ORIGINAL ARTICLE

The cystine/cysteine cycle: a redox cycle regulating susceptibility versusresistance to cell death

A Banjac1, T Perisic1, H Sato2,6, A Seiler1, S Bannai2, N Weiss3, P Kolle3, K Tschoep4, RD Issels4,PT Daniel5, M Conrad1,7 and GW Bornkamm1,7

1GSF-Forschungszentrum fur Umwelt und Gesundheit, Institut fur Klinische Molekularbiologie und Tumorgenetik, Munchen,Germany; 2Department of Biochemistry, Institute of Basic Medical Sciences, University of Tsukuba, Tsukuba, Ibaraki, Japan;3Medizinische Poliklinik Innenstadt der Ludwig-Maximilians-Universitat, Munchen, Germany; 4Medizinische Klinik III derLudwig-Maximilians-Universitat, GSF-Klinische Kooperationsgruppe Hyperthermie, Munchen, Germany and 5MedizinischeKlinik mit Schwerpunkt Hamatologie und Onkologie, Charite, Humboldt Universitat, Berlin, Germany

The glutathione-dependent system is one of the keysystems regulating cellular redox balance, and thus cellfate. Cysteine, typically present in its oxidized form cystinein the extracellular space, is regarded as the rate-limitingsubstrate for glutathione (GSH) synthesis. Cystine istransported into cells by the highly specific amino-acidantiporter system xc

�. Since Burkitt’s Lymphoma (BL)cells display limited uptake capacity for cystine, and arethus prone to oxidative stress-induced cell death, we stablyexpressed the substrate-specific subunit of system xc

�,xCT, in HH514 BL cells. xCT-overexpressing cellsbecame highly resistant to oxidative stress, particularlyupon GSH depletion. Contrary to previous predictions,the increase of intracellular cysteine did not affect thecellular GSH pool, but concomitantly boosted extracel-lular cysteine concentrations. Even though cells weredepleted of bulk GSH, xCT overexpression maintainedcellular integrity by protecting against lipid peroxidation,a very early event in cell death progression. Our resultsshow that system xc

� protects against oxidative stress notby elevating intracellular GSH levels, but rather creates areducing extracellular environment by driving a highlyefficient cystine/cysteine redox cycle. Our findings showthat the cystine/cysteine redox cycle by itself must beviewed as a discrete major regulator of cell survival.Oncogene (2008) 27, 1618–1628; doi:10.1038/sj.onc.1210796;published online 10 September 2007

Keywords: cystine–glutamate exchange; glutathione me-tabolism; lipid peroxidation; redox regulation; system xc

�

Introduction

Redox regulation of cell cycle progression and cell deathhas attracted remarkable interest in recent years (Arnerand Holmgren, 2000). A variety of enzymatic systemsare involved in the maintenance of intracellular redoxhomeostasis including the glutathione and thioredoxin-dependent systems. Starting to dissect their functionalredundancy, we have created mice with targeteddeficiencies for both cytosolic and mitochondrial thio-redoxin reductases (Txnrd1 and 2, Conrad et al., 2004;Jakupoglu et al., 2005). Thereby, we could demonstratethat Txnrd2 indeed efficiently protects cells against thedetrimental effects of GSH depletion (Conrad et al.,2004).

Glutathione is present in cells in millimolar concen-tration and is considered as the major natural antioxi-dant, protecting cells from oxidative stress (Meister,1995). Availability of cystine/cysteine is the rate-limitingstep in GSH synthesis (Bannai and Tateishi, 1986; Ishiiet al., 1987). Cysteine is transported into cells via neutralamino-acid transport systems, whereas cystine, thepredominant form in plasma, extracellular body fluidsand cell culture medium, is carried by the anionicamino-acid transport system, system xc

� (Bannai andTateishi, 1986). Expression of system xc

� is fairly low inmany cell types as firstly depicted for murine Blymphocytes. Provision of b-mercaptoethanol (2-ME)or other sulfhydryl-containing compounds is thus aprerequisite for the survival of those cells in vitro(Broome and Jeng, 1973; Metcalf, 1976). Thiol-contain-ing compounds form mixed disulfides and therebyrelease cysteine which eventually enters cells via neutralamino-acid transport systems. The mixed disulfides inturn enter the cell via a transport system for bulkyamino acids, releasing cysteine intracellularly. Thiolcompounds are, however, not required if B cells are co-cultured with irradiated fibroblasts. Fibroblasts have ahigh uptake capacity for cystine, and upon intracellularreduction, provide cysteine to co-cultured B cells (Falket al., 1998).

Limited uptake capacity for cystine is a phenomenonneither restricted to murine cells nor to B cells. T cells

Received 22 February 2007; revised 4 July 2007; accepted 20 August

2007; published online 10 September 2007

Correspondence: Dr M Conrad, Institute of Clinical Molecular Biologyand Tumor Genetics, GSF-Research Centre, Marchioninistr. 25,Munich, Bavaria 81377, Germany.E-mail: marcus.conrad@gsf.de6Current address: Department of Bioresources, Faculty of Agriculture,Yamagata University, Tsuruoka, Yamagata 997-8555, Japan.7These authors contributed equally to this work.

Oncogene (2008) 27, 1618–1628& 2008 Nature Publishing Group All rights reserved 0950-9232/08 $30.00

www.nature.com/onc

are dependent on the supply of cysteine by antigen-presenting cells (Droge et al., 1991; Kuppner et al.,2003). Proliferation of many human and murinelymphoma and leukemia cell lines are dependent onfree thiols or a feeder layer of irradiated fibroblasts(Falk et al., 1993, 1998). Likewise, neurons show limiteduptake capacity for cystine which may lead to lowintracellular GSH concentrations and high susceptibilityto oxidative stress-induced cell death (Murphy et al.,1989, 1990).

The xc� cystine/glutamate-exchange transporter is a

heterodimer composed of the xCT light chain conferringthe specificity of the amino-acid exchange reaction, andthe 4F2 heavy chain, a ubiquitously expressed cellsurface component shared with several other amino-acidtransport systems (Sato et al., 1999; Verrey et al., 2000).The xCT light chain consists of 12 putative transmem-brane domains and is linked to the 4F2 heavy chainthrough an extracellular disulfide bond (Sato et al.,1999). Transcription of the xCT gene and xc

� transportactivities are induced by oxidative stress, mediated byelectrophilic agents, depletion of cystine and by oxygen(Bannai et al., 1989). xCT-deficient mice are viable andfertile indicating that the supply of cysteine can becompensated by other routes during normal develop-ment. By contrast, fibroblasts isolated from xCT�/� micecan only be cultivated if the culture medium issupplemented with 2-ME or N-acetylcysteine (NAC,Sato et al., 2005).

To study the contribution of system xc� to the redox

balance in proliferation and prevention of human B cellsfrom oxidative stress-induced cell death, we haveestablished a BL cell line that stably expresses xCT lightchain. Our data demonstrate that elevated expression ofxCT efficiently protects BL cells from oxidative stress-induced cell death even under conditions of cellularGSH depletion.

Results

L-cystine uptake is strongly increased inxCT-overexpressing cells and is sensitive toinhibition by L-glutamateHuman and murine xCT light chain were cloned into theexpression vector p141CAG-3SIP and used for stableexpression in HH514 BL cells. For each gene, threeindependent clones were selected with puromycin andconstantly maintained under selection pressure. Expres-sion levels were analysed by northern blotting of totalRNA, isolated from empty vector-transfected and xCT-overexpressing cells. By using an xCT-specific cDNAprobe, strong hybridization signals, corresponding tothe bicistronic xCT-IRES-puromycin acetyl transferasemRNA of about 3.5 kb, were detected (Figure 1b). Theendogenous xCT mRNA of 12.5 kb became visible afterextended exposure (Figure 1b). L-cystine uptake activitywas determined in vector- and xCT-transfected cells.Figure 1c shows a time course of L-cystine uptake(0.1 mCi per sample) in non-transfected cells, vector-

transfected cells and hxCT-overexpressing cells for thetime intervals of 1, 2 and 3min. L-cystine uptake wasvirtually undetectable in untransfected and vector-transfected cells, whereas xCT-transfected cells showedan uptake activity of more than 2 nmolmin�1 per mgprotein, which was linear at least for the first 3min.Uptake of L-cystine was efficiently inhibited in thepresence of 2.5mM L-glutamate, indicating that L-cystineuptake is solely mediated by the cystine–glutamateexchange transporter (Figure 1d). Seeding BL cells atlow cell density generates oxidative stress which can beovercome by the addition of antioxidant supplements(Falk et al., 1993; Brielmeier et al., 1998). To investigatewhether overexpression of human xCT provides agrowth advantage to HH514 cells, a critical cell densitywas determined that discriminates between cell survival/proliferation and cell death. Cells were plated in96-well plates in serial dilutions from 10 000 cells perwell (100 000 cellsml�1) down to 20 cells per well.For untransfected and vector-transfected controlcells, the critical cell density was 50 000 cellsml�1,whereas overexpression of human xCT supportedcell growth to a density as little as 6000 cellsml�1

(Figure 1e).

xCT-overexpressing HH514 cells are highly resistant toL-buthionine-sulfoximine- (BSO) mediated cell deathNext, we examined whether xCT-overexpression sup-ports growth of BL cells even under strongly limitingGSH conditions. To this end, HH514 cells werecultivated in the presence of various L-buthionine-sulfoximine (BSO) concentrations, and the number ofviable and dead cells was determined over a period of 8days (Figures 2a and b). We used BSO in our studies,since BSO specifically inhibits g-glutamyl-cysteinesynthetase (g-GCS), the rate-limiting enzyme in GSHanabolism, and thus causes rapid depletion of intracel-lular GSH (Griffith, 1982). While control cells alreadydied at the lowest BSO concentration within 48 h,proliferation of xCT-overexpressing cells was sloweddown in a dose-dependent manner and cells remainedviable even in the presence of 100 mM BSO (Figure 2b).BSO treatment did not impair the uptake capacity ofcells for cystine (Figure 2c). Addition of 2.5mM GSHrescued the BSO-mediated decrease in proliferation rate,again with lesser efficiency at high BSO concentrations,ruling out a toxic side effect of BSO (Figure 2d).

A decrease in the proliferation rate of BSO-treatedcells might be caused by an increase in the number ofcells exiting the cell cycle and entering G0 or G1, by anincreased death rate in the BSO-treated culture thatdiminishes the number of cells without affecting the cellcycle distribution, or both. To discriminate betweenthese possibilities, the cell cycle distribution of xCToverexpressing cells was studied that had been treatedwith 50 mM BSO. As shown in Figure 2e, the ratio ofcells in G1 to those in G2 remained unaffected, while thenumber of dead cells (sub-G1) increased from 6.3 to25% in the presence of BSO. The number of cells in Sphase decreased to some extent suggesting that cells in

The cystine/cysteine redox cycle

A Banjac et al

1619

Oncogene

S phase are more prone to oxidative stress-induced celldeath.

To correlate intracellular GSH levels with survivaland proliferation, cells were treated with differentconcentrations of BSO, and GSH concentrations weredetermined after 24 h of incubation, a time point whenthe number of viable control cells had not yet decreased.As shown in Figure 3a, no relevant change in GSHlevels of xCT-overexpressing cells and control cells was

detectable at any given BSO concentration. Mostimportantly, empty vector-transfected cells treated with5 mM BSO died within 48 h (Figure 2a), whereas xCT-overexpressing cells continued to proliferate under thesame conditions (Figure 2b), despite equal low GSHlevels (Figure 3a). Moreover, xCT-overexpressing cellssustained cell survival even at significantly lower GSHlevels due to higher BSO concentrations as compared tocontrol cells treated with only 5 mM BSO (Figure 3a). Inaddition, intracellular GSH became even undetectableby HPLC analysis in BSO-treated cells irrespective ofxCT-expression levels (Figure 3b). These unexpectedfindings strongly argue against a putative, criticalthreshold of GSH that discriminates between cellsurvival and cell death. Our data imply that HH514cells can survive and proliferate at strongly decreasedGSH levels provided that sufficient L-cystine uptake isguaranteed by system xc

�.

Intra- and extracellular L-cysteine levels are stronglyaugmented in xCT-overexpressing cellsSince increased cystine uptake activity rescued xCT-overexpressing cells from BSO-induced cell death, andGSH did not play the suspected role in maintaining cellsurvival, we focussed on the fate of intracellular cystine.As expected, a consistent increase of intracellularcysteine by a factor of about 6 was observed in xCT-overexpressing cells (Figure 4a). However, when westudied extracellular total mercaptan levels, we observedremarkably higher levels in xCT-overexpressing cellsregardless of BSO treatment (Figure 4b). In contrast,only marginal levels were secreted by control cells. HPLCanalysis revealed that secreted mercaptans consistpredominantly of cysteine (Figure 4c). These datasupport the notion that system xc

� drives a redox cycleconsisting of cystine import, intracellular reduction ofcystine to cysteine, cysteine secretion and reoxidation tocystine in the extracellular environment.

Figure 1 Stable overexpression of human xCT in HH514 cellspromotes cystine uptake and allows cells to grow at cell densitiesnon-permissive for control cells. (a) A northern blot showing highexpression of human xCT (arrow) in two transfected HH514 cellclones (lanes 3 and 4) and low expression in two cell clonestransfected with the empty vector (lanes 1 and 2). Panel (b) is a longexposure of (a) to visualize endogenous 12.5 kb xCT-mRNA(arrowhead). (c) Uptake of L-[14C]Cystine was measured in non-transfected cells (’), empty vector-transfected cells (J) andhxCT-overexpressing cells (E) for the indicated time intervals. (d)L-cystine uptake was inhibited by glutamate. L-cystine uptake wasmeasured for 1min in the absence (filled bars) or presence (emptybars) of 2.5mM glutamate (Glu) resulting in 87% inhibition byglutamate. A clear decrease was also noticed in empty vector-transfected cells. (e) Cell survival was monitored three weeks afterseeding. The critical cell density for survival and proliferation ofnon-transfected BL cells (’) and cells transfected with the emptyvector (m) was between 100 000 and 50 000 cellsml�1. (E) In thepresence of 100mM a-thioglycerol (a-TG), and 3mM pyruvate non-transfected BL were capable of proliferating at cell densities of200 cellsml�1. Overexpression of human (�) and murine xCT (x) inBL cells supported growth at densities of about 6000 cellsml�1 andhigher. Results represent mean values of four independentexperiments with duplicate measurements7s.d.

The cystine/cysteine redox cycle

A Banjac et al

1620

Oncogene

xCT-overexpression protects HH514 cells from hydrogenperoxide induced cell deathTo address whether xCT overexpression is able toprotect HH514 cells also from other death-inducingstimuli, xCT overexpressing HH514 and control cells weretreated with increasing concentrations of hydrogenperoxide. As shown in Figure 5a, xCT-overexpres-sing cells were more resistant to hydrogen peroxidetreatment than untransfected or mock-transfectedcontrol cells.

To study susceptibility of xCT-overexpressing cellsand control cells to killing through death receptors,

expression of CD95 and of the death receptors DR4 andDR5 was studied by flow cytometry (Figure 5b).Expression of neither CD95, nor DR4 and DR5 wasaffected by overexpression of xCT. CD95 was expressedat very low levels on HH514, mock-transfected andxCT-transfected cells, whereas the majority of cellsstained positive for DR4 and about half of the cells forDR5 (Figure 5b). Treatment of the cells with CD95/Fasligand (FasL) did not induce cell death consistent withthe low CD95 expression on the cell surface. Mock-transfected and xCT-transfected HH514 cells were alsoresistant to killing by TRAIL independently of the

Figure 2 xCT-overexpressing BL cells are highly resistant to cell death induced by GSH-depletion. (a) Vector-transfected control cellsdied already at concentrations as little as 5mM. The inset represents a magnification of the cell numbers at days 0, 1 and 2. The scalecorresponds to 0.05� 106 cellsml�1 per bar. The columns in the inset reflect, from left to right, the same increasing BSO concentrationsas depicted for the symbols (from top to bottom). (b) Dose-dependent impairment of proliferation of xCT-overexpressing BL cells inthe presence of 0–100mM BSO. (c) To exclude the possibility that BSO interferes with cystine uptake through system xc-, L-cystineuptake activity was measured in the absence (filled bars) and presence (empty bars) of BSO. No significant difference in xCT activitywas observed in xCT-overexpressing and vector-transfected cells at a BSO concentration of 5mM. Results represent mean values offour independent experiments with duplicate measurements7s.d. (d) Glutathione (2.5mM) supplementation efficiently rescued celldeath induced by BSO treatment. (e) Cell cycle analysis of xCT-overexpressing cells revealed that cells treated with 50 mM BSO do notexit the cell cycle but rather die. Note the increase in cells in sub-G1 (M1), while cells in G1 (M2) and S phase (M3) decrease.

The cystine/cysteine redox cycle

A Banjac et al

1621

Oncogene

expression of xCT (Figure 5c). Of note, xCT-over-expression did not modulate the expression of a varietyof anti- or proapoptotic genes (Supplementary Figure1), and did not protect the cells from epirubicin-inducedcell death (Figure 5c).

The cystine/cysteine cycle protects cells fromBSO-induced lipid peroxidationTime-course experiments revealed that cell death be-came apparent in control cells 24 h after BSO treatment(Figure 6). To address whether dying cells showhallmarks of classical apoptosis, cells were stained withAnnexin V and propidium iodide (PI). Annexin V-positive,PI-negative apoptotic (pre-necrotic) cells could not bedetected at any time point after BSO treatment

(Supplementary Figure 2), indicating that BSO-treatedcells do not undergo a classical form of apoptosis(Nicotera and Melino, 2004; Melino, 2005).

Figure 3 Glutathione levels are not altered by xCT overexpres-sion. (a) Total cellular GSH levels were measured (Tietze) in hxCT-overexpressing (filled bars) and vector-transfected cells (emptybars) after 24 h cultivation in the absence/presence of various BSOconcentrations. (b) Semi-quantitative determination of intracellularcysteine and GSH by HPLC in hxCT-overexpressing and controlcells in the absence/presence of 10 mM BSO.

Figure 4 Intra- and in particular extracellular cysteine levels arestrongly increased in xCT-overexpressing cells. (a) Intracellularcysteine levels are approximately six-fold increased in overexpres-sing versus vector-transfected cells (filled bars) regardless of BSOtreatment (50mM BSO, 24 h; empty bars). (b) Total extracellularmercaptans are highly elevated in the supernatant of xCT-over-expressing cells. The data represent mean values of three independentexperiments with duplicate measurements7s.d. (c) HPLC analysisconfirmed that cysteine is the secreted mercaptan in the culturemedium of hxCT-overexpressing cells in the absence (�BSO) as wellas in the presence of BSO (þBSO).

The cystine/cysteine redox cycle

A Banjac et al

1622

Oncogene

We examined additional cell death initiating orpromoting parameters such as lipid peroxidation,accumulation of intracellular ROS, breakdown of themitochondrial membrane potential, caspase activationand DNA fragmentation (Figure 6 and SupplementaryFigure 3). Lipid peroxidation clearly preceded all othercellular processes, starting already 8 h after addition ofBSO (Figure 6). Lipid peroxidation became prominentin the vast majority of cells 12 h of BSO treatment,clearly before ROS accumulation and breakdown of themitochondrial membrane potential occurred. An in-crease in intracellular ROS was observed 24 h after BSOaddition, a time point when lipid peroxidation hadprogressed in virtually all cells (Figure 6). Breakdown ofthe mitochondrial membrane potential ensued after lipidperoxidation and ROS accumulation (Figure 6), sug-gesting that loss of mitochondrial integrity initiates theexecution phase of BSO-induced cell death associatedwith broad caspase activation and DNA fragmentation(Supplementary Figure 3).

If lipid peroxidation and subsequent breakdown ofmembrane integrity trigger BSO-induced cell death,antioxidants specifically targeting lipid membranes

should protect cells from BSO-induced cell death. Asshown in Figure 7a, treatment of control cells withvitamin E fully protected HH514 cells from lipidperoxidation and cell death. xCT-overexpressing cellsbut not vector-transfected HH514 cells were protectedfrom lipid peroxidation (Figure 7b). In line with the datadescribed in Figure 2, lipid peroxidation increased invector-transfected cells with increasing BSO concentra-tion, and likewise, protection from lipid peroxidation byxCT overexpression was less complete at higher BSOconcentration (Figure 7b).

Discussion

To investigate the role of cystine uptake for theregulation of cell survival and proliferation, we estab-lished HH514 BL cell clones that stably overexpresshuman xCT light chain of the xc

� cystine/glutamateexchange transporter. Our data show that overexpres-sion of xCT light chain promotes a strong increase in theuptake activity for L-cystine and protects HH514 cells

Figure 5 xCT-overexpression confers resistance to hydrogen peroxide, but does not influence receptor-mediated apoptotic pathways.(a) Cells were incubated with increasing concentrations of H2O2 and the number of viable cells was determined 48 h after treatment. (b)Expression of CD95 receptor and TRAIL receptors DR4 and DR5 was determined by flow cytometry on a single cell level.Representative histograms are shown on a log10 logarithmic scale for relative receptor expression of the death receptor (black) orcontrol stained cells (white). The T-ALL Jurkat and the lymphoblastoid line SKW6.4 served as positive controls. Percentages indicaterelative numbers of positively stained cells based on marker sets relative to the control stained cells. (c) Mock and hxCT overexpressingcells were exposed for 72 h to epirubicin, or the death ligands TRAIL (TNF-related apoptosis inducing ligand, APO-2L) and Fasligand (CD95L, APO-1L) at a dose of 100 ngml�1. Cell death was determined by flow cytometric analysis of genomic DNAfragmentation on a single cell level. Means7s.d. (n¼ 3) from a typical experiment are shown.

The cystine/cysteine redox cycle

A Banjac et al

1623

Oncogene

from oxidative stress-induced cell death. The protectiveeffect of xCT is not associated with changes in theexpression level of various pro- or antiapoptoticgenes.

Availability of cysteine is regarded as the substrate-limiting step in GSH synthesis (Ishii et al., 1987), and wetherefore investigated whether increased cystine uptakeimpacts the intracellular GSH pool. We included BSO,the most powerful inhibitor of GSH synthesis, in ouranalyses to correlate survival and proliferation ofcontrol and xCT-overexpressing cells with intracellularGSH levels. Side effects of BSO could be excluded asaddition of GSH rescued the effect of BSO on survivaland proliferation of HH514 cells. At high BSOconcentrations, the protective effect of xCT overexpres-sion on cell survival was less complete suggesting thathigh BSO concentrations select for high xCT expressionand cells with lower xCT expression are killed by BSO inthis experimental setting. Unexpectedly, xCT-over-expression does not fuel intracellular GSH levels eitherin the absence or in the presence of a wide range ofdifferent BSO concentrations (Figure 3). In fact, theincreased cystine uptake leads to about six-fold increase

in intracellular and to an even more pronouncedincrease in extracellular cysteine levels. These highcysteine levels are obviously sufficient to sustain cellgrowth of xCT-overexpressing cells even under condi-tions of strongly reduced GSH levels. Hence, our resultsdo not support the idea of a putative threshold of GSHthat decides on cell survival or cell death. In fact, systemxc� drives a distinct cystine/cysteine cycle that efficiently

protects cells from detrimental lipid peroxidation, a veryearly event in the onset of BSO-induced cell deathstarting about 8 h after the addition of BSO. Lipidperoxidation is initiated much earlier than ROSaccumulation (24 h) and breakdown of the mitochon-drial membrane potential (30 h) suggesting that lipidperoxidation is the causal and initiating event in thecourse of cell death induction. This is strongly supportedby our finding that cell death induced by BSO-mediatedGSH depletion is completely rescued by the lipophilicantioxidant vitamin E. Breakdown of the mitochondrialmembrane potential is a comparatively late event thatpresumably initiates the execution phase of cell deathassociated with broad caspase activation, DNA frag-mentation and Annexin V/PI staining.

Figure 6 Lipid peroxidation precedes BSO-induced cell death. BL cells were treated with BSO for the indicated time points. Cellviability was assessed by morphology (FSC/SSC), lipid peroxidation by C11-BODIPY-staining, cellular ROS levels by DCF andmitochondrial membrane potential by JC-1 staining.

The cystine/cysteine redox cycle

A Banjac et al

1624

Oncogene

The bulk antioxidant function of cellular GSH canthus be substituted by the cystine/cysteine cycle, thoughit cannot be ruled out at present whether small amountsof GSH are indispensable for other aspects in cellularfunctions. For instance, it was shown recently that GSHis essential for cytosolic Fe–S cluster formation in yeast(Sipos et al., 2002). Independent evidence that bulkGSH is dispensable for cell proliferation at least in vitrohas been provided by genetic means. Targeted inactiva-tion of g-GCS in mice revealed that GSH is essential forembryonic development, but unexpectedly not requiredfor proliferation of blastocyst-derived cell lines in thepresence of N-acetylcysteine (Shi et al., 2000). In vitroexperiments with CaCo2 cells showed that the extra-cellular redox potential defined by the ratio of extra-cellular cysteine to cystine is critical for cell proliferationand independent of GSH (Jonas et al., 2002; Sonodaet al., 2004).

The importance of the cystine/cysteine cycle for theresistance of cancer cells to chemotherapy has beenrecognized only recently (Okuno et al., 2003). Activationof the xCT gene in cancer cells by cytotoxic drugs maybe a mechanism contributing to chemotherapy resis-tance in vivo, a possibility that apparently deservesfurther studies. It is another important issue whether thelimitation of cystine uptake, as observed in various celllines, has any pathophysiological impact. As proposedby Ye and Sontheimer (1999), in the central nervoussystem, deregulated activity of the cystine/cysteine cycle,as observed in human malignant glioma cells, leads to adetrimental increase in secreted glutamate associated

with increased glutamate toxicity, neuronal death andseizures. In the hematopoietic system, the particularsensitivity of the cells to ionizing radiation may berelated to a limited uptake capacity of cystine. Limitedactivity of the cystine/cysteine cycle may also benecessary for an orchestrated regulation of apoptosisin the immune system. Low activity of system xc

� in Band T cells may be a prerequisite to render the immunesystem highly dynamic. Redox regulation also appearsto be crucially important to fine tune the T helper cellresponse towards Th1 or Th2 (Peterson et al., 1998).

In the present work, we provide strong evidence thatthe cystine/cysteine redox cycle, driven by system xc

�, ischaracterized by slightly increased intracellular andexceedingly high extracellular cysteine concentrations,and efficiently protects cells from oxidative stress-induced cell death. The previous paradigm of augmen-ted GSH levels due to increased cystine availability isnot supported by our findings. In fact, we provideexperimental evidence that increased cysteine uptakedoes not alter cellular GSH levels to a significant extent;hence, the cystine/cysteine cycle by itself might beregarded as a major redox system regulating cell survivaland cell death.

Materials and methods

Cell line and chemicalsThe BL cell line HH514 cells was cultured in RPMI 1640medium containing 10% fetal bovine serum (FBS) (Biochrom,Berlin, Germany), 100Uml�1 penicillin, 100 mgml�1 strepto-mycin and 2mM glutamine. L-[14C]Cystine was purchased fromAmersham Corp. (Freiburg, Germany) and other chemicalsfrom Sigma (Deisenhofen, Germany), Invitrogen (Karlsruhe,Germany), and ICN (ICN Biomedicals GmbH, Eschwege,Germany). Restriction and DNA-modifying enzymes andDNA linkers were obtained from MBI Fermentas (Vilnius,Lithunia) and Biolabs GmbH (Schwalbach, Germany).

Cloning of mouse and human xCT into an expression vector andstable expression in HH514 cellsThe BclI site of pSPORT-mxCT was converted into an EcoRIsite and the HindIII site into a NheI site. Murine xCT cDNAwas transferred as a HindIII/NheI fragment into the vectorp141CAG-3SIP driving expression from a CMV enhancer-chicken-b-actin promoter (Niwa et al., 2002; Okita et al.,2004). For cloning of human xCT, a synthetic linker carryingEcoRI, BstBI, BglII and NheI sites (MCa: 50-ATTTCATTCGAACGG AGATCTTG-30; MCb: 50-CTAGCATGA TCTCCGTTCGAATG-30) was cloned in p141CAG-3SIP digestedwith EcoRI and HindIII. Human xCT cDNA was excised froma pBluescript SK clone as a ClaI/BglII fragment and insertedinto the modified p141CAG-3SIP vector between the BstBIand BglII sites. Transfections of the cell line HH514 with xCT-expressing or empty vectors were performed as describedpreviously (Brielmeier et al., 1998). Stably transfected cloneswere selected with puromycin at a final concentration of2 mgml�1.

Northern blot analysisTotal RNA was isolated using Qiagen RNeasy Midi Kitaccording to the manufacturer’s instructions (Qiagen, Hilden,

Figure 7 BSO-induced lipid peroxidation and cell death can beprevented by vitamin E (a) and xCT overexpression (b). Lipidperoxidation was monitored by C11-BODIPY staining.

The cystine/cysteine redox cycle

A Banjac et al

1625

Oncogene

Germany). Total RNA (10 mg) per cell clone were blotted anddetection of human and murine xCT-expression was per-formed with radiolabelled human- and mouse-specific cDNAprobes.

Measurement of L-cystine transport activityUptake of L-cystine was measured as described (Novogrodskyet al., 1979; Sato et al., 1995).

Viability assayCells were plated in 6-well plates at 1� 105 cellsml�1 in 3mlstandard medium supplemented with different BSO concen-trations. The number of viable cells was determined by trypanblue exclusion or PI staining by flow cytometry.

Glutathione content (Tietze)Total glutathione (GSH and GSSG) was extracted with 5%trichloroacetic acid from cells grown in 10 or 200ml flasks.Total glutathione was measured as outlined previously(Bannai and Ishii, 1982). The method is based on the reductionof 5,50-dithiobis-2-nitrobenzoic acid by GSH which in turn isreduced by glutathione reductase (Tietze, 1969). IntracellularGSH was additionally determined by HPLC analysis (seebelow).

Determination of total thiols in the mediumTotal mercaptans secreted into the cell culture medium weredetermined as described previously (Bannai and Ishii, 1982).GSH was used as a standard.

HPLC determination of extra- and intracellular cysteineDetermination of cysteine in cell culture supernatants wasperformed by a HPLC after derivatization of the freesulfhydryl group with SBDF (7-fluoro-benzo-2-oxa-1,3-dia-zole-4-sulphonate) as described previously with slight mod-ifications (Feussner et al., 1997). A measure of 2� 106 cellswere washed, resuspended in 2ml of serum-free medium andincubated for 2 h at 37 1C. Cells were collected by centrifuga-tion and 100 ml of the supernatant were mixed with 100ml 2M

borate buffer (containing 5mM Na2EDTA, pH 10.05 and100 ml SBDF (1mgml�1 in borate buffer)) and incubated at60 1C for 60min. The reaction was performed without anyreducing agent to detect free cysteine only. The reaction wasstopped by incubation on ice for 5min. The sample wasdeproteinized by adding 200ml HClO4 containing 0.5M EDTAand centrifuged. A volume of 50 ml of the supernatant wasinjected in a HPLC column and analysed as describedpreviously. Cysteine was used as a standard.

For the determination of intracellular cysteine levels, cellpellets were resuspended and sonified in 500ml borate buffer(2M, pH 9.5). After centrifugation, 200 ml of the supernatantwas reduced by adding 20ml of tri-n-butylphosphine for 30minat 4 1C, deproteinized using HClO4, and centrifuged again. Avolume of 100ml of the supernatant was mixed with 250ml of2M borate buffer and thiol groups were derivatized usingSBDF. HPLC analysis was performed as described above.

Quantification of lipid peroxidation and intracellular reactiveoxygen species (ROS)Lipid peroxidation was measured by FACS analysis usingC11-BODIPY. ROS were detected by flow cytometry usingDCF according to the manufacturer’s instructions (MolecularProbes, Inc., OR, USA).

Determination of the mitochondrial membrane potentialIn all, 2� 105 cells were collected and stained with 5,50,6,60-tetrachloro-1,10,3,30-tetraethyl-benzimidazolycarbocyanin iodide(JC-1; Biotium, Inc., Hayward, CA, USA) at a final concent-ration of 2.5 mgml�1 in 500ml PBS as described by von Haefenet al. (2004). After incubation for 30min at 37 1C, cells werewashed and resuspended in 200 ml PBS. The mitochondrialmembrane potential was measured by flow cytometry usinga FACScan (Becton Dickinson, Heidelberg, Germany) equip-ped with CellQuest software. Data are presented as per-centage of cells with lowered membrane potential (DCm).

Measurement of DNA fragmentationTo measure genomic DNA fragmentation, the nuclear DNAcontent was determined by flow cytometry (Gillissen et al.,2003). Overall, 2� 105 cells were pelleted in a 96-well U-bottom plate and fixed in 200ml of 2% (v v�1) formaldehyde inPBS on ice for 30min. After fixation, DNA was precipitatedwith 100% ethanol for 15min, pelleted and resuspended inPBS containing RNaseA (40 mgml�1). Following incubation at37 1C for 30min, the pellet was resuspended in 200 ml ofpropidium iodide (PI, 50 mgml�1) and incubated overnight inthe dark at 4 1C. Nuclear DNA fragmentation was quantifiedby flow cytometric determination of hypodiploid DNA. Datawere collected and analysed using FACScan flow cytometerwith CellQuest software.

ImmunofluorescenceSurface expression of death receptors was determined viadirect immunofluorescence by the use of a FACScan flowcytometer (Becton-Dickinson, Heidelberg, Germany) on asingle cell level. PE-labelled antibodies for CD95/Fas (cloneDX2), TRAIL receptor 1/DR4 (clone DJR1), Trail R2/DR5(clone DJR2-4) and a PE-labelled non-binding control anti-body (clone P3, mouse IgG1 k) were from eBioscence (SanDiego, CA, USA). Histograms are shown for 10 000 eventsacquired in a 256-channel resolution. Percentages of positivecells were determined by setting markers in relation to controlstained cultures.

Protein extraction and immunoblottingCells were washed twice with PBS and lysed in buffercontaining 10mM Tris–HCl pH 7.5, 300mM NaCl, 1% TritonX-100, 2mM MgCl2, 5mM EDTA, 1mM pepstatin, 1 mM

leupeptin and 0.1mM phenylmethylsulfonyl fluoride. Proteinconcentration was determined using the bicinchoninic acidassay (Pierce, Rockford, IL, USA), and equal amounts ofprotein (usually 20 mg per lane) were separated by SDS–PAGE. Immunoblotting was performed as described (Wiederet al., 2001). For protein visualization the following antibodieswere used: the mouse monoclonal antibodies anti-Bax YTH-2D2 (1:10 000, Trevigen, Gaithersburg, MA, USA) and anti-Bcl-2 100/D5 (1:100) (Novocastra, Newcastle upon Tyne,UK), and the rat monoclonal antibody anti-Bcl-w (16H12,1:1000) (Alexis). Anti-PUMA (N-19, 1:100), anti-Bid (C-20,1:1000) and anti-Nbk/Bik (N-19, 1:1000) were polyclonal goatantibodies from Santa Cruz. All other reagents were poly-clonal rabbit antibodies used in a dilution of 1:1000. Anti-Bakantibodies were purchased from Dako, anti-Bcl-x and anti-Bad antibodies from Transduction Laboratories, anti-Bim andanti-p53 from PharMingen, and anti-b-actin antibodies fromSigma. Protein bands were detected using the enhancedchemiluminescence system (Amersham Buchler, Braunsch-weig, Germany).

The cystine/cysteine redox cycle

A Banjac et al

1626

Oncogene

Detection of active caspasesActive caspases were determined by the FLICA apoptosisdetection kit (Alexis GmbH, Gruenberg, Germany) accordingto the manufacturer’s instructions. The methodology is basedon the covalent binding of a fluorochrome-labelled inhibitor ofcaspases (FLICA) to the active site (Ekert et al., 1999). Theseinhibitors (VDVAD-AFC for caspase 2; DEVD-AFC forcaspase 3; IETD-AFC for caspase 8 and LEHD-AFC forcaspase 9) are cell permeable and non-cytotoxic. The numberof cells with active caspase was determined by FACScan flowcytometry using the CellQuest software. FITC-labelled FLI-CAs were used for staining and positive cells were measured inthe FL-1 green channel.

Measurement of cell death by Annexin-V-FITC and propidiumiodide stainingCell death was determined by staining cells with Annexin-V-FITC and counterstaining with PI. Cells were washed twicewith cold PBS and resuspended in buffer containing 10mM

N-(2-hydroxyethyl)piperazin-N0-3(propansulfonic acid)/NaOH,pH 7.4, 140mM NaCl, 2.5mM CaCl2. Next, 5 ml of Annexin-V-FITC (BD PharMingen, Heidelberg, Germany) and 10 mlPI (20 mgml�1, Sigma-Aldrich) were added. Analyses were

performed using FACScan and CellQuest analysis software(Becton Dickinson, Heidelberg, Germany).

Statistical analysisComparison between groups was made by the Student’s t-test.A difference between groups with Po0.05 was consideredsignificant.

Acknowledgements

We thank A Richter for expert technical assistance, A Klopferfor technical advice and J-M Bechet for critical reading ofthe manuscript. We are grateful to T Schroeder for thep141CAG-3SIP vector, W Droege and A Roscher for helpfuldiscussions. This work was supported by a grant of DeutscheForschungsgemeinschaft (Priority Programme ‘Biology ofSelenoproteins’) to GWB and MC, by a short-term fellowshipof Deutsche Akademische Austauschdienst (DAAD), andJapanese Society for Promotion of Science (JSPS) to HS andMC for working in Munich and in Tsuruoka, respectively.GWB was additionally supported by Fonds der ChemischenIndustrie.

References

Arner ES, Holmgren A. (2000). Physiological functions ofthioredoxin and thioredoxin reductase. Eur J Biochem 267:6102–6109.

Bannai S, Ishii T. (1982). Transport of cystine and cysteine andcell growth in cultured human diploid fibroblasts: effect ofglutamate and homocysteate. J Cell Physiol 112: 265–272.

Bannai S, Tateishi N. (1986). Role of membrane transport inmetabolism and function of glutathione in mammals.J Membr Biol 89: 1–8.

Bannai S, Sato H, Ishii T, Sugita Y. (1989). Induction ofcystine transport activity in human fibroblasts by oxygen.J Biol Chem 264: 18480–18484.

Brielmeier M, Bechet JM, Falk MH, Pawlita M, Polack A,Bornkamm GW. (1998). Improving stable transfectionefficiency: antioxidants dramatically improve the outgrowthof clones under dominant marker selection. Nucleic AcidsRes 26: 2082–2085.

Broome JD, Jeng MW. (1973). Promotion of replication inlymphoid cells by specific thiols and disulfides in vitro.Effects on mouse lymphoma cells in comparison with spleniclymphocytes. J Exp Med 138: 574–592.

Conrad M, Jakupoglu C, Moreno SG, Lippl S, Banjac A,Schneider M et al. (2004). Essential role for mitochondrialthioredoxin reductase in hematopoiesis, heart development,and heart function. Mol Cell Biol 24: 9414–9423.

Droge W, Eck HP, Gmunder H, Mihm S. (1991). Modulationof lymphocyte functions and immune responses by cysteineand cysteine derivatives. Am J Med 91: 140S–144S.

Ekert PG, Silke J, Vaux DL. (1999). Caspase inhibitors. CellDeath Differ 6: 1081–1086.

Falk MH, Hultner L, Milner A, Gregory CD, Bornkamm GW.(1993). Irradiated fibroblasts protect Burkitt lymphoma cellsfrom apoptosis by a mechanism independent of bcl-2. Int JCancer 55: 485–491.

Falk MH, Meier T, Issels RD, Brielmeier M, Scheffer B,Bornkamm GW. (1998). Apoptosis in Burkitt lymphoma cellsis prevented by promotion of cysteine uptake. Int J Cancer75: 620–625.

Feussner A, Rolinski B, Weiss N, Deufel T, Wolfram G,Roscher AA. (1997). Determination of total homocysteine in

human plasma by isocratic high-performance liquid chro-matography. Eur J Clin Chem Clin Biochem 35: 687–691.

Gillissen B, Essmann F, Graupner V, Starck L, Radetzki S,Dorken B et al. (2003). Induction of cell death by the BH3-only Bcl-2 homolog Nbk/Bik is mediated by an entirely Bax-dependent mitochondrial pathway. EMBO J 22: 3580–3590.

Griffith OW. (1982). Mechanism of action, metabolism, andtoxicity of buthionine sulfoximine and its higher homologs,potent inhibitors of glutathione synthesis. J Biol Chem 257:13704–13712.

Hattori H, Imai H, Furuhama K, Sato O, Nakagawa Y.(2005). Induction of phospholipid hydroperoxide glu-tathione peroxidase in human polymorphonuclear neutro-phils and HL60 cells stimulated with TNF-alpha. BiochemBiophys Res Commun 337: 464–473.

Ishii T, Sugita Y, Bannai S. (1987). Regulation of glutathionelevels in mouse spleen lymphocytes by transport of cysteine.J Cell Physiol 133: 330–336.

Jakupoglu C, Przemeck GK, Schneider M, Moreno SG, MayrN, Hatzopoulos AK et al. (2005). Cytoplasmic thioredoxinreductase is essential for embryogenesis but dispensable forcardiac development. Mol Cell Biol 25: 1980–1988.

Jonas CR, Ziegler TR, Gu LH, Jones DP. (2002). Extracellularthiol/disulfide redox state affects proliferation rate in ahuman colon carcinoma (Caco2) cell line. Free Radic BiolMed 33: 1499–1506.

Kuppner MC, Scharner A, Milani V, Von Hesler C, TschopKE, Heinz O et al. (2003). Ifosfamide impairs theallostimulatory capacity of human dendritic cells byintracellular glutathione depletion. Blood 102: 3668–3674.

Meister A. (1995). Glutathione metabolism. Methods Enzymol251: 3–7.

Melino G. (2005). Discovery of the ubiquitin proteasomesystem and its involvement in apoptosis. Cell Death Differ12: 1155–1157.

Metcalf D. (1976). Role of mercaptoethanol and endotoxin instimulating B lymphocyte colony formation in vitro.J Immunol 116: 635–638.

Murphy TH, Miyamoto M, Sastre A, Schnaar RL, Coyle JT.(1989). Glutamate toxicity in a neuronal cell line involves

The cystine/cysteine redox cycle

A Banjac et al

1627

Oncogene

inhibition of cystine transport leading to oxidative stress.Neuron 2: 1547–1558.

Murphy TH, Schnaar RL, Coyle JT. (1990). Immature corticalneurons are uniquely sensitive to glutamate toxicity byinhibition of cystine uptake. FASEB J 4: 1624–1633.

Nicotera P, Melino G. (2004). Regulation of the apoptosis-necrosis switch. Oncogene 23: 2757–2765.

Niwa H, Masui S, Chambers I, Smith AG, Miyazaki J. (2002).Phenotypic complementation establishes requirements forspecific POU domain and generic transactivation functionof Oct-3/4 in embryonic stem cells. Mol Cell Biol 22:1526–1536.

Nonn L, Williams RR, Erickson RP, Powis G. (2003). Theabsence of mitochondrial thioredoxin 2 causes massiveapoptosis, exencephaly, and early embryonic lethality inhomozygous mice. Mol Cell Biol 23: 916–922.

Novogrodsky A, Nehring Jr RE, Meister A. (1979). Inhibitionof amino acid transport into lymphoid cells by the glutamineanalog L-2-amino-4-oxo-5-chloropentanoate. Proc NatlAcad Sci USA 76: 4932–4935.

Okita C, Sato M, Schroeder T. (2004). Generation ofoptimized yellow and red fluorescent proteins with distinctsubcellular localization. Biotechniques 36: 418–422, 424.

Okuno S, Sato H, Kuriyama-Matsumura K, Tamba M, WangH, Sohda S et al. (2003). Role of cystine transport inintracellular glutathione level and cisplatin resistance inhuman ovarian cancer cell lines. Br J Cancer 88: 951–956.

Peterson JD, Herzenberg LA, Vasquez K, Waltenbaugh C.(1998). Glutathione levels in antigen-presenting cells mod-ulate Th1 versus Th2 response patterns. Proc Natl Acad SciUSA 95: 3071–3076.

Sato H, Fujiwara K, Sagara J, Bannai S. (1995). Induction ofcystine transport activity in mouse peritoneal macrophagesby bacterial lipopolysaccharide. Biochem J 310: 547–551.

Sato H, Shiiya A, Kimata M, Maebara K, Tamba M,Sakakura Y et al. (2005). Redox imbalance in cystine/

glutamate transporter-deficient mice. J Biol Chem 280:37423–37429.

Sato H, Tamba M, Ishii T, Bannai S. (1999). Cloning andexpression of a plasma membrane cystine/glutamate ex-change transporter composed of two distinct proteins. J BiolChem 274: 11455–11458.

Shi ZZ, Osei-Frimpong J, Kala G, Kala SV, Barrios RJ,Habib GM et al. (2000). Glutathione synthesis is essentialfor mouse development but not for cell growth in culture.Proc Natl Acad Sci USA 97: 5101–5106.

Sipos K, Lange H, Fekete Z, Ullmann P, Lill R, Kispal G.(2002). Maturation of cytosolic iron–sulfur proteins requiresglutathione. J Biol Chem 277: 26944–26949.

Sonoda Y, Aiba N, Utsubo R, Koguchi E, Hasegawa M,Kasahara T. (2004). Induction of antioxidant enzymes byFAK in a human leukemic cell line, HL-60. Biochim BiophysActa 1683: 22–32.

Tietze F. (1969). Enzymic method for quantitative determina-tion of nanogram amounts of total and oxidized glutathione:applications to mammalian blood and other tissues. AnalBiochem 27: 502–522.

Verrey F, Meier C, Rossier G, Kuhn LC. (2000). Glycopro-tein-associated amino acid exchangers: broadening the rangeof transport specificity. Pflugers Arch 440: 503–512.

von Haefen C, Gillissen B, Hemmati PG, Wendt J, Guner D,Mrozek A et al. (2004). Multidomain Bcl-2 homolog Bax butnot Bak mediates synergistic induction of apoptosis byTRAIL and 5-FU through the mitochondrial apoptosispathway. Oncogene 23: 8320–8332.

Wieder T, Essmann F, Prokop A, Schmelz K, Schulze-OsthoffK, Beyaert R et al. (2001). Activation of caspase-8 in drug-induced apoptosis of B-lymphoid cells is independent ofCD95/Fas receptor-ligand interaction and occurs down-stream of caspase-3. Blood 97: 1378–1387.

Ye ZC, Sontheimer H. (1999). Glioma cells release excitotoxicconcentrations of glutamate. Cancer Res 59: 4383–4391.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc).

The cystine/cysteine redox cycle

A Banjac et al

1628

Oncogene