Brief Exposure to Hyperoxia Depletes the Glial Progenitor Pool and Impairs Functional Recovery after Hypoxic-Ischemic Brain Injury

Abstract

Patterns of hypoxic-ischemic brain injury in infants and children suggest vulnerability in regions of white matter development, and injured patients develop defects in myelination resulting in cerebral palsy and motor deficits. Reperfusion exacerbates the oxidative stress that occurs after such injuries and may impair recovery. Resuscitation after hypoxic-ischemic injury is routinely performed using 100% oxygen, and this practice may increase the oxidative stress that occurs during reperfusion and further damage an already compromised brain. We show that brief exposure (30 mins) to 100% oxygen during reperfusion worsens the histologic injury in young mice after unilateral brain hypoxia—ischemia, causes an accumulation of the oxidative metabolite nitrotyrosine, and depletes preoligodendrocyte glial progenitors present in the cortex. This damage can be reversed with administration of the antioxidant ebselen, a glutathione peroxidase mimetic. Moreover, mice recovered in 100% oxygen have a more disrupted pattern of myelination and develop a static motor deficit that mimics cerebral palsy and manifests itself by significantly worse performance on wire hang and rotorod motor testing. We conclude that exposure to 100% oxygen during reperfusion after hypoxic-ischemic brain injury increases secondary neural injury, depletes developing glial progenitors, interferes with myelination, and ultimately impairs functional recovery.

Introduction

Hypoxic-ischemic brain injury is one of the most common causes of death and long-term neurologic morbidity in both the neonatal and pediatric populations (Volpe, 2001). Resuscitation during reperfusion is important in determining the magnitude of additional injury that occurs after the initial insult. Secondary injury after ischemic stress is at least partially mediated by the presence of free radicals, which is exacerbated in the presence of high oxygen levels (Halliwell, 1992). Despite this, it is standard practice in neonates, children, and adults to resuscitate and recover patients after hypoxic-ischemic injury with 100% oxygen (Kattwinkel et al, 1999).

There are conflicting data regarding whether resuscitation with 100% oxygen after hypoxic injury is harmful in animal models (Munkeby et al, 2004; Zwemer et al, 1995). The reasons for this include the heterogeneity of models tested and an inability to establish a link between the type of resuscitation used and the functional neurologic outcome that goes beyond biochemical measurements of potentially harmful intermediates. Randomized controlled trials in human subjects are sparse, although evidence suggests improved outcomes in neonates resuscitated with room air after asphyxial birth depression (Saugstad et al, 1998). A recent meta-analysis concluded that room air resuscitation resulted in a lower mortality rate but recommended that additional studies are needed on long-term developmental outcome (Rabi et al, 2007).

Periventricular leukomalacia is the most commonly observed central nervous system abnormality in premature infants after hypoxic-ischemic injury, and although the mechanisms underlying this are still under investigation, damage to developing white matter is associated with long-term neurologic disability (Volpe, 2001). One potential mechanism is the vulnerability of myelinating progenitors to injury from oxidative stress. In vivo, cerebral hypoxia and ischemia lead to the production of excessive reactive oxygen and nitrogen species, resulting in oxidative damage to immature oligodendrocytes (Haynes et al, 2003). In vitro, immature oligodendrocytes exhibit selective vulnerability to depletion of antioxidants and exposure to exogenous free radicals (Back et al, 2002). Finally, long-term exposure to hyperoxia impairs brain development and function, in part, mediated by nitric oxide and other nitrating agents (Sirinyan et al, 2006).

Brain development is far from complete at the time of birth, as greater than two-thirds of human brain growth occurs postnatally (Dobbing and Sands, 1973). Brain growth after birth is almost exclusively glial in nature in both humans and rodents, and completion of white matter development and myelination in humans is ongoing in the first few years of life. The genesis of myelinating oligodendrocyte progenitor cells peaks around 2 weeks of age in rodents, whereas the majority of actual myelination occurs between 3 and 6 weeks (Sauvageot and Stiles, 2002). Therefore, central nervous system insults that occur at a time when there are abundant immature oligodendrocytes may be amplified because those cells are particularly sensitive to the effects of oxidative stress after hypoxic-ischemic insults.

The primary purpose of this study was to determine whether recovery in 100% oxygen after hypoxic-ischemic brain injury would exacerbate neuronal injury and impair functional recovery in young mice. Given what is currently known about the development of myelination and oligodendro-genesis in the rodent brain, we chose to subject young (postnatal day 14) mice to the Rice—Vannucci model of hypoxic-ischemic injury followed by recovery in either room air or 100% oxygen for various time points from 15 mins to 4 h (Rice et al, 1981). By using a transgenic mouse model that marks neural and glial progenitor populations with green fluorescent protein (GFP), we were able to distinguish what cell types were most vulnerable to hyperoxic recovery-mediated injury (Yu et al, 2005). We observe that a relatively brief exposure to hyperoxia after hypoxic-ischemic injury results in increased histologic injury, generation of oxidative damage, depletion of cortical oligodendrocyte progenitors, and functional motor deficits.

Materials and methods

Mice

All protocols involving the use of animals were approved by the Institutional Animal Care and Research Advisory Committee (IACRAC) at University of Texas Southwestern Medical Center at Dallas. Transgenic mice expressing GFP under control of the nestin promoter and second intron (nestin-GFP) previously generated in our lab were back-crossed into the CD1 background with wild-type CD1 mice (Charles River Laboratories, Wilmington, MA, USA) for eight generations. Offsprings were genotyped according to a previously published protocol, and nestin-GFP transgenic animals were used for all experiments (Yu et al, 2005). Pups were housed under a 12:12-h light—dark cycle, with food and water available throughout the study. A total of 242 mice were used for all of the experiments.

Hypoxic-Ischemic Injury and Recovery

Postnatal day 14 littermates were exposed to hypoxic-ischemic injury for 30 mins with 8% oxygen as an adaptation of the Rice—Vannucci model and randomized to either recovery in room air or 100% oxygen (Rice et al, 1981). Pups were anesthetized for < 10 mins with a mixture of isoflurane (4% for induction and 1.5% for maintenance), 30% oxygen, and balance nitrogen. Under sterile technique, a midline neck incision was made, and the right common carotid artery was exposed, isolated from the nerve and vein, and permanently ligated with 5–0 surgical silk. Temperature was maintained with a warming blanket at 36.5°C to 37.5°C. The incision was closed with a surgical wound clip and the pups were recovered for 1 h in their cage. The small percentage of mice suffering a stroke after ligation alone (about 5%) were euthanized and not included in the study. Animals were then placed in a constructed acrylic glass chamber with separate partitions on a warming blanket maintained at 37°C and exposed to premixed gas of 8% oxygen (balance nitrogen) at 2 L/min for 30 mins. Mice were then randomized to either return to their cage to recover in room air or to remain in the chamber to recover in 100% oxygen for defined time periods of 15 mins (n = 20), 30 mins (n = 13), 1 h (n = 15), 2 h (n = 10), or 4 h (n = 10), after which they were returned to the cage. Oxygen concentrations were continuously monitored with an indwelling oxygen sensor (Teledyne Inc., Thousand Oaks, CA, USA). Sham-operated animals underwent the same operative procedure except that the carotid artery was not isolated and ligated, and they were not exposed to hypoxia. All mice were killed 72 h after injury for histologic scoring.

Ebselen Administration

Mice underwent carotid ligation and exposure to 8% oxygen for 30 mins as above and then randomized to receive intraperitoneal injection of ebselen (2-phenyl-1,2-benzoisoselenazole-3(2H)-one; Cayman, Ann Arbor, MI, USA), a glutathione mimetic and scavenger for peroxynitrite, (20 mg/kg) (n = 30) or phosphate-buffered saline (PBS) (n = 29) immediately after the hypoxic period (Briviba et al, 1996; Takasago et al, 1997). A stock solution of ebselen was dissolved in DMSO (dimethyl sulfoxide) and stored at −20°C, then diluted in sterile PBS for injection. Volumes injected ranged from 0.05 to 0.1 mL. All mice then received 100% oxygen for 2 h and were returned to their cages. Another group (n = 17) received ebselen and was recovered in room air. Mice were killed at 72 h and scored for injury severity.

Western Blot

Animals were divided into six groups: sham with no exposure to hyperoxia (n = 6), sham with 2 h exposure to 100% oxygen (n = 3), injury with normoxic recovery (n = 10), and three groups that underwent hypoxia—ischemia for 30 mins followed by increasing amounts of 100% oxygen recovery time (1 h (n = 11), 2 h (n = 8), and 4 h (n = 9)). Animals were killed at 24 h, and the ipsilateral-injured hemisphere was dissected for analysis. Individual brains were homogenized and cleared of cellular debris by centrifugation for separate blotting. The supernatant was separated on 10% SDS-PAGE (sodium dodecyl sulfate-polyacrylamide gel electrophoresis) and Western blotted with mouse monoclonal anti-nitrotyrosine antibody (1:1,000; Millipore, Billerica, MA, USA). The same membrane was stripped with 0.2 N NaOH and probed again with a mouse monoclonal anti-B-tubulin antibody (1:1000; Sigma, St Louis, MO, USA). Images were obtained using a Kodak Imaging Station 4000R and underwent densitometry analysis using Kodak Molecular Imaging Software 4.0.3. All bands on anti-nitrotyrosine-labeled blots were quantified together and adjusted on the basis of the analysis of anti-B-tubulin density.

Histology and Immunohistochemistry

Three days after injury, animals were anesthetized with a ketamine/xylazine mixture and perfused with 20 mL of PBS followed by 20 mL of 4% PFA (paraformaldehyde) in PBS. Brains were dissected and postfixed overnight at 4°C in 4% PFA. Whole brains were blocked in 3% agarose and cut into 50 μm coronal sections on a vibratome (Leica VT 1000S).

Sections were stained with cresyl violet (Sigma-Aldrich, St Louis, MO, USA) to assess neuronal morphology after injury. Brains were scored independently by two blinded investigators using a previously described scoring system and discrepancies were resolved by taking a mean of the two assigned scores (Sheldon et al, 1998). Eight regions of the brains were scored: the anterior, middle, and posterior cortex, CA1, CA2, CA3, and dentate gyrus of the hippocampus, and caudate putamen, with the contralateral side serving as a reference for uninjured tissue. Each region was given a score from 0 to 3: 0 = no detectable neuronal cell loss; 1 = small focal areas of neuronal cell loss; 2 = columnar damage in the cortex involving predominantly layers II to IV or moderate cell loss in the hippocampus; and 3 = cystic infarction. The score for each region was summed for a final score ranging from 0 to 24. Animals that died before randomization were excluded from the study; animals that died after randomization (and before perfusion at 3 days) were included in their assigned group and given a maximal injury score of 24 if, upon post-mortem examination, they showed gross evidence of ipsilateral cerebral infarction.

For immunostaining, free floating 50 μm vibratome sections were permeabilized with 0.3% Triton X-100 PBS, washed three times, and blocked with 5% normal donkey serum for 2 h. Slices were incubated overnight at 4°C with primary antibodies: rat anti-PDGFRα (1:250; BD Pharmingen, Franklin Lakes, NJ, USA), rabbit anti-GFP (1:500; Molecular Probes, Carlsbad, CA, USA), rabbit anti-cleaved caspase 3 (1:100; Cell Signaling Technologies, Danvers, MA, USA), and rabbit anti-nitrotyrosine (1:250; Millipore). Colabeling experiments were performed with anti-PDGFRα and anti-nitrotyrosine. Samples were washed again with PBS three times and incubated in secondary antibodies: Cy2 donkey anti-rabbit, Cy3 donkey anti-mouse, and Cy5 donkey anti-rat (1:200; Jackson ImmunoResearch, West Grove, PA, USA) for 2 h at room temperature. Slides were cover-slipped with Fluoromount-G (Southern Biotech, Birmingham, AL, USA) and visualized and photographed using confocal microscopy (Zeiss, Oberkochen, Germany). For the combination of Fluoro-Jade C staining and immunocytochemistry, sections were immunostained as above with anti-cleaved caspase 3, washed with distilled water, and incubated in 0.06% potassium permanganate solution for 5 mins. They were then stained with 0.0004% Fluoro-Jade C (Millipore) for 15 mins, air-dried, and photographed using confocal microscopy. Representative midbrain sections were chosen randomly from ebselen-treated and control mice (100% oxygen only), and the Fluoro-Jade C-positive cells were counted in the CA3 region of the hippocampus.

Staining with anti-myelin basic protein (MBP) was performed in a similar manner, with free-floating slices permeabilized with 0.3% Triton X-100 PBS, washed three times, and blocked with 5% normal goat serum at room temperature for 2 h. Slices were then incubated overnight at 4°C in the primary antibody mouse anti-MBP (1:10; Millipore). Slices were washed and incubated with the secondary antibody biotinylated goat anti-mouse (1:500; Vector, Burlingame, CA, USA) for 2 h at room temperature. Samples were washed with PBS three times, incubated in 0.3% hydrogen peroxide in methanol for 30 mins, and then incubated in Avidin: Biotinylated Enzyme Complex (ABC) solution (Vector) for 2 h. After washing with PBS, slices were added to a well-containing diaminobenzidine (Vector) for 90 secs, washed again, and plated to slides. The slides were then placed in 70%, 95%, and 100% ethanol baths, cover-slipped with Permount (Fisher, Pittsburgh, PA, USA), and visualized with light microscopy (Olympus BX50) and photographed with a color CCD camera using ACT-1 software.

Fluorescent-Activated Cell Sorting

Two litters of nestin-GFP mice with identical birth dates were randomized to 1 of 3 groups: sham (n = 4), normoxia recovery (n = 5), or hyperoxic recovery (n = 5) for 2 h. After 3 days, animals were killed by CO₂ asphyxiation and the brain was freshly dissected. For each mouse, the cortex and hippocampus were separated from the ipsilateral (injured) side of the brain. Tissue was minced with scissors, triturated with 500 μL DDP (dispase, DNase I, papain), and then incubated for 15 mins in a 37°C water bath. The tissue was then retritrated in DDP until homogeneity was achieved. Cells were spun down for 30 mins at 16,000 r.p.m. and washed three times with FACS (fluorescent-activated cell sorting) buffer (DMEM (Dulbecco's modified Eagle's medium)/F12 and 10% FBS (fetal bovine serum)). Cells were then filtered with a 70 μm sterile filter and incubated with primary antibody (rat anti-CD11B and hamster anti-CD81 (BD Pharmingen) for 20 mins each. Between antibodies, cells were washed two times with FACS buffer. Finally, cells were fixed with 10 μL of 4% PFA. A BD FACS Calibur or LSRII machine was used for phenotyping. Each experiment was repeated in quadruplet.

Rotorod and Wire Hang Analysis

Postnatal day 14 mice underwent randomization into 1 of 3 groups: sham injury (n = 17), injury with normoxic recovery (n = 15), or injury with hyperoxic recovery for 2 h (n = 13). Animals survived and underwent rotorod testing (Letica, Bioseb, France) at postnatal day 28 (2 weeks after injury) and again at postnatal day 42 (4 weeks after injury) to assess vestibulomotor function. Mice were placed on the rotorod, which accelerated from 4 to 40 r.p.m. over a 2-min time period, and latency to fall was recorded. Mice underwent four trials on 2 consecutive days by an investigator blinded to the experimental group, with 10 to 15 mins in between trials, yielding a total of 16 trials. After completion of the second rotorod testing period, mice were tested for gross motor strength using the wire hang test. Briefly, mice were suspended by their forepaws on a thin metal wire over a cushioned box and could use their hind legs to steady and support themselves. Latency to fall was recorded up to 2 mins.

Physiologic Description of the Model

The respiratory rates of postnatal day 14 sham and ligated mice were monitored before, during, and after hypoxia, with both normoxia and hyperoxic recovery (n = 6 in each group). A separate experiment was performed using adult mice (n = 20) to determine venous blood gas composition during hypoxia and recovery. Two-month-old mice underwent unilateral carotid artery ligation and were then exposed to 8% oxygen, as previously described. After 30 mins of exposure to hypoxia, mice were removed from the hypoxia chamber and venous blood was obtained within 30 secs via a submaxillary vascular bundle using the Goldenrod lancet (Medipoint, Mineola, NY, USA). Blood gas data were obtained using an i-STAT machine (Abbot, East Windsor, NJ, USA), and mice were then recovered in either room air or 100% oxygen. After 30 mins of recovery, another blood gas was obtained.

Statistics

All statistical analysis was performed using SPSS 14.0. When data were not normally distributed, nonparametric tests (Mann—Whitney U-test for two independent samples and Kruskal—Wallis for comparison of multiple groups) were used to determine significance. Results are expressed as medians and interquartile ranges (25th to 75th percentiles) and box plots are used to graphically display data distributions. Normally distributed data were analyzed by ANOVA (analysis of variance) and are presented as means (±s.d.). All post hoc analyses were performed using a Bonferroni correction. A general linear repeated measures ANOVA model incorporating both within-subjects and between-subjects comparisons was used to analyze motor learning data. Pillai's Trace test was used to determine significance between groups.

Results

Exposure to Brief Periods of Hyperoxia after Hypoxic-Ischemic Injury Worsens Histologic Injury

To establish whether exposure to 100% oxygen after injury results in exacerbation of cerebral injury, mice underwent hypoxic-ischemic injury for 30 mins and were recovered in either room air or 100% oxygen. Mice were then histologically examined with Nissl (cresyl violet) staining 72 h after injury using a validated scoring system (Sheldon et al, 1998). Figure 1 shows a representative section from the normoxia group with a mild injury, detectable only in the CA1 layer of the hippocampus (Figures 1A to 1E), whereas hyperoxia-exposed mice sustained significantly more injury, represented in Figure 1 as a large unilateral cortical infarct with significant injury in the CA1, CA2, and CA3 areas of the hippocampus (Figures 1F to 1J). Exposure to normoxia (n = 24) resulted in a median injury score of 1 (0 to 13; 25th% to 75th%), whereas exposure to 1 h of hyperoxia (n = 15) yielded a median injury score of 19 (2 to 24), and exposure to 2 h of hyperoxia (n = 10), a median score of 20 (13.25 to 24) (Figure 1K). Analysis of variance by Kruskal—Wallis was statistically significant (P = 0.003). On post hoc analysis, the injured group recovered in room air was significantly different from both the injured group recovered in 1 h of hyperoxia (P = 0.045) and the injured group recovered in 2 h of hyperoxia (P = 0.003). There was no difference between the two hyperoxia-exposed groups. As this established that hyperoxia-exposed mice have a worse injury 72 h after the insult, we next examined earlier time points to detect evidence of oxidative stress.

Figure 1

Injury scoring and histologic representation of median scores. (A) Low-magnification view of mildly injured mouse, injury score = 1. Higher magnification views of ipsilateral (B and D) and contralateral (C and E) hippocampus and the CA1 region of the hippocampus show a mild thinning of the cellular layer and scattered pyknotic nuclei, indicative of apoptosis. (F) In low magnification of severely injured mouse, injury score = 19, the arrows identify the large cortical infarct. Higher magnification views of ipsilateral (G and I) and contralateral (H and J) hippocampus and the CA1 region show extensive necrosis and apoptosis present throughout the section on the ipsilateral side. (K) Box plot of injury scoring for three groups; dark line represents median, and outer limits of boxes are Tukey's hinges, approximately 25th and 75th percentiles. Post hoc analysis revealed a significant difference between the group recovered in 1 h hyperoxia compared with the group recovered in room air (P = 0.045) and between the group recovered in 2 h hyperoxia group and the group recovered in room air (P = 0.003). There was no significant difference between 1 and 2 h of hyperoxia. Scale bars = 1 mm (A and F), 200 μm (B, C, G, and H), and 50 μm (D, E, I, and J).

Oxidant Tissue Injury is Increased in the Hyperoxia-Exposed Brain after Hypoxic-Ischemic Injury

To determine whether hyperoxia during reperfusion results in increased oxidative injury to neuronal tissue, we examined ipsilateral-injured hemispheres from normoxia-exposed and hyperoxia-exposed mice for evidence of protein nitrosylation, an established method for measuring oxidative injury (Tan et al, 1998). Nitrotyrosine is formed when tyrosine is nitrated by peroxynitrite, a toxic-free radical formed when superoxide and nitric oxide combine. Exposure for 1, 2, or 4 h consistently shows the formation of high-molecular-weight protein nitrosylation products from increased formation of the nitric oxide-derived species (Figure 2A). When quantified using densitometry, hyperoxic recovery for either 1 h (n = 11, 260, 190 ± 74,500 AU (arbitrary units)), 2 h (n = 8, 278,700 ± 103,940 AU), or 4 h (n = 9, 260,540 ± 62,130 AU) results in a statistically significant quantitative increase in nitrotyrosine accumulation compared with the mice recovered in room air (n = 10, 134,160 ± 61,040 AU) (Figure 2B). There was no difference between the group injured and exposed to normoxia when compared with sham-operated mice (n = 6, 89,020 ± 74,590 AU) or sham mice exposed to 100% oxygen for 2 h (n = 3, 79,010 ± 48,450 AU). There was also no difference between the three groups that were injured and exposed to increasing times of hyperoxia. This suggests that ischemia and reperfusion with highly oxygenated blood cooperate to increase oxidative stress that leads to more oxidative injury in the brain. To better characterize which cell types were most vulnerable to the effects of hyperoxic recovery, we performed immunohistochemistry with anti-nitrotyrosine and markers for early glial progenitors. Nitrotyrosine staining was concentrated in the most severely injured portions of the brain and colocalized predominantly with PDGFRα (platelet-derived growth factor receptor-α)-expressing cells (Figures 2C to 2E). PDGFRα expression represents an immature oligodendrocyte precursor stage in the developing brain, suggesting that these progenitors are susceptible to accumulating oxygen-mediated radicals (Ellison and de Vellis, 1994).

Figure 2

Increasing accumulation of the oxidative metabolite nitrotyrosine and localization to glial progenitors. (A) Western blot of anti-nitrotyrosine (ANT) shows increasing amounts of nitrotyrosine with increasing exposure to hyperoxic recovery. An additional nitrosylated product is seen in hyperoxia-exposed groups in this representative blot. (B) Quantification of ANT staining using densitometry analysis reveals significantly higher arbitrary density units in the hyperoxia-exposed mice compared with sham and normoxia-exposed controls (*P < 0.05, ANOVA with Bonferroni correction on post hoc analysis). (C to E) ANT is present in many cortical cells, and is it localized to the most severe area of injury and colocalizes with PDGFRα, a marker for immature glial cells. Scale bar = 35 μm.

Exposure to as Little as 30 mins of Hyperoxia after Injury Worsens Histologic Injury, and Injury can be Reversed with the Antioxidant Glutathione Peroxidase Mimetic Ebselen

After identifying that exposure to either 1 or 2 h of 100% oxygen during recovery increases injury susceptibility, we sought to determine the boundaries of toxicity by exposing injured mice to various times of 100% oxygen after hypoxia—ischemia. We found a consistent increase in histologic injury in mice exposed to hyperoxic recovery between 30 mins and 4 h compared with normoxic recovery. However, there was no difference with increasing duration of hyperoxic recovery, suggesting that longer exposure to 100% oxygen does not produce further injury (Figure 3A). Interestingly, mice exposed to only 15 mins of hyperoxic recovery (n = 13) trended toward a worse injury with a median injury score of 2.5 (1.5 to 18; 25th% to 75th%) but did not differ statistically from normoxic-recovered mice with a median injury score of 1 (0 to 13), suggesting a threshold effect to oxygen toxicity during reperfusion that is mitigated if the exposure to hyperoxia is brief and limited to less than 15 to 30 mins.

Figure 3

Exposure to as little as 30 mins of hyperoxia after injury worsens histologic injury, and injury can be reversed with the glutathione peroxidase mimetic ebselen. (A) Box plot of injury scoring for cumulative injury groups, including ebselen rescue; dark line represents median, and outer limits of boxes are Tukey's hinges, approximately 25th and 75th percentiles. Post hoc analysis revealed a significant difference between the groups recovered in 30 mins, 2 h, and 4 h of hyperoxia when compared with the normoxia-recovered control group (*P < 0.05). There was no significant difference between mice that were exposed to 100% oxygen for 2 h after injury and rescued with ebselen treatment when compared with normoxia-recovered controls. There was no significant difference between mice recovered in room air and those that received ebselen and recovered in room air. (B and C) Cleaved caspase-3, a marker of apoptosis, is present in only scattered amounts in hippocampal (CA1) cells of mice that received ebselen, but is abundant in those that did not receive ebselen. (D and E) Fluoro-Jade C, a marker of degenerating neurons, is present in scattered amounts in hippocampal (CA1) cells of mice that received ebselen, but is abundant in those that did not receive ebselen. Scale bar = 25 μm.

Our previous data suggest that brief hyperoxic recovery after hypoxia—ischemia increases neuronal injury and is associated with increases in free radical oxidative damage in early glial progenitors. To determine whether the increased injury observed with hyperoxic recovery is mediated through increasing cellular oxidative stress, we randomized mice after hypoxic-ischemic injury to immediately receive either intraperitoneal ebselen, a glutathione peroxidase mimetic, or saline, with both groups receiving subsequent exposure to 2 h of hyperoxia. Brains were examined 72 h after injury using the previously described scoring system. Mice that received saline (n = 29) scored significantly worse with a median injury score of 10 compared with the mice that received ebselen (n = 30), which had a median injury score of 5 (P < 0.001, Mann—Whitney U-test). Although antioxidant therapy did not provide complete protection, the median histologic score for this group was not statistically different than the mice recovered in room air (Figure 3A). When mice recovered in room air were administered ebselen (n = 17), their median injury score of 1 was same as the group of mice injured and recovered in room air without ebselen. Thus, protection provided by ebselen, a known antioxidant, indicates that oxidative stress and free radical toxicity are critical components that underlie the increased injury severity observed with hyperoxic recovery.

To confirm that ebselen provided protection from neuronal injury, we performed immunostaining against cleaved caspase-3, a marker of apoptotic cell death, and Fluoro-Jade C, a stain for degenerating neurons, on sections obtained from ebselen-treated mice and saline-treated controls. Fluoro-Jade C staining was significantly greater in the neuronal pyramidal layers of the CA3 region of the hippocampus of saline-treated mice (n = 5, 54 ± 15) compared with those that received ebselen (n = 5, 13 ± 9; P = 0.001) (Figures 3B to 3E), lending additional support to the notion that increased oxidative stress promotes further secondary neural injury after ischemia.

Glial but not Neuronal Progenitors are Selectively Vulnerable to Hyperoxic Exposure during Reperfusion after Hypoxia—Ischemia

Many cell types within the brain participate in the injury response, including neuronal and glial progenitor cells, astrocytes, and microglia (Urrea et al, 2007). The developing brain is unique, in that it comprises large populations of as yet undifferentiated neural and glial progenitor populations. It is currently unclear how different progenitor cell types differ in their adaptive response to injury. Nestin is expressed broadly in the developing nervous system and in many cell types, including reactive glia, muscle, and endothelial cells (Kawaguchi et al, 2001). As the brain matures, nestin expression in neural tissue is mainly apparent in neural progenitors and is restricted to two neurogenic regions, the subventricular zone and the dentate gyrus in nestin-GFP transgenic mice (Yu et al, 2005). We observe that by postnatal day 14, GFP-expressing neural progenitors have localized to the subgranular layer in the hippocampus and subventricular zone of the lateral ventricle. However, young mice in our nestin-GFP transgenic line show GFP expression in not only neural progenitors in the subgranular layer of the dentate gyrus and subventricular zone but also in immature glial progenitors expressing PDGFRα in the cortex (Figures 4A to 4F). This allows us to reliably distinguish the effects of hyperoxic recovery on GFP-expressing neural progenitors in the hippocampus from GFP-expressing glial populations in the cortex.

Figure 4

Glial progenitors are located in the cortex and pluripotent progenitors are found in the hippocampus in the immature mouse brain. (A and D) Under the control of the nestin promoter, GFP is expressed in stem and progenitor cells in the cortex and hippocampus. (B and E) PDGFRα, a marker for immature glial cells, is present in the cortex and sporadically in the molecular layer and hilus of the hippocampus, but not in the granular layer. (C and F) GFP and PDGFRα colocalize in the cortex, but not the granular layer of the hippocampus, which localizes glial progenitor cells to the cortex and pluripotent stem and progenitor cells to the hippocampus. Scale bars = 50 μm (A to C) and 100 μm (D to F).

We used flow cytometry and endogenous GFP to mark neuronal and glial progenitors, CD11b antibody to mark microglia (Wirenfeldt et al, 2005), and CD81 antibody to mark both microglia and astrocytes (Dijkstra et al, 2000). Neural progenitor cells (GFP+, CD11b-) or (GFP+, CD81+) from the hippocampus were depleted after hypoxia—ischemia injury in both normoxic and hyperoxic-recovered mice (P < 0.001), but did not differ from each other (Figures 5A to 5G). However, glial progenitor cells that are GFP + /CD11b- or GFP + /CD81 + in the cortex were only depleted in the hyperoxic-recovered group compared with the sham group, whereas normoxic recovery was not different from sham controls (P = 0.019) (Figures 5H to 5N). Cortical microglia that are CD11b + /GFP- are decreased in hyperoxic-recovered mice compared with sham animals (6.2% ± 1.7% versus 9.8% ± 2.8%, P = 0.036) as well, but not in the group recovered in room air. There were no statistically significant differences between groups with regard to the hippocampal microglia cell population or the cortical or hippocampal astrocyte cell populations (GFP-/CD81+).

Figure 5

Flow cytometry shows that glial progenitors in the cortex are depleted after injury and recovery with hyperoxia. Images of hippocampal dissections (A to F) visually depict a reduction in progenitor cell population (GFP-positive and CD11b-negative in A, C, and E; GFP-positive and CD81-positive in B, D, and F). Quantification in panel G reveals that pluripotent stem and progenitor cells are depleted after injury in both groups, recovery with room air and hyperoxia (P < 0.001). Flow cytometry images of cortical dissections (H to M) visually depict a marked reduction in glial progenitor cells (GFP-positive and CD lib-negative in H, J, and L; GFP-positive and CD81-positive in I, K, and M) in the injured group recovered in hyperoxia, but not room air. Quantification in panel N reveals a significant difference between groups (P < 0.05). In box plot panels G and N, the dark line represents median and outer limits of boxes are Tukey's hinges, approximately 25th and 75th percentiles.

Myelination of White Matter Tracts is Disrupted and Disorganized after Hyperoxic Recovery in Injured Mice

At postnatal day 14, although oligodendrocyte progenitors are abundant, there is relatively little mature myelination that has occurred in the rodent brain, and myelination increases over the next 2 to 4 weeks to reach adult levels (Sauvageot and Stiles, 2002). We sought to determine whether the increase in oxidative injury and decrease in glial progenitors that we observed after exposure to hyperoxia at postnatal day 14 result in dysfunctional myelin production. Mice that were injured at postnatal day 14 were killed 4 weeks later and examined for mature myelination with MBP immunohistochemistry. Figures 6A to 6C show that although myelination does not appear grossly disrupted in hyperoxic-recovered mice, when examined at higher power, the ipsilateral (injured) myelin staining pattern is reduced and disorganized compared with the contralateral (uninjured) side, suggesting that injury to early progenitors leads to long-term defects in myelination. This reduced MBP staining is most prominent in the cortical-thalamic tracts, where cortical motor neurons project to the spinal cord.

Figure 6

Injured mice have a more disrupted myelination pattern with myelin basic protein staining and perform significantly worse on behavior testing with rotorod and wire hang. Low-magnification views of an injured mouse recovered in hyperoxia do not reveal marked differences between the ipsilateral (injured) and the contralateral (uninjured) sides (A). Higher magnification views comparing the ipsilateral (B) with the contralateral (C) side reveal a disrupted and disorganized MBP staining pattern in the cortical-thalamic tracts. (D) Results of wire hang motor testing reveal a significantly shorter latency to fall in hyperoxia-exposed mice when compared with sham-injured mice (*P < 0.01). (E to H) In rotorod testing, examining the mean latency to fall 2 weeks after injury, there is a significant difference between hyperoxia-exposed mice and sham controls, but not in normoxia-exposed controls (E). Four weeks after injury, when myelination is expected to be complete, examining the mean latency to fall reveals a significant difference between hyperoxia-exposed mice and normoxia-exposed controls (F) (P = 0.045). A general linear repeated measures ANOVA model was performed on data from the two testing periods, 2 weeks after injury (G) and 4 weeks after injury (H). There were no differences between the slopes of each group during either time point, indicating that motor learning was not significantly different between groups. Data are presented as means with error bars representing s.e.m. Pillai's Trace was used to determine statistical significance between slopes. Scale bars = 1 mm (A) and 200 μm (B and C).

Exposure to Hyperoxia during Recovery Leads to Increased Functional Motor Deficits and a Static Motor Lesion

As hyperoxic-recovered mice show increased histologic injury, depletion of myelin progenitors, and impaired myelination of cortical-spinal neurons, we next tested whether recovery in 100% oxygen, would affect functional recovery after hypoxic-ischemic injury. Strength testing using the wire hang test was performed at postnatal day 42 and revealed that the hyperoxic-recovered mice performed significantly worse (median time to fall 16 secs, range = 4 to 20; 25th% to 75th%) than the sham-injured mice (49 secs, range = 29 to 69) (Figure 6D). However, there was no difference between normoxic recovery and hyperoxic recovery groups, or between the sham-operated mice and normoxic-recovered mice.

Rotorod testing was performed at postnatal days 28 to 29 and 42 to 43 (2 and 4 weeks after injury). At postnatal days 28 to 29, we observed a significant decrease in latency to fall among the injured mice recovered in 100% oxygen (47 secs, mean = 32 to 62) compared with sham-injured controls (73 secs, mean = 59 to 80; P = 0.001); however, there was no difference between the hyperoxia-recovered mice and the normoxia-recovered controls (Figure 6E). However, when animals were tested again at postnatal days 42 to 43, we found a significant difference between the normoxic recovery (79 secs, mean = 51 to 88) and the hyperoxic recovery group (41 secs, mean = 29 to 50, P = 0.045) (Figure 6F). This difference became apparent primarily owing to a failure in improvement in motor function in hyperoxic-recovered mice that was observed in both the sham and normoxic-recovered animals. Improvement in latency to fall times during development is not unexpected, as significant myelination and vestibulomotor development continues to occur during the first 6 weeks in the rodent. To exclude defects in motor learning as a cause for lack of improvement in rotorod testing, we next examined the data for evidence of motor learning by evaluation with a general linear model and multivariate analysis of repeated measures. There were no differences between the slopes of each group during either testing period, indicating that motor learning was not significantly different between groups (Figures 6G and 6H).

Physiologic Description of the Model

Sham and ligated mice had similar respiratory rates before hypoxia exposure (137 ± 5 versus 134 ± 4 breaths per min). Both groups had a marked increase in respiratory rate when exposed to hypoxia (192 ± 8 and 196 ± 5 breaths per min), which was significantly different when compared with preexposure respiratory rates (P < 0.001). During recovery in room air, both sham and ligated mice returned to a prehypoxia respiratory rate (132 ± 7 and 128 ± 7 breaths per min), as did both groups recovered in 100% oxygen (134 ± 8 and 134 ± 11 breaths per min). There was a significant difference between both recovery groups and mice undergoing hypoxia (P < 0.001), but no difference in any group between sham and ligated mice.

After 30 mins of hypoxia, mice (n = 15) had a pH of 7.29 ± 0.03 with a pCO₂ of 47 ± 6. Mice recovered in room air (n = 9) had lower pH of 7.23 ± 0.03 with a higher pCO₂ of 53 ± 6. Mice recovered in 100% oxygen had a similar pH to room air-recovered mice of 7.20 ± 0.05 with a pCO₂ of 58 ± 5. There was a statistically significant difference in pH between the hypoxic mice and both recovery groups (P = 0.005 for room air group and P < 0.001 for hyperoxia group). There was also a significant difference in pCO₂ between the hypoxic mice and both recovery groups (P = 0.039 for room air group and P < 0.001 for hyperoxia group). There was no statistically significant difference between either recovery group for pH or for pCO₂.

Discussion

Hypoxic-ischemic brain injury is an important cause of neurodevelopmental disabilities in childhood, as the immature brain is both vulnerable to secondary injury and may be limited in its ability to undergo self-repair. Oxidative stress is one important factor that contributes to secondary neural injury, and despite many years of study in the context of brain injury, there are currently no clinically accepted ways of attenuating its effects. One common tenet in clinical resuscitation research after hypoxic-ischemic insults includes not only reestablishment of normoxia but also administration of hyperoxic gas, presumably to ensure a vigorous resuscitation (Niermeyer et al, 2000). In fact, many investigators advocate for global hyperoxia to the brain after injury in the hope of optimally oxygenating marginalized brain tissue (Sunami et al, 2000). This study suggests that such a practice may not be beneficial and may, in fact, result in increased injury and worse functional outcome.

Recovery in hyperoxia after hypoxic-ischemic injury has deleterious effects on cerebral metabolism and inflammation, although long-term developmental and functional outcomes are unknown (Dohlen et al, 2005). In this study, we show that young mice sustain significant increases in histologic brain injury after unilateral hypoxic-ischemic brain injury when recovered in 100% oxygen compared with room air. We observed substantial increases in neuronal injury after as little as 30 mins of hyperoxic recovery, which could easily encompass a reasonable resuscitation period in humans. Minimal exposure to hyperoxia during resuscitation is not integrated into current models of resuscitation, and our results suggest that limiting exposure to hyperoxic recovery is crucial in preventing further secondary injury.

We show that hyperoxic recovery increases the oxidative stress and free radical production resulting in nitrosylation of proteins in injured brain tissue. Furthermore, oligodendrocyte progenitor cells, which are known to be sensitive to hypoxia—ischemia and free radical injury, appear to be most susceptible to accumulating oxidative damage (Back et al, 2002; Tan et al, 1998). Reactive oxygen species are generated in a variety of ways after oxidative stress, depending on the stage of ischemia and state of reperfusion (Abramov et al, 2007). The production of reactive oxygen species appears to be a dynamic process that is driven by several factors, and defense against free radical damage is mediated by two major groups: antioxidant enzymes (superoxide dismutases, glutathione peroxidases, and catalase) and low-molecular-weight antioxidants (ascorbate, glutathione, vitamin E, and coenzyme O) (Bayir et al, 2006).

Oxidative stress upregulates antioxidant enzymes in adult rodents after various forms of neuronal injury, including traumatic brain injury, ischemic injury, and hypoxic preconditioning (Arthur et al, 2004; Goss et al, 1997; Guegan et al, 1998). Glutathione peroxidase, which detoxifies peroxynitrite, is thought to be crucial in limiting oxidative injury (Sheldon et al, 2004). Importantly, the immature brain contains significantly less glutathione peroxidase activity than the mature brain and lacks the compensatory increase seen in adult animals after neural injury (Bayir et al, 2006; Sheldon et al, 2004). The immature brain also has significantly less superoxide dismutase activity than the adult brain but increased levels of catalase (Bayir et al, 2006). This suggests that the immature developing brain, unlike the adult brain, may be uniquely vulnerable to increases in oxidative injury during resuscitation. Ebselen is a selenium-containing organic antioxidant that mimics the action of glutathione peroxidase and has been shown to improve outcomes in various models of ischemic injury in animals and humans (Takasago et al, 1997; Yamaguchi et al, 1998). We show here that the administration of ebselen immediately after a hypoxic-ischemic insult markedly attenuates the increased injury induced during hyperoxic recovery. This supports our hypothesis that hyperoxic exposure mediates increasing neural injury through mechanisms related to increased oxidative stress and toxicity.

Stem and progenitor cells in the developing brain are abundant and continually add functional neurons and other cell types to the brain throughout life (Miles and Kernie, 2006). In humans, the newborn and immediate neonatal periods are critical for normal oligodendrocyte formation and ultimate myelination, because this is the stage at which oligodendrocyte progenitors are most abundant (Back et al, 2002). Mechanisms underlying recovery may include not only progenitor-mediated replacement of lost cells but also changes in dendritic arborization, spine density, and synaptogenesis (Miles and Kernie, 2006). Developing progenitor cells are believed to be more vulnerable to hypoxic-ischemic injury and oligodendrocyte progenitor cells, in particular, are more sensitive to oxidative stress than mature oligodendrocytes (Back et al, 2002). The data presented here show that glial progenitors accumulate nitrotyrosine and are preferentially depleted after hyperoxic recovery, which is consistent with in vitro studies, suggesting that oligodendrocyte progenitor cells are sensitive to oxidative injury (Back et al, 1998).

A disruption in myelination, whether from cerebral anoxia, cerebral hemorrhage, or infection, can result in static motor lesions and underlie cerebral palsy in humans (Hoon, 2005). Here, we show that hyperoxic recovery results in disrupted myelination, and even moderately injured mice show irregular and decreased myelin deposition, particularly in the cortical-thalamic motor tracts. In addition, we show the developmental significance of this impaired myelination where exposure to hyperoxic recovery causes a static motor deficit similar to that seen in cerebral palsy. It may also be possible that injury to projecting cortical neurons and axons inhibited proper myelination in these areas. Although based on previous data that glial progenitors are depleted and accumulate oxidant injury after injury, we believe that observed defects in myelination are due, at least in part, to selective destruction of oligodendrocyte progenitors, which is exacerbated by hyperoxic recovery.

In conclusion, we establish that exposure to hyperoxia for even brief periods of time after hypoxic-ischemic brain injury results in an increased histologic injury that is accompanied by increased oxidative stress, depletion of oligodendrocyte progenitors, and impaired functional recovery from injury. Hypoxia—ischemia damages some cells irreversibly, and although other cells are destined for survival, there apparently exists a range of injured and vulnerable cells that may or may not survive, depending on the amount of secondary injury that ensues. Increasing oxidative stress with excessive exposure to oxygen favors cell death. Further studies are needed to better characterize the mechanisms underlying this sensitivity to injury and continued efforts are needed to determine the best possible concentration of oxygen required for optimal resuscitation while limiting secondary injury.