Hypoxic reduction in cellular glutathione levels requires mitochondrial reactive oxygen species

Kyle D. Mansfield, M. Celeste Simon, Brian Keith


When exposed to hypoxia (1.5% O2), several cell types have been shown to increase production of reactive O2 species derived from the mitochondrial electron transport chain (mtROS). The general physiological consequences of hypoxic mtROS production are not completely understood, although several groups have demonstrated that mtROS promote the stabilization and activity of hypoxia inducible factor-1α (HIF-1α) transcription factor, alter cardiac myocyte contractility, and modulate Na+-K+-ATPase activity. To investigate the effects of hypoxia-induced mtROS on general cellular oxidative metabolism, we measured the levels of glutathione, a major cellular antioxidant, in response to hypoxic treatment. Our data indicate that HEK293 and Hep3B cells exposed to 1.5% O2 exhibit a time-dependent decrease in cellular glutathione stores and concomitant inhibition of glutathione biosynthesis, which correlates to impaired transport of the substrate cystine. Using a combination of ROS scavengers, mitochondrial electron transport inhibitors, and mitochondrial DNA-deficient ρ0 cells, we demonstrate that this decrease in cellular glutathione levels is mediated by hypoxia-induced mtROS. Intriguingly, this effect is also inhibited by cyclohexamide but is not dependent on HIF-mediated transcription. Overall, these data suggest a novel HIF-independent role for mitochondrial ROS in regulating glutathione synthesis, and hence cellular oxidative homeostasis, during hypoxic exposure.

  • hypoxia
  • mitochondria
  • hypoxia-inducible factor-1α
  • cystine

the ability to adapt to changes in O2 concentration is essential for normal development, as well as for recovery from ischemic conditions (1, 39, 45, 47). Multicellular organisms have evolved complex mechanisms to perceive changes in O2 tension and mount appropriate adaptive responses (9, 43). Although much has been learned in recent years about transcriptional responses to hypoxia (40, 44, 58, 60), the consequences of physiological hypoxia on general intracellular reduction-oxidation (redox) status is not completely understood.

Several studies have demonstrated that cells exposed to hypoxic conditions (e.g., 1.5% O2) produce increased levels of reactive O2 species (ROS) derived from the mitochondrial electron transport chain (mtROS) (12, 13, 18, 19, 37, 55). These mtROS have been implicated in a number of cellular processes, including altered transcriptional activity (2, 12), myocardiocyte contraction (55), and Na+-K+-ATPase activity (15). Reports from several laboratories have demonstrated that mtROS can regulate hypoxia-inducible factor (HIF), a transcriptional activator composed of an α- and a β-subunit that regulates genes involved in cellular and systemic responses to O2 deprivation. Generation of mtROS under hypoxic conditions can lead to the stabilization of the regulatory HIF-α subunits (2, 3, 12, 13, 42), although these results are controversial (48, 54) and may be influenced by cell type and absolute O2 concentration (see discussion). Several groups have also presented evidence that the redox-regulated proteins Ref-1 and thioredoxin regulate HIF activity (11, 24, 33, 56, 57).

The generation of mtROS under hypoxia is likely to alter intracellular redox status: specifically, mtROS production may affect the level or function of molecules that maintain the reducing environment of the cytosol. The tripeptide glutathione (Glu-Gly-Cys) is present in millimolar concentrations in cells, and the thiol group of the cysteine residue acts as an important reductant in a large number of biochemical reactions (17, 23, 34). Glutathione can detoxify xenobiotics through irreversible conjugation and can protect protein structure and function through reversible thiol-disulfide exchange (31). Glutathione also acts as an intracellular redox buffer; for example, intracellular hydrogen peroxide (H2O2) formed under oxidative stress is reduced by glutathione peroxidase with concomitant conversion of reduced glutathione (GSH) to the oxidized form (GSSG) (17). This effectively detoxifies H2O2 and protects the cell from oxidative damage. Typically, GSSG is then converted back to GSH via glutathione reductase or is effluxed from the cell and degraded by extracellular γ-glutamyl transpeptidase (16, 17). The oxidation state and intracellular concentration of glutathione are therefore useful indicators of oxidative stress in cells.

To address the connection between hypoxia, mtROS, and cellular redox status, we measured the levels of the antioxidant glutathione as a function of mitochondrial respiration in HEK293 and Hep3B cells. Whereas a few previous studies reported that glutathione levels are lowered by hypoxia in certain cell types (10, 36, 38), the underlying molecular mechanisms leading to these changes was not determined. In this report, we demonstrate that this decrease is dependent on the generation of mtROS and correlates to impaired cystine uptake and inhibition of glutathione synthesis. Moreover, treatment with cyclohexamide blocks the drop in glutathione levels, indicating that this response requires de novo protein synthesis and may be part of a long-term adaptive response to hypoxia. These data describe a novel HIF-independent molecular response regulated by hypoxia-induced mtROS.


Tissue culture cells and reagents.

HEK293 and Hep3B cell lines were maintained in DMEM (CellGro), 10% fetal bovine serum (GemCell), 2 mM glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin, and 1 mM sodium pyruvate. Hypoxia was defined as 1.5% O2. For glutathione and Western blot assays, hypoxia was maintained by use of an In Vivo 2 hypoxic workstation (Ruskinn Technologies, Leeds, UK), whereas reporter assays were carried out in an IG750 variable O2 tissue culture incubator (Jouan). All pharmacological reagents were purchased from Sigma Chemical. Cell cultures were pretreated with the indicated drug concentrations for ∼15 min before hypoxic treatment.

Rho zero (ρo) generation.

To generate rho zero (ρo) cells, HEK293 or Hep3B cells were treated with ethidium bromide (50 or 100 ng/ml) in complete DMEM supplemented with 0.1 mM MEM nonessential amino acids (GIBCO), 2.5 mM sodium pyruvate, and 50 μg/ml uridine (30). Cultures were maintained for 2–6 wk, during which time ρo status was monitored by cytochrome oxidase subunit II (COXII) PCR or Southern blot and respirometry. Cells were considered ρo when there was no detectable COXII by PCR or Southern blot, and, more importantly, no mitochondrial-dependent O2 consumption. The O2 consumption rate was measured in a water-jacketed respirometer chamber by using a polarographic O2 electrode (Cameron Instrument, Port Arkansas, TX) at 37°C.

Glutathione assay.

Total cellular glutathione was measured via a modified Tietze recycling method adapted for microplate readers (5). Cells were plated at 0.5 × 106/well of a six-well plate and allowed to recover overnight. After appropriate treatments, cells were washed in PBS and then lysed in 100–150 μl of buffer (100 mM NaPO4, 1 mM EDTA, pH 7.5) containing 0.1% Triton X-100 and frozen at −80°C until analysis. After cells were thawed, cellular debris was removed via centrifugation for 10 min at maximum speed in a microcentrifuge. To measure total glutathione, proteins were precipitated with sulfosalicylic acid at a final concentration of 1%. Samples were then spun for 10 min in a microcentrifuge to pellet proteins, and supernatant was diluted 1:20 in buffer before being measured. To assay GSSG, supernatants were diluted 1:2 with buffer, and 3 μl of 2-vinylpyridine were added to 100 μl of sample and incubated for 15 min at room temperature to conjugate GSH. For all measurements, 50-μl triplicates of each sample were used for glutathione determination. The reaction was started by adding 100 μl of reaction mixture [0.3 mM 5,5′-dithio-bis(2-nitrobenzoic acid), 0.2 mM NADPH, 1 mM EDTA, and 2 U/ml GSH reductase in a 100 mM sodium phosphate buffer (pH 7.5)] to 50 μl of sample. The change in absorbance at 412 nm was then monitored for 5 min with a Spectra Max 190 microplate reader (Molecular Devices, Sunnyvale, CA), and the glutathione concentration was determined by comparing the rate of color change with that of a GSSG standard curve. Samples were normalized to protein concentration determined by the Pierce BCA assay.

Western blots.

Whole cell lysates were either used directly from glutathione samples or prepared separately in whole cell extract buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 5 mM EDTA, 0.1% SDS, 1 mM phenazine methosulfate, and complete protease inhibitor; Roche Molecular Biochemicals). Equal amounts of protein were electrophoresed on an acrylamide gel, transferred to nitrocellulose (Bio-Rad), and immunoblotted according to standard protocols using 5% nonfat dry milk in Tris-buffered saline with 0.1% Tween 20. Blots were stained with Ponceau S to ensure equal loading. α-Human HIF-1α (Transduction Laboratories; 610959) and α-mouse HIF-1α (Novus; NB100–105) were both used at 1:250 dilution and detected by a 1:2,000 dilution of horseradish peroxidase-conjugated α-mouse secondary antibody (Cell Signaling Technologies) followed by enhanced chemiluminescence detection (Amersham). α-Glutamylcysteine synthase (GCS) heavy subunit (Neomarkers) was used at 1:500 dilution and detected by a 1:2,000 dilution of horseradish peroxidase-conjugated α-rabbit secondary antibody (Cell Signaling Technologies) followed by enhanced chemiluminescence detection.

Reporter assay.

The wild-type hypoxia response element (HRE)-luciferase reporter constructs consisted of a trimerized 24-mer containing 18 bp of sequence from the phosphoglycerate kinase (PGK) promoter including the HRE (5′-tgtcacgtcctgcacgactctagt, HRE underlined) and an 8-bp linker sequence followed by a 50-bp minimal tyrosine kinase promoter in a pGL2-basic backbone (Promega). The mutant HRE had the ACG of the HIF-1 binding site mutated to CAT, abolishing binding, as well as a point mutation that abolished a BsgI restriction site for diagnostic purposes.

Cells were transfected with either wild-type or mutant HRE and TK luciferase constructs by use of Fugene and were incubated overnight. Cells were then replated into six-well plates and allowed to recover overnight. The following day, plates were treated and placed under hypoxia for 8 h or overnight. Cells were then lysed and analyzed via the dual luciferase assay kit (Promega) according to the manufacturer's instructions.

ROS measurements.

Cells were plated at 0.5 × 106/well of a six-well plate and allowed to recover overnight. Cells were washed in PBS and then treated with 1 ml of PBS containing 20 μM 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (carboxy-H2DCFDA), in the presence or absence of 50 μg/ml cyclohexamide. After 15 min of pretreatment, cells were exposed to normoxia or hypoxia for 6 h. Cells were washed from the plate in the incubation medium, and 300 μl were transferred to a black opaque 96-well plate. Samples were excited at 488 nm, and fluorescence emission was measured at 530 nm by use of a Fluoromax-2 fluorometer equipped with a microplate reader. Hypoxic samples were harvested under hypoxic conditions in the Ruskinn In Vivo2 hood, and the plate was kept sealed before being read.

Glutathione synthesis rates.

Glutathione synthesis rates were measured as previously described (46). Briefly, cells were plated as in the standard glutathione assay described above. After appropriate treatments, cells were then treated with 5 mM diethyl maleate for 30 min to deplete cellular glutathione stores. The cells were then washed two times with PBS and incubated at 37°C under normoxic conditions for the times indicated. Cells were harvested at the various time points and analyzed for glutathione concentration as described above.

Northern blot analysis.

HEK293 or Hep3B cells were plated at 2.5 × 106 cells/10-cm dish and allowed to recover overnight. After 6 h of treatment, total RNA was extracted by use of TRIzol reagent (Life Technologies). RNA immobilized on nylon membranes (Amersham) was hybridized with 32P-labeled probe specific for GCS light subunit as previously described (14) and was detected by phosphoimaging.

Cystine uptake assay.

Cystine uptake rates were measured as previously described (6). Cells were plated 1 × 106 cells/well of a six-well culture dish and allowed to recover overnight. After appropriate treatments, the plates were washed two times with PBS. They were then incubated in 1 ml of uptake medium (PBS with Ca, Mg, and 0.1% glucose) with 2 μCi l-[35S]cystine (Amersham) for 2 min at 37°C. The uptake was stopped by rinsing the cells three times with ice-cold PBS. The cells were then lysed in 0.5 N NaOH, and the amount of radioactivity taken up was determined by scintillation counting.

Data and statistical analysis.

Unless stated otherwise in the text, all data shown are representative examples of three to five separate experiments, which all gave similar results. All glutathione values were corrected for protein content and then normalized to the normoxic control levels (HEK293 = 57.8 ± 13 pmol/μg protein; Hep3B = 185 ± 24 pmol/μg protein). Values shown are means ± SE of three separate samples. Statistically significant differences between samples were determined by one-way ANOVA followed by a Bonferroni multiple-comparisons test.


Effect of hypoxia on cellular glutathione.

To determine the effects of hypoxia on intracellular glutathione pools, cells were exposed to hypoxic conditions for 0–8 h, and the amount of glutathione in total cellular extracts was determined. As shown in Fig. 1, exposure of HEK293 or Hep3B cells to 1.5% O2 caused a time-dependent decrease in total cellular glutathione, with HEK293 cells exhibiting a decrease of almost 50% by 6 h. The change in HEK293 glutathione concentration was significant after 2 h and continued in a linear fashion (R2 = 0.98; P < 0.05). Whereas total cellular glutathione decreased dramatically, the GSH-to-GSSG ratio did not change significantly over the same time course (Hep3B: normoxia = 20.54, hypoxia = 19.89; HEK293: normoxia = 780, hypoxia = 881). In contrast, HEK293 cells altered their GSH-GSSG ratio in response to treatment with exogenous H2O2 (1 mM) with an almost 10-fold change (HEK293 = 78) within 30 min.

Fig. 1.

Hypoxia causes a time-dependent decrease in total cellular glutathione (GSX). HEK293 (293) and Hep3B (H3B) cells were exposed to 1.5% O2 for the indicated times and subsequently analyzed for total cellular glutathione and normalized to protein concentration. Values are expressed as a percentage of the normoxic levels. ***P < 0.05.

Role of mtROS in hypoxic glutathione decrease.

ROS are generated from complex III of the mitochondrial electron transport chain when molecular O2 gains an electron from ubisemiquinone (Fig. 2A; Refs. 8, 52). It was recently shown that hypoxic treatment increases production of these mtROS, which in turn activate downstream effects including stabilization of HIF-1α subunits. We therefore examined the role of mtROS in hypoxia-mediated changes in glutathione levels. As shown in Fig. 2B, treatment with 100 μM of the antioxidant pyrrolidinedithiocarbamate (PDTC) inhibited the decrease in glutathione after 6 h of hypoxic exposure. Similar results were obtained with the antioxidant ebselen. In these experiments, 6 h of hypoxia treatment lowered HEK293 cellular glutathione levels to 65 ± 9% of normoxic values, whereas treatment with 20 μM ebselen maintained hypoxic glutathione levels at 94 ± 6% of the normoxic values.

Fig. 2.

Mitochondrial reactive O2 species (mtROS) are responsible for the hypoxic glutathione decrease. A: outline of the electron flow through the electron transport chain. Boxes indicate the sites of action of the various inhibitors used in this study. PDTC, pyrrolidinedithiocarbamate; ROS, reactive O2 species; Cyt C, cytochrome c. B: HEK293 cells were pretreated for 15 min with 100 μM PDTC and then exposed to 1 μg/ml antimycin A (Anti A or AA) or 1.5% O2 for 6 h and analyzed for total cellular glutathione. Values are expressed as a percentage of the untreated normoxic control (Ctrl). Hyp, hypoxia. C: HEK293 cells were pretreated for 15 min with 0.01 to 1 μg/ml rotenone and then exposed to 6 h of 1.5% O2. Norm, normoxia. ***P < 0.05 vs. normoxic control, +++P < 0.05 vs. hypoxia, ###P < 0.05 vs. antimycin A.

Antimycin A, a mitochondrial inhibitor that prevents transfer of electrons from ubisemiquinone to ubiquinone (Fig. 2A), was used to generate mtROS independently of hypoxia, as previously described (12). Treatment of HEK293 with 1 μg/ml antimycin A for 6 h decreased glutathione levels to a similar extent as hypoxia. PDTC also prevented the antimycin A-mediated decrease (Fig. 2B), providing evidence that the reduction in glutathione was a result of mtROS generation and not simply due to decreased ATP production resulting from electron transport dysfunction. Cells were also treated with rotenone, a mitochondrial inhibitor that blocks the flow of electrons from complex I, disrupting the electron transport chain upstream of ubisemiquinone, therefore blocking hypoxic mtROS production (Fig. 2A) (12). Treatment of hypoxic HEK293 cells with 10 ng/ml to 1 μg/ml rotenone prevented the O2-dependent decrease in glutathione (Fig. 2C). To ensure that the effects seen were due to the specific effects of rotenone on mitochondrial electron transport, HEK293 cells were also treated with the complex III inhibitor myxothiazol (Fig. 2A). In these experiments, hypoxia reduced glutathione levels to 70 ± 9% of normoxic control values after 6 h, whereas treatment with 100 ng/ml myxothiazol completely prevented the decrease (106 ± 4%).

Cells can also be rendered deficient in mitochondrial electron transport by depleting mitochondrial DNA (1). These ρ0 cells are selected by prolonged culture in low levels of ethidium bromide and are dependent on pyruvate and uridine for growth. Because they lack mitochondrial DNA, including the COXII gene encoding the cytochrome oxidase subunit II, these cells display defective electron transport. We generated ρ0 HEK293 cells by culturing them in 50 or 100 ng/ml ethidium bromide for 3 wk, after which we observed a dramatic loss of COXII DNA as determined by both PCR and Southern blot analysis (Fig. 3A). The loss of COXII DNA correlated with a substantial decrease in respiration, as measured by cellular O2 consumption (Fig. 3B). The small amount of residual O2 consumption was most likely due to nonrespiratory activities, because it was insensitive to treatment with cyanide (data not shown).

Fig. 3.

Rho zero (ρ0) cells fail to decrease glutathione under hypoxia. After HEK293 cells were cultured in 0, 50, or 100 ng/ml ethidium bromide (EtBr) for 3 wk, the presence of mitochondrial DNA (A) and respiration (B) were assessed. For respiration (B), 2.5 million cells were analyzed for O2 consumption. Whole cell DNA was then extracted from the cells, 10 μg were analyzed for cytochrome oxidase DNA content via Southern blot (A, top blot), and 100 ng were assessed via PCR (A, middle blot). HEK293 (C) and Hep3B (D) ρ0 cells were then exposed to 6 h of 1.5% O2 and analyzed for total cellular glutathione. ***P < 0.05 vs. respective normoxic control.

Once the ρ0 status of the cells was confirmed, we tested the effect of hypoxia on glutathione levels. As shown in Fig. 3C, ρ0 HEK293 cells exhibited an attenuated hypoxia-mediated drop in glutathione levels. Specifically, wild-type cells displayed ∼40% less intracellular glutathione when exposed to 6 h of hypoxia, whereas ρ0 (50 ng/ml ethidium bromide) cells displayed a 25% drop, and ρ0 (100 ng/ml ethidium bromide) showed no significant change. Similar results were seen with Hep3B ρ0 cells (Fig. 3D). Together, these results argue strongly that the hypoxic decrease in glutathione levels is dependent on mtROS generated from the electron transport chain.

Effect of lowering glutathione levels on HIF stabilization and transcriptional activity.

We reasoned that a substantial drop in cellular glutathione concentration might have effects on other redox-sensitive processes, including stabilization of HIF-1α. In a previous report, cells treated with antimycin A increased the production of mtROS, although the level was insufficient to stabilize HIF-1α under normoxic conditions. However, pretreatment with the glutathione synthesis inhibitor dl-buthionine-[S,R]-sulfoximine (BSO) potentiated the effects of antimycin A and stabilized HIF-1α protein (13). These results raise the possibility that decreased glutathione levels may be a critical downstream mediator of hypoxic mtROS signaling and that this reduction in glutathione levels may directly or indirectly contribute to HIF-1α stabilization in hypoxic cells. To determine whether reduced intracellular glutathione is sufficient to stabilize HIF-1α in the absence of mtROS production, we treated hypoxic HEK293 ρ0 cells with BSO. As shown in Fig. 3C, control HEK293 ρ0 cells failed to lower glutathione levels when placed at 1.5% O2; moreover, these cells did not stabilize HIF-1α protein (Fig. 4A), as expected. In contrast, treatment of ρ0 cells with 1 mM BSO under hypoxia effectively decreased glutathione levels by almost 95% (hypoxia = 31.2 ± 2.2 pmol/μg protein; hypoxia + BSO = 2.0 ± 0.9 pmol/μg protein) but did not restore hypoxic HIF-1α stabilization (Fig. 4A). These cells did, however, retain the ability to stabilize HIF-1α when incubated with the iron chelator desferrioxamine, which directly inhibits HIF-1α turnover, presumably by inhibiting the prolyl hydroxylases that target HIF-1α for degradation (25, 26).

Fig. 4.

Modulation of cellular glutathione does not affect hypoxia-inducible factor (HIF). A: 25 μg of whole cell extracts from HEK293 ρ0 cells were probed for HIF-1α levels via Western blot after 4 h of treatment with 1.5% O2, 1 mM dl-buthionine-[S,R]-sulfoximine (BSO), or 100 μM desferrioxamine (DFX). A nonspecific (NS), Ponceau S-stained band was used as a loading control. B: HEK293 cells were transfected with wild-type (WT) or mutant (Mut) hypoxia response element (HRE) reporters and a TK luciferase control plasmid. Cells were treated with 100 μM or 1 mM BSO and placed under 1.5% O2 for 8 h. Lysates were analyzed for dual luciferase activity, and the reporter firefly luciferase was normalized to the control renilla luciferase activity. C: lysates from luciferase assays were analyzed for glutathione, corrected for protein content, and expressed as means ± SE normalized to the normoxic control. ***P < 0.05 vs. normoxic control.

To investigate a potential role for glutathione levels in modulating HIF transcriptional activity in wild-type cells, we next examined the effects of BSO treatment on HIF-mediated transcription of an HRE-luciferase reporter gene construct in HEK293 cells. Hypoxic treatment induced a 2.0-fold increase in wild-type HRE transcription after 16 h but had no effect on a mutated HRE reporter construct (Fig. 4B). Treatment with either 100 μM or 1 mM BSO significantly lowered glutathione (Fig. 4C) but had no effect on the level of HIF-mediated transcription; again, hypoxia elicited a 2.0-fold increase in luciferase activity (Fig. 4B). Together, these data indicate that the effects of mtROS on HIF activity are independent of decreased glutathione levels.

Role of protein synthesis in the hypoxic reduction of glutathione concentration.

Hypoxic exposure induces a number of transcriptional and translational changes that might account for the hypoxic glutathione decrease (58). To determine the role of protein synthesis in this glutathione response, HEK293 cells were treated with 50 μg/ml cyclohexamide and then exposed to hypoxia for 6 h. Cyclohexamide maintained glutathione levels in hypoxic cells (Fig. 5A), suggesting a role for de novo protein expression in the hypoxia-mediated glutathione decrease. This inhibition was not due to a change in mtROS formation (Fig. 5B): hypoxia caused a significant increase in ROS production, as seen previously (12, 13), which was inhibited by the antioxidant PDTC but unaffected by the cyclohexamide treatment (Fig. 5B).

Fig. 5.

Cyclohexamide (Cyclo) blocks hypoxic glutathione decrease but not ROS generation. HEK293 cells were pretreated for 15 min with 50 μg/ml cyclohexamide or 100 μM PDTC before being treated for 6 h with 1.5% O2. A: glutathione content was normalized to the normoxic control. B: cells were analyzed for ROS production by incubating with 5-(and-6)-carboxy-2′,7′-dichlorodihydrofluorescein diacetate (carboxy-H2DCFDA; 20 μM) in PBS and monitoring dihydrofluorescein (DCF) fluorescence. N, normoxia; H, hypoxia. ***P < 0.05 vs. normoxic control.

To investigate whether HIF-mediated transcription played a role in regulating cellular glutathione changes, HEK293 cells were treated for 6 h with a variety of hypoxic mimetics, including the iron chelators 1,10-phenanthroline (PhTr) and desferrioxamine, as well as cobalt chloride. Although all treatments resulted in stabilization of HIF-1α, there was no significant change in the total cellular glutathione levels (Fig. 6A). Thus induction of HIF activity is not itself sufficient to induce a glutathione decrease.

Fig. 6.

Hypoxic glutathione decrease is HIF independent. A: HEK293 cells were treated for 6 h with hypoxia mimetics 1,10-phenanthroline (PhTr) (1 mM), DFX (0.1 and 1 mM), or cobalt chloride (CoCl) (100 or 500 μM). Extracts were analyzed for glutathione and protein; 25 μg of extract were then analyzed via Western blot for the induction of HIF-1α protein. B: BPRC cells null for arylhydrocarbon receptor nuclear translocator (ARNT) (S3 and S7) or with ARNT restored (A1 and A4) were pretreated with 50 μg/ml cyclohexamide (Cyclohex) and then exposed to 6 h of 1.5% O2. Extracts were analyzed for glutathione, corrected for protein content, and normalized to the respective normoxic control. ***P < 0.05 vs. respective normoxic control.

Although not sufficient, it was still possible that HIF-mediated transcription was necessary for this effect. To test this idea, we used the BPRC1 hepatoma cell line, which is deficient in the HIF-1β [arylhydrocarbon receptor nuclear translocator (ARNT)] subunit and is consequently unable to direct HIF-mediated transcription. Previous reports have shown that expression of Arnt cDNAs in BPRC1 cells restores HIF-mediated target gene expression (35). We generated independent stably transformed BPRC1 clones containing either ARNT-expressing plasmids or vector controls. Control (S3 and S7) and ARNT-expressing (A1 and A4) cell lines were exposed to 8 h of hypoxia, and total glutathione was measured. As shown in Fig. 6B, all four cell lines exhibited a decrease in their total cellular glutathione, irrespective of their ability to carry out HIF-mediated transcription. Thus both the ARNT-deficient cell lines (S3 and S7) as well as the ARNT-expressing cell lines (A1 and A4) showed an ∼25% decrease in total cellular glutathione. This effect was not due to a difference between cell types because cyclohexamide was able to prevent the decrease in all four lines, as was treatment with the glutathione peroxidase mimetic ebselen (data not shown).

Hypoxic inhibition of glutathione biosynthesis and cystine transport.

The decrease in cellular glutathione could be achieved through a number of molecular mechanisms. Hepatic cells are known to efflux glutathione under conditions of oxidative stress, and all cells appear to degrade glutathione extracellularly (16, 28, 29, 46). Hypoxia did not cause a measurable efflux of glutathione into the media or extracellular matrix of HEK293 or Hep3B cells, even in the presence of acivicin, a compound known to block extracellular glutathione degradation (data not shown). To determine whether hypoxia instead affected glutathione synthetic rates in our system, we used the thiol-conjugating compound diethyl maleate to deplete glutathione as previously described (46). The rate of glutathione synthesis was then measured as a function of time. When HEK293 cells were exposed to 6 h of hypoxia before diethyl maleate treatment, they exhibited an approximately twofold lower rate of glutathione synthesis than normoxic cells (Fig. 7A). Inclusion of the ROS scavenger PDTC during the hypoxic exposure restored the rate of glutathione synthesis to normoxic levels, supporting a role for ROS production.

Fig. 7.

Hypoxia inhibits glutathione synthesis by restricting cystine uptake. A: HEK293 cells were pretreated for 15 min with 1 mM PDTC before being exposed to 6 h of 1.5% O2. Cells were then treated for 15 min with 2.5 mM diethyl maleate (DEM) to decrease cellular glutathione stores. Cells were then reoxygenated, DEM was washed off, and samples were harvested at indicated times for glutathione and protein determination. HEK293 and Hep3B cells were exposed to 6 h of 1.5% O2. B: 30 μg of whole cell extracts were probed for α-glutamylcysteine synthase (GCS) catalytic subunit protein levels (GCS-H) via Western blot using a nonspecific (NS) band to ensure equal loading. C: 12.5 μg total RNA were probed for GCS regulatory subunit mRNA levels (GCS-L) via Northern blot using 28S RNA as a loading control. D: Hep3B cells were exposed to 6 h of 1.5% O2, and the rates of l-[35S]cystine uptake were compared with normoxic controls. ***P < 0.05 vs. normoxia.

Glutathione synthesis rates can be affected by a number of factors, including 1) changes in the levels or activity of the rate-limiting synthetic enzyme γ-GCS and 2) availability of the substrate cysteine, which is transported into cells as cystine (34). GCS is a heterodimer consisting of both a heavy (73 kDa) catalytic subunit and a light (30 kDa) regulatory subunit. Expression levels of the GCS catalytic and regulatory subunits were unchanged in Hep3B and HEK293 cells after 6 h of hypoxic treatment, as determined by Western and Northern blot analyses (Fig. 7, B and C). However, Hep3B cells exposed to 6 h of 1.5% O2 showed an approximate 10-fold decrease in radiolabeled cystine uptake (Fig. 7C). These results suggest that decreased cystine transport under hypoxia results in lowered intracellular cysteine stores and a decreased ability to synthesize and maintain normoxic levels of glutathione.


A great deal of work has demonstrated that mammalian cells respond to changes in ambient O2 levels through a complex series of molecular events. Particular attention has been focused on transcriptional responses to hypoxia, many of which are mediated by the HIF family of proteins. Studies on the O2-dependent degradation of the HIF-α subunits led to the recent identification of evolutionarily conserved prolyl hydroxylases, which modify specific proline residues in HIF-α proteins, targeting them for ubiquitination and degradation in the 26S proteasome under normoxic conditions (25, 26, 59). Treating cells with hypoxia, or with prolyl hydroxylase inhibitors, interferes with HIF-α degradation and permits cells to activate HIF target genes that mediate adaptive responses. HIF target genes include those encoding erythropoietin, vascular endothelial growth factor, GLUT1, and glycolytic enzymes. It is becoming increasingly clear that these prolyl hydroxylases are critical components of the cellular O2 sensor(s), and intense efforts are underway to elucidate their regulation.

Changes in ambient O2 levels also impact mitochondrial function directly. Schumacker's group and others have demonstrated that hypoxic conditions can induce the formation of ROS derived from the mitochondrial electron transport chain (2–4, 12, 13, 18, 19, 32, 37, 42, 55). Although the use of pharmacological inhibitors raises the potential for nonspecific effects, the consistent results obtained when using a variety of inhibitors and respiration-deficient ρ0 cells argue strongly that mitochondrial electron flow and consequent ROS formation underlie the effects we observe. Intriguingly, several published reports corroborate our results, indicating that mitochondria and mtROS (2, 3, 12, 13, 22, 42) are necessary for HIF-α stabilization in a number of different cell types, although other groups have reported contradictory data (48, 54). The nature of this apparent discrepancy is not yet clear, but it may reflect the existence of multiple O2 sensing pathways, any one of which may be rate limiting depending on the specific cell type in question or the severity of the hypoxic treatment (42). A direct comparison of hypoxic mtROS levels and respiration rates in the various cell lines employed in these studies might help clarify this issue.

One might expect increased mtROS production in hypoxic cells to produce a change in the intracellular redox state. Our results indicate that total cellular glutathione levels decrease significantly in hypoxic cells, and in a time-dependent manner. Moreover, intact mitochondrial electron transport and the formation of mtROS are both required for this effect. Previous data have suggested that glutathione synthesis from methionine is inhibited under hypoxic conditions because of a decrease in cellular ATP levels and consequent inability to form S-adenosylmethionine (46); however, several factors appear more consistent with mtROS playing a central role in regulating hypoxic glutathione levels. For example, the synthesis of glutathione from methionine is apparently restricted to hepatocytes; hence, the kidney-derived HEK293 cells used in our studies instead rely on cysteine to generate glutathione. In addition, electron transport inhibitors such as rotenone can lower intracellular ATP levels yet prevent the hypoxic glutathione decrease. Further evidence that this effect is not entirely ATP dependent comes from experiments showing that antimycin A lowers GSH synthesis. Although antimycin A can also lower ATP levels, the fact that antimycin A-induced glutathione depletion was completely reversed by the ROS scavengers PDTC and ebselen strongly suggests that hypoxic changes in glutathione levels are a direct result of mtROS generation and cannot be explained solely by decreased ATP levels.

Our results indicate that glutathione synthesis is inhibited by hypoxia, caused (at least in part) by decreased uptake of extracellular cystine and consequent reduction in intracellular cysteine stores. Our data confirm a previous report showing hypoxic inhibition of cystine transport in human fibroblasts (7) and identify mtROS as the intracellular signal mediating this effect. One of our more intriguing findings is that cyclohexamide treatment could prevent hypoxic glutathione depletion. mtROS formation was unaffected by cyclohexamide, suggesting that protein synthesis is required to downregulate glutathione levels. We propose that one or more proteins synthesized during hypoxia negatively regulates cystine uptake, perhaps by inhibiting amino acid transporter activity or by regulating their levels, possibly through altered endocytosis as described for the plasma-membrane Na+-K+-ATPase in hypoxic alveolar cells (15). Which proteins mediate these effects and how they are regulated by mtROS are subjects for future studies.

The cellular consequences of lowering glutathione during hypoxia remain unclear and may depend on cell type and local environment. In our experiments, the reduction in glutathione did not appear to greatly facilitate HIF-1α stabilization or transcriptional activity. However, cellular redox changes may play a role in other adaptive responses to hypoxic stress. Examples of this have been observed in endochondral chondrocytes and more recently in dendritic cells and macrophages (36, 38). In chondrocytes, hypoxia decreases cellular glutathione levels, which in turn appears to facilitate cellular maturation (50). In addition to cell type-specific adaptive responses, lowered cellular glutathione levels may have the general effect of preparing cells for apoptosis under conditions of extreme oxidative stress (23, 53).

Interestingly, whereas the cell types mentioned above show a decrease in intracellular glutathione, hypoxic lung alveolar epithelial cells display an increase in cellular glutathione levels. In these cells, it was demonstrated that the resultant change in redox potential can modulate the activity of the transcription factors HIF-1α and NF-κB (20, 21, 41). These differences underscore the diversity in cellular responses to physiological stimuli such as hypoxia and the myriad of consequences that changes in cellular redox can have. It is interesting to note that lung epithelia are unusual in that they experience an almost immediate shift from physiological O2 levels (∼3%) to atmospheric O2 levels (21%) at birth. It is perhaps not surprising that changes in redox potential in these cells may be regulated differently than in kidney or liver cells. Given that cellular redox changes have been implicated in fields as diverse as apoptosis (23, 27, 53), transcriptional activity (10, 20, 49), cellular signaling (17), and human disease states (51), it is likely that changes in intracellular glutathione levels will be subject to complex regulation in different cell types. It will be of considerable interest to determine how mtROS affect the physiology of these different cell types by modulating intracellular glutathione.

Our data demonstrate that, in the hepatic Hep3B and HEK293 kidney cell lines, hypoxia reduces the concentration of glutathione, a major cellular antioxidant. This occurs through a reduction in cystine uptake and a concomitant decrease in glutathione synthesis. Moreover, we provide new mechanistic insight into this effect by demonstrating its dependence on both mtROS and HIF-independent protein synthesis. These results also expand the spectrum of cellular responses modulated by hypoxia-generated mtROS.


This work was supported by an National Institutes of Health Training Grant Program in Lung Biology and Respiratory Physiology (K. D. Mansfield, Grant HL-07027); the American Heart Association (K. D. Mansfield, Grant 0315339U); the Howard Hughes Medical Institute (M. C. Simon); The National Institutes of Health (M. C. Simon, Grant 66310); and the Abramson Family Cancer Research Institute, University of Pennsylvania (K. D. Mansfield, M. C. Simon, B. Keith).


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract