Global Incomplete Cerebral Ischemia Produces Predominantly Cortical Neuronal Injury
Background and Purpose We determined the neuropathologic damage in a canine model of global incomplete ischemia commonly used in a variety of physiological experiments.
Methods We induced 20 minutes of incomplete ischemia in dogs (n=9) by increasing intracranial pressure via intraventricular infusion of artificial cerebrospinal fluid to maintain a cerebral perfusion pressure of 10 mm Hg while keeping body temperature at 38°C during and immediately after ischemia. After a 7-day recovery period, animals were perfusion-fixed for neuropathology. In hematoxylin and eosin preparations, ischemic neuronal injury was assessed, neurons were counted, and percentage of cell damage was determined.
Results No focal neurological deficits or overt seizures were observed during the 7-day recovery period. In superior temporal gyrus, 49±11% and 70±10% damage (mean±SEM) was observed in layer III pyramidal cells in the crown and sulcus, respectively. All neocortical regions examined showed neuronal damage in layers III and/or V. In hippocampus, 59±11% damage of pyramidal neurons occurred in CA1, with dorsal (septal) hippocampus showing more injury than ventral (temporal) portions. The caudate nucleus (head) exhibited 27±7% neuronal injury. In cerebellar cortex (anterior lobule), 70±7% damage of Purkinje cells occurred, but different folia of cerebellum showed varying degrees of injury. Brain stem and thalamus were minimally affected despite reduced blood flow. Inflammatory changes (leukocytic infiltration and neuronal incrustations) were observed, but only when neuronal degeneration was severe. Pancellular necrosis and infarction did not occur.
Conclusions This animal model of ischemia causes reproducible neuronal injury primarily in cortical regions without pancellular necrosis and infarction. Damage to subcortical areas is less severe than to cortical areas, despite comparable reductions in regional cerebral blood flow. Therefore, in the presence of regionally uniform but incomplete cerebral ischemia, neocortical and hippocampal pyramidal neurons and cerebellar cortical Purkinje cells are more likely than subcortical neurons to degenerate; alternatively, pyramidal and Purkinje neurons degenerate before neostriatal neurons in this model. This neuronal degeneration may represent an intrinsic cellular mechanism without major contribution of cytotoxic pathways associated with inflammation.
After cerebral ischemia, varying degrees of brain damage occur that range from selective neuronal death to infarct. The mechanisms of selective neuronal damage are of interest because this form of cellular injury may exhibit some component that is reversible, whereas with infarct, brain parenchyma is irretrievably lost. Thus, it is important to develop and study reproducible models of selective neuronal injury.
Large-animal models of cerebral ischemia offer several advantages in the study of ischemic neuronal injury. For example, easier control of physiological variables is attained, and the larger brain size makes neuropathologic evaluation easier. Furthermore, in studying regional selective vulnerability, it is important to use animal models that closely simulate the situation of global ischemia encountered in humans; thus, these models should include reproducible neuropathologic features of cerebral ischemia in humans. For instance, cerebellar Purkinje cells are highly vulnerable to ischemia, and in humans, selective loss of Purkinje cells often occurs after severe global ischemia.1 Reproducible Purkinje cell death can be obtained from several rodent models of brain ischemia.2 3 However, these models are not widely used. The most popular rodent models of forebrain ischemia spare vulnerable neurons in cerebellum, whereas many large-animal models of global ischemia provide reproducible cerebellar damage.
The dog model of global incomplete ischemia has been used in a variety of physiological studies in past years, but the neuropathology in this model has not been described. This model causes a uniform decrease in cerebral blood flow throughout all supratentorial and infratentorial structures during the ischemic period.4 5 In addition, after prolonged periods of ischemia, there is greater than 80% recovery of high-energy phosphates and good initial recovery of somatosensory evoked potentials.4 5 On the basis of these metabolic and physiological data, we hypothesized that the neuropathologic damage is unlikely to be severe in dogs subjected to global incomplete ischemia and that this model may be suitable for use in long-term outcome studies. The aim of the present study was to determine the distribution and severity of neuropathology in a dog model of global incomplete ischemia and to evaluate whether this model causes selective neuronal damage in vulnerable regions.
Materials and Methods
After placement of a peripheral intravenous catheter, dogs (n=10) were anesthetized with fentanyl (50 μg/kg IV) and pentobarbital (10 mg/kg IV) followed by a continuous pentobarbital infusion (3 mg/kg per hour IV). Pancuronium bromide (0.1 mg/kg IV) was injected for muscle paralysis. The trachea was intubated, and the lungs were mechanically ventilated. A femoral artery was cannulated for blood pressure measurement and blood sampling. The left temporalis muscle over the lambdoidal suture was retracted from the skull. After placement of a small burr hole, a thermistor was inserted between bone and dura to monitor epidural temperature. After placement of a second small burr hole, a Silastic ventricular drain catheter with multiple side ports was inserted into the lateral ventricle for measuring intracranial pressure (ICP) and infusing artificial cerebrospinal fluid (CSF).4 5 6 Arterial line placement, intraventricular catheter placement, and production of ischemia were all performed using strict sterile precautions. The body of the dog was wrapped in a plastic bag and placed on a blanket perfused with recirculating warm water, and heat lamps were used if necessary to maintain epidural temperature at 37°C to 38°C. End-tidal CO2 was monitored, and ventilation was adjusted to maintain Paco2 at 35 to 40 mm Hg. Mean arterial pressure and ICP were continuously recorded. Inspired O2 was approximately 30% to maintain oxyhemoglobin saturation. Supplemental oxygen was administered until approximately 30 minutes after ischemia (when the animal was extubated). Although prolonged postischemic hyperoxia may exacerbate neurological dysfunction, 30 minutes of postischemic hyperoxia does not influence neurological outcome.7
Incomplete ischemia was produced in dogs (n=9) by infusing sterile artificial CSF into the lateral ventricle to produce an ICP 10 mm Hg below the animal’s mean arterial blood pressure. The artificial CSF was prewarmed and maintained at a temperature of 38°C. Three dogs served as sham controls. A moderate pressor response results from this procedure, but ICP can be easily manipulated to help keep cerebral perfusion pressure constant at 10 mm Hg for 20 minutes. Cerebral blood flow during ischemia ranges from 2 to 11 mL/min per 100 g.4 This is sufficient to reduce cerebral O2 uptake, flatten the somatosensory evoked potentials, and reduce phosphocreatine and ATP levels.4 5 To start reperfusion, the CSF pressure reservoir is disconnected and ICP allowed to normalize. Arterial blood pressure and ICP were measured continuously with Statham pressure transducers. Arterial Po2, Pco2, and pH were measured with Radiometer BMS3 electrodes and analyzer. Arterial oxygen content, saturation, and hemoglobin concentration were determined with a Hemoximeter3. Arterial glucose concentration was measured with a Yellow Springs glucose/lactate analyzer. All analyzers were calibrated routinely throughout each experiment as per our previous work.6 After the ischemic episode, all catheters were removed, bleeding was controlled, and skin incisions were sutured before awakening the animal from anesthesia.
In animals surviving ≥24 hours after ischemia, a neurological deficit scoring system for dogs adapted from Bircher and Safar8 was used to document functional neurological recovery. The neurological deficit evaluation was a single score comprising the assessment of five components of neurological function: (1) level of consciousness, (2) respiratory pattern, (3) cranial nerve function, (4) motor and sensory function, and (5) behavior. All of these components are weighted equally. The neurological deficit evaluation provides a single score that can range from a low of 20, which is the best possible, to a high of 62. Neurological scoring was performed daily during the postischemia recovery period by an unbiased observer.
At 7 days after ischemia, dogs were anesthetized with pentobarbital and anticoagulated with heparin, and their brains were perfused with PBS followed by 4% paraformaldehyde. After removal, each brain was divided midsagittally, and the hemispheres were cut systematically into 1-cm-thick coronal slabs from anterior to posterior. From the left hemisphere, the midhippocampus and surrounding temporal neocortex, frontal cortex, caudate (head), and cerebellum were sampled consistently for paraffin histology and hematoxylin and eosin staining. Our preliminary studies confirmed that in this model of ischemia neuronal damage is similar bilaterally. Five areas were evaluated quantitatively for ischemic injury: the hippocampus (CA1), the crown of the superior temporal gyrus, the sulcus of the superior temporal gyrus, the caudate nucleus, and the anterior cerebellar lobule at the midline. We selected these areas because in other models of ischemia these regions are damaged. In addition, these regions can be sampled and examined consistently. Purkinje cells in the cerebellum were counted in four fields at ×200 magnification in the rostral cerebellar folia (anterior/rostral vermis) against the pons and above the fourth ventricle. The left hippocampal CA1 region was counted at the midhippocampal level in six fields at ×600. Pyramidal neurons in layer III of the superior temporal gyrus were counted at ×600 in six fields at the crown (moving toward the sulcus) and six fields at the sulcus. The caudate was examined in five fields at ×1000. In sections stained with hematoxylin and eosin, the number of neurons were counted and the percentage of neurons with ischemic damage was determined using the criteria of microvacuolar change within the cell body, peripheralization of chromatin, perinuclear eosinophilia, or perikaryal shrinkage with eosinophilic cytoplasm. Ischemic neuronal damage was determined by a blinded investigator (L.M.) who evaluated the five regions of interest separately. The percentage of neuronal damage in each region was averaged for each animal, and a group mean was calculated. The number of neurons per square millimeter was calculated for each region in both the ischemic and nonischemic animals. Group averages for the neurological deficit evaluation scores were obtained for each postischemic day.
Nine dogs underwent 20 minutes of global incomplete ischemia. Arterial blood gas values immediately before ischemia were pH, 7.42±0.02; Pco2, 31±2; and Po2, 243±57. Immediately before ischemia, blood glucose was 3.98±0.16 mmol/L and hemoglobin was 14.0±0.6 g/dL. Epidural temperature during ischemia was 37.8±0.3°C.
All animals survived for 7 days after ischemia. No focal neurological deficits or overt seizures were observed during the 7-day recovery period. The neurological score increased to 25±4 on postischemic day 1. By postischemic day 2 the neurological deficit score was similar to the preischemic baseline score.
Selective neuronal injury was observed throughout cerebral cortex and was present in frontal, parietal, and temporal neocortices (Fig 1A⇓ and 1B⇓). Neuronal damage in neocortex had a laminar distribution, with deeper layers affected more than superficial layers. Damage occurred primarily in layers III and V and was sporadic in layers II and VI. The severity of postischemic neuropathology was determined as a function of changes in neuronal density relative to control neuronal densities and as a function of remaining neurons that showed ischemic cell injury. At 7 days after ischemia, there was approximately a 45% reduction in neuronal counts in layer III of superior temporal cortex (Fig 2⇓). Of the remaining layer III pyramidal cells, 49±11% and 70±10% neurons displayed evidence of ischemic change in the crown and sulcus, respectively. The severity of neuronal injury varied not only in relation to neocortical lamination but also position within the cortical mantle. For example, there was a gradient of increasing neuronal ischemic change on moving from the crown to deep within the sulcus (Fig 1A⇓). In several animals there was greatly decreased neuronal density, and in some animals (n=2) Betz cells in the motor cortex were severely damaged. Inflammatory changes throughout all cell layers were observed in three dogs with many extravascular monocytes and lymphocytic incrustations, but infarction was not detected in any animal.
The neuropathology within the hippocampal formation was consistent with cerebral ischemia (Fig 1C⇑). The CA1 subfield exhibited approximately a 53% decline in neuronal density at 7 days after the ischemic insult (Fig 2⇑). Of the remaining pyramidal cells, 59±11% neurons displayed evidence of ischemic change. Cell counts in both hemispheres showed that hippocampal damage was equal bilaterally (data not shown). Vacuolization was observed in the pyramidal cells of CA1, and virtually all damaged neurons showed a shrunken, eosinophilic perikaryon and nuclear pyknosis but no morphological evidence of karyolysis at 7 days after ischemia. Within CA1, the damage increased on moving from CA2 toward the subiculum. Injury in CA1 of ventral (temporal) hippocampus was different from injury in CA1 of dorsal (septal) hippocampus, with dorsal hippocampus showing more injury (Fig 1C⇑ and 1D⇑). Neuronal density was severely decreased after ischemia in CA4 (data not shown), and CA2 showed focal zones of damage to pyramidal neurons. There was variable damage in CA3. In dentate gyrus, granule cells were highly vulnerable to ischemia. Damage in this area was primarily confined to the inferior blade (Fig 1F⇑); there was less complete damage to the superior blade. In three of nine animals, perivascular cuffing was noted throughout the hippocampus with mild inflammatory changes and extravascular lymphocytes.
In cerebellum, there was a slight decrease (18%) in Purkinje cell count 7 days after the ischemic insult (Fig 2⇑). Evidence of ischemic change was displayed in 70±7% of the Purkinje cells (Fig 1G⇑). The deep cerebellar nuclei showed some evidence of nuclear pyknosis, but the cell bodies were for the most part free of eosinophilic changes and vacuolization. Different folia of cerebellum showed varying degrees of injury. Brain stem was minimally affected despite reduced blood flow. Little or no evidence of ischemic cell change was observed in midbrain, pontine, and medullary nuclei.
The basal ganglia and thalamus exhibited less damage than neocortex, hippocampus, and cerebellum. In the caudate nucleus, there was 27±7% damage to striatal neurons (Fig 2⇑). Both large (presumably cholinergic interneurons) and medium-sized neurons showed vulnerability. The globus pallidus and substantia nigra manifested mild neuronal changes. Thalamus was unremarkable with the exception of evidence for ischemic cell change within the lateral geniculate nucleus.
We evaluated the neuropathology in a large-animal model of incomplete cerebral ischemia. This model provides reasonable neurological recovery allowing for long-term survival experiments. In our model of global incomplete ischemia, infarction is either not present or is incipient at 7 days of recovery, and selective neuronal damage in the regions examined occurs with high reproducibility. The ischemic neuronal damage is present in brain regions previously found to be selectively vulnerable to cerebral ischemia. However, in this model of global incomplete cerebral ischemia, we report that the brain damage is predominantly cortical and that less severe neuronal damage is produced in subcortical regions. The regional distribution of neuronal damage that predominates in this model (ie, neocortical and hippocampal but not subcortical) is consistent with the absence of focal neurological deficits during the 7-day recovery period. Therefore, in the presence of regionally uniform but incomplete cerebral ischemia, neocortical and hippocampal pyramidal neurons and cerebellar Purkinje cells are more likely than subcortical neurons to degenerate; alternatively, pyramidal and Purkinje neurons degenerate before neostriatal neurons in this model. This neuronal degeneration may represent an intrinsic cellular mechanism without major contribution of cytotoxic pathways associated with inflammation. On the basis of previous metabolic and physiological data derived from this model indicating that the probability of recovery of cerebral function is relatively high,4 5 we hypothesized that the neuropathologic damage would not be severe in dogs subjected to global incomplete ischemia. However, we did find relatively severe neuronal damage in this model. The differences between the metabolic/physiological, neurological, and neuropathologic outcomes may be reconciled by the finding that, at 7 days after ischemia, the regional and cellular distribution of damage was highly selective without the occurrence of global necrosis, regional infarction, and pancellular injury.
Selective neuronal damage occurs in many regions after ischemia. Previous investigators have documented that certain neuronal populations are more susceptible to ischemic cell damage and that the order of regional vulnerability may be ranked.9 In our ischemic model at 7 days of recovery, the damage is uniform throughout cortical regions previously noted to be selectively vulnerable, but it is less severe in subcortical regions such as striatum. Because regional blood flow decreases are comparable in all areas of the brain in our model,4 5 the neuronal damage obtained thus reflects the intrinsic cell type–specific susceptibilities to ischemic insults. Pancellular necrosis does not occur in our dog model at 7 days after ischemia, a finding that is consistent with the absence of focal neurological deficits and seizures. In addition, the lack of thalamic neuropathology and modest damage to basal ganglia is consistent with the absence of seizures. However, this study only examines one time point, and we speculate that changes in the neurological outcome may be biphasic. If neurons undergo progressive delayed neuronal death and possibly transynaptic neurodegeneration with longer survival times, neurological deficits may be manifested at later recovery times. Alternatively, because the damaged regions were predominantly cortical but not subcortical, our neurological examination may not have sufficient sensitivity for detecting subtle abnormalities in neocortical, hippocampal, and cerebellar function. Specific behavioral tests for learning and memory, rather than neurological tests for brain stem and motor function, may be more useful for this model. The current neurological deficit score has been used to successfully assess differences in outcome after cardiac arrest.10 However, damage to both hippocampus and cerebellum may occur without serious impairment of the neurological deficit score if the neocortex is spared. Neocortical damage in our animals occurred primarily in layers III and V with sporadic damage in other layers. Thus, our neurological examination results are consistent with previous studies.
The rodent and gerbil models of incomplete ischemia are the primary ischemic models with which to compare our results. Our model is more invasive than those in rodents. However, our histological results indicate that neuropathology due to the placement of an intraventricular catheter is minimal. In rat forebrain ischemia, blood flow differences may occur during ischemia, making cerebellar damage poorly reproducible9 and causing some laterality of damage.9 11 In gerbils, there may be heterogeneity in the hemispheric response to unilateral carotid artery occlusion,12 and bilateral carotid artery occlusion does not always produce complete ischemia. Both rat and gerbil models provide well-defined pathology, especially in hippocampus, neocortex, and caudate.13 14 15 16 In the present model of incomplete ischemia, there is uniform damage throughout the neocortex without laterality. However, caudate damage with the present model was less severe than with the rat and gerbil models. Neurological recovery is good with all the models described, with focal neurological deficits infrequently reported. However, seizure frequency is lower with the present model, especially compared with the gerbil studies. Although the two- and four-vessel models of forebrain ischemia have been extensively used to investigate the pathophysiology of ischemic brain injury, other rodent models of global ischemia using CSF pressure elevations have also been infrequently studied.17 18 19 In one study, the early neuropathologic consequences of a 60-minute period of incomplete ischemia were assessed, and significant cortical pathology was reported.17
We have found additional neuropathologic changes occurring with global incomplete ischemia in dogs that are not frequently reported in other animal models of ischemia. However, we currently cannot empirically explain the mechanisms for these neuropathologic observations and thus resort to brief speculation. For example, within the neocortex, neuronal damage is more severe in the cortical mantle of sulci than in gyral crests. This topographical distribution of injury is reminiscent of hypoxic-ischemic brain injury in human neocortex1 and may be caused by a variety of mechanisms, including regional differences in blood flow, connectivity, or edema formation. These differences in neuropathology in the dog model will be useful for ascertaining mechanisms underlying the influence of neuronal position on ischemic injury in neocortex. This feature cannot be explored in the lissencephalic rodent/gerbil neocortex. Furthermore, granule cells of the dentate gyrus are reproducibly damaged in this model of ischemia, and the superior and inferior blades of the granule cell layer of the dentate gyrus are differentially vulnerable. Regional cerebral blood flow and metabolism as well as connectivity differences may explain these observations. Alternatively, this difference in susceptibility may represent model artifact, with the inferior blade of the dentate being exposed to greater compressional forces during ischemia due to proximity to the ventricles. Delayed neuronal death in the dorsal (septal) hippocampus was greater than in the ventral (temporal) portions. This dorsal-ventral difference in hippocampal vulnerability may be related to differences in connectivity and intrinsic metabolism.
In summary, we present a neuropathologic study of a dog model of global incomplete brain ischemia in which the physiology has been well defined in previous studies. In this model, frank infarct does not occur, and regionally selective neuronal damage occurs with high reproducibility and reflects patterns of injury that are comparable to those observed in humans. This model provides reasonable neurological recovery allowing for long-term survival experiments. This animal model will be useful for identifying early mechanisms and types of ischemic neuronal death without the confounding influences of postischemic pancellular necrosis.
This study was supported in part by grants from the US Public Health Service National Institutes of Health (NS 01380, NS 20020, and AG 07914). The authors thank Rosie Cousins, Dawn Spicer, and Ann Price for their tireless technical assistance and Lee Palmer for her help in preparing this manuscript.
Reprint requests to Frederick E. Sieber, MD, Department of Anesthesiology and Critical Care Medicine, The Johns Hopkins Hospital, 600 N Wolfe St, Meyer 8-134, Baltimore, MD 21287-7834.
- Received May 1, 1995.
- Revision received July 25, 1995.
- Accepted July 28, 1995.
- Copyright © 1995 by American Heart Association
Graham DI. Hypoxia and vascular disorders. In: Adams JH, Cuchen LW, eds. Greenfields Neurophysiology. London, UK: Arnolds; 1992:153-268.
Mossakowski MJ, Zelman IB. Dynamics of late neuronal changes in global cerebral ischemia. In: Ito U, Kirino T, Kuroiwa T, Klatzo I, eds. Maturation, Phenomenon in Cerebral Ischemia. Tokyo, Japan: Springer-Verlag; 1990:65-74.
Kirino T, Nakagomi T, Tamura A. Reversibility of damage to rat cerebellar Purkinje cells following ischemia. In: Ito U, Kirino T, Kuroiwa T, Klatzo I. Maturation, Phenomenon in Cerebral Ischemia. Tokyo, Japan: Springer-Verlag; 1990:49-56.
Sieber FE, Koehler RC, Brown PR, Eleff SM, Traystman RJ. Diabetic chronic hyperglycemia and cerebral pH recovery following global ischemia in dogs. Stroke. 1994;25:1449-1455.
Palmon SC, Sieber FE, Brown PR, Koehler RC, Eleff SM, Traystman RJ. Poor hemodynamic and metabolic recovery after global incomplete cerebral ischemia associated with short term diabetes in dogs. J Cereb Blood Flow Metab. 1995,15:673-680.
Hurn PD, Koehler RC, Norris SE, Blizzard KK, Traystman RJ. Dependence of cerebral energy phosphate and evoked potential recovery on end-ischemic pH. Am J Physiol. 1991;260:H532-H541.
Mickel HS, Vaishnav YN, Kempski O, von Lubitz D, Weiss JF, Feuerstein G. Breathing 100% oxygen after global brain ischemia in mongolian gerbils results in increased lipid peroxidation and increased mortality. Stroke. 1987;18:426-430.
Ginsberg MD, Busto R. Rodent models of cerebral ischemia. Stroke. 1989;20:1627-1642.
Pulsinelli WA, Waldman S, Rawlinson D, Plum F. Moderate hyperglycemia augments ischemic brain damage: a neuropathologic study in the rat. Neurology. 1982;32:1239-1246.