Overexpression of bcl-x^LProtects Astrocytes from Glucose Deprivation and Is Associated with Higher Glutathione, Ferritin, and Iron Levels 

All chemicals were obtained from Sigma (St. Louis, MO) unless otherwise specified.

Primary astrocyte cultures were prepared from postnatal (days 1–3) Swiss Webster mice (Simonsen, Gilroy, CA) as previously described. 32All procedures were carried out according to a protocol approved by the Stanford University animal care and use committee, in keeping with the National Institutes of Health guide. In brief, neocortices were dissected free of meninges, treated with 0.09% trypsin, triturated and plated as a single-cell suspension in Falcon Primaria 24-well plates (Becton Dickinson, Lincoln, IL) at a density of two hemispheres per multiwell, in Eagle's Minimal Essential Medium (Gibco, Grand Island, NY), supplemented with 10% equine serum (Hyclone, Logan, UT), 10% fetal bovine serum (Hyclone), 21 mM (final concentration) glucose and 10 ng/ml epidermal growth factor. The cultures were infected with retroviral vectors encoding bcl-x  ^L, bcl-2 , or lacZ , the Escherichia coli  gene encoding β-galactosidase and selected with G418 as previously described. 19Cells were used for experiments between days 21 and 45 in vitro . Experiments were performed at least three times on cultures from a minimum of three different dissections.

The mouse bcl-x  ^Lgene was cloned by reverse-transcriptase polymerase chain reaction from Swiss Webster mouse RNA, and both strands were sequenced for verification. The bcl-x  ^Lsequence was then inserted into the LXSN retroviral backbone. 33Retroviruses to express human bcl-2  or E. coli β-galactosidase in the MPZen vector 34were provided by Dr. David Vaux. We have previously found that after antibiotic selection, more than 95% of the astrocytes express the gene of interest. 19 

Immunostaining of cultures for ferritin was accomplished after fixing in methanol:acetone (1:1) at −20°C for 20 min. The cells were air-dried, rinsed in phosphate-buffered saline (PBS), and incubated in a polyclonal antiferritin antibody (Boehringer Mannheim, Indianapolis, IN) diluted 1:100 in PBS containing 5% fetal calf serum and 0.09% sodium azide for 2 h at room temperature. The cells were rinsed in PBS and incubated in a horseradish peroxidase-linked antirabbit antibody (Amersham, Arlington Heights, IL) diluted 1:100, for 1 h. After rinsing in PBS, antibody binding was visualized by exposing to diaminobenzidine and urea H²O²for 5 min. No staining was observed if the primary antibody was omitted. Counterstaining was with Gill no. 3 hematoxylin. Astrocytes expressing β-galactosidase were stained with X-gal (Molecular Probes, Eugene, OR).

Expression of ferritin or bcl-x  ^Lwas estimated by immunoblotting. Equal amounts of protein as determined by the bicinchoninic acid method (Pierce, Rockford, IL), 30 μg per condition, were separated on a 12.5% polyacrylamide gel and electrotransferred onto Immobilon polyvinylidene fluoride membrane (Millipore Corp., Bedford, MA). Gels and membranes were stained with 0.15% Coomassie blue and Ponceau S red solutions respectively, to ensure equal loading and transfer of protein. Membranes were blocked with 5% nonfat dry milk in water with 0.1% Tween 20 and 0.02% sodium azide overnight, incubated for 2 h with 1:1000 polyclonal rabbit antihuman ferritin antibody (Boerhringer Mannheim) or 1:1000 polyclonal rabbit anti–bcl-x  (PharminGen, San Diego, CA), washed with 1% bovine serum albumin in PBS and incubated with 1:2000 horseradish peroxidase linked antirabbit antibody (Amersham), 1% bovine serum albumin in PBS, for 1 h. Immunoreactive bands were visualized with the enhanced chemiluminescence detection system (Amersham). After visualizing bcl-x  the blot was striped in distilled H²O for 10 min, 0.2% NaOH for 5 min, and H²O for 5 min; blocked overnight in nonfat milk; and reprobed using 1:1,000 goat polyclonal anti-actin antibody (Santa Cruz Biotechnology, Santa Cruz, CA) in 1% bovine serum albumin. The membrane was washed and then incubated with 1:3,000 horseradish peroxidase linked antigoat antibody (Santa Cruz Biotechnology) for 1 h followed by enhanced chemiluminescence detection. Densitometric quantitation was performed using a Gel Doc 1000 densitometer (BioRad Laboratories, Hercules, CA) with Molecular Analyst software.

The method of Hill and Switzer 35was used with modifications. 36Cells were fixed, stained in Perls’ solution, and counterstained with Gill no. 3 hematoxylin.

Total cellular iron content in the cultures was measured 37about 25 days after infection with the retroviral vectors. The cells were washed three times with 0.9% NaCl, then harvested in 1% sodium dodecyl sulfate, 50 mM Tris (p  H 7.5). The protein content was determined using bicinchoninic acid, and 300 μg total protein in 1 ml was assayed for each condition, in triplicate. The colorimetric reaction was initiated by addition of 0.5 ml 0.6 N HCl, 2.25% KMnO⁴for 2 h at 60°C, followed by 0.1 ml reagent B(ferrozine 3.25 μM, neocuproine 6.54 μM, L-ascorbic acid 1 M, and ammonium acetate 2.54 M). Absorbance at 562 nm was measured after 30 min at room temperature. Iron standards were prepared using iron (II) ethylenediammonium sulfate (Alfa ASAR, Ward Hill, MA). The level of iron was calculated from a standard curve and expressed in nanomoles per milligram protein.

Injury paradigms were carried out as previously described. 19 

Thirty percent H²O²was diluted in balanced salt solution (BSS^5.5) at p  H 7.4, containing (in mM) glucose 5.5, NaCl 116, CaCl²1.8, MgSO⁴0.8, KCl 5.4, NaH²PO⁴1, NaHCO³14.7, and HEPES 10, and phenol red 10 mg/l to a final concentration of 400 μM.

Cultures were deprived of all substrate by replacing the culture medium with BSS⁰, identical to BSS^5.5but lacking glucose.

Cultures were transferred to an anaerobic chamber (Forma Scientific, Marietta, OH) with an atmosphere of 5% CO², 10% H², and 85% N². The culture medium was replaced with warmed, deoxygenated BSS⁰. Oxygen tension was < 0.2%, monitored with an oxygen electrode (Microelectrodes, Bedford, NH). Oxygen glucose deprivation (OGD) was ended by adding glucose to a final concentration of 5.5 mM and returning the cultures to the normoxic incubator. The medium was sampled at the end of OGD and 24 h later to measure lactate dehydrogenase (LDH).

Astrocyte injury was evaluated morphologically by phase-contrast light microscopy and was quantitated by measuring the activity of LDH released into the culture medium. 32LDH release is appropriate in assessing these injuries, as they appear necrotic in nature, lacking evidence of nuclear fragmentation or DNA laddering (data not shown). Total LDH release corresponding to complete astrocyte death was determined at the end of each experiment following freezing at −70°C and rapid thawing.

Reduced GSH was measured before and after glucose deprivation (GD) for 7 h as described previously 38using a modification of the method of Fernandez-Checa and Kaplowitz. 39The concentration of GSH was calculated from standard curves and expressed as nanomoles per milligram protein. Preincubation with the GSH synthesis inhibitor L-buthionine sulfoximine (BSO; 100 μM) was for 12 h in growth medium. GSH was then measured immediately, or the astrocytes were subjected to GD in BSS⁰without added BSO. LDH release was measured after 28 h.

Loss of cis-parinaric fluorescence, as an indicator of lipid peroxidation, has been shown to correlate with the generation of thiobarbituric acid reactive substances. 40The cis-parinaric assay was performed on astrocyte suspensions as previously described. 38,40,41 

We assayed the superoxide anion in two ways. (1) The colorimetric assay using iodonitrotetrazolium violet, described by Bagchi et al. 42was performed as previously described. 19(2) Dihydroethidium (Molecular Probes) was also used to determine relative superoxide levels according to the method of Bindokas et al.  43The probe is oxidized by superoxide to ethidium. Dihydroethidium was prepared as a 3.2 mM stock in dimethylsulfoxide stored under N²at −20°C. The cells were washed with BSS^5.5twice. The dihydroethidium (6 μM) was added in the dark. The cells were suspended after incubating 1 h in the dark. The fluorescence of ethidium was measured at 510 nm excitation, 590 nm emission.

Differences between groups underwent analysis of variance followed by the Student–Newman–Keuls test for multiple comparisons, or t  test if there were only two conditions, using Sigmastat (SPSS, Chicago, IL). Values given are mean ± SEM, with the number of cultures tested indicated in the figures.

Transfection of primary cultured astrocytes with the mouse bcl-x  ^Lgene using a retroviral vector resulted in markedly higher levels of expression of this protein (fig. 1) compared with the level seen in astrocytes expressing β-galactosidase or mock-infected cells. Astrocytes overexpressing bcl-x  ^Lwere significantly protected from H²O²(400 μM) injury (fig. 2) as assessed morphologically and by LDH release (fig. 3). Four hours after addition of 400 μM H²O², astrocytes overexpressing bcl-x  ^Lsuffered only half the cell death seen in β-galactosidase–expressing controls, and protection was stable at 24 h. Because we have found the half-life of peroxide in astrocyte cultures to be about 15 min, with the release of LDH occurring primarily in the first 5 h, 37the protection seen with bcl-x  ^Lappears to be long-lasting. In contrast, if exposed to OGD, bcl-x  ^L–overexpressing astrocytes were less injured than β-galactosidase–expressing controls only acutely at the end of a mild insult, 6 h of OGD, but this protection was lost after 24 h of reoxygenation and refeeding (table 1). Similarly, if exposed to a more severe insult, 8 h OGD, no protection was seen immediately at the end of the period of OGD or 24 h later.

Protection from GD alone was significant. After 24 h deprivation bcl-x  ^Lastrocytes suffered 22 ± 6% death compared with 81 ± 6% in β-galactosidase–expressing controls (P < 0.01;fig. 4A). Injury in bcl-x  ^Loverexpressing cells was still less even at 30 h. The level of GSH was significantly higher in the astrocytes that overexpress bcl-x  ^Lcompared with mock-infected or β-galactosidase–expressing controls (P < 0.01;table 2) under normal growth conditions. After GD for 7 h, the level of GSH decreased in each group but remained higher in bcl-x  ^Lthan β-galactosidase–expressing or mock-infected controls (table 2). To determine the importance of this elevation of GSH to protection by bcl-x  ^Lwe reduced the levels of GSH in all groups so that they were equal prior to starting the injury paradigm. Preincubation with BSO (100 μM) for 12 h in growth medium resulted in equivalent, reduced GSH levels in bcl-x  ^L–expressing astrocytes and controls (table 2). Astrocyte cultures preincubated with or without BSO for 12 h in growth medium were then deprived of glucose. Although the degree of protection was reduced compared with the extent of protection without BSO pretreatment, the bcl-x  ^L–overexpressing cells still suffered significantly less injury than controls (fig. 4B).

The level of superoxide in the bcl-x  ^L–overexpressing cells was higher than in controls under normal growth conditions, whether determined by reaction with iodonitrotetrazolium violet (fig. 5) or hydroethidium (data not shown). After GD for 7 h, the levels of superoxide were decreased to about 40% of the level seen in growth medium, but the level in bcl-x  ^L–overexpressing cells was still higher than in β-galactosidase expressing (1.6 times) or mock-infected control (two times) astrocytes (fig. 5). Levels of lipid peroxidation were assessed with cis-parinaric acid and found to be similar in all three groups of astrocytes under normal growth conditions (data not shown).

bcl-x  ^L–overexpressing astrocytes (fig. 6A) showed greater staining for iron than β-galactosidase–expressing cells (fig. 6B) or mock-infected cells (fig. 6C). Perl's reagent revealed a granular distribution of iron staining. Colorimetric determination of iron content showed significantly higher levels in bcl-x  ^Land bcl-2  infected astrocytes compared with β-galactosidase expressing (twofold) and mock-infected controls (fourfold;table 3).

In situ  immunohistochemistry showed greater staining for ferritin in bcl-x  ^L–overexpressing astrocytes (fig. 6D) than in β-galactosidase–expressing (fig. 6E) or mock-infected controls (fig. 6F). Semiquantitative assessment of ferritin by immunoblot showed two bands: the heavy (H-) subunit (about 22 kDa) and the light (L-) subunit (about 19 kDa). Increased amounts of both H- and L-subunits were seen in both bcl-2 – and bcl-x  ^L–overexpressing cells (fig. 7); control mock-infected cells expressed little of the L-subunit. The level of H-subunit in bcl-x  ^L–overexpressing cells was about three times that in β-galactosidase–expressing or mock-infected controls by densitometry (fig. 7A); the L-subunit increased about sixfold with bcl-x  ^Loverexpression or more than threefold with bcl-2  overexpression.

There are similarities between bcl-x  ^Land bcl-2  in structure, intracellular localization, and effect. 26,44,45However, bcl-x  ^Land bcl-2  also differ. 9–11The former has been suggested to be the predominant protective protein in gerbil hippocampus. 18We investigated the ability of bcl-x  ^Lto protect primary astrocytes in culture from ischemia-like injury. It provided robust protection of primary astrocytes from H²O²exposure and GD but was relatively ineffective against combined OGD. This protection is associated with increases in at least two aspects of cellular antioxidant defense: increased glutathione levels and increased ferritin levels. The increased levels of GSH were found to contribute to the extent of protection from glucose deprivation, because pretreatment with BSO reduced the extent of protection about 50%. This is consistent with prior observations that hypoglycemia represents an oxidative stress, 38which is worsened by reduction of GSH levels and ameliorated by artificially elevating GSH levels. 38OGD is not worsened by reducing GSH levels. 19GSH is an important hydrophilic-free radical scavenger that also participates in enzymatic free-radical scavenging. Although several reports document increased GSH levels in neural cells and fibroblasts overexpressing bcl-2 , 21,46there are fewer data on the effects of bcl-x  ^Lon GSH levels. Two recent studies in lymphocytes report that bcl-x  ^Lcan preserve GSH levels and intracellular thiols and block induction of apoptosis. 47,48 

There is no evidence that either bcl-2  or bcl-x  ^Lacts directly as an antioxidant. The mechanism by which they influence levels of reactive oxygen intermediates and endogenous cellular antioxidants is not yet understood. Superoxide is primarily a byproduct of mitochondrial function, and bcl-2  and bcl-x  ^Laffect mitochondrial function under stress conditions. 26,49The possibility that bcl-2  has a prooxidant effect that induces a cellular antioxidant response was first suggested by Steinman 20and is consistent with our prior observations on bcl-2  overexpression in astrocytes. 19We report here that bcl-x  ^Lincreases superoxide levels yet is associated with improved antioxidant defense—higher levels of ferritin and GSH. Both of these antioxidants are induced by oxidative stress, so the observed increases if bcl-2  or bcl-x  ^Lis overexpressed imply that there may be some level of oxidative stimulus leading to up-regulation under normal growth conditions. That the potential oxidative stress is compensated for is suggested by the inability to detect increased lipid peroxidation. Although no assay is completely specific, we used two different assays to confirm a change in superoxide. The dihydroethidium assay has been shown by Bindokas et al.  not to be affected by many other common oxidants including ·OH, NO, ⁻ONOO, H²O², hypochlorite, or singlet O². 43 

Although superoxide levels were increased, the superoxide radical is generally considered harmful if metal ions, such as Fe³⁺, are reduced by superoxide and then catalyze formation of hydroxyl radicals from H²O². 50The increased level of ferritin, which binds Fe³⁺, observed with bcl-x  ^Lmay prevent increased hydroxyl radical production despite the higher superoxide levels. That the cells are not suffering greater oxidative stress is suggested by the observed lack of difference in the extent of lipid peroxidation. This finding is reminiscent of the results of Hockenbery et al. , 25who found no lipid peroxidation in bcl-2 –overexpressing cells in the absence of exogenous stress.

Although the precise mechanism of action of bcl-2 –related proteins is still not known, recent studies have focused on the role of these proteins in mitochondrial function, 49,51their ability to associate in protein complexes, and their ability to form pores in membranes. 1Several reports suggest that the relative levels of expression of the pro- and antiapoptotic members of the bcl-2  gene family may play an important role in determining the sensitivity of the brain to ischemia. Many of these genes can homodimerize and heterodimerize with each other. The relative levels of expression of these proteins may determine whether a cell undergoes apoptosis or not. Postischemic alterations in expression, relatively increased levels of bax in cells that go on to die, and increased levels of bcl-2  or bcl-x  ^Lin cells that survive have been reported in models of both global 14,22,23,52and focal ischemia. 7,8,53 

We report here the novel observation that overexpression of bcl-x  ^Land bcl-2  is associated with elevated ferritin levels. Brain ferritin is composed of H- and L-subunits. 54The ferritin observed in our primary cultured astrocytes was predominantly the H-subunit. With bcl-x  ^Lor bcl-2  overexpression the amount of ferritin in the cells increased. In addition to higher levels of the H-subunit, significant expression of the L-subunit was found. The roles suggested for H-subunit rich ferritin are as a stress or early-response protein allowing rapid iron binding, as an antioxidant, 55and in cell proliferation. 56In contrast, L-subunit–rich ferritin, such as is found in liver, is thought to be primarily a storage protein. 57Both subunits show largely post-transcriptional control. However, after iron administration, both subunits show transcriptional induction, with the L-subunit having a greater response. 54The observed accumulation of the L subunit may reflect the higher levels of iron seen in bcl-x  ^L– and bcl-2 –overexpressing astrocytes. The H-subunit has ferroxidase activity 57and has been found to be upregulated in oligodendrocytes after hypoxia, through what is thought to be an oxidative stress-induction mechanism. 58It is possible that the increased levels of superoxide found in astrocytes overexpressing bcl-x  ^Lor bcl-2  induce ferritin H, which allows the cells to capture more iron and subsequently leads to induction of ferritin L. Further studies are required to determine the validity of this hypothesis.

Iron has been detected by histochemical staining in many cell types of the CNS, including neurons, microglia, oligodendroglia, and astrocytes. A large proportion of the soluble brain iron was shown to be present in ferritin. 57Iron can catalyze oxidant-induced cell injury, and iron can be released from ferritin by acid and reduction. 59Ferrous iron in the presence of oxygen can catalyze the production of active oxygen species. 50Recent findings indicate that an increase in ferritin can compensate for increased iron levels and prevent accumulation of protein carbonyls, evidence of protein oxidation. 60By sequestering iron, ferritin may prevent iron-catalyzed lipid peroxidation injury. Although excess iron can be deleterious, a certain level of iron is essential for cell growth and viability. Iron depletion was found to induce apoptosis in human leukemic cells. 61Ferritin may both provide the iron required to sustain oxidative metabolism and protect the cell from oxidant mediated injury. 60 

The authors thank Xiao Yun Sun for expert technical assistance and Dr. David Vaux for the gift of bcl-2  and lacZ  retrovirus.