NMDA-Receptor Activation but Not Ion Flux Is Required for Amyloid-Beta Induced Synaptic Depression

Alzheimer disease is characterized by a gradual decrease of synaptic function and, ultimately, by neuronal loss. There is considerable evidence supporting the involvement of oligomeric amyloid-beta (Aβ) in the etiology of Alzheimer’s disease. Historically, AD research has mainly focused on the long-term changes caused by Aβ rather than analyzing its immediate effects. Here we show that acute perfusion of hippocampal slice cultures with oligomeric Aβ depresses synaptic transmission within 20 minutes. This depression is dependent on synaptic stimulation and the activation of NMDA-receptors, but not on NMDA-receptor mediated ion flux. It, therefore, appears that Aβ dependent synaptic depression is mediated through a use-dependent metabotropic-like mechanism of the NMDA-receptor, but does not involve NMDA-receptor mediated synaptic transmission, i.e. it is independent of calcium flux through the NMDA-receptor.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative disorder of modern society. It affects the elderly and causes impaired memory formation and loss of higher cognitive functions [1]. Brains of AD patients usually contain two pathological protein agglomerates: extracellular plaques composed of amyloid-beta (Ab) peptides and intracellular neurofibrillary tangles composed of hyper-phosphorylated tau protein.
Through studies conducted on human patients and animal AD models, it is now widely accepted that Ab is implicated in the cause of AD [2]. Elevated levels of this peptide disrupt glutamatergic synapses in the hippocampus and cortex [3][4][5]. Neurons exposed to Ab have impaired long-term potentiation [6,7] and increased depression of synaptic transmission, most likely by activating pathways involved in synaptic long-term depression [8,9]. Interestingly, deficits in synaptic transmission appear before the formation of amyloid plaques in AD transgenic mouse models [10]. Furthermore, the magnitude of memory and learning disabilities in AD patients does not correlate well with the amyloid plaque burden [11]. These findings suggest that plaque formation may not be the primary cause for synaptic failure in early AD. Consequently, over the last years oligomeric Ab has been identified as one of the main factors for the changes found in synaptic transmission in AD [12,13].
Most experimental models currently used in AD research chronically incubate neuronal cultures with Ab, by either exposing neurons to oligomeric Ab for several hours or by using animal models overexpressing Ab. This approach, although generally accepted for mimicking the conditions occurring during AD, makes it challenging to analyze the acute effects of Ab on synaptic transmission. Given that neurons can regulate synaptic strength with mechanisms such as homeostatic plasticity, it becomes difficult to dissect primary from secondary effects on synaptic transmission by Ab, i.e. the neuron can counter-regulate the synaptic insult caused by Ab. To address the immediate effects of Ab on synaptic transmission, we used an acute perfusion protocol of Ab oligomers in hippocampal slice cultures. Our results show that acutely applied Ab oligomers promote synaptic depression in hippocampal CA1 pyramidal neurons within 15 min. This depression is mediated by glutamate dependent activation of NMDA receptors but appears to be independent of NMDAreceptor mediated ion flux. In summary, we propose that Ab oligomers promote synaptic depression through NMDA-receptors acting in a metabotropic rather than ionotropic fashion.

Acute Ab Perfusion Leads to NMDA-receptor Dependent Synaptic Depression
Previous studies in which we participated have shown that the overexpression of the amyloid precursor protein (APP) in hippocampal neurons leads to synaptic depression in organotypic hippocampal slice cultures [14]. This depression was dependent on the activation of NMDA-receptors and further analysis revealed that APP expression activates a LTD-like intracellular signaling cascade. Given that the overexpression of APP leads to a chronic exposure of neurons to Ab, we wanted to analyze this depression in a more acute and more controlled experimental setup.
We therefore acutely perfused oligomeric Ab (0.1 mM, Fig. 1a) and analyzed its effects on evoked field potentials. After having established a stable baseline recording for 20 min, we perfused the slice cultures with oligomeric Ab for 20 min. As seen in Figure 1 b and c, Ab oligomers induced a significant synaptic depression within 10 min of exposure. We did not observe a difference in the fiber volley amplitude upon Ab perfusion ( Fig. 1 b and c), indicating that the acute Ab-dependent synaptic depression does not act presynaptically. To further analyze the synaptic mechanisms of acute Ab induced synaptic depression, we recorded evoked basal synaptic transmission in CA1 neurons in whole-cell patch clamp configuration (260 mV holding potential) by stimulating the Schaffer collaterals. After 3 min of baseline recording, we perfused slices with oligomeric Ab. As seen in Figure j, we observed a statistically significant synaptic depression within 10-15 minutes after Ab application. This depression was specific for Ab oligomers, since the scrambled form of Ab (randomized Ab amino acid sequence peptide), the monomeric Ab as well as vehicle alone had no effect on basal synaptic transmission ( Fig. 2a and 2b).
To further characterize the basis of Ab induced changes in synaptic transmission, we recorded synaptic miniature events in the presence of Ab. We perfused slice cultures for 20 min with Ab or vehicle, added TTX for 10 min and recorded miniature events (Fig. 3a). As seen in Figure 3b, we observed no changes in miniature amplitudes or frequency in the presence of Ab and TTX. The unchanged miniature frequency is consistent with our data showing that Ab does not affect fiber-volley amplitudes ( Fig. 1b and c) or paired-pulse facilitation (Figure 3c and d), indicating that the presynaptic site is not involved in acute Ab dependent synaptic depression. However, the lack of change in miniature amplitudes while having previously observed synaptic depression in field recordings (Fig. 1b) as well as whole-cell patchclamp recordings (Fig. 2a), suggested that Ab mediated synaptic depression is dependent on synaptic stimulation. To test this likely due to Ab already being present at the synapse when stimulation is resumed, i.e. slices are already perfused for 10 min with Ab). Taken together, we conclude that Ab can rapidly induce synaptic depression, but only in the presence of sufficient synaptic activity.
Given that synaptic activity was necessary for the acute effect of Ab on synaptic depression, we wondered whether this depression would be sensitive to NMDA-receptor (NMDA-R) activation as it was shown with chronic APP expression [14]. We therefore added the competitive NMDA-R specific antagonist APV to the bath solution 10 min prior to recording. Interestingly, APV completely blocked acute Ab dependent synaptic depression (Fig. 4). This result was surprising, since the NMDA-R is known to be blocked by magnesium ions when the cell is at resting potential or, as in our case, clamped at 260 mV resting potential. To further analyze which of the two NMDA-R subunits (GluN2A or GluN2B) would be responsible for Ab dependent synaptic depression, we repeated the experiment in the presence of Ifenprodil, a GluN2B specific inhibitor. As shown in Figure 4 Ifenprodil, like APV, blocked Ab dependent synaptic depression. Taken together, we could show that the acute perfusion of oligomeric Ab leads to an activity dependent synaptic depression involving GluN2B containing NMDA-Rs, even if the cell is clamped at resting potential and therefore preventing ion influx through NMDA-receptors.

Acute Ab Dependent Synaptic Depression does not Require NMDA-receptor Dependent Ion Flux
Given that Ab induces NMDA-R dependent synaptic depression at a holding potential of 260 mV, we speculated that oligomeric Ab could act by loosening the magnesium block that normally prevents NMDA-R dependent transmission at resting potential, i.e. Ab would increase NMDA-R dependent synaptic currents at resting potential.
To analyze changes in NMDA-R dependent synaptic transmission, we recorded isolated NMDA-R responses by employing the AMPA/Kainate-receptor inhibitor CNQX. As seen in Figure 5a, CNQX completely blocks AMPA-receptor dependent transmission. After AMPA-receptor dependent transmission was blocked, we perfused oligomeric Ab for 15 min and recorded evoked synaptic responses. However, we did not observe any increase in NMDA-R mediated synaptic transmission, as we would have expected if Ab could increase ion leakage through the NMDAreceptor.
Given this result, we wondered whether small local depolarization in dendritic spines at the moment of AMPA-receptor activation would be sufficient to affect NMDA-receptor dependent synaptic transmission in the presence of Ab, i.e. Ab would weaken the magnesium block only under depolarized conditions. Since the sensitivity of our experimental setup in Figure 5a would be insufficient to detect these changes, we decided to compare evoked NMDA-R currents at different holding potentials with and without oligomeric Ab. We first recorded synaptic AMPA-receptor dependent transmission without exposure to Ab, followed by the perfusion of Ab for 15 minutes and the addition of the AMPA/ Kainate-receptor inhibitor CNQX thereafter. With AMPAreceptors blocked, the remaining NMDA-receptor dependent synaptic transmission was recorded at different holding potentials (260 to +20 mV in 10 mV increments). Each of the recorded NMDA-receptor responses was then normalized to the correspondent AMPA-receptor dependent response obtained during the first 3 minutes of recording without Ab or CNQX. As seen in Figure 5b, Ab does not change NMDA-R dependent synaptic transmission at different holding potentials, indicating that Ab does not act by loosening the magnesium block in NMDA-Rs. Furthermore, we did not observe a change in NMDA-R dependent transmission when the neuron was clamped at +10 mV or +20 mV, indicating that Ab does not change NMDA-R dependent synaptic transmission in the absence of magnesium block. Taken together, we concluded that oligomeric Ab does not affect the ion flux via the opening of NMDA-Rs.
So far, it appears that synaptic activity (Fig. 3e) and GluN2B containing NMDA-R activation (Fig. 4) are necessary for the effect of Ab on synaptic depression, but that Ab is not changing the ion flux through the NMDA-Rs. Furthermore, ion flux through the NMDA-receptor is virtually absent at 260 mV resting potential (Fig. 5b). These findings led us to hypothesize that Ab in combination with synaptic activity is activating a NMDA-R dependent intracellular signaling pathway even in the absence of ion flux through the NMDA-R, i.e. the NMDA-R would act like a metabotropic receptor. To test this hypothesis, we made use of MK-801, a use dependent NMDA-R ion channel blocker. Different from APV, which interferes with the binding of glutamate to the NMDA-R, MK-801 binds within the pore of NMDA-Rs. We therefore expected no effect on Ab dependent synaptic depression if NMDA-Rs were to act in a metabotropic like fashion. We incubated slice cultures for 12 hours with MK-801 prior to recording the effects of Ab on synaptic transmission. As seen in Fig. 6a and b, Ab was still able to induce synaptic depression in presence of MK-801. To exclude that 12 hours of MK-801 incubation would be insufficient to block NMDAreceptor dependent synaptic transmission, we recorded in a separate experiment evoked synaptic currents at 260 mV and +40 mV holding potential after MK-801 incubation. As shown in Fig. 6c, at positive holding potentials, neurons treated with MK-801 show a complete block of NMDA-R responses at +40 mV compared to untreated controls (the remaining response is composed of the AMPA-R response). In summary, we could show that acute Ab exposure leads in the presence of MK-801 to the induction of synaptic depression. This depression engages a use-dependent but ion-flux independent activation of GluN2B containing NMDA-Rs. Our results therefore point toward a new metabotropic-like mechanism in NMDA-R dependent signal transduction.

Discussion
Amyloid-beta is considered to be one of the most prominent factors in the development of Alzheimer's disease. Transgenic AD animal models and in-vitro models with chronic over-expression of Ab have broadened our understanding of the physiological changes occurring during Alzheimer disease. However, in all of these models, it is hard to distinguish whether the observed effects are directly caused by Ab or whether they are reflecting the neurons counter-response to the primary Ab insult. To shed light on this question, we acutely perfused hippocampal slice cultures with oligomeric Ab. Within 20 minutes of Ab perfusion we observed a NMDA-R dependent synaptic depression. To our surprise, the involved NMDA-receptors act in this scenario like metabotropic receptors, i.e. an NMDA-receptor mediated ion-flux is not necessary for Ab induced synaptic depression.

The NMDA-receptor Acting as a Metabotropic Receptor
Our results suggest that NMDA-receptors, in the presence of Ab, can act as metabotropic receptors; thus, upon ligand binding, these receptors can activate intracellular signaling cascades in the absence of ion flux. While this concept of NMDA-receptor activity appears to be rather new, it is however not unprecedented. Several laboratories described intracellular effects of glutamate binding on NMDA-receptors in the absence of ion flux. In 2001, the group of Gary Westbrook showed that the Tyrosine residues in GluN1/ GluN2A containing NMDA-receptors are dephosphorylated in a use-dependent fashion without involving ion flux [15]. This dephosphorylation would ultimately lead to an endocytosis of synaptic NMDA-receptors. The group of Michael Salter found that extracellular glycine binding to the NMDA-receptor primes the receptor for endocytosis by recruiting intracellular AP2 to the receptor complex. Again, the effects were independent of ion flux through the NMDA-receptor [16]. The group of Roberto Malinow has shown that the exchange of GluN2B by GluN2A in the CA1-CA3 hippocampal synapse is driven by spontaneously released glutamate, even in the absence of synaptic NMDAreceptor dependent currents [17], indicating that the NMDAreceptor might act in a metabotropic fashion. Finally, a more recent publication by the group of John Wang described the activation of extracellular signal-regulated protein kinase (ERK) through the co-activation of metabotropic glutamate receptor 5 (mGluR5) and NMDA-receptors [18]. Interestingly, the authors were able to block ERK activation by applying the NMDAreceptor antagonist APV (APV prevents glutamate binding and thereby activation of NMDA-receptors), while the non-competitive NMDA-receptor channel blocker MK-801, had no effect on ERK activation.
Our results furthermore, shed new light on the AD medication memantine. Memantine, like MK-801, is an open channel blocker targeting the ion pore of NMDA-receptors and thereby reducing the ion flux [19]. Given that oligomeric Ab would activate the NMDA-receptor in a metabotropic-like fashion, memantine should have no or very limited effects on Ab induced synaptic depression. Our results could therefore contribute in explaining why memantine was ineffective in treating early stages of AD [20].

Ab Induced Synaptic Depression
Our data support the notion that while Ab induced LTD-like synaptic depression is dependent on NMDA-receptor activation (i.e. ligand binding; Fig. 4), it appears to be independent of calcium flux through the NMDA-receptor (Fig. 6). At this moment, we can only speculate how the NMDA-receptor, acting in a metabotropic fashion, might induce synaptic depression.
Recently, several groups have shown that Ab induced synaptic depression by facilitating NMDA-receptor dependent LTD or mGluR dependent LTD [9,21]. Furthermore, blocking the Figure 5. Ab has no effect on NMDA-receptor mediated ion conductance. (A) Normalized EPSC amplitudes plotted against time in presence of CNQX and Ab. Neurons were patched and held at 260 mV and responses were evoked throughout the duration of the experiments. Following baseline recording, CNQX was applied to the bath to block AMPA-receptor dependent currents (arrow on the left). Following the sharp decrease in the EPSC amplitude, Ab oligomers were applied (arrow on the right). Averaged traces taken during baseline recording (min 2), after CNQX application (min 12) and after Ab application (min 25) are shown above the graph. There is no evident difference between currents recorded immediately before Ab application and currents recorded 10 minutes following Ab application (n = 6). Color legend: black: currents recorded before CNQX application; red: currents recorded before Ab application; blue: currents recorded after Ab application. Scale bars: horizontal 0.1 s; vertical 10 pA. (B) Normalized NMDA-receptor I/V curve recorded in presence or absence of Ab oligomers. Evoked synaptic responses of CA1 pyramidal neurons were recorded at 260 mV. Next, CNQX was bath applied to block AMPA-receptor currents. Following AMPA-receptor blockade, either vehicle or Ab was bath applied. 15 min after Ab application, evoked EPSCs (now dependent on NMDA-receptor responses) were recorded at different holding potentials (steps of 10 mV increments from 260 to +20 mV holding potential). Calculating the ratio of NMDA-receptor responses vs. the initial AMPA-receptor response was used to normalize the NMDA-receptor currents from different experiments. No noticeable difference in NMDA-receptor responses was found between neurons treated with vehicle vs. Ab oligomers (n = 5 per group). doi:10.1371/journal.pone.0065350.g005 NMDA-receptor with the competitive antagonist APV, also blocks Ab dependent induction of synaptic depression [14,22], while the chronic expression of the amyloid-precursor protein was also able to occlude mGluR dependent LTD induction [14]. All of these results suggest that Ab can mimic LTD-like pathways, leading to the chronic induction of synaptic depression. In general, the literature suggests that NMDA-receptor dependent LTD and mGluR-dependent LTD are largely based on different cellular mechanisms and induction pathways [23]. However, it is also known that mGluR5 is able to modulate the outcome of NMDAreceptor mediated LTD (e.g. [24]). Ab can act as an extracellular scaffold for mGluR5, leading to an increased clustering of mGluR5 in synapses accompanied by an increased concentration of intracellular calcium [24]. Hence, it is possible that Ab dependent synaptic depression involves the co-activation of NMDA-receptors and mGluR5. Of interest is hereby the aforementioned study by Yang et al., which showed that the activation of mGluR5 and NMDA-receptors in the absence of NMDA-receptor dependent ion flux is able to induce ERK5 activation [18,25]. ERK activation on the other hand is necessary for the induction of mGluR dependent LTD [25]. Beside mGluR5, Ab was described to bind the alpha7 nicotinic acetylcholine receptor [26], a receptor that was later on shown to activate the striatal-enriched tyrosine phosphatase (STEP) [27]. Again, STEP appears to be necessary for the induction of mGluRdependent LTD. Through its effect on NMDA-receptor and AMPA-receptor dephosphorylation and subsequent receptor internalization, STEP was described as a more general modulator of synaptic plasticity [28]. Finally, Ab was described to bind EphB2, a receptor that directly interacts and modulates NMDAreceptor surface expression and activity [29]. However, the involvement of EphB2 in synaptic plasticity remains to be characterized. Taken together, Ab is able to interact with several receptors, some of them, like mGluR5 and alpha7 nicotinic receptor, are directly involved in regulating synaptic plasticity. Which of these receptors together with the NMDA-receptor is responsible for the effects of Ab on synaptic depression remains a topic for future investigation.

Ab Preparation
Ab 1-42 (Bachem) was prepared according to Barghorn et al [30]. Oligomeric Ab was prepared fresh every three days and stored at 4uC when not used. As control, we prepared the same solution but with scrambled Ab (Covance) or without any peptide at all. We also performed an additional control, employing nonoligomerized monomeric Ab. To analyze the formation of oligomeric Ab, 100 ng of Ab were loaded on a 10-20% discontinuous gradient Tris-Tricine gel. Western blot analysis was performed using the 6E10 anti-Ab antibody (Millipore).

Slice Preparation
Hippocampal slices were prepared from p7 rat pups as previously described in [31,32] and maintained in culture for 5 to 7 days. For electrophysiological recording, slices were transferred to a submerged recording chamber maintained at 32uC.
EPSCs were evoked by stimulating the Schaffer collaterals at 0.2 Hz through two bipolar tungsten electrodes, connected to a stimulus isolation unit (A.M.P.I.). The electrodes were positioned ,150 and ,250 mm away from the recording site. This setting allowed us to stimulate two independent synaptic inputs; results from both pathways were then averaged and counted as n = 1. Given that slices were only used once for recordings, n reflects the number of neurons as well as the number of slices (usually from 4-6 different animals per experiment). All recordings were performed at 32uC. Baseline recordings lasted for 3 to 5 minutes, following perfusion of either Ab or the control vehicle. After application of drugs or Ab, neurons were usually recorded for a minimum of 20 minutes and the results compared to the baseline recordings.
For the I/V profile, neurons were patched and held at 260 mV and a baseline of 3 minutes was recorded. Following baseline recording, CNQX was then bath applied for 5 min in order to block AMPA receptor dependent transmission (AMPA receptor dependent currents were no longer observed). Next, either Ab or the control vehicle was bath applied for 20 minutes. Neurons were then stepwise depolarized from 260 mV to +20 mV in order to determine the amplitude of NMDA receptor responses recorded at each holding potential. We recorded a minimum of 2 min per stepping potential. Responses from Ab-treated and untreated neurons were analyzed and displayed as the ratio of the NMDA response at a specific holding potential normalized by the corresponding AMPA receptor response recorded before at 260 mV.
For the recordings of miniature events, either Ab or vehicle was bath applied prior to the beginning of the experiment. As above, 20 minutes were allowed for Ab or vehicle to take effect, followed by TTX (1 mM) bath application. 5 minutes after drug application, neurons were patched and miniature events recorded for a minimum of 10 minutes. Miniature events were then analyzed offline employing the Clampfit 10.2 software.

Drugs
Drugs were usually applied to the bath solution prior to the beginning of the experiment (unless otherwise specified). We employed the AMPA-receptor antagonist CNQX (Tocris, 20 mM), the NMDA-receptor specific antagonists AP-V (TOCRIS, 100 mM) and Ifenprodil (Abcam Biochemicals, 10 mM). Also, in some experiments we used the NMDA-receptor blocker MK-801 (Tocris, 25 mM). MK-801 was added to the cultured slices the night before and MK-801 was present at the same concentration during the recording. For miniature events analysis, experiments were conducted in presence of the Na + channels blocker TTX (Alomone labs, 1 mM).

Statistical Analysis
Results are displayed as bar diagrams or scatter graphs, representing the experimental mean; the error bars represent the standard error. For statistical analysis we used the non-parametric Wilcoxon-Mann-Whitney test.
Note added in proof: While this manuscript was under review, the group of Roberto Malinow employing a different experimental approach reported comparable results with similar conclusions [33].

Author Contributions
Conceived and designed the experiments: AT AD SL JB. Performed the experiments: AT AD SL CB. Analyzed the data: AT AD SL JB. Wrote the paper: AT JB.