Choline Release and Inhibition of Phosphatidylcholine Synthesis Precede Excitotoxic Neuronal Death but Not Neurotoxicity Induced by Serum Deprivation*

N -Methyl- D -aspartate (NMDA) receptor overactiva- tion has been proposed to induce excitotoxic neuronal death by enhancing membrane phospholipid degrada-tion. In previous studies, we have shown that NMDA releases choline and reduces membrane phosphatidylcholine in vivo . We now observed that glutamate and NMDA induce choline release in primary neuronal cortical cell cultures. This effect is Ca 2 1 -dependent and is blocked by MK-801 (( 1 )-5-methyl-10,11-dihydro-5H-dibenzo[ a,d ]cyclohepten-5,10-imine hydrogen maleate). In cortical neurons, the NMDA receptor-mediated choline release precedes excitotoxic cell death but not neuronal death induced by either osmotic lysis or serum deprivation. Glutamate, at concentrations that release arachidonic acid, does not release choline in cerebellar granule cells, unless these cells are rendered susceptible to excitotoxic death by energy deprivation. The NMDA-evoked release of choline is not mediated by phospholipases A 2 or C. Moreover, NMDA does not activate phos- pholipase D in cortical cells. incubator. Neuronal cell death glutamate. Studies— Cell death was assessed using PI is the plasma membrane of cells. Injury to the cytoplasmic membrane allows the entry of which, by interacting with nuclear DNA, yields a bright red fluorescence. PI fluorescence was in 24 plates using a CytoFluor Analysis— Results are expressed as mean 6 S.E. of at least three separate experiments with triplicate samples in each. Statistical significance of differences was examined using independent t tests or using one-way ANOVA when required. Post hoc multiple comparisons were performed using Student-Newman-Keuls tests.

lated to underlie the neuronal death observed in brain ischemia, traumatic brain injury and in some chronic neurodegenerative disorders, including Alzheimer's and Huntington's diseases (3)(4)(5). It has been established that Ca 2ϩ influx through the N-methyl-D-aspartate (NMDA) subtype of glutamate receptors is an initial event that plays a central role in glutamate-evoked neuronal excitotoxicity. However, the immediate Ca 2ϩ -dependent pathogenic targets that trigger NMDA receptor-mediated neurotoxicity are not fully understood.
Excitotoxic neuronal death induced by glutamate has been hypothesized to occur via both necrosis and apoptosis, depending on the intensity and duration of glutamate exposure (6 -8). Apoptosis is considered to be an active process of cell destruction characterized by nuclear chromatin condensation, cytoplasmic shrinking, dilated endoplasmic reticulum, and membrane blebbing followed by phagocytosis. In contrast, necrotic cell death is considered to be a passive process of cell death produced by cell swelling, injury to cytoplasmic organelles, and membrane lysis. Nonetheless, in neurons, apoptosis and necrosis represent only the extreme ends of a wide range of possible morphological and biochemical deaths (8,9). Mature cortical or granule cell cultures exposed to mild excitotoxic insults undergo a cell death that is mediated by posttranslational activation of caspases and exhibits apoptotic morphology but does not require new RNA and protein synthesis (7,10). As the exposure time or concentration of NMDA receptor agonist increases, excitotoxic cell death changes its shape from apoptotic to necrotic. However, despite the fact that apoptosis and necrosis may share some biochemical mechanisms, the major distinction between them is that in necrosis there is cytoplasmic membrane breakdown and release of cell contents that causes an inflammatory reaction in vivo. We hypothesized that the cytoplasmic membrane breakdown associated with necrosis, instead of being a passive process of cell destruction, is produced by alteration of biochemical processes involved in membrane building.
Choline-containing phospholipids such as phosphatidylcholine (PC) and sphingomyelin (SM) are major structural components of cell membranes. It has been postulated that the cytoplasmic membrane damage associated with excitotoxic neuronal cell death is caused by increased hydrolysis of membrane phospholipids by various phospholipases (11). Previous results obtained in our laboratory using in vivo microdialysis demonstrated that NMDA receptor activation increases extracellular choline (Cho) levels (12). The NMDA-evoked increase in extracellular Cho concentration was associated with a reduction of membrane phosphatidylcholine and with delayed excitotoxicity in prefrontal cortex. In the present experiments, we investigated the mechanisms involved in the increase of extracellular Cho induced by NMDA. The objectives were to determine whether choline release is directly related with excitotoxic cell death and to assess whether NMDA receptormediated choline release results from membrane phospholipid degradation. We now report that the increase in extracellular Cho observed after NMDA receptor activation precedes necrotic but not apoptotic cell death. Moreover, the increase in extracellular Cho induced by NMDA receptor activation is produced by inhibition of phosphatidylcholine synthesis and not by an increase in phosphatidylcholine hydrolysis.

EXPERIMENTAL PROCEDURES
Cell Culture-Primary cultures of brain cortical and cerebellar granule neurons were performed essentially as described previously (13). Cortical and cerebellar granule cells were obtained from Harlan Sprague-Dawley E18 rat fetuses and 7-postnatal day rat pups, respectively. Frontal-lateral cortical lobes or cerebellum were dissected, and cells were chemically dissociated in the presence of trypsin and DNase I and plated in poly-L-lysine-coated (10 g/ml) wells. Cells were seeded in 24-well plates at a density of 9 ϫ 10 5 cells/cm 2 in 750 l of basal Eagle's medium supplemented with 0.1 mg/ml gentamicin, 2 mM L-glutamine, 25 mM KCl, and 10% heat-inactivated fetal bovine serum. In experiments measuring phospholipase D activity, cells were plated in six-well plates at the same density. For cortical cell cultures, medium was supplemented with D-glucose to reach a final concentration of 25 mM. Cytosine ␤-D-arabinofuranoside (10 M) was added to cerebellar and cortical cells 24 and 72 h after seeding, respectively. Cultures were incubated at 37°C in a 5% CO 2 , 95% air atmosphere. Medium remained unchanged until experiments were performed (8 -10 days in vitro). For experiments using Hoechst 33342 or propidium iodide (PI) staining, cells were plated at lower concentrations (4.5 ϫ 10 5 cells/cm 2 ) to facilitate cell counting under microscope.
Treatments-The effect of NMDA or GLU on Cho release and cell death was measured in both cortical and cerebellar granule cells. For experiments in cortical cell cultures, medium was removed, cells were washed twice, and incubations with different treatments were performed in 300 l of modified Locke Hepes buffer without Mg 2ϩ (MLH) (154 mM NaCl, 3.6 mM NaHCO 3 , 2.3 mM CaCl, 5.6 mM KCl, 5.6 mM D-glucose in 5 mM Hepes buffer, pH 7.4). In cells treated with serum deprivation (SD), medium was replaced by fresh medium without fetal bovine serum. In experiments examining the effects of phospholipase inhibitors, these compounds were added 20 min before treatment with NMDA. Cho levels were measured in the incubation media at the indicated times after each treatment. Neuronal cell death was determined with PI staining at the indicated times after treatment. In experiments with cerebellar granule cells, medium was removed and saved for later use as conditioned media. Cells were preincubated for 40 min with MLH with or without D-glucose. Previous studies have shown that a 40 min incubation with MLH without glucose renders cerebellar granule neurons susceptible to GLU toxicity (14). After this pretreatment, both control and energy-compromised cerebellar granule cells were exposed to either GLU or vehicle for 30 min in 300 l of MLH. At the end of this period, Cho was measured in the incubation buffer and cultures were washed twice with 1-ml aliquots of Locke Hepes buffer containing Mg 2ϩ (1.2 mM). Conditioned medium was added to the cells and cultures were returned to the incubator. Neuronal cell death was determined 24 h after exposure to glutamate.
Cell Viability Studies-Cell death was assessed using PI staining. PI is excluded by the plasma membrane of viable cells. Injury to the cytoplasmic membrane allows the entry of PI, which, by interacting with nuclear DNA, yields a bright red fluorescence. PI fluorescence was measured in 24 well plates using a CytoFluor 2350 scanner (Millipore, Barcelona, Spain) with 530 nm (25 nm band pass) excitation and 645 nm (40 nm band pass) emission filters. The percentage of nonviable cells was measured using a modification of the method described by Rudolph et al. (15). In brief, a background fluorescence reading F (Ϫ1) was obtained immediately after addition of PI (30 M). Baseline fluorescence was measured immediately after treatment F (0) . Subsequent fluorescence readings were obtained at different times after the beginning of treatment. At the end of the experiment, cells were permeabilized with 375 M digitonin for 10 min at 37°C to obtain the maximum fluorescence corresponding to 100% of cell death (F (max) ). Percentage of cell death was calculated as follows: % cell death ϭ 100 ϫ (X Ϫ F (Ϫ1) )/ (F (max) Ϫ F (Ϫ1) ), where X is fluorescence at any given time. Cells were kept in the incubator between measurements. In some experiments, the percentage of dead cells was measured counting the number of PIstained cells and dividing by the total number of cells using simulta-neous fluorescence and phase contrast observation in an epifluorescence microscope. In these experiments, cells were incubated with 10 M PI for 30 min and fixed in 3.7% paraformaldehyde for 20 min at room temperature before the addition of a final glycerol protective layer.
Determination of Cho-Extracellular Cho levels were measured in the incubation medium using a high pressure liquid chromatograph coupled to an enzyme reactor and an electrochemical detector (BAS, West Lafayette, IN). Cho was separated in a BAS/Sepstik microbore column with a mobile phase consisting of 50 mM NaH 2 P0 4 , 0.005% Kathon, and 0.5 mM EDTA, pH 8.5 (adjusted with NaOH), at a flow rate of 140 l/min. Cho was enzymatically converted to hydrogen peroxide by a postcolumn enzyme reactor containing immobilized Cho oxidase. The resulting hydrogen peroxide was detected on a platinum electrode at ϩ500 mV (versus an Ag/AgCl reference electrode, BAS-LC4B). Results were expressed in pmol/mg of protein. Protein content was measured in each well using the BCA protein assay reagent (Pierce).
Measurement of Cells with Apoptotic Morphology-Nuclear condensation or pyknosis, a hallmark of cells undergoing apoptotic cell death, was examined in permeabilized cells with Hoechst 33342 staining. Following treatments, cells were washed twice with phosphate-buffered saline and fixed for 5 min in ice-cold methanol. Once fixed, cells were incubated with Hoechst 33342 (0.1 g/ml) for 15 min at room temperature. Cells were then washed three times with phosphate-buffered saline, and a final Mowiol protective layer was added. Cells with pyknotic nuclei and total cells were counted (3 fields per well) by direct observation using an epifluorescence microscope.
Assay of Phospholipase D Activity-Phospholipase D (PLD) (EC 3.1.4.4) activity was assayed in the presence of alcohol using the transphosphatidylation reaction, which is catalyzed specifically by PLD (16). PLD normally hydrolyses phosphatidylcholine producing free choline and phosphatidic acid. In the presence of a primary alcohol, the enzyme catalyzes a transphosphatidylation reaction that transfers the phosphatidyl moiety to the alcohol, forming a phosphatidylalcohol (17). The concentrations of ethanol generally used to determine PLD activity (18) have been reported to inhibit NMDA receptor activation (19). In fact, ethanol and butanol induced a concentration-dependent inhibition of NMDA-evoked Cho release (results not shown). Thus, to measure PLD activity after NMDA receptor activation, we used butanol at a concentration (0.15%) that did not inhibit NMDA-evoked Cho release but allowed sufficient production of phosphatidylbutanol (PtdBut). Cells were prelabeled for 4 h in 1.5 ml of MLH with [ 32 P]orthophosphate (5 Ci/well). Then, cells were incubated for 1 h with 1.5 ml of MLH containing 0.15% butanol and vehicle, 10 M ionomycin, or 100 M NMDA. Incubation buffer was removed and cells were sonicated in ice-cold HCl:methanol (1:100 (v/v)). Lipid extraction was performed by adding chloroform and H 2 O (final proportion of methanol/chloroform/ H 2 O, 2:2:1 (v/v/v)) and separating phases by centrifugation (10 min at 400 g). The lower organic phase was removed, dried under vacuum and dissolved in 10 l of chloroform/methanol (1:1 (v/v)). [ 32 P]PtdBut was separated from other phospholipids in silica TLC plates (HPTLC 60, 10 ϫ 20 cm, Merck) using ethyl acetate/2,2,4-trimethylpentane/acetic acid/water (13:2:3:10 (v/v/v/v)) as a mobile phase. Phospholipid spots were identified by exposure to iodine and co-migration with standards. Radioactivity co-migrating with PtdBut and radioactivity present in the total lipid fraction was determined. [ 32 P]PtdBut production was expressed as a percentage of total labeled phospholipids.
Determination of PC and SM Synthesis-To determine the effect of NMDA on de novo synthesis of PC and SM, we measured incorporation of [methyl-3 H]Cho into membrane PC and SM. As a positive control, we used hemicholinium 3 (HCh-3) a choline kinase inhibitor. Cells were incubated for 1 h in MLH with [methyl-3 H]choline chloride (0.5 Ci/ well) in the presence of vehicle, 1 mM HCh-3, or 100 M NMDA. Immediately after incubation, medium was removed, ice-cold HCl:methanol (1:100 (v/v)) was added and the methanolic cell suspension was sonicated. Lipid extraction was performed by adding chloroform and H 2 O (final proportion of methanol/chlorofom/H 2 O, 2:2:1 (v/v/v)) and separating phases by centrifugation (10 min at 400 ϫ g). The lipids present in the lower organic phase were dried under vacuum centrifugation and dissolved in 10 l of chloroform/methanol (4:1 (v/v)). Lipids were separated on silica TLC plates (HPTLC 60, 10 ϫ 20 cm, Merck) using chloroform/methanol/acetic acid/water (75:45:3:1 (v/v/v/v)) mobile phase. Phospholipids were identified by exposure to iodine and comigration with standards. The spots identified as PC and SM were scraped, and the radioactivity was measured. Results are expressed as dpm of 1), and phases were separated by centrifugation. The aqueous phase was dried under vacuum centrifugation. Water soluble metabolites were dissolved in 50% ethanol/water. Choline, phosphocholine and CDP-choline were separated on silica gel chromatographic plates (TLC 60, 20 ϫ 20 cm) and developed in the solvent methanol/0.9% NaCl/ ammonium hydroxide (50:50:5 (v/v/v)). Lipids in the chloroform phase were separated in the solvent, chloroform/methanol/acetic acid/water (75:45:3:1 (v/v/v/v)). The spots were revealed by iodine vapor and scraped, and the radioactivity was measured by scintillation counting.
Data Analysis-Results are expressed as mean Ϯ S.E. of at least three separate experiments with triplicate samples in each. Statistical significance of differences was examined using independent t tests or using one-way ANOVA when required. Post hoc multiple comparisons were performed using Student-Newman-Keuls tests.  (Fig. 1B). Removal of extracellular Ca 2ϩ inhibited the increase in extracellular Cho evoked by NMDA, indicating that this effect is dependent on extracellular Ca 2ϩ entry (Fig. 1B).

NMDA Induces a Concentration and Ca
The NMDA-evoked Increase in Extracellular Cho Precedes Excitotoxic Cell Death but Not Cell Death Induced by Trophic Factor Withdrawal in Cortical Cells-Previous studies in cortical cell cultures have shown that NMDA induces excitotoxic cell death that is predominantly necrotic when the culture medium contains physiological ion concentrations (20). To determine whether the increase in extracellular Cho induced by NMDA is produced before or after cell death, we performed time course studies. First, the effects of NMDA on excitotoxic neuronal death and extracellular Cho levels were investigated in serum-free Locke Hepes buffer (Figs. 1 and 2), because both serum and basal Eagle's medium contain Cho. In addition, exposure of cortical cell cultures to fresh serum-containing medium triggers excitotoxicity (21). Fig. 2A shows the time course of neuronal cell death induced by NMDA in serumdeprived cortical cell cultures. A significant cell death was observed in serum-deprived controls (38 Ϯ 1%, after 24 h of serum deprivation). Simultaneous treatment with SD and with 100 M NMDA for 24 h produced 57 Ϯ 1% cell death, an effect that is significantly different from the observed in cultures treated with SD alone. Thus, exposure of cortical cells to 100 M NMDA for 24 h produced a cell death of approximately 19%. Whereas both SD and NMDA were neurotoxic, the cell death induced by SD was significant after the first half-hour of treatment compared with controls not deprived of serum. In contrast, cell death induced by NMDA was not significant until 6 h of continuous treatment, compared with controls deprived of serum for 6 h ( Fig. 2A). Exposure of cortical cell cultures to 100 M NMDA induced a marked increase in extracellular concentration of Cho. This effect was observed 30 min after addition of NMDA and reached a maximum after 4 h of continuous treatment with NMDA. 30 min, 1 h, 2 h, and 4 h after addition of NMDA, extracellular Cho was increased approximately 2-, 6-, 12-, and 14-fold, respectively (Fig. 2B). Exposure of control cells to SD alone did not modify extracellular Cho levels during the first 4 h; however, a small but significant increase in extracellular Cho levels was observed between 4 and 24 h of treatment (Fig. 2B). The NMDA-evoked increase in extracellular Cho was observed well before cytoplasmic membrane breakdown associated with necrotic cell death. Excitotoxicity induced by NMDA was not significant until after 6 h of continuous treatment ( Fig. 2A). However, the increase in extracellular Cho occurred within the initial 4 h after the addition of NMDA (Fig.   2B). In contrast, trophic factor withdrawal produced a significant cell death but did not modify extracellular Cho levels. Cell death induced by SD was significant after 30 min of treatment ( Fig. 2A), but no significant increase was observed in extracellular Cho until 24 h of SD (Fig. 2B).
To assess whether the NMDA-evoked Cho release depends on the presence of extracellular Cho and to determine whether it is produced only before excitotoxic cell death, we investigated the effect of NMDA (100 M) on extracellular Cho levels in cortical cell cultures incubated with serum-containing conditioned medium. In these experiments, NMDA was added directly to conditioned medium that contains Cho and serum. The results obtained confirmed that the increase in extracellular Cho levels precedes excitotoxic cell death but not cell death induced by growth factor withdrawal (Figs. 3). No significant cell death was observed after 1 h of continuous treatment with either NMDA or SD (Fig. 3A). However, a pronounced cell death was observed 24 h after continuous exposure to both treatments (43 Ϯ 3% and 56 Ϯ 4% for SD and NMDA, respectively) (Fig. 3A). Moreover, exposure of cortical cells for 1 or 24 h to 100 M NMDA in conditioned medium produced an increase in extracellular Cho of 79 Ϯ 12 and 426 Ϯ 69 pmol/well over control levels at 1 and 24 h, respectively (Fig. 3B). In contrast, SD did not significantly modify extracellular Cho content after either 1 or 24 h of treatment (Fig. 3B).
To gain further insight on the type of cell death that is preceded by Cho release, we measured the number of condensed nuclei (pyknosis), an index of apoptosis. SD for 12 h produced a significantly higher pyknosis than the observed after 12 h of treatment with NMDA (26 Ϯ 2% and 12 Ϯ 2%, respectively (Fig. 4A)).
GLU Increases Extracellular Cho Only in Cerebellar Granule Cells That Undergo Excitotoxic Cell Death-Energy-compromised cerebellar granule cells are sensitive to excitotoxic cell death but are resistant to excitotoxicity when there is enough energy supply (14,22,23). To investigate whether the increase in extracellular Cho was directly linked to cell death, we assessed whether NMDA receptor overactivation was able to induce Cho release in cerebellar granule cells resistant to excitotoxicity. Incubation of energy-compromised cerebellar gran- ule cells with glutamate for 30 min killed 61 Ϯ 3% of the cells, 24 h after treatment. In contrast, in the presence of glucose, 30 min incubation with glutamate did not produce significant cell death compared with controls (19 Ϯ 3% of dead cells 24 h after treatment with glutamate compared with 18 Ϯ 2% cell death in control cells) (Fig. 5A). Glutamate released Cho only in energycompromised cells, in which exposure to glutamate leads to cell death. Neither glucose deprivation nor incubation with glutamate in the presence of glucose induced significant Cho release or cell death (Fig. 5).
The Increase in Extracellular Cho Induced by NMDA Is Not Associated with Hydrolysis of Membrane Phospholipids-To determine whether NMDA-evoked Cho release originates from hydrolysis of membrane phospholipids we investigated whether phospholipase A 2 (PLA 2 ) and phospholipase C inhibitors modify this effect of NMDA. The contribution of PLA 2 to the NMDA-evoked increase in extracellular Cho was investigated in primary cortical cell cultures using the PLA 2 inhibitors 4-bromophenacyl bromide (10 M) and aristolochic acid (100 M). None of these inhibitors antagonized the effect of NMDA on extracellular Cho levels. In fact, aristolochic acid potentiated the effects of NMDA on Cho release (Fig. 6A). Furthermore, D-609 (100 M), an inhibitor of phospholipase C, did not modify the effect of NMDA on extracellular Cho (Fig.  6A, inset). The contribution of PLD to the NMDA-evoked increase in extracellular Cho was investigated by measuring PtdBut, the product of the PLD-catalyzed transphosphatidyla-tion reaction in the presence of butanol. Ionomycin, a compound that activates PLD, increased the amount of PtdBut. However, NMDA did not significantly modify PtdBut levels (Fig. 6B).  Time course analysis of incorporation of [methyl-3 H]Cho into membrane PC indicated that inhibition of PC synthesis by NMDA is already significant 10 min after the initiation of treatment (Fig. 8D). However, analysis of the intermediates of the CDP-choline pathway (Fig. 8, B and C) did not reveal significant differences between control and NMDA-treated cells. Thus, CDP-choline labeling was identical in NMDAtreated cells versus control (Fig. 8C). Labeling of phosphocholine was also identical in control and treated cells in most time points, with the exception of 60 min after treatment with NMDA, when there is a small but significant accumulation of [methyl-3 H]phosphocholine compared with control cells (Fig.  8B). In contrast, NMDA treatment induces a marked accumulation of intracellular [methyl-3 H]Cho (Fig. 8A). DISCUSSION The present work demonstrates that NMDA receptor-mediated release of Cho precedes excitotoxic cell death but not cell death induced by SD. We found that, in cortical cell cultures, both NMDA and GLU induce a concentration-dependent increase in extracellular Cho (Fig. 1A). This effect is dependent on the presence of extracellular Ca 2ϩ and it is blocked by (ϩ) MK-801 (10 M), a selective antagonist of the NMDA subtype of glutamate receptors (24), indicating that Ca 2ϩ entry through NMDA receptors is necessary for the excitatory amino acid induced increase of extracellular Cho (Fig. 1B).

NMDA Inhibits de Novo Incorporation of [methyl-3 H]Cho into Membrane Phosphatidylcholine and Sphingomyelin-Be
We observed that the increase in extracellular Cho evoked by NMDA is previous to excitotoxic cytoplasmic membrane damage and cell death (Figs. 2 and 3). Exposure of cortical cells to NMDA in Locke Hepes buffer produces a significant increase in extracellular Cho 30 min after addition of NMDA, whereas no significant cell death was observed until 6 h of continuous treatment with NMDA (Fig. 2). This difference in time course between NMDA-evoked Cho release and cell death indicates that Cho release is an early event in the process of excitotoxic cell death.
To determine whether the absence of Cho and serum in the incubation medium has an influence on the effect of NMDA on Cho release and cell death, we performed experiments in which NMDA was directly added to the medium in which cells had grown (conditioned medium) that contains serum and Cho. Thus, in these experiments, the effect of NMDA could be isolated from that of SD. As in experiments performed in Locke Hepes buffer, we found that addition of 100 M NMDA to conditioned medium increased extracellular Cho (79 Ϯ 12 pmol/well over control values) but did not produce cell death during the first 1 h of exposure (Fig. 3). Continuous exposure of cortical cells to NMDA for 24 h increased extracellular Cho levels further and induced a cell death of 56 Ϯ 4% (Fig. 3). These results confirm those obtained in experiments in which treatment with NMDA was performed simultaneously with SD and provide further evidence showing that NMDA receptormediated Cho release precedes excitotoxic cell death.
Subjecting cortical cells to SD caused an amount of cell death that was equivalent to the observed after treatment with GLU or NMDA (Fig. 3A). However, extracellular Cho did not increase during SD (Figs. 2B and 3B). Under our experimental conditions, cell death induced by NMDA was mostly necrotic (Fig. 4). This is in agreement with recent evidence showing that in cortical cells, even delayed excitotoxic cell death, previously considered apoptotic, is in fact necrotic (20). In contrast, cell death induced by SD was mostly apoptotic (Fig. 4), in accordance with previous findings in cortical cell cultures showing that cell death induced by SD is dependent on protein synthesis and fulfills morphological criteria for apoptosis (25). The difference between SD and NMDA in releasing Cho before cell death indicates that Cho release discriminates between necrotic cell death induced by excitotoxicity and apoptotic cell death induced by SD.
To determine whether NMDA receptor-mediated Cho release is directly related with excitotoxicity or is just an epiphenomenon, we investigated the effects of NMDA and GLU in cerebellar granule cell cultures. Cerebellar granule cells are resistant to excitotoxic cell death (3). However, energy deprivation makes cerebellar granule cells susceptible to excitotoxicity by GLU (14). To test the hypothesis that there is a direct link between Cho release and excitotoxicity, we investigated whether the neurotoxic effect of GLU in energy-compromised cerebellar granule cells was preceded by an increase in extracellular Cho. Exposure of cerebellar granule cells to NMDA (500 M) for 1 h (results not shown) or GLU (20 M) for 30 min, in the presence of a physiological concentration of glucose, did not produce significant cell death 24 h after treatment (Fig.  5A). Under these conditions, neither NMDA (results not shown) nor GLU released Cho (Fig. 5B), but both NMDA and GLU released arachidonic acid (13). In contrast, exposure of energy-compromised cerebellar granule cells to GLU for 30 min increased extracellular Cho levels by 56% during GLU exposure and induced 61 Ϯ 3% cell death 24 h after treatment with GLU (Fig. 5). These results indicate that glutamate receptor activation is followed by Cho release only if the activation leads to neurotoxicity. Based on these findings, we can conclude that NMDA receptor-mediated Cho release is directly related with excitotoxic cell death.
These results in cerebellar granule cells are in contrast with those observed in cortical cells, in which NMDA receptor activation increases extracellular Cho and produces subsequent cell death even in the presence of physiological concentrations of glucose. The mechanism responsible for the high sensitivity of cortical cells to release Cho after NMDA receptor activation may account for the greater susceptibility of these cells to excitotoxicity compared with cerebellar granule cells. The difference between cortical and cerebellar granule cells to release Cho after NMDA receptor activation is consistent with our previous results in vivo showing that NMDA increases dialysate Cho levels in prefrontal cortex but not in cerebellum (12).
The extracellular Cho measured after exposure of cells to NMDA in a Locke Hepes buffer that does not contain Cho indicates that this Cho must proceed from the cell. In brain, Cho is stored almost completely in a metabolized form (26). Intracellular free Cho concentration is kept low because once Cho is inside the cell, it is rapidly phosphorylated by choline kinase (26,27). In fact, in our experimental conditions, osmotic lysis of cortical cells did not produce a significant increase in free Cho (75 and 78 pmol/well Cho in control and lysed cells, respectively). Thus, we investigated whether NMDA-evoked Cho release is produced by breakdown of Cho containing compounds. The main compounds containing Cho in neurons are acetylcholine, in cholinergic cells, and the phospholipids PC and SM (28,29). Neither cortical nor cerebellar granule cell cultures showed detectable levels of extracellular acetylcholine even in the presence of high concentrations of the acetylcholinesterase inhibitor neostigmine (not shown), indicating that acetylcholine is not the source of Cho released by NMDA. Our previous studies showed that sustained activation of NMDA receptor in vivo is associated with a significant decrease of membrane PC (12). Thus, we investigated whether the phospholipases PLA 2 , PC-phospholipase C and PLD that are involved in PC breakdown are also implicated in the effect of NMDA on Cho release. NMDA receptor activation has been shown to induce arachidonic acid release in primary cultures of striatum (30), hippocampus (31), and cerebellar granule cells (13,32,33). Phospholipase A 2 is the primary effector enzyme responsible for NMDA receptor-evoked release of arachidonic acid in neuronal cultures (30,31,33). In addition, NMDA receptor activation has been reported to increase phospholipase C activity (34 -36).
The PLA 2 inhibitors 4-bromophenacyl bromide and aristolochic acid did not inhibit NMDA-evoked Cho release (Fig. 6A). This observation, combined with the fact that cerebellar granule cells exposed to GLU, in the presence of glucose, release arachidonic acid (13) but not Cho (Fig. 5B), strongly indicates that PLA 2 is not involved in the release of Cho induced by NMDA. Likewise, the phosphatidylcholine-specific phospholipase C inhibitor D-609 (37) did not inhibit the increase in extracellular Cho induced by NMDA (Fig. 6A, inset). Finally, we investigated the possibility that Cho released by NMDA could be produced by activation of PLD. There is evidence indicating that activation of PLD by glutamate is exclusively mediated through metabotropic receptors (18,38,39). Accordingly, we found that NMDA does not activate PLD in neuronal cortical cell cultures (Fig. 6B). However, the concentrations of alcohol generally used to determine PLD activity by the transphosphatidylation reaction (18) inhibit NMDA receptor activation (19). It could therefore be possible that the concentration of primary alcohol used in the enzymatic assay might block the effect of NMDA on PLD activity. This possibility was ruled out in our experiments because we used a concentration of alcohol in the transphosphatidylation reaction that does not significantly inhibit NMDA-evoked Cho release. Based on these results, we conclude that the NMDA receptor-mediated increase in extracellular Cho is not produced by activation of membrane phospholipid degradation.
Because the increase in extracellular Cho is not produced by NMDA receptor-induced phospholipid catabolism, inhibition of phospholipid synthesis by NMDA emerged as an alternative mechanism to explain the NMDA-evoked Cho release. A decrease in PC synthesis with no significant alteration in PC breakdown would lead to a decrease in membrane PC content and also a subsequent increase in extracellular Cho, because if conversion of phosphocholine to PC is inhibited, Cho leaks outside the cell after rapid dephosphorylation. Indeed, the finding that NMDA causes a marked inhibition of PC synthesis (Fig. 7) provides a strong support for this interpretation. Furthermore, consistent with this interpretation is the fact that inhibitors of phospholipid hydrolysis potentiate the effect of NMDA on Cho release (Fig. 6A). Inhibitors of phospholipid catabolism diminish the cellular supply of diacylglycerol, an essential and limiting substrate in the last step in PC biosynthesis, inhibiting, in turn, the biosynthesis of PC. Thus, the potentiation of NMDA-evoked Cho release by inhibitors of phospholipid hydrolysis can be attributed to an enhanced inhibition of PC synthesis produced by these compounds. Moreover, recent studies in our laboratory have shown that NMDA does not inhibit Cho uptake (40). Therefore, the NMDA-evoked inhibition of PC synthesis must be produced by an action on one or several of the enzymes of the Kennedy cycle (41).
Analysis of incorporation of [methyl-3 H]Cho into the intermediates of the CDP-choline pathway in the presence and absence of NMDA indicated that the reduction of radiolabeled membrane PC is associated with an NMDA-evoked accumulation of intracellular free [methyl-3 H]Cho (Fig. 8A). The amount of radiolabeled phosphocholine and CDP-choline was not different between NMDA-treated cells and controls (Fig. 8, B and  C). These results indicate that inhibition of PC biosynthesis by NMDA is not due to the inability of the cells to take up choline. It is unlikely that this effect of NMDA is produced by NMDAevoked energy depletion because treatment with this compound does not reduce the radiolabeled products of energy-dependent enzymes such as choline kinase (EC 2.7.1.32) and CTP:phosphocholine cytidylyltransferase (EC 2.7.7.15) to levels below those of control cells. The accumulation of labeling in the intracellular pool of free Cho suggests that NMDA inhibits PC synthesis by selectively inhibiting choline kinase activity. In support of this interpretation is the fact that NMDA does not produce a significant accumulation of phosphocholine or CDPcholine (Fig. 8, B and C). However, the possibility still exists that the increase in intracellular [methyl-3 H]Cho is produced by a rapid dephosphorylation of phosphocholine and/or hydrolysis of CDP-choline. To test this possibility, we are currently performing pulse-chase experiments (41) to identify the enzyme or enzymes responsible for the inhibition of PC synthesis by NMDA.
In summary, this study shows that NMDA receptor activation, but not SD, induces Cho release that precedes and is directly linked to excitotoxic necrotic cell death. In addition, the present results indicate that NMDA induces excitotoxic necrotic cell death by inhibiting PC synthesis and not by increasing membrane phosphatidylcholine breakdown. These findings provide evidence for a new mechanism of excitotoxicity that may discriminate between necrotic and apoptotic neuronal cell death. Moreover, we propose that NMDA receptor-mediated inhibition of phosphatidylcholine synthesis is a key early event in the excitotoxic cascade that leads to necrotic cell death.