Oligodendrocyte excitotoxicity determined by local glutamate accumulation and mitochondrial function

Authors


Address correspondence and reprint requests to Dr Frances E. Jensen, Division of Neuroscience, Children's Hospital, Harvard Medical School, 300 Longwood Avenue, Boston, MA 02115, USA. E-mail: frances.jensen@tch.harvard.edu

Abstract

Developing oligodendrocytes (OL precursors, pre-OLs) express α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) subtype glutamate receptors (AMPARs) and are highly vulnerable to hypoxic-ischemic or oxygen-glucose deprivation (OGD)-induced excitotoxic injury, yet the mechanisms of injury remain unclear. Here we investigated the role of glutamate accumulation and mitochondrial function in OGD-induced pre-OL toxicity in vitro. Bulk glutamate concentration in the culture medium did not increase during OGD and OGD-conditioned medium did not transfer toxicity to naïve cells. Facilitation of glutamate diffusion by constant agitation of the culture reduced, while inhibition of glutamate diffusion by increasing medium viscosity with dextran enhanced, OGD-induced pre-OL injury. Depletion of extracellular glutamate by the glutamate scavenging system, glutamate-pyruvate transaminase plus pyruvate, attenuated pre-OL injury during OGD. Together these data suggest that local glutamate accumulation is critical for OGD toxicity. Interestingly, under normoxic conditions, addition of glutamate to pre-OLs did not cause receptor-mediated toxicity, but the toxicity could be unmasked by mitochondrial impairment with mitochondrial toxins. Furthermore, OGD caused mitochondrial potential collapse that was independent of AMPAR activation, and OGD toxicity was enhanced by mitochondrial toxins. These data demonstrate that pre-OL excitotoxicity is exacerbated by mitochondrial dysfunction during OGD. Overall, our results indicate that OGD-induced pre-OL injury is a novel form of excitotoxicity caused by the combination of local glutamate accumulation that occurs without an increase in bulk glutamate concentration and mitochondrial dysfunction. Therapeutic strategies targeting local glutamate concentration and mitochondrial injury during hypoxia-ischemia may be relevant to human disorders associated with pre-OL excitotoxicity.

Abbreviations
AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

AMPAR

AMPA receptor

CTZ

cyclothiazide

GluR

glutamate receptor

HPLC/MS/MS

high performance liquid chromatography coupled with tandem mass spectrometry

FCCP

p-trifluoromethoxy carbonyl cyanide phenylhydrazone

JC-1

5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbiocyanine iodine

MBP

myelin basic protein

NBQX

6-nitro-7-sulfamoylbenzo[f]quinoxaline-2,3-dione

OGD

oxygen-glucose deprivation

OL

oligodendrocyte

pre-OL

oligodendrocyte precursor

PVL

periventricular leukomalacia

TUNEL

terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling

Glutamate is a major excitatory neurotransmitter in the brain, but can also cause excitotoxic injury to the nervous system under many pathological circumstances. Glutamate excitotoxicity is indeed a common mechanism underlying the pathogenesis of many nervous system disorders, and has been primarily studied in neurons (Rothman and Olney 1986; Choi 1988; Olney 1989; Choi 1992; Lipton and Rosenberg 1994). However, emerging evidence indicates that oligodendrocytes (OLs) share with neurons a high vulnerability to excitotoxic injury (Yoshioka et al. 1995; Yoshioka et al. 1996; Matute et al. 1997; McDonald et al. 1998; Liu et al. 1999; Fern and Moller 2000; Follett et al. 2000; Liu et al. 2002a; Deng et al. 2003; Rosenberg et al. 2003). Although the main function of OLs is the formation and maintenance of the myelin sheath around neuronal axons, much attention has focused recently on the possibility that these cells are capable of responding to or influencing neuronal activity and that these functions may be essential for neuroprotection and repair (Baumann and Pham-Dinh 2001; Deng and Poretz 2003).

Hypoxia-ischemia causes excitotoxic injury to OLs in vivo and in vitro (Yoshioka et al. 1995; Yoshioka et al. 1996; Matute et al. 1997; McDonald et al. 1998; Fern and Moller 2000; Follett et al. 2000; Deng et al. 2003). Hypoxic-ischemic brain injury in the premature infant results in selective cerebral white matter damage, a disorder termed periventricular leukomalacia (PVL) (Volpe 2001; Volpe 2003). PVL is the leading cause of cerebral palsy in premature infants, yet no specific therapy currently exists for this serious pediatric problem (Volpe 2001; Volpe 2003). PVL primarily involves developing OLs (OL precursors, pre-OLs) that populate the developing white matter during the period of greatest risk for the lesion (Back et al. 2001). We have previously shown that pre-OLs are more sensitive to excitotoxic death than mature OLs (Follett et al. 2000; Deng et al. 2003; Rosenberg et al. 2003), and pre-OL excitotoxicity is implicated in the pathogenesis of PVL (Volpe 2001; Follett et al. 2000; Deng et al. 2003; Deng et al. 2004; Follett et al. 2004). Oxygen-glucose deprivation (OGD) has been shown to cause glutamate release via reverse glutamate transport in pre-OLs, leading to excitotoxic pre-OL death (Fern and Moller 2000; Deng et al. 2003). Neuronal death under hypoxic-ischemic conditions has been shown to be dependent upon extracellular glutamate accumulation leading to excessive glutamate receptor (GluR) activation and induction of excitotoxic cascades (Rothman and Olney 1986; Choi 1988; Olney 1989; Choi 1992; Lipton and Rosenberg 1994). In contrast, it is not clear whether extracellular glutamate accumulation is responsible for cell death in hypoxic-ischemic or OGD-induced pre-OL injury.

Previous studies demonstrate that pre-OL excitotoxicity is mediated by Ca2+-permeable α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptors (AMPARs) (Deng et al. 2003; Liu et al. 2002a; Rosenberg et al. 2003). Mitochondria are the major intracellular sink for Ca2+, and Ca2+ overload causes mitochondrial dysfunction in neurons (Nicholls et al. 1999; Reynolds 1999). However, the role of mitochondrial function in pre-OL excitotoxicity remains unclear.

In the present study, we investigate the role of glutamate accumulation and mitochondrial function in pre-OLs excitotoxicity. We first report that OGD-induced excitotoxic pre-OL injury is caused by local glutamate accumulation that occurs on a scale not affecting bulk glutamate concentration. We next show that due to receptor desensitization glutamate does not elicit receptor-mediated excitotoxicity in pre-OLs under normoxic conditions, but interestingly this toxicity can be unmasked by mitochondrial impairment. Furthermore, we demonstrate that pre-OL excitotoxicity is exacerbated by mitochondrial dysfunction during OGD. These insights into the excitotoxic mechanisms of pre-OL injury are relevant to both understanding of the pathogenesis and design of therapeutic strategies for white matter disorders, such as PVL.

Materials and methods

Cell culture

Highly enriched pre-OLs were obtained from newborn Sprague-Dawley rat brains using a selective detachment procedure, as described in detail elsewhere (McCarthy and de Vellis 1980; Deng et al. 2003; Rosenberg et al. 2003; Deng et al. 2004). Pre-OLs were maintained in a chemically defined Dulbecco's Modified Eagle's Medium (Invitrogen Corporation, Carlsbad, CA, USA) containing d-glucose (25 mm), l-glutamine (4 mm), sodium pyruvate (1 mm), human apo-transferrin (50 µg/mL), bovine pancreatic insulin (5 µg/mL), sodium selenium (30 nm), hydrocortisone (10 nm), d-biotin (10 nm), bovine serum albumin (1 mg/mL), recombinant human platelet derived growth factor-AA (10 ng/mL), and basic fibroblast growth factor (10 ng/mL) for 6 days with fresh medium change every 2 days. All experiments were performed with cultures maintained in the chemically defined medium without the supplemental growth factors. Cultures were routinely characterized by immunocytochemical detection of the expression of developmental stage-specific OL markers, A2B5 (progenitors), O4 (later-stage precursors), O1 (immature OLs), and myelin basic protein (MBP) (mature OLs). A representative pre-OL culture had the following composition: 95% A2B5 +, 90% O4 +, 4% O1 +, and 1% MBP +. All cultures contained less than 2% of glial fibrillary acidic protein-positive astrocytes and essentially non-detectable CD11 + microglia.

Oxygen-glucose deprivation

Cultures were switched to the same medium that was deoxygenated and lacked glucose (Invitrogen), and transferred to an anaerobic chamber filled with 95% N2 plus 5% CO2 at 37°C (Deng et al. 2003; Deng et al. 2004). Following OGD for 2 h, d-glucose as a concentrated stock solution made in the glucose-free medium was added back to the cultures to a final concentration of 25 mm, and the cultures were returned to a normoxic 5% CO2 incubator at 37°C. Cell death was assessed at 24 h by measuring the extent of lactate dehydrogenase release from the cells into medium, as previously described (Deng et al. 2003; Deng et al. 2004).

Determination of glutamate concentration

Glutamate concentration was measured by high performance liquid chromatography coupled with tandem mass spectrometry (HPLC/MS/MS). Authentic glutamic acid was used to generate standard curves, and N-methyl glutamate (final concentration = 500 ng/mL) was included in each measurement as the internal standard. Culture media collected before and after OGD or working solutions of glutamate standard were subjected to protein precipitation by addition of 3 volumes of acetonitrile. Supernatants were dried under a stream of nitrogen, and reconstituted in 100 µL acetonitrile. Reconstituted samples were injected into an Agilent 1100 HPLC system (Agilent, Palo Alto, CA, USA) with a Leap CTC PAL refrigerated autosampler (Leap Technologie, Carrboro, NC, USA), a Thermo Betasil Silica-100 (100 × 2.1 mm, 5 µm) column (Thermo Electron Corporation, Waltham, MA, USA), an isocratic gradient of 30% mobile phase A (1% formic acid in water) and 70% mobile phase B (1% formic acid in acetonitrile), and the flow rate of 400 µl/min. Glutamate was detected by MS/MS with a Sciex API 4000 Qtrap system (Applied Biosystems/MDS Sciex, Concord, Ontario, Canada) equipped with electrospray ionization and multiple-reaction monitoring transition. The data were processed by the computer program analyst version 1.4.1 (Applied Biosystems/MDS Sciex).

45Ca2+ uptake.

Cultures were incubated with 45CaCl2 (8 µCi/mL) at 22°C for 10 min, then washed with Hank's balanced salts solution (Invitrogen) and lysed with 1% Triton-X 100 (Sigma, St Louis, MO, USA). Radioactivity in the whole lysate was counted by liquid scintillation (Deng et al. 2003; Deng et al. 2004).

Measurement of mitochondrial dysfunction.

5,5′,6,6′-Tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbiocyanine iodine (JC-1, Invitrogen) was used to measure mitochondrial membrane potential in pre-OLs (Baud et al. 2004). JC-1 monomer fluoresces at 510–527 nm (green). When JC-1 is concentrated by actively respiring mitochondria, it aggregates and fluoresces at 590 nm (red). JC-1 (1 µg/mL) was added directly into the culture medium 20 min before microscopic cellular imaging or quantitative fluorescence ratiometric measurement with plate-reader-based assays.

Terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling (TUNEL) staining and DNA fragmentation assays

An in situ apoptosis detection kit (Roche, Indianapolis, IN, USA) was used for TUNEL staining according to the protocols provided by the manufacturer. Cells were also stained with Hoechst 33258 to reveal nuclear fragmentation in situ. Alternatively, DNA fragmentation was determined by agarose gel electrophoresis following DNA extraction from cultured pre-OLs using a commercially available apoptotic DNA-ladder kit from Roche.

Data analysis.

All experiments were performed in triplicate, and all data represent the mean ± SEM values of three independent experiments. Statistical differences were assessed by one-way analysis of variance (anova) with Tukey post hoc analysis for multiple comparisons. Student's t-test was used when only two independent groups were compared. Statistical significance was determined at p < 0.05.

Results

OGD does not increase glutamate concentration in the culture medium and OGD-conditioned medium does not transfer toxicity to naïve cells

Hypoxia-ischemia in vivo and OGD in vitro cause extracellular glutamate accumulation from neurons (Benveniste et al. 1984; Perez Velazquez et al. 1997; Izumi et al. 2002). Pre-OLs, like neurons, are highly vulnerable to OGD-induced injury (Fern and Moller 2000; Deng et al. 2003; Rosenberg et al. 2003). To address whether OGD-induced pre-OL injury is due to the accumulation of glutamate or other toxic factors in the culture medium, we first examined whether OGD-conditioned medium would transfer toxicity to naïve cells. Oxygen-glucose deprivation (2 h) caused significant cell death at 24 h as compared to the control culture (Fig. 1), while transferring culture medium harvested 2 h after OGD to naïve cells failed to cause any toxicity (Fig. 1). Furthermore, transferring OGD-conditioned medium harvested 4, 6, 8, and 10 h after OGD (2 h) also was not toxic to naïve cells (data not shown).

Figure 1.

 Representative pre-OL cell images under the normoxic condition (control), 24 h afer OGD, or after transferring OGD-conditioned culture medium. OGD-conditioned culture medium did not transfer toxicity to naïve cells.

To specifically address the role of glutamate accumulation in OGD-induced pre-OL injury, we established a highly sensitive HPLC/MS/MS method to measure glutamate concentration. The glutamate concentration in the fresh normal DMEM-based glucose-free culture medium (Invitrogen) was 1.96 ± 0.10 µm. We measured glutamate concentrations in the culture medium 0, 2, 4, 6, 8, 10 h after exposure of the cells to OGD (2 h), when cell death was not yet apparent. The extracellular glutamate concentration immediately after OGD for 2 h was 1.94 ± 0.17 µm. Oxygen-glucose deprivation did not increase bulk glutamate concentration in the medium at any examined time point (Fig. 2), but rather appeared to decrease the concentration to 0.90 ± 0.07, 0.80 ± 0.10, 0.70 ± 0.17, 0.54 ± 0.07, and 0.58 ± 0.06 µm (n = 3) at 2, 4, 6, 8, and 10 h after OGD for 2 h, respectively.

Figure 2.

 Glutamate concentration in the culture medium does not increase during OGD to pre-OLs. Extracellular glutamate was measured before and 0, 2, 4, 6, 8, 10 h after exposure of pre-OLs to OGD (2 h) by HPLC/MS/MS. Values (µM) are mean ± SEM of culture media collected from three cultures for each data point.

Local glutamate accumulation is responsible for OGD-induced pre-OL injury

Previous studies demonstrate that OGD causes glutamate release via reverse transport in pre-OLs (Fern and Moller 2000; Deng et al. 2003). Given the lack of bulk extracellular glutamate accumulation, we aimed to determine whether local glutamate accumulation during OGD might be responsible for pre-OL death. As it is not possible to directly measure local glutamate concentration, we employed several experimental manipulations to the OGD paradigm of pre-OL injury to test the hypothesis that glutamate accumulation occurs locally. First, we examined whether experimental manipulations to alter glutamate diffusion in the medium could alter OGD-induced pre-OL toxicity. We found that OGD-induced death was significantly decreased by constant agitation of the culture (facilitation of glutamate diffusion) compared to standard OGD conditions (p < 0.001) (Fig. 3a). Consistent with previous data (Deng et al. 2003; Deng et al. 2004), OGD-induced toxicity was blocked by the AMPAR antagonist 6-nitro-7-sulfamoylbenzo[f]quinoxaline-2,3-dione (NBQX, 50 µm) (Fig. 3a). Furthermore, the injury was significantly enhanced (p < 0.05) when increasing medium viscosity with 5% 40-kDa dextran (inhibition of glutamate diffusion) (Fig. 3b). The enhancement in toxicity by addition of dextran was blocked by NBQX (50 µm) (Fig. 3b), demonstrating the specificity for AMPARs. The bulk glutamate concentrations in the medium measured at any examined time point following OGD was not different with or without constant agitation of the culture and with or without 5% 40-kDa dextran in the medium (data not shown). In addition, introduction of a glutamate scavenging system (20 units/mL glutamate-pyruvate transaminase plus 1 mm pyruvate which converts glutamate to 2-ketoglutarate) into the medium markedly reduced OGD (2 h)-induced cell death at 24 h (p < 0.001) (Fig. 3c), while pyruvate (1 mm) alone had no effect (Fig. 3c). Taken together, these results indicate that OGD-induced pre-OL toxicity is due mainly to local glutamate accumulation that occurs on a scale not affecting bulk glutamate concentration.

Figure 3.

 OGD-induced excitotoxic pre-OL injury is caused by local glutamate accumulation. (a) Facilitation of glutamate diffusion by constant agitation of the culture attenuated OGD (2 h)-induced pre-OL injury at 24 h. (b) Inhibition of glutamate diffusion by increasing medium viscosity with 5% 40-kDa dextran exacerbated OGD (2 h)-induced pre-OL injury at 24 h, and the enhancement in toxicity was blocked by the AMPAR antagonist NBQX (50 µm). (c) Depletion of extracellular glutamate by the glutamate scavenging system, glutamate-pyruvate transaminase (GPT) + pyruvate, reduced OGD (2 h)-induced cell death at 24 h. Data represent mean ± SEM of three separate experiments. *p < 0.05, **p < 0.01 vs. OGD.

Under normoxic conditions, exogenous glutamate fails to cause receptor-mediated toxicity to pre-OLs because of receptor desensitization, but causes non-receptor-mediated toxicity at high concentrations

As glutamate is the putative excitotoxin during hypoxia-ischemia or OGD, we studied the concentration-dependence of glutamate toxicity to pre-OLs by applying various concentrations of glutamate in the culture medium in normoxic conditions (Fig. 4a). Cell survival was measured at 24 h. Glutamate did not cause any toxicity to normoxic pre-OLs unless very high concentrations of glutamate (millimolar) were added. The ED50 (50% effective dose) of glutamate toxicity to pre-OLs was 2.6 ± 0.4 mm. Because glutamate can cause receptor-dependent or independent toxicity (Oka et al. 1993), we examined the mechanisms of glutamate toxicity to pre-OLs. The AMPAR antagonist NBQX (50 µm) was not protective against pre-OL toxicity induced by a very high concentration of glutamate (5 mm) (Fig. 4B), indicating that this toxicity is receptor-independent. High concentrations of glutamate are known to cause intracellular cystine depletion by acting on the glutamate/cystine exchanger on the plasma membrane. Cystine is a precursor of glutathione, which is the major cellular antioxidant, high concentrations of glutamate may thus cause oxidative injury. Addition of cystine (2 mm) or the antioxidant vitamin E (30 µm) was significantly protective in the paradigm of toxicity induced by glutamate (5 mm) (Fig. 4b), indicating that high concentrations of glutamate cause non-receptor-mediated oxidative injury to pre-OLs.

Figure 4.

 Glutamate causes receptor-independent toxicity at high concentrations and receptor-dependent injury to pre-OLs when receptor desensitization is blocked. (a) Dose–response of glutamate toxicity to pre-OLs. Glutamate (0–1 m) was applied to pre-OLs, and cell survival was assayed at 24 h. (b) A high concentration of glutamate (5 mm) caused receptor-independent oxidative injury via inhibition of the glutamate/cystine exchanger. **p < 0.01 vs. glutamate (5 mm). (c) A lower concentration of glutamate (0.5 mm) caused no apparent toxicity to pre-OLs due to receptor desensitization. ***p < 0.001 vs. glutamate (0.5 mm). (d) The AMPAR receptor desensitization blocker CTZ potentiated glutamate (0.5 mm)-induced 45Ca2+ uptake in a dose-dependent manner, and NBQX (50 µm) blocked the effect. Data represent mean ± SEM of three separate experiments.

Interestingly, a lower concentration of glutamate (0.5 mm) caused no toxicity to normoxic pre-OLs (Fig. 4c). However, the AMPAR desensitization blocker cyclothiazide (CTZ, 50 µm) could unmask the toxicity, and the AMPAR antagonist NBQX (50 µm) totally blocked this toxicity (Fig. 4c). In addition, CTZ potentiated glutamate-induced 45Ca2+ uptake in a dose-dependent manner, and NBQX blocked this 45Ca2+ uptake (Fig. 4d), consistent with the previous demonstration that AMPAR-mediated pre-OL toxicity is Ca2+ dependent (Fern and Moller 2000; Deng et al. 2003; Rosenberg et al. 2003). Thus, under normoxic conditions, AMPAR desensitization prevents glutamate from causing receptor-mediated pre-OL toxicity, but high concentrations of glutamate cause non-receptor-dependent oxidative toxicity in pre-OLs.

Mitochondrial impairment can unmask receptor-mediated glutamate toxicity under normoxic conditions

The lack of receptor-mediated excitotoxicity to pre-OLs caused by glutamate under normoxic conditions raises the issue of whether pre-OL injury during OGD can be solely due to receptor-mediated glutamate toxicity in the absence of other modifying factors. We first examined whether certain factors that might exacerbate glutamate toxicity were present in the culture medium during OGD. As stated above, OGD-conditioned medium harvested 0, 4, 6, 8, and 10 h after OGD (2 h) was not toxic to naïve pre-OLs. Furthermore, addition of glutamate (0.5 or 1 mm) to the conditioned medium failed to elicit any toxicity to pre-OLs (data not shown), indicating that OGD-induced pre-OL injury is not due to secreted factors in the medium that can modulate or unmask glutamate toxicity, and that the modifying factors for OGD-induced pre-OL injury may be cell-intrinsic.

We hypothesized that mitochondrial dysfunction might be one such modifying factor that can aggravate pre-OL excitotoxicity, as previous neuronal studies revealed that elevation of extracellular glutamate to levels that are not sufficient to induce neuronal damage can do so when mitochondrial energy metabolism is compromised (Nicholls et al. 1999; Reynolds 1999). We tested whether glutamate could cause receptor-mediated toxicity in pre-OLs when mitochondrial function is compromised. As shown in Fig. 4(c), addition of glutamate (0.5 mm) caused no apparent pre-OL toxicity under baseline normoxic conditions. However, glutamate (0.5 mm) toxicity was unmasked by prior treatment (24 h) with the mitochondrial respiratory chain uncoupler p-trifluoromethoxy carbonyl cyanide phenylhydrazone (FCCP, 100 nm) or the mitochondrial respiratory chain inhibitor sodium cyanide (NaCN, 1 µm), each of which at the specified concentration had no effect on pre-OL survival (Fig. 5a). The unmasking of glutamate toxicity by FCCP (100 nm) or NaCN (1 µm) was blocked by NBQX (50 µm), indicating that the injury is receptor-dependent (Fig. 5a). However, higher concentrations of either FCCP or NaCN did cause toxicity to pre-OLs (Fig. 5b and c). The AMPAR antagonist NBQX (50 µm) blocked or only partially attenuated FCCP or NaCN-induced toxicity with levels of protection being dependent upon the concentrations of FCCP or NaCN (Fig. 5b and c), indicating that high concentrations of FCCP or NaCN also elicited AMPAR-independent toxicity to pre-OLs.

Figure 5.

 Role of mitochondrial dysfunction in glutamate toxicity to pre-OLs. (a) Receptor-mediated glutamate (0.5 mm) toxicity was unmasked by mitochondrial impairment with FCCP (100 nm) or NaCN (1 µm), each of which itself had no effect on survival. (b, c) FCCP-(b) or NaCN- (c) induced toxicity was attenuated by the AMPAR antagonist NBQX (50 µm). Data represent mean ± SEM of three separate experiments. **p < 0.001 vs. glutamate (0.5 mm).

OGD causes mitochondrial potential collapse, and mitochondrial dysfunction enhances OGD toxicity to pre-OLs

As receptor-mediated glutamate toxicity to pre-OLs can be unmasked and enhanced by mitochondrial impairment, we investigated the role of mitochondrial function in OGD injury to pre-OLs (Fig. 6). We first examined whether OGD causes mitochondrial potential collapse in pre-OLs. We used JC-1 to probe mitochondrial function by imaging the monomeric form (green) and the aggregate form (red) of JC-1. A decrease in red fluorescence and an increase in green fluorescence in pre-OLs were seen after OGD for 2 h, indicating mitochondrial potential collapse (Fig. 6a). Interestingly, the AMPAR antagonist NBQX (50 µm) had no effect on this mitochondrial impairment (Fig. 6a), indicating that OGD causes mitochondrial potential collapse in a receptor-independent manner. FCCP (1 µm) treatment for 10 min was used as a positive control of mitochondrial potential collapse in these experiments. Next, to quantitatively assay JC-1 signals, we established a method using a multiwell fluorescence plate reader to perform ratiometric measurements with JC-1 by comparing the ratio of the monomeric form to that of the aggregate form. The quantitative data obtained from this ratiometric assay showed that OGD causes mitochondrial impairment independently of AMPAR activation (Fig. 6b), consistent with the findings of the microscopic cellular imaging (Fig. 6a).

Figure 6.

 Role of mitochondrial dysfunction in OGD-induced pre-OL injury. (a, b) OGD caused mitochondrial potential collapse measured by JC-1 either with microscopic cellular imaging (a) or by quantitative radiometric fluorescence assays on a plate reader (b). **p < 0.01, ***p < 0.001 vs. control. (c) OGD toxicity was exacerbated by mitochondrial impairment with FCCP (100 nm) or NaCN (1 µm). Data represent mean ± SEM of three separate experiments. *p < 0.05, **p < 0.01 vs. OGD.

We further examined the role of mitochondrial function in OGD injury by determining whether mitochondrial dysfunction enhances OGD toxicity to pre-OLs. We treated pre-OLs with sublethal doses of the mitochondrial toxin FCCP (100 nm) or NaCN (1 µm) 24 h before exposure of the cells to OGD (2 h), and cell survival was assayed 24 h later. As shown in Fig. 5(a), FCCP (100 nm) or NaCN (1 µm) for 24 h was not toxic to pre-OLs. However, FCCP (100 nm) or NaCN (1 µm) potentiated OGD-induced pre-OL death, and the enhancement in toxicity was prevented by NBQX (50 µm) (Fig. 6c). Taken together, these results indicate that mitochondrial impairment enhances receptor-dependent pre-OL injury during OGD.

Cell death occurs in a non-apoptotic manner

AMPAR activation in pre-OLs causes a robust increase of Ca2+ in mitochondria of pre-OLs, and massive mitochondrial Ca2+ overload often leads to necrotic cell death (Itoh et al. 2000; Itoh et al. 2002; Matute et al. 2002; Deng et al. 2003; Sanchez-Gomez et al. 2003). However, it is not known whether OGD-induced pre-OL death occurs by necrosis or apoptosis. Firstly, control, OGD-treated, and staurosporine-treated pre-OLs were stained with Hoechst 33258 to reveal nuclear fragmentation in situ, and cells were also subjected to TUNEL staining. Cultures at 10 h after a 2-h OGD lacked TUNEL + staining and nuclear fragmentation as compared to the positive-control cultures treated by the apoptosis inducer staurosporine (1 µm) (Fig. 7a). The paucity of TUNEL staining and nuclear fragmentation was also noted in cultures 2, 4, 6, 8 h following OGD (data not shown). Secondly, DNA fragmentation was determined by agarose gel electrophoresis following DNA extraction from cultured pre-OLs. Cultures at 10 h after OGD lacked DNA fragmentation as compared to staurosporine-treated cultures (Fig. 7b). Thirdly, we determined the effect of caspase inhibitors on OGD-induced pre-OL death. The caspase 3 inhibitor acetyl-aspartyl-glutamyl-valyl-aspart-1-aldehyde (DEVD), the caspase 1 inhibitor acetyl-tyrosyl-valyl-alanyl-aspart-1-aldehyde (YVAD), and the pan-caspase inhibitor N-benzyloxycarbonyl-valyl-alanylaspart-(OMe)-fluoromethylketone (ZVAD) (Enzyme Systems Products, Dublin, CA, USA) (each at 50 µm) were all unable to attenuate OGD-induced pre-OL death (Fig. 7c). Thus, cell death is non-apoptotic, as demonstrated by structural, biochemical, and pharmacological criteria.

Figure 7.

 OGD-induced cell death occurs in a non-apoptotic manner. (a) Lack of TUNEL + staining and nuclear fragmentation in OGD-treated cultures compared to staurosporine (1 µm)-treated cultures. (b) Lack of DNA fragmentation in OGD-treated cells compared to staurosporine (1 µm)-treated cells. (c) Caspase inhibitors had no effect on OGD-induced pre-OL injury.

Discussion

The pattern of hypoxic-ischemic brain injury is highly age-dependent. In term infants, the injury predominantly affects cerebral cortex with characteristic neuronal involvement, but in premature infants, the injury selectively affects white matter with prominent OL involvement, as seen in PVL lesions. Pre-OLs are the main cell type that is injured in PVL, and appear to be the major cellular substrate of PVL. Like neurons, these cells are highly vulnerable to hypoxic-ischemic injury mediated by over-activation of GluRs. Unlike neurons, in which glutamate excitotoxicity is mediated primarily by N-methyl-D-aspartate (NMDA) type GluRs, pre-OLs predominantly express non-NMDA type GluRs in vivo and in vitro and are exquisitely vulnerable to excitotoxicity mediated by these receptors (Follett et al. 2000; Deng et al. 2003; Rosenberg et al. 2003). However, the mechanisms of hypoxic-ischemic or OGD-induced pre-OL injury remain unclear. In this paper, we report that (i) OGD does not increase bulk glutamate concentration in pre-OL cultures; (ii) increase in glutamate concentration alone does not cause receptor-mediated toxicity in pre-OLs in normoxic conditions, yet glutamate toxicity can be unmasked by mitochondrial impairment; (iii) OGD causes mitochondrial membrane potential collapse independently of AMPAR activation, and mitochondrial impairment exacerbates OGD toxicity; and (iv) pre-OL death occurs in a non-apoptotic manner during OGD. Overall, these results suggest unexpectedly that OGD-induced pre-OL toxicity is caused by the combination of local glutamate accumulation that occurs without an increase in bulk glutamate concentration and mitochondrial dysfunction. These data provide new insights into excitotoxic mechanisms of pre-OL injury and are relevant to human disorders, such as PVL, in which pre-OL injury plays an important role.

Role of local glutamate accumulation in pre-OL excitotoxicity

Abundant evidence has revealed that brain damage associated with cerebral ischemia and hypoglycemia is in large part the consequence of the extracellular accumulation of excitatory amino acids (Rothman and Olney 1986; Choi 1988; Olney 1989; Choi 1992; Lipton and Rosenberg 1994). Glutamate transporters are the major regulator of extracellular glutamate concentration. Hypoxia-ischemia has been shown to cause glutamate release via reverse glutamate transport in neurons (Benveniste et al. 1984; Perez Velazquez et al. 1997; Izumi et al. 2002), leading to extracellular glutamate accumulation. Using OGD to simulate hypoxia-ischemia in pre-OL cultures, previous studies have demonstrated that OGD causes glutamate release via reverse glutamate transport in pre-OLs (Fern and Moller 2000), resulting in cellular toxicity by activation of Ca2+-permeable AMPARs (Deng et al. 2003). In the present study, we show that experimental manipulations of glutamate diffusion can alter pre-OL excitotoxicity. Facilitation of glutamate diffusion by constant agitation of the culture reduced OGD-induced toxicity, while inhibition of glutamate diffusion by addition of dextran into the medium increased OGD-induced toxicity. These experimental manipulations in the OGD paradigm demonstrated specificity for glutamate because glutamate depletion with a glutamate scavenging system was protective and also implicated a role for AMPARs because NBQX blocked the toxicity induced by medium viscosity increase by addition of dextran. We further demonstrated that OGD-conditioned medium did not transfer toxicity to naïve cells and that OGD toxicity was not accompanied by an increase in bulk glutamate concentration in the medium. Instead, glutamate concentration in the medium appeared to decrease after OGD, presumably as a result of glutamate transport and metabolism under the hypoxic-ischemic conditions. Taken together, our data support the possibility that local glutamate accumulation is responsible for OGD toxicity to pre-OLs, suggesting that devising approaches to inhibit glutamate release or to facilitate glutamate diffusion may be useful in treating disorders associated with hypoxic-ischemic pre-OL injury.

Receptor-dependent and -independent glutamate toxicity

Our results indicate that glutamate is capable of exerting both receptor-dependent and receptor-independent toxicity to pre-OLs. Due to receptor desensitization, glutamate does not cause receptor-dependent toxicity to pre-OLs under normoxic conditions. Receptor-dependent glutamate toxicity is seen when mitochondrial energy metabolism is compromised. In addition, high concentrations of glutamate cause oxidative damage that is not mediated by GluRs. Both receptor-dependent and receptor-independent mechanisms of pre-OL injury may play a role in the pathogenesis of hypoxic-ischemic white matter injury in the developing brain.

Mitochondrial function and pre-OL excitotoxicity

Both brief and prolonged lethal activation of neuronal GluRs involve an abrupt and persistent depolarization of mitochondria, suggesting that early mitochondrial damage is a critical downstream event in excitotoxicity (White and Reynolds 1996). However, our results suggest that OGD can lead to mitochondrial dysfunction independently of AMPAR activation. This mitochondrial impairment may help to enhance pre-OL vulnerability to glutamate excitotoxicity. Mitochondrial dysfunction impairs intracellular Ca2+ handling, and thus in turn aggravates OGD-induced injury to pre-OLs.

A better understanding of the mechanisms of injury to pre-OLs is important in devising potential therapeutic approaches for cerebral white matter disorders, such as PVL (Volpe 2001). OL excitotoxicity may also be involved in the pathogenesis of demyelinating diseases, such as multiple sclerosis (Pitt et al. 2000; Matute et al. 2001), and in white matter injury in stroke (Stys 2004; Liu et al. 2002b). Our study demonstrates a critical role of local glutamate accumulation on OL injury, identifies the aggravating effect of mitochondrial dysfunction on OL excitotoxicity, and thus provides new insights into the pathogenesis of many nervous system diseases associated with OL injury.

Acknowledgements

We thank Hong Wang, Yuming Zhang, Ping Xu, and Ling Dong for technical assistance, and Nikolaus Sucher for suggestions. This work was supported by grants from the National Institute of Health, R01 NS31718 (to F.E.J), P01 NS38475 (to J.J.V., P.A.R and F.E.J), T32 AG00222 (to W.D), the William Randolph Hearst Fund (to W.D), the United Cerebral Palsy Research and Education Foundation (to F.E.J and W.D), and the Mental Retardation Research Center grant P30 HD18655 (to J.J.V).

Ancillary