alpha-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid subtype glutamate receptor (AMPAR) endocytosis is essential for N-methyl-D-aspartate-induced neuronal apoptosis.

Excessive activation of the N-methyl-d-aspartate subtype glutamate receptor (NMDAR) is thought to be involved in mediating programmed cell death (apoptosis) in numerous central nervous diseases. However, the underlying mechanisms remain unknown. We report here that stimulation of NMDARs activates intracellular signaling cascades leading to apoptosis and facilitates clathrin-dependent endocytosis of alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid subtype glutamate receptors (AMPARs). Both broad spectrum inhibitors of clathrin-dependent endocytotic processes and a specific inhibitor of AMPAR endocytosis selectively inhibit NMDA-induced apoptosis without affecting apoptosis produced by staurosporine. These results demonstrate that clathrin-dependent endocytosis of AMPARs is an essential step in NMDAR-mediated neuronal apoptosis. Our study not only identifies a previously unsuspected step in NMDA-induced apoptosis but also demonstrates that AMPAR endocytosis, in addition to attenuating synaptic strength as previously demonstrated in models of synaptic plasticity, may play a critical role in mediating other important intracellular pathways.

Neuronal apoptosis induced by activation of NMDARs 1 has been postulated to underlie the loss of neurons and neuronal function in many central nervous system disorders (1)(2)(3)(4)(5). However, the mechanisms linking NMDAR activation to neuronal apoptosis remain unclear, although excessive calcium influx has been proposed as a primary underlying processes (1,3,4). Interestingly, several recent studies have indicated that NMDAR activation can also lead to significant facilitation of clathrinmediated endocytosis of AMPARs, leading to long term depression of AMPAR-mediated synaptic transmission (6,7). Given the evidence that endocytosis of EGF receptors, in addition to attenuating receptor activation on the plasma membrane, is essential to some signaling pathways mediated by EGF receptors (8), we set out to investigate whether the stimulated endocytosis of AMPARs constitutes an essential step in the signaling pathway leading to NMDA-induced neuronal apoptosis.

EXPERIMENTAL PROCEDURES
Primary Culture of Hippocampal Neurons-Hippocampal cultures were prepared from embryonic E18 Sprague-Dawley rats and grown in culture medium (Neurobasal medium containing B-27 supplement and 0.5 mM glutamine (Invitrogen)). The medium from mature 14-day in vitro neurons was removed and replaced with 100 M NMDA plus 10 M glycine in extracellular solution (10 mM HEPES, pH 7.35, 140 mM NaCl, 25 mM glucose, 5.4 mM KCl, 1.3 mM CaCl 2 , osmolarity: 310 -320 mosM) for 1 h at 37°C prior to restoring neurons to the culture medium. Twenty-four hours after NMDA/glycine application, neurons were processed using cell death assays. NMDA-induced [Ca 2ϩ ] i responses were evoked and measured using methods described previously (9).
Apoptosis Assays-NMDA-induced apoptosis was quantified using either a Cell Death Detection Elisa Plus Kit (Roche Applied Sciences) or using TdT-mediated addition of biotinylated 11-dUTP to the free 3Ј-OH ends of DNA (Chemicon). Absorbance readings for both assays were determined using a spectrophotometric microplate reader. Data analysis and expression relative to appropriate controls were performed according to the manufacturer's instructions for each kit.
Propidium Iodide (PI) Staining of Nuclei-After the induction of apoptosis, cells were fixed with 4% paraformaldehyde, 4% sucrose for 10 min followed by ice-cold acetone for 1 min, and then stained with 20 mg/ml PI in Dulbecco's phosphate-buffered saline for 30 min and viewed with a Leica fluorescence microscope to identify condensed nuclei. Cells with condensed nuclei were counted as apoptotic, and the percentage of apoptotic cells to the total number of cells was calculated to give a semiquantitative analysis, expressed as percentage of apoptosis.
Peptide Treatments-Synthetic peptides, GluR2 3Y or GluR2 3A , were incubated with a carrier protein (Pep-1) (10) at a ratio of 1:20 in Dulbecco's modified Eagle's medium (Invitrogen) at 37°C in a humidified atmosphere containing 5% CO 2 for 30 min to allow the formation of GluR2/Pep-1 complexes. Hippocampal neurons (12-14 days in vitro) were then overlaid with the preformed complex to reach a final GluR2 peptide concentration of 1 M and further incubated for 1 h before experiments commenced.
ELISA Cell-surface Receptor Assay-Quantification of cell-surface AMPA or NMDARs was performed by a colorimetric cell-ELISA assay essentially as described previously (11), using monoclonal antibodies * This work was supported by Canadian Institute of Health Research (CIHR) and Heart and Stroke Foundation of British Columbia and Yukon (HSFO/BC&Y) (to Y. T. W.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ These authors contributed equally to the work. against the extracellular domains of GluR2 or GluR1 or NR1 (Chemicon, 1 g/ml). Data are expressed as absorbance measurement under non-permeabilized conditions/absorbance measurement under permeabilized conditions.
Transferrin Receptor Endocytosis Assay-Hippocampal neurons were incubated with 2 mg/ml Alexa-A488-conjugated transferrin (Molecular Probes) for 30 min at 37°C in the presence or absence of endocytosis inhibitors. Internalized receptors were then visualized with a Leica fluorescence microscope.

RESULTS
NMDA treatment (100 M plus 10 M glycine, 1 h) induced a time-dependent increase in caspase-3 activity, a biochemical indicator of neuronal apoptosis (12), as detected by ELISA assay of DEVD-p-nitroanilide cleavage (Fig. 1A). The increase peaked 12-24 h after the treatment, at which time the majority of neurons were either dying or dead, exhibiting the hallmarks of apoptotic cell death, including DNA laddering as demonstrated by gel electrophoresis of extracted DNA (Fig. 1B), and nuclear condensation with disintegrating processes shown by nucleus staining with PI ( Fig. 1C) or the intercalating DNA dye, Hoechst 33258 (bisbenzimide, data not shown). The degree of neuronal apoptosis was also quantified by measuring internucleosomal cleavage of DNA with both 11-dUTP (Fig. 1E) and histone biotinylation assays (Fig. 1F). Furthermore, the NMDA-induced apoptosis was a result of specific activation of NMDARs, as it was fully blocked by the NMDAR antagonist, APV (50 M; Fig. 1D), but not inhibited by blocking AMPARs with DNQX (10 M; Fig. 1D). Therefore, in accordance with previous reports (12), the NMDA treatment paradigm used in the present study produced neuronal apoptosis.
To investigate the role of NMDA-induced endocytosis in mediating neuronal apoptosis, we first examined the effect of hypertonic sucrose, a well characterized inhibitor of clathrindependent endocytosis that prevents the assembly of clathrincoated pits (11,13). As shown in Fig. 1E, sucrose treatment (400 mM) dramatically reduced NMDA-induced apoptosis. To more specifically target clathrin-dependent endocytosis, we employed a dynamin-derived, myristoylated peptide (myr-Dyn). This peptide is membrane-permeable and effectively inhibits clathrin-mediated endocytosis (14,15) by blocking the recruitment of dynamin to clathrin-coated pits by amphiphysin (16). We found that myr-Dyn (10 M) was as effective as hypertonic sucrose in reducing NMDA-induced apoptosis (Fig. 1, E and F). In contrast, control Dyn peptides, both non-myristoylated (membrane-impermeable) Dyn (Dyn; Fig. 1E) and scrambled myr-Dyn (s-myr-Dyn; Fig. 1F), had little effect. To determine whether endocytosis is a general requirement for neuronal apoptosis, we next tested the effect of these inhibitors on a well characterized model of neuronal apoptosis induced by treating neurons with the kinase inhibitor, staurosporine (STS; 100 nM, 1 h) (12). As shown in Fig. 1E, both endocytosis inhibitors failed to significantly alter STS-induced neuronal apoptosis. These results demonstrate that a clathrin-dependent endocytotic process is necessary for NMDAR-mediated, but not STS-mediated, apoptosis.
Ca 2ϩ overload following NMDAR activation has been suspected as a primary causal factor leading to neuronal apoptosis (1,3,4). However, as summarized in Fig. 2A, sucrose at concentrations that inhibited endocytosis and apoptosis did not significantly alter NMDA-evoked transient [Ca 2ϩ ] i responses measured with the Ca 2ϩ dye, Fura-2. The fact that inhibition of endocytosis blocked NMDA-induced apoptosis without affecting its [Ca 2ϩ ] i responses strongly suggests that intracellular increases in Ca 2ϩ concentrations, although necessary (3, 4), may not be sufficient to produce NMDA-induced apoptosis.
Activation of certain forms of caspases, such as caspase-3 and caspase-7 (17) (also see Fig. 1A) has been implicated in NMDA-induced neuronal apoptosis. We therefore investigated the effects of inhibiting endocytosis on NMDA dependent activation of caspase-3. NMDA treatment dramatically increased the level of the activated form of caspase-3, as demonstrated by Western blots using an antibody that specifically recognizes AMPAR Endocytosis Is Essential for NMDA-induced Neuronal Apoptosis 41268 activated/cleaved caspase-3, and this effect was efficiently inhibited by the membrane-permeable myr-Dyn (Fig. 2B). The serine/threonine kinase Akt/PKB has previously been shown to be critically involved in protecting neurons from apoptotic cell death (18), and inhibition of this kinase activity has previously been suspected of being involved in NMDAR-mediated apoptosis (19). We therefore investigated whether the endocytosis process plays a critical role in the inhibition of Akt activity by determining the level of Akt phosphorylation at serine 473, a residue whose phosphorylation is required for full activation of Akt (20). As shown in Fig. 2C, treatment of neurons with NMDA resulted in a significant reduction in Ser 473 -phosphorylated Akt and hence Akt activity without altering the levels of total Akt. This reduction in Akt activity was largely prevented by the inhibition of endocytosis with hypertonic sucrose. In striking contrast, sucrose treatment had no effect on the reduction of Akt phosphorylation following STS treatment (Fig. 2C). Thus, stimulated endocytosis appears to be an obligatory step that is downstream of rising [Ca 2ϩ ] i and upstream of caspase activation and Akt inhibition in NMDA-induced neuronal apoptosis.
Several recent studies in the synaptic plasticity field have demonstrated that NMDAR activation can selectively stimulate the clathrin-mediated endocytosis of AMPARs (6,7,(21)(22)(23). Consistent with these previous studies, we observed a significant reduction of cell-surface AMPA, but not NMDA, receptors following the NMDA paradigm used here. Since the reduction in AMPARs was observed by measuring cell-surface expression of both the GluR2 subunit (Fig. 3, A and B) and the GluR1 subunit (64 Ϯ 2 and 51 Ϯ 1% of the total receptors on the cell surface for control and NMDA-treated neurons, respectively; n ϭ 5, p Ͻ 0.001), it appears to be a decrease in total cell-surface AMPARs and not a replacement of GluR2-containing with GluR2-lacking receptors as a result of GluR2-dependent endocytosis. The AMPAR reduction was blocked by the endocytosis inhibitor myr-Dyn but not the control peptide, Dyn ( Fig. 3A), indicating the involvement of facilitated clathrin-dependent endocytosis. We have demonstrated previously that stimulated AMPAR endocytosis requires the GluR2 subunit (11), more specifically the short amino acid sequence between residues tyrosine 869 and glutamic acid 879 within the carboxyl-terminal region (21,24). We have therefore reasoned that a short peptide, which comprises the required sequence YKEGY-NVYGIE (termed GluR2 3Y ), may function as a dominant inhibitor of NMDAR-mediated AMPAR endocytosis (24). As predicted, when the GluR2 3Y peptide (1 M) was delivered into cultured neurons by mixing it with a carrier peptide (Pep-1) (10) 1 h prior to and during the NMDA treatment, the NMDAinduced reduction of cell-surface AMPARs was completely prevented (Fig. 3B). To be sure that the blockade by this peptide was not due to nonspecific effects on the endocytotic process, we examined its effect on transferrin receptor endocytosis, a well characterized clathrin-mediated receptor endocytosis (13). As shown in Fig. 3C, internalization of fluorescently labeled transferrin as a result of transferrin receptor endocytosis was eliminated by hypertonic sucrose, but not by GluR2 3Y ϩ Pep-1, applied to these neurons 1 h prior to and during the transferrin incubation. Thus, GluR2 3Y does not indiscriminately affect clathrin-mediated endocytosis.
If NMDA-induced apoptosis requires AMPAR endocytosis, C, endocytosis inhibition specifically disrupts the NMDA-induced, but not the STS-induced, reduction in Akt phosphorylation. Cell lysates from neurons treated as indicated were first probed with an antibody specific to Akt phosphorylated on serine 473, the active form of the enzyme (P-Akt). Membranes were then stripped and re-probed with an anti-Akt antibody. Blots from four individual experiments were imaged and quantified. The histogram represents Akt phosphorylation relative to total Akt. **, p Ͻ 0.01 when compared with the respective control group (n ϭ 4 -6).
FIG. 3. NMDA stimulates clathrin-mediated AMPAR, but not NMDAR, endocytosis. Cultured neurons were pretreated with either dynamin peptides or GluR2 3Y for 30 min followed by treatment with NMDA (100 M plus 10 M glycine) for 1 h in extracellular solution before being assayed for cell-surface receptor expression. A, ELISAbased cell-surface receptor assays for NMDA and AMPARs (see "Experimental Procedures"). NMDA treatment induced a significant reduction in cell-surface AMPAR but not NMDAR, and AMPAR internalization was prevented by pretreatment of neurons with the myr-Dyn peptide (myr-Dyn; 10 M), but not the membrane impermeable control, Dyn (Dyn; 10 M). B, NMDA-induced AMPAR internalization is blocked by a GluR2 carboxyl tail-derived peptide (GluR2 3Y ; 1 M) delivered into neurons by mixing it with a carrier peptide Pep-1 (Pep-1ϩGluR2 3Y ). **, p Ͻ 0.01 compared with control; n ϭ 36 -72 wells from three separate experiments for each group. C, GluR2 3Y does not block transferrin receptor endocytosis. Fluorescent images (top panels) were overlaid with differential interference contrast images (bottom panels). Neuronal uptake of fluorescently labeled transferrin was blocked by clathrindependent endocytosis inhibitor hypertonic sucrose but not by Pep-1 ϩ GluR2 3Y .

AMPAR Endocytosis Is Essential for NMDA-induced Neuronal Apoptosis 41269
one would expect that the GluR2 3Y peptide should be effective in preventing NMDA-induced apoptosis. Consistent with our prediction, GluR2 3Y plus Pep-1 significantly reduced NMDAinduced apoptosis as quantified by the histone biotinylation assay (Fig. 4A), and by PI nuclear staining (Fig. 4, B and C). Neither GluR2 3Y nor Pep-1 alone nor Pep-1 plus the control peptide GluR2 3A (in which the three tyrosine residues were replaced with alanines (24)) had any detectable effect on NMDA-induced apoptosis (Fig. 4A). Similar to the general blockade of the clathrin-mediated endocytotic process with either sucrose or myr-Dyn, interference with AMPAR endocytosis by GluR2 3Y did not alter STS-induced neuronal apoptosis (Fig. 4B). Taken together, our results provide strong evidence for an obligatory requirement for AMPAR endocytosis in mediating NMDA-induced neuronal apoptosis.

DISCUSSION
In the present study we have found that clathrin-dependent AMPAR endocytosis is specifically required for NMDA-induced, but not STS-induced, apoptosis of hippocampal neurons maintained in primary culture. Blocking endocytosis has no effect on NMDA-induced Ca 2ϩ responses but prevents both NMDA-induced activation of caspase-3 and inhibition of Akt phosphorylation. Since AMPAR endocytosis is known to be an intracellular Ca 2ϩ -dependent process (11,21,23), it is likely that Ca 2ϩ influx through NMDARs may function as a trigger leading to the facilitation of AMPAR endocytosis, which may in turn initiate a cascade causing neuronal apoptosis. Thus, the present study has identified AMPAR endocytosis as a critical link between NMDA-induced Ca 2ϩ i overload and intracellular cascades leading to apoptosis.
A large body of evidence accumulated recently has established a central role for stimulated clathrin-dependent AMPAR endocytosis in reducing receptor-mediated synaptic transmission, contributing to the expression of certain forms of well characterized synaptic plasticity including long term depression in both hippocampus (11,25) and cerebellum (26). The requirement for AMPAR endocytosis in NMDA-induced apoptosis demonstrated in the present study further suggests that, in addition to the decrease in receptor-mediated synaptic transmission, endocytosed AMPARs also play previously unsuspected roles in intracellular signaling pathways. The mediation of NMDA-induced apoptosis is one example. The detailed mechanisms by which endocytosed AMPARs contribute to the initiation of downstream apoptotic cascades remain to be determined. However, analogous to the recently reported retinal cell apoptosis induced following stimulated endocytosis of rhodopsin (27)(28)(29) and activation of EGF receptor-mediated mitogen-activated protein kinase signaling by endocytosed EGF receptors (8), it is possible that endocytosed AMPARs, in response to NMDA stimulation, may form, with some of the clathrin-coat protein components, a signaling complex that initiates the apoptotic cascade. Nevertheless, the present study provides the first evidence for the involvement of stimulated AMPAR endocytosis in NMDA-induced neuronal apoptosis, thereby unveiling a novel pathway in mediating cell apoptosis. NMDAR-mediated apoptosis is one of the primary causes leading to the selective loss of neurons in both neurodegenerative diseases and acute brain insults, and clathrin-dependent AMPAR endocytosis is a fundamental process in the control of efficacy of synaptic transmission at glutamatergic synapses. By linking the endocytosis of AMPARs to neuronal apoptosis, our study suggests the components of the AMPAR endocytotic pathway may be novel therapeutic targets for the development of a new generation of neuroprotective agents.