Key words: Amino acids, excitatory: aspartate; glutamate; glycine. Anesthetics, intravenous: pentobarbial. Anesthetics, volatile: isoflurane. Brain: cerebral protection; gamma-aminobutyric acid. Complications, ischemia: cerebral. Measurement techniques: microdialysis. Temperature: hypothermia. Toxicity: excitotoxicity.

IN recent years, mild to moderate hypothermia has been shown to protect the central nervous system against hypoxia and ischemia. [1-5] Because this cannot be explained easily by metabolic suppression, investigators have examined the effects of temperature on the postischemic extracellular concentrations of various neurotransmitters, particularly excitatory amino acids such as glutamate, aspartate, and glycine. While the exact mechanisms of "excitotoxicity" are still being investigated, there is no question that these compounds play an important pathophysiologic role. [6-8] In 1989 Busto demonstrated that while extracellular glutamate concentrations (as determined by microdialysis) increased almost fivefold during 10 min of forebrain ischemia, reducing cerebral temperature to 33 degrees C limited this to at most a twofold increase. [9] This has now been confirmed by several laboratories. [10-16] We also have examined the effects of a wider range of temperatures on postischemic glutamate concentrations and found them to be reduced in a temperature-related fashion. [17] Because it is unlikely that hypothermia can actively augment reuptake, these diminished extracellular concentrations almost certainly reflect a decreased rate of release.

Anesthetic agents, particularly the barbiturates and possibly isoflurane, also have been shown to protect against some ischemic insults, [18-25] although not to the same degree as hypothermia. [5] Like hypothermia, attention has been directed at the possible role of excitatory amino acids in anesthetic protection. Results have been contradictory. Koorn et al. showed that isoflurane and etomidate could blunt the increase in extracellular dopamine that resulted from temporary forebrain ischemia, [26,27] whereas Patel et al. have shown that isoflurane and etomidate could reduce ischemic glutamate release. [28,29] By contrast, Illeivich et al. were unable to show any differences in glutamate release between animals anesthetized with halothane, isoflurane, pentobarbital, or propofol (although propofol blunted the increase in glycine). [14] .

The current experiments were undertaken as an extension of our studies on hypothermia. As others, we used microdialysis to study the extracellular concentrations of excitatory amino acids and gamma-aminobutyric acid (GABA) after the onset of severe global cerebral ischemia (cardiac arrest). We specifically compared the effects of high-dose isoflurane and pentobarbital with those of a low-dose halothane anesthetic, and included a group of rats subjected to moderate hypothermia (t = 28 degrees C). These hypothermic animals have cerebral metabolic rates equal to those in the two high-dose anesthetic groups. [30] Finally, we also recorded cortical DC (direct current) potentials, in an effort to define the onset of terminal ischemic depolarization, an event which appears to either trigger or at least coincide with the rapid-rise phase of postischemic glutamate release.

All aspects of this study were approved by the University of Iowa Animal Care and Use Committee.

Male Sprague-Dawley rats, weighing 325-350 g, were used. Animals were fasted overnight (with free access to water) until anesthetized. Anesthesia was induced with 4% halothane in oxygen in a closed plastic box. After tissue infiltration with 1% lidocaine, a tracheotomy was performed and mechanical ventilation was started with an inspired mixture of 1.5% halothane in oxygen, using a tidal volume of 2.5-3.5 ml and a rate of 32-45 breaths/min. A femoral artery and vein were cannulated, again after tissue infiltration with 1% lidocaine, and were used for continuous blood pressure monitoring, arterial blood sampling, and the infusion of fluids and drugs, respectively. Muscle relaxation was achieved with 0.2 mg intravenous pancuronium given as needed. The animal was then turned prone and the head fixed in a stereotactic frame (David Kopf Instruments, Tujunga, CA). The scalp was infiltrated with 1% lidocaine and reflected laterally to expose the calvarium. A 2 x 2 mm left frontal craniectomy (located 2 mm rostral to the bregma and 2 mm lateral to the midline) and a similar-sized left parietal craniectomy (located [nearly equal] 7 mm caudal to the bregma and 5 mm lateral to the midline) were made with a high-speed drill. The drilling site was irrigated with cool saline to avoid thermal trauma, and care was taken to leave the dura mater intact. The inspired halothane concentration was then reduced to [nearly equal] 0.8% (as verified with a Datex Anesthetic Agent Monitor 222, DATEX Instrumentarium, Helsinki, Finland), combined with 50% N²O in oxygen. A continuous infusion of lactated Ringer's solution was started at the rate of 1 ml/h, and 6% hetastarch or rat donor blood was given as necessary to maintain mean arterial pressure (MAP) > 80 mmHg. Platinum needle electrodes were inserted into the temporalis muscles bilaterally to permit the recording of a single biparietal electroencephalogram (EEG). Temperature was recorded in each animal in two locations: a needle thermistor (YSI Model No. 524, Yellow Springs Instrument, Yellow Springs, OH) was placed into the pericranial tissue adjacent to the craniotomy and a YSI Model 401 probe was inserted rectally.

After these preparations, a small slit was made in the dura mater of the left frontal craniectomy, and a saline-filled glass micropipette with tip diameter of 5 micro meter was inserted 0.5 mm into the cortical surface, using a micromanipulator. Care was taken to avoid damage to cortical vessels. An Ag/AgCl wire in the barrel served as the electrical contact and an Ag/AgCl rod was inserted into the dorsal neck muscles as the reference. The DC potential between these electrode was measured using a Grass 7P122 amplifier (Grass Instrument, Quincy, MA) equipped with a Grass H1P5 high-impedance input probe. Then, through a slit in the dura mater of the left parietal craniectomy, a microdialysis probe (CMA/11; membrane 4-mm long, 0.24-mm OD, CMA, Stockholm, Sweden) was inserted 4 mm into the left parietal cortex. The probe was perfused by means of a microinjection pump (EP-60, Eicom, Kyoto, Japan) at a constant flow rate of 2 micro liter/min. The perfusion medium consisted of Ringer's solution with the following ion content in mM: 147 Sodium sup +, 4 Potassium sup +, and 2.3 Calcium sup ++, 155.6 Chlorine sup -, pH 6.0, 285 mOsm/kg H²O. In vitro recovery of probes were estimated to be between 10% and 20% for glutamate. This did not vary significantly with temperature.

Hematocrit, pH, partial arterial pressure of carbon dioxide, and partial arterial pressure of oxygen were determined intermittently (System 1306 pH/Blood Gas Analyzer, Instrumentation Laboratory, Lexington, MA), and ventilation was adjusted if necessary to maintain partial arterial pressure of oxygen > 100 mmHg and partial arterial pressure of carbon dioxide between 35 and 42 mmHg. All blood gas values were measured at 37 degrees C and reported without temperature correction.

When preparation was complete, pericranial temperature was adjusted to and maintained at target values of 38 degrees C or 28 degrees C, using a warming blanket in normothermic animals or an ice-water perfused water-jacket placed around animal's body in hypothermic animals. Fifty units of heparin was administered intravenously to all rats. All animals were allowed to recover for 90 min from any damage caused by probe implantation, and the target temperature was maintained for at least 60 min. At the end of this interval, a 10-min (20-micro liter) baseline sample was collected. The animals were then killed with 0.5 ml saturated potassium chloride given intravenously. The subsequent times (in s) to the appearance of an isoelectric EEG and to terminal ischemic depolarization (TD) were recorded. A 20-micro liter dialysate sample was then collected 11, 21, and 31 min after appearance of depolarization. The 1-min delay was introduced to minimize the amount of predepolarization dialysate that would be present in the first sample (i.e., to account for "dead space" in the dialysis tubing). Pericranial temperatures were maintained at the prearrest values throughout the ischemic period. Animals were not resuscitated.

Pentobarbital. Animals were prepared in the same manner as described earlier, except both femoral veins were cannulated, one for the infusion of fluids and the other for continuous infusion of pentobarbital. After the surgical preparation, anesthesia were maintained with 0.8% halothane in 50% N²O/O². The pericranial temperature was maintained at 38 degrees C using a warming blanket. Thirty minutes after insertion of the dialysis probe, an intravenous infusion of pentobarbital started at the rate of 5 mg *symbol* kg sup -1 *symbol* min sup -1. This was continued until a deep burst suppression or an isoelectric EEG was obtained (about 5 min). The pentobarbital infusion was thereafter reduced to 0.5-1.5 mg *symbol* kg sup -1 *symbol* min sup -1, and adjusted to maintain a similar EEG pattern (either isoelectricity or occasion bursts). Pericranial temperature was maintained at 38 degrees C, and MAP was kept > 80 mmHg by infusion of 6% hetastarch or rat donor blood if necessary. The baseline microdialysate sample was collected 90 min after microdialysis probe insertion. Intravenous potassium chloride (0.5 ml) was then given and, after TD was recorded, three consecutive dialysis samples were collected at 10-min intervals as described earlier.

Isoflurane. Animals were prepared as described earlier. Thirty minutes after dialysis probe insertion, halothane was discontinued, and anesthesia was thereafter maintained with 2.3-2.4% isoflurane in 50% N²O/O², adjusted to maintain an isoelectric EEG (or a tracing with occasional bursts of activity). Pericranial temperature was maintained at 38 degrees C, and MAP was kept > 80 mmHg by infusion of 6% hetastarch or rat donor blood if necessary. The baseline and postarrest microdialysate samples were collected as described earlier.

Microdialysis and High-performance Liquid Chromatography. Dialysis samples were analyzed using a high-performance liquid chromatography system (EP-10, Eicom, Japan) equipped with an ODC-C18 column (particle diameter 5 micro meter, column 6 mm ID, 150 mm long, Eicompack MA-5ODS, Eicom, Kyoto, Japan) and an electrochemical detector (ECD-100, Eicom, Kyoto, Japan), using o-phthalaldehyde precolumn derivatization. Phosphate buffer (80% 0.1 M) and 20% methanol was run at the rate of 1.0 ml/min as the mobile phase. Amino acid concentrations were determined using a calibration curve determined by external standards. Results are reported as directly measured dialysate concentrations, with no attempt made to correct for probe recovery rates.

Data Analysis. All physiologic variables were compared between groups using a factorial analysis of variance. Changes in dialysate concentrations versus time and versus temperature/anesthetics were examined using two-factor analysis of variance, with temperature/anesthetic as the between-groups variable, and with time treated as a repeated measure. Specific post hoc testing was done with a Fisher's least significant difference test.

Physiologic values recorded just before circulatory arrest are listed in Table 1. Mean arterial pressure and partial arterial pressure of oxygen values were higher in the two halothane-anesthetized groups (38 degrees C and 28 degrees C), whereas blood glucose was highest in the hypothermic animals. There were no significant differences in partial arterial pressure of carbon dioxide or pH level. The time to EEG isoelectricity was significantly longer in hypothermic animals. (Because the EEG was already isoelectric in the pentobarbital and isoflurane groups, no results are given.) Similarly, depolarization times were greater in all three of the experimental groups (halothane/hypothermia, pentobarbital, isoflurane), with the longest latency seen in hypothermic rats. All of these results are similar to those described previously. [30] .

Dialysate concentrations of glutamate, aspartate, glycine, and GABA are shown in Figure 1. There was a progressive increase in the concentrations of all four compounds with time after the onset of depolarization. Glutamate, aspartate, and glycine concentrations were lower in hypothermic animals (glutamate at all post-depolarization collection intervals, aspartate at the 20- and 30-min collection times, glycine at the 30-min interval, one-way analysis of variance), whereas there were no significant differences among the three normothermic groups (control, isoflurane, and pentobarbital). gamma-aminobutyric acid could not be detected under baseline conditions in any group, but increased progressively with increasing durations of ischemia. Like glutamate, the lowest GABA concentrations were seen in hypothermic animals. However, at all three post-depolarization intervals, concentrations in both the pentobarbital and isoflurane groups were lower than in normothermic control animals (and greater than in hypothermic animals).

The ability of hypothermia to protect against focal and global cerebral ischemia has been well demonstrated in the animal laboratory, and deep hypothermia is of unquestioned value in humans undergoing procedures involving circulatory arrest. The ability of anesthetics to protect the central nervous system has been less consistent. There is a large body of data showing that barbiturates can reduce the volume and/or severity of injury resulting from a focal ischemic insult or from incomplete global ischemia (for reviews, see references 31 and 32). The data concerning isoflurane are less convincing. Several studies of incomplete global ischemic have suggested some protective value, [23,24,33] but others examining focal ischemia have failed to find evidence of benefit. [20,22,34,35] Neither the barbiturates nor isoflurane have any demonstrated value in the face of severe global ischemia, such as that associated with a cardiac arrest. [5,36-38] .

The mechanism(s) by which hypothermia and anesthetics might protect the brain have long been a topic of discussion and investigation. For many years, it was believed that simple reductions in cerebral metabolic rate were sufficient to explain protection. However, as reviewed in the introduction, the value of mild hypothermia and the contrasting efficacy of barbiturates versus isoflurane (both of which reduce cerebral metabolic rate to an equivalent degree) have called this explanation into question. [5,39] In the late 1980s, attention was turned to the role of excitatory amino acid neurotransmitters. This was triggered by the finding by Busto et al. that mild hypothermia attenuated the increase in extracellular glutamate that accompanied severe forebrain ischemia. [9] This has been confirmed in numerous laboratories. [10-16] While the exact role of excitatory neurotransmitters in ischemic injury is still the subject of intense debate and investigation, there is no question that these compounds play a role in outcome, because drugs that block either their release or their action at various receptors have been shown to improve neurologic outcome. It is hence reasonable to conclude that the effects of hypothermia on extracellular excitatory amino acid concentrations during ischemia is linked to protective efficacy.

The mechanism by which hypothermia slows the increase in extracellular glutamate is not clear. In normal (nonischemic) brain, glutamate and aspartate release from storage vesicles is linked to synaptic activity, and is both a Calcium sup ++ and adenosine triphosphate-dependent process. However, glutamate release occurring after complete energy failure is independent of Calcium sup ++. Currently, the prevailing hypothesis is that postischemic release is mediated by reversal of the normal Sodium sup + -dependent glutamate/aspartate reuptake carriers, [40,41] which also implies that the increase in concentration is due both to enhanced release and inhibited reuptake. This is supported by our recent experiments demonstrating a temperature-related decrease in glutamate release after circulatory arrest. [17] Using these data, we calculated a hypothermic Q10 for postarrest glutamate release of [nearly equal] 3.5, a value too large to be the result of simple "diffusion," and indirectly supporting a carrier-mediated process. The mechanisms for ischemic glycine and GABA release are not known. However, their differing Q10 values ([nearly equal] 1.5 and 6.1, respectively) suggest different carrier-mediated mechanisms.

Anesthetics are well known to inhibit depolarization-induced vesicular glutamate release via the normal Calcium sup ++ /adenosine triphosphate-dependent paths. [42,43] The current experiments were designed to ensure that this process was not operative. This was done by the production of complete cerebral ischemia, an event that results in energy failure (e.g., adenosine triphosphate depletion) within several minutes. In addition, dialysis collection was delayed until terminal depolarization was observed. This event (TD) is typically associated with a sudden, rapid increase in interstitial glutamate concentrations, by a mechanism that must be energy independent and that is known to be independent of extracellular Calcium sup ++. [44-49] Our results indicate that under these conditions, neither pentobarbital nor isoflurane have any measurable effects. While it can only be speculation, we believe it possible that this difference may explain the greater protective efficacy of hypothermia better than differences in the direct metabolic effects of the different interventions.

One major question concerns the differences between our results and those of Patel et al. [28] These authors measured glutamate release in the cortex and hippocampus via microdialysis in normothermic rats subjected to 10 min of forebrain ischemia, and observed that isoflurane in both low and high doses substantially reduced measured glutamate concentrations. In fact, the concentrations observed in high-dose animals (EEG burst suppression) were similar to those seen in animals cooled to 34 degrees C. We believe there are several likely explanations. One unlikely cause might be strain of animals; we used Sprague-Dawley rats, whereas Patel et al. used Wistar rats. Another possibility concerns their use of a temporary ischemic event (ischemia followed by reperfusion). With reperfusion, glutamate concentrations rapidly normalize. When the duration of ischemia is limited, the extracellular concentration of glutamate during ischemia can be reduced by two mechanisms. The first is a delay in the time until the start of glutamate release, something that can be estimated by the time until TD. The second is a reduction in the absolute rate of release. Isoflurane does prolong the interval between the onset of ischemia and TD; in our experiment by almost 60 s. However, given a 10-min period of ischemia, this should produce roughly a 10% reduction in glutamate concentrations, not enough to explain experimental differences. We must then conclude that the differences are more likely owing to the rates of release. A more likely explanation is the severity of ischemia. Patel et al. employed a model of forebrain ischemia produced by bilateral carotid occlusion combined with hypotension to a MAP [nearly equal] 30 mmHg. This model does not reduce CBF to zero. [50] Furthermore, under these conditions, even small changes in MAP may result in changes in CBF sufficient to markedly influence the delivery of energy substrate, and which, under ischemic conditions, might have important effects of tissue viability. [50] This also may have an effect on the rate of energy failure. From previous studies, we know that depolarization times are much longer under the conditions of incomplete ischemia used by Patel et al. [51,52] If energy failure is incomplete, "normal" vesicular glutamate release pathways may still be active. This is supported by the much higher glutamate concentrations measured by Patel et al. suggesting that perhaps more than one glutamate release pathway may be operating. Because vesicular release is sensitive to anesthetics, it might explain the differences between our two studies. It might also indicate some protective value for anesthetics under conditions of incomplete ischemia. [23] .

Based on our results, it is interesting to calculate the hypothetical times until ischemic injury. Glutamate becomes cytotoxic at extracellular concentrations of [nearly equal] 100 micro Meter. [53] Given a recovery rate for our dialysis probes of [nearly equal] 20%, we can calculate that toxic extracellular concentrations would be reached in [nearly equal] 8.9 min in our normothermic control animals. For pentobarbital and isoflurane-anesthetized animals, the respective times would be 11.4 and 11.3 min, a difference entirely caused by the greater time until the onset of TD, and a difference that would be unlikely to be of clinical value. By contrast, for hypothermic animals, the increased time to TD, combined with the slower rate of release yields a time of 39.1 min. Again, while this must remain speculation, such calculations may partly explain the greater efficacy of hypothermia as a protective method.

In summary, when rats are subjected to total, irreversible cerebral ischemia via circulatory arrest, and when extracellular glutamate concentrations are measured from the onset of terminal depolarization, neither pentobarbital nor isoflurane have a significant impact on the increase in excitatory amino acid concentrations, although they do delay the onset of TD. By contrast, hypothermia to 28 degrees C has a major inhibitory effect on the increase of all measured compounds.

The authors thank Dr. Tsuyoshi Maekawa, Chairman, Department of Critical Care and Emergency Medicine, Yamaguchi University, and Dr. John Tinker, Head, Department of Anesthesia, University of Iowa, for their support of this work; David S. Warner, M.D., for his help during the development of this protocol; and EICOM Inc., Kyoto, Japan, for loaning the high-performance liquid chromatography equipment and the microdialysis perfusion pumps.