample sizes n1 and n2, and the significance value p in Table 1. Significance levels are *p < 0.05, **p < 0.01, ***p < 0.001. All experiments were performed with parallel controls from the same animal/preparation.
Summary of raw data values, n numbers, and statistical data for all experiments
Drugs and chemicals.
Chemical compounds used in intracellular and extracellular solutions were obtained from Sigma-Aldrich. Kainic acid was obtained from Tocris Bioscience. The ligand Dec-DETA, designed to stabilize Aβ in an α-helical conformation, was synthesized as described by Nerelius et al. (2009).
Expression and purification of recombinant proteins.
Met-Aβ1-42 was expressed in Escherichia coli BL21 from synthetic genes and purified in batch format using ion exchange and passed through a 30000 Da Vivaspin concentrator filter (Sartorius Stedim Biotech GmbH) to remove large aggregates (Walsh et al., 2009). Purified peptide was concentrated to 50–100 μm, aliquoted in low-bind Eppendorf tubes (Axygene) and stored at −20°C. The peptide concentration was determined using an extinction coefficient of 1400 M−1cm−1 at 280 nm. Aβ1–42 was thawed on ice and diluted to the desired concentration for the respective experiment. Monomeric Aβ1-42 was obtained by lyophilizing concentrated Aβ1-42 peptide, dissolving it in 7 m guanidine hydrochloride and subjecting it to a size exclusion chromatography column. Fibrillized Aβ1-42 was obtained by placing 10 μm monomeric Aβ1-42 at 37°C with 500 rpm agitation for 2 h. Thioflavin-T fluorescence was measured to ensure that Aβ1-42 had fully fibrillized.
Because purified monomeric Aβ1–42 consists essentially of only one molecule species, its mol-per-volume (molarity) concentration can easily be established. Mixed and fibrillar Aβ1-42 solutions on the other hand side are of a highly dynamic molecule composition with molarity and number of different molecule species contained in these solutions being essentially stochastic. The molarity of mixed and fibrillar Aβ1-42 solutions was calculated based on monomeric Aβ1-42. In general we advocate the use of weight-per-volume concentration instead of mol-per-volume with 50 nm Aβ1-42 being equal to 226 ng/ml (molecular weight, Aβ1-42 = 4614 g/mol) because weight-per-volume stays constant, whereas molarity changes during aggregation.
The ProSPC BRICHOS chaperone was made as previously described by Willander et al. (2012). To ensure that ProSPC BRICHOS is free from contaminant, it was subjected to an endotoxin removal column (Thermo Scientific), passed through a 0.22 μm diameter Millex-GV filter (Millipore) and stored at −20°C.
Spider silk proteins (spidroins) convert from a soluble state dominated by helical and random structures to insoluble fibers during silk formation (Simmons et al., 1996) and amyloid-like fibrils have been found in spider silk (Kenney et al., 2002). We have found that a domain from Araneus ventricosus minor ampullate spidroin (Chen et al., 2012) forms typical amyloid-like fibrils, as judged by Thioflavin-T fluorescence, electron microscopy, and Congo red staining. Such fibrils were made by incubating 10 μM of A. ventricosus MiSp domain under quiescent conditions at 25°C in 50 mm sodium acetate buffer, pH 5.2, for 8 h. The fibrils were diluted to 50 nm and studied as described for Aβ fibrils.
Electron microscopy.
Aliquots of 2 μl were taken from the different Aβ1-42 preparations, adsorbed on copper grids and negatively stained with 2.5% uranyl acetate in 50% ethanol. Mixed Aβ1-42 sample (non-SEC purified) stored at −20°C was thawed on ice before being adsorbed on grid. Monomeric Aβ1-42 was adsorbed on a grid directly after SEC purification and after fibrillization. The samples were examined and photographed using a Hitachi H7100 microscope operated at 75 kV.
Results
Effects of Aβ on hippocampal gamma oscillations
Previously we have shown that synthetic Aβ1-42 (Aβ hereafter) reduces hippocampal gamma oscillations (Nerelius et al., 2009) and septal theta synchronization (Leão et al., 2012). Using high-purity recombinant Aβ (Walsh et al., 2009) we first characterized the concentration and incubation time-dependence of the Aβ-induced degradation of hippocampal gamma oscillations. Horizontal hippocampal slices from C57BL/6 mice were incubated with varying Aβ concentrations for varying lengths of time. Subsequent to incubation, gamma oscillations were induced by superfusing slices with 100 nm kainate (KA). LFP recordings in area CA3 revealed control gamma oscillations (no prior incubation with Aβ) of 1.24 × 10−09 ± 1.31 × 10−10 V2 power (n = 19; Fig. 1; Table 1a).
Effects of Aβ on hippocampal gamma oscillations. A, Example traces and example power spectra of control gamma oscillations and their degradation by increasing concentrations of Aβ. B, Summary box plot of gamma oscillation degradation by increasing concentrations of Aβ. C, Example traces and example power spectra of control gamma oscillations and their degradation by increasing exposure times to 50 nm Aβ. D, Summary box plot of gamma oscillation degradation by increasing exposure times to 1 μm Aβ and 50 nm Aβ (inset).
Keeping the incubation time constant at 180 min we proceeded to test the effect of varying concentrations of Aβ (mixed Aβ = containing a mix of all Aβ conformations). Although incubation with the very low concentration of 2 nm Aβ (subphysiological for CNS; Sjögren et al., 2001; for review, see Frankfort et al., 2008) resulted in a nonsignificant reduction of gamma oscillation power (Table 1b), a physiological concentration of 50 nm led to a significant reduction (to 20% of control; Table 1c; Fig. 1A,B). Further increase in Aβ concentration resulted in further gamma power reduction (500 nm Aβ: to 14% of control; 1 μm Aβ: to 1.6% of control; Table 1d,e; Fig. 1A,B). The EDTA-containing solution Aβ was dissolved in had no effect on hippocampal gamma oscillations in the absence of Aβ (Table 1f).
In a parallel set of experiments, we kept the Aβ concentration constant at 1 μm and varied the incubation time from 15 to 180 min. Even a short Aβ application time of 15 min resulted in a very strong reduction of gamma oscillation power (to 11% of control; Table 1g; Fig. 1D). Progressively longer application times resulted in further gamma power reduction (30 min: to 11% of control; 60 min: to 4% of control; 120 min: to 1.7% of control; 180 min: to 1.6% of control; Table 1c,h–j; Fig. 1D).
Finally, we confirmed that the physiological Aβ concentration of 50 nm yields a significant and acute reduction of gamma oscillation power after only 15 min of incubation (to 55% of control; Table 1k; Fig. 1C,D, inset), which could not be washed-out by superfusion with 100 nm KA solution for 180 min (Table 1l). Because of its physiologically relevant concentration and acute effect, the “50 nm Aβ for 15 min” paradigm is well suited for the investigation of cellular and network mechanisms underlying the acute Aβ-induced degradation of gamma oscillations and was used for all subsequent LFP experiments (interface recording chamber for LFP and patch-clamp experiments in submerged recording chamber 1 μm Aβ was used to offset the method-dependent lower-signal amplitude in submerged conditions).
To control for potential unspecific peptide effects and ascertain sequence-specificity of the primary agent, Aβ, we performed experiments with synthetic sequence-inverted Aβ42-1. To control for possible different effects/efficacy of our recombinantly derived Aβ versus the synthetic sequence-inverted Aβ we also performed a control experiment with synthetic Aβ1-42. Synthetic Aβ (50 nm, 15 min incubation) resulted in a significant reduction of KA-induced gamma oscillation power to 43% (KA: 5.73E-09 ± 1.75E-10 V2; n = 8; Aβ: 2.40E-09 ± 4.85E-10 V2; n = 8; Mann–Whitney U test: U = 64, p = 0.00016). The power reduction caused by synthetic Aβ is not significantly different from the power reduction caused by recombinantly derived Aβ (both 50 nm, 15 min incubation; Mann–Whitney U test: U = 60, p = 0.384). In contrast, synthetic sequence-inverted Aβ (50 nm, 15 min incubation) failed to reduce the power of KA-induced gamma oscillation (to 102%; KA: 6.90E-09 ± 9.91E-10 V2; n = 8; Aβ: 7.03E-09 ± 1.96E-10 V2; n = 8; Mann–Whitney U test: U = 40, p = 0.44). Therefore the described Aβ-induced effects on gamma oscillations are sequence-specific to Aβ1-42.
To underscore the relevance of our in vitro model for cognition and gamma oscillations in vivo, we also tested a second mechanism of gamma oscillation induction in vitro for possible Aβ effects: gamma oscillations were induced by 20 μm carbachol in the absence of Aβ to establish control values for gamma oscillation power (9.86E-09 ± 3.32E-10 V2; n = 7). Other slices were incubated for 15 min in an Aβ solution (50 nm; same protocol as for KA-induced gamma oscillations) before induction of gamma oscillations by 20 μm carbachol (3.32E-09 ± 2.40E-10 V2; n = 7). Our data show that carbachol-induced gamma oscillations are significantly reduced to 34% of control by Aβ (Mann–Whitney U test: U = 49, p = 0.0006) and that the described Aβ-induced reduction of gamma oscillations is not specific to kainate-induced gamma oscillations.
Aβ desynchronizes generation of pyramidal cell APs
Desynchronization of AP firing in various cell types is a possible cause for degradation of network rhythms (Andersson et al., 2010, 2012; Leão et al., 2012). Recording pyramidal cell (PC) AP in area CA3 in cell-attached patch-clamp mode and concomitant extracellular gamma oscillations (100 nm KA) we found that application of Aβ (1 μm) led to a steady degradation of gamma oscillation power (to 80 ± 2.3% of control, n = 26, p < 0.0001; Fig. 2A) while leaving gamma oscillation frequency unchanged (control: 31.05 ± 0.95 Hz; Aβ: 31.59 ± 1.17 Hz; n = 26, p = 0.720; Fig. 2B). Concomitantly to the extracellular changes Aβ led to an increase in AP firing in PCs (to 159 ± 26% of control, n = 12, p < 0.05; control: 3.50 ± 0.63 Hz, Aβ: 4.90 ± 0.71 Hz; Fig. 2C). Analysis showed that although the mean phase of AP firing stayed the same (δAP phase = 0, n = 26; Fig. 2D), the firing window became wider consistent with a desynchronization of PCs (AP phase Gaussian fit half-width 211 ± 31% of control, n = 12, p < 0.005; Fig. 2E,F). Note that both degradation of gamma oscillation power, increase of PC AP firing and increase of AP phase Gaussian half-width follow the same kinetics and plateau 10–15 min after Aβ application.
Aβ desynchronizes PC firing. A, Degradation of gamma oscillation power over time by Aβ. B, Gamma oscillation frequency remains unchanged by Aβ. C, AP frequency of PCs is increased by Aβ. D, AP firing phase remains unchanged by Aβ. E, F, AP firing window increases after exposure to Aβ.
Aβ alters excitation/inhibition balance independently from network activation state
Gamma oscillations depend on a tightly regulated equilibrium between inhibition and excitation in the neuronal network. Alterations to this balance are a possible cause to altered network rhythms. We therefore investigated the effect of Aβ (1 μm) on excitatory and IPSCs in CA3 PCs. EPSCs (Vh = −70 mV) and IPSCs (Vh = 0 mV) were recorded from PCs in an activated neuronal network (100 nm KA present). Analysis showed that EPSC charge transfer increased in the presence of Aβ (to 123 ± 6% of control, n = 6, p < 0.05), whereas IPSC charge transfer decreased (to 72 ± 6.7% of control, n = 6, p < 0.05; Fig. 3A). Changes to postsynaptic current charge transfer can be caused by either changes in current frequency and/or amplitude. Our experiments showed that Aβ caused an increase of both EPSC frequency (control: 19.15 ± 0.25 Hz; Aβ: to 116 ± 5% of control, n = 6, p < 0.05) and amplitude (control: 51.04 ± 0.21 pA; Aβ: to 115 ± 3% of control, n = 6, p < 0.001; Fig. 3A), while causing a decrease of both IPSC frequency (control: 42.67 ± 0.09 Hz; Aβ: to 73 ± 10% of control, n = 6, p < 0.05) and amplitude (control: 124.21 ± 0.35 pA; Aβ: to 68 ± 10% of control, n = 6, p < 0.05; Fig. 3A).
Aβ alters balance between excitation and inhibition. A, In an activated network (100 nm KA), Aβ increases EPSC charge transfer and decreases IPSC charge transfer. These changes are based on an Aβ-induced increase of EPSC frequency and amplitude and a decrease of IPSC frequency and amplitude. B, In a quiescent network, Aβ increases EPSC charge transfer and decreases IPSC charge transfer. These changes are based on an Aβ-induced increase of EPSC frequency and amplitude and a decrease of IPSC frequency and amplitude.
To ascertain that the Aβ-induced alteration of excitatory/inhibitory balance is not dependent on the activation state of the neuronal network, we repeated our EPSC/IPSC recordings in quiescent slices (no KA present). Both spontaneous EPSCs and spontaneous IPSCs recorded under quiescent conditions showed the same response to Aβ as postsynaptic currents in the activated network. Spontaneous EPSC charge transfer increased to 129 ± 7.6% of control (n = 13, p < 0.005) with both frequency and amplitude increasing (control frequency: 24.86 ± 0.09 Hz; Aβ frequency: to 122 ± 7.7% of control; control amplitude: 17.00 ± 0.14 pA; Aβ amplitude: to 111 ± 4.5% of control; both p < 0.05, both n = 13; Fig. 3B). Spontaneous IPSC charge transfer decreased to 71 ± 5.1% of control (n = 14, p < 0.0001) with both frequency and amplitude decreasing (control frequency: 33.54 ± 0.12 Hz; Aβ frequency: to 81 ± 5% of control, p < 0.001; control amplitude: 14.93 ± 0.07 pA; Aβ amplitude: to 76 ± 4.5% of control, p < 0.0005; both n = 14; Fig. 3B).
Consistent with the above results cumulative probability distributions analysis showed that amplitudes of EPSCs increase moderately while their inter-event-intervals decrease in both the activated and the quiescent network state. In contrast, the amplitudes of IPSCs decrease and their IEIs increase in both the activated and the quiescent network state. Finally, we conducted paired-pulse experiments in CA3 pyramidal cells (no KA present) with mossy-fiber pathway stimulation. The ratio between the first to second EPSP slope decreased from 1.082 ± 0.109 in control conditions to 0.872 ± 0.1119 after 20 min exposure to 1 μm Aβ (n = 65, p < 0.01). This suggests that Aβ modifies presynaptic release dynamics, which may contribute to the observed increase in EPSC amplitude.
Severity of Aβ effect increases with fibrillization
Knowledge of which step in the misfolding and aggregation sequence of Aβ is the (most) cytotoxic and responsible for the cellular and network changes seen in AD remains sparse and contradictory. Arguments have been advanced for soluble Aβ oligomers rather than fully formed Aβ fibrils (Klein et al., 2001; Nimmrich and Ebert, 2009). Using purified Aβ preparations that contained either monomeric Aβ, fibrillar Aβ or a mix of all conformations (mixed Aβ; all 50 nm, 15 min incubation; Fig. 4A) we tested their ability to degrade gamma oscillations. Our experiments showed that with increasing presence of fibrillized Aβ forms in the incubation solution the level of gamma oscillation degradation also increases when compared with controls (Table 1m): monomeric Aβ: to 66% of control; mixed Aβ: to 35% of control; fibrillar Aβ: to 19% of control; Table 1n–p; Figure 4B,C). The results of this experiment suggest that the fibrillar form of Aβ is the most effective in eliciting detrimental network and cellular effects. Because it has been shown that secondary nucleation into oligomeric species is mediated by Aβ monomers that bind to the surface of fibrils, and that this is a major source of toxic species (Cohen et al., 2013) we conclude that fibrillar Aβ may be toxic as such, but may also cause toxicity by nucleating the formation of oligomers.
Severity of Aβ effect increases with fibrillization. A, Electron micrographs of Aβ1-42 preparations negatively stained with uranyl acetate. Monomeric: in the SEC-purified Aβ monomer, no aggregated material was found. Mixed: the non-SEC purified Aβ contained aggregated material but also scarce fibrils. Fibrillar: after incubation of monomeric SEC-purified Aβ at 37°C, fibrils were abundant. B, Example traces and example power spectra of control gamma oscillations and their degradation by monomeric, mixed, and fibrillar Aβ. C, Summary box plot of gamma oscillation degradation by monomeric, mixed, and fibrillar Aβ.
To exclude the possibility that the strong degradation of gamma oscillations we see in the presence of fully fibrillized Aβ is due to unspecific fibril effects, we conducted a control experiment in which fibrils derived from a spider silk protein domain were substituted for Aβ fibrils. Incubating hippocampal slices with 50 nm spider silk-derived fibrils for 15 min as well as 180 min did not alter the size of subsequently induced gamma oscillations (Table 1q,r). We conclude that the strong degradation of gamma oscillations by Aβ fibrils is Aβ-specific and not due to unspecific fibril effects.
Prevention of gamma degradation by interference with Aβ-folding and aggregation
Several compounds are currently available that can interfere with the Aβ-folding and aggregation sequence at different points (Schenk et al., 2012). We used two such compounds to further determine the (most) cytotoxic conformation of Aβ. Dec-DETA is a designed ligand that shifts the equilibrium between the α-helical and the β-sheet secondary peptide conformation of Aβ toward the α-helical form (Nerelius et al., 2009; Ito et al., 2012). Because the β-sheet form is necessary for both potential pore formation (Jang et al., 2010; Sepulveda et al., 2010) as well as fibril formation, Dec-DETA limits both by interfering at a very early point in Aβ's folding and aggregation sequence. When incubating slices for 15 min with monomeric Aβ, mixed Aβ, or fibrillar Aβ (all 50 nm) in the presence of Dec-DETA (250 nm), we found that the presence of the ligand completely prevented the degradation of gamma oscillations by monomeric Aβ (Table 1s) and mixed Aβ (Table 1t), and partially prevented degradation by fibrillar Aβ (to 63% of control; Table 1u) (Fig. 5A,B). Control experiments with Dec-DETA-only incubation showed no change of gamma oscillation power (Table 1v) compared with control gamma oscillations.
Prevention of gamma degradation by interference with Aβ folding and aggregation. A, Example traces and example power spectra of control gamma oscillations and prevention of their degradation by Aβ in the presence of a designed ligand. B, Summary box plot of prevention of gamma oscillation degradation by monomeric, mixed, and fibrillar Aβ in the presence of a designed ligand. C, Example traces and example power spectra of control gamma oscillations and prevention of their degradation by Aβ in the presence of a chaperone. D, Summary box plot of prevention of gamma oscillation degradation by monomeric, mixed, and fibrillar Aβ in the presence of a BRICHOS chaperone.
The second compound used was recombinant human BRICHOS domain from ProSPC, which delays aggregation of Aβ into β-sheet oligomers and fibrils in vitro (Willander et al., 2012; Knight et al., 2013) and in vivo (Hermansson et al., 2014). Compared with Dec-DETA ProSPC BRICHOS limits fibril formation by interfering at a later point in Aβ's folding and aggregation sequence. When incubating slices for 15 min with monomeric Aβ, mixed Aβ, or fibrillar Aβ (all 50 nm) in the presence of ProSPC BRICHOS (1 μm), we again found that the presence of the ligand completely prevented the degradation of gamma oscillations by monomeric Aβ (Table 1w) and mixed Aβ (Table 1x), and partially prevented degradation by fibrillar Aβ (to 48% of control; Table 1y; Fig. 5C,D). Control experiments with ProSPC BRICHOS-only incubation showed no change of gamma oscillation power (Table 1z) compared with control gamma oscillations.
It is interesting to note that the effectiveness of both Dec-DETA as well as ProSPC BRICHOS to prevent degradation of gamma oscillations follows the effectiveness of the particular Aβ solution to induce it (in the presence of Dec-DETA: monomeric Aβ to mixed Aβ: U = 85.5, n1 = 14, n2 = 8, p = 0.042; mixed Aβ to fibrillar Aβ: U = 87.0, n1 = 14, n2 = 8, p = 0.035; in the presence of ProSPC BRICHOS: monomeric Aβ to mixed Aβ: U = 82.0, n1 = 15, n2 = 8, p = 0.169; mixed Aβ to fibrillar Aβ: U = 63.0, n1 = 8, n2 = 8, p = 0.00031; Fig. 5B,D).
Preservation of excitatory/inhibitory balance by interference with Aβ folding and aggregation
We have shown that designed ligands that interfere with Aβ-folding (Dec-DETA) and aggregation (ProSPC BRICHOS) prevent the Aβ-induced degradation of gamma oscillations. Because Aβ also alters the balance between excitation and inhibition on the cellular level, we proceeded to test whether Dec-DETA and ProSPC BRICHOS act to protect this balance against Aβ-induced changes, explaining their protective effect on extracellular gamma oscillations. We therefore investigated the effect of Aβ (1 μm) in the presence of either ligand on inhibitory and EPSCs in CA3 PC. EPSCs (Vh = −70 mV) and IPSCs (Vh = 0 mV) were recorded from PCs in an activated neuronal network (100 nm KA present). Analysis showed that both EPSC and IPSC charge transfer remained unaltered in the presence of Aβ and Dec-DETA (5 μm; EPSCs: 95 ± 4% of control, n = 5, p > 0.37; IPSCs: 97 ± 4% of control, n = 5, p < 0.49; Fig. 6A). Likewise, in the presence of ProSPC BRICHOS (1 μm) EPSC and IPSC charge transfer remained unaltered (EPSCs: 99 ± 3% of control, n = 5, p > 0.69; IPSCs: 102 ± 7% of control, n = 4, p < 0.79; Fig. 6B). We conclude that the protective effect of Dec-DETA and ProSPC BRICHOS on gamma oscillation power is founded on the compounds ability to prevent the Aβ-induced alteration of the excitatory/inhibitory balance in the hippocampal network.
Preservation of excitatory/inhibitory balance by interference with Aβ folding and aggregation. A, In an activated network (100 nm KA), a designed ligand prevents the Aβ-induced increase in EPSC charge transfer and decrease in IPSC charge transfer. B, In an activated network (100 nm KA), a BRICHOS chaperone prevents the Aβ-induced increase in EPSC charge transfer and decrease in IPSC charge transfer.
Discussion
In this study, we investigated the cellular mechanisms responsible for the degrading effect of Aβ on AD-relevant functional network dynamics, namely gamma oscillations, as well as means to prevent the neurotoxic effect of Aβ. By doing so we have explored the functional link between the main culprit for the cognitive deficits observed in AD patients, Aβ, and a brain rhythm that underlies higher cognitive functions and is known to be significantly degraded in AD patients, who at the same time suffer from deficiencies in their cognitive faculties (Ribary et al., 1991; Uhlhaas and Singer, 2006; Jelic and Kowalski, 2009). Furthermore, we have linked the level of Aβ toxicity to the initially available concentration of fibrillar over monomeric Aβ forms and have demonstrated that compounds that interfere with Aβ aggregation can prevent Aβ neurotoxicity.
Gamma oscillations are a type of brain rhythm that is particularly important for higher brain functions, such as cognition, attention, and memory (Singer, 1993). As a rhythmic electrical brain activity its generation and maintenance is dependent on the synchronization of action potential firing of different cell classes and the tightly regulated balance of excitation and inhibition in the neuronal circuitry. It is therefore only consistent that the degrading effect of Aβ on gamma oscillation strength is caused, as our data show, by an Aβ-induced desynchronization of AP firing in pyramidal cells and a general shift of the balance between excitation and inhibition in the hippocampal circuitry.
Our data show an increasing Aβ-induced degradation of gamma oscillations with increasing concentration of initially available fibrillar over monomeric Aβ forms. Importantly, most of the degrading effect on gamma oscillations is caused by nanomolar concentrations of Aβ, which is comparable to physiological Aβ concentrations, and contrasts favorably with the concentrations in the micromolar range used in many other studies. Likewise, the designed ligand Dec-DETA and the chaperone domain BRICHOS from ProSPC used to prevent Aβ toxicity demonstrate a decreasing prevention effectiveness with increasing concentration of initially available fibrillar over monomeric Aβ forms. Although both results overtly point to the fibrillar form of Aβ as the major source of the peptides toxicity, and as such, appear contrary to the oligomeric hypothesis of AD this is not necessarily the case: arguably it is possible that Aβ fibrils have some inherent toxicity. However, given their sheer size it seems unlikely that they are able to penetrate our hippocampal slice preparation readily enough to explain a profound degradation of cellular parameters and gamma oscillations after only 15 min of exposure. Penetration of single cells seems even less likely. However, it is plausible that a higher initial concentration of fibrillar Aβ translates into a higher concentration of oligomeric Aβ because of the deconstruction of fibrils into oligomers and the induction of oligomer production by secondary nucleation (Cohen et al., 2013).
Our study provides correlative evidence for a link between Aβ-induced effects on synaptic currents and AD-relevant neuronal network oscillations. Our findings of the coupling of Aβ toxicity to initially available concentration of fibrillar over monomeric Aβ forms, together with other recent findings on Aβ secondary nucleation mechanisms (Cohen et al., 2013), open the door to combine research obtained on the molecular Aβ level with functional electrophysiological investigations in AD-relevant brain activity. Finally, our study underscores the promising approach to prevent Aβ toxicity by stabilizing the peptide's α-helical form (Nerelius et al., 2009) or preventing its aggregation into fibrillar form (Knight et al., 2013; Hermansson et al., 2014) by providing evidence that these potentially therapeutic approaches indeed prevent the underlying cellular causes of Aβ toxicity for functional network dynamics, namely action potential desynchronization and shift of excitatory/inhibitory balance in the neuronal network.
Footnotes
This work was supported by a KID PhD studentship grant (F.R.K.), a Swedish Brain Foundation postdoctoral fellowship (M.Z.), the Swedish Research Council (A.F., J.J., R.S., J.P.), Alzheimerfonden (A.F., D.H.), the Swedish Brain Power program (B.W.) and the Swedish Medical Association (A.F.), the Swedish Brain Foundation (A.F.), and the Strategic Program in Neurosciences at the Karolinska Institute (A.F., B.W.).
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr André Fisahn, Neuronal Oscillations Laboratory, Center for Alzheimer Research, Department of Neurobiology, Care Sciences and Society, Karolinska Institute, Retzius väg 8, 17177 Stockholm, Sweden. andre.fisahn{at}ki.se
