Aβ induces astrocytic glutamate release, extrasynaptic NMDA receptor activation, and synaptic loss

Inflammatory cells, including microglia and astrocytes, are thought to contribute to damage in AD, in part via glutamate excitotoxicity (2, 12). For example, exposure to oligomeric Aβ or conditioned medium from microglial cultures incubated with Aβ has been reported to decrease glutamate reuptake from astrocytes in brain slices and cultures (13–16), but whether Aβ also induces local release of toxic glutamate levels onto neurons remains unknown. To study this question, we used a FRET-based glutamate sensor system (SuperGluSnFR) (17) to detect the local concentration of glutamate contiguous to astrocytes after exposure to Aβ peptides. To ensure the close apposition of the sensor probe (consisting of GluSnFR-transfected HEK 293 cells) to the astrocytes, the sensor cells were genetically engineered to coexpress neuroligin (18) (Fig. 1A). Unlike prior methods, the FRET GluSnFR technique allows unprecedented spatial and temporal resolution of local glutamate concentration on a subsecond timescale. This resolution was important for comparing extrasynaptic glutamate levels with the rapid electrophysiological effects of Aβ observed in subsequent patch-clamp recordings.

In mixed neuronal/astrocytic or astrocyte cultures, the addition of glutamate itself resulted in an increased normalized FRET ratio, with a standard curve revealing sensitivity in the dynamic range of 1 to ∼100 µM, as reported previously (17), and similar to the glutamate sensitivity of native NMDARs (19). Within seconds of exposure to picomolar concentrations of naturally occurring Aβ prepared from human postmortem AD brain by a method modified from Selkoe and colleagues (1, 15, 20) or to nanomolar concentrations of oligomerized (but not monomeric) synthetic Aβ_1–42, we observed local increases in glutamate. Potentially, both neurons and glia contribute to glutamate release in our mixed cultures, so we also tested the response to Aβ in pure astrocyte cultures. In this case, the change in glutamate concentration was on the order of 30 µM and occurred in ∼40% of the fields of astrocytes examined (Fig. 1 B–D; n = 370 responding cells quantified in 26 experiments). (For details of Aβ preparations, see Materials and Methods and Fig. S1.) Although naturally occurring Aβ yielded robust responses, synthetic Aβ_1–42 yielded significant increases in local glutamate with as little as 325 pM of an oligomerized preparation in both rat and human astrocyte cultures (Fig. S1D). HPLC analysis validated these results, revealing a small but significant rise in glutamate in the medium bathing the astrocytes after Aβ exposure (Fig. S1E). Depleting microglia from the astrocyte cultures using l-leucine methyl ester did not influence the level of glutamate (Fig. S1F), arguing for a direct effect of Aβ_1–42 on astrocytes under our conditions. Additionally, low-micromolar Aβ_25–35 also engendered local increases in glutamate release from astrocytes, as detected by the FRET GluSnFR technique (Fig. S1G). Control experiments showed that the transfected HEK cells did not respond to Aβ in the absence of astrocytes.

Because some forms of astroglial glutamate release are Ca²⁺ dependent (21), we asked whether oligomeric Aβ_1–42-induced glutamate release from astrocytes was dependent on Ca²⁺ influx. Indeed, when Ca²⁺ was omitted from the extracellular medium, Aβ failed to induce glutamate release, whereas the addition of Ca²⁺ immediately restored glutamate release (Fig. 1E). Absence of Ca²⁺ did not affect the sensitivity of glutamate sensor cells, because responses to exogenously applied glutamate were present in nominally Ca²⁺-free solutions.

Aβ_1–42 binds with high affinity to the α7 nicotinic acetylcholine receptor (α7nAChR) (22), a ligand-gated ion channel with high Ca²⁺ permeability that has been implicated in the pathology of AD (23). Because activation of α7nAChRs can increase intracellular Ca²⁺ ([Ca²⁺]_i) in astrocytes (24), and glutamate release was calcium dependent, we next asked if α-bungarotoxin (α-Bgtx), a highly selective α7-antagonist, could inhibit Aβ-induced glutamate release from astrocytes. When oligomerized Aβ was applied to mouse astrocytes in the presence of 100 nM α-Bgtx, glutamate release was almost totally abrogated (Fig. 1F). Moreover, astrocytes obtained from α7nAChR-knockout mice released very little glutamate in response to Aβ, as monitored with the FRET GluSnFR probe (Fig. 1G). The decrease in Aβ-induced glutamate release was mirrored by a reduction in [Ca²⁺]_i in α7nAChR-knockout astrocytes as compared with WT (Fig. 1H). These findings support the notion that in addition to the reported inhibition of glutamate reuptake, Aβ induces release of glutamate from astrocytes, mediated at least in part by α7nAChRs. Aβ has been shown to bind to group I metabotropic glutamate receptors (25), but antagonists to these receptors manifested little or no effect on oligomeric Aβ-induced glutamate release in our system (Fig. S1H).

A caveat to these findings lies in the fact that astrocytic receptors and their pharmacological properties can change in culture. Hence, to vet our results showing extracellular glutamate accumulation in response to Aβ and its pharmacological properties in more intact systems, we performed experiments in vivo in animal models of AD during microdialysis. Using transgenic mice expressing human amyloid precursor protein (hAPP tg), we found basal glutamate levels were increased compared with nontransgenic littermates and that this increase was largely abrogated and was not statistically different from WT after crossing with α7nAChR-null mice (Fig. 1I and Fig. S1I). Our results show that extracellular glutamate accumulates not only in culture but also in living brain in the presence of oligomerized Aβ in an α7nAChR-dependent manner (23). As a corollary, under our culture conditions as well as in vivo, the high density of transporters adjacent to synaptic sites would be expected to clear excessive glutamate released by Aβ away from synaptic receptors, although extrasynaptic receptors still might be activated (but see ref. 26). Hence, this premise was queried next.

Hippocampal microcultures contain a few or even a single neuron that synapses on itself to form an autapse; neurons are cultured in isolation or on a tiny bed of astrocytes (Fig. S2 A and B). This preparation allows rapid access of exogenous Aβ and applied drugs to neural tissue and also simultaneous recording of eNMDARs and sNMDARs with a single patch electrode. In hippocampal autaptic cultures, we found that picomolar naturally occurring Aβ-soluble oligomers, nanomolar oligomeric Aβ_1–42 (but not monomeric), or low-micromolar Aβ_25–35 (but not Aβ_35–25) induced a tonic inward current within tens of seconds of application in ∼55% of neurons (Fig. 2 A–G; n = 90). In excitatory neurons in the nominal absence of extracellular Mg²⁺, the inward current was inhibited by the NMDAR antagonists (d-)-2-amino-5-phosphonovalerate (APV) or memantine (Fig. 2F) or by the combination of an NMDAR antagonist and an AMPA receptor (AMPAR) antagonist (Fig. 2 A–C). Prior work had shown that the tonic inward current in this preparation represents activation of extrasynaptic glutamate receptors (3, 5, 27, 28). To confirm this finding, we performed recordings after pharmacological isolation of extrasynaptic currents using the published protocol of first activating excitatory synaptic currents electrically followed by the addition of dizocilpine (MK-801) to block these synaptic responses (29) (Fig. S2 C and D). After isolating extrasynaptic currents in this manner, we found that Aβ_1–42 (containing 250-nM oligomers) resulted in increased eNMDAR activity similar to that seen with the application of low-micromolar glutamate (Fig. S2E).

When we studied autapses formed by inhibitory neurons, which release GABA rather than glutamate (Fig. S2F), we also observed a tonic inward current engendered by oligomerized Aβ_1–42 that was antagonized at least partially by 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f]quinoxaline (NBQX) plus memantine (Fig. 2D). This finding is consistent with the notion that glutamate was being released predominantly by astrocytes rather than neurons, because these inhibitory neurons do not release glutamate. If so, then the addition of α-Bgtx to these neuronal/astrocyte cultures should inhibit oligomerized Aβ_1–42-induced currents by inhibiting glutamate release from the astrocytes via blocking α7nAChRs. To test this premise in a manner that was pathophysiologically relevant to human AD, we used neurons and astrocytes derived from human induced pluripotent stem cells (hiPSCs). Consistent with the notion that astrocytic glutamate release is mediated by α7nAChRs, we found in this preparation that NMDAR antagonists inhibited oligomerized Aβ_1–42-induced currents, as did 100 nM α-Bgtx (Fig. S2 G and H).

Another important consideration is that memantine, like other NMDAR open-channel blockers, might block α7nAChR channels (30) and hence prevent release of glutamate from the astrocytes. We tested this possibility using our FRET probe and indeed found memantine produced a small decrement in glutamate release from astrocytes, but the effect did not reach statistical significance (Fig. S1J). Hence, this action could not account for the memantine effect, because substantial glutamate release remained. Nonetheless, this secondary effect of memantine on α7nAChRs potentially might contribute to the drug’s ability to limit eNMDAR currents, because, at least in theory, glutamate release from astrocytes might be inhibited to a degree, in addition to memantine’s direct blocking of eNMDAR-operated channels.

Importantly, in neurons in which synaptic currents in addition to extrasynaptic responses were monitored quantitatively after the addition of TTX, we found a decrement in mEPSC frequency and a smaller or no decline in mEPSC amplitude within minutes of Aβ exposure (Fig. 2 E and H–M). The significant decrease in mEPSC frequency suggested a presynaptic deficit, but functional loss of synapses in response to oligomeric Aβ under these conditions was possible also. We knew these autaptic cells were well clamped, because the miniature synaptic currents reversed at or near 0 mV, as expected for excitatory cation-mediated responses under our conditions. Interestingly, in neurons that did not manifest any extrasynaptic glutamatergic current in response to Aβ, we did not observe a subsequent decrease in mEPSC frequency. Furthermore, depletion of astrocytes from the cultures largely abrogated these effects of oligomeric Aβ (Fig. S2I), even though excitatory neurons have been shown to release synaptic glutamate in response to Aβ (31). The fact that the Aβ-induced current was greatly abated in the absence of astrocytes also was consistent with the notion that the predominant effect on extrasynaptic current observed here did not result from direct action of Aβ on neuronal NMDARs. Although these experiments do not rule out a direct effect of Aβ on neurons, they do indicate that the major effects observed under our conditions were dependent on the presence of astrocytes.

Additionally, in the absence of TTX we observed that cells responding to Aβ with an inward extrasynaptic current manifested subsequent evoked EPSCs with smaller AMPAR- and NMDAR-mediated components than in control (Fig. S2J). This result might reflect the development of silent synapses, endocytosis of AMPARs, or sNMDAR depletion resulting from EphB2 binding of exogenous Aβ, as previously reported (10, 32, 33). However, coupled with the very significant decrease in mEPSC frequency that we observed in the hippocampal autaptic preparation, our results also are consistent with the notion of rapid compromise or functional loss of the synapse after Aβ exposure. Hence, we further investigated this possibility next.

Using Fura-2, we performed Ca²⁺ imaging experiments in mixed neuronal/astrocytic cultures after exposure to Aβ. We observed an increase in neuronal Ca²⁺ in response to nanomolar oligomerized Aβ_1–42 or micromolar Aβ_25–35 (but not to nonoligomerized Aβ_1–42 or Aβ_35–25) that was largely abrogated by 5–10 µM memantine and its more potent adamantane nitrate derivative, NitroMemantine (Fig. 3 A and B and Fig. S3 A–C) (34). At this concentration in this preparation, we previously have shown that memantine and NitroMemantine preferentially block eNMDARs while relatively sparing sNMDARs (3, 28). As a control, this effect of Aβ also was largely blocked in cultures depleted of astrocytes. Taken together with the foregoing results, these findings are consistent with the notion that Aβ induced release of glutamate from astrocytes, which in turn activated neuronal eNMDARs.

Excessive influx of Ca²⁺ via NMDARs activates neuronal nitric oxide synthase, which generates toxic levels of NO (35, 36). NO has been shown to contribute to synaptic spine loss after Aβ exposure, at least in part via mitochondrial actions of S-nitrosylated dynamin-related protein 1 (Drp1) after transnitrosylation from cyclin-dependent kinase 5 (Cdk5) enzyme (37, 38). Accordingly, in our cultures, in addition to a rise in neuronal Ca²⁺ levels, Aβ induced an increase in NO, as monitored with diaminofluorescein (DAF) fluorescence imaging (38, 39). Both memantine and NitroMemantine prevented this Aβ-induced increase in NO (Fig. 3C and Fig. S3B). Notably, NitroMemantine was significantly more effective than memantine in abrogating the increase in Ca²⁺ and toxic NO response, consistent with its more effective tonic blockade of eNMDARs, as previously suggested electrophysiologically (34). Confirming the involvement of eNMDARs in this process, we also found similar changes in Ca²⁺ and NO in response to oligomeric Aβ after pharmacological isolation of extrasynaptic currents using the published protocol of activating excitatory synaptic currents (by antagonizing inhibitory currents with the GABA antagonist bicuculline) followed by the addition of MK-801 to block the excitatory synaptic responses (Fig. S3 D and E) (3, 27, 28).

Previously, Selkoe and colleagues demonstrated that soluble oligomeric Aβ depressed long-term potentiation induced by high-frequency stimulation while enhancing long-term depression induced by electrical (low-frequency) stimulation (1, 15, 20, 40). Considering that accumulation of extracellular glutamate induced by high levels of oligomeric Aβ, as observed here, might underlie changes in synaptic function and plasticity, we next investigated the effect of Aβ on synaptic depression. To do so, we studied synaptic transmission at the Schaffer collateral–CA1 pathway of the hippocampus using electrophysiological recording of synaptic field potentials in acute hippocampal slices. We observed that as little as 50 nM oligomeric (but not 1 µM monomeric) Aβ_1–42 induced a gradual depression of field excitatory postsynaptic potentials (fEPSPs), outlasting the application of Aβ (Fig. 4A; n = 12). Although the input–output relationship also was affected by oligomeric Aβ_1–42, paired-pulse facilitation remained largely unaffected (Fig. 4 B and C).

In some ways, this effect of Aβ was reminiscent of chemical long-term depression, whereby glutamate induces synaptic depression that is modulated, at least in part, by eNMDARs (41–43). Therefore, we asked whether Aβ-induced synaptic depression might be mediated through eNMDARs by applying memantine, which we have shown at low-micromolar concentrations blocks eNMDARs to a relatively greater degree than sNMDARs (3, 28). We found that 10 µM memantine blocked induction of synaptic depression by 50 nM Aβ_1–42 (Fig. 4D; n = 4). Moreover, 500 nM Aβ_25–35 (but not Aβ_35–25) induced synaptic depression that was inhibited by 10 µM memantine or 5 µM NitroMemantine (Fig. S4 A and B). This latter finding is consistent with the greater potency of NitroMemantine at eNMDARs over memantine (Fig. 3 B and C and Fig. S3).

Next, we hypothesized that if eNMDARs indeed contributed to Aβ-induced synaptic depression, then activation of extrasynaptic receptors by exogenous NMDA (5–50 μM) should recapitulate the synaptic depression caused by Aβ and should be inhibited by low-micromolar memantine or NitroMemantine; in fact, we found this to be the case (Fig. S4C).

Given the evidence for activation of eNMDARs after acute application of Aβ, one might expect persistent eNMDAR-mediated activity from long-term accumulation of oligomerized Aβ in the AD brain. Hence, we asked if eNMDARs were tonically overstimulated in transgenic AD models of Aβ overexpression (33). Indeed, compared with WT littermate controls, we found a significant increase in basal inward current in neurons from the CA1 region of hippocampal slices prepared from hAPP-J20 tg mice in nominal absence of extracellular Mg²⁺ (Fig. 5; n = 15, P < 0.01). Pharmacological inhibition by the AMPAR antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) plus the NMDAR antagonist 3-[(R)-2-carboxypiperazin-4-yl]-propyl-1-phosphonic acid (CPP) showed that this basal current was induced by glutamate (Fig. 5 A, B, and E and Fig. S4D). Moreover, application of 100 nM α-Bgtx largely abrogated the basal current, consistent with the notion that α7nAChRs were mediating release of glutamate in the slices, as encountered earlier in our culture and in vivo AD models. Additionally, CNQX plus CPP blocked mEPSCs in these slices, indicating the glutamatergic nature of these synaptic currents (Fig. 5 A–D).

Although the small basal glutamatergic currents in some control WT slices (e.g., Fig. 5A) could be ascribed to leakage or damage, other control slices manifested virtually no basal current (Fig. 5F). Moreover, memantine (10 µM), which is known to inhibit eNMDARs preferentially over sNMDARs (3, 28), substantially blocked the basal current in hAPP-J20 slices (Fig. 5 G and H) but had little or no effect on mEPSC amplitude or frequency in WT slices even with incubation periods of more than 1 h (Fig. 5I). In contrast, after ∼30 min of perfusion with memantine, mEPSCs frequency increased but amplitude and kinetics remained relatively unaffected in hAPP-J20 slices (Fig. 5J). These findings are consistent with the notion that the loss of synaptic function observed with chronic exposure to Aβ might be partially reversible on a relatively short time scale if excessive eNMDAR activity were inhibited, in this case by memantine.

To obtain evidence independent of memantine that the basal current was indeed mediated by eNMDAR activation, we took advantage of the recent report that glycine is the predominant coagonist of eNMDARs. Thus, by degrading glycine in slice preparations, the enzyme glycine oxidase (GO) can inhibit eNMDAR-mediated responses. In contrast, enzymes that degrade d-serine, such as d-amino acid oxidase (DAAO) or d-serine deaminase (DsdA), can inhibit sNMDAR-mediated responses (44, 45). We used this approach to provide further proof that the basal inward current seen in the hAPP-J20 slices (and associated with our measurement of extracellular glutamate in these mice by HPLC) is caused by eNMDAR activity, because GO but not DAAO largely blocked this current (Fig. S5 A and B).

Next, we studied mEPSCs in hippocampal slices from hAPP-J20 and WT littermates in more detail. We observed a significant decrease in frequency (P < 0.01), a small but insignificant decrease in amplitude, and no change in kinetics of mEPSCs in hAPP-J20 slices compared with WT (Fig. 5 K–O; n = 12).

We sought a heuristic explanation for the decrease in mEPSC frequency that occurred in response to Aβ-induced eNMDAR activity in hippocampal autapses and in hAPP-J20 hippocampal slices, as well as for the partial recovery of mEPSC frequency in hAPP-J20 tg slices treated with memantine. One possibility is that the decrease in mEPSC frequency reflects initial synaptic dysfunction that subsequently leads to synapse loss. We reasoned that, if the initial decrease in mEPSC frequency induced by eNMDAR activity truly represented the initial phase of synaptic damage, then the molecular pathway(s) underlying this damage should be engaged early on and that pharmacological blockade of eNMDARs should inhibit these molecular pathways and give protection from subsequent morphological loss of dendritic spines. Such spine loss has been observed to occur ∼16 h after exposure to oligomerized Aβ (1), although an initial decrease in spine volume has been reported as early as 10 min (10). Here, we found that treatment of our mixed neuronal/glial cultures with the standard pharmacological protocol to activate eNMDARs selectively after blocking sNMDARs (using bicuculline exposure followed by MK-801, washout, and then exposure to low concentrations of NMDA) (3, 27, 28) triggered an increase in tau and hyperphosphorylated tau and in caspase-3 activation (Fig. 6 A and C). When the cultures were exposed to oligomerized Aβ rather than low-dose NMDA, phospho-tau increased more dramatically than tau (Fig. 6B). Importantly, these same pathways involving phospho-tau and caspase-3 previously were shown to be involved in oligomeric Aβ-induced abnormal excitability and synaptic spine loss (10, 11, 14, 46–56). For example, the relatively specific inhibitor of caspase-3 activity, z-DEVD-fmk, blocks a pathway leading to dendritic spine shrinkage via activation of calcineurin, which results in dephosphorylation and internalization of synaptic AMPARs (56). As would be expected for a mechanism involving eNMDAR activation, this effect of Aβ was inhibited by memantine and to an even greater degree by NitroMemantine (Fig. 6B). In contrast to eNMDARs, selective pharmacological activation of sNMDARs led to a decrease in tau phosphorylation and caspase-3 activation (Fig. 6 A and C).

To assess the contribution of eNMDAR activity to Aβ-induced synapse loss, we next exposed hippocampal slices to oligomerized Aβ and then treated them with memantine or its improved derivative NitroMemantine to block eNMDARs selectively while relatively sparing sNMDARs (3, 28). In fact, at equimolar concentrations NitroMemantine (34) is more effective than memantine not only in blocking eNMDAR activity but also in sparing synaptic activity (Fig. 3 B and C, Fig. S3, and Fig. S4 A, B, and E). We found that treatment with these eNMDAR-selective antagonists ameliorated the effect of Aβ on synaptic loss, with NitroMemantine manifesting a larger and more significant protective effect than memantine, as monitored morphologically in YFP-labeled dendritic spines (Fig. 6 D and E). To confirm that this protective effect was mediated by eNMDARs, we used a second approach to inhibit eNMDARs vs. sNMDARs by bathing hippocampal slices in glycine and d-serine--degrading enzymes (44, 45, 57). We found that only inhibition of eNMDARs with glycine oxidase significantly protected from spine loss induced by oligomerized Aβ (Fig. S5C). Taken together, these results suggest that early events associated with hyperactivation of eNMDARs by Aβ indeed may be linked to the subsequent loss of synapses in AD.

To test further the effect of inhibiting eNMDARs relative to sNMDARs with memantine and NitroMemantine, we treated the triple transgenic (3× tg) AD mouse model for 3 mo starting at 6 mo of age. This mouse model of AD manifests early synaptic dysfunction and cognitive deficits in the presence of soluble Aβ and abnormal tau protein before the formation of frank amyloid plaques and tau tangles (58). We used this mouse because we wanted to analyze the effect of soluble oligomeric Aβ, rather than plaques and tangles, and, unlike other AD mouse models, there were few if any plaques or tangles when the mice were treated. Moreover, this mouse model displayed only very mild loss of neurons in the neocortex and hippocampus, despite a dramatic loss of synapses. Hence, neuronal loss could not account for the decrease in synapses. By quantitative confocal immunofluorescence microscopy in both the cortex and hippocampus, after treatment we observed a significant increase in synaptic and dendritic density by synaptophysin and microtubule-associated protein 2 (MAP2) staining, respectively (Fig. 7 A–E). As in the hippocampal slice preparation, NitroMemantine demonstrated an effect superior to that of memantine. Additionally, in behavioral studies of hippocampal function, NitroMemantine-treated (but not memantine-treated) 3× tg AD mice displayed significantly improved function on the location-novelty recognition test (Fig. 7F).

Mixed neuronal/glial rat cerebrocortical cultures and purified rat, mouse, and human astrocytes were prepared following standard protocols with some modifications (37). For studying autapses, microisland rat hippocampal cultures were prepared as previously described (SI Materials and Methods) (28).

To generate hiPSCs from human dermal fibroblasts, we used an integration-free reprogramming method that used electroporation of three episomal expression vectors collectively encoding six reprogramming factors, namely OCT3/4, SRY-box containing gene 2 (SOX2), Kruppel-like factor 4 (KLF4), l-MYC, LYN28, and p53-shRNA) (64). hiPSC colonies were maintained on mouse embryonic fibroblast feeders and were validated for pluripotency, trilineage differentiation capability, and karyotypic stability as previously described (65) (SI Materials and Methods).

Using FRET microscopy of the SuperGluSnFR probe (17), we monitored glutamate release from mixed neuronal/glial and pure astrocyte cultures (SI Materials and Methods).

Recombinant glycine oxidase (GO) and d-serine deaminase were purified as described (57). d-Amino acid oxidase) was purchased from a commercial source (Calzyme). All enzymes were kept frozen and were dissolved immediately before use.

Human synthetic Aβ_1–42 (GenicBio or Anaspec) was dissolved following established procedures. Naturally occurring soluble Aβ dimers and trimers were prepared by size-exclusion chromatography from postmortem human AD cortex, with minor modifications of the procedure described previously (1, 15, 20, 40). By ELISA, this preparation contained 220 pM of Aβ before 1:4 dilution for use in experiments at a final concentration of 55 pM. For all Aβ preparations, concentration and quality were assessed by ELISA and Western blot analysis (SI Materials and Methods).

The Aβ_1–42 oligomeric concentration was analyzed as reported previously (SI Materials and Methods) (66).

For hippocampal autaptic cultures, whole-cell recordings were performed on single neurons located on microislands of one or more astrocytes. Recordings were performed on 14–26 d in vitro cultures at room temperature using a patch-clamp amplifier (Axopatch 200B or MultiClamp 700A; Molecular Devices). Drugs were administered via a fast valve-controlled perfusion system (Lee Company). All antagonists were purchased from Tocris Bioscience unless otherwise noted. The NitroMemantine derivative used in these studies is the lead drug candidate designated YQW-036/NMI-6979 (34). For data acquisition and analysis, signals were filtered, digitized, and stored in a computer (Dell) using PClamp v.10 software (Axon Instruments). Display and analysis of data distributions were carried out using a statistical software package (Origin 7, OriginLab Corp.) (SI Materials and Methods).

Intracellular nitric oxide [NO]_i and calcium [Ca²⁺]_i concentrations in cells in primary cultures were measured using DAF FM diacetate (2.5 µM) and Fura-2/AM (4 μM), respectively (SI Materials and Methods).

To assess the effect of eNMDARs selectively, we initially activated sNMDARs by a brief (7–10 min) application of bicuculline (50 µM) to block inhibitory transmission and then blocked the sNMDARs with the long-lasting NMDAR inhibitor MK-801 (10 µM), as previously described (3, 5, 28). After bicuculline and MK-801 were washed out from the dish, NMDA or Aβ (to release glutamate from astrocytes) was added to elicit eNMDAR-dependent signaling.

Thy1-YFPH transgenic mice (8–10 d old) were used to prepare organotypic hippocampal slices using the interface method. Dendritic spine density was evaluated as described previously (37, 40) (SI Materials and Methods).

To determine the in situ integrity of the presynaptic and dendritic complex of the hippocampus and neocortex, 40-µm-thick vibratome sections were cut from paraformaldehyde-fixed brain and were immunolabeled with mouse monoclonal antibodies against synaptophysin (SY38; 1:500; Millipore) and MAP2 (1:100; Millipore) (SI Materials and Methods).

sNMDARs and eNMDARs (by exposure to 100 μM NMDA) or oligomerized Aβ_1–42 (by exposure to 250 μM NMDA) were stimulated for 15 min to 1 h; then cell lysates were prepared as described previously (3). Western blots were probed for phospho-tau (AT8; Thermo), tau (TAU-5; Millipore), cleaved caspase-3 (Cell Signaling), and actin (Millipore) (SI Materials and Methods).

The novel object exploration tests included the location novelty recognition test and the object novelty recognition tests to assess spatial learning as well as object learning (SI Materials and Methods).

After 30 min incubation in oligomerized Aβ_1–42, culture medium was boiled and cleared by centrifugation (10,000 × g for 10 min) to remove insoluble materials. Subsequently, samples were diluted in 20 mM borate buffer at pH 9.0 and were derivatized for 1 min with N-tert-butyloxycarbonyl-l-cysteine and o-phthaldialdehyde. Samples then were separated in a 5-mm C18 reverse-phase column (220 × 4.6 mm) Sheri-5 (Brownlee), and glutamate was monitored by fluorescence (334 nm excitation and 433 nm emission) using an l-7485 detector (Hitachi).

Mice were treated with memantine or NitroMemantine as previously described (SI Materials and Methods) (3).

All procedures were conducted in accordance with Institutional Animal Care and Use Committee (IACUC) guidelines. Microdialysis procedures were performed as previously described (67) on 12-mo-old and 20- to 24-mo-old male and female mice of the following genotypes: hAPP tg, α7KO, and hAPP tg/α7KO (23). Dialysate samples were analyzed as previously reported (68) (SI Materials and Methods).

A Student t test was used for two-way comparisons, and an ANOVA with Tukey's HSD test was used for multiple comparisons. For cumulative probability curves, comparisons were made using the Kolmogorov–Smirnov test. Results are expressed as mean ± SEM.