Selective γ-Aminobutyric Acid Type A Receptor Antagonism Reverses Isoflurane Ischemic Neuroprotection

The following studies were approved by the Duke University Animal Care and Use Committee (Durham, North Carolina).

Organotypic hippocampal slices were prepared from postnatal day 15 and 16 Sprague-Dawley rat pups (Zivic Laboratories, Pittsburgh, PA) as described by Stoppini et al .20and Sullivan et al .21with modifications. This age of pup was chosen because it has achieved an adult phenotype of the GABA^Areceptor.22Pups were anesthetized with 5% halothane, dipped in 70% ethanol, and decapitated. The hippocampi were removed and placed in chilled Gey's balanced salt solution with 5.5 mg/ml glucose, 6 g/ml HEPES, 50 mm adenosine, and 0.038 mg/ml ketamine. The hippocampi were transversely cut (300-μm thickness) using a tissue chopper (Siskiyou Design Instruments, Grants Pass, OR). Slices were placed on 30-mm-diameter membrane inserts (Millicell-CM; Millipore, Bedford, MA) and cultured in six-well sterile plates with 1.2 ml culture medium. This system allowed one surface of the slice to be continuously exposed to the gas phase, while the opposing surface was exposed to the culture medium. The culture medium consisted of 50% minimal essential medium (Invitrogen, Carlsbad, CA), 25% Earle's balanced salt solution (Invitrogen) and 25% Hyclone heat inactivated horse serum (Perbio, Cell Culture Division, South Logan, UT) with 4.125 g/l glucose, 50 mg/l Mg²SO⁴, 50 μm adenosine, 250 μg/l Fungizone™ (Invitrogen Corp., Grand Island, NY), and 25 U/ml penicillin. The medium was exchanged after the second and fifth days in culture to fresh medium without adenosine.

Slices were cultured for 7 days in a humidified incubator maintained at 37°C in 5% CO²–room air. Slices were then randomly assigned to experimental conditions. After completion of exposure to experimental conditions, the slices were incubated for 3 days.

Three days after exposure to experimental conditions, slices were imaged using a Leica inverted microscope (2.5×) (Wetzlar, Germany). The slices were placed in culture medium containing 5 μm Sytox (Molecular Probes, Eugene, OR). Sytox is a high-affinity fluorescent nucleic acid stain specific for cells with compromised plasma membranes and thus does not penetrate live cells.23,24After 1 h exposure to Sytox, fluorescent digital images were captured using a CoolSnap digital camera (Image Processing Solutions, North Reading, MA) at an excitation wavelength of 490 nm and emission of 590 nm. The parameters for imaging were standardized for all slices. Using an operator-controlled cursor, CA1 was outlined and fluorescence intensity was recorded. Background fluorescence was determined by manually outlining a region on the slice known to be acellular. This background value was subtracted from the CA1 fluorescence intensity, resulting in a corrected and standardized measurement of cell death. An observer made all observations blind to experimental condition.

Using the above system, an OGD dose-response curve was generated first. The culture medium was changed to an exposure medium (glucose-free 37°C Hanks balanced salt solution; Invitrogen Corp.). Culture plates were then placed in a thermoregulated chamber (37°C). The gas phase was altered to 95% N²–5% CO²flowing into the incubated chamber at 10 l/min. Pilot studies were performed which determined that this flow rate caused a reduction of the oxygen concentration to less than 5% within 8 min, after which flow was reduced to 1 l/min for the balance of the exposure interval. Oxygen concentration in the effluent gas was continuously monitored with a medical gas analyzer to assure it remained under 5% for the balance of the exposure interval. Timing of OGD duration was started after the chamber reached 5% O²concentration. OGD was terminated by exchange of the exposure medium to normal oxygenated culture medium and restitution of the ambient atmosphere to 5% CO²–room air at 37°C after return to the incubation chamber. Slices (n = 15 or 16 per condition) were exposed to 0–120 min of OGD. Three days after OGD, the slices were exposed to Sytox, and fluorescence was measured 1 h later. A dose-dependent effect of OGD duration was observed (0 min, 8 ± 9; 15 min, 30 ± 60; 45 min, 63 ± 90; 60 min, 157 ± 188; 90 min, 224 ± 166; 120 min, 322 ± 253 relative fluorescence intensity units; P < 0.0001). Sixty minutes of OGD was used for all subsequent experiments.

Next, a dose-response curve was constructed for isoflurane protection against 60 min of OGD. Slices were exposed to 0–3.5% isoflurane delivered via  a calibrated vaporizer in a 3-l airtight chamber. Isoflurane was delivered using the oxygen-deprivation gas (95% N²–5% CO²) as the carrier at 10 l/min until the oxygen concentration decreased to less than 5%, after which the carrier gas flow rate was reduced to 1 ml/min. Effluent isoflurane and oxygen concentrations were continuously monitored with a medical gas analyzer. For each condition, 7–18 slices were examined. A main effect for isoflurane concentration was observed (0%, 220 ± 173; 0.5%, 282 ± 220; 1%, 173 ± 196; 1.5%, 11 ± 10; 2%, 24 ± 24; 2.5%, 4 ± 8; 3.5%, 79 ± 92 relative fluorescence units; P < 0.0001). Depending on the experiment, either 1% or 2% isoflurane was used for subsequent studies.

Using these baseline conditions, we examined the effects of GABA^Aand GABA^Breceptor antagonism or agonism (and their interactions) in OHSs exposed to OGD with or without isoflurane. Isoflurane was delivered as described above. GABA^Aand GABA^Breceptor agonists and antagonists were added to the system with change of culture medium to exposure medium (see Results for respective concentrations). This treatment was reversed at the termination of OGD when the exposure medium was changed to culture medium. OHSs remained in culture for 3 days and were then imaged as described above.

Male Sprague-Dawley rats (aged 8–10 weeks; Harlan Sprague-Dawley, Inc., Indianapolis, IN) were fasted from food but allowed free access to water for 12 h before the experiments. The animals were anesthetized with 5% isoflurane in oxygen. After orotracheal intubation, the lungs were mechanically ventilated (30% O²–balance N²) to maintain normocapnia. The inspired isoflurane concentration was reduced to 1–2%. Surgery was performed with an aseptic technique. The tail artery was cannulated and used for blood pressure monitoring and blood sampling. Via  a ventral neck incision, the right jugular vein was cannulated with a silicone catheter for drug infusion and blood withdrawal. The common carotid arteries were encircled with suture. The vagus nerves and cervical sympathetic plexus were left intact. Muscle paralysis was provided by a 1-mg intravenous bolus of succinylcholine, repeated as necessary to allow control of ventilation during ischemia. Pilot studies had been performed to assure that rats would not exhibit an escape response in the absence of succinylcholine given the respective anesthetic regimens. Bilateral cortical electroencephalogram was continuously monitored throughout the experiment from active subdermal electrodes positioned over the parietal cortex bilaterally, a reference electrode placed on the nasion, and a ground lead positioned in the tail.

A 22-gauge needle thermistor (model 524; YSI Co., Yellow Springs, OH) was percutaneously placed adjacent to the skull beneath the temporalis muscle, and pericranial temperature was servoregulated (model 73ATA Indicating Controller; YSI Co.) at 37.5°± 0.1°C by surface heating or cooling. Heparin (50 U) was given intravenously. Exposure to isoflurane during anesthesia induction and surgical preparation lasted approximately 1 h.

Rats were randomly assigned to one of four groups. In three groups, delivered isoflurane was adjusted to 1.4% in 30% O²–balance N². In the remaining group, isoflurane was discontinued, and the inspiratory gas mixture was changed to 70% N²O in 30% O². In this group, fentanyl was given as a 10-μg/kg intravenous bolus followed by 25 μg · kg⁻¹· h⁻¹. The anesthetic doses were chosen on the basis of previous studies that showed short-term (i.e ., 3–5 days) outcome differences, as a function of these anesthetic conditions, in rats and mice subjected to 10 min of severe forebrain ischemia.2,3,6Anesthetic conditions were established 30 min before ischemia, during which time rats in all groups were allowed to stabilize physiologically. Arterial blood gases and hematocrit were measured 10 min before ischemia and at 10 and 60 min after ischemia. Blood glucose was measured at 10 min before ischemia.

The rats were treated as follows:

• Fentanyl-nitrous oxide: Anesthesia was maintained with fentanyl-nitrous oxide throughout ischemia as described above. Thirty minutes before ischemia, rats were given 1 ml intravenous 0.9% NaCl (n = 11).

• Isoflurane: Isoflurane, 1.4%, was continued throughout ischemia. Thirty minutes before ischemia, rats were given 1 ml intravenous 0.9% NaCl (n = 12).

• Low-dose bicuculline: Isoflurane, 1.4%, was continued throughout ischemia. Thirty minutes before ischemia, rats were given 1 mg/kg bicuculline ((−)-bicuculline methiodide; Sigma-Aldrich, Co., St. Louis, MO) in 1 ml intravenous 0.9% NaCl (n = 11).

• High-dose bicuculline (n = 13): Isoflurane, 1.4%, was continued throughout ischemia. Thirty minutes before ischemia, rats were given 2 mg/kg intravenous bicuculline in 1 ml NaCl, 0.9%.

Pilot studies had been performed on rats not subjected to ischemia. A dose de-escalation study was performed wherein rats were treated with intravenous bicuculline in the presence of 1.4% isoflurane. The high dose in this experiment was selected as the dose sufficient to cause minor electroencephalographic changes without evidence of epileptiform electroencephalographic activity. This was confirmed by absence of motor convulsions in the absence of neuromuscular blockade.

Ischemia was induced by withdrawal of 6–10 ml blood from the jugular catheter so as to reduce mean arterial blood pressure to 25–30 mmHg.25,26The carotids were then occluded with aneurysm clips, and a timer was started. Ischemia persisted for 10 min and was confirmed by the presence of an abrupt onset of electroencephalographic isoelectricity.

To terminate ischemia, shed blood was reinfused, and the aneurysm clips were removed. NaHCO³(0.3 mEq intravenous) was given to counteract systemic acidosis. The catheters were removed. The wounds were infiltrated with 1% lidocaine and closed with suture. Respective anesthetics were continued after ischemia for variable intervals of time (isoflurane, 110 min; fentanyl, 80 min; nitrous oxide, 110 min) to ensure that animals in both groups would recover the righting reflex at approximately 2 h after onset of reperfusion. Upon recovery of spontaneous ventilation, the trachea was extubated. Temperature regulation and electroencephalographic recording were continued until recovery of the righting reflex, at which time the thermistor and electroencephalographic electrodes were removed. Animals were allowed to recover in an oxygen-enriched environment (fraction of inspired oxygen = 0.4) and were then returned to their cages with free access to food and water.

After completion of the ischemia protocol, rats were allowed to recover for 5 days. On the final day of recovery, with the observer blinded to group assignment, functional tests were performed according to an established protocol, including assays of prehensile traction and balance beam performance.27,28The functional score was graded on a 0–9 scale (best score = 9).

After functional evaluation, the rats were anesthetized with isoflurane, and the brains were fixed in situ  by intraaortic infusion of buffered 10% formalin. After 24 h, the brains were removed and stored in 10% formalin. Paraffin-embedded brain sections were serially cut (5 μm thick) and stained with acid fuchsin-celestine blue. With the investigator blinded to group assignment, injury to the hippocampal CA1 sector (bregma −4.0 mm) was evaluated by light microscopy. Both viable and nonviable neurons were counted. Viable neurons were considered to be those with a blue hue and also having an intact plasma membrane and visible nucleus. Nonviable neurons were considered to be those that were pyknotic with an acidophilic reddish hue.29The total number of neurons (viable + nonviable) was calculated for each animal. The percentage of alive CA1 neurons was calculated as viable neurons/total neurons × 100 for each hemisphere. Because damage in the caudoputamen and neocortex is less discrete than that in hippocampal CA1, damage in the caudoputamen and neocortex was graded where the septal nuclei are widest, using a crude damage index where 0 = no observed histologic change, 1 = 1–5% neurons with pathologic changes, 2 = 6–50% neurons damaged, 3 = 51–100% of neurons damaged, and 4 = infarction.30By convention, values from the hemisphere with the greatest damage in each animal were used for the final analysis.

An additional six rats were studied as operative shams. These rats underwent procedures identical to those described above. Three rats were assigned to the isoflurane group, and three rats were assigned to the isoflurane plus 2 mg/kg bicuculline group. These rats underwent all anesthesia, surgery, and recovery procedures described above, with the exception of omission of carotid occlusion and systemic hypotension. The purpose of studying these shams was to determine whether the bicuculline dose studied in the ischemic injury was capable of inducing neuronal necrosis in the absence of ischemia.

Rats were subjected to isoflurane anesthesia and surgical preparation as described above. The rats were then mounted in a stereotactic head frame. The scalp was incised, and a 2-mm burr hole was drilled in the right parietal bone using continuous saline irrigation to avoid thermal injury in the brain. Using a micromanipulator, a glass microelectrode (tip diameter approximately 5 μm and intraparenchymal shaft diameter approximately 20 μm) filled with 4 m NaCl and containing an Ag-AgCl wire was inserted into the CA1 sector of the hippocampus (3.5 mm posterior to 2.0 mm left lateral, bregma, and 2.7 mm ventral to the cortical surface). Two screws were introduced into the skull, and the assembly of screws and glass microelectrode was secured to the skull with cyanoacrylate glue. The reference electrode was an Ag-AgCl disc (type E5SH; Grass Instruments Co., Quincy, MA) applied with electrode cream (EC2; Grass Instruments Co.) to shaved skin on the animal's neck. The direct current (DC) potential between electrodes was recorded using a H1P5 high-impedance input probe attached to a 7P122 Low Level DC amplifier (Grass Instruments Co.).

The animal was then dismounted from the frame and laid on its side. This was necessary because the ensuing fentanyl-nitrous oxide anesthetic was not sufficient to allow continued positioning in the stereotactic frame. The animals were then randomly assigned to one of three groups:

• Fentanyl-nitrous oxide (n = 5): Anesthesia was maintained with fentanyl-nitrous oxide throughout ischemia as described in Experiment 1. Thirty minutes before ischemia, rats were given 1 ml intravenous 0.9% NaCl.

• Isoflurane (n = 5): Isoflurane, 1.4%, was continued throughout ischemia. Thirty minutes before ischemia, rats were given 1 ml intravenous 0.9% NaCl.

• High-dose bicuculline (n = 6): Isoflurane, 1.4%, was continued throughout ischemia. Thirty minutes before ischemia, rats were given 2 mg/kg intravenous bicuculline in 1 ml NaCl, 0.9%.

Ischemia was then induced as described for experiment 2. DC potential, electroencephalogram, and mean arterial pressure were continuously recorded during ischemia and reperfusion until full repolarization was established. The animals were then killed with an overdose of isoflurane anesthesia. Correct position of electrode placement was verified at necropsy after injection of dye via  the microelectrode. DC potential shift after onset of ischemia and recovery of DC potential after onset of reperfusion were recorded on a strip chart recorder. These recordings were later evaluated by an observer, blinded to treatment group, to measure time to DC potential shift after onset of ischemia (defined as a 50% deflection from baseline) and recovery of DC potential after onset of reperfusion (defined as a 50% recovery to baseline potential).

For experiment 1, the primary dependent variable was relative Sytox fluorescence intensity in OHS CA1. Values were compared by one-way analysis of variance. When indicated by a significant F ratio, the Scheffé test was used to identify between group differences. For experiments 2 and 3, the primary dependent variables were percentage of alive CA1 neurons and time to onset of DC potential shift, respectively. One-way analysis of variance and the Scheffé test were used to compare percentage of alive CA1 neurons, time to onset of DC potential shift, and physiologic values. Nonparametric data (functional scores, and cortical and caudoputamen crude damage index values) were compared with the Kruskal-Wallis H statistic. When significant, post hoc  analysis was performed using the Mann-Whitney U statistic. Parametric values are presented as mean ± SD. Nonparametric values are reported as median ± interquartile range. A P  value less than 0.05 was considered statistically significant.

All experiments included 13–16 OHSs per experimental condition.

In the first experiment, the GABA^Breceptor antagonist phaclofen was added to 2% isoflurane in OHSs exposed to 60 min of OGD. Although isoflurane completely blocked cell death (i.e ., completely inhibited the OGD-induced increase in Sytox fluorescence), this was not affected by any concentration of phaclofen tested. Phaclofen treatment in the absence of OGD and isoflurane had no effect on Sytox fluorescence (fig. 1).

The GABA^Aantagonist bicuculline (9 × 10⁻⁵m) markedly attenuated isoflurane protection against OGD-induced Sytox fluorescence (P < 0.0001). Addition of phaclofen (9 × 10⁻⁴m) to bicuculline had no additional effect (P = 0.525). Sytox fluorescence in organotypic hippocampal slices not exposed to OGD or isoflurane was not increased by bicuculline alone, phaclofen alone, or bicuculline plus phaclofen (fig. 2).

The GABA^Areceptor agonist muscimol caused a dose-dependent reduction in OGD-induced Sytox fluorescence in the absence of isoflurane (fig. 3). Isoflurane (1%) reduced Sytox fluorescence by approximately 80% (P = 0.0001). This effect was similar to that provided by 4 × 10⁻⁵m muscimol (P = 0.99). The combination of 1% isoflurane and 4 × 10⁻⁵m muscimol was no different than either agent used alone (P = 0.99 for both tests).

Preischemic electroencephalogram recordings showed greater slowing with 1.4% isoflurane versus  fentanyl-nitrous oxide. Addition of bicuculline caused modest electroencephalographic activation, but no spindle or seizure activity was observed before or during ischemia or for 120 min after onset of reperfusion (fig. 4).

Physiologic values are reported in table 1. As expected, preischemic and postischemic mean arterial pressure values were lower in the isoflurane groups versus  fentanyl-nitrous oxide (P < 0.05). Preischemic body weight was similar among groups. There was a main effect for day 5 postischemic body weight (P = 0.009), with the isoflurane group having greater body weight than both the fentanyl-nitrous oxide (P = 0.04) and isoflurane plus 2 mg/kg bicuculline (P = 0.03) groups.

Functional scores are reported in figure 5. There was a difference among groups (P = 0.0004). The isoflurane group had better scores (8 ± 1) than the fentanyl-nitrous oxide (6 ± 1.75; P = 0.002), isoflurane plus 1 mg/kg bicuculline (6.5 ± 1.5; P = 0.04), and isoflurane plus 2 mg/kg bicuculline (5 ± 2; P = 0.0006) groups. There was no difference between the fentanyl-nitrous oxide versus  1 mg/kg bicuculline (P = 0.08) or 2 mg/kg bicuculline (P = 0.21) groups.

Figures 6 and 7depict histologic outcome. There was a main effect for treatment group in CA1 (P < 0.0001). The isoflurane group had a greater percentage of alive CA1 neurons (61 ± 25) versus  the fentanyl-nitrous oxide (15 ± 7; P = 0.003) or 2 mg/kg bicuculline (21 ± 15; P = 0.006) groups. Damage in the 1 mg/kg bicuculline group was intermediate (44 ± 32). There was no difference between fentanyl-nitrous oxide and the 2 mg/kg bicuculline groups (P = 0.96). A similar pattern was present for the caudoputamen (P = 0.02). There was no effect of treatment group on cortical injury (P = 0.13).

In sham-operated animals anesthetized with isoflurane in the presence or absence of 2 mg/kg bicuculline, there was no detectable injury (i.e ., acidophilic neurons) in any animal.

Physiologic values were similar to those in experiment 2 and therefore are not reported. Isoflurane reduced the time to depolarization relative to fentanyl-nitrous oxide (P < 0.0001). However, addition of 2 mg/kg bicuculline to isoflurane did not alter time to depolarization versus  isoflurane (P = 0.42). Time to repolarization was greater in the isoflurane group versus  both the fentanyl-nitrous oxide and isoflurane plus 2 mg/kg bicuculline groups (P < 0.0001), with the latter two groups having similar values (P = 0.71). See figure 8.

Isoflurane reduces ischemic brain damage, at least in the acute phase of recovery.1–6Although isoflurane is known to interact with the NMDA receptor in vitro  9–11,13and there is in vivo  evidence that isoflurane neuroprotection might be attributable to interactions with the same receptor system,12this theory was challenged by both in vitro  analysis in primary mixed neuronal-glial cell cultures and the serendipitous in vivo  observation that trimethaphan reversed isoflurane neuroprotection.3,13Because trimethaphan is classically considered a ganglionic blocker, subsequent work evaluated the potential role of isoflurane in modulating the stress response to ischemia. Several subsequent studies failed to demonstrate an association between volatile anesthetic and adrenergic or neuroendocrine events that could explain the positive effect of isoflurane on ischemic outcome.31–33 

There is considerable evidence that the GABA^Areceptor is a principal molecular site of volatile anesthetic action.34,35We were particularly intrigued by the finding that isoflurane inhibits the response of rat mechanosensory thalamocortical relay neurons to whisker stimulation, but this effect could be specifically reversed by local iontophoretic administration of bicuculline.36These data, and recognition that trimethaphan also is a GABA^Areceptor antagonist,14led to the hypothesis that isoflurane neuroprotection might be predominantly mediated by effects on the GABA^Areceptor. This was first tested in an OHS model of simulated ischemia, in which isoflurane protection has been demonstrated.15Coadministration of bicuculline reversed the protective effect of isoflurane. The current study has replicated that finding and extended it with the observation that the same effect was absent for the GABA^Bantagonist phaclofen. Further, the effect of isoflurane could be mimicked by the selective GABA^Aagonist muscimol.

Based on the above findings, a study was performed to determine whether reversal of isoflurane neuroprotection by a selective GABA^Aantagonist could be demonstrated in vivo . Indeed, bicuculline dose-dependently reversed the beneficial effects of isoflurane on hippocampal CA1 damage resulting from severe forebrain ischemia. This was associated with worsened behavioral outcome. Taken together, potentiation of Cl⁻influx via  the activated GABA^Aionophore would seem to be a sufficient mechanism of action to explain in the in vivo  neuroprotective properties of isoflurane.

Enhanced Cl⁻influx causes hyperpolarization of the postsynaptic neuron, resulting in inhibition of excitatory neurotransmitter-mediated depolarization in normal brain. Although most anesthetics slow depletion of high-energy phosphates at the onset of ischemia,7,37little is known about the specific neurophysiologic interaction between pharmacologic hyperpolarization and effects on intracellular energy state in nervous tissue deprived of metabolic substrate. However, Patel et al .38showed that isoflurane and pentobarbital, both of which act at the GABA^Areceptor, reduce the frequency of peri-infarct depolarizations in the rat focal ischemic cortex. It is also known from different models of acute brain injury that volatile anesthetics must be present during the ischemic injury to provide protection,39,40temporally implicating effects on neurotransmission. Ischemic depolarization is associated with massive intracellular Ca²⁺influx,41and the duration of depolarization has been associated with the magnitude of intracellular Ca²⁺accumulation in vitro .42We therefore hypothesized that the mechanism by which GABA^Apotentiation provides ischemic neuroprotection is via  reduction of the time to onset of ischemic depolarization. This would be consistent with previous observations that isoflurane reduces Ca²⁺influx in OHSs stimulated by glutamate,10not necessarily by NMDA receptor antagonism, but instead by potentiation of inhibitory afferentation.

In the current in vivo  experiment, isoflurane delayed the time to ischemic depolarization by 57 s relative to fentanyl-nitrous oxide. A delay of time to ischemic depolarization as brief as 90 s has been shown to cause measurable changes in ischemic histologic outcome.43However, unlike histologic and behavioral outcome, time to depolarization was not altered by 2 mg/kg bicuculline. Further, the onset of repolarization after reperfusion was more rapid with either fentanyl-nitrous oxide or isoflurane plus 2 mg/kg bicuculline than with isoflurane alone. Therefore, the mean duration of ischemic depolarization was shortest with isoflurane plus 2 mg/kg bicuculline, although this treatment group portrayed the same poor outcome as that observed in rats treated with fentanyl-nitrous oxide. We therefore conclude that modulation of duration of ischemic depolarization in this model of 10 min of severe forebrain ischemia does not explain the beneficial effect of isoflurane on outcome.

We have observed a similar phenomenon when studying the effects of hypothermia in rats subjected to near-complete forebrain ischemia.44Hypothermia (31°C) delayed the time to ischemic depolarization by 4 min. However, prolongation of the ischemic insult by 4 min did not reverse the histologic neuroprotective effect of hypothermia. These data indicate that times to onset and duration of ischemic depolarization do not necessarily reflect protective potential. This was further amplified by the work of Wang et al .,45who examined acute hippocampal slices exposed to hypoxia in the presence or absence of thiopental. Thiopental doubled the time to depolarization but, more interesting, profoundly reduced intracellular Ca²⁺accumulation, even after depolarization occurred. Therefore, the data in our experiment demonstrating delayed time to onset of depolarization by isoflurane, and the failure of bicuculline to reverse that, indicate a mechanism of isoflurane protection that is reversible by bicuculline but is independent of duration of ischemic depolarization. The above findings also suggest that our in vivo  data may be consistent with another in vitro  study. Gray et al .46examined OHSs subjected to 60 min of hypoxia with or without isoflurane. Isoflurane allowed only moderate hypoxia-induced intracellular Ca²⁺concentration increases (as opposed to much greater increases in the absence of isoflurane), and this was associated activation of the mitogen-activated kinase p42/44 pathway and antiapoptotic gene expression. Additional in vivo  study examining effects of isoflurane on Ca²⁺influx will help to establish this mechanism and its relation to GABA^A–ergic potentiation.

There is a potential weakness in the design of both the in vitro  and in vivo  experiments presented herein. It can be postulated that bicuculline worsened ischemic outcome independent of interaction with isoflurane at the GABA^Areceptor such that instead of reversing isoflurane protection, bicuculline initiated damage mechanisms in parallel to ischemia, thereby overriding isoflurane protection. For example, the duration of aortic occlusion required to cause paraplegia in awake rabbits was increased by treatment with intravenous muscimol but was decreased by a subepileptogenic dose of bicuculline.47Presumably, these drugs modulated endogenous γ-aminobutyric acid-mediated afferentation. Similarly, Bickler et al .15reported a small but statistically insignificant increase in OHS damage when slices were exposed to OGD in the presence of bicuculline alone. We observed no direct toxicity of 2 mg/kg bicuculline in our sham-operated rats. During ischemia, interactions between bicuculline and neuronal physiology might be different. Our study design provided no opportunity to observe independent worsening of ischemic outcome by bicuculline because the duration of ischemia (10 min) in fentanyl-nitrous oxide anesthetized animals was sufficient to cause near-maximal damage in our primary dependent variable, i.e ., CA1 cell death (fig. 6). Such a test presumably could be achieved by studying shorter durations of ischemia. However, some effect of bicuculline alone should persist if it serves to inhibit protective effects of endogenous γ-aminobutyric acid.48 

In conclusion, we examined interactions between γ-aminobutyric acid receptor antagonism and isoflurane in an OHS model of simulated in vitro  ischemia and in an in vivo  model of severe forebrain ischemia. In vitro , antagonism of the GABA^A, but not GABA^B, receptor reversed isoflurane neuroprotection, and the protective effect of isoflurane was mimicked by treatment with the GABA^Aagonist muscimol. In vivo , the relative neuroprotection caused by isoflurane compared with fentanyl-nitrous oxide was dose-dependently reversed by bicuculline. Despite this, bicuculline did not alter the delay in onset of ischemic depolarization caused by isoflurane. These data are consistent with the hypothesis that potentiation of the GABA^Areceptor by isoflurane, at least in part, accounts for its neuroprotective properties. Beneficial downstream intracellular consequences of this potentiation require further definition.