Synaptotoxicity of Alzheimer Beta Amyloid Can Be Explained by Its Membrane Perforating Property

The mechanisms that induce Alzheimer's disease (AD) are largely unknown thereby deterring the development of disease-modifying therapies. One working hypothesis of AD is that Aβ excess disrupts membranes causing pore formation leading to alterations in ionic homeostasis. However, it is largely unknown if this also occurs in native brain neuronal membranes. Here we show that similar to other pore forming toxins, Aβ induces perforation of neuronal membranes causing an increase in membrane conductance, intracellular calcium and ethidium bromide influx. These data reveal that the target of Aβ is not another membrane protein, but that Aβ itself is the cellular target thereby explaining the failure of current therapies to interfere with the course of AD. We propose that this novel effect of Aβ could be useful for the discovery of anti AD drugs capable of blocking these “Aβ perforates”. In addition, we demonstrate that peptides that block Aβ neurotoxicity also slow or prevent the membrane-perforating action of Aβ.


Introduction
Alzheimer's disease (AD) is a progressive and irreversible neurodegenerative brain disorder that leads to major debilitating cognitive deficits in the elderly. It is now believed that the cellular and molecular alterations that cause brain dysfunctions are slow in onset and that it probably takes several years to develop the full blown disease [1]. Surprisingly, the cellular and molecular mechanisms that induce AD are largely unknown, deterring the development of effective modifying or symptomatic therapies. Thus, attempts to alleviate and stop AD symptoms are actually based on compensating synaptic deficits and blocking intracellular signaling cascades [2]. However, the results are minor because at clinical stages the AD brain is already too deteriorated.
It is accepted that the toxic effects of Ab, one etiological agent in AD, depend on dimer formation and subsequent oligomerization which include diverse structural forms [3]. Thus, blocking dimerization reduces aggregation and the ensuing peptide toxicity [4]. The working hypothesis of AD is that excess of Ab either i) binds to membrane receptors affecting their functions [5], ii) interferes with signaling cascades [6][7][8] or iii) directly disrupts neuronal membranes causing pore formation leading to alterations in ionic homeostasis [9]. Although the latter is an attractive hypothesis because it could explain several effects of Ab in brain neurons, it is largely unknown if this can also occur in native brain neuronal membranes. Additionally, the existence of this mem-brane phenomenon will reveal that the target of Ab is not another membrane protein, but that Ab itself is the cellular target and explain the failure to interfere with the course of AD. In agreement with this idea, atomic force microscopy (AFM) in lipid environments and molecular dynamic analysis have shown the presence of molecular entities with inner diameters in the 1.5-2.6 nm range [10,11] which were similar to those generated by other peptidergic molecules known to form pores in cell membranes, such as amylin and a-synuclein [12].
For many years it has been recognized that several peptides with differing structures such as gramicidin, amphotericin and a-latrotoxin can alter membrane permeability after inducing pore formation [13,14]. Additionally, it is known that antifungal antibiotics are toxic because they can attach to the cell wall and steadily disrupt permeability. Electrophysiologists have utilized gramicidin and amphotericin for more than 20 years to perforate cell membranes and record whole cell ionic currents with the patch clamp technique [13,15,16]. In the patch perforated mode, the membrane is disrupted thereby making holes that allow the continuous flow of ionic currents under the patch pipette. Here, we report that Ab has a rapid and potent perforating property in neuronal membranes. We postulate that these perforations increase intracellular calcium leading to synaptic transmission failure [17]. Based on this membrane property, similar to gramicidin and amphotericin, we have defined Ab as a perforating toxic agent, rather than a classical pore-forming agent. The cell attached mode of the patch clamp technique allows for  stable measurements at the single molecule level. Thus, activation  of most single voltage or ligand activated channel proteins can be  adequately time resolved with open kinetics and complex cellular  regulations [16,18]. Alternatively, small peptides can produce minute disturbances in membrane stability causing a low resistance pathway under the membrane patch, known as a ''perforated recording configuration'' [19]. In this study, cellattached recordings in hippocampal neurons were stable using a control solution in the patch pipette (i.e. .30 min). For example, the application of a 5 mV voltage pulse induced a stable, fast current arising from a partly compensated electrode capacitance (Fig. 1A). This current was markedly altered when 500 nM of preaggregated Ab (see methods) was added into the patch pipette solution and allowed to diffuse to the underlying membrane (Fig. 1B). For example, the traces in figure 1B clearly show that the amplitude and time course of the capacitive current increased with time and this was similar to that induced by gramicidin (Fig. 1C). Fig. 1D obtained with a fluorescent form of Ab and with Western blot analysis shows that the peptide was able to associate with neuronal membranes at times when it was producing membrane perforation (i.e. 15-30 min). The data also show that the time it took to form the perforated configuration by Ab was dependent on the peptide concentration (Fig. 1E). For example, it took nearly 40 min to establish perforated recordings with 1 nM Ab and less than 10 min with micromolar concentrations. Ab (2.2 mg/ml, 500 nM) was more potent and rapid than gramicidin in forming the perforated configuration. However, the perforated configuration formation with gramicidin (100 mg/ml) can take more than 30 min [20]. Interestingly, Ab 1-42 produced similar effects in membrane charge and input resistance as those of Ab 1-40 (Fig.  S1B,C). Furthermore, the Ab-dependent actions were demonstrated by their blockade with the Ab WO2 antibody that recognizes residues 1 to 5 of Ab (Fig. S1B,C).

Ab perforates hippocampal neuron membranes
The analysis of the charge transferred during the capacitative response showed that the effect of Ab was similar to those of gramicidin and amphotericin, two peptides commonly used to perforate neuron membranes (Fig. 1F). On the other hand, the effect of Ab 1-40 was not produced by the reverse Ab peptide, supporting the idea that Ab aggregation leads to membrane damage. Finally, the membrane currents induced by Ab were very similar to those induced by positive pressure in the whole cell configuration, suggesting that they can transfer significant charge, equivalent to that induced by positive pressure, which is believed to completely burst the membrane under the patch pipette [18]. Ab displayed a ''gramicidin-like'' behavior in neuronal membranes Peptides that perforate cell membranes can form pathways which are more or less selective to cations or anions [15,16]. Gramicidin and amphotericin, for example, are used to record GABA A and glutamatergic whole-cell currents, respectively, because while the former generates mainly cationic pores in the membrane, the latter is somewhat more selective for anions. Consistent with a time dependent membrane perforation process, the application of extracellular GABA or glutamate only 30 s after GV seal formation was unable to induce detectable membrane currents. This demonstrates the existence of a high resistance pathway between the membrane and the patch pipette containing 500 nM Ab. After 15 minutes of Ab application to the patch membrane, on the other hand, extracellular applications of both neurotransmitters induced membrane currents, demonstrating the formation of pathways in the membrane capable of conducting ionic currents through the Ab-containing pipette ( Fig. 2A, lower traces). Additionally, the data show that Ab induced perforated patches in a time-dependent manner making it possible to detect synaptic currents arising from synapses distant to the recording patch electrode (Fig. 2B). These results overwhelmingly show that Ab is acting in a similar fashion to other pore forming peptides (i.e. gramicidin, amphotericin) well known to perforate neuronal membranes. Additionally, these novel results are appealing because they show that Ab resembles other well known potent cytotoxic compounds, providing a novel molecular mechanism for neuronal toxicity of the Ab peptide.
We next studied the current-voltage (I-V) relationships [21] to compare the ionic selectivity properties of perforates produced by Ab (20 min of application) with those of gramicidin and amphotericin, known to form cationic and anionic selective pores, respectively. The application of GABA, the agonist for the GABA A Cl 2 current present in hippocampal neurons, showed that Ab behaved like gramicidin, but not like amphotericin (Fig. 2C). For example, the Cl 2 current recorded with Ab in the pipette reversed direction near the expected equilibrium potential for Cl 2 in the perforated mode [18]. Amphotericin, on the other hand, which dissipates the Cl 2 gradient, reversed the GABA A current at 0 mV. The data also shows that the AMPA current reversed close to 0 mV with the three perforating peptides (Fig. 2D), which is near the expected value for a non-selective cationic channel.
Ab action on conductance is not mediated by a classical channel-like behavior Some biophysical studies have indicated that Ab can increase intracellular calcium and membrane conductance in artificial lipid bilayers and clonal cell lines [22][23][24], but demonstration of actual channel or pore formation in native brain membranes has been inadequately resolved. In our experiments, it was clear that Ab was able to induce an increase in membrane noise before establishing the perforated configuration in hippocampal neurons. However, the noisy nature of the neuronal membrane related to activation of endogenous channels precluded us from studying the pore properties in more detail. To circumvent this, we recorded from HEK 293 cells before the formation of a perforated configuration. Recordings done in more than 40 cells did not show membrane events reminiscent of typical single channel behavior, in the sense of having well structured open and closing behavior, in the presence of Ab. Therefore, the noisy nature of the microscopic current events produced by Ab did not allow for a good discrimination between conformational states or to construct open and shut distributions (Fig. 3A). Nevertheless, plots of all-point current distributions from different patches showed multiple levels of peak conductance (20062, 260640, 360660, 440620 and 680660 pS) supporting the occurrence of multiple membrane disruptions by Ab (Fig. 3B) more than formation of a single unitary channel. In parallel experiments, we found that fluorescent Ab was able to associate to cell membranes giving a morphological correlation to the membrane-perforating actions of the peptide (Fig. 3C). Furthermore, in some patches, Ab produced a large transient increase in membrane current (1000-2000 pS) which we interpreted as spontaneous breakage-resealing of the membrane produced by Ab that sometimes resulted in a whole cell configuration (Fig. 3F-H). Interestingly, Ab was able to bind widely to HEK 293 cells and also caused the generation of a perforated configuration in these cells, as indicated by parallel monitoring of membrane capacitance (Fig. S3 and Fig. 3C). These data, therefore, indicate that Ab affects the membrane inducing a range of current responses which are different from those of membrane channels, which have well defined conductance and time distributions due to their gating properties [25,26]. For instance, the analysis of single channel currents associated with a1 glycine receptors provided a way of comparing a typical ion channel having a single channel conductance of 9263 pS with Ab activity (Fig. 3D,E). The previous experiments showed small and large microscopic membrane current events induced by Ab. While the smaller Ab perforations might exhibit a degree of ion selectivity (Fig. 2C), it is likely that the large ones might allow the entry of other molecules into the cell, which can be examined using fluorescent probes loaded in the pipette.

The amyloid pore allows entry of a large molecule into neurons
The data in figure 4 show combined patch clamp-imaging recordings using patch pipettes filled with ethidium bromide in the presence and absence of Ab (Fig. 4A,C). From this data, it is evident that a large (M.W. 394.3, ,1.3 nm van der Waals diameter, PDB ID: 2ZOZ) organic molecule can enter the neuron in parallel with the process of electrical membrane perforation (Fig. 4A, D). In the absence of Ab in the pipette (Fig. 4C), or with Na7, an Ab-pore blocking peptide [27] (Fig. 4B,E), ethidium bromide was unable to enter into the cell. The size of this fluorescent molecule allows us to place a diameter of at least 1.5 nm for the large perforation induced by Ab, which agrees with previous AFM data [10,11]. Clearly, these large perforations might cause enormous homeostatic consequences for neuronal functions.

The amyloid perforates can be inhibited by small peptides
One of the main issues related to the toxicity of Ab in brain neurons is the identification of potential targets for the development of pharmaceutics capable of blocking its effects. In line with the idea that pore formation is relevant to Ab toxic actions, it was reported that the increase in Ca 2+ influx and lipid bilayer conductance was blocked by two small peptides [27]. Interestingly, these peptides also inhibited Ab-induced cell death [28]. Furthermore, the increase in charge transfer and entry of ethidium bromide into the cell was well inhibited by this peptide (Fig. 4B,E). In addition, we found that the Na7 peptide produced a blockade of Ab effects on membrane resistance (1/G) in hippocampal neurons in a concentrationdependent fashion (Fig. 5A). In experiments using hippocampal neurons loaded with fluo-4, Ab produced a reversible increase in intracellular calcium, demonstrating the diffusible nature of Ab. Additionally, this increase was antagonized by Na7 (Fig. 5B), but not by other blockers of ligand-gated or voltage-dependent calcium channels, suggesting that this effect was mainly mediated by Abinduced membrane perforation.
Consistent with the critical role of calcium in synaptic transmission and in agreement with recently published data [17], we found that 500 nM of Ab enhanced the release of synaptic vesicles from hippocampal neurons. This synaptic facilitation was blocked by the presence of the Na7 peptide suggesting the participation of Ab perforation in this phenomenon (Fig. 5C). Na7 and Na4a, another structurally related peptide (Fig. S1A), but not the inactive analogs Na13 and Na15 [29], also antagonized the delayed synaptotoxic effects of Ab on synapsin I and SV2, two vesicular proteins (Fig. 5D), and in addition altered membrane charge and resistance (Fig. S1B,C). In conclusion, the data indicate that the perforating effects of Ab are associated to microscopic structures resembling small fibrils (Fig. S2A), but not to unstructured forms of Ab (Fig S2B). This data might be important for future pharmacological applications in terms of the neurotoxic activity of Ab. Furthermore, these results strongly suggest that Ab perforations are involved in synaptic dysfunction mediated by Ab oligomers [17].

Discussion
Only recent studies have dealt with the action of Ab at concentrations without overt neurotoxicity on synaptic properties [3]. Although controversy exists on whether Ab can up or down regulate specific components of synaptic transmission, several studies in rodent hippocampus showed that Ab alters pre and postsynaptic components governing LTP, NMDA-and AMPA neurotransmissions and calcium homeostasis. No definitive mechanism is available to explain this variety of effects [30][31][32], hindering the development of anti Ab therapies. On the other hand, due to the urgency of generating disease-modifying therapeutics to treat people suffering from AD, we believe that the present data provide novel insights into innovative strategies to interfere with the toxic processes likely initiated at the neuronal membrane level [29]. Future studies should decipher the characteristics of Ab perforate formation in brain membranes. For example, although pore forming peptides have been in use for more than 40 years, most of their mechanisms for membrane insertion, pore formation and membrane conductance initiation have remained largely undetermined [13]. For example, from lipid bilayer studies, it was postulated that gramicidin required simultaneous insertion of two monomers on opposite faces of the lipid bilayer to perforate the membrane. However, this phenomenon might not occur in biological membranes. Our most recent experiments have shown that gramicidin forms oligomeric complex structures in aqueous solution and induces membrane perforations, similar to Ab, rather than single channel currents in native cell membranes (unpublished results). Nevertheless, because Ab can internalize rapidly [33], it might break the membrane inserting itself in both faces.
In agreement with the data in the present study, AFM and molecular dynamic studies of Ab pores in bilayers support the presence of diverse, small and large molecular entities that possibly correspond to the functional perforation described in this study. The Ab inner pore diameter appears to be much larger (at least 2.6 nm) than ion selective channels, which have an estimated diameter of 0.6 nm [21]. Overall, studies with AFM, molecular simulations and single channel conductances suggest a high range of pore sizes [34], and provide additional support to the idea that the phenomenon of insertion and conductance of Ab are very complex. The proposal of a complex pore structure is consistent with a recent study that proposed a model for the Ab pore [42]. Additionally, because these conducting Ab entities appear to lack most regulatory mechanisms (i.e. post transductional modifications, inactivation, membrane anchoring, stable pore size) important for channel gating, we believe that they do not behave as classical ion channels to allow selective ion permeation. Since these membrane disruptions are important for neuronal toxicity, their blockade would be expected to inhibit synaptotoxicity, neurodegeneration and subsequently AD pro- gression. Furthermore, this membrane permeabilization action of Ab is in agreement with the vesicular depletion recently reported [17].
Interestingly, the actions of Ab show strong similarities, although to a lesser extent, to the effect of a-latrotoxin (LTX) on neurotransmission [35]. For example, after a strong enhancement of synaptic transmission, LTX induced vesicle depletion and diminution in miniature potentials by a pore forming mechanism, having conductance and kinetic properties very similar to those of pores formed by Ab in lipid bilayers [14].
In summary, our working model to explain the toxicity of Ab in Alzheimer's disease proposes the existence of diverse membrane structures that can progress from a small, ion selective pore, to a large membrane perforation (Fig. 6). All these Ab perforations are capable of producing a wide range of toxic effects ranging from synaptotoxicity to cell death.

Ethics Statement
All animals were handled in strict accordance with the Animal Welfare Assurance (permit number 2008100A) and all animal work was approved by the appropriate Ethics and Animal Care and Use Committee of the University of Concepcion.

Cultures
Hippocampal neurons were obtained from 18 day pregnant mouse embryos (C57BL/J6) or Sprague-Dawley rat embryos as previously described [36] in accordance with NIH recommendations. Human Embryonic Kidney 293 cells (HEK) were cultivated in D-MEM (Dulbecco's Modified Eagle Medium, Life Technologies, Inc. USA) supplemented with 10% fetal bovine serum (Life Technologies Inc. USA.) and streptomycin-penicillin (200 units each, Life Technologies Inc. USA). Cells were maintained with 5% CO 2 at 37uC. HEK 293 cells were kindly provided by Dr. Olate (University of Concepcion) and have been previously described in the lab [41].

Amyloid Aggregation
Human Ab 1-40 labeled with Rhodamine Green at its Nterminus and unlabeled were purchased from Anaspec (CA, USA) and Tocris (MO, USA), respectively. Ab 1-40 was dissolved in DMSO (10 mg/ml) and stored in aliquots at 220uC. For the preparation of Ab aggregates (80 mM), aliquots of peptide stock (250 mg in 25 ml of DMSO) were added to 700 ml of PBS (Gibco, USA) and continuously agitated (200 RPM at 37uC) for 90 minutes and stored at 4uC. Ab 1-40 Rhodamine Green (Abs/ Em = 502/527 nm) was dissolved in DMSO (4 mg/ml) and immediately stored in aliquots at 220uC.

Recordings
Patch pipettes having a resistance between 1 and 3 MV were prepared from filament-containing borosilicate micropipettes. Currents were measured with the whole-cell patch-clamp technique at a holding potential of 260 mV using an Axopatch 200B (molecular devices, USA) amplifier as previously described [37,38]. Perforated recordings were obtained as follows: the perforating agent was added into the pipette solution and a 5 mV pulse was used to monitor the formation of the perforation. Gramicidin and amphotericin were used at 100 mg/ml. Short applications of Ab, GABA (100 mM) and AMPA (100 mM) were done via lateral motion of a multi-pipette array (approx. 200 mm in diameter). Some experiments involved an external solution without added calcium, Na7 (20-100 mM), Na4a (20 mM) or the inactive peptides Na13 and Na15 (20 mM).

Western Blots
Standard Western blotting procedures were followed. Equal amounts of protein were separated on 10% SDS-PAGE gels. Protein bands were transferred onto nitrocellulose membranes, blocked with 5% milk and incubated with a primary antibody using the following concentrations: anti-Ab (NAB228, Santa Cruz Biotechnology, CA, USA) 1:500, anti-Synapsin I (AB1543, Chemicon, MA, USA) 1:1000, anti-SV2 (Developmental Studies Hybridoma Bank, IA, USA) 1:200. Immunoreactive bands were Figure 6. The scheme is a simplified model for association, micro and macro perforation induced by Ab in cellular membranes. A, aggregation and binding (association) of Ab to the neuronal membrane B, smaller perforations are associated to a selective ion influx (gramicin-like ion influx). C, larger perforations allow the entry of large molecules, which include EtBr (,1.3 nm). All these Ab effects are blocked by application of anti-Ab antibody. doi:10.1371/journal.pone.0011820.g006 visualized with ECL plus Western Blotting Detection System (PerkinElmer, MA, USA).

Intracellular Calcium Imaging
Neurons were loaded with Fluo-4 AM (1 mM in pluronic acid/ DMSO, Molecular Probes, Eugene, OR, USA) for 30 min at 37uC. The neurons were then washed twice with external solution and incubated for 30 min at 37uC. The cells were mounted in a perfusion chamber that was placed on the stage of an inverted fluorescent microscope (Eclipse TE, Nikon, USA). The cells were briefly (200 ms) illuminated using a computer-controlled Lambda 10-2 filter wheel (Sutter Instruments, USA). Regions of interest (ROI) were marked in a field having usually more than 10 cells. Images were collected at 2-5 s intervals during a continuous 5-min period. The imaging was carried out with a SensiCam camera (PCO, Germany) using Axon Instruments Workbench 2.2 software. The calcium channel inhibitors used were conotoxin (1 mM), agatoxin (1 mM), nifedipine (3 mM), CNQX (4 mM) and D-AP5 (50 mM).

FM1-43 Loading and Unloading
Presynaptic vesicles were labeled by exposure to FM1-43 (15 mM, Molecular Probes, USA) during a high-K + depolarization for 5 min and immediately washed, as previously described [39,40]. Coverslips were mounted on a rapid switching flow perfusion chamber with an inverted fluorescent microscope (Eclipse TE, Nikon, USA) equipped with a 1006 objective (oil immersion, NA 1.4). Depolarization-dependent destaining was induced by bath perfusion with 30 mM K + (equiosmolar replacement of Na + ).

Immunocytochemistry
Hippocampal neurons treated during 15 minutes with 500 nM fluorescent Ab were fixed for 15 min with 4% paraformaldehyde and permeabilized with 0.1% triton X-100 in PBS and incubated with anti-MAP2 1:300 (Santa Cruz Biotechnology, CA, USA). Secondary anti-rabbit IgG (Jackson ImmunoResearch Laboratories, PA) conjugated with Cy3 was used at 1:500 for 2 hours.

Calculation of Ethidium Bromide Size
The van der Waals diameter of EtBr was measured with Swiss PDBviewer using atomic coordinates for the crystal structure of the ethidium-bound form of the multi-drug binding transcriptional repressor CgmR (PDB ID: 2ZOZ).

Data Analysis
Non-lineal analysis was performed using Origin (Microcal). Membrane charge was analyzed by integrating the transient capacitative current after subtracting the pipette capacitance. The values are expressed as mean 6 SEM (standard error mean). Statistical differences were determined using Student's t test or ANOVA. The experiments were performed in triplicate. Figure S1 Blockade of Ab induced membrane disruption by small peptides. A, sequence of Ab and mini peptides used in this study (NA7, NA4a, NA13 and NA15). B-C, shows the effect of Ab (500 nM) and Ab plus mini peptides (20 mM) on the transferred membrane charge and resistance, respectively. The bars are means 6SEM. * denotes a P,0.05. Found at: doi:10.1371/journal.pone.0011820.s001 (0.80 MB TIF) Figure S2 Perforating actions of Ab were associated to the presence of fibril-like structures. A, the upper electron micrograph shows active structures labeled with 5 nm gold-particles. B, the current trace show that these structures caused membrane perforations in rat hippocampal neurons. Lower panels show a more globular Ab structure that was found to be inactive. Data is typical from 6 experiments.