AMPA protects cultured neurons against glutamate excitotoxicity through a phosphatidylinositol 3-kinase-dependent activation in extracellular signal-regulated kinase to upregulate BDNF gene expression


Address correspondence and reprint requests to Ann M.Marini, Department of Neurology and Neuroscience, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814, USA. E-mail:


The signal transduction and molecular mechanisms underlying α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA)-mediated neuroprotection are unknown. In the present study, we determined a major AMPA receptor-mediated neuroprotective pathway. Exposure of cerebellar granule cells to AMPA (500 µm) + aniracetam (1 µm), a known blocker of AMPA receptor desensitization, evoked an accumulation of brain-derived neurotropic factor (BDNF) in the culture medium and enhanced TrkB-tyrosine phosphorylation following the release of BDNF. AMPA also activated the src-family tyrosine kinase, Lyn, and the downstream target of the phosphatidylinositol 3-kinase (PI3-K) pathway, Akt. Extracellular signal regulated kinase (ERK), a component of the mitogen-activated protein kinase (MAPK) pathway, was also activated. K252a, a selective inhibitor of neurotrophin signaling, blocked the AMPA-mediated neuroprotection. The involvement of BDNF release in protecting neurons by AMPA was confirmed using a BDNF-blocking antibody. AMPA-mediated neuroprotection is blocked by PP1, an inhibitor of src family kinases, LY294002, a PI3-K inhibitor, or U0126, a MAPK kinase (MEK) inhibitor. Neuroprotective concentrations of AMPA increased BDNF mRNA levels that was blocked by the AMPA receptor antagonist, 1,2,3,4-tetrahydro-6-nitro-2,3-dioxo-benzo[f]quinoxaline-7-sulfonamide (NBQX). The increase in BDNF gene expression appeared to be the downstream target of the PI3-K-dependent activation of the MAPK cascade since MEK or the PI3-K inhibitor blocked the AMPA receptor-mediated increase in BDNF mRNA. Thus, AMPA receptors protect neurons through a mechanism involving BDNF release, TrkB receptor activation, and a signaling pathway involving a PI3-K dependent activation of MAPK that increases BDNF expression.

Abbreviations used

α-amino-3-hydroxy-5-methyl-4-isoxazole propionate


brain-derived neurotropic factor


cAMP response element binding protein


extracellular signal-regulated kinase


mitogen-activated protein kinase


MAPK kinase


3-[4,5-dimethylthiozol-2-yl]-2,5-diphenyltetrazolium bromide






phosphatidylinositol 3-kinase

Glutamate is the major excitatory neurotransmitter in the mammalian CNS and acts through two types of receptors, ionotropic and metabotropic (Hollman et al. 1989; Schoepp et al. 1999). Glutamate has also been implicated in the pathophysiology of hypoxic/ischemic neuronal damage (Olney et al. 1971; Choi 1995). The ionotropic glutamate receptor subtypes, N-methyl-d-aspartate (NMDA) (Choi 1995) and α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptors are thought to play a major role in the pathophysiology of hypoxic–ischemic neuronal damage (Michaelis 1998; Yamada et al. 1998). Moreover, AMPA/kainate antagonists protect hippocampal neurons against transient global ischemia more effectively than NMDA receptor antagonists (Sheardown et al. 1990). Protecting neurons in the CNS against hypoxic–ischemic neuronal injury will reduce neuronal cell death incurred by the injury, and increasing neuronal resistance to hypoxic–ischemic injury is one way to avert cell death. Thus, understanding fundamental survival mechanisms that exist in neurons is important in developing treatments that confer increased resistance.

Calcium-permeable AMPA receptors are expressed in cultured cerebellar granule cells (Hack and Balazs 1995; Savidge and Bristow 1997; Marini et al. 1999). AMPA receptor activation is neurotoxic to Purkinke and Golgi cells in the cerebellum and hippocampal neurons (Garthwaite and Garthwaite 1991) as well as in cortical cultures (Jensen et al. 2001). Influx of calcium through AMPA receptors may or may not be required for its neurotoxicity (Choi 1995; Friedman et al. 2003) but appears to be required for synaptic plasticity effects (Perkinton et al. 1999). However, AMPA is not excitotoxic to cultured cerebellar granule cells (Marini et al. 1999; Banaudha and Marini 2000). In fact, AMPA + aniracetam (Copani et al. 1992) protects neurons against the excitotoxic and apoptotic effects of glutamate cultured cerebellar granule cells (Banaudha and Marini 2000).

Previous studies suggest that AMPA increases brain-derived neurotrophic factor (BDNF) expression in cortical and cerebellar granule cells (Zafra et al. 1990; Lauterborn et al. 2000; Legutko et al. 2001). BDNF is a member of the neurotrophin family of trophic factors. Neurotrophins are known to be essential regulators of neuronal development, function, survival, and plasticity (Sofroniew et al. 2001). In cerebellar granule cells, BDNF protects vulnerable neurons against glutamate-mediated excitotoxicity (Lindholm et al. 1993; Marini et al. 1998). Because cerebellar granule cells express TrkB receptors, but not other Trks (Marini et al. 1998), we hypothesized that ionotropic AMPA receptors may protect neurons against excitotoxicity via BDNF release.

BDNF, by binding to and activating its high-affinity receptor TrkB, initiates a signal transduction cascade that involves recruitment of Shc and the Grb-2–Sos complex. Activation of Ras, Raf, mitogen-activated protein kinase (MAPK) kinase (MEK), Rsk, and cAMP response element binding protein (CREB) promote neuronal survival (for review, see Kaplan and Miller 2000; Segal and Greenberg 1996). BDNF has been shown to rescue neurons against different types of injury including excitotoxic cell death (Ernfors et al. 1990; Hyman et al. 1991; Oppenheim et al. 1992; Ip et al. 1993; Lindholm et al. 1993; Cheng and Mattson 1994; Schabitz et al. 1997; Wu and Partridge 1999).

In the cerebellum, the src family protein tyrosine kinase, Lyn, interacts with AMPA receptors to activate the MAPK signaling pathway to enhance BDNF mRNA levels, which may contribute to synaptic plasticity by regulating the expression of BDNF (Hayashi et al. 1999). AMPA receptors can also activate MAPK through a G protein-coupled mechanism in cultured cortical neurons (Wang and Durkin 1995; Hawes et al. 1996) or striatal neurons (Perkinton et al. 1999). AMPA receptor-mediated activation of phosphatidylinositol 3-kinase (PI3-K) appears to be required for MAPK activation (Lopez-Ilasaca et al. 1997; Perkinton et al. 1999). Because AMPA receptor-mediated processes may use similar mechanisms to protect neurons, we hypothesized that parallel signal transduction pathways are used by AMPA receptors to protect cerebellar granule neurons against glutamate excitotoxicity. Excitotoxicity is defined here as the influx of calcium from a neurotoxic concentration of glutamate acting through NMDA receptors (Novelli et al. 1988; Arundine and Tymianski 2003).

In this study, we report that AMPA exerts its major neuroprotective effect through an activity-dependent release of BDNF thereby enhancing the functional state of the TrkB receptor. The neuroprotective signaling cascade activated by AMPA involves the src family protein kinase, Lyn, and a PI3-K-dependent increase in activation of extracellular signal-regulated protein kinase to increase BDNF gene transcription.

Materials and methods

Preparation of cultured cerebellar granule cells

Cerebellar granule cells were prepared from postnatal day 8 Sprague–Dawley rat pups as described previously (Marini et al. 1989). Briefly, cerebella freed of meninges were minced and subjected to low speed centrifugation. The pellets from 20 rat pups were subjected to trypsinization, followed by inactivation of the trypsin by the addition of soybean trypsin inhibitor. Cells were dissociated by a series of triturations and recovered by centrifugation. The final pellet was reconstituted in basal medium eagle containing Earle's salts, 25 mm KCl, 10% fetal calf serum, and 2 mm l-glutamine. No antibiotics were added, and the plating density was 1.8 × 106 cells/mL. The cells (2-mL) were seeded in Nunc dishes (35-mm) coated with poly-l-lysine (1.0 µg/mL). The dishes were placed in a 37°C humidified incubator containing 5% CO2. Cytosine arabinoside (10 µm) was added 18–24 h later to inhibit the proliferation of non-neuronal constituents. On day in vitro (DIV) 7, glucose (100 µL of a 100 mm solution) was added to each culture dish to maintain survival. The culture medium was not replaced throughout the cultivation period. Granule cell neurons were used on DIV 8 for all experiments unless otherwise specified. Granule cell neurons represent about 95% of the total population of cells in the culture.

This research was conducted according to the principles set forth in the National Institutes of Health (NIH) Publication no. 85–23, Guide for the Care & Use of Laboratory Animalsand the Animal Welfare Act of 1986, as amended. Every effort was made to minimize the number of animals.

Determination of neuronal survival

Neuronal viability was quantified using the 3-[4,5-dimethylthiozol-2-yl]-2,5-diphenyltetrazolium bromide (MTT) assay as described previously (Lin and Long 1996). This colorimetric assay measures the reduction of MTT by mitochondrial succinate dehydrogenase to an insoluble, colored, formazan product. Because reduction of MTT can only occur in metabolically active cells, the level of activity is a measure of the viability of the cells. MTT was added to the dishes at a final concentration of 0.3 mg/mL for 30 min at 37°C. The cells were then solubilized in 70% isopropanol acidified with 0.1 N HCl and the optical density measured at 595 nm. Percentage neuronal survival was determined using untreated neurons.

Exposure of cerebellar granule cells to drugs and neurotoxins

All experiments were carried out with granule cell neurons on DIV 8 unless otherwise specified, and neuronal viability was determined 24 h later by the MTT assay. Aniracetam (1 µm) was added 5 min prior to the addition of AMPA (500 µm). Cycloheximide (1 µg/mL), the MEK inhibitor, U0126 (10 µm), the PI3-K inhibitor, LY294002 (10 µm), and the src kinase inhibitor, PP1 (6 nm), were added 1 h prior to the addition of AMPA and aniracetam. The AMPA antagonist, 1,2,3,4-tetrahydro-6-nitro-2,3-dibenzo[f]quinoxaline-7-sulfonamide (NBQX) [5 µm], was added 5 min prior to the addition of AMPA and aniracetam. Glutamate was added 4 h after the addition of aniracetam and AMPA in those experiments requiring the addition of this excitotoxin. K252a (10 nm), an inhibitor of the neurotrophin tyrosine kinase (Lee et al. 2002), was added 24 h prior to the addition of AMPA and aniracetam and 28 h prior to the addition of glutamate as indicated in the text. Neuronal viability was always determined 24 h after the addition of glutamate. The NMDA competitive receptor antagonist, 2-aminophosphovaleric acid (500 µm), was added together with AMPA and aniracetam followed by replacement with culture medium from naïve neurons along with AMPA + aniracetam. It should be noted that there was no difference in AMPA-mediated neuroprotection in the presence of the NMDA receptor antagonist indicating that protection was mediated by AMPA receptors (results not shown).

Quantitative 5′ nuclease assay of BDNF-specific mRNA

Cerebellar granule cells were treated with AMPA (500 µm) + aniracetam (1 µm) with or without various inhibitors as indicated in the text. The culture medium was aspirated and the granule cells were disrupted with Trizol (Invitrogen, Carlsbad, CA, USA) at room temperature and protected from light. The culture dishes were scraped, and the solubilized cells (≈107 cells/mL) were placed in 1.5 mL sterile microcentrifuge tubes. The samples were then placed on dry ice and stored at − 80°C. Total RNA was isolated according to the manufacturer's instructions. This ‘real-time’ fluorescence detection was performed using a 7700 Sequence Detector (Applied Biosystems, Foster City, CA, USA). Sequences of the PCR amplification primers and the detection probe for quantifying rat BDNF mRNA levels were based on the polypeptide-coding portion of the rat BDNF gene (GenBank accession number NM_012513) from total RNA of the granule cell neuron cultures. Sequences of the PCR primers were as follows: forward: 5′-AACCATAAGGACGCGGACTT-3′, reverse: 5′GGAGGCTCCAAAGGCACTT-3′. The sequence of the detection probe was 5′-FAM-ACTTCCCGGGTGATGCTCAGC-TAMRA-3′.

Western blot analysis

Phosphorylated [activated] ERK (Thr202/Tyr204) 1/2 (p44/p42)], total ERK1/2, phosphorylated (Ser473) [activated] Akt and total Akt, phosphorylated (Tyr 507) [activated] Lyn and total Lyn were analyzed in lysates from cerebellar granule cells on DIV 6–8. Cultured neurons were treated with a maximum neuroprotective concentration of AMPA (500 µm) + aniracetam (10 µm) at various times at 37°C in a 95% air/5% CO2 humidified incubator. At the indicated time, the culture medium was removed and the cells washed twice with ice-cold Locke's buffer containing: 154 mm NaCl; 5.6 mm KCl; 1 mm MgCl2; 2.3 mm CaCl2; 5.6 mm d-glucose; 8.6 mm HEPES; 1 mm Vanadate; pH 7.4. Cerebellar granule cells were disrupted in 200 µL of lysis buffer (1% Nonidet P-40, 20 mm Tris, pH 8.0, 137 mm NaCl, 10% glycerol, 1 mm phenylmethylsulfonyl fluoride, 0.15 units/mL aprotinin, 20 µm leupeptin, 1 mm sodium vanadate at 4°C). After removal of cellular debris by centrifugation, protein levels in the lysates were measured by the Bradford Coomassie Blue colorimetric assay (Bio-Rad Laboratories, Hercules, CA, USA) and equalized accordingly. Samples were boiled in the presence of sample buffer [2% socium dodecyl sulfate (SDS), 100 mm dithiothreitol, 10% glycerol, 0.25% bromphenol blue, pH 6.8] for 5 min. Samples (20 µg/lane) were electrophoresed through 10% SDS polyacrylamide gels and proteins were transferred to nitrocellulose membranes (0.22 µm, Schleicher and Schuell, Keene, NH, USA). After blocking with 5% non-fat dry milk dissolved in Tris-buffered saline containing 0.2% Tween-20 (TBST) for 30 min, the immobilized proteins were incubated overnight at 4°C with specific antibodies: (antiphosphorylated ERK1/2 1 : 1000, anti-ERK1/2 1 : 1000, antiphosphorylated Akt 1 : 1000, anti-Akt 1 : 1000, antiphosphorylated Lyn, anti-Lyn 1 : 1000; Cell Signaling, Beverly, MA, USA). On the following day, the membranes were washed with TBST (2 × 30 min), and then incubated with goat anti-rabbit IgG (1 : 10 000) conjugated to horseradish peroxidase (HRP; Pierce, Rockville, MD, USA) for 1 h. The secondary antibody to determine total ERK1/2 was goat anti-mouse IgG conjugated to HRP. Blots were washed in TBST (2 × 30 min), and exposed to enhanced chemiluminescence (ECL) reagent for 1–5 min. Excess reagent was removed and the blots exposed to photographic film (Kodak Biomax) according to the manufacturer's recommendations (Amersham Pharmacia Biotech, Piscataway, NJ, USA). Total protein levels of ERK1/2, Akt, Lyn, or actin were obtained to ensure equal loading of proteins. Blots were stripped with Restore Western Blot Stripping Buffer (Pierce, Rockford, IL, USA), washed with TBST followed by probing the blot with anti-ERK, anti-Akt, anti-Lyn or antiactin (Sigma-Aldrich, St Louis, MO, USA) overnight at 4°C. Blots were washed with TBST for 30 min followed by incubation with the appropriate secondary antibody linked to HRP for 1 h. Immunoreactive bands were visualized using the ECL detection according to the manufacturer's recommendations (Amersham Pharmacia Biotech). Immunoactive bands were quantitated by image analysis using a scanner (Hewlett-Packard) and Scanalytics image analysis software. When more than one band was detected by the antibodies (pERK1 and ERK2), each band was scanned separately and quantitated.

Because total Lyn levels increased upon exposure to AMPA, all blots probed with anti-Lyn were stripped and reprobed with antiactin to demonstrate equal loading of protein in each lane.

BDNF two-site enzyme immunoassay

The quantitative two-site enzyme immunoassay for the determination of BDNF was performed using an immunoassay system from Promega (BDNF Emax immunoassay system). In brief, neurons (1.8 × 106 cells/mL) were exposed to AMPA (500 µm) in the presence of aniracetam (1 µm) for various times, and the medium was then collected and concentrated using a Centricon (Amicon) concentrator as described previously (Marini et al. 1998).

TrkB-tyrosine phosphorylation

Analysis of TrkB-tyrosine phosphorylation was carried out essentially as described (Marini et al. 1998). Briefly, granule cell neurons were treated for 10 min with human recombinant BDNF (Promega, Madison, WI, USA), AMPA (500 µm) + aniracetam (1 µm) for 10 min and 1 h. The neurons were disrupted in 150 µL of lysis buffer and protein concentration was measured by the Bradford Coomassie Blue colorimetric assay. Lysates were incubated with pan-trk antibody (1 : 1000; Santa Cruz, Biotechnology, Santa Cruz, CA, USA) followed by protein A-Sepharose precipitation (Amersham Pharmacia Biotech) at 4°C for 2 h. The immunoprecipitates were washed with lysis buffer and water before resuspension in 10 µL of sample buffer (2% SDS, 100 mm dithiothreitol, 10% glycerol, 0.25% bromphenol blue) followed by electrophoresis through 7.5% polyacrylamide gels. After transfer to nitrocellulose membranes, the immobilized proteins were incubated overnight at 4°C with antiphosphotyrosine monoclonal antibody 4G10 (1 : 2500; Upstate Biotechnology, Lake Placid, NY, USA). Blots were subsequently washed and incubated with HRP-conjugated goat IgG (1 : 10 000; Chemicon, Temecula, CA, USA). Immunoreactive bands were analyzed by using ECL detection according to the manufacturer's recommendations (Amersham Pharmacia Biotech). The blot was stripped and reprobed with anti-TrkB antibody (1 : 2000, Upstate Biotechnology, Lake Placid, NY, USA) to demonstrate that equal amounts of Trk protein were present in each lane.

Statistical analysis

All values are presented as mean ± SD or percentage of control (untreated). Either the Student's t-test or anova with post-hoc comparisons using Bonferroni–Dunn were used for analysis; p < 0.05 is considered significant.


Glutamate, AMPA, NBQX, aniracetam, U0126, K252a, and the chemicals used for preparation of the cerebellar granule cells were obtained from Sigma-Aldrich (St Louis, MO, USA). The src-protein tyrosine kinase inhibitor, PP1, was purchased from Biomol (Plymouth Meeting, PA, USA). The BDNF-blocking antibody is a chicken polyclonal antibody purchased from Promega (Madison, WI, USA). All others chemicals were obtained from commercial sources. All drugs were dissolved in sterile water with the exception of K252a, NBQX, PP1, and U0126, which were reconstituted in dimethyl sulfoxide. Aniracetam was diluted in 50% ethanol as described previously (Banaudha and Marini 2000).


Cycloheximide blocks AMPA receptor-mediated neuroprotection

Glutamate is neurotoxic to cerebellar granule cells acting on NMDA receptors (Novelli et al. 1988; Marini and Paul 1992) and kills about 50% of the neurons. We showed previously that the neuroprotective activity of AMPA (500 µm) + aniracetam is time- and concentration-dependent against glutamate-mediated excitotoxicity (Banaudha and Marini 2000). Because AMPA receptor-mediated neuroprotection is time-dependent, we determined whether a neuroprotective protein(s) is involved by using the protein synthesis inhibitor cycloheximide. Note that AMPA + aniracetam in the presence of glutamate (AMPA + Ani + Glu) protected most of the neurons (> 97%) against glutamate excitotoxicity, whereas pre-treatment with cycloheximide blocked this protective effect (AMPA + Ani + Cyclo + Glu) resulting in loss of ≈50% of the neurons (Fig. 1). Cycloheximide alone had no effect on neuronal survival and did not protect neurons against glutamate-mediated excitotoxicity (Cyclo + Glu). These data suggest a requirement for protein synthesis in the neuroprotective effect elicited by AMPA.

Figure 1.

Cycloheximide blocks AMPA-mediated neuroprotection. Cultured rat cerebellar granule cells on DIV 8 were exposed to cycloheximide (Cyclo; 1 µg/mL) for 1 h followed by the addition of AMPA (500 µm) + aniracetam (1 µm; AMPA + Ani) for 4 h. An excitotoxic concentration of glutamate (Glu; 100 µm) was then added and neuronal viability was quantified 24 h later using MTT. Data are expressed as the mean ± SD. This experiment was performed in triplicate and repeated three times (n = 9). **p < 0.01, glutamate versus untreated neurons. **p < 0.01, AMPA + Ani + Glu versus Glu; AMPA + Ani +Cyclo + Glu versus AMPA + Ani + Glu.

AMPA + aniracetam increase BDNF mRNA levels

We tested the hypothesis that AMPA receptor activation promotes survival against glutamate excitotoxicity by enhancing BDNF mRNA expression and release of BDNF. Our first approach was to measure AMPA-mediated BDNF mRNA using the quantitative 5′ nuclease assay (Taqman®). While four promoters of the rat BDNF gene are known to be active in brain (Timmusk et al. 1993), and that the rat BDNF gene is composed of at least four 5′ untranslated exons (1–4), each of the primary transcripts are differentially spliced to a single 3′ coding exon that encodes the entire sequence for the mature polypeptide. The primers and probes we used were designed to detect the coding exon 5-containing transcripts, which is common to all spliced forms of BDNF mRNAs. Addition of AMPA in the presence of aniracetam increased BDNF-specific mRNA levels resulting in a broad peak of activation beginning within 30 min and lasting at least up to 6 h (Fig. 2). This increase is blocked by the AMPA receptor antagonist, NBQX, suggesting that it is specifically mediated by activation of AMPA receptors (see Results of inhibitors below and Fig. 10B).

Figure 2.

AMPA + aniracetam increases BDNF-specific mRNA levels. Cerebellar granule cells were treated with AMPA (500 µm) + aniracetam (1 µm) for various times (0–24 h). At the indicated time, the medium was removed and the cells disrupted with 5 m guanidinium isothiocyanate and total RNA isolated as described previously (Marini et al. 1998). BDNF-specific mRNA levels were quantified using the 5′ nuclease assay as outlined. Note that there are some peaks of BDNF expression. This experiment was performed in triplicate and repeated 13 times (n = 39). *p < 0.01 and **p < 0.001 compared to untreated cells.

Figure 10.

Induction of BDNF mRNA by AMPA + aniracetam is blocked by LY294002 and U0126. Cultured cerebellar granule cells were treated with medium alone or in presence of AMPA (500 µm) + aniracetam (1 µm) for 4 h in the absence or presence of either LY294002 (a, 10 µm) or U0126 (b, 10 µm). The cells were harvested and RNA isolated. Exon 5-specific BDNF mRNA levels were quantified using the 5′ nuclease assay as outlined in Methods. AMPA + aniracetam (AMPA + Ani) reproducibly increased BDNF mRNA at least 2-fold. NBQX was incorporated as an additional control to show that the induction of BDNF mRNA was strictly mediated by AMPA receptors. Addition of either NBQX (5 µm) or LY294002 (10 µm) did not significantly affect basal BDNF-specific mRNA levels compared with the untreated control. There was a small but statistically significant reduction in baseline exon 5-specific BDNF mRNA levels in the presence of U0126 (b) compared to untreated neurons. The increase in BDNF mRNA by AMPA + aniracetam is abolished by U0126, and LY294002. These results suggest that the PI3-K-dependent increase of the MAPK pathway regulates BDNF gene expression. The fluorescence of the mean of each experiment was compared with glyceraldehydes phosphate dehydrogenase (GAPDH) mRNA and is expressed as relative fluorescence. Each experiment was performed at least in triplicate using two different preparations of neurons (n = 9). **p < 0.01, AMPA + Ani versus untreated cells and LY294002 + AMPA + Ani versus AMPA + Ani. **p < 0.01, NBQX + AMPA + Ani versus AMPA + Ani, and U0126 + AMPA + Ani was compared with AMPA + Ani. *p < 0.05 compared with untreated cells.

AMPA + aniracetam evokes the release of BDNF protein

To examine whether AMPA + aniracetam increased the release of BDNF from the neurons, culture medium was collected at different time intervals, concentrated, and BDNF protein levels were determined using a two-site immunoassay. The BDNF level in culture medium from untreated neurons was approximately 50 pg/mL, which was well above the detection limits of this assay. Analysis of BDNF levels in the culture medium from AMPA + aniracetam-treated neurons showed an increase of about 1.5-fold at 5 min (Table 1). BDNF levels were further increased by AMPA + aniracetam at 1, 3 and 6 h in medium from AMPA + aniracetam-treated neurons, indicating that AMPA elicits a time-dependent increase in BDNF protein release compared to media from untreated neurons.

Table 1.  Activation of AMPA receptors increases BDNF protein levels in rat cerebellar granule cell medium
TreatmentBDNF (pg/mL)
  1. Cerebellar granule cells were treated for 5 min, 1 h, 3 h, and 6 h with culture medium either alone or containing AMPA (500 µm) + aniracetam (1 µm). Culture medium was collected, concentrated (Centricon) to 100 µL, and assayed in duplicate using the BDNF two-site immunoassay (Promega, Madison, WI). Data are expressed as mean ± SEM of at least two independent experiments (n = 6). *p < 0.05 versus untreated (Student's t-test).

Untreated 5 min47.6 ± 3.1
AMPA + aniracetam 5 min79.2 ± 13*
Untreated 1 h67.9 ± 5.9
AMPA + aniracetam 1 h123.8 ± 22*
Untreated 3 h71 ± 6.3
AMPA + aniracetam 3 h131 ± 8.4*
Untreated 6 h92.7 ± 4.8
AMPA + aniracetam 6 h158 ± 9.6*

AMPA enhances TrkB receptor activation

We showed previously that when BDNF (100 ng/mL) was added to culture medium of cerebellar granule cells for 24 h, about 30% of the neurons were protected from an excitotoxic concentration of glutamate (100 µm) acting on NMDA receptors (Marini et al. 1998). Neither nerve growth factor (NGF; 100 ng/mL) nor neurotrophin-3 (NT-3; 100 ng/mL) were effective in preventing glutamate excitotoxicity, which was not surprising as granule cell neurons express TrkB receptors, the cognate receptor for BDNF (Marini et al. 1998). We also showed that cerebellar granule cells only express functional TrkB receptors (Marini et al. 1998). To determine whether AMPA receptor survival-promoting activity involved a BDNF autocrine loop, TrkB-tyrosine phosphorylation was examined after AMPA + aniracetam treatment. A maximum neuroprotective concentration of AMPA (500 µm) in the presence of aniracetam (1 µm) increased TrkB-tyrosine phosphorylation above basal levels within 10 min and a further increase in TrkB-tyrosine phosphorylation is observed at 1 h (Fig. 3). AMPA + aniracetam increased TrkB phosphorylation by 50% ± 2% (mean ± SEM) over baseline by scanning densitometry (data not shown).

Figure 3.

AMPA + aniracetam activates TrkB receptor phosphorylation in cerebellar granule cells. Cerebellar granule cells were exposed to medium alone (c), BDNF (b, 100 ng/mL) for 10 min, or AMPA (500 µm) + aniracetam (1 µm; AMPA + Ani) for 10 min (10′) and 1 h (1 h). The neurons were then lysed and protein precipitated with pan-trk antibody before immunoblotting with a phosphospecific Trk antibody. Upper panel, tyrosine phosphorylation of TrkB induced by BDNF and AMPA + aniracetam is indicated by the arrowhead. Lower panel, the blot was stripped and re-incubated with an antibody specific for the extracellular domain of TrkB to confirm that similar amounts of TrkB proteins were present in each lane.

AMPA + aniracetam neuroprotection requires activation of TrkB receptors

Addition of AMPA (500 µm) + aniracetam (1 µm) protected all of the vulnerable neurons against the excitotoxic effects of glutamate (100 µm; Fig. 4). However, pre-treatment with K252a (10 nm), a selective inhibitor of Trk receptor activity (Lee et al. 2002), completely blocked the neuroprotective activity of AMPA + aniracetam, suggesting that activation of TrkB receptors was required for AMPA to mediate its neuroprotective effect (Fig. 4). However, K252a itself did not adversely effect neuronal survival (data not shown), indicating that TrkB activation was not required for basal survival. Further, addition of K252a to medium containing AMPA + aniracetam (data not shown) or glutamate did not affect neuronal survival supporting the observation that K252a had no direct effect on basal survival (Fig. 4).

Figure 4.

AMPA + aniracetam requires activation of TrkB receptors. Cultured cerebellar granule cells were treated with AMPA + aniracetam (1 µm; AMPA + Ani), and K252a (10 nm) + AMPA (500 µm) + aniracetam (1 µm; k252a + AMPA + Ani). K252a was added 24 h prior to addition of drugs. AMPA + aniracetam were always exposed to neurons for 4 h followed by the addition of glutamate (Glu; 100 µm) for 24 h. Data are presented as the percent neuronal survival (mean ± SD; n = 12). **p < 0.01, glutamate versus untreated cells. **p < 0.05, AMPA + Ani + Glu versus Glu alone; k252a + AMPA + Ani + Glu versus AMPA + Ani + Glu.

AMPA receptor-mediated neuroprotection is attenuated by the BDNF-blocking antibody

To confirm the initial finding that AMPA protects neurons through BDNF, we employed a BDNF-blocking antibody (see Materials and methods). Pre-treatment of cerebellar granule cell neurons with this antibody (1 µg/mL for 3 h) attenuated AMPA (500 µm) + aniracetam (1 µm)-mediated neuroprotection (Fig. 5). In contrast, treatment of the neurons with the identical immunoglobulin concentration did not affect basal neuronal survival or survival of neurons treated with an excitotoxic concentration of glutamate. Taken together, these data support the K252a results suggesting that BDNF released into the medium following AMPA receptor activation protects neurons against glutamate excitotoxicity.

Figure 5.

The BDNF-blocking antibody attenuates AMPA + aniracetam-mediated neuroprotection. Cerebellar granule cells were exposed to medium alone or medium containing chicken IgG (IgG, 1 µg/mL), or the BDNF-blocking antibody (BDNF Ab; 1 µg/mL) for 3 h. Neurons were also treated with glutamate (Glu, 100 µm) alone for 24 h. Cells were exposed to AMPA (500 µm) + aniracetam (1 µm) for 4 h in the presence or absence of the BDNF-blocking antibody followed by the addition of glutamate. Neuronal viability was determined 24 h later using the MTT assay. Note that the BDNF-blocking antibody attenuated the protective effect mediated by AMPA + aniracetam. Data are expressed as percent neuronal survival (mean ± SD; n = 9). **p < 0.01, Glutamate was compared with untreated cells. **p < 0.01, AMPA + Ani + Glu versus Glutamate. *p < 0.05, BDNF Ab + AMPA + Ani + Glu versus AMPA + Ani + Glu.

Effect of src-tyrosine kinase, MAPK and PI3-K inhibitors on the neuroprotective activity of AMPA + aniracetam

We used selective inhibitors to determine the role of various signal transduction pathways in AMPA + aniracetam-mediated neuroprotection. To determine whether the activation of Lyn, the MAPK, and the PI3-K pathways mediated signals critical for AMPA receptor neuroprotection, we used PP1 a src-tyrosine protein kinase inhibitor (6 nm; Hanke et al. 1996; Lee et al. 2002), U0126, a MAPK kinase (MEK) inhibitor (10 µm) (Nicole et al. 2001), and LY294002 a PI3-K inhibitor (10 µm; Dudek et al. 1997). The addition of AMPA (500 µm) + aniracetam (1 µm) protected vulnerable granule cell neurons against glutamate toxicity. However, pre-treatment with PP1 (Fig. 6a), Ly294002 (Fig. 6b) or U0126 (Fig. 6c) blocked AMPA + aniracetam-mediated neuroprotection. None of the inhibitors alone affected basal neuronal survival (data not shown). None of the inhibitors affected neuronal survival in the presence of glutamate or AMPA + aniracetam (data not shown), indicating a lack of a direct effect. Taken together, these results support the idea that activation of Lyn, the PI3-K, and the MAPK pathways can protect neurons against glutamate-mediated excitotoxicity.

Figure 6.

The src-tyrosine kinase Lyn, MAPK, and PI3-K pathways mediate AMPA + aniracetam survival to protect neurons against glutamate toxicity. Cultured cerebellar granule cells were pre-treated with PP1 (a, 6 nm), LY294002 (b, 10 µm) or U0126 (c, µm), for 1 h. AMPA (500 µm) + aniracetam (1 µm) were then added followed by the addition of an excitotoxic concentration of glutamate (100 µm) for 24 h. Neuronal viability was quantified using the MTT assay. Individually, the inhibitors did not affect basal survival (data not shown) or toxicity mediated by glutamate (n = 12). **p < 0.01, glutamate versus untreated cells, AMPA + Ani + Glu versus glutamate alone, and the inhibitors (PP1, LY294002, U0126) + AMPA + Ani + Glu were compared to AMPA + Ani + Glu.

AMPA + aniracetam activates a signaling pathway involving the src-tyrosine kinase Lyn, the PI3-K pathway, and ERK1/2, a component of the MAPK cascade

Based on the inhibitor findings, we determined whether AMPA receptors activate Lyn, a src-tyrosine protein kinase, ERK1/2, components of the MAPK pathway, and Akt, a downstream target of the PI3-K pathway. AMPA (500 µm) + aniracetam (1 µm) resulted in the activation of Lyn within 5 min (Fig. 7a). Activation of Akt, and ERK1/2 by AMPA + aniracetam occurred within 10 min and lasted at least 6 h (Figs 7b and c). The time course of ERK1/2 and Akt activation for BDNF is shown for comparison and the quantitation of pERK2 and pAkt are shown in each of the lower panels (Figs 7d and e). Increases in pERK1 were also observed (data not shown). Although the extent of the increase for pERK2 and pAkt is higher for BDNF, the temporal profiles between BDNF and AMPA + aniracetam are remarkably similar. These results provide additional support for the role of BDNF release by AMPA + aniracetam-mediated neuroprotection.

Figure 7.

PA + aniracetam activate the src-tyrosine kinase lyn, Akt, and ERK1/2. Cultured cerebellar granule cells were treated with AMPA (500 µm) + aniracetam (1 µm) for the indicated times. The cultures were lysed and immunoblotted with the specific antibody as described in the Methods section. For the Lyn experiments, cultured cerebellar granule cells (DIV 8) were treated with medium alone (no addition), or in the presence of AMPA (500 µm) + aniracetam (1.0 µm;[MPA +Ani)for 5 and 10 min. The blots were probed with antiphospho (activated) Lyn (PLyn), antitotal Lyn (Lyn) and antiactin (Actin) antibodies (a). The activation of Akt by AMPA + aniracetam (b) and BDNF (d) in cultured cerebellar granule cells was determined at various time intervals. Following transfer to nitrocellulose membranes the immobilized proteins were probed with antiphospho (activated) Akt (pAkt), and antitotal Akt (Akt). The time course of ERK1/2 activation by AMPA + aniracetam (c) and BDNF (e) were performed in an identical fashion with the exception that antiphospho (activated) ERK1/2 (pERK1/2), antitotal ERK1/2 (ERK1/2) were used to probe the membrane. Quantitation of the activation of pAkt and pERK for AMPA + as well as BDNF is shown in the lower panels of each respective figure. Each data point represents the percent pAkt or pERK2 compared to control from two independent experiments performed in duplicate (n = 4). *p < 0.05 and **p < 0.01 by Student's t-test.

Effect of the PI3-K pathway inhibitor, LY294002, on activation of ERK1/2

Previous results showed that the PI3-K pathway activates the MAPK pathway upstream of Sos and Ras (Hawes et al. 1996; Lopez-Ilasaca et al. 1997). It is also known that neurotrophins are able to regulate neuronal survival through the PI3-K and MAPK pathways, although the relative contributions of these pathways differ for different types of neurons (Hetman et al. 1999; Vaillant et al. 1999). We therefore tested the hypothesis that the AMPA + aniracetam–BDNF autocrine loop may protect neurons through a PI3-K-dependent activation of MAPK that leads to enhanced BDNF mRNA levels to increase the release of BDNF protein over time. Pre-treatment with LY294002 blocked the ERK1/2 activation by AMPA + aniracetam; LY294002 did not affect baseline levels of pERK2 (Fig. 8a). Wortmannin, another PI3-K inhibitor, also blocks AMPA + aniracetam-induced activation of ERK1/2 (Fig. 8b). These data indicate a role for PI3-K in the activation of ERK1/2 possibly by affecting BDNF release and/or gene transcription.

Figure 8.

Activation of the PI3-K pathway is required to increase ERK1/2 phosphorylation by AMPA + aniracetam. Cultured rat cerebellar granule cells were treated with medium alone (lane 1) or in the presence of AMPA (500 µm) + aniracetam (1 µm; lanes 2–3) for 10 min in the absence (lane 2) or presence of LY 294002 (20 µm) added 1 h prior to the addition of AMPA + aniracetam (lane 3). Cells were extracted and immunoblotted with active (phosphorylated) ERK1/2 antibody (Thr202/Tyr204; 1 : 1000) and then incubated with horseradish peroxidase-conjugated goat IgG (1 : 2000). Immunoactive bands were visualized using a chemiluminescence method (upper panel). The blot was stripped and reprobed with anti-ERK1/2 (1 : 1000) and then incubated with horseradish peroxidase-conjugated goat IgG (1 : 2000) followed by detection by chemiluminescence. Quantitation of the results is shown in the lower panel. Data are represented as the percentage pERK2 (phosphoERK2) compared to control from two independent experiments (n = 4). *p < 0.05 and **p < 0.01 by Student's t-test. The identical experiment was performed with the exception that Wortmannin (30 µm) was used to inhibit the PI3-K pathway (b). Wortmannin also inhibited the activation of ERK1/2 mediated by AMPA (500 µm) + aniracetam (1 µm).

Effects of the PI3-K and MAPK pathways on BDNF release from cultured cerebellar granule cells

Because neurotrophins activate the PI3-K and MAPK pathways (Kaplan and Miller 2000), we determined whether the PI3-K or MAPK inhibitors affected BDNF release in the presence or absence of AMPA + aniracetam. Pre-treatment with the MAPK inhibitor U0126 itself slightly increased (1.5-fold) BDNF levels in the culture medium at 5 min, the initial phase of the time course, but did not affect BDNF release after 6 h (Fig. 9). An initial increase in BDNF release at 5 min was also observed in U0126 + AMPA + aniracetam-treated neurons but this was not significantly different than AMPA + aniracetam-treated cultures (Fig. 9). In addition, U0126 pre-treatment failed to alter BDNF levels after 6 h. However, LY294002, the PI3-K inhibitor, reduced AMPA receptor-mediated BDNF release significantly at 5 min and 6 h (Fig. 9). These results suggest that the PI3-K pathway is required for the activity-dependent release of BDNF, whereas the MAPK is not. These data support the BDNF-blocking antibody findings that release of BDNF is required for AMPA-mediated neuroprotection.

Figure 9.

The effect of U0126 and LY294002 on BDNF release. Cultured cerebellar granule cells were treated with U0126 (10 µm) or LY294002 (20 µm) for 1 h and then either alone or in combination with AMPA (500 µm) + aniracetam (1 µm) for 5 min (upper panel) or 6 h (lower panel). The culture medium (4 mL) from two dishes of three separate experiments was collected, concentrated and assayed for BDNF protein. The data are expressed as percentage control ± SD. *p < 0.05 control versus AMPA + Ani or U0126. #p < 0.01 AMPA + Ani versus LY294002 + AMPA + Ani; n = 6.

Inhibition of the PI3-K and MAPK pathways block AMPA receptor-mediated increases in BDNF gene expression

The neuroprotective activity of AMPA receptors may activate the ERK1/2 and PI3-K pathways to increase BDNF gene expression. The addition of AMPA (500 µm) + aniracetam (1 µm) enhanced BDNF mRNA levels by about twofold (Fig. 10). Pre-treatment of cultured neurons with either the inhibitor LY294002 (10 µm; Fig. 10a) or the MEK inhibitor U0126 (10 µm; Fig. 10b) completely blocked the increase in BDNF mRNA levels. Addition of LY294002, the PI3-K inhibitor, or the AMPA/kainate antagonist, NBQX, did not affect baseline BDNF mRNA levels, indicating that AMPA receptor activation is not required for basal expression of BDNF mRNA. In contrast, there was a small but statistically significant decrease in basal expression of BDNF mRNA in neurons treated with U0126 (Fig. 10b).


In the present study, we show for the first time that activation of ionotropic AMPA receptors protect cultured granule cell neurons by releasing BDNF protein in the culture medium, which in turn activates its receptor, TrkB in an autocrine fashion. This autocrine loop leads to the activation of the downstream survival-promoting intermediates Lyn, Akt, and ERK1/2. We suggest that AMPA + aniracetam exerts its neuroprotective effects via the increased synthesis and release of BDNF into the culture medium, which then activates TrkB to promote survival in response to the glutamate challenge. Addition of a BDNF-blocking antibody markedly reduced this protective effect, and provides additional support for this hypothesis. We have previously demonstrated that a BDNF autocrine loop is a major neuroprotective mechanism used by ionotropic NMDA glutamate receptors against glutamate toxicity (Marini et al. 1998). Thus, both classes of ionotropic glutamate receptors appear to use a common mechanism to protect themselves against glutamate toxicity.

Our results demonstrate that activation of AMPA receptors in the presence of aniracetam evokes a time-dependent increase in BDNF protein in the culture medium. The release of BDNF in untreated neurons likely reflects the activation of ionotropic glutamate receptors by glutamate in the culture medium that is released by the neurons under depolarizing conditions (unpublished results). We show an early (by 5 min) increase in BDNF protein in the neuronal culture medium after addition of AMPA + aniracetam. This accumulation is greater by 1 h and is associated with an increase in BDNF mRNA within 30 min. These results suggest that, in addition to release, AMPA + aniracetam affects BDNF gene expression. The relatively rapid increase in BDNF mRNA levels (30 min) may be due to pre-formed mRNA that is stabilized through an unidentified mechanism (Mobarak et al. 2000). Thus, AMPA + aniracetam elicits an early and a late phase of BDNF protein release coupled with an increase BDNF mRNA. These results are similar to those obtained with NMDA receptor-mediated neuroprotection in the rapid release of BDNF protein, although the increase in BDNF mRNA occurred at 3 h (Marini et al. 1998). Interestingly, the PI3-K inhibitor LY294002 attenuated BDNF release, which provides additional data supporting a role for BDNF release in protecting neurons by AMPA + aniracetam. This result is also consistent with its inhibitory effect on BDNF gene expression. Curiously, U0126 blocked AMPA neuroprotection without affecting release even though it effectively blocks BDNF gene expression. It is possible that AMPA + aniracetam increases BDNF levels above baseline and U0126 blocks this increase. However, the effect of U0126 on total BDNF levels after administration of AMPA + aniracetam is unknown.

We show that a maximum neuroprotective concentration of AMPA + aniracetam elicits a rapid increase in TrkB phosphorylation (within 10 min). The increase in TrkB phosphorylation is blocked by the AMPA/kainate receptor antagonist NBQX (results not shown). Because AMPA is a highly selective agonist of AMPA receptors (Dingledine et al. 1999), our results suggest that the enhanced TrkB receptor activation is specifically activated by BDNF, which is released by AMPA receptor activation. Furthermore, K252a, blocked AMPA + aniracetam-mediated neuroprotection, and addition of the BDNF-blocking antibody also attenuated AMPA receptor-mediated neuroprotection. Taken together, these results suggest that BDNF plays a major role in AMPA + aniracetam-induced neuroprotection.

The signal transduction pathway by which AMPA + aniracetam mediates the increase in BDNF gene expression is unknown. It has been shown previously that the activation of MAPK by AMPA occurs in a PI3-K-dependent fashion (Perkinton et al. 1999). Thus, these two pathways may play a role in the increase in BDNF gene expression mediated by AMPA + aniracetam. We first showed that AMPA + aniracetam activates the src-tyrosine kinase Lyn, which is consistent with previous findings that Lyn is activated by and associates with AMPA receptors (Hayashi et al. 1999). In addition, pre-treatment of granule cell neurons with the src-tyrosine kinase inhibitor, PP1, blocks AMPA + aniracetam-mediated neuroprotection, suggesting that the association of Lyn with AMPA receptors may play an important role in the neuroprotective activity of AMPA by increasing BDNF mRNA (Hayashi et al. 1999) or by anchoring AMPA receptors through its known interaction with cytoskeletal proteins in the cerebellum (Umemori et al. 1992; Helmke and Pfenninger 1995; Prinetti et al. 2001). The neuroprotective serine/threonine protein kinase, Akt, a downstream target of the PI3-K pathway, is also activated, as is ERK1/2, a component of the ras/MAPK pathway, which may promote survival through its known effect on the antiapoptotic protein, Bcl-2 (Liu et al. 1999). Activation of this pathway is likely mediated through the activation of TrkB receptors by BDNF release (Kaplan and Miller 2000). We also show a selective inhibitor of the PI3-K pathway, LY294002, blocked the neuroprotective activity of AMPA. We demonstrated that LY294002 and Wortmannin, another PI3-K inhibitor, blocked the AMPA + aniracetam-mediated increase in ERK1/2, again consistent with previous findings describing a PI3-K-dependent activation of MAPK (Perkinton et al. 1999). Our results also suggest that these two pathways play a role in promoting the survival of in cerebellar granule cells against glutamate excitotoxicity.

We hypothesize that PI3-K and ERK1/2 may play a role in the increase in BDNF gene expression. To test this hypothesis, we quantified BDNF mRNA levels in AMPA + aniracetam-treated neurons in the presence or absence of the identical inhibitors shown to block the neuroprotective activity. We show that pre-treatment with either the MEK or PI3-K inhibitor abolished the AMPA + aniracetam-mediated increase in BDNF mRNA. These results suggest that activation of the PI3-K and MAPK pathways increases BDNF mRNA. These results are in accordance with previous studies where calcium influx increases BDNF mRNA levels by activating the transcription factor CREB (Tao et al. 1998). Both PI3-K and MAPK via the appropriate downstream intermediates (Akt and Rsks, respectively) are capable of phosphorylating or activating CREB (Du and Montminy 1998; Riccio et al. 1999). Whether the PI3-K pathway is required to activate the MAPK pathway to activate BDNF gene expression requires further study.

AMPA receptor-mediated signaling can directly increase BDNF mRNA levels through CREB activation. However, it is plausible that other transcription factors are also likely to be activated by AMPA. Thus, the release of BDNF and activation of TrkB is likely to be at least partly responsible for activating BDNF gene expression. This idea comes from our own work showing that TrkB activation leads to increased NF-κB DNA binding activity in BDNF-treated cerebellar granule cells, an event that is critical for NMDA receptor-mediated neuroprotection (Lipsky et al. 2001). In addition, BDNF mRNA expression induced by kainic acid, which increases intracellular calcium levels through AMPA/kainate receptors, in transgenic mice overexpressing a dominant negative form of TrkB was significantly lower in the transgenic mice compared to their wild-type littermates (Saarelainen et al. 2001). These data support the idea that activation of TrkB receptors, at least partly, regulates the BDNF gene by a positive feedback mechanism.

In summary, AMPA receptors regulate BDNF mRNA levels through a signal transduction mechanism that involves contributions from AMPA and TrkB receptor activation. The activation of AMPA receptors results in the release of BDNF, which in turn binds to and activates TrkB receptors in an autocrine fashion. This autocrine loop activates a signal transduction pathway involving a PI3-K-dependent increase of the MAPK pathway to enhance BDNF gene transcription to protect vulnerable neurons against glutamate-mediated excitotoxicity.