Calcium flux-independent NMDA receptor activity is required for Aβ oligomer-induced synaptic loss

Synaptic loss is one of the major features of Alzheimer's disease (AD) and correlates with the degree of dementia. N-methyl-d-aspartate receptors (NMDARs) have been shown to mediate downstream effects of the β-amyloid peptide (Aβ) in AD models. NMDARs can trigger intracellular cascades via Ca2+ entry, however, also Ca2+-independent (metabotropic) functions of NMDARs have been described. We aimed to determine whether ionotropic or metabotropic NMDAR signaling is required for the induction of synaptic loss by Aβ. We show that endogenous Aβ as well as exogenously added synthetic Aβ oligomers induced dendritic spine loss and reductions in pre- and postsynaptic protein levels in hippocampal slice cultures. Synaptic alterations were mitigated by blocking glutamate binding to NMDARs using NMDAR antagonist APV, but not by preventing ion flux with Ca2+ chelator BAPTA or open-channel blockers MK-801 or memantine. Aβ increased the activity of p38 MAPK, a kinase involved in long-term depression and inhibition of p38 MAPK abolished the loss of dendritic spines. Aβ-induced increase of p38 MAPK activity was prevented by APV but not by BAPTA, MK-801 or memantine treatment highlighting the role of glutamate binding to NMDARs but not Ca2+ flux for synaptic degeneration by Aβ. We further show that treatment with the G protein inhibitor pertussis toxin (PTX) did not prevent dendritic spine loss in the presence of Aβ oligomers. Our data suggest that Aβ induces the activation of p38 MAPK and subsequent synaptic loss through Ca2+ flux- and G protein-independent mechanisms.

Alzheimer's disease (AD) is clinically characterized by cognitive impairments caused by massive neuronal degeneration and synaptic loss. The reduction in synapse numbers is the best neuropathological correlate to the degree of dementia in AD. 1 Besides synaptic alterations, the levels of soluble oligomeric forms of β-amyloid peptide (Aβ) but not plaques correlates best with memory loss in AD. 2 Accumulating evidence indicates that transgenically produced Aβ or the treatment with Aβ oligomers decrease dendritic spine density, [3][4][5][6] impair long-term potentiation (LTP), 7 facilitate long-term depression (LTD) 8 and induce aberrant spine morphology. 5,9 Although the signaling cascades coupling Aβ with synaptic degeneration are incompletely understood, experimental evidence suggests an essential role for N-methyl-D-aspartate receptors (NMDARs). Oligomeric Aβ can bind to dendritic spines and treatment with NMDAR antibodies abolishes Aβ binding. 10 Pharmacological inhibition of NMDAR activity also mitigates the pathological effect of Aβ on synapses. [4][5][6]11 NMDARs are ionotropic receptors permeable for cations and controlled by a voltage-dependent Mg 2+ block that is removed after membrane depolarization by α-amino-3-hydroxy-5methyl-4-isoxazolepropionic acid receptors (AMPARs). Upon glutamate binding to the NR2 subunit of NMDARs, cations including Ca 2+ enter the cell. It has been thought for a long time that the levels of Ca 2+ influx through NMDARs determine the induction of either LTP (high Ca 2+ influx or LTD (mild Ca 2+ influx). 12 Nevertheless, a recent study showed that Ca 2+ flux is not essential for the induction of NMDAR-LTD, whereas glutamate binding to the receptor is required. 13 NMDAR signaling independently of ion flux has already been proposed to regulate NMDAR phosphorylation and endocytosis. 14,15 Further, the subunit switch between NR2B and NR2A NMDARs is driven by glutamate in the absence of NMDAR currents. 16 However, the role of ion flux for synaptic loss in AD still remains to be elucidated.
We show that the Aβ-induced pre-and postsynaptic loss is mediated by glutamate binding to NMDARs, independent of ion influx.

Results
To determine the mechanisms of synaptic loss by Aβ, we cultured hippocampal slices from arcAβ-transgenic mice and infected them with neurotropic Sindbis virus expressing EGFP to visualize single neurons. Neurons in transgenic slices showed reduced dendritic spine densities. Treatment with the NMDAR antagonist D-APV, which blocks the glutamatebinding sites, completely abolished spine loss (Figures 1a  and b). This is in agreement with previous findings that the glutamate-binding site antagonist CPP rescued spine loss in APP-transgenic cultures. 5 As NMDAR signaling has been reported to be mediated through Ca 2+ influx, we sought to determine the role of Ca 2+ influx for Aβ-induced loss of dendritic spines. To this end, slices were treated with NMDAR open-channel blockers memantine or MK-801 ( Figure 1c) or with Ca 2+ chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid (BAPTA; Figure 1e). Neither memantine nor MK-801 nor BAPTA treatment restored spine density in transgenic cultures (Figures 1d and f). Ca 2+ flux through NMDARs requires the removal of the Mg 2+ block within the receptor pore, which is achieved upon membrane depolarization by AMPARs. To analyze the involvement of AMPARs, we treated cultures with AMPAR antagonist CNQX, but did not observe any effect on spine numbers in transgenic cultures (Figures 1g and h). Concentrations of inhibitors were chosen according to previous reports demonstrating highest degree of specificity and/or therapeutic relevance (APV and MK-801; 17 memantine 18,19 ).
To ensure that memantine, MK-801 and BAPTA preparations at the used concentrations are functional and can block NMDARs and the entry of Ca 2+ , despite having no protective effect, we performed synaptic activation experiments in the presence of these compounds. Synaptic activation induced phosphorylation of extracellular signal regulated kinases (phospho-ERK, p-ERK), which is in agreement with previous studies 6,20 (Figure 1i). The presence of APV, memantine, MK-801 or BAPTA fully prevented ERK activation, confirming the functionality of these compounds (Figure 1j). Of note, memantine blocked synaptic activation although it has been described to preferentially block extrasynaptic over synaptic activity at the used concentration. 21 As shown previously, spine loss in arcAβ-transgenic cultures can be prevented in the presence of anti-Aβ antibodies, 6 confirming that Aβ but not APP or any other cleavage product is responsible for the observed effects on spines. Hence, our data indicate that glutamate binding to NMDARs rather than Ca 2+ flux mediates Aβ-induced dendritic spine loss.
To confirm the morphological spine data, we determined whether Aβ also affects the levels of pre-and postsynaptic proteins and analyzed PSD-95 (postsynaptic) and synaptophysin (presynaptic) levels in lysates of non-transgenic and arcAβ-transgenic cultures (Figure 2a). Compared with controls, the levels of both proteins were strongly reduced in transgenic cultures. Treatment with APV but not with memantine or MK-801 rescued the reduction of protein levels ( Figure 2b). Likewise, treatment with BAPTA did not affect loss of synaptic proteins in transgenic cultures (Figures 2c and d).
Synaptic activity has been shown to increase the production of Aβ and, inversely, preventing synaptic activity can reduce Aβ production. 22 To exclude that the protective effect of APV is simply based on Aβ reduction, we quantitatively measured the levels of Aβ40 in the supernatant of arcAβ-transgenic cultures treated with the respective NMDAR antagonists (Figure 2e) under identical conditions as in the previous experiments. NMDAR antagonists did not significantly alter Aβ levels in the medium of transgenic cultures.
This indicates that, in addition to dendritic spine loss, reductions in pre-and postsynaptic protein levels are caused by NMDAR functions, independent of Ca 2+ flux.
A recent study showed that activation of p38 MAPK is essential for Ca 2+ -independent metabotropic function of NMDARs. 13 We analyzed whether p38 MAPK is also involved in the Aβ effects on synapses and examined the activity of p38 MAPK in lysates from non-transgenic and arcAβ-transgenic cultures (Figure 3a). Increased levels of phosphorylated (active) p38 MAPK (p-p38) were observed in transgenic cultures. After treatment with APV, the levels of p-p38 were reduced to control levels, whereas memantine or MK-801 treatment had no effect (Figure 3b). To ascertain that active p38 MAPK mediates synaptic deficits caused by Aβ, we treated cultures with the p38 MAPK inhibitor SB239063. Treatment abolished spine loss in arcAβ-transgenic cultures (Figures 3c and d). To examine whether a general increase in synaptic activity in transgenic cultures contributes to increased p-p38 MAPK levels, we treated non-transgenic and arcAβ-transgenic cultures with bicuculline and 4-aminopyridine (Figures 3e and f). Synaptic activation increased levels of p-ERK. No difference in p-ERK levels were observed between transgenic and non-transgenic cultures, indicating that arcAβ-transgenic slices display no general increase in synaptic activity. Further, synaptic activation does not affect the activity of p38 MAPK.
This data suggests that Aβ induces the activity of p38 MAPK, which mediates the loss of dendritic spines. This effect does not depend on Ca 2+ influx or general synaptic activation.
Oligomeric Aβ is considered to be one of the main toxic Aβ species in the AD brain. So far, we used cultures from arcAβtransgenic mice to determine the effects of Aβ on synapses in the presence of other APP processing products (Figures 1-3). Although arcAβ mice show early formation oligomeric Aβ in vivo, 23 the role of oligomeric Aβ for our findings requires further investigations. To conclusively validate the role of Aβ oligomers, we treated non-transgenic cultures with defined preparation of Aβ42 oligomers at sublethal concentrations ( Figure 4). Oligomer preparations contained mostly mono-, triand tetramers as determined by silver stained SDS gel and western blot (Figure 4k), which is in agreement with previous studies. 24,25 Scrambled Aβ, subjected to the same oligomerization protocol as Aβ42, did not aggregate, as expected. Treatment with Aβ oligomers but not scrambled Aβ reduced dendritic spine density to a similar extent as observed in transgenic cultures (compare Figure 4 and   (Figures 4i and j). Aβ did not cause cell death at the used concentration (Figure 4l).
This indicates that oligomeric Aβ, similar to transgenically produced Aβ, exerts its toxic properties on synapses via NMDAR signaling, independent of Ca 2+ influx.
To determine whether Ca 2+ flux-independent synaptic loss depends on G protein signaling, we treated slices with oligomeric Aβ and pertussis toxin (PTX), an inhibitor of the heterotrimeric G i/o protein family, at concentrations described before in slice cultures. 26 PTX administration did not prevent spine loss caused by Aβ (Figures 5a and b) suggesting that Aβ-induced synaptic loss does not require a PTX-sensitive subgroup of G proteins.

Discussion
In this study, we have examined the role of Ca 2+ flux for Aβinduced loss of dendritic spines and pre-and postsynaptic proteins. Our data show that NMDAR-dependent ion flux is not required for synaptic loss, whereas binding of glutamate to the  significances indicate differences to the respective non-tg control; for p-ERK/total ERK significances indicate differences to the respective non-activated culture). non-tg, nontransgenic; tg, arcAβ transgenic; Mem, memantine; MK, MK-801; p-ERK, phospho-ERK; p-p38, phospho-p38 MAPK; syn, synaptic activation NMDAR is essential for coupling Aβ with synaptic degeneration.
NMDARs have been thought to signal exclusively ionotropic, regulating intracellular signaling via Ca 2+ transmission. However, recent evidence indicates that NMDARs can signal metabotropically, that is, independent of ion flux. The group of Roberto Malinow showed that induction of NMDAR-LTD via activation of p38 MAPK is based on metabotropic signaling. 13 Further, the induction of LTD by Aβ can occur in the absence of Ca 2+ transmission. 17,27 Together, these data suggest that glutamate binding to NMDARs may induce a conformational change that subsequently activates intracellular signaling cascades even in the absence of Ca 2+ flux. This possibly does not exclude an additional role of intracellular Ca 2+ , because the injection of Ca 2+ chelators into neurons prevents LTD induction. 13,28 In agreement, oligomeric Aβ has been shown to increase intracellular Ca 2+ levels by mobilizing Ca 2+ from the ER rather than promoting influx of Ca 2+ from the extracellular space. 29 Our data show that the induction of specific metabotropiclike NMDAR signaling pathway by Aβ, which is not induced by general synaptic activation, causes downstream phosphorylation of p38 MAPK. Active p38 MAPK is key player in NMDAR-and mGluR-dependent LTD 30,31 and mediates AMPAR endocytosis. 32 A study by Yang et al. 33 described an intracellular pathway based on the co-activation of mGluR5 and NMDARs, also independent of Ca 2+ flux. Further, mGLuR5 has been implicated in mediating toxic effects of Aβ at synapses. 34,35 Thus, a co-activation of mGluR5 and NMDARs may cause downstream activation of p38 MAPK followed by synaptic loss. However, treatment with the G protein inhibitor PTX did not prevent spine loss in our model, which renders the involvement of mGluRs unlikely.
In previous studies, we showed that caspase-3 and calcineurin are essential for the loss of spines by Aβ. 5,6 Caspase-3 can be activated by p38 MAPK. 36,37 Further, D'Amelio and colleagues reported caspase-3-and calcineurin-mediated synaptic dysfunction in APP-transgenic mice. Importantly, they observed that caspase-3-activated calcineurin by proteolytic cleavage in a Ca 2+ -independent manner, supporting our finding that Aβ-induced synaptic dysfunction can occur in the absence of Ca 2+ flux.
An important finding in our study is the lack of synaptic protection by memantine as memantine is the only clinical approved NMDAR antagonist for treatment of AD patients. Memantine, at clinically relevant low μM concentrations, is a low-affinity, uncompetitive open-channel blocker with a relatively high off-rate. 38 High-affinity NMDAR antagonists may be toxic after long exposure due to block of synaptic transmission. However, memantine has been suggested to be more tolerable because of blocking mainly over-excitation of the receptor rather than its physiological activity. As the high offrate of memantine could still allow Ca 2+ influx into the cell, we confirmed the data using a second open-channel blocker MK-801 and the Ca 2+ chelator BAPTA. Further, all used compounds could fully block ERK phosphorylation after synaptic activation. Interestingly, treatment of slices with even the high-affinity inhibitors D-APV or MK-801 or with BAPTA did not show any side effects on spines.
Despite having no protective effect for Aβ-induced synaptic loss in our study, memantine may be more beneficial  with respect to other Aβ effects. A recent study showed that injection of low-molecular-weight (LMW) oligomers into mice caused persistent memory impairment and synaptic loss, whereas injection of high-molecular-weight (HMW) oligomers resulted in neuronal oxidative stress and reversible cognitive deficits but no synaptic loss. Memantine treatment could rescue only the effects of HMW but not LMW Aβ oligomers, 39 further indicating that memantine may not be beneficial with respect to Aβ-induced synaptic alterations. However, memantine protected against the induction of oxidative stress by oligomeric Aβ 10 and studies from our lab showed that memantine, at the concentration used in this study, prevented the increase in tau phosphorylation by Aβ (unpublished observations).
Because synaptic loss occurs early in the disease process, 40 our data my contribute to explain why memantine is ineffective in treating early-staged mild AD patients. 41 Together, our data establish a Ca 2+ flux-and G protein-independent NMDAR signaling pathway coupling Aβ toxicity with p38 MAPK activation and synapse loss, suggesting pharmacological inhibition of this pathway as a potent mechanism to prevent Aβ-mediated early synaptic loss. Hippocampal slice cultures. ArcAβ-transgenic mice were obtained as described. 23 All animal experiments were performed in accordance with the guidelines of the Swiss veterinary cantonal office. Hippocampal slice cultures were prepared and cultured as described. 42 In short, 6-7-day-old transgenic and nontransgenic C57BL/6 mice were decapitated, brains were removed, hippocampi were isolated and cut into 400-μm thick slices. Slices were cultured in culture medium (minimum essential medium Eagle with HEPES modification, 25% basal medium with Earle's modification, 25% heat-inactivated horse serum, 2 mM glutamine, 50 units per ml penicillin, 50 μg/ml streptomycin, 0.6% glucose, pH 7.2). Culture medium was exchanged every second or third day. On DIV 11, culture medium was replaced by low-serum Nb-N2 medium (Neurobasal medium, 0.5% heat-inactivated horse serum, 2 mM glutamine, 50 units per ml penicillin, 50 μg/ml streptomycin, 0.6% glucose, 1 × N2 supplement, pH 7.2) to ensure more defined condition during analysis. For spine analysis, slice cultures were infected with Sindbis virus expressing EGFP on DIV 12 in culture and fixed on DIV 15 with 4% paraformaldehyde/sucrose. For protein analysis, uninfected slices were lysed on DIV 15 in culture.

Treatments.
To determine inhibitor effects in transgenic cultures, slices were treated with respective inhibitors from DIV 11-15. To analyze effects of oligomeric Aβ, slices were treated with Aβ oligomers or scrambled Aβ from DIV 11-15. To assess the effects of inhibitors on cultures exposed to oligomeric Aβ, slices were treated with Aβ oligomers and the respective inhibitor in parallel from  For treatment with PTX, slices were exposed to PTX from DIV 13-15.
Dendritic spine analysis. To determine dendritic spine density, virus solution was diluted to achieve 1-10 infected neurons per slice to allow imaging of single dendritic fragments. Analysis of dendritic spine density was performed using Leica SP2 CLSM equipped with 63 × objective (NA: 1.2) and 488-nm Argon laser. Apical dendritic segments in CA1 stratum radiatum were imaged with size of 30 × 30 μm (512 × 512 pixel, voxel size: 0.05813 × 0.05813 × 0.25 μm). Image stacks were processed to maximum projections, and dendritic spine density was determined using ImageJ.
Spine imaging and counting were performed blinded (without the researcher knowing the mouse genotype or culture treatment).
Synaptic activation protocol. Stimulation of synaptic activity was adapted from Tackenberg et al. 6 Cultures were pretreated with APV, memantine, MK-801 or BAPTA for 12 h before activation. Then, cultures were exposed to neurobasal medium containing 1 mM 4-AP, 25 mM bicuculline and the respective inhibitor for 20 min. Control cultures were treated with neurobasal medium containing identical DMSO concentrations as above but devoid of 4-AP, bicuculline and inhibitors.
Silver staining. Oligomeric Aβ42 preparations were analyzed by SDS-PAGE. The gel was left overnight in fixing solution (40% EtOH, 10% acetic acid), sensitized in 0.017% sodium thiosulfate for 2 min, impregnated in 0.27% silver nitrate solution (including 0.37% formaldehyde) for 30 min and developed in 0.03 M sodium carbonate (supplemented with 0.15% formaldehyde and 0.02% sodium thiosulfate). The reaction was stopped in 3% glacial acid.
Aβ oligomer preparations. Synthetic Aβ42 peptide was obtained from American peptide. Preparation of Aβ42 oligomers (Abeta-derived diffusible ligands) was carried out as previously described. 43 Cold 1,1,1,3,3,3-hexafluro-2-propanol (HFIP) was added to Aβ42 peptide to a final concentration of 1 mM. HFIP was evaporated overnight, peptides dried for 10 min in a speedvac and stored at − 80°C. Peptides were resuspended in DMSO at 5 mM concentrations. Neurobasal medium without phenol red was added to achieve a peptide concentration of 100 μM and incubated for 24 h at 4°C. Higher aggregates, for example, fibrils were removed by centrifugation at 14 000 g for 10 minutes at 4°C and the supernatant was used for experimental procedures. Aβ42 oligomer preparations were analyzed by silver staining and western blot for each experiment.