Hypoxia, hyperoxia, ischemia, and brain necrosis

Levine series.

The rats in these series (n = 135) were divided into hypoxia only, hypoxia with ischemia, and ischemia only groups.

To induce hypoxia only, after ∼15 minutes, the inspired oxygen was lowered to give an arterial PaO₂ of 25 mm Hg. Tidal volume was adjusted when necessary throughout the experiment to maintain PaCO₂ in the normal range. To counteract systemic acidosis, a peripheral IV infusion of sodium bicarbonate (1 M, 0.84 mL/hour) was given during hypoxia to maintain blood pH > 6.9. Hypoxia led immediately to a marked reduction in BP. Because of the known sensitivity of ischemic damage to small differences in BP,16 the mean arterial BP (MABP) was fixed at 30 mm Hg for 15 minutes (see figure 1) by adjusting the fractional inspiratory oxygen concentration (FiO₂). Hypoxia was stopped by reoxygenation with 100% O₂, and rats were extubated when spontaneous breathing was strong and constant. Arterial blood was sampled before, during, and after hypoxia.

To study hypoxia and ischemia simultaneously, ligation of the right common carotid artery was combined 1 day later with reduction in inspired oxygen content to achieve the desired PaO₂, and simultaneous exsanguination as needed to clamp MABP to the desired level without the use of pharmacologic agents. Two series (table 1) were done, one varying MABP at a constant PaO₂ and the second varying PaO₂ at a clamped MABP. In one series of five groups, hypoxia was induced to a PaO₂ of 25 mm Hg by lowering the FiO₂, and when this induced the MABP to spontaneously fall to 30, 40, 50, 60, or 70 mm Hg, the FiO₂ was adjusted to maintain the desired MABP for 15 minutes. In a separate series of four groups, hypoxia was induced to a PaO₂ of 25, 35, or 45 mm Hg, or > 100 mm Hg (normoxia), by adjusting the FiO₂. The PaO₂ = 25, MABP = 30 mm Hg group was a member of both series, as shown in table 1. MABP was kept at 30 mm Hg for 15 minutes by regulating the FiO₂ in the PaO₂ 25, 35, and 45 mm Hg groups, or by rapid withdrawal or reinfusion of blood via the central venous catheter in the PaO₂ > 100 mm Hg group (8.4 ± 0.3 mL blood withdrawn). In the PaO₂ 25, 35, and 45 mm Hg groups, blood withdrawal was also performed as in the PaO₂ > 100 mm Hg group (6.0 ± 0.5 mL blood withdrawn), if MABP did not fall to 30 mm Hg within 45 minutes during hypoxia. After this period of time, the shed blood was reinfused, and rats were reoxygenated with 100% oxygen. The rats were extubated and returned to their cages after recovery from anesthesia.

In all groups, the animals were allowed access to food and water in the postoperative period. After 1 week, they were anesthetized with 4% halothane in 40% O₂ and 60% N₂O, and perfused transcardially with 4% phosphate-buffered formaldehyde solution. The brains were removed on the following day and cut into 5-mm slices, processed in graded ethanols and xylol, and embedded in paraffin.

Each brain was serially sectioned (8 μm thick) at 500-μm intervals with a sliding microtome (model Hn 40, Reichert-Jung), from a rostral limit of bregma +2.2 mm to bregma −6.8 mm caudally. Animals that died spontaneously (see Results) before the 1-week survival period were excluded from analysis. The sections were stained with hematoxylin and eosin, and examined under a light microscope.

To quantitate brain damage, the infarcted area on sections of each of 18 coronal levels was traced using microcomputer-based digital image analysis (Jandel video analysis software JAVA, version 1.41, Jandel Scientific Inc., San Rafael, CA). Because significant atrophy was also present in the ischemic hemisphere, total tissue loss was calculated by summing the areas of cortical necrosis, striatal necrosis, and atrophy (difference between the hemispheres). Volumes of hemispheric tissue loss were determined by integrating the areas of damage in the anteroposterior dimension, from the rostral to the caudal section. Interindividual variation in brain size was controlled for by dividing the volume of tissue loss by the contralateral hemispheric volume, expressing damage as a percent16 rather than as raw data in mm³. The hippocampus was sampled at six coronal levels encompassing the entire septotemporal axis, and the percentage of necrotic neurons in CA1 was calculated. All data were expressed as mean ± standard error (SEM). Physiologic variables (tables 1 and 2⇓) were analyzed using one-way analysis of variance (ANOVA) followed by Scheffé test for individual group comparisons. The percentages of tissue volume loss and CA1 necrotic neurons were analyzed using the Kruskal-Wallis test followed by the Mann-Whitney U test, and by arcsine transformation followed by ANOVA.

Autoradiographic measurement of CBF.

Nineteen male Wistar rats were divided into four groups: normal control (n = 4), hypoxia only (n = 5), ischemia only (n = 6), and hypoxia with ischemia (n = 4). Hypoxia, ischemia, and the combination of the two were produced by methods identical to those described above for the histologic series, except that the PaO₂ was regulated to 35 mm Hg in the hypoxia and hypoxia with ischemia groups. By the methods described above, MABP was controlled to ≈30 mm Hg, except in controls. Local CBF (lCBF) was measured using 4-iodo-¹⁴C-antipyrine (¹⁴C-IAP) as described by Sakurada et al.17 Briefly, femoral venous and arterial lines were placed to allow injection and sampling, respectively, of radioisotope for CBF measurement. Fifty μCi of ¹⁴C-IAP (50 mCi/mmol, Amersham Laboratories, Buckinghamshire, England) in 0.5 mL of 0.9% saline was infused via the femoral vein over a period of 30 seconds by a Harvard 944 infusion pump after at least 10 minutes of stable MABP at 30 mm Hg. Arterial blood samples were taken every 5 seconds to assess ¹⁴C activity. Immediately after a circulation time of 30 seconds in the normoxia and ischemia groups, or 60 seconds in the hypoxia and hypoxia with ischemia groups (due to their slower circulation), the rats were decapitated. Brains were removed rapidly and frozen in isopentane cooled to −60 °C with dry ice, and stored in a freezer at −70 °C until prepared for autoradiography. At that time, brains were cut into 20-μm sections in a cryostat (Frigocut 2800, Leica Instruments, Germany) and were exposed to x-ray film (Hyperfilm βmax, Amersham) for 7 days, together with ¹⁴C-methyl methacrylate standard (range 31 to 874 nCi/g). Plasma ¹⁴C radioactivity was measured with a scintillation counter (Beckman LS 6800 series, liquid scintillation system, Beckman Instruments, Irvine, CA). The density of the autoradiograms was measured using an image analysis system (JAVA version 1.41, Jandel Scientific Inc.) via computer-based densitometry. LCBF was calculated by using a tissue-blood partition coefficient of 0.8 and the equation of Sakurada et al.17

Oxygenation in MCA ischemia.

A total of 30 male Wistar rats were equally divided into normoxia (40% O₂ in ventilation gas), hypoxia (PaO₂ controlled to 45 mm Hg by varying the FiO₂), and hyperoxia (100% O₂ in inspired gas) groups. Core body temperature and ipsilateral temporalis muscle temperature were monitored and maintained at 37.5 °C using a thermistor-regulated servo-controlled heating blanket and overhead lamp during MCA occlusion.

MCA occlusion was done as previously reported.16 The 3-0 nylon suture was advanced until a feeling of faint resistance was encountered. PaO₂ was adjusted to the desired levels in each group. MABP was also controlled to 60 mm Hg. This was done by rapid withdrawal or reinfusion of blood via the central venous catheter in the hyperoxia and normoxia groups (5.5 ± 0.4 and 5.1 ± 0.5 mL, respectively), and by adjusting the FiO₂ in the hypoxia group. Suxamethonium chloride was injected IV to immobilize animals. After 80 minutes occlusion, the nylon suture was removed and the external carotid artery was tied permanently. MABP and PaO₂ were restored to preischemic levels by reinfusion of blood and adjusting the FiO₂, respectively. Halothane was discontinued and the animals were allowed to awaken.

One week after MCA occlusion, rats were perfused-fixed and the brain was removed, paraffinized, and serially sectioned (8 μm thick) at 500-μm intervals with a sliding microtome. The sections were stained with hematoxylin-eosin, and polygons for necrosis in the neocortex and in subcortical structures (white matter, deep gray) were traced separately, as were the areas of the ipsilateral and contralateral hemispheres. Brain tissue loss was then quantified as mm³ by integrating necrosis and atrophy in the third, orthogonal axis. Volumes were finally normalized to the opposite hemisphere to control for variation in absolute brain size, being expressed as % of the hemisphere.

Physiologically monitored/controlled Levine preparation.

Severe hypoxia caused immediate cardiac hypotension (see figure 1), and preliminary experiments revealed that it was very difficult for rats to survive hypoxia at PaO₂ levels <20 mm Hg owing to immediate cardiac death. Forty-four rats died within 2 days of the insult; 10 of these were in the hypoxia only group, where mortality was 43%. The rest were from hypoxia with ischemia groups, and died likely due to delayed cardiogenic shock8 (5 of these 44 died following crescendo seizures). In the variable ischemia groups, at a fixed PaO₂ of 25 mm Hg, the mortality was 0%, 9%, 47%, 36%, and 53% at MABP 70, 60, 50, 40, and 30 mm Hg, respectively. In the groups with a fixed MABP of 30 mm Hg, the mortality was 29%, 18%, and 17% at PaO₂ > 100, 45, and 35 mm Hg, respectively.

Physiologic variables are shown in table 1. MABP was well controlled to the desired values by adjusting the FiO₂. The PaO₂ levels during control of MABP and the duration of hypoxia showed no differences among groups.

Hypoxia of PaO₂ = 25 mm Hg, despite the concurrent BP of 30 mm Hg, caused some slowing of EEG frequencies with increased amplitude (see figure 1), but failed to cause necrosis in any of 13 animals (figure 2). However, necrosis in the cortex, striatum, and hippocampus was seen in all hypoxia plus ischemia groups, and of 12 rats with ischemia only, 4 had brain damage (see figure 2). No necrotic neurons were found in any nonischemic hemisphere, even those contralateral to hemispheres severely damaged by hypotensive ischemia and severe hypoxia (see figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Representative hematoxylin-eosin–stained sections from caudate nucleus (left) and hippocampus (right). There is no neuronal necrosis in either the caudate (A) or hippocampus (B) with hypoxia only, whereas a small amount of infarction of the caudate (C) and necrosis in CA1 (D) is seen with ischemia only. Caudate pan-necrosis, indicated by acidophilia of the neuropil and loss of the striatal bundles of white matter (E), is seen with hypoxia plus ischemia, which also causes necrosis of all CA1 pyramidal neurons, CA3 pyramidal neurons, and the dentate gyrus (F). Many inflammatory cells infiltrate the hippocampus, but contralaterally in the same animal, no neuronal necrosis is seen in either striatum (G) or hippocampus (H). Bars = 250 μm.

The degree of ischemic brain damage at a clamped level of hypoxia (PaO₂ = 25 mm Hg) was strongly influenced by BP, increasing in a linear fashion as MABP was lowered in steps through 70, 60, 50, 40, and 30 mm Hg, respectively (figure 3). Although no hemispheric tissue loss (cortical/subcortical pan-necrosis or atrophy) was seen at 70 mm Hg, hippocampal CA1 necrosis was already seen at this BP, confirming that neurons of the hippocampus are the most sensitive within the cerebral hemisphere.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Quantitated necrosis at graded levels of mean arterial blood pressure (MABP) with fixed PaO₂ = 25 mm Hg in the cerebral hemisphere (A) and the CA1 pyramidal cells (B). At a fixed MABP = 30 mm Hg, quantitated necrosis is shown in the hemisphere (C) and the CA1 pyramidal neurons (D) at graded levels of PaO₂. Total brain damage increased with reduction of MABP (A) and PaO₂ (B), the effect of hypotension being more clearly graded than hypoxia. Necrotic volumes were traceable because of the characteristically sharp demarcation of ischemic necrosis (see figure 6). Hemispheric volumes were calculated by integrating the coronal planar areas in the anteroposterior dimension. Atrophy represents nontraceable tissue lost, obtained by subtracting the ipsilateral from the contralateral hemisphere. The volume of tissue loss is referenced to the contralateral hemispheric volume (see text) to give the percentage of normal hemisphere damaged. *p < 0.05 and **p < 0.005 versus MABP 70 mm Hg group, ***p = 0.0001 versus PaO₂ > 100 mm Hg group.

The degree of hypoxia at a clamped BP influenced brain damage less clearly. At a fixed MABP of 30 mm Hg, the percentage of hemispheric tissue loss generally increased, but not always in a clear linear way, as the PaO₂ was lowered from >100 to 45, 35, and 25 mm Hg (see figure 3). Common carotid artery ligation caused hemispheric pan-necrosis and CA1 neuronal necrosis even at PaO₂ >100 mm Hg, and both were augmented by hypoxia, the results achieving significance at PaO₂ = 25 mm Hg (see figure 3). Cortical necrosis proved most sensitive to PaO₂, being eliminated entirely (see figure 3) at arterial oxygen tensions above 100 mm Hg.

Autoradiographic measurement of CBF.

Carotid ligation led to a drop in ipsilateral CBF to the range of 15 to 35 mL/100 g/minute in most brain structures (for supplementary data, please visit our website at www.neurology.org and access the data through the Table of Contents page of the issue). Contralateral to the ligation, at a MABP of 30 mm Hg, CBF values were similar to those in the hypoxia only group at this same BP. Hypoxic hypotension to a MABP of 30 mm Hg led to CBF values lower than normoxic animals at a MABP of 110 mm Hg. With carotid ligation, the addition of hypoxia to ischemia led to a further fall in local CBF in all brain regions. Global CBF in the hemisphere showed no evidence of hypoxic vasodilatation, but was reduced by hypoxia superimposed on ischemia, and to a lesser degree by the ischemic procedure alone (figure 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Global hemispheric cerebral blood flow (CBF) values calculated as arithmetic means of all structures measured. Hypoxia with hypotension to 30 mm Hg shows reduced blood flow from controls. Hemispheric blood flow ipsilateral to carotid occlusion is depressed, with no increase in CBF with superadded hypoxia, at the same blood pressure. Mean ± SEM. ***p = 0.0001, **p = 0.01, *p = 0.05 versus control.

Transient MCA occlusion.

Physiologic measures (table 3), apart from arterial oxygen tensions, were the same between groups. Hypoxia, as predicted from the Levine experiments, worsened ischemic brain damage after MCA occlusion to 24.31 ± 4.67% from 16.57 ± 6.98% of the hemisphere with normoxia (figure 5). Normobaric hyperoxia, on the other hand, protected the brain from ischemic injury, with only 7.50 ± 1.86% hemispheric damage (see figure 5). Hyperoxic protection in the cerebral cortex was remarkable, cortical necrosis being almost completely abolished (figure 6).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Quantitated necrosis in cortex and striatum, as well as atrophy and total brain damage (sum of necrosis and atrophy), normalized to the contralateral hemisphere. Deep gray matter receives marginal benefit from high oxygen tensions, but cortical necrosis is all but eliminated (0.34 ± 0.22% with hyperoxia, 7.97 ± 4.63% with normoxia, and 12.82 ± 3.08% with hypoxia). **p < 0.01, ANOVA on arcsine-transformed percentage data.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Representative hematoxylin-eosin–stained sections from middle cerebral artery (MCA) occlusion experiments show how arterial oxygen levels influence MCA infarction in the cerebral cortex. Normoxia (A) (PaO₂ = 120.5 ± 4.1 mm Hg) produces a smaller infarct than hypoxia (B) (PaO₂ = 46.5 ± 1.4 mm Hg), which produces a large infarct with a richer neutrophilic infiltration. Hyperoxia (C) (PaO₂ = 213.9 ± 5.8 mm Hg) converts the infarct into selective neuronal necrosis (inset), visible as microglial rod cells (dendritic phagocytosis) and acidophilic neuronal perikarya. There is no traceable pan-necrosis. The characteristic sharp border (arrows) of cerebral infarction, which allows quantification, is seen in the normoxic and hypoxic rats. Scale bar = 500 μm (inset, 150 μm).