Acute Insult of Ammonia Leads to Calcium-dependent Glutamate Release from Cultured Astrocytes, an Effect of pH*

Hyperammonemia is a key factor in the pathogenesis of hepatic encephalopathy (HE) as well as other metabolic encephalopathies, such as those associated with inherited disorders of urea cycle enzymes and in Reye's syndrome. Acute HE results in increased brain ammonia (up to 5 mm), astrocytic swelling, and altered glutamatergic function. In the present study, using fluorescence imaging techniques, acute exposure (10 min) of ammonia (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{NH}_{4}^{+}{/}\mathrm{NH}_{3}\) \end{document}) to cultured astrocytes resulted in a concentration-dependent, transient increase in [Ca2+]i. This calcium transient was due to release from intracellular calcium stores, since the response was thapsigargin-sensitive and was still observed in calcium-free buffer. Using an enzyme-linked fluorescence assay, glutamate release was measured indirectly via the production of NADH (a naturally fluorescent product when excited with UV light). \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{NH}_{4}^{+}{/}\mathrm{NH}_{3}\) \end{document} (5 mm) stimulated a calcium-dependent glutamate release from cultured astrocytes, which was inhibited after preincubation with 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid acetoxymethyl ester but unaffected after preincubation with glutamate transport inhibitors dihydrokainate and dl-threo-β-benzyloxyaspartate. \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{NH}_{4}^{+}{/}\mathrm{NH}_{3}\) \end{document} (5 mm) also induced a transient intracellular alkaline shift. To investigate whether the effects of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{NH}_{4}^{+}{/}\mathrm{NH}_{3}\) \end{document} were mediated by an increase in pHi, we applied trimethylamine (TMA+/TMA) as another weak base. TMA+/TMA (5 mm) induced a similar transient increase in both pHi and [Ca2+]i (mobilization from intracellular calcium stores) and resulted in calcium-dependent release of glutamate. These results indicate that an acute exposure to ammonia, resulting in cytosolic alkalinization, leads to calcium-dependent glutamate release from astrocytes. A deregulation of glutamate release from astrocytes by ammonia could contribute to glutamate dysfunction consistently observed in acute HE.

Hyperammonemia consequently leads to increased concentrations of ammonia, up to 5 mM, in the brain. This high level of brain ammonia is a key factor in the pathogenesis of central nervous system dysfunction in acute and chronic liver failure. The nature and severity of the central nervous system disorder mainly depend upon the degree and acuteness of the onset of hyperammonemia (1). Acute liver failure (ALF) 1 resulting from viral infections or toxic liver injury is a life-threatening condition where hepatic encephalopathy (HE) develops rapidly and mortality rates are high due to brain stem herniation caused by increased intracranial pressure, a fatal consequence of cytotoxic brain edema. Excess ammonia is toxic to the brain resulting in deleterious effects, by both direct and indirect mechanisms, on cerebral metabolism and neurotransmission.
Over the past 10 years, there has been an increasing body of evidence demonstrating that ammonia toxicity is involved in alterations of glutamatergic synaptic regulation which is implicated in the pathophysiology of HE in ALF. Several reports have consistently described increased extracellular concentrations of brain glutamate in different models of experimental ALF (2)(3)(4)(5); however, neither the cell type nor the underlying release mechanisms have been identified. One possible explanation for the increased extracellular glutamate may be ammonia's inhibitory effects on the glutamate transporter system in astrocytes. It has been shown that ammonia inhibits glutamate uptake into astrocytes in vitro (6) and decreases protein and gene expression of the glutamate transporter GLT-1 (EAAT-2) in the frontal cortex of rats with ALF (7). The role of ammonia in the glutamatergic dysfunction demonstrated in HE is supported with a positive correlation between extracellular brain concentrations of glutamate and arterial ammonia concentrations in ALF in rats (4). In addition, using mild hypothermia as a treatment in rats with ALF, extracellular brain glutamate concentrations were normalized concomitantly with a lowering of brain ammonia (8).
Glutamate has been demonstrated to be an important signaling molecule for neuron-glia communication. Astrocytes express receptors and transporters for glutamate and recently have also been demonstrated to contain the protein machinery necessary to release glutamate by exocytosis through vesicles (9) and a fusion-related mechanism (10,11). Overall, astrocytes have many characteristics that were previously considered exclusive for neurons and are therefore actively involved in cell signaling by releasing glutamate. Astrocytic glutamate release is calcium-dependent and can be triggered by any ligand that stimulates an increase in [Ca 2ϩ ] i , such as bradykinins (12), prostaglandins (13), and ATP (14,15). Even a spontaneous [Ca 2ϩ ] i increase leads to glutamate release from astrocytes (16).
A rapid increase in ammonia results in an increase in pH i (intracellular alkanization) in all cell types, including astrocytes (17). It has been also demonstrated that intracellular alkalinization is accompanied with an increase in [Ca 2ϩ ] i in cultured acinar cells (18), in enthothelial cells (19), in pituitary cells (20), and in neurons (21). Furthermore, ammonia-induced intracellular alkalinization has been demonstrated to increase [Ca 2ϩ ] i in microglia initiating Ca 2ϩ release from thapsigarginsensitive stores (22).
The purpose of our study was to investigate whether ammonia-induced intracellular alkalinization could have an effect on [Ca 2ϩ ] i signaling in astrocytes and furthermore study whether the effects of ammonia could play a pathophysiological role in glutamate release from astrocytes.

EXPERIMENTAL PROCEDURES
Preparation of Cultured Astrocytes-Astrocytes were prepared from cortex of newborn NMRI mice as described previously (23). Briefly, cortical tissue was carefully dissected from blood vessels and meninges, rinsed with basal medium eagle solution, and incubated with 0.05% trypsin, 0.02% EDTA solution for 8 min at 37°C, trypsinized, and gently triturated with a fire-polished pipette in the presence of 0.05% DNase (Worthington). After washing cells twice, cells were cultured in 75-cm 2 plates on poly-L-lysine-coated coverslips using cultured basal medium Eagle's/10% fetal calf serum. One day later, cultures were washed twice with Hanks' balanced salt solution to remove cellular debris and maintained for 4 days. After reaching subconfluent state, cellular debris, microglia cells, oligodendrocytes as well as their early precursor cells were dislodged by manual shaking and removed by washing with Hanks' balanced salt solution. The purity of the astrocytes was routinely determined by immunofluorescence using a polyclonal antibody against glial fibrillary acidic protein (DAKO, Hamburg, Germany), a specific astrocyte marker. The cultures showed more than 90% cells positive for glial fibrillary acidic protein. Measurements were made from cells between days 11 and 15.
Solutions-All solutions were freshly prepared from refrigerated stock solutions. The standard bath solution was composed of 150 mM NaCl, 5.4 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, 10 mM glucose, pH adjusted to 7.4 by NaOH. To obtain calcium-free solution, CaCl 2 was omitted, replaced with equivalent amount of MgCl 2 and 1 mM of EGTA (calcium chelator) was added. For ammonia and trimethylamine solutions, NH 4 Cl or TMACl were added to the solution by replacing an equivalent amount of NaCl.
Fluorescence Imaging System-Cells were visualized under water immersion with a ϫ40 objective (numerical aperture 0.9) from a perfusion chamber mounted on a stage with an upright microscope (Axioskop FS, Zeiss, Oberkochen, Germany). Cells were superfused with standard physiological bath solution, and cell stimulation was achieved by changing the perfusate with ammonia or trimethylamine solution. Excitation light was provided by a monochromator (TILL Photonics, Munich, Germany), and the fluorescence emission was captured by a cooled CCD camera (SensiCam; PCO, Kelheim, Germany) and digitized by an image processing system (TILLVision; TILL Photonics, Munich, Germany). The monochromator and CCD camera were controlled by TILLVision software, which was used for image analysis. Ratio images were collected at intervals every 1 s.
Measurement of Extracellular Glutamate-Glutamate levels were detected using an enzymatic assay (13,24,25). In the presence of glutamate, GDH reduces NAD ϩ to NADH (Fig. 6 (inset), Reaction 1), a product that fluoresces when excited by UV light (360 nm). Provided that GDH and NAD ϩ are present in the cell medium, any glutamate released from the astrocytes into the medium will be detected as an increase in NADH fluorescence. NADH production can be amplified by adding glutamate pyruvate transaminase (GPT) and alanine (see Fig. 6 (inset), Reaction 2) to the medium. The newly formed ␣-ketoglutarate (from Reaction 1), driven by excess of supplied alanine, produces glutamate and pyruvate, forming a cycle and amplifying the NADH production (fluorescence).
To estimate the extracellular levels of glutamate, known concentrations of glutamate were applied by pipette using cell-free buffer in the presence of GDH (or GDH ϩ GPT), NAD ϩ , and alanine. The flow of the solution was stopped to allow NADH to accumulate in a concentration-and time-dependent manner. With GDH alone, the amplitude of the fluorescent signal reached a peak after 20 -30 s and sustained. With GDH ϩ GPT, the amplitude of the fluorescent signal never reached a plateau, and therefore to be able to compare both assays, fluorescence was measured after 10 min for all experiments. The approximate concentration of glutamate released from cells was estimated from the respective standard curves.
Data are expressed as NADH formation (percentage of base line), where base line represents the fluorescence level of the optical field before cell stimulation, the sum of fluorescence emitted from GDH ϩ NAD ϩ along with basal NADH (either as a contaminant or because of enzymatic activity). Data were collected from at least three different cultures.
Measurement of Intracellular Calcium-Cultured astrocytes on coverslips were incubated with 5 M fura-2/acetoxymethylester (AM) for 20 min in physiological buffer in the dark at room temperature. Cells were then washed with physiological buffer and stored in the dark for an additional 20 min to ensure fura-2/AM hydrolysis. Coverslips with loaded cells were transferred to the perfusion chamber and visualized under the microscope. Fura-2, a ratiometric dye, was excited with UV light at 340 and 380 nm, and the emission was measured at 530 Ϯ 10 nm.
The [Ca 2ϩ ] i was calculated from the ratio of fluorescence recorded at 340 and 380 nm excitation wavelengths using the equation of Grynkiewicz et al. (26) where R ϭ F 340 /F 380 , R min is the fluorescence ratio of calcium-free/fura-2 and R max is the ratio of calcium-bound/fura-2. The constant K d ␤ was determined empirically. The system was calibrated in situ by employing an ionomycin-based intracellular calibration procedure as described previously (27). The parameters K d ␤, R min , and R max characterizing the system were 1.6, 0.2, and 2.1 M, respectively.
Measurement of Intracellular pH-For measurements of pH i the H ϩ -sensitive fluorophore 2Ј,7Ј-bis-(carboxyethyl)-carboxyfluorescein (BCECF) was used. Cultured astrocytes on coverslips were incubated at room temperature in the dark with a 5 M concentration of the membrane-permeant BCECF/AM in physiological buffer for 15 min and then washed and stored in the dark for an additional 15 min to ensure BCECF/AM hydrolysis; BCECF, a ratiometric dye was measured at excitation wavelengths 440 and 488 nm. Emission was measured at 530 Ϯ 10 nm. An acid pH shift is demonstrated with a decrease in BCECF fluorescence. Calibrations of pH i with different pH solutions (6.0, 6.5, 7.0, 7.5, and 8.0) were made in the presence of high K ϩ (105 mM) solution containing nigericin (10 M) (28). Each calibration was repeated three times with 20 -30 cells from different cultures.
Glutamate Transporter Inhibitors-To block glutamate release, cells were pretreated, for 5 and 10 min prior to the experiments, with DL-threo-␤-benzyloxyaspartate (TBOA), an EAAT-1 glutamate transporter inhibitor, and dihydrokainate (DHK), an EAAT-2 glutamate transporter inhibitor at a final concentration of 10 and 100 M, respectively. TBOA and DHK are nonsubstrate glutamate transporter inhibitors. An important functional distinction between substrate and nonsubstrate inhibitors is that substrate inhibitors can induce release of excitatory amino acids by heteroexchange or exchange with an intracellular transporter substrate. Nonsubstrate inhibitors block uptake reversal through transporters by binding at the extracellular surface and therefore are not transported (29), preventing glutamate from being transported into and out of the cell.
An increase in [Ca 2ϩ ] i peaked within ϳ1 min and returned to base line within 5-8 min in the continuous presence of NH 4 ϩ / NH 3 ( Fig. 2A). The absolute amplitude increase was, on average, 66.3 Ϯ 4.4 nM (n ϭ 217) (Fig. 2D). Higher concentrations of NH 4 ϩ /NH 3 (10 and 20 mM) elicited a response with similar time course but with a higher amplitude, whereas NH 4 ϩ /NH 3 at 1 mM did not stimulate a significant increase in [Ca 2ϩ ] i (Fig. 2D) ϩ /NH 3 in physiological and calcium-free bathing solutions. NH 4 ϩ /NH 3 triggered similar calcium responses in the calcium-free bathing solution as compared with physiological (control) solution with respect to amplitude and time course and was effective at 5, 10, and 20 mM ( Fig. 2B) but not 1 mM (Fig. 2D). To support the view that NH 4 ϩ /NH 3 triggered calcium release from internal calcium stores, we used a paradigm to deplete endoplasmic reticulum (ER) stores; in the presence of thapsigargin (500 nM), a blocker for calcium transport into the ER stores, we applied ATP (100 M) to further deplete ER stores. As shown before (30), ATP triggered a large increase in [Ca 2ϩ ] i , and typically all subsequent metabotropic responses were abolished in the continuous presence of thapsigargin. When NH 4 ϩ /NH 3 (5 mM) was applied following this paradigm and still in the presence of thapsigargin, it failed to elicit a calcium response (n ϭ 91) (Fig. 3). We conclude that NH 4 ϩ /NH 3 triggers calcium release from thapsigargin-sensitive intracellular stores.
TMA ϩ /TMA, Another Weak Base, Mimics NH 4 ϩ /NH 3 -triggered [Ca 2ϩ ] i Increase and Alkaline Shift-An increase of NH 4 ϩ / NH 3 has been described to result in an alkalinization of the cell plasma in most cells, including astrocytes (31). We therefore used another weak base, trimethylamine chloride (TMA ϩ / TMA), to mimic the alkaline shift and test the impact of this alkalinization on [Ca 2ϩ ] i signaling. We measured intracellular pH using the cell-permeant pH-sensitive fluorophore BCECF/ AM.

. NH 4 ؉ /NH 3 stimulates a concentration-dependent transient increase in [Ca 2؉
i in cultured astrocytes in both physiological and calcium-free buffer.  (Fig. 5A). The amplitude was comparable with that of the NH 4 ϩ /NH 3 -triggered response, namely 69.9 Ϯ 5.3 nM (n ϭ 132) in physiological (control) and 65.9 Ϯ 4.8 nM (n ϭ 147) in calcium-free buffer (Fig. 5A). We then applied the above described paradigm to depleted ER stores, namely by adding thapsigargin to the bath (500 nM) and briefly applying ATP (100 M). When TMA ϩ /TMA (5 mM) was added (still in the presence of thapsigargin), it did not trigger a significant increase in [Ca 2ϩ ] i (Fig. 5B). Taken together, these data indicate that alkaline shifts trigger the release of calcium from cytoplasmic thapsigargin-sensitive stores.
An Enzyme-linked Fluorescence Assay Can Record Low Levels of Glutamate in the Presence of NH 4 ϩ /NH 3 -Glutamate released from cultured astrocytes has previously been recorded using an enzyme-linked fluorescence assay (13,24,25). With the enzyme GDH and the substrate NAD ϩ in the extracellular medium, glutamate produces ␣-ketoglutarate, NH 4 ϩ , and NADH ( Fig. 6 (inset), Reaction 1). Formation of NADH was recorded as a fluorescence signal when excited at 360 nm, and therefore changes in fluorescence reflect changes in glutamate levels in the bath. However, application of NH 4 ϩ /NH 3 results in product inhibition, decreasing the formation of NADH. Testing the degree of inhibition, different concentrations of NH 4 ϩ /NH 3 were applied with various concentrations of glutamate in a medium (without cells) containing NAD ϩ and GDH. NH 4 ϩ /NH 3 inhibited the production of NADH in a concentration-dependent manner, at a given (known) concentration of glutamate (Fig. 6A). At 5 M glutamate, NADH formation was barely detectable.
To increase the sensitivity to detect NADH, an enzymatic loop was produced to amplify the NADH production. This was accomplished by adding a second enzyme GPT and alanine ( Fig. 6 (inset), Reaction 2).
To demonstrate the difference between a nonamplified (GDH only) and an amplified (GDH ϩ GPT) response, typical traces are shown in Fig. 6B with 50 M of glutamate given in the absence and presence of NH 4 ϩ /NH 3 (5 mM). In the absence of NH 4 ϩ /NH 3 , a significant increase in NADH formation with GDH ϩ GPT (102%) was found compared with GDH alone (33%). NH 4 ϩ /NH 3 (5 mM) inhibited NADH formation in both assays, GDH and GDH ϩ GPT; however, a significant rise in NADH formation was demonstrated with GDH ϩ GPT (58%) compared with GDH (10%) (Fig. 6B). In the presence of GDH alone, NADH formation reached a plateau within ϳ3 min with or without NH 4 ϩ /NH 3 (5 mM). Since NAD ϩ is present in excess, this plateau is most probably due to the limited amount of glutamate (i.e. all glutamate given has been converted to ␣ketoglutarate). With GDH ϩ GPT, a continuous increase in NADH formation is demonstrated within the observation time of up to 20 min. Eventually, a plateau will be reached when all NAD ϩ or alanine is converted, but this obviously did not occur within 20 min. This implies that NADH formation must be measured with respect to time, and we chose 10 min as our temporal end point. NH 4 ϩ /NH 3 at concentrations of 1, 5, 10, 20 mM again caused a dose-dependent inhibition; however, the increase in NADH fluorescence was clearly detectable (Fig.  6A). Thus, with the combined enzyme system, GDH ϩ GPT, the level of NADH formation can be measured in response to application of glutamate as low as 500 nM in the presence of GDH ϩ GPT demonstrating the inhibition of NH 4 ϩ /NH 3 at 0, 1, 5, 10, and 20 mM at different glutamate concentrations is shown in Fig. 6C.
Due to the strong inhibition of NH 4 ϩ /NH 3 at 10 and 20 mM, NH 4 ϩ /NH 3 at 5 mM was further used to study the effects of ammonia on glutamate release from cultured astrocytes. Furthermore, 5 mM NH 4 ϩ /NH 3 is the pathophysiological concentration found in the brain in acute HE (32).
Application of NH 4 ϩ /NH 3 and TMA ϩ /TMA Triggers Glutamate Release from Cultured Astrocytes-With GDH ϩ GPT, NAD ϩ , and alanine present in the medium, NH 4 ϩ /NH 3 (5 mM) application to cultured astrocytes resulted in an increase in NADH formation as a measure of glutamate release. In physiological buffer, the basal NADH formation within the 10-min recording period was 9.3 Ϯ 1.5% (n ϭ 19 coverslips). In the presence of NH 4 ϩ /NH 3 (5 mM), it was significantly higher (23.7 Ϯ 2.5%, n ϭ 26 coverslips, p Ͻ 0.01). NADH formation occurred with a small delay after application of NH 4 ϩ /NH 3 (5 mM). We assume that, in the beginning, NADH levels are too low to be detected by our system, but with further amplification, NADH levels surpassed the threshold of our detection system (Fig. 7A). No significant increase in NADH was seen when a similar volume of physiological buffer (without NH 4 ϩ / NH 3 ) was applied as a control (n ϭ 11 coverslips). Furthermore, NH 4 ϩ /NH 3 (5 mM) did not increase NADH production in the absence of an enzyme-linked fluorescence system (n ϭ 5 coverslips) in the extracellular medium.
FIG. 6. Enzyme-linked fluorescence assay to record glutamate levels in the bath. A, the schematic diagram in the inset illustrates the detection assay. To detect levels of glutamate in the bathing solution, GDH (ϳ70 units/ml) and NAD ϩ (1 mM) are added to the extracellular medium. Upon glutamate release, GDH reduces NAD ϩ to NADH, which fluoresces when excited at 360 nm (reaction 1; see inset). NADH formation recorded as fluorescence indirectly reflects released glutamate concentrations. In the presence of GDH and NAD ϩ , NH 4 ϩ /NH 3 (1, 5, 10, and 20 mM) concentration-dependently inhibits (through product inhibition) the production of NADH with different concentrations of applied glutamate. Detectability of NADH formation is very low with 5 M glutamate. NADH formation can be amplified with the addition of GPT (ϳ23 units/ml) and alanine (1 mM) (reaction 2; see inset). With amplification, quantification of NADH formation is measured in relation to time. In all experiments, 10 min was the standard time used. Amplifying NADH formation with the addition of GDH ϩ GPT, in the presence of NAD ϩ and alanine, NH 4 ϩ /NH 3 (1, 5, 10, and 20 mM) again concentration-dependently inhibits the production of NADH with different concentrations of applied glutamate; however, levels of NADH are much higher as compared with levels without GPT and alanine. Detectability of glutamate increased to as low as 500 nM in the presence of 1 and 5 mM NH 4 ϩ /NH 3 . B, typical traces demonstrating the significant increase in NADH formation (with 50 M glutamate) with GDH ϩ GPT versus GDH in the absence (control) or the presence of NH 4 ϩ /NH 3 (5 mM). Note the plateau formation of NADH with only GDH (limited by glutamate), whereas with GDH ϩ GPT the loop is closed and amplification continues. C, logarithmic standard curve of NADH formation in the presence of different concentrations of NH 4 ϩ /NH 3 along with different concentrations of glutamate.
Approximate Concentrations of Glutamate Released upon Stimulation with TMA ϩ /TMA and NH 4 ϩ /NH 3 -Using the standard curves as shown in Fig. 6 for NH 4 ϩ /NH 3 and Fig. 8 for TMA ϩ /TMA, the approximate glutamate levels in the bathing solution after a 10-min stimulation were calculated. As shown in Fig. 9, a 10-min NH 4 ϩ /NH 3 stimulation resulted in an increase in glutamate between 4 and 8 M. For TMA ϩ /TMA, we estimated a similar increase in the bathing solution. With both weak bases, glutamate release was not affected when astrocytes were pretreated with TBOA ϩ DHK. However, glutamate release was inhibited when astrocytes were preincubated with BAPTA/AM. There were no significant differences in the amount of released glutamate between the NH 4 ϩ /NH 3 (5 mM) and TMA ϩ /TMA (5 mM) groups (Fig. 9).

Ammonia-triggered Release of Ca 2ϩ from Intracellular ER
Stores Is Due to a pH Shift-Our data indicate that NH 4 ϩ /NH 3 triggers a calcium release from intracellular ER stores, since 1) the response was similar in calcium-free buffer solution as compared with physiological (control) buffer and 2) the response was blocked after depleting internal calcium stores with thapsigargin, a potent inhibitor of the Ca 2ϩ -ATPase of FIG. 8. TMA ؉ /TMA stimulates NADH formation from cultured astrocytes. A, as described in the legend of Fig. 5, we determined the NADH formation in the presence and absence of TMA ϩ /TMA (5 mM). It is obvious that TMA ϩ /TMA (5 mM) did not influence the formation of NADH. B, TMA ϩ /TMA (5 mM) stimulated a significant increase in NADH formation compared with control in both physiological (TMA ϩ / TMA (5 mM); 40.4 Ϯ 3.8% versus control; 8.9 Ϯ 1.1%; *, p Ͻ 0.01) and calcium-free buffer (TMA ϩ /TMA (5 mM); 45.0 Ϯ 6.6% versus control; 10.7 Ϯ 1.9%; *, p Ͻ 0.01). Astrocytes preincubated with BAPTA/AM (5 M, for 15 min) did not demonstrate any significant increase in NADH formation upon TMA ϩ /TMA (5 mM) stimulation in calcium-free medium (BAPTA ϩ TMA ϩ /TMA (5 mM); 11.9 Ϯ 2.8% versus control; 10.7 Ϯ 1.9%). Astrocytes preincubated with glutamate transporters TBOA (10 M) and DHK (100 M) did not influence TMA ϩ /TMA (5 mM)-stimulated glutamate release (TMA ϩ /TMA (5 mM) ϩ TBOA ϩ DHK; 45.8 Ϯ 3.2% versus control; 13.6 Ϯ 2.8%; *, p Ͻ 0.01). TMA ϩ /TMA (5 mM) did not increase NADH production in the absence of GDH ϩ GPT. There was no significant difference in glutamate release within the TMA ϩ /TMAtreated groups or the control groups. Data are expressed as means Ϯ S.E. Significant difference between groups was calculated using a oneway analyisis of variance and post hoc Tukey's test. Differences were considered when p was Ͻ0.05. the ER. A similar mechanism has been described for microglial cells (22). NH 4 ϩ /NH 3 (5 mM) stimulated a transient increase in pH i in both physiological and calcium-free buffer. It is commonly known that NH 4 ϩ /NH 3 application induces an increase in pH i in many different cell systems. This alkaline shift is simply due to the rapid permeation of the gaseous NH 3 into the cytosol and the subsequent formation of a new NH 4 ϩ /NH 3 equilibrium according to the Henderson-Hasselbach equation rendering the cytosol alkaline (17). We therefore used another weak base, TMA ϩ /TMA, which at 5 mM also stimulated a similar increase in pH i as NH 4 ϩ /NH 3 (5 mM) in both physiological and calciumfree buffer; however, pH i returned to base line much more slowly. This can be easily explained by the fact that NH 4 ϩ , but not TMA ϩ , can pass through barium-sensitive K ϩ channels (33). In other words, the increase of pH i with NH 4 ϩ /NH 3 (5 mM) is to some extent attenuated by an influx of NH 4 ϩ ions through K ϩ channels (34). In support of this, when NH 4 ϩ /NH 3 (5 mM) was applied to cultured astrocytes along with BaCl 2 (a potassium channel blocker), a transient increase in pH i was observed; however, the return time to base line was longer, an effect similar to that seen upon TMA ϩ /TMA (5 mM) application. Since TMA ϩ /TMA (5 mM) mimicked the increases in pH i and [Ca 2ϩ ] i of NH 4 ϩ /NH 3 (5 mM), it supports the view that the calcium-dependent glutamate release by NH 4 ϩ /NH 3 (5 mM) is triggered by a cytosolic alkaline shift.
Our Assay Detects Glutamate Release in the Presence of NH 4 ϩ /NH 3 -Glutamate release can be measured with techniques involving preloading cells with radiolabeled glutamate and subsequently measuring radiolabeled glutamate release after stimulus or with high performance liquid chromatography. Both techniques have the disadvantage that temporal resolution is lost and glutamate reuptake affects quantitative results. Alternatively, glutamate levels can be determined by an enzymatic assay; GDH deaminates glutamate to form ␣ketoglutarate in concert with a conversion of NAD ϩ to NADH. Here any glutamate released from the astrocytes will be deaminated, and glutamate reuptake will be prevented. The concentration of the naturally fluorescent NADH is therefore proportional to the released glutamate. However, in this reaction, NH 4 ϩ is a by-product, and therefore when applying NH 4 ϩ /NH 3 , product inhibition occurs concentration-dependently and consequently inhibits NADH production. To overcome this inhibition, another enzyme, GPT, along with alanine, was added to amplify the NADH production, allowing for increased sensitivity for glutamate. With this amplifying system, glutamate release was measured quantitatively with respect to time and could be compared at a given NH 4 ϩ /NH 3 concentration. We used 5 mM NH 4 ϩ /NH 3 throughout the study, since 5 mM ammonia 1) demonstrates a significant increase in [Ca 2ϩ ] i , 2) is the pathophysiological concentration found in HE, and 3) does not display strong inhibition in the detection of glutamate (measured as low as 500 nM) with the amplifying enzymatic assay.
TMA ϩ /TMA (5 mM) did not have any inhibitory effects on the amplifying enzymatic assay and therefore resulted in higher NADH formation as compared with NH 4 ϩ /NH 3 (5 mM). When calculating the glutamate concentrations from the measured NADH levels using the respective NADH standard curves for NH 4 ϩ / NH 3 and TMA ϩ /TMA, similar concentrations of glutamate were released upon NH 4 ϩ /NH 3 and TMA ϩ /TMA application (Fig. 8). Potential Mechanism of Glutamate Release-Glutamate can be released from astrocytes, and several mechanisms have been proposed. Swelling induced opening of ion channels or reversal of glutamate transporters can occur, and these mechanisms are independent of an increase in [Ca 2ϩ ] i . An increase in cytosolic [Ca 2ϩ ] has also been shown to result in glutamate release, and this release activity was proposed to occur as a vesicular or fusion-mediated release (9 -11). Several ligands acting on receptors that trigger an increase in [Ca 2ϩ ] i in astrocytes initiate the release of glutamate from the cells. These include bradykinin (12), ATP (14,15), and glutamate (13,35) and even spontaneous increases in cytosolic [Ca 2ϩ ] (16).
In our experiments, a 10-min application of NH 4 ϩ /NH 3 (5 mM) triggered an efflux of glutamate in both physiological and calcium-free medium. However, increasing the intracellular buffering capacity with BAPTA/AM, the NH 4 ϩ /NH 3 (5 mM)-stimulated glutamate release was attenuated, strongly suggesting that ammonia stimulates glutamate release from astrocytes in a calcium-dependent manner.
Cell swelling-induced glutamate release has been demonstrated in cortical astrocytes exposed to a hypo-osmotic medium (36). Since astrocytic swelling is a pathological characteristic observed in acute HE, it was important to test whether glutamate release was due to cell swelling. In our study, using the isosbestic point of fura-2 (where the probe is Ca 2ϩ -insensitive and the emitted fluorescence is independent of intracellular calcium concentration), NH 4 ϩ /NH 3 (5 mM) application did not lead to astrocytic swelling (data not shown). Furthermore, with BAPTA/AM having no effect on cell swelling, if NH 4 ϩ /NH 3 (5 mM)-stimulated glutamate release was a result of cell swelling, glutamate release would have persisted when astrocytes were pretreated with BAPTA/AM, which was not the case. Therefore, the ammonia-induced glutamate release is not due to cell swelling. In support, Albrecht's group has shown that a 10-min treatment with 5 mM ammonia did not produce cell swelling in cultured rabbit Muller cells (37) or cultured cortical astrocytes (38). This further suggests that ammonia-induced astrocytic swelling may appear to develop with longer treatments of ammonia but not during a 10-min insult. FIG. 9. Approximate concentration of glutamate released upon stimulation with 5 mM NH 4 ؉ /NH 3 or TMA ؉ /TMA. Using the respective standard curves for NH 4 ϩ /NH 3 as displayed in Fig. 5 and TMA ϩ /TMA as displayed in Fig. 7, approximate concentrations of released glutamate were calculated. Within the control and BAPTA/AM groups, the approximate released glutamate is within the range of 280 -490 nM. Within the stimulated groups (NH 4 ϩ /NH 3 (5 mM) and TMA ϩ /TMA (5 mM)), the calculated concentration of released glutamate ranged from 4 to 8 M. Data are expressed as means Ϯ S.E. Significant difference between groups was calculated using a one-way analyisis of variance and post hoc Tukey's test. Differences were considered when p was Ͻ0.05.
Astrocytes contain high affinity Na ϩ -dependent glutamate transporters to regulate the concentration of extracellular glutamate. GLT-1 and GLAST, more commonly referred to as EAAT-2 and EAAT-1 respectively, are located in rat forebrain and cerebellum at birth. Szatkowski et al. (39) demonstrated that glutamate transporters reverse when insufficient energy is available to regulate the membrane potential. There is increasing evidence that an energy impairment develops in acute HE (40). We therefore tested the effect of glutamate transport inhibitors on NH 4 ϩ /NH 3 (5 mM)-triggered glutamate release. TBOA, a nonsubstrate EAAT-1 inhibitor, and DHK, an effective nonsubstrate EAAT-2 inhibitor, did not affect the amount of glutamate release following NH 4 ϩ /NH 3 (5 mM) application, suggesting that ammonia stimulated glutamate release is not due to the reversal of glutamate transporters. NH 4 ϩ /NH 3 (5 mM) induced an increase in [Ca 2ϩ ] i (ϳ65 nM; Fig. 2D), which, according to Parpura and Haydon (41), is sufficient to trigger glutamate release. In their study, a ⌬[Ca 2ϩ ] i of 56 nM released enough glutamate to result in a slow inward current on neighboring neurons. In addition, further increased [Ca 2ϩ ] i leads to more glutamate release, suggesting a dose-response relationship (16). Taken together, our data point to a exocytotic glutamate release from astrocytes in response to NH 4 ϩ /NH 3 . Astrocytes Release a Substantial Amount of Glutamate in Response to NH 4 ϩ /NH 3 or TMA ϩ /TMA-We estimated the glutamate levels in the bathing solution after release from astrocytes to between 4 and 8 M for NH 4 ϩ /NH 3 (5 mM) and TMA ϩ / TMA (5 mM). If we assume that the astrocytes on our coverslip have a volume of 300 ϫ 10 Ϫ12 m 3 (assuming a continuous layer on the coverslip with an average thickness of 5 m) and a given bathing volume of 300 l (equal to 300 ϫ 10 Ϫ9 m 3 ), the intracellular astrocyte concentration prior to release would be 4 -8 mM. This is about in the range as previously estimated, namely at 1-10 mM (42). It also implies that in the central nervous system extracellular space, glutamate would increase substantially. Assuming that astrocytes make up 30% of the brain volume and the extracellular space is about 20% (43), the same amount of glutamate released would lead to an extracellular glutamate concentration of 6 -12 mM. Even if glutamate transporter activity might attenuate this high level, millimolar glutamate levels are realistic to assume, and at this concentration glutamate can result in hyperexcitability.
Significance of NH 4 ϩ /NH 3 -triggered Glutamate Release for HE in ALF-Astrocyte dysfunction has been assumed to be an important event in the pathologic cascade of HE. Swelling of the astrocytes is a major consequence that leads to brain edema, intracranial pressure, and fatal brain stem herniation. Increased extracellular brain glutamate has been consistently observed in different animal models of ALF where brain edema and increased intracranial pressure prevails. Our data indicate that NH 4 ϩ /NH 3 could trigger the release of glutamate from astrocytes, possibly leading to an increase in extracellular brain glutamate and consequently resulting in glutamatergic dysfunction and an overstimulation of NMDA receptors on neurons (44). This would result in hyperexcitability, and it is known that seizures are not uncommon in patients with ALF. The astrocytic end feet contact the blood capillaries and are thus the first elements to be exposed to elevated NH 4 ϩ /NH 3 from the blood. The level of NH 4 ϩ /NH 3 used in our study is in the range observed after ALF (32). For this to occur in vivo, NH 4 ϩ /NH 3 fluctuations at the astrocyte membrane in the millimolar range would have to arise; however, precise temporal NH 4 ϩ /NH 3 transients in vivo data are not available. Ammonia fluctuations and subsequently cytosolic alkaline shifts are larger at the onset of ALF, suggesting that a deregulation of glutamate release by ammonia from astrocytes may be an early phenomenon and in addition one of the sources leading to this increased extracellular brain glutamate/glutamatergic dysfunction consistently found in different models of ALF (45).