FGF-2 Potentiates Ca2+-Dependent Inactivation of NMDA Receptor Currents in Hippocampal Neurons

Adam L. Boxer, Herman Moreno, Bernardo Rudy, Edward B. Ziff


Peptide growth factors such as the neurotrophins and fibroblast growth factors have potent effects on synaptic transmission, development, and cell survival. We report that chronic (hours) treatment with basic fibroblast growth factor (FGF-2) potentiates Ca2+-dependentN-methyl-d-aspartate (NMDA) receptor inactivation in cultured hippocampal neurons. This effect is specific for the NMDA-subtype of ionotropic glutamate receptor and FGF-2. The potentiated inactivation requires ongoing protein synthesis during growth factor treatment and the activity of protein phosphatase 2B (PP2B or calcineurin) during agonist application. These results suggest a mechanism by which FGF-2 receptor signaling may regulate neuronal survival and synaptic plasticity.


At excitatory, glutamatergic synapses, release of the neurotransmitter glutamate stimulates a family of ionotropic glutamate receptors. These receptors are transmembrane proteins that are assembled from four to five structurally related integral membrane subunits. One member of this family is theN-methyl-d-aspartate receptor (NMDAR), a cation channel that is activated by simultaneous binding of the ligand, glutamate, and depolarization of the postsynaptic cell (reviewed byHollmann and Heinemann 1994). The receptor is composed of a ubiquitous NR1 subunit and one of four NR2 subunits termed NR2A-D (reviewed by Mori and Mishina 1995). Under physiologically normal conditions, entry of Ca2+ions through the NMDAR triggers biochemical and electrophysiologic changes that underlie synaptic plasticity. However, overactivation of NMDARs, which may follow anoxia or glucose deprivation, has pathological consequences and is an important component of a large number of CNS disorders including excitotoxic cell death (Lipton and Rosenberg 1994). Excitotoxic death is thought to result from excessive Ca2+ influx through the receptor following pathological receptor activation (reviewed by Mody and MacDonald 1995).

Peptide growth factors are regulators of CNS development, synaptic plasticity, and neuronal survival after ischemia (reviewed inBaird 1994; Eckenstein 1994; Lewin and Barde 1996) and they exert both short and long term effects on synaptic physiology (reviewed in Finkbeiner 1996;Katz and Shatz 1996; Thoenen 1995). The expression of the growth factor, basic fibroblast growth factor (FGF-2), in the CNS increases during early postnatal development (Caday et al. 1990; Kuzis et al. 1995). In the adult CNS, the concentration of FGF-2 exceeds those of certain neurotrophins, such as nerve growth factor (NGF) (Eckenstein 1994). These findings suggest that FGF exerts important long-term effects in the adult as well.

Studies of excitotoxic cell death show that long term FGF-2 treatment can prevent NMDAR-mediated cell death both in vitro (Férnandez-Sánchez and Novelli 1993;Mattson et al. 1989) and in vivo (Fisher et al. 1995; Kirschner et al. 1995; Koketsu et al. 1994; MacMillan et al. 1993; Nozaki et al. 1993a,b). FGF can also regulate the internal cellular Ca2+ concentration after NMDAR activation (reviewed in Lindholm 1994), and its neuroprotective effects correlate with an attenuation of NMDAR-induced increases in [Ca2+]i (Cheng and Mattson 1991; Mattson et al. 1989) and with the expression of a putative NMDAR protein (Mattson et al. 1993) and the Ca2+-binding protein, calbindin D28 (Mattson et al. 1991; Peterson et al. 1996). However, it is not known if FGF-2 affects NMDAR electrophysiologic function.

To determine whether FGF-2 can regulate the function of the NMDAR ion channel, we have carried out electrophysiological and molecular biological studies in cultured rat hippocampal neurons. Treatment with FGF-2 for hours to days significantly enhanced the ability of extracellular Ca2+ to inactivate NMDAR currents. Enhancement of receptor inactivation required protein synthesis during FGF-2 treatment, and the activity of calcineurin during agonist application, but did not involve changes in NMDAR protein subunit expression.


Cell culture

Hippocampal cultures were prepared similarly to those used byMattson et al. (1991). E18 Sprague–Dawley embryos (pregnant rats obtained from Taconic) were dissected from rats anesthetized with pentobarbital sodium and placed in ice-cold, Ca2+/Mg2+-free phosphate-buffered saline (PBS), containing 0.6% (wt/vol)d-glucose. Hippocampi were dissected and incubated in 1 × trypsin/EDTA (Gibco) plus 10 μg/ml pancreatic DNAse I (Boehringer) at room temperature for 15 min. Cells were dissociated by trituration through a series of fire-polished pasteur pipets and seeded on polyornithine (50–100 μg/ml, [MW 30,000–70,000], Sigma)/laminin (10 μg/ml; natural mouse, Gibco) at ∼1 × 106 cells per 60 mm dish. Culture medium consisted of minimum essential medium containing Earle's salts andl-glutamine (Gibco No. 11095-080), supplemented with 10% heat-inactivated (30[min], 56°C) fetal bovine serum, 20 mM KCl, 1 mM Na Pyruvate, 1% (vol/vol) penicillin-streptomycin, and 0.6% (wt/vol) d-glucose and was changed 4 h and 3 days after plating. Mitotic inhibitors were not added because these became neurotoxic in the presence of FGF-2. Cells were grown in a 5% CO2-98% humidity incubator in 3 ml of media. In some initial experiments, medium was changed to neurobasal medium (Gibco) plus 1 × B27 supplement (Gibco) and 0.5 mMl-glutamine after 4 h. Growth in this medium suppressed glial cell growth in control cultures but not in cultures grown in the presence of FGF-2. Results obtained in neurobasal/B27 were similar to those in serum containing medium. Growth factors were added directly to the medium from stock solutions prepared and stored as suggested by the manufacturer [R + D Systems for BDNF, FGF-2 (157aa. isoform) and NGF; Promega or R + D systems for NT-3]. Cultures were used at 8–10 days in vitro, at which time ∼40% of cells were glia (as judged by GFAP immunofluorescence), and little neuronal cell death was evident.


Recording electrodes were formed on a Narishige PP-83 puller from 1.5 mm diam hard glass (A-M Systems, No. 6030) and polished on a Narishige MF-83 microforge. Pipet solution contained (in mM) 155 Cs MeSO3, 10 Cs HEPES, 5 BAPTA, 0.4 CaCl2, 2 MgCl2, and 5 Na ATP (pH 7.3 with CsOH). Standard ECS contained (in mM) 167 NaCl, 2.4 KCl, 10 Na HEPES, 1 CaCl2, 0.01–0.02 glycine, and 10 d-glucose (pH 7.3 with NaOH). In 5 mM CaCl2 ECS, NaCl concentration was reduced to 157 mM. In some experiments 1 μM TTX (Research Biochemicals, RBI) was included, and did not produce any observable effect on glutamate receptor currents. When filled with pipet solution, electrodes had resistances of 3–7 MΩ. Typical seal resistances were 2–4 GΩ. Whole cell recordings (Hamill et al. 1981) were carried out at 22–24°C in tissue culture dishes on the stage of an inverted microscope with the use of an Axopatch 200A amplifier (Axon Instruments). Neurons were identified on the basis of morphology and voltage-activated sodium conductances at the beginning of each recording (see below). Series resistance was 70–90% compensated and most of the capacitance was cancelled. Access resistance was monitored with a 20-ms, 10-mV hyperpolarizing pulse at the start of each episode. Currents were filtered at 2 kHz, digitized, stored on hard disk, and analyzed with pClamp 6 software (Axon Instruments). Curve fitting was done by Chebeshev method using pClamp 6 software or least squares method using DeltaGraph Pro, with equations taken from pClamp 6 manual. Statistical analysis was performed with StatView 4.5 (Abacus Concepts) on a Macintosh computer. Data are expressed as the mean ± SE.

Agonist and antagonist application

Recordings were carried out in 6 cm tissue culture dishes and were perfused continuously by peristaltic pump at a rate of 2–3 ml/min (total bath volume was ∼2 ml). Agonists and antagonists were dissolved in extracellular solution and applied by pressure ejection (1–300 mmHg) from a flow pipe ∼50 μm from the cell, using a computer-controlled solenoid valve, 12 reservoir drug application device (DAD-12; Adams and List Associates, Westbury, NY). The solution exchange time was estimated to be ∼200–300 ms based on the rise time of kainate-evoked currents and perfusion of 10% normal ECS. After gaining whole cell access, cells were held at −90 mV and stepped in 10 mV increments to +30 mV to verify neuronal phenotype (large Na-channel current, not shown). All cells were also subject to an 100 μM kainateI-V analysis to verify cell type dependence (Ozawa et al. 1991). Growth factors were not present in the ECS except as indicated. Extracellularly applied drugs were present only during the inactivation protocol (see below), except cycloheximide which was present in the bath, wash and inactivation solutions, but not in the agonist solutions. Intracellularly dialyzed drugs were present throughout the experiment. All agonists and antagonists were obtained from RBI except ω-agatoxin IVA (Alamone Labs, Jerusalem), cycloheximide (Sigma), calcineurin autoinhibitory peptide (Bachem), control calcineurin inhibitory peptide (generous gift of B. Perrino and T. Soderling, Vollum Institute), and phalloidin (Boehringer).

Molecular biology

RT-PCR analysis of NMDAR subunit mRNA was carried out by the method of Sheng et al. (1994). Total RNA was isolated from rat tissue or cultured cells by the guanidinium isothiocyanate/SDS method and digested with RNase-free DNAse I (Boehringer, 1 U/2.5 μg) for 1 h at 37°C and phenol/chloroform extracted, ethanol precipitated, and resuspended in water. cDNA was synthesized with the Invitrogen cDNA cycle kit. cDNA reaction (1–2 μl) was included in 50 μl PCR reactions containing 60 mM Tris · Cl, 15 mM NH4Cl, 1–7 mM MgCl2(empirically optimized), and 2 mM each dNTP, plus 250 nM 5′ and 3′ oligo and 1.5 U Amplitaq (Perkin-Elmer) polymerase. In some reactions, 8.3 μCi alpha 32P -dCTP was included. The number of amplification cycles to remain in the linear range was empirically determined for each oligonucleotide pair. Reactions were visualized on 1.8% agarose gels, stained with ethidium bromide, or for phosphorimager (Molecular Dynamics) quantitation, on 4% nondenaturing polyacrylamide gels. Identity of bands was verified by size and by in situ hybridization of adult rat brain sections (C. Kentros, A. Boxer, B. Rudy, and E. B. Ziff, unpublished observations). Oligonucleotide primers used to amplify NR1, N1, and C1 exons were identical to those described in Sheng et al. (1994). Neurofilament-L primers amplified nucleotides 1174–1719 of the rat sequence: 5′ oligo, TGGACATCGAGATTGCAGC; 3′ oligo, GGTTGGTGATGAGGTTGACC. NMDAR2A primers amplified nucleotides 4076–4453 of the rat sequence: 5′ oligo, GGATTAACCGACAGCACTCC; 3′ oligo, ATGATGCTTGACCTCAAGG. NMDAR2B primers amplified nucleotides 3960–4347 of the rat sequence: 5′ oligo, GCATTCCTACGACACCTTCG; 3′ oligo, GACCACCACTGGCTTATTGG.

For western analysis, cells were washed twice with PBS and harvested in ice-cold RIPA buffer containing protease and phosphatase inhibitors (Sheng et al. 1994) for 5–15 min. Protein concentration was determined by Bradford Assay (BioRad) and equal amounts were loaded onto SDS-polyacrylamide gels, run, and transferred to nitrocellulose by standard methods. Blots were probed with antibody and visualized using an ECL Kit (Amersham) according to manufacturer's directions. Antibodies were used according to manufacturer's directions as follows: mouse monoclonal anti-calbindin D28 and anti-neurofilament L (Sigma); rabbit polyclonal antisera to NMDAR1, NMDAR2A; and rabbit polyclonal NMDAR1, GluR1, GluR2/3, and GluR4 antisera from (Chemicon).


Cells were grown in eight-well, permanox chamber slides (Nunc), coated with polyornithine and laminin as above. Cells were washed twice with PBS and fixed in Vilim's fixative: 4% paraformaldehyde, 0.2% picric acid, 0.1 M Na2PO4, and 3% sucrose for 15–30 min. Cells were then washed three times in PBS for 5 min, permeabilized in 0.1% Triton X-100 + 0.1% BSA for 10–30 min, and blocked with 5% normal goat serum (Sigma) in PBS for 30 min or overnight. Immunocytochemistry was performed as directed with a Vectastain Elite immunoperoxidase kit. Antibodies were purchased from Chemicon.


Ca2+ concentration and activity-dependent inactivation of NMDA-evoked currents in FGF-2 treated cells

The NMDAR undergoes several forms of activity-dependent current inactivation (reviewed by Jones and Westbrook 1996). Two of these forms, glycine-dependent and glycine-independent desensitization, are induced by glutamate and NMDA in the presence of low or high concentrations of the co-agonist glycine, respectively. A third form of inactivation, Ca2+-dependent inactivation (CDI), is induced by the increase in [Ca2+]i (Legendre et al. 1993; Medina et al. 1996) that follows the entry of Ca2+ through the NMDAR receptor or other channels. CDI involves the Ca2+-dependent binding of calmodulin to the C0 domain of the NR1 C terminus and the displacement of α-actinin, which reduce NMDAR currents (Ehlers et al. 1996; Krupp et al. 1999; Wyszynski et al. 1997; Zhang et al. 1998). CDI is also sensitive to [Ca2+]o (Legendre et al. 1993).

To determine whether FGF-2 changes activity-dependent physiological properties of NMDARs, whole cell patch-clamp recordings (Hamill et al. 1981) were elicited from pyramidal shaped neurons that had been placed in culture for 8–11 days and subjected to different growth factor treatments (Fig. 1). Cells were analyzed that displayed 100 μM kainate-evoked I-Vprofiles characteristic of type I, glutamatergic cells (Fig.1 G) (Ozawa et al. 1991). NMDA-evoked currents were detected in all cells tested. Cultures were treated with 10 ng/ml FGF-2 for 24–48 h, or for controls were left untreated with growth factor, or were treated with 10 ng/ml BDNF for 72–120 h before recording.

Fig. 1.

Basic fibroblast growth factor (FGF-2) induces an accumulative inactivation of N-methyl-d-aspartate receptor (NMDAR) but not α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR). A: representative currents evoked from a V H = −60 mV in 1 mM Ca2+ ECS in a control cell and a cell treated with FGF-2 for 24 h by 3-s pulses of 50 μM NMDA at 2 min−1. Shown are the currents during the first pulse,I t=0s (+) and the pulses atI t=195s (▾) andI t=300s (⋄).B: normalized mean ± SE currents, measured as the average response during 3-s pulses at 2 min−1 for control cells (n = 11), FGF-2 (24–48 h)-treated cells (n = 8), and BDNF (120 h)-treated cells (n = 6). C: recovery of 50 μM NMDA-evoked currents in a representative FGF-2-treated cell (24 h). Cell was stimulated with 50 μM NMDA in 1 mM Ca2+ ECS for 10 min (as in A) (left traces) and stimulated with 50 μM NMDA in Ca2+-free (1.5 mM EGTA) ECS for 10 min (right traces). Arrows indicate progression of time. D: normalized mean ± SE currents in control (n = 3) and FGF-2 (n = 5) cells during stimulation with 3-s 50 μM NMDA pulses 2 min−1 in Ca2+-free ECS after 20 pulses in 1 mM Ca2+ (as in A). E: examples of currents evoked by 3-s application of 100 μM kainate in 2 mM Ca2+ ECS (2 min−1). FGF-2 cell treated for 24 h. ⋄, I t=0s; ▾, I t=1600s.F: normalized mean ± SE currents, measured as the average response during the 3-s 100 μM kainate application, normalized to the first response for experiments such as the one shown in E. G: traces are from 2 representative cells held at 0 mV and stepped from −70 mV to +50 mV in 10 mV increments with 1-s application of 100 μM kainate at each potential.

In the first experiments, we exposed FGF-2-treated cells, BDNF-treated cells, and control cells to a 10-min conditioning pulse train consisting of 3-s pulses of 50 μM NMDA in 1 mM Ca2+ ECS applied for 2 min−1. Each agonist application was preceded by a 20-ms, 10-mV hyperpolarizing pulse from the holding potential of −60 mV to monitor access resistance. In addition to CDI, decreases in NMDAR currents can take place in an irreversible, Ca2+-independent, ATP-sensitive manner (washout) (MacDonald et al. 1989; Wang et al. 1996), or in a reversible, Ca2+-dependent, ATP-sensitive manner (rundown) (Rosenmund and Westbrook 1993a,b). In this experiment a concentration of ATP (5 mM) sufficient to prevent washout and retard rundown was included in the pipet solution.

During the administration of the pulse train, all cells displayed an inactivation of NMDA-evoked current (average amplitude during 3 s pulses) whose magnitude increased with successive stimulations. However the inactivation was greater in FGF-2 treated cells than in control or BDNF-treated cells (Fig. 1, A and B). After 10 min of stimulation, control and BDNF-treated cell currents decreased to 75.2 ± 6.4% and 76.6 ± 4.9% of initial values, respectively [not significantly different, analysis of variance (ANOVA), Sheffé post hoc test], whereas currents in FGF-2-treated cells decreased to a much greater extent, to 31.5 ± 6.4% (P < 0.001) of initial values. This showed that FGF-2 treated cells displayed enhanced activity-dependent current inactivations relative to untreated or BDNF treated controls during a NMDA pulse train that were highly statistically significant at physiological concentrations of Ca2+ (1 mM Ca2+).

The decay of the NMDA-evoked currents in FGF-2-treated cells was best fit by a single exponential with a time constant of 323 ± 79 s (n = 8). To determine whether extracellular Ca2+ was required to maintain this current inactivation, the ECS bathing the cell was switched to a nominally Ca2+-free (1.5 mM EGTA) ECS after the 10 min of stimulation in 1 mM Ca2+ ECS and cell stimulation with 50 μM NMDA was continued. Figure 1 C shows that a representative FGF-2-treated cell displayed almost complete recovery of NMDAR currents after 10 min in Ca2+-free ECS. On average (Fig. 1 D), NMDA-evoked currents recovered to ∼90% of initial, 1 mM Ca2+ ECS, values in 3 min for control cells (n = 4), and in 6 min for FGF-2 treated cells (n = 4). This demonstrates that the inactivation of NMDAR currents during the administration of the pulse train requires extracellular Ca2+, is greater in FGF-2 treated cells than in BDNF-treated or untreated control cells, and is reversible in both control and FGF-2 treated cells when extracellular Ca2+ is removed. Furthermore, the decline of the current in the presence of extracellular Ca2+, as well as the recovery of the current in the absence of extracellular Ca2+, involves relatively slow processes (τ1/2 = ∼300 s) .

Mechanism of the FGF-2 effect

To determine the mechanism of action of FGF-2, we analyzed several parameters of FGF-2 effects, including the specificity of the effects for FGF-2 relative to other growth factors and the specificity for the NMDAR versus the α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR). We also determined the length of exposure to FGF-2 necessary to alter NMDAR currents, the roles of new protein synthesis and of the Ca2+-regulated phosphatase, calcineurin (PP2B), and of voltage regulated Ca2+channels. We asked if FGF-2 acts through modification of NMDAR subunit composition. Finally we examined the effects of FGF-2 on currents elicited by single agonist pulses.

Specificity for NMDAR and chronic FGF-2 treatment

To determine whether AMPARs are also regulated by FGF-2, 3-s pulses (2 min−1) of 100 μM kainate were applied to cells in 1–2 mM Ca2+ ECS. After 25 min of stimulation, there was no difference in the amount of inactivation in kainate-evoked currents in control (49.6 ± 9.0% relative to initial value) versus FGF-2 treated cells (49.6 ± 9.0%) (Fig. 1, E and F). Thus the effects of FGF-2 treatment are specific to the NMDAR versus the AMPAR.

To determine the length of FGF-2 treatment required to potentiate the inactivation of NMDAR currents, 0.5-s test pulses (1 min−1) of NMDA were applied to cells, before and after 50 conditioning pulses 3 s each (2 min−1) of 10 μM l-glutamate in 1 mM Ca2+ ECS (see Fig.2). The responses of cells treated with FGF-2 for different times were compared. Figure 2 A shows NMDA-evoked currents recorded from a representative control and FGF-2 (24 h) treated cell. Peak NMDA-evoked currents were reduced to 66 and 68% of initial values (50 and 100 μM; compare preconditioning pulse records with postconditioning) by the conditioning pulses in the FGF-2 treated cell, but not in the control cell. NMDA-evoked currents decreased to 71.3 ± 3.6% of initial values in cells that had been treated with FGF-2 for 24 h (Fig. 2 B). Control cells displayed NMDA-evoked currents that were 116 ± 11.4% of initial values and thus showed no current change. The effect in the FGF-2 treated cells was significant (P ≤ 0.006, ANOVA, Fisher's PLSD post hoc test), and also occurred in cells treated with FGF-2 for 4 h and 120 h before recording, but not in cells treated with FGF-2 only during the 25 min of conditioning pulses (0 h). A similar result was obtained when the conditioning pulses used NMDA (10 μM) instead of glutamate (data not shown). These data suggest that the greater Ca2+-dependent NMDAR current inactivations seen in FGF-2 treated cells require >0.5 h of growth factor treatment, and are not an acute effect of FGF-2.

Fig. 2.

Increased NMDAR inactivation is not an acute effect of FGF-2 treatment nor a general response to neurotrophic factor exposure.A: representative currents evoked in 1 mM Ca2+ ECS by 0.5-s pulses of NMDA before (pre-) and after (post-) 50 3-s (2 min−1) conditioning pulses of 10 μMl-glutamate. FGF-2 cell treated for 24 h. Approximately 25 min elapsed between pre- and post- traces. The lesser reduction in NMDA-evoked currents in these experiments compared with that observed in earlier experiments (Fig. 1) is consistent with the use of a lower concentration of agonist in the conditioning pulses.B: mean ratio (post-/pre-) of NMDA-evoked (5, 10, 50, and 100 μM, combined) peak currents after a 10-μM glutamate pulse train (as above) in cells treated with FGF-2 (10 ng/ml), for the indicated amount of time. 0 h signifies growth factor present only during conditioning pulses (25 min). Control cells (n = 7), 0 h FGF-2 (n = 8), 4 h FGF-2 (n = 7), 24 h FGF-2 (n = 7), 120 h FGF-2 (n = 7). *, P ≤ 0.006 vs. control, analysis of variance (ANOVA), Fisher's PLSD post hoc test. Absence of * indicates P> 0.1. Comparison of effects of 10 ng/ml FGF-2, BDNF or NT-3 treatment on mean (5, 10, 50, 100 μM, combined) NMDA-evoked peak current ratios after 10 μM glutamate conditioning pulses (as above). Control cells (n = 7), BDNF cells (n = 8), NT-3 cells (n = 5). *, P ≤ 0.006 vs. control, ANOVA, Fisher's PLSD post hoc test. Absence of * indicatesP > 0.1. D: representative western blot of total cellular extracts of cultured cells following the treatments indicated in the figure (repeatedtwice). Equal amounts of protein loaded, parallel blot probed with anti-neurofilament L as a control for amount of neuronal protein. Hippocampus indicates total extract of adult hippocampus.

To rule out the possibility that the effects of FGF-2 are general effects of RTK stimulation, cells were treated with 10 ng/ml of BDNF or NT-3 for 120 h and subject to the above assay. Only FGF-2-treated cells displayed an inactivation of NMDA-evoked currents after conditioning pulses with glutamate, although both BDNF and FGF-2 were able to induce the expression of the Ca2+ binding protein, Calbindin D28 (Ghosh and Greenberg 1995;Vicario-Abejon et al. 1995) (Fig. 2, C andD). These data confirm the results of Fig. 1 that BDNF does not induce the effect and indicate that the inactivation of NMDA-evoked currents is a specific effect of activation of the FGF RTK (FGF receptor 1 or FGFR1) (Eckenstein 1994) and not a general effect of neurotrophic growth factor RTK activation.

Dependence on Ca2+ and calcineurin and ongoing protein synthesis

To address biochemical aspects of the mechanism, conditioning pulses of glutamate were applied under different extracellular ionic conditions or in the presence of different pharmacological agents. Cells were then tested with 0.5-s pulses of 50 and 100 μM NMDA (response ratios averaged, Fig.3 C). As above, the inactivation of the current was dependent on repetitive activation of the NMDAR and was not a nonspecific effect of whole cell dialysis (MacDonald et al. 1989) because conditioning pulses lacking agonist did not decrease the peak current ratio in FGF-2 treated cells. Also consistent with previous results, the inactivation of the current was dependent on Ca2+ entry because application of the conditioning pulses in Ca2+-free (1.5 mM EGTA) ECS abolished the effect (P = 0.05). The inactivation did not result from rundown, a Ca2+-dependent decrease in NMDAR currents involving local depolymerization of the actin cytoskeleton, with a time course similar to the FGF-2-potentiated inactivation of NMDAR currents (Rosenmund and Westbrook 1993b). Rundown is blocked by dialysis of the cell with 1 μM phalloidin, a filamentous actin stabilizing drug. However, inclusion of phalloidin in the pipet solution had no effect on the NMDA-evoked inactivation.

Fig. 3.

Effects of FGF-2 treatment require ongoing protein synthesis and calcineurin activity. A: 50-μM NMDA-evoked peak current ratios for cells treated with FGF-2 in the absence and presence of 1 μM cycloheximide for 24 h. *, P < 0.031 vs. control, two tailed t-test. Initial mean peak current densities (pApF−1) were similar with each treatment (not shown). Controls (n = 6), FGF-2 (n = 7), cycloheximide (n = 4), and FGF-2 + cycloheximide (n = 9).B: western blot demonstrates effective block of FGF-2 induced change in gene expression, relative to NF-L control for total neuronal protein (blot stripped and reprobed). C: mean, peak, 100-μM NMDA-evoked current ratios after 50 conditioning pulses of 10 μM glutamate in cells with the indicated modifications to the extracellular solution (ECS) or pipet solution (ICS). *,P < 0.05, ANOVA, Fisher's PLSD post hoc test, **,P < 0.01. FGF-2 untreated (n = 6), FGF-2 treated with mock conditioning pulses (n= 3), FGF-2 (n = 7), 0 mM Ca2+ ECS (n = 6), phalloidin (1 μM, n= 5), PP2B inhibitory peptide (10 μM, n = 7), control peptide (10 μM, n = 5), nifedipine (50 μM, n = 4), ω-conotoxin GVIA (10 μM,n = 4), and ω-agatoxin IVA (200 nM,n = 4).

Ca2+-dependent inactivations of NMDAR function may be mediated by the Ca2+-dependent protein phosphatase 2B, calcineurin (Lieberman and Mody 1994;Tong and Jahr 1994; Tong et al. 1995). To address whether calcineurin was required for the FGF-2 effect, a specific calcineurin inhibitor peptide that mimics the autoinhibitory domain (10 μM) (Hashimoto et al. 1990) was included in the pipet. The peptide significantly inhibited the FGF-2–treated cell Ca2+-dependent inactivation of peak 50 μM NMDA-evoked current (P < 0.01), whereas a control inhibitor peptide with two mutant amino acid residues had no effect. This shows that calcineurin is required for the Ca2+-dependent NMDAR current inactivation potentiated by FGF-2-treatment.

In Fig. 2 B, >25 min of cell treatment with FGF-2 (the time of FGF-2 treatment for t = 0 h) was necessary for FGF-2's effects on NMDAR currents. The requirement for extended treatment raised the possibility that FGF-2 was acting indirectly, perhaps through the induction of the synthesis of a new protein. To determine the requirement for new protein synthesis, neurons were treated with FGF-2 (10 ng/ml) for 24 h in the absence and presence of 1 μM cycloheximide and assayed for changes in NMDAR inactivation. Figure 3 A shows that the effect of FGF-2 on peak 50 μM NMDA-evoked current ratios is abolished in the presence of cycloheximide (also observed at all NMDA concentrations tested, data not shown). This concentration of cycloheximide is sufficient to block the FGF-2-induced increase in calbindin D28 protein level, as assayed by western blot (Fig. 3 B). Thus the FGF-2 effect requires ongoing protein synthesis during FGF-2 exposure.

Independence from voltage gated calcium channels and from NMDAR subunit change

It has been shown that FGF-2 increases the expression of L-type voltage-gated Ca2+ channels (Shitaka et al. 1996). Increased Ca2+ fluxes through voltage-gated channels in neurons in the distal neurites could contribute to the altered NMDAR responses observed in FGF-2 treated cells. However, specific inhibitors of L-type (50 μM nifedipine), N-type (10 μM ω-conotoxin GVIA), and P/Q-type Ca2+ channels (200 nM ω-agatoxin IVA), either alone or in combination (data not shown) had no effect on the inactivation of NMDAR currents in FGF-2-treated cells (Fig.3 C). This indicates that these voltage-gated Ca2+ channels do not have a role in the FGF-2 effect.

The requirement for ongoing protein synthesis during FGF-2 treatment suggests that a change in gene expression may be required for the FGF-2 effect. To determine whether the FGF-2 effect results from changes in the expression of NMDAR protein subunits (Hollmann and Heinemann 1994), the levels of these subunits were assayed in control cells and cells treated with FGF-2 for 120 h. No significant FGF-2-induced changes were detected in NR1 (constitutively spliced portion of transcript), NR1 (N1 exon), NR1 (C1 exon), or NR2A mRNA levels, as analyzed by RT-PCR. There was a ∼30% decrease in NR2B mRNA levels, relative to an NF-L internal control (Fig.4 Bii), however western analysis (Fig. 4 C, ii and iii) failed to demonstrate a change in NR2B protein levels. There were also no changes in protein levels or distribution induced by FGF-2 treatment, as assayed by immunocytochemistry (NR1-full protein, NR2A/B, GluR1, GluR2/3, andGluR4; data not shown). FGF-2 treatment induced an increase in calbindin D28 protein expression (Figs. 2-4 Cvi), but there were no changes in calcineurin α subunit levels (Fig.4 Cvii). Significant levels of NR1, NR2B, and GluR1–3, but not N2RA protein (Fig. 4 Cii) were detected in our cultures, consistent with reports of NMDAR subunit mRNA expression in cortical neuron and glial cultures grown under similar conditions (Zhong et al. 1994). NMDAR NR1 subunit mRNA contained approximately a 20:1 ratio of N1-exon (absent):N1-exon (present) and a 20:1 ratio of C1-exon (present):C1-exon (absent) splice variants (Sheng et al. 1994) in both control and FGF-2 treated cells (Fig. 4 A, ii and iii). These results strongly suggest that FGF-2 does not act by altering the levels of known neuronal glutamate receptor proteins.

Fig. 4.

FGF-2 treatment does not alter NMDAR mRNA or protein levels.A: autoradiogram of 32P-labeled, semi-quantitative RT-PCR for NMDAR1 mRNA in control cells, cells treated with 10 ng/ml FGF-2 for 120 h, or adult rat brain total RNA. Top panel (i.): cDNA coamplified with oligos to a nonalternatively spliced region of NMDAR1 and neurofilament L internal control. As a control for sensitivity of the PCR reaction serially diluted (1:4, 1:2, and 1:1) samples of adult rat brain cDNA in lanes 3–5, respectively. Lane 6 amplified with NF-L oligos only. Middle (ii.) andbottom (iii.) panels: equal amounts of cDNA amplified with oligos spanning the N1 and C1 alternative exons of NMDAR1. N1+ or C1+ indicates presence of alternative exon, N1− or C1− indicates its absence. Splice variant ratios were approximately 20:1 N1−:N1+ (n = 5 experiments) and 20:1 C1+:C1− (n = 3 experiments) in both controls and FGF-2 treated cells. B,i.: cDNAs, as above, amplified with specific oligos to NMDAR2A. Repeated once with identical result. ii.: cDNAs co-amplified with oligos to neurofilament L and NMDAR2B, with adult rat brain cDNA serial dilution controls. Lane 6, E17 rat hippocampus cDNA amplified under identical conditions. Lane 7, adult rat brain cDNA amplified with NMDAR2B oligos only. There was a ∼30% decrease in NMDAR2B mRNA levels, relative to NF-L, in FGF-2 treated cells (n = 2 experiments). C: western blots of equal amounts (∼20 μg) of protein extracts from control and 10 ng/ml FGF-2 (120 h) treated cultures, probed with indicated antibodies. ii.: lane 3 contains 10 μg of adult rat hippocampus protein as a control to show lack of NMDAR2A protein in the cultures. In v., top band is GluR3, bottom band, GluR2. All blots were stripped and reprobed with an antibody to NF-L to control for neuronal protein levels, and were repeated at least once, with identical results.

FGF-2 alters currents elicited by single agonist pulses

To determine if the effects of FGF-2 could be observed during single NMDA pulses, cells were held at −60 mV and a pulse of 50 μM NMDA was applied via pressure ejection for 9 or 3 s in Ca2+-free (1.5 mM EGTA) or in 1 mM Ca2+-containing, Mg2+-free extracellular solution (ECS), containing 20 μM glycine. In 0 mM or 1 mM extracellular Ca2+, there was no significant difference (P > 0.1, Student's t-test, two tailed) in the peak amplitude of the NMDA-evoked currents in FGF-2-treated versus control cells (Fig.5 A, i–iv). In ECS containing 1 mM Ca2+, the currents inactivated after the initial peak. In the presence of 50 μM NMDA for 3 s, there was a greater inactivation of the current in FGF-2-treated (34.4 ± 5.6%) relative to control cells (21.0 ± 2.2%) in 1 mM Ca2+ extracellular solution (P = 0.06; Fig. 5 B). This suggested that the effect of FGF-2 treatment is observed during single pulses and is enhanced with increasing concentrations of Ca2+. To determine whether a higher extracellular Ca2+ concentration would accentuate the difference in NMDAR currents during single pulses in FGF-2 treated versus control cells, 9-s pulses of 50 μM NMDA were applied in ECS containing 5 mM Ca2+ to control and FGF-2 treated cells, as above. Elevation of the extracellular Ca2+ concentration (Fig. 5 A, v andvi) led to a larger inactivation of the current in both FGF-2-treated and control cells. Although the peak, NMDA-evoked currents were not significantly different under these conditions [630.6 ± 56.6 (FGF) vs. 734.5 ± 66.7 pA (control),P > 0.1, Student's t-test, two tailed], the mean I 3s and mean current at 9 s (I 9s) were significantly (P = 0.05 at 3 s and P = 0.007 at 9 s) reduced in FGF-2 treated cells to a greater extent than in control cells. The mean I 9s in FGF-2-treated cells was ∼50% of controls: 205.9 ± 52.8 versus 438.1 ± 61.2 pA. This difference was not related to a difference in cell size because the mean whole cell capacitance of the cells was similar under these conditions (24.2 ± 3.3 vs. 27.8 ± 1.2 pF). The difference in I 9s was caused by a greater degree of time-dependent current inactivation in the FGF-2-treated than in control cells (Fig. 5 B): 69.6 ± 5.7% versus 42.3 ± 3.2%, after 9 s (P = 0.0004; I 3s's also significantly different, P = 0.0007). There was no difference in relative mean current amplitude responses to 1-s applications of 1–500 μM NMDA (P > 0.1, ANOVA; Fig. 5 C), suggesting that FGF-2 treatment does not alter the receptor's agonist affinity. These data show that NMDARs in FGF-2 treated hippocampal neurons display a greater Ca2+-dependent inactivation of NMDAR currents during single 3- or 9-s applications of NMDA in comparison with these receptors in control cells.

Fig. 5.

FGF-2 induces an increase in Ca2+ -dependent NMDAR inactivation during single NMDA pulses. A: currents evoked in neurons held at −60 mV by 9-s application of 50 μM NMDA in 0 mM Ca2+ ECS (i. andii.), 3-s application of 50 μM NMDA in 1 mM Ca2+ ECS (iii. and iv.), or 9-s application of 50 μM NMDA in 5 mM Ca2+ ECS (v. and vi.) to cells treated with FGF-2 for 24 h (right records) and to untreated cells (left records). B: percent inactivation (I maxI min/I max × 100) during 50-μM NMDA application for 3 or 9 s in indicated [Ca2+] ECS. n = 8 andn = 9 for control and FGF-2 treated cells in 1 mM Ca2+ ECS, and n = 16 experiments from 9 control cells and n = 16 experiments from 9 FGF-2-treated cells at 5 mM Ca2+ ECS. +,P = 0.06; *, P < 0.001 vs. control, two tailed t-test. C: normalized mean NMDA-evoked current dose-responses in 1 mM Ca2+ ECS from n = 7 control cells and n= 8 FGF-2 (24–48 h)-treated cells. Responses represent the average current response during a 1-s application of the indicated concentration of agonist. The currents are normalized to the value at 500 μM normalized to the control SE. The error bars at 500 μM represent the SE of the FGF-2 treated cells. There was no significant difference in response between control and FGF-2 treated cells at any concentration (ANOVA).


A novel action of basic FGF-2 on embryonic hippocampal neurons is described here. We show that FGF-2 treatment of 8 d.i.v. hippocampal neurons enhances the capacity of the NMDAR to undergo a form of Ca2+-dependent inactivation. The change is specific for both FGF-2 and the NMDA receptor. The change requires new protein synthesis during growth factor treatment, which is consistent with a need for >25 min of FGF-2 stimulation. Changes in NMDAR subunit composition, however, do not appear to be involved. The inactivation is dependent on extracellular Ca2+, and is reversible in the absence of extracellular Ca2+. The inactivation was highly statistically significant following a train of pulses administered under physiological concentrations of extracellular Ca2+ ([Ca2+] = 1 mM). The magnitude of the current inactivation relative to the initial current during a pulse train became greater with successive pulses. This suggests that, the inactivation may be a consequence of increased accumulation of intracellular Ca2+ during the train. In agreement, during single pulses, the extent of the FGF-2 effect was elevated by increases in Ca2+ in the bath. The inactivation, however, does not depend on voltage regulated Ca2+ channels. A possible target for Ca2+ is the Ca2+-dependent protein phosphatase, calcineurin, whose activity was required for receptor inactivation.

Forms of use-dependent decrease of NMDAR function

The NMDAR undergoes several forms of activity-related decrease in function which are distinguished from one another by their dependence on agonist and Ca2+ (reviewed by Jones and Westbrook 1996). Two forms of desensitization may take place following glutamate (or NMDA) binding (Benveniste et al. 1990; Mayer et al. 1989; Vyklicky et al. 1990). In the first, glycine-independent desensitization, a high (micromolar) concentration of the co-agonist glycine, limits receptor entry into the most highly desensitized states. At suboptimal glycine concentrations (in the nanomolar range), the receptor undergoes a greater, glycine-dependent desensitization. Because the FGF-2 induced effects studied here take place in the presence of 10 μM glycine, they are unlikely to involve the first of these processes, glycine-dependent desensitization. Moreover, because both control and FGF-2 treated cells showed identical current profiles when stimulated with 50 μM NMDA and 10 μM glycine in the absence of extracellular Ca2+, the extent of the second desensitization process, glycine-independent desensitization which takes place under these conditions, is not altered by FGF-2. Finally, the FGF-2 effect is unlikely to involve either rundown, a decrease in conductance resulting from depolymerization of the actin cytoskeleton (Rosenmund and Westbrook 1993a,b) because it was not affected by phalloidin, a stabilizer of actin filaments, or washout because of washout's irreversibility (MacDonald et al. 1989; Wang et al. 1996).

Relationship of Ca2+-dependent inactivation to the FGF-2 effect

The NMDAR undergoes yet another process, CDI (Krupp et al. 1999; Legendre et al. 1993; Mayer and Westbrook 1985; Medina et al. 1994;Zilberter et al. 1991). CDI is a glycine-insensitive inactivation whose rate (τinact = 4.7 s at [Ca2+]0 = 1.3 mM) increases with increasing [Ca2+]oand which plateaus at 45–50% (Legendre et al. 1993). CDI is diminished by buffering intracellular Ca2+with BAPTA (Legendre et al. 1993). Following CDI, if [Ca2+]o is lowered, currents recover biphasically, with kinetic components of τ1/2 = 0.5–5 s and of τ1/2 = 10–50 s (Legendre et al. 1993; Medina et al. 1994, 1996). This recovery is closely correlated with the decay of [Ca2+]i (Medina et al. 1996). Because inactivation develops on the inner part of the plasma membrane (Medina et al. 1996) and because tonic intracellular perfusion of the cell with elevated Ca2+ occludes inactivation (Legendre et al. 1993), CDI is thought to be induced by elevation of [Ca2+]I rather than by ligand binding alone.

The activity-dependent inactivation of NMDA receptor currents described here in FGF-2 treated cells resembles CDI in several respects. These include its Ca2+-dependence, its reversibility, its time course and its kinetic parameters. At 0 μM [Ca2+]o, there was no difference between FGF-2 treated and control cell current profiles. As [Ca2+]o was increased, current inactivations were seen in both cell groups, but were greater for FGF-2 treated cells. Thus the FGF-2 effect was manifested only in the presence of extracellular Ca2+ and is larger in 5 mM than in 1 mM Ca2+, both characteristics of CDI. Furthermore, the current inactivation reversed fully in the absence of extracellular Ca2+, as does CDI. Also, the time constant of the current inactivation during an individual agonist pulse, ∼1–5 s, is similar to that of CDI (Legendre et al. 1993; Medina et al. 1994), when the two are measured under comparable conditions, but is greater than that of glycine-independent desensitization (∼100 ms) (Benveniste et al. 1990).

During administration of a train of agonist pulses to FGF-2 treated cells, we observed inactivation of mean receptor currents with an apparent τ1/2 = 323 s. The currents recovered when the cells were exposed to 0 mM Ca2+ ECS, with an apparent τ1/2 of ∼360 s. These values for τ1/2 are apparent time constants that depend on the parameters of the protocol. However, they indicate that the FGF-2 potentiated effect has a long time constant component, as does CDI. The Ca2+-dependence, reversibility and kinetic parameters of the FGF-2 potentiated effect all suggest that either CDI itself, or another process with similar properties, is enhanced in cells which have been treated with FGF-2.

Role of calcineurin, α-actinin, and Ca2+ homeostasis

FGF-2 could alter the biochemical response of the NMDAR to Ca2+ fluxes. Peptide inhibitor experiments indicate that potentiation by FGF-2 requires calcineurin. Calcineurin has been implicated previously in receptor control (Lieberman and Mody 1994; Tong and Jahr 1994; Tong et al. 1995), although the dependence of inactivation on a phosphatase has been questioned (Legendre and Westbrook 1990; Medina et al. 1996; Rosenmund and Westbrook 1993a). The effects of FGF-2 also require new protein synthesis prior to agonist stimulation of the receptor. The new protein could function as a receptor regulatory factor that enhances the ability at Ca2+ to control receptor currents. However, calcineurin itself is not likely to be the newly synthesized protein since FGF-2 did not alter its level.

Recently, it has been shown (Halpain et al. 1998;Wyszynski et al. 1997) that brief (5 min) treatment of cultured hippocampal neurons with NMDA induces an increase in the levels of calcineurin in spines, and a calcineurin-dependent collapse of spine structure that is associated with the depolymerization of F-actin. These findings raise the possibility that the calcineurin-dependent enhancement of receptor auto-regulation by FGF-2 described here involves a calcineurin-dependent reorganization of the actin cytoskeleton. Indeed, the association of the actin cytoskeleton with the NMDA receptor has been shown to affect channel currents and is controlled by Ca2+ through Ca2+-calmodulin, which displaces α-actinin from NR1 (Ehlers et al. 1996; Krupp et al. 1999; Zhang et al. 1998). Krupp et al. (1999) have shown that CDI is abolished when the C0 region of C terminal domain of NR1, which contains the binding site for α-actinin (Wyszynski et al. 1997), is truncated. Krupp et al. (1999)suggest that the receptor, when “latched” to the actin cytoskeleton via α-actinin, is in the active state. Possibly the requirement that we observe for protein synthesis during treatment with FGF-2 involves expression of a protein that makes the receptor more susceptible to dissociation from α-actinin by Ca2+- and calcineurin-dependent process.

FGF-2 may also act by altering the regulation of intracellular Ca2+. Indeed, FGF-2 induces L-type Ca2+-channels in fetal hippocampal neurons (Shitaka et al. 1996). However, this is an unlikely basis for the action of FGF-2 since voltage gated Ca2+ channel (VGCC) blockers did not attenuate the FGF-2 effect. The induction by NMDA, the dependence on [Ca2+]o, and the lack of sensitivity to VGCC blockers, all suggest that the NMDAR rather than VGCC is the major point of Ca2+ entry. Neither the role of endoplasmic Ca2+ stores, shown to be important in dorsal horn neurons (Kyrozis et al. 1996), nor the role of retrieval mechanisms were studied.

Alteration of Ca2+ homeostasis by FGF-2 and a consequent accumulation of intracellular Ca2+ could also increase inactivation. Perhaps the Ca2+ elevation produced by the conditioning train is larger or longer lasting in the FGF-2 treated cells, inactivating the receptor. Increased levels of Ca2+ could enhance the Ca2+-calmodulin dependent displacement of α-actinin discussed above. Increased receptor inactivation and calbindin induction, which was also observed in FGF-2 treated cells, may be viewed as components of a more general protective cellular response to increased intracellular Ca2+ transients. However, induction of calbindin is unlikely to be the basis for the effects on the NMDA receptor because BDNF did not alter receptor regulation although it induced calbindin. Also, there is no evidence that the calbindin in these cells is in sufficient proximity to the synaptic membrane of dendritic spines to buffer acute rises in peri-synaptic calcium concentration.

Physiological significance of FGF-2-induced increases in NMDAR inactivation

Glutamate receptor inactivation is a process that maintains physiological levels of signaling while preventing toxicity (Brorson et al. 1995; David et al. 1996;Zorumski and Thio 1992; Zorumski et al. 1989). Treatment with FGF-2 protects neurons from toxic insults involving NMDAR activation in vitro, and in vivo reduces focal ischemic infarct size (Lin and Finklestein 1997). Our results suggest that an enhancement of the capacity of the NMDAR to inactivate may contribute to the neuroprotective effect of FGF-2. The high levels of FGF proteins and the ubiquitous neuronal expression of the FGF receptor, FGFR1, in adult CNS neurons are consistent with such a role (Caday et al. 1990; Eckenstein 1994;Kuzis et al. 1995; Yazakai et al. 1994).

NMDAR desensitization and inactivation are important components of basic synaptic physiology (reviewed by Jones and Westbrook 1996), and their changes may influence the threshold for synaptic modification (Abraham and Bear 1996). Developmental increases in FGFR signaling could regulate synaptic development or plasticity by shifting the threshold for synaptic plasticity to favor LTD or related events.


We thank M. Chesler and P. Osten for critical reading of the manuscript; S. Burden, B. Perrino, and M. Sheng for helpful discussions; B. Perrino and T. Soderling for the control calcineurin inhibitory peptide; and M. Sheng for antibodies. A. L. Boxer is a trainee at the New York University Medical Center Medical Scientist Training Program and E. B. Ziff is an Investigator at the Howard Hughes Medical Institute.

This research was supported by National Institutes of Health Grants NS-30989 to B. Rudy and AG-13620 to E. B. Ziff.

Present address of H. Moreno: Dept. of Neurology, SUNY Brooklyn College of Medicine, Brooklyn, NY 11203-2098.


  • Address for reprint requests: E. B. Ziff, Howard Hughes Medical Institute, New York University Medical Center, Dept. of Biochemistry, 550 First Ave., New York, NY 10016.

  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract