Address correspondence and reprint requests to Dr. B. Jacquemont at Laboratoire de Neuro- Virologie Moléculaire, Faculté de Médecine Laënnec, rue Guillaume Paradin, 69372 Lyon cédex 08, France. E-mail : jacquemo @ laennec.univ-lyon1.fr
Abstract : Astrocytes play a key role by catabolizing glutamate from extracellular space into glutamine and tricarboxylic acid components. We previously produced an astrocytic cell line that constitutively expressed glutamic acid decarboxylase (GAD67), which converts glutamate into GABA to increase the capacity of astrocytes to metabolize glutamate. In this study, GAD-expressing astrocytes in the presence of glutamate were shown to have increased energy metabolism, as determined by a moderate increase of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide reduction, by an increased ATP level, and by enhanced lactate release. These changes were due to GAD transgene expression because transient expression of a GAD antisense plasmid resulted in partial suppression of the ATP level increase. These astrocytes had an increased survival in response to glucose deprivation in the presence of glutamate compared with the parental astrocytes, and they were also able to enhance survival of a neuronal-like cell line (PC12) under glucose deprivation. This protection may be partially due to the increased lactate release by GAD-expressing astrocytes because PC12 cell survival was enhanced by lactate and pyruvate under glucose deprivation. These results suggest that the establishment of GAD expression in astrocytes enhancing glutamate catabolism could be an interesting strategy to increase neuronal survival under hypoglycemia conditions.
To increase glutamate metabolism in astrocytes, we previously transduced an astrocyte cell line with a retrovirus expressing the GAD transgene, directed by the glial fibrillary acid protein promoter, chosen because glial fibrillary acidic protein is one of the genes most commonly activated during focal injury (Mucke et al., 1991). The astrocytic clones were able to express functional GAD and release GABA in the presence of extracellular glutamate. The GAD transgene seems to be more efficient at catabolizing glutamate than the other glutamate catabolic pathways (Sacchettoni et al., 1998). Because GABA can be also metabolized in the TCA cycle (Balázs et al., 1970 ; Hassel et al., 1998), we have now examined how the presence of the GAD transgene affects the energetic metabolism of astrocytes [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) reduction, ATP levels, and lactate release] and the effect of the establishment of this new astrocytic glutamate metabolism on astrocyte and neuronal survival under glucose deprivation. PC12 cells, a rat pheochromocytoma cell line derived from glutamate-resistant chromaffin cells, which can differentiate into neurons, were used.
MATERIALS AND METHODS
C8 cells (also designated C8s) are a spontaneously immortalized astrocytic cell line, obtained from the 8-day-old post-natal mouse cerebellum and described as Bergman-Golgi-like cells (clone type II) (Alliot and Pessac, 1984). C8 astrocytic clones (13H and 13J), stably expressing the G418 resistance gene and rat GAD67 directed by the glial fibrillary acidic protein promoter, previously generated in our laboratory (Sacchettoni et al., 1998), were designated C8-GAD67. C8 and the GAD-expressing clones were grown in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum and split 1 : 6 twice a week. Except during experiments, the clones were maintained in the presence of 200 μg/ml G418 (Gibco-BRL) and reselected every 3 weeks with 400 μg/ml G418.
PC12 cells (Greene and Tischler, 1976) were maintained in DMEM (GibcoBRL) containing 6% fetal calf serum (Sigma) and 6% horse serum (Sigma). They were fed every 2 days and split 1 : 10 once a week.
To facilitate survival assessment and avoid contact between different populations, cocultures on separated supports were used. Astrocytes (1.4 × 106) were seeded into 100-mm-diameter culture dishes containing 18-mm-diameter glass coverslips (4.5 × 104 astrocytes per coverslip). After 1 day, the coverslips were removed, inverted, and introduced into 100-mm-diameter wells containing PC12 cells (5 × 104 cells per well). A small paraffin foot was applied to each coverslip to raise it slightly above the bottom of the well, thus avoiding contact between the cell lines (Bartlett and Banker, 1984). The use of this support has several advantages over the traditional use of a membrane insert as it allows maximal exchange between cultures because of the absence of a porous membrane and, in addition, survival assessment can be carried out exclusively on PC12 cells once the astrocyte-coated coverslips are removed. A check for contamination of the PC12 cells by astrocytes showed this to be <1%, a negligible value that was not taken into consideration in the experiments.
The cells were gently washed three times with Tris-buffered salt solution without glucose (120 mM NaCl, 5.5 mM KCl, 0.8 mM MgCl2, 1.8 mM CaCl2, and 25 mM Tris-HCl, pH 7.6) and then exposed to various test solutions, as described below. For astrocyte survival assays, the cells were fed with glucose-free DMEM containing 10% dialyzed fetal calf serum. For PC12 cell survival assays, the cells were fed with glucose-free DMEM containing 6% fetal calf serum and 6% horse serum.
The GAD67 antisense plasmid was constructed as described previously (Sacchettoni et al., 1998), with the HindIII fragment containing GAD cDNA (Erlander et al., 1991) being inserted in the opposite direction in the pKS vector plasmid and directed by the human cytomegalovirus promoter. Cells (1.4 × 106) were seeded into 100-mm2 cell culture flasks ; 24 h later, the cells were washed with fresh serum-free medium (OptiMEM ; GibcoBRL) and transfected. For transfection, 6 μg of DNA and 6 μl of Lipofectamine (GibcoBRL) were diluted with 200 μl of OptiMEM, and the solution was gently mixed and then incubated for 30 min before addition of culture medium. After 12 h of incubation, the medium was removed, and the cells were fed with fresh maintenance medium for 24 h. Finally, 1 mM glutamate was added for a further 24 h. The coverslips were removed from the flasks, and the ATP content of the cells on the coverslips was measured, whereas the flask contents were used to measure GAD activity, used as a control of antisense plasmid efficiency. Transfection efficiency, determined using lacZ directed by the human cytomegalovirus promoter, was roughly 50% by X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactoside) staining.
Trypan blue exclusion
The culture medium was removed and replaced by 0.1% trypan blue solution in phosphate-buffered saline for 3 min at room temperature. Viable cells in 10-15 randomly chosen fields per well (field = 0.58 mm2) were counted at 200-fold magnification using bright-field and phase-contrast microscopy for blue cells and total cells, respectively.
The MTT colorimetric assay is based on the reduction of the tetrazolium salt, MTT, which results in a blue formazan product (Mosmann, 1983). MTT, which reacts with almost all dehydrogenases and cofactors of the respiratory chain, is an indicator of mitochondrial activity (Berridge and Tan, 1993). The culture medium from the astrocytes or PC12 cells was replaced by a solution of 0.75 mg/ml MTT (Merck) in phosphate-buffered saline containing 25 mM glucose. After 2 h of incubation at 37°C, the medium was removed, 1 ml of dimethyl sulfoxide was added, and the plates were shaken for 30 min at room temperature to solubilize the blue formazan crystals. Reduction was measured on a spectrophotometer by the difference between optical density at 550 and 690 nm.
Intracellular ATP levels were assayed using a luciferin/luciferase-based method (DeLuca and McElroy, 1984), based on the oxidation of luciferin by luciferase in the presence of ATP. The cells were washed with cold phosphate-buffered saline and extracted with 100 μl of cell lysate buffer. Then, 1 μl of the lysate was diluted to 100 μl with water and mixed with 100 μl of luciferase/luciferin reagent (high-sensitivity ATP bioluminescence assay ; Boehringer). After 10 s, the light emitted was recorded using a luminometer at 562 nm and integrated over 5 s.
GAD activity was assayed radiometrically, as previously described (Denner and Wu, 1985). Cell extracts were incubated for 90 min at 37°C in a total value of 50 μl of 50 mM potassium phosphate buffer (pH 7) containing 1 mM 2-amino-methylisothiouronium hydroxybromide, 200 mM pyridoxal 5′-phosphate, 1 mM dithiothreitol, 0.15 μCi of [1-14C]glutamic acid (45 mCi/mmol), and 0.25 μmol of glutamic acid. The 14CO2 produced was trapped on hyamine-presaturated Whatman filters and counted. Rat brain homogenates were used as the control. The protein concentration was determined using the method of Bradford (1976).
Using amyloglucosidase, glycogen was converted to glucose, which was quantified using hexokinase and glucose-6-phosphate dehydrogenase (Pellerin et al., 1997). In brief, 24 h after 106 cells were plated, they were removed, spun down, lysed with 0.4 M perchloric acid, sonicated, and neutralized with 1 M KHCO3. Aliquots were first incubated with or without 0.1 mg/ml amyloglucosidase (Sigma) for 2 h at 40°C and then with 0.1 M Tris (pH 8) containing 3 mM MgCl2, 300 μM ATP, 38 μM NADP, 0.004 mg/ml hexokinase, and 0.002 mg/ml glucose-6-phosphate dehydrogenase for 30 min at room temperature, and the optical density of the supernatant was read at 340 nm. The amount of glucose produced was calculated by subtracting the results for the nonhydrolyzed sample from those for the sample hydrolyzed with amyloglucosidase. All values were normalized to the protein content, determined using the method of Lowry et al. (1951).
Quantification of glucose, lactate, and GABA in medium
Glucose was quantified using hexokinase/glucose-6-phosphate dehydrogenase as described above.
Lactate was quantified using lactate dehydrogenase (bovine heart), which, in the presence of excess NAD, converts lactate into pyruvate, which was trapped with hydrazine. The NADH formed is a measure of the amount of lactate originally present and is quantified spectrophotometrically at 340 nm (Walz and Mukerji, 1988). Values were normalized to the protein content of the cells (Lowry et al., 1951).
GABA in media was quantified by HPLC with fluorescence detection. Samples were immediately filtered through a membrane (Vivaspin 500 ; Vivascience) and then forzen at -80°C. Dialysates were treated with naphthalene-2,3-dicarboxaldehyde in the presence of CN- (de Montigny et al., 1987).
Data are mean ± SD values. All survival results were converted to a percentage of the value in control cells. Statistical analysis was carried out by one-way factorial ANOVA followed by Duncan's test.
Evidence for GABA synthesis in GAD-expressing astrocytes
In a previous article, we reported that GAD-expressing astrocytes are able to synthesize and release small amounts of GABA when glutamate is present in the culture medium (Sacchettoni et al., 1998). In the present study, GAD-expressing astrocytes were incubated with 1 mM glutamate or with 50 μMγ-vinyl GABA (GVG ; Hoechst Marion Roussel) alone or with the combination of 1 mM glutamate plus 50 μM GVG. GVG is an irreversible inhibitor of GABA transaminase (Löscher, 1980). We used a high concentration of GVG as it is les effective on astrocytes than on neurons (Larsson et al., 1986). GABA released from astrocytes in the medium was quantified by HPLC after various times (Table 1). As previously demonstrated, in the absence of glutamate, no GABA release was seen. In the presence of GVG, a small amount of GABA release was detected. In the presence of glutamate, the amount of GABA release was of the same order and increased >30 times with addition of GVG.
Table 1. Kinetics of GABA release from GAD-expressing astrocytesND, not done.
Increased energetic metabolism in GAD67-expressing astrocytes
Redox state and ATP levels. The parental astrocytes (C8) or GAD-expressing astrocytes (C8-GAD67) were incubated in the presence of 1 mM glutamate for either 1 or 24 h, and then MTT reduction was measured (Mosmann, 1983). The results, expressed as a percentage of that in the absence of glutamate, are shown in Fig. 1 (upper panel). In the absence of glutamate and in the presence of glutamate for 1 h, no difference was seen between the two sets of astrocytes, whereas after 24 h of glutamate treatment, MTT reduction was unchanged for the control astrocytes (94.3 ± 4.2%) but moderately increased for the GAD-expressing astrocytes (111.2 ± 5.7%). In other experiments, MTT values were 81.1 ± 3.7 compared with 119.3 ± 7.6 pmol/mg and 91 ± 5.4 compared with 115.6 ± 7.1 pmol/mg, respectively.
To verify that this increase corresponded to free available energy, ATP levels were measured. Cells, incubated as above, were lysed, and their ATP levels were determined using a luciferin/luciferase-based assay (DeLuca and McElroy, 1984). In the absence of glutamate, ATP levels were higher in C8 cells (201 ± 3 pmol/mg) than in C8-GAD67 cells (157 ± 10.5 pmol/mg). However, the GAD-expressing cells showed a marked increase in ATP levels after 1 h of glutamate treatment and a twofold increase after 24 h (280.8 ± 10.1 pmol/mg), whereas C8 cells only showed a small change (205.8 ± 6.5 pmol/mg ; Fig. 1, lower panel). This increase was reproducible. In other experiments, ATP values were 274 ± 12 compared with 205 ± 5.1 and 306 ± 11 compared with 200 ± 4.1 pmol/mg, respectively, and were seen using two different clones (13J and 13H) of GAD-expressing astrocytes. Thus, in the presence of glutamate, energy levels, determined by two separate methods (redox state and ATP levels), were higher in GAD-expressing astrocytes than in the parental cells.
Relationship between GAD expression and ATP levels.
To ascertain whether ATP stimulation was directly related to GAD activity, GAD expression was inhibited using a GAD antisense vector. C8-GAD67 cells were transfected with 6 μg of the original vector without cDNA (pKS) or with the vector containing GAD cDNA in the 3′-5′ orientation (GAD67-AS). After 36 h, 1 mM glutamate was added to the test cells ; 24 h later, the cells were lysed, and their GAD activity and ATP levels were measured (Fig. 2). In the absence of glutamate, GAD activity, determined by decarboxylation of [14C]glutamic acid, was inhibited by the anti-GAD vector (37.9 ± 5.7 compared with 90.9 ± 4.8 nmol/h/mg ; Fig. 2, upper panel), whereas no change in ATP levels was seen (157.1 ± 10.5 compared with 158.9 ± 29.4 pmol/mg ; Fig. 2, lower panel). In another experiment, ATP values were 188 ± 25.5 compared with 173.5 ± 24.5 pmol/mg. However, in the presence of glutamate, the anti-GAD vector inhibited GAD activity to the same extent (38.1 ± 0.5 compared with 89.3 ± 3.2 nmol/h/mg ; Fig. 2, upper panel) and also inhibited the increase in ATP levels (278.9 ± 7.8 compared with 206.9 ± 27.6 pmol/mg ; Fig. 2, lower panel). In another experiment, values decreased from 304.5 ± 8.5 to 226 ± 24.2 pmol/mg. Thus, the anti-GAD vector blocked GAD expression in both the presence or absence of glutamate but effected ATP levels only in the presence of glutamate.
Medium glucose and lactate levels and cellular glycogen levels.
Because glycolysis and glycogen are very important in astrocytes (Wiesinger et al., 1997), we measured glucose and lactate levels in the medium and glycogen levels in the cells in the presence of glutamate.
C8 and C8-GAD67 astrocytes were incubated in the presence of 25 mM glucose and 1 mM glutamate. After 1, 24 or 36 h, the amounts of glucose and lactate in the medium were measured using hexokinase/glucose-6-phosphate dehydrogenase and lactate dehydrogenase, respectively. The results, shown in Fig. 3, indicate that similar glucose consumption occurred in C8-GAD67 and C8 cells (167 ± 4.2 and 175 ± 1.9 μmol/mg, respectively, after 24 h and 108 ± 0.7 and 104 ± 0 μmol/mg, respectively, after 36 h) and that more lactate was progressively relased by C8-GAD67 than C8 cells (13.23 ± 0.68 and 8.98 ± 0 μmol/mg, respectively, after 24 h and 39.35 ± 2.75 and 21.51 ± 0.29 μmol/mg respectively, after 36 h). With another C8-GAD67 clone (13J), lactate release was 31.94 ± 0.34 μmol/mg after 36 h and 36.54 ± 2.15 μmol/mg after 48 h. Thus, GAD-expressing astrocytes produce more lactate but do not seem to use more glucose than control astrocytes.
Glycogen in cell lysates was quantified using the hexokinase/glucose-6-phosphate dehydrogenase assay after hydrolysis with amyloglucosidase. In confluent cells, the amount of glycogen was 44 ± 13 nmol of glucosyl units/mg of protein in GAD-expressing astrocytes and 39 ± 11 nmol of glucosyl units/mg of protein in control astrocytes.
GAD67-expressing astrocyte survival under glucose deprivation
The establishment of the GAD transgene in these cells allows more efficient energetic metabolism and lactate release. To examine the survival of these astrocytes under hypoglycemic conditions, the same number of C8 or C8-GAD67 cells was incubated for 24 h with decreasing concentrations of glucose (25-0 mM) in the presence or absence of 1 mM glutamate, and the number of surviving astrocytes was determined by trypan blue exclusion (Fig. 4). C8 cells in the absence or in the presence of glutamate were more sensitive to hypoglycemia than C8-GAD67 cells. Moreover, glutamate increases the resistance of C8-GAD67 cells to hypoglycemia.
In the next experiment, C8 or C8-GAD67 cells were incubated for 24 h with increasing concentrations of glutamate (0.01-10 mM) in the absence of glucose, and the number of surviving cells was measured as above (Fig. 5). Only a small percentage of C8 or C8-GAD67 cells survived, but even at the lowest glutamate concentrations, C8-GAD67 cells were more than twice as resistant, and, at glutamate concentrations >0.1 mM, resistance increased progressively with glutamate concentration, reaching ~60% at 10 mM glutamate.
Thus, under glucose deprivation conditions, GAD-expressing astrocytes survived better than control astrocytes, and their resistance was increased by the presence of glutamate, whereas this was not the case for C8 cells. Thus, glutamate, catabolized by the product of the GAD transgene, appears to compensate for a lack of glucose.
GAD67-expressing astrocytes enhance PC12 cell survival under glucose deprivation
Because GAD-expressing astrocytes were themselves protected against glucose deprivation in the presence of glutamate, it was possible that they might also protect neurons. To test this, we used PC12 cells, a pheochromocytoma cell line derived from glutamate-resistant chromaffin cells and considered a neuronal-like cell line (Greene and Tischler, 1976).
C8 and C8-GAD67 cells were cocultured with PC12 cells on separated supports. After 24 h, the cells were subjected either to normal conditions or to glucose deprivation in the presence or absence of 1 mM glutamate for a further 18 or 24 h, and then the surviving PC12 cells were quantified by trypan blue exclusion. The 24-h results are shown in Fig. 6, in which it can be seen that PC12 cells in the presence of either C8 or C8-GAD cells were resistant to 1 mM glutamate but sensitive to glucose deprivation (28.1 ± 3.8 and 29.4 ± 2.5% survival, respectively). In the presence of glutamate, this toxicity was not reversed by the parental astrocytes but was partially reversed by the GAD-expressing astrocytes (19.6 ± 2.9 and 50.3 ± 2.5% survival, respectively). In other experiments, values for C8 and C8-GAD cells were 32.2 ± 9.6 and 38.2 ± 4.8% in the absence of glutamate and 36.6 ± 6.7 and 70.8 ± 6.8% in the presence of glutamate, respectively. After 18 h, PC12 cells were more resistant, whereas the protection provided by GAD-expressing astrocytes was the same as that at 24 h. Thus, under conditions of glucose deprivation, PC12 cell survival can be enhanced by coculture with GAD-expressing astrocytes in the presence of glutamate.
Lactate and pyruvate enhance PC12 cell survival under glucose deprivation
Because the GAD-expressing astrocytes released lactate, to check whether the energetic metabolites themselves could protect PC12 cells, the latter were incubated alone for 18 h in the presence of 1 mM glutamate under conditions of glucose deprivation in the presence or absence of 9 mM lactate or pyruvate, and then the surviving cells were counted by trypan blue exclusion (Fig. 7). Like in the presence of astrocytes (Fig. 6), PC12 cells were sensitive to glucose deprivation (36.8 ± 1.4%), but when lactate or pyruvate was added, they were partially rescued, with the proportion of cells rescued being similar for the two additions (49.6 ± 0.8 and 51.3 ± 2.1%, respectively, for lactate and pyruvate). In another experiment, values were 46 ± 5.8% for glucose deprivation and 59.8 ± 2.8 and 60.6 ± 1.7% for lactate and pyruvate, respectively. Thus, lactate and pyruvate are both able to improve PC12 cell survival under glucose deprivation.
The aims of this study were to analyze the energetic metabolism of GAD-expressing astrocytes in the presence of glutamate and their effect on astrocyte and neuron survival under glucose deprivation. Increased MTT reduction and ATP levels, both evidence of enhanced energetic metabolism, were seen in the GAD-expressing astrocytes. In the presence of glutamate, these astrocytes were able to enhance their survival and PC12 cell survival under glucose deprivation. In these cells, glucose consumption was similar to that in control astrocytes, whereas lactate release was higher, suggesting that these protective effects involve intermediate metabolites, such as lactate.
Consequences of the establishment of a GAD transgene in astrocytes
GABA, the major inhibitory neurotransmitter synthesized by GAD, is also an intermediate of energy metabolism, being converted into succinate, a TCA cycle metabolite, by GABA transaminase and succinic semialdehyde dehydrogenase, both of which are found in astrocytes (Erecińska and Silver, 1990). The GABA pathway of the TCA cycle, the GABA shunt, is active in GABAergic neurons. The contribution of this pathway to the total TCA activity was initially underestimated (Balázs et al., 1970), but recently, using 13C NMR spectroscopy, its contribution was found to be approximately the same as that of the 2-oxoglutamate dehydrogenase pathway (Hassel et al., 1998). In addition, in primary astrocytes, glutamate dehydrogenase has been shown to be involved in glutamate degradation, whereas transamination is important in glutamate biosynthesis (Westergaard et al., 1996). When these cells are incubated with concentrations of glutamate <200 μM, most of the glutamate is metabolized into glutamine, and the glutamine excess is exported to neurons (Waniewski and Martin, 1986). It has recently been reported that induction of glutamine synthetase in astrocytes protects neurons against glutamate (Gorovits et al., 1997). However, at higher glutamate levels, the proportion of glutamate metabolized in the TCA cycle progressively increases, owing mainly to the action of glutamate dehydrogenase, which requires less energy than glutamine synthetase (McKenna et al., 1996). Thus, the requirement of ammonia and energy demands are major factors determining the metabolic fate of glutamate in astrocytes (Sonnewald et al., 1997).
C8 parental astrocytes, like primary astrocytes (Walz and Mukerji, 1988 ; Silver and Erecińska, 1997), produce lactate from glucose. GAD-expressing astrocytes also produce lactate but at a higher rate for the same glucose consumption. In primary astrocytes, the presence of a high extracellular glutamate concentration increases glycolysis and regulates energy metabolism by causing lactate release both in vitro (Pellerin and Magistretti, 1994 ; Tsacopoulos and Magistretti, 1996) and in vivo (Demestre et al., 1997), but in GAD-expressing astrocytes, glycolysis was not increased in the presence of 1 mM glutamate. As in our cells, glutamate-independent glycolysis has been seen in DI TNCI cells, another established astrocyte cell line (Pellerin et al., 1997) but the uptake of glutamate in these cells seems inefficient, and we have previously observed that during the first hour of exposure glutamate is taken up by both C8 and GAD-expressing astrocytes, excluding impairment of glutamate transporters (Sacchettoni et al., 1998). Moreover, enzymatic quantification of extracellular glutamine (using glutaminase and glutamate dehydrogenase) did not show any difference between C8 and GAD-expressing astrocytes (data not shown), indicating that the production of lactate does not result from differences in glutamine.
The astrocytes store the brain glycogen (Wiesinger et al., 1997) ; an increase in extracellular glutamate content enhances glycogen contents and reduces glucose utilization (Swanson et al., 1990). Thus, it is possible that a large part of the lactate release detected from GAD-expressing astrocytes might have been derived from glycogen, which might be present at higher levels in these cells owing to a more efficient glutamate metabolism. However, glycogen levels in C8 and GAD-expressing astrocytes were similar, being 250 times lower than the amount of lactate released and of the same order as that seen in primary astrocytes (Pellerin et al., 1997). Thus, the higher production of lactate in our conditions does not result from glycogen levels.
GAD-expressing astrocytes also showed a moderate increase of MTT reduction and higher ATP levels. Glutamate uptake is normally energy-consuming because the Na+/K+-ATPase pump, which maintains the potassium, sodium, and calcium gradients, consumes 20% of astrocytic ATP production (Silver and Erecińska, 1997). Na+ exchange could occur via reverse transport of GABA transporters because GABA was released. However, glutamate metabolism via the TCA cycle is an energy- and ammonia-producing pathway (Yu et al., 1992 ; Sonnewald et al., 1997). In our experiments, in the presence of glutamate, ATP levels in the C8 parental astrocytes remained stable, as previously seen in primary astrocytes (Yudkoff et al., 1986), whereas in GAD-expressing astrocytes, ATP levels increased to a plateau after 1 h. Thus, the presence of the GAD transgene leads to more efficient glutamate metabolism directed toward the TCA cycle. This was confirmed by the fact that the GAD antisense vector only inhibited the increase in ATP levels in the presence of glutamate, indicating a direct relationship between ATP levels and GAD activity. The pathways producing pyruvate and lactate from TCA can involve either phosphoenol-pyruvate carboxykinase and pyruvate kinase or malic enzyme, all of which are active in astrocytes (Alves et al., 1995 ; Wiesinger et al., 1997). Thus, lactate release from GAD-expressing astrocytes could be an indirect consequence of an efficient TCA cycle related to transgene expression via malic enzyme because ATP excess should inhibit various glycolytic enzymes, including pyruvate kinase (Wood, 1968).
Another possibility to explain lactate release would be the involvement of transaminases. As shown in Table 1, we observed a large release of GABA into the medium after 24 h in the presence of the combination of 1 mM glutamate and 50 μM GVG, a GABA transaminase inhibitor, suggesting that GAD and GABA transaminase are very active in this cell. In addition, GABA release was detected even in the absence of glutamate. In GAD-expressing astrocytes, the presence of extracellular glutamate, which stimulates Ca2+ entry, probably enhances GAD activity, as in synaptosomes (Erecińska et al., 1996). This newly synthesized GABA can stimulate GABA transaminase, which is already stimulated by GAD, as a decarboxylation-dependent transamination (Porter et al., 1985). Other transaminases, including those for branched-chain amino acids, are probably activated, and consequently pyruvate, which is converted into lactate, is produced from alanine. A recent study of [U-13C]glutamate metabolism indicates that addition of GABA to primary cultures of cortical astrocytes increases lactate synthesis (McKenna et al., 1998). Thus, as proposed (Hassel et al., 1997), the TCA cycle may operate more efficiently and/or more rapidly and produce more intermediate metabolites.
In conclusion, we hypothesize that this transgene, directed by the glial fibrillary acidic protein promoter, should allow the establishment of a new GABA shunt that enhances glutamate metablism and is indirectly responsible for NADH and ATP synthesis and lactate release. These effects probably involve phosphoenol-pyruvate carboxykinase, pyruvate kinase, and malic enzyme, all of which are active in astrocytes, and also transaminases, which are directly or indirectly stimulated by GABA synthesis.
Neuronal survival in the presence of GAD transgene-expressing astrocytes under glucose deprivation
The second hypothesis developed in this article is that the increased energetic metabolism of GAD-expressing astrocytes in the presence of glutamate may enhance neuronal survival. GAD-expressing astrocytes enhance PC12 cell survival under glucose deprivation. GAD-expressing astrocytes weakly increase the NADH/NAD ratio and more efficiently ATP levels, which are decreased in hypoglycemia (Erecińska and Silver, 1990). This neuronal protection requires the presence of glutamate, the substrate of the GAD enzyme. Thus, the synthesized GABA may be responsible either directly, as an intermediate metabolite, or indirectly, by interacting with its receptor after its release (Saji and Reis, 1987). Addition of GABA or a GABA agonist (muscimol) to the culture medium did not protect PC12 cells against hypoglycemia, and addition of a GABA antagonist (bicuculline) did not affect protection by astrocytes (data not shown). Thus, a paracrine effect of GABA is likely excluded in PC12 cell survival.
In the absence of glutamate, GAD-expressing astrocytes were more resistant than parental astrocytes to glucose deprivation (Fig. 4), and the presence of glutamate increased the resistance of GAD-expressing astrocytes to hypoglycemia. Because previous data indicate that glutamate uptake is stimulated in astrocytes during glucose deprivation or ischemia (Stanimirovic et al., 1997), glutamate probably penetrates into these cells where it may be used by the GAD transgene to compensate for the GABA decrease. In the brain during hypoglycemia, GABA content falls to a very low level (Paulsen and Fonnum, 1988), and in PC12 cells, the deleterious effects of glutamate are potentiated by hypoglycemia (Pereira et al., 1998). Thus, under glucose deprivation in the presence of glutamate, GAD-expressing astrocytes are more resistant and can enhance PC12 cell survival.
Many genes have been reported to be able to decrease neuronal death due to hypoglycemia. Genes coding for glucose transporters, transferred in neurons using defective herpesvirus, can protect neurons against hypoglycemia and ischemic insults (Ho et al., 1995 ; Lawrence et al., 1995). A similar result was obtained using the calbindin vector with a decreased Ca2+ concentration (Meier et al., 1997) and pyridoxal phosphate, which is required for GAD activity (Geng et al., 1997). Our results indicate that genetically modified astrocytes can produce more lactate and that this lactate may help PC12 cells to survive under hypoglycemic conditions, as previously seen on rat hippocampal slices (Izumi et al., 1994), in which lactate or pyruvate may serve as an alternative substrate for energy metabolism under conditions of glucose deprivation (Bröer et al., 1997 ; Izumi et al., 1997). Moreover, during glucose deprivation, intracellular calcium levels increase in hippocampal neurons as a result of increasing oxidative damage (Mattson et al., 1993), and α-ketoacids, such as pyruvate, produced by astrocytes, are able to protect neurons against hydrogen peroxide toxicity (Desagher et al., 1997). Thus, it would be interesting to test the efficiency of GAD-expressing astrocytes under conditions of oxidative stress and other insults, such as hypoxia or hypoxia plus hypoglycemia, because a reduction in brain energy levels, manifested by a fall in ATP levels, is seen in all of these injuries (Erecińska and Silver, 1990). Moreover, it cannot be ruled out that others factors released by astrocytes such as neurotrophic factors could also participate in neuronal survival (Ridet et al., 1997). Thus, we can speculate that the increased survival of genetically modified astrocytes can also enhance one or more of these factors, resulting in increased protection of neuronal-like and neuronal cells. In conclusion, the establishment of the GAD transgene in astrocytes confirms a relationship not only between glutamate metabolism and ATP production, but also between GABA and ATP levels. Astrocytic energy levels appear to be an interesting pathway in determining the extent of neuronal survival.