Anesthetics and Mild Hypothermia Similarly Prevent Hippocampal Neuron Death in an In Vitro Model of Cerebral Ischemia

Hippocampal brain slices were prepared from 2- to 3-month-old 200- to 250-g Sprague-Dawley rats using methods approved by the Committee on Animal Research, University of California, San Francisco. After decapitation during 2% halothane anesthesia, hippocampi were rapidly removed and placed in ice-cold artificial cerebrospinal fluid (aCSF). The aCSF used to prepare and maintain the slices contained NaCl 116 mm, NaHCO³26.2 mm, KCl 5.4 mm, CaCl²1.8 mm, MgCl²0.9 mm, NaH²PO⁴0.9 mm, and glucose 5.6 mm, pH 7.40, bubbled with 5% CO²–95% O². Both hippocampi were sectioned with a tissue chopper into 400-μm transverse slices, yielding 10–16 slices with intact CA1, CA3, and dentate cell body fields. Following sectioning, slices remained in oxygenated aCSF at room temperature for about 1 h. Within this time, our slices recovered synaptic function as measured by CA1 population spike amplitude and excitatory postsynaptic potentials.

Slices were subjected to simulated ischemia by transferring them with a plastic pipette to glass vials containing 23 ml glucose-free aCSF bubbled with 95% N²–5% CO²containing 1 mm dithionite, an oxygen absorbent. The P^O2of this solution was less than 0.1 mmHg as measured using a Clark oxygen electrode (Cameron Instrument Co., Port Aransas, TX). The vials containing the slices were immersed in a water bath so that aCSF temperature could be controlled to 37 ± 0.2°C. The temperature of the aCSF in the vials was monitored with a thermocouple thermometer. After the desired period of simulated ischemia, slices were removed and recovered in oxygenated aCSF at 37°C for 5 h. This recovery period was used to allow the development of cell injury and death that may not be evident immediately following an ischemic insult.

Experimental treatments present during simulated ischemia included 0.7% and 2.0% isoflurane added to the gas mixtures used to bubble the vials containing the brain slices. A calibrated isoflurane vaporizer was used to prepare these gas mixtures, and isoflurane concentration in the gas bubbling the slices was determined in the gas headspace in the vial and adjusted with an infrared anesthetic analyzer (Datex AS3, Helsinki, Finland). Isoflurane concentration in the aCSF was measured with a gas chromatograph. Less than 5% loss occurred between the vaporizer and the liquid in the vial. The anesthetic containing gas was bubbled in the vials 30 min prior to introduction of the slices. The other treatments were present during the entire time of the in vitro  anoxia–zero glucose stress.

Following recovery, slices were fixed in 10% formalin or 4% paraformaldehyde in buffered saline overnight at 4°C. Further handling included paraffin embedding, sectioning, and staining with cresyl violet, methylene blue–azure II, or hematoxylin and eosin. Paraffin-embedded slices were sectioned in the same transverse direction as the original slices to permit examination of the CA1, CA3, and dentate cell body fields in the same slice. Specimens were examined by a single blinded observer to determine percentage of cell loss (death) and percentage of cells exhibiting morphologic changes. The interior region of the slice (≈ 100 μm from the edge) was examined to avoid areas subjected to slicing trauma during preparation. Under 1000× magnification, the percentage of dead neurons (holes the size and shape of neuron cell bodies;fig. 1) and the percentage of morphologically damaged neurons were counted. A reticle in the ocular of the microscope was used to count neurons in the same size area in each cell body field. The area circumscribed by the reticle was approximately 50 × 75 μm. Two sections of each slice were examined and results averaged. Neuronal damage was determined by the presence of one of the following characteristics: cell swelling, vacuolization (commonly associated with light-staining cytoplasm), or presence of shrunken, darkened nuclei. 15,16 

Comparison of cell death among different treatments was done using chi-square analysis because binary scoring was used (i.e. , we determined the percentage of remaining and lost neurons or damaged or pristine neurons). Because our study involved the comparison of multiple treatment groups, we used a multiple comparison procedure for chi-square tests 17to identify which groups were significantly different from expected. Statistical significance was assumed at P < 0.05. N values shown in the figures indicate the total number of brain slices examined in each group; 2–3 slices per individual animal were used. Each experimental trial included a control anoxia–aglycemia group, resulting in greater cumulative numbers of slices in those groups. In addition, slices treated with 50 μm sodium pentothal were included as positive controls with other neuroprotectants tested during the 20-min anoxia–aglycemia groups, similarly resulting in a higher N value.

In vitro  ischemia of 10 or 20 min followed by a 5-h recovery period produced a characteristic pattern of injury in hippocampal slices that depended on the cell type and temperature of the stress. CA1 neurons were most susceptible to injury, with 35 ± 6.9% (mean ± SD) cells dead and 33 ± 10% of the remaining neurons morphologically damaged following 20 min in vitro  ischemia at 37°C (figs. 1A, 1B, 2, and 3). In contrast, neuron cell bodies in the CA3 and dentate were less damaged by 20 min of in vitro  ischemia at 37°C. Following ischemia at 40°C, more extensive cell loss was observed in the CA3 and dentate (fig. 2). In some slices in the 40°C group, extensive cell loss appeared uniformly distributed throughout the different cell body regions. Mild hypothermia (33–34°C) prevented cell loss in all three areas (figs. 1F and 2). Because of this pattern of injury, we focused the remainder of the study on CA1 neurons.

The swelling and cell death of CA1 neurons produced by 10 or 20 min of in vitro  ischemia was completely prevented by 10 μm MK-801, a noncompetitive NMDA receptor antagonist (P < 0.001;fig. 3). CA1 cell damage during in vitro  ischemia in this model is therefore dependent on processes involving activation of the NMDA receptor.

Isoflurane reduced cell loss and number of swollen CA1 neurons exposed to either 10 or 20 min of in vitro  ischemia (figs. 1C, 1D, and 3). Both 0.7 and 2.0% isoflurane produced similar degrees of protection. Sodium thiopental (50 μm, ≈ 1 minimum alveolar concentration equivalent) similarly reduced cell loss and swelling in both 10- and 20-min ischemia groups (figs. 1E and 3).

Because cell death from in vitro  ischemia in the hippocampal slices appeared to be NMDA receptor dependent, we tested whether the protection afforded by isoflurane is mediated by inhibition of glutamate receptors. Glutamate (1 mm) present for 20 min during aerobic conditions (slices bubbled with 95% O²–5% CO²) killed about one fourth of the CA1 neurons (fig. 4). This toxicity was blocked completely by 10 μm MK-801, indicating that NMDA receptors play a role in glutamate toxicity, as was the case in ischemia. Isoflurane (2%) prevented glutamate-induced cell death, suggesting that a mechanism for the observed protective effects of isoflurane in our model is attenuation of glutamate-mediated excitotoxicity.

Our results show that the volatile anesthetic isoflurane, the intravenous anesthetic sodium pentothal, and mild hypothermia reduce cell death to similar degrees in an in vitro  brain slice model of cerebral ischemia in which acute cell death is largely due to NMDA receptor activation. The results also suggest that the mechanism of neuroprotection afforded by isoflurane involves attenuation of glutamate excitotoxicity.

Volatile anesthetics reduce infarct volume in in vivo  focal ischemia in various rodent models, 1–4as well as protecting against excitotoxic (NMDA) injury in cultured neurons. 14To our knowledge, this is the first report showing that a volatile anesthetic attenuates neuronal cell loss in a brain slice model. Several previous studies of anesthetic neuroprotection in in vitro  ischemia in brain slices have used measures such as rise in intracellular calcium concentration, 10release of glutamate, 9or failure of synaptic transmission 12to indicate cell damage without direct knowledge that cell death would necessarily follow. Cell death in hypoxic–ischemic brain insults involves a cascade of events which may lead to early cell death by necrosis and delayed cell death with features of necrosis or apoptosis. 13,18,19Our model only allowed examination of the consequences of early events in this cascade.

An important finding of this study was that isoflurane and sodium pentothal showed protection similar to mild hypothermia. Mild hypothermia is markedly protective in animal models of cerebral ischemia and provides protection against delayed cell death. 20The mechanisms by which mild hypothermia lead to protection against cell damage from cerebral ischemia have not been definitively determined but may involve the attenuation of glutamate accumulation and excitoxicity, 5mechanisms not unlike those associated with anesthetics.

We found that 0.7% isoflurane was just as effective in reducing neuron death as a dose of isoflurane (2%) associated with a nearly silent electroencephalogram in humans. 21Thus, electroencephalographic silence or profound reduction of metabolic rate are not synonymous with brain protection. 6,7,22Our data suggest that the cerebroprotective qualities of isoflurane may reside in the inhibition of events associated with glutamate excitotoxicity. This protection is apparently achieved as well by 0.7% isoflurane as by 2% isoflurane. Because only those two concentrations were studied, we cannot infer a dose–effect relation.

Cell injury and death in our model of in vitro  ischemia depends on NMDA receptor activation as at least one mechanism. We also showed that isoflurane attenuated injury caused by glutamate. Together, these data suggest that the protective effects of isoflurane during ischemia may occur through attenuation of NMDA receptor–dependent processes. This proposal is consistent with data showing that isoflurane decreases NMDA receptor activity. 10,11In addition, volatile anesthetics may increase glutamate uptake. 23Thus, it can be suggested that volatile anesthetics such as isoflurane may produce their protective effects via  a mechanism involving the attenuation of glutamate toxicity, either by reducing its release, 8,9enhancing its uptake, or inhibiting postsynaptic glutamate receptor activity. Models of cerebral ischemia in intact rodent models show similar dependence on glutamate excitoxicity. Compounds such as the NMDA receptor antagonist MK-801 protect against cerebral injury in focal ischemia, 2,24,25transient global ischemia, 26and in vitro  models of ischemia. 27Protective effects include suppression of spreading depression, a component of which is glutamate receptor–dependent. 28This conclusion does not exclude other neuroprotective actions mediated by inhibition of AMPA or kainate glutamate receptors or sodium channels.

The volatile anesthetics and sodium pentothal also produce significant GABAergic inhibition of CA1 neurons in the hippocampus. 29Augmentation of GABA currents by these agents would indirectly decrease the activation of NMDA receptors on CA1 neurons because of membrane hyperpolarization in the presence of these anesthetics. 30Taken together, it is possible that a combination of direct and indirect attenuation of NMDA receptor–dependent processes accounts for the neuroprotective efficacy of isoflurane and sodium thiopental on CA1 neurons.

Cell death in hypoxic–ischemic injury involves a cascade of injury that may terminate in delayed cell death. It is therefore probable that we underestimated the eventual extent of cell death by fixing brain slices for only 5 h after the insult. Had the slices a longer time to develop injury, some of the neuroprotective effects may have disappeared. We have addressed these shortcomings by developing an organotypic hippocampal slice culture model of delayed cell death following in vitro  ischemia. In slices exposed to 1% isoflurane, we observed reduction of cell death in CA1 72 h after simulated ischemia (unpublished observations). In the current model, injury and death were equally severe after 10 and 20 min of ischemia. This is probably due to the fact that significant injury occurs in this model after just 5 min of anoxia. 31Therefore, both 10 and 20 min of ischemia should be considered to produce severe injury, which makes the observed protection all the more striking. In vitro  models differ in many ways from the intact brain; therefore, extrapolation of our results to whole animals is difficult. Among the many ways in which in vitro  models differ from the intact brain is that during simulated ischemia, the P^CO2, [H⁺], and lactic acid levels do not change as they would in vivo  due to diffusion away from the slice. However, this in vitro  model permits comparison of the neuroprotective potency of various anesthetics without some of the problems associated with in vivo  models, such as differences in cerebral blood flow among treatment groups. In addition, it is easier to use pharmacologic manipulations to discern mechanisms of neuroprotection in brain slices than in intact animals, and variables such as brain tissue oxygen tension, temperature, and chemical environment can be precisely controlled. An advantage over cultured neurons is that neurons in hippocampal slices have similar sensitivity to hypoxia as neurons in the intact brain. Five to ten min of in vitro  hypoxia causes cell death in CA1 neurons in hippocampal slices that is manifest within 5 h following the insult. 31 

We have shown that isoflurane and sodium pentothal at clinical concentrations attenuate cell injury and death in an in vitro  model of cerebral ischemia. The mechanism of injury in the model and the mechanism of protection afforded by isoflurane may involve attenuation of glutamate excitotoxicity.