The cystine/glutamate exchanger (xCT) provides intracellular cyst(e)ine for production of glutathione, a major cellular antioxidant. Using xCT overexpression and underexpression, we present evidence that xCT-dependent glutathione production modulates both neuroprotection from oxidative stress and cell proliferation. In embryonic and adult rat brain, xCT protein was enriched at the CSF–brain barrier (i.e., meninges) and also expressed in the cortex, hippocampus, striatum, and cerebellum. To examine the neuroprotective role of xCT, various non-neuronal cell types (astrocytes, meningeal cells, and peripheral fibroblasts) were cocultured with immature cortical neurons and exposed to oxidative glutamate toxicity, a model involving glutathione depletion. Cultured meningeal cells, which naturally maintain high xCT expression, were more neuroprotective than astrocytes. Selective xCT overexpression in astrocytes was sufficient to enhance glutathione synthesis/release and confer potent glutathione-dependent neuroprotection from oxidative stress. Moreover, normally nonprotective fibroblasts could be re-engineered to be neuroprotective with ectopic xCT overexpression indicating that xCT is a key step in the pathway to glutathione synthesis. Conversely, astrocytes and meningeal cells derived from sut/sut mice (xCT loss-of-function mutants) showed greatly reduced proliferation in culture attributable to increased oxidative stress and thiol deficiency, because growth could be rescued by the thiol-donor β-mercaptoethanol. Strikingly, sut/sut mice developed brain atrophy by early adulthood, exhibiting ventricular enlargement, thinning of the cortex, and shrinkage of the striatum. Our results indicate that xCT can provide neuroprotection by enhancing glutathione export from non-neuronal cells such as astrocytes and meningeal cells. Furthermore, xCT is critical for cell proliferation during development in vitro and possibly in vivo.
- system xc−
- phase 2 detoxification enzymes
- oxidative glutamate toxicity
- cystine deprivation
- oxidative stress
The major cellular antioxidant glutathione (GSH) is an important line of defense against oxidative stress, and its deficiency can sensitize the brain to injury (Meister and Anderson, 1983; Mizui et al., 1992; Bobyn et al., 2002). GSH synthesis is limited by availability of the sulfhydryl amino acid cysteine, which is readily oxidized to cystine in the extracellular milieu of the brain. High-affinity cystine uptake by the Na+-independent cystine–glutamate exchange transporter (system xc−) is a rate-limiting step for GSH synthesis in various brain cell types (Miura et al., 1992). In particular, fetal brain cells (Sagara et al., 1993a), immature cortical neurons (Murphy et al., 1990), oligodendrocyte precursors (Back et al., 1998), HT22 hippocampal cell line (Li et al., 1997), and gliomas (Chung et al., 2005) are uniquely vulnerable to cystine deprivation or competitive inhibition of system xc− by excessive extracellular glutamate concentrations (oxidative glutamate toxicity). Prolonged oxidative glutamate toxicity leads to gradual GSH depletion, oxidative stress, and apoptosis (Ratan et al., 1994).
Structurally, system xc− is composed of a light-chain subunit (xCT, encoded by the Slc7a11 gene), which confers substrate specificity (Sato et al., 1999), and a glycosylated heavy-chain subunit (4F2hc or rBAT) common to the transporter family (Mastroberardino et al., 1998; Wang et al., 2003). Basal xCT expression is highest at the CSF and blood–brain barrier, suggesting a role in redox buffering of the CSF and plasma (Sato et al., 2002). Importantly, xCT is also expressed in neurons and astrocytes of the cerebral cortex (Pow, 2001; Melendez et al., 2005; Burdo et al., 2006) and may contribute to the coupling of GSH and other sulfhydryl species between these cell types. Enhancement of astrocyte–neuron GSH coupling is coordinated by the stress-inducible transcription factor Nrf2, which upregulates xCT and other GSH synthesis/release machinery to constitute a defense mechanism against oxidative stress (Sasaki et al., 2002; Shih et al., 2003). In this GSH coupling pathway, astrocytes use xCT and other transport mechanisms to uptake cyst(e)ine for GSH synthesis (Cho and Bannai, 1990; Dringen et al., 2000; Wang and Cynader, 2000; Allen et al., 2002). GSH is then exported from astrocytes and degraded back to cysteine in the extracellular space for neuronal uptake. Mature neurons primarily uptake cysteine using system xAG (cysteine-permeable, Na+-dependent glutamate transporter) (Shanker et al., 2001; Chen and Swanson, 2003), whereas immature neurons exclusively uptake cystine via xCT (Murphy et al., 1990).
Although xCT function may be essential for GSH production by individual brain cell types, its role in GSH coupling within heterogeneous neuron–astrocyte populations is unknown. Furthermore, the role of xCT has been difficult to conclusively study because its antagonist pharmacology can overlap with glutamate receptors (Patel et al., 2004). Here, we tested the hypothesis that enhanced xCT activity is sufficient to confer neuroprotection by promoting GSH synthesis and delivery from non-neuronal support cells (astrocytes and meningeal cells) to immature neurons. Conversely, we examined whether brain cells derived from sut/sut mice, which express nonfunctional xCT, experience increased oxidative stress because of chronic impairment of cystine uptake.
Materials and Methods
All chemicals were purchased from Sigma Canada (Oakville, Ontario, Canada) unless stated otherwise.
Mammalian cell culture.
All experiments were approved by the University of British Columbia Animal Care Committee and were conducted in strict accordance with guidelines set by the Canadian Council on Animal Care. All rats were obtained from the University of British Columbia Animal Care Facility. Enriched astrocyte cultures were prepared from the cerebral cortices of postnatal day 0 (P0) to P2 Wistar rat pups using the papain dissociation method, as described previously (Shih et al., 2003). Enriched meningeal cultures from leptomeninges of P0–P2 Wistar rat pup brains (collected from the surface of cerebral and cerebellar cortices) and fibroblast cultures from the eviscerated bodies of embryonic day 18 (E18) Wistar rat fetuses (head removed) were similarly prepared using the papain dissociation method. All non-neuronal cell types were grown in culture for 7 d and used for experiments before 10 d in vitro (DIV). The various non-neuronal cells had different growth rates, with fibroblasts and meningeal cells proliferating approximately three times faster than astrocytes, based on MTT [(2)-3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] turnover and protein concentrations assays. Therefore, when comparing neuroprotection by non-neuronal cells, we seeded astrocytes at a three times higher density to achieve comparable cell densities at the time of glutamate exposure.
Immature cortical cultures were prepared from the cerebral cortices of E18 Wistar rat fetuses, as described in detail previously (Shih et al., 2003). Importantly, immature cortical cultures were first plated in B27-supplemented Neurobasal medium (NBM) at 1 × 106 cells/ml (Invitrogen, Gaithersburg, MD) to ensure neuronal health at an early stage. After 1 DIV, NBM was exchanged for minimum essential medium (MEM; Invitrogen) supplemented with 5.5 g/L d-glucose, 2 mm glutamine, 10% fetal bovine serum (FBS; HyClone, Logan, UT), 1 mm pyruvate, 100 U/ml penicillin, and 0.1 mg/ml streptomycin (MEM-pyr). This medium change was required to reduce excessive antioxidant levels from the B27-supplemented medium for subsequent oxidative challenge. At neuronal age 2 DIV, various non-neuronal cell types (collected 24 h after infection with adenovirus, for some experiments) were trypsinized and transplanted directly into naive (no contact with virus) immature cortical cultures. Buthionine sulfoximine (BSO) experiments in neuron–astrocyte coculture required a setup in which astrocytes were physically separated from neurons. This setup consisted of naive immature cortical cultures (2 DIV) prepared in 24-well plates and infected astrocytes separated in collagen-coated culture plate inserts (Millipore, Bedford, MA), as described in detail previously (Shih et al., 2003). At 3 DIV, toxicity treatments were initiated (see below). At 4 DIV, the cocultures were washed and fixed for quantification of neuronal viability (see below).
Human embryonic kidney 293 cells (HEK293; American Type Culture Collection, Manassas, VA) were grown and transiently transfected as described previously (Shih et al., 2003).
Plasmids and adenoviral constructs and infections.
Recombinant adenoviral vectors were constructed using the Cre-lox system (Canadian Stroke Network core facility, University of Ottawa, Ottawa, Ontario, Canada) (Hardy et al., 1997). Briefly, the mouse xCT cDNA (a gift from Dr. S. Bannai, University of Tsukuba, Tsukuba, Japan) was excised from the pcDNA3.1+ vector using the restriction enzymes NotI and EcoRI (Sato et al., 1999; Shih and Murphy, 2001). The cDNA was then subcloned into the adenovirus construct. C-terminal epitope-tagged xCT (HA-xCT) was generated by subcloning the xCT cDNA upstream and in frame with the hemagglutinin (HA) coding sequence within the GW1 vector (a gift from Dr. A. el-Husseini, University of British Columbia) using the HindIII and EcoRI restriction enzymes in the vector. Infections were performed as described previously (Shih et al., 2003). In this study, a multiplicity of infection of 200 plaque-forming units/cell was used. All viruses carried a green fluorescent protein (GFP) cDNA driven by its own cytomegalovirus promoter for monitoring of viral infection.
For all toxicity studies, the cortical cultures were used in their immature state (<4 DIV) when a full complement of ionotropic glutamate receptors was not yet expressed. Previous control studies confirmed that NMDA receptor-dependent excitotoxicity is not involved in the oxidative glutamate toxicity paradigm (Shih et al., 2003). MEM-pyr was replaced with MEM supplemented with 5.5 g/L d-glucose, 2 mm glutamine, 5% FBS, 100 U/ml penicillin, and 0.1 mg/ml streptomycin (MEM–5% FBS) containing the indicated concentrations of l-glutamate or H2O2. Cells were exposed to all toxins for 24 h before evaluation of neuronal viability.
Determination of neuronal viability.
Immature cortical cultures were immunostained with antibodies for anti-neuron-specific enolase (NSE), followed by Texas Red-conjugated secondary antibodies (see below). Neuronal viability was evaluated by manual counting of cells positively labeled for NSE, as described previously (Shih et al., 2003).
xCT antibody production.
The rabbit xCT polyclonal antibody was generated using a synthetic eight-branch multiple antigen peptide corresponding to a 15 aa sequence in the N-terminal region of the mouse xCT protein (VATISKGGYLQGNMS) (Szumlinski et al., 2004). Animals were immunized five times, and the final antiserum was purified by affinity-column chromatography using the original peptide. xCT-specific bands were absent in blots probed with preimmune serum and when the antibody was preadsorbed with peptide (data not shown).
Western blot analysis and immunocytochemistry.
Cell cultures were washed with PBS, collected in harvest buffer containing PBS with a commercial mixture of protease inhibitors (Roche Biochemicals, Burlington, NC), and sonicated for 10 s to make a crude lysate. For tissue collection from E18 rat brains, leptomeninges were first taken from the surface of the cerebral and cerebellar cortices. Major brain regions were then dissected out using tweezers. Tissue collection from adult rats was similar, except the entire brain was sliced into 1-mm-thick coronal sections to aid identification of structures. All tissue samples were collected on ice and homogenized in harvest buffer with 10 strokes of a Dounce homogenizer. For Western blot-positive controls, astrocyte cell lysates were collected 24 h after infection with Ad-xCT or Ad-Nrf2. Protein concentration was measured using the bicinchonic acid method (Pierce, Rockford, IL). Samples diluted in loading buffer (7 mg/ml dithiothreitol, 5% β-mercaptoethanol (β-ME), 6% SDS, 30% glycerol, 0.38 m Tris, pH 6.8, and pyronin Y) were denatured by boiling for 2 min before gel loading. Dithiothreitol and β-ME were omitted from the loading buffer for nonreducing conditions. For SDS-PAGE, 10% acrylimide gels were used to run all samples at 10 μg per lane. Antibody reactivity was detected using ECL chemiluminscence substrate (Amersham Biosciences, Piscataway, NJ).
For immunocytochemistry after toxicity experiments, cultures were washed three times with 37°C PBS and fixed with 2% paraformaldehyde for 10 min. Fixed cells were washed three times and incubated with primary antibodies overnight, followed by fluorescent secondary antibodies for 1 h at room temperature. Immunostained cells were washed three times and mounted in Fluoromount-G (Southern Biotechnology, Birmingham, AL).
The antibodies used in this study include the following: anti-GFP from mouse (1:1000; Boehringer Mannheim, Indianapolis, IN), anti-NSE from rabbit (1:2000; Polysciences, Warrington, PA), anti-mouse Alexa Fluor 488 from goat (1:2000; Invitrogen), anti-rabbit Texas Red from goat (1:2000; Invitrogen), anti-actin from goat (1:1000; Santa Cruz Biotechnology, Santa Cruz, CA), anti-HA (1:1000; Babco, Berkeley, CA), anti-xCT (1:250; custom made), anti-4F2hc (1:100; Santa Cruz Biotechnology), anti-rabbit horseradish peroxidase (HRP) from sheep (1:5000; Amersham Biosciences), anti-mouse HRP from sheep (1:5000; Amersham Biosciences), and anti-goat HRP from donkey (1:5000; Santa Cruz Biotechnology)
Total intracellular GSH assay and effluxed GSH assay.
Total intracellular and media GSH was quantified by the method of Tietze (1969). These methods were described in detail previously (Shih et al., 2003). The monochlorobimane (mCBi) (Invitrogen) plate reader assay of GSH content in astrocytes and meningeal cells was performed as described previously (Sun et al., 2005).
Radioactive substrate uptake assays.
xCT uptake activity was measured in cultured cells as described previously (Shih and Murphy, 2001; Shih et al., 2003). Briefly, cultures were preincubated in Na+-free HBSS for 10 min at 37°C and then incubated with Na+-free HBSS containing 1.6 Ci/ml l-[3H]glutamate or 1 Ci/mL l-[35S]cystine (Amersham Biosciences) and the indicated concentrations of any unlabeled competitors for an additional 20 min at 37°C. Then the cells were washed three times with ice-cold Na+-free HBSS and lysed with 0.5% Triton X-100 in 0.1 m phosphate buffer. Radioactivity was determined by liquid scintillation counting and normalized to protein concentration for each sample. For acute uptake studies, cells were papain dissociated from E18 Wistar rat meninges and cortex (with the cell culture protocol), immediately resuspended in Na+-free HBSS, and divided into Eppendorf tubes (1 × 106 cells/tube). The cells were pelleted at 2000 × g and resuspended in 300 μl of radioactive substrate solutions for 20 min at 37°C. Cells were washed by spinning down and resuspending three times with 1 ml of ice-cold Na+-free HBSS and lysed with 100 μl of 0.5% Triton X-100 in 0.1 m phosphate buffer. Perchloric acid was added (3% final concentration) to separate the acid-soluble and protein fractions. The protein precipitate was removed by centrifugation (14,000 × g for 10 min), and the acid-soluble fraction (containing free amino acids and GSH) was measured for radioactivity.
sut/sut mice breeding pairs were obtained from Dr. R. Swank (Roswell Park Cancer Institute, Buffalo, NY) (Chintala et al., 2005). The C3H/HeSnJ control background strain was obtained from The Jackson Laboratory (Bar Harbor, ME). All mice were maintained at the University of British Columbia Animal Care Facility in a 12 h light/dark cycle with food and water ad libitum.
Brain morphology measurements.
sut/sut and C3H/HeSnJ mice, ∼3.5 months of age, were deeply anesthetized with euthanyl (Bimeda-MTC) and perfused transcardially with room temperature PBS, followed by ice-cold 4% paraformaldehyde in PBS. The brains were postfixed overnight in paraformaldehyde, cryoprotected with 30% sucrose for 2 d, and cryosectioned. Cresyl violet staining was performed with standard protocols on 40 μm sections mounted on SuperFrost Plus slides (Fisher Scientific, Houston, TX) and scanned at 600 dpi on a desktop scanner (Epson 1660). Brain measurements were performed on scanned images using Image J software (version 1.33u; National Institutes of Health, Bethesda, MD). Ventricle, striatum, and hippocampal areas were measured bilaterally and averaged for each bregma. The measured hippocampal area encompassed all major layers including the stratums oriens, pyramidale, radiatum, moleculare, and dentate gyrus. Cortical width was also measured bilaterally between the apex of the corpus callosum to the pial layer of the cortex (primary motor cortex region).
All experiments were repeated at least three times. Results are presented as the mean ± SEM. Statistical analysis of raw data was performed with Prism 2.0 (Graph Pad, San Diego, CA). Experimental groups were compared by one-way ANOVA, two-way ANOVA, or Student's t test (*p < 0.05, **p < 0.01, and ***p < 0.001).
Characterization of an anti-xCT polyclonal antibody
To examine xCT protein expression in rodent brain and to validate our overexpression assays, we developed a polyclonal antibody specific for N-terminal amino acids (7–22) of the mouse xCT protein, a region predicted to be both hydrophilic and intracellular by membrane topology analysis (Fig. 1A) (Sato et al., 1999; Szumlinski et al., 2004; Melendez et al., 2005). Antibody specificity was first verified by probing astrocyte cultures overexpressing xCT (Ad-xCT) or its transcriptional regulator Nrf2 (Ad-Nrf2) in Western blot (Sasaki et al., 2002; Shih et al., 2003). With both Ad-xCT- and Ad-Nrf2-infected cultures, the antibody detected three bands corresponding to xCT at 35, 55, and 105 kDa (Fig. 1B, top blot). These bands were barely detectable in Ad-GFP-infected controls. The 35 kDa band corresponded to the monomeric form of xCT. The 105 kDa band, which was detectable using both xCT and 4F2hc antibodies, was only present under nonreducing conditions and represented the xCT plus 4F2hc disulfide-linked heterodimer. Monomeric 4F2hc migrated at 80 kDa under reducing conditions (Fig. 1B, bottom blot) (Burdo et al., 2006). Interestingly, Nrf2 overexpression increased 4F2hc protein expression, suggesting possible xCT and 4F2hc coregulation in astrocytes (Gochenauer and Robinson, 2001). A fainter 55 kDa band (xCT-mod) was consistently detected in xCT- and Nrf2-overexpressing astrocytes but was unlikely to be an xCT homodimer because it was too low in molecular weight and could not be disrupted by reducing conditions (Fig. 1A, top blot). The density of this band was increased with xCT and Nrf2 overexpression, unlike the nonspecific band at 95 kDa, suggesting that it is derived from the xCT transcript and may be an alternatively spliced or translationally modified form distinct from 35 kDa xCT (Kim et al., 2001). No heterodimerization was observed between 4F2hc and xCT-mod.
To further verify the migration of xCT at 35 kDa and 55 kDa, we overexpressed HA-tagged xCT (HA-xCT) in HEK293 cells and examined its migration in Western blot using an HA-specific antibody. Consistent with adenoviral xCT and Nrf2 overexpression, the majority of HA-xCT was detected at 35 kDa, and a minor component was observed at 55 kDa. Furthermore, the 105 kDa HA-xCT-4F2hc heterodimer could be disrupted under reducing conditions, but not the 55 kDa band (supplemental Fig. 1A, available at www.jneurosci.org as supplemental material).
In vivo regional distribution of xCT protein
The antibody was used to determine the regional tissue distribution of xCT protein in the immature (E18) and adult rat brain by Western blot. No immunostaining was performed because of the possibility that the nonspecific 95 kDa band may interfere. In the immature brain, both 35 kDa xCT and xCT-mod were robustly expressed in the meninges (Fig. 1C) (Sato et al., 2002; Burdo et al., 2006). xCT-mod was predominantly expressed in major regions of the immature brain including the cortex, hippocampus, striatum, and cerebellum. In adult tissues, meningeal expression of 35 kDa xCT and xCT-mod remained high (Fig. 1D). However, a developmental transition in xCT protein was observed because 35 kDa xCT was now predominant in all major brain regions, whereas xCT-mod expression decreased. Consistent with mRNA analysis studies by Sato et al. (2002), high levels of xCT protein expression were observed in the adult periventricular region containing the ependymal cells of the lateral ventricles, but no protein was detected in liver.
Further confirming xCT transcriptional modulation by Nrf2, meningeal tissues from adult Nrf2−/− mice showed decreased expression of the 35 kDa xCT band, compared with wild-type littermates (supplemental Fig. 1B, available at www.jneurosci.org as supplemental material) (Chan et al., 1996; Sasaki et al., 2002). A modest reduction in xCT protein was also observed in the cortex, but not in other brain tissues examined. The expression level of xCT was also verified using Na+-independent l-[35S]cystine and l-[3H]glutamate uptake assays (both can be used as substrates) on enzymatically dissociated E18 rat meningeal and cortical tissue. Indeed, Na+-independent l-[35S]cystine uptake was 23.2 ± 5.7-fold higher in meningeal cells than in cortical cells and was completely inhibited by 1 mm nonradiolabeled glutamate. l-[3H]Glutamate uptake was 6.4 ± 2.1-fold higher in meningeal cells than in dissociated cortex and was completely inhibited by 1 mm nonradiolabeled cystine. The relative difference between meningeal and cortical tissues was higher when using cystine as a substrate and may have been related to higher basal levels of glutamate uptake by other transporters.
Meningeal cells exhibit enhanced GSH synthesis and neuroprotection of immature neurons
The immature brain contains high levels of GSH compared with the adult brain, which may be important for cell proliferation and neuronal viability during development (supplemental Fig. 2A,B, available at www.jneurosci.org as supplemental material) (Lowndes et al., 1994). We hypothesized that xCT-mediated GSH production could serve an important neuroprotective role in the immature brain and that augmentation of this pathway could confer increased protection from oxidative stress. We examined this possibility using relatively purified primary culture systems. Neuroprotection in culture was specifically defined as an increased viability of NSE-positive immature neurons after toxicity treatment.
Consistent with higher xCT activity in meningeal tissue, cultured meningeal cells showed enhanced maximal fluorescence (Fmax) when assayed with a selective enzyme-linked marker for GSH, monochorobimane (mCBi) (Fmax: meninges, 1.54 ± 0.19 vs astrocytes, 1.08 ± 0.15; *p < 0.05, Student's t test) (Chatterjee et al., 1999; Sun et al., 2005), indicating increased GSH content compared with cultured cortical astrocytes (Fig. 2A). Indeed, cultured meningeal cells exhibited an approximate threefold higher l-[35S]cystine uptake than astrocytes. Meningeal l-[35S]cystine uptake fit the pharmacological profile of system xc− because it was strongly blocked by its competitive inhibitors l-glutamate, l-quisqualic acid, and homocysteic acid, partially blocked by l- and d-aspartate, but not blocked by kainic acid (Fig. 2C). Accordingly, application of cystine to the mCBi assay led to a more pronounced upregulation in GSH synthesis (∼10-fold greater slope) in meningeal cells compared with cortical astrocytes (slope of fluorescence increase: meninges, 0.096 ± 0.018 vs astrocytes, 0.009 ± 0.002; *p < 0.05, Student's t test) (Fig. 2B).
In our previous studies, we found that GSH release from Nrf2-overexpressing astrocytes led to potent neuroprotection of immature neurons during oxidative glutamate toxicity (Shih et al., 2003). Meningeal cells naturally exhibit high levels of Nrf2 activity (A. Y. Shih and T. H. Murphy, unpublished observation), which may explain their high level of xCT expression, GSH content, and GSH synthesis. Thus, we reasoned that meningeal cells would be highly neuroprotective. We cocultured various densities of meningeal cells, astrocytes, and peripheral fibroblasts with immature cortical neuron cultures and exposed the cocultures to oxidative glutamate toxicity (3 mm glutamate for 24 h). Indeed, cultured meningeal cells offered widespread neuroprotection and were significantly more protective than astrocytes seeded at the same densities (2.5–10% of the total cell number) (Fig. 3). In contrast, peripheral fibroblasts were relatively non-neuroprotective at any plating density tested. Glutamate exposure caused no obvious meningeal cell, astrocyte, or fibroblast death.
xCT overexpression is sufficient to enhance GSH synthesis and confer neuroprotection from oxidative stress
Because Nrf2 upregulates numerous antioxidant/detoxification genes (in addition to those involved in GSH production), neuroprotection by meningeal cells could be attributable to other Nrf2-regulated pathways. Thus, we examined whether increased xCT activity alone was sufficient to confer neuroprotection from oxidative stress using ectopic xCT overexpression. 4F2hc was not concomitantly overexpressed because it had the potential to affect the transport activities of other glycoprotein-associated transporters and may be unnecessary because of adequate endogenous 4F2hc expression (Shih and Murphy, 2001). Ad-xCT-infected astrocyte cultures exhibited fivefold higher Na+-independent l-[3H]glutamate uptake over Ad-GFP-infected controls, which fit the pharmacological profile defined for system xc−, as described above (Fig. 4A). Ad-xCT infection of astrocytes also increased total intracellular GSH by ∼1.7-fold and GSH in the astrocyte-conditioned medium by ∼1.6-fold compared with Ad-GFP control (Fig. 4B,C).
To examine the neuroprotective potential of increased xCT activity, we transplanted Ad-GFP- or Ad-xCT-infected astrocytes into naive (no virus treatment) immature cortical neuron cultures at 0–2% of the total cell number (Shih et al., 2003). The use of astrocyte transplantation avoids direct infection of neurons and isolates the specific contribution of infected astrocytes. These cocultures were then exposed to oxidative glutamate toxicity. Ad-xCT-infected astrocytes transplanted at only 2% of the total cell number conferred dose-dependent and widespread neuroprotection (Fig. 4D,E). At these seeding densities, Ad-GFP-infected astrocytes offered no neuroprotection. GSH synthesis was necessary for neuroprotection because pretreatment of Ad-xCT-infected astrocytes with the selective GSH synthesis inhibitor BSO (in separate membrane-delimited inserts) completely blocked neuroprotection (Fig. 4F). Importantly, Ad-xCT-infected astrocytes also conferred neuroprotection during direct exposure to the reactive O2 species, H2O2, which does not rely on xCT inhibition to induce oxidative stress (Fig. 4G). Strikingly, Ad-xCT infection was even sufficient to enhance neuroprotective potential of fibroblasts, a cell type that is not associated with CNS neuroprotection (Fig. 5A,B). Fibroblasts infected with the Ad-GFP control virus were not neuroprotective.
Loss of xCT function inhibits astrocyte and meningeal cell proliferation in vitro and leads to brain atrophy in vivo
Recent studies have selectively linked the mouse subtle gray pigmentation mutant phenotype (sut/sut) to a truncation mutation in the xCT gene (Slc7a11), leading to expression of a nonfunctional xCT protein (Chintala et al., 2005). Melanocytes isolated from sut/sut mice have extremely low levels of GSH and cannot be grown under normal culture conditions unless in the presence of β-ME, a reducing agent (Chintala et al., 2005). β-ME circumvents loss of xCT function by decreasing oxidative stress and reducing extracellular cystine to cysteine for uptake by other non-xCT transport systems. Consistent with these results, we found that astrocytes cultured from sut/sut mouse pups failed to proliferate unless cultured with β-ME from the day of plating (0 DIV) (Fig. 6). In contrast, astrocytes derived from the control C3H/HeSnJ (+/+) background strain proliferated normally with and without β-ME (Fig. 6). Meningeal cells from sut/sut mice were also unable to proliferate without β-ME (data not shown). Interestingly, our preliminary results suggest that β-ME is only necessary during the initial stages of culturing. Removal of β-ME after 1 week of culture did not affect sut/sut astrocyte viability or GSH content (data not shown), suggesting that immature proliferating astrocytes rely on xCT, whereas mature astrocytes adapt to use other mechanisms of cyst(e)ine uptake or induce antioxidant factors that allow them to cope with stressful conditions. Unfortunately, the requirement of β-ME early in culture to prevent astrocyte cell death would make it difficult to interpret the role of xCT loss of function in direct protection of neurons in a coculture experiment (as in Figs. 4 and 5).
Our results show that xCT is widely expressed in the mature brain and its function is important for astrocyte and meningeal cell proliferation in vitro. We therefore reasoned that loss of xCT activity could alter the development of the sut/sut mouse brain. We compared gross brain morphology on coronal sections between adult sut/sut and +/+ mice (Fig. 7A). When examined at comparable ages (+/+, 13.5 ± 1.2 weeks; sut/sut, 13.7 ± 0.1 weeks), sut/sut mice exhibited no difference in body weight compared with controls (+/+, 28.0 ± 0.5 g; sut/sut, 28.1 ± 1.7 g). However, brain size was significantly reduced indicating atrophy in certain brain tissues (Fig. 6B) (reduced hemispheric area; **p < 0.0057). Indeed, sut/sut mice exhibited pronounced enlargement of the ventricles (Fig. 6C) (***p < 0.0001), accompanied by thinning of the cortex (Fig. 6D) (*p < 0.0059) and shrinkage of the striatum (Fig. 6E) (*p < 0.023). A trend toward shrinkage of hippocampus was observed, particularly in the posterior regions. However, this difference was not significant in statistical analysis (Fig. 6F) (p = 0.34).
We show that enhanced xCT activity in astrocytes is sufficient to increase GSH synthesis/release and protect cocultured neurons during oxidative stress-based toxicity in vitro. Meningeal cells naturally exhibit high levels xCT-mediated GSH production and may provide a reservoir for GSH in vivo. Fibroblasts, a cell type not associated with CNS neuroprotection, could be re-engineered to be neuroprotective with xCT overexpression. Our results also reveal a role for xCT in brain cell proliferation in vitro and possibly in vivo. Astrocytes and meningeal cells derived from sut/sut mice show greatly reduced proliferation because of increased oxidative stress. sut/sut mice show brain atrophy by early adulthood, exhibiting enlarged ventricles, cortical thinning, and striatal shrinkage.
Role of xCT at the CSF–brain barrier
xCT protein was most highly expressed at structures forming the CSF–blood–brain barrier (meninges and periventricular zone), in agreement with previous studies (Sato et al., 2002; Burdo et al., 2006). Based on this unique localization, Bannai and colleagues (Sato et al., 2002) proposed a role for xCT in recycling cystine to cysteine in the CSF and plasma to maintain redox balance. This is plausible because cystine is the predominant sulfhydryl species extracted from plasma by brain (Wang and Cynader, 2000). Furthermore, cystine levels in the CSF and plasma are very low (micromolar range), suggesting that a high-affinity transport system such as xCT would be necessary (Gjessing et al., 1972; Perry et al., 1975). Indeed, targeted deletion of the mouse xCT gene led to higher cystine/cysteine ratios in blood plasma in vivo (Sato et al., 2005).
Consistent with xCT localization, our previous examination of GSH distribution in brain slices also revealed the highest GSH levels in cells of the leptomeninges and ventricular ependyma, even when compared with cortical astrocytes (Sun et al., 2006). Because GSH efflux is dependent on intracellular GSH content, these sites may support neuronal viability by releasing GSH to distant sites within the brain (Sagara et al., 1996). Here we reconstitute this mechanism in vitro, with meningeal cells conferring widespread protection of immature neurons in a coculture system. Although there are several steps in the pathway to GSH synthesis, we show that xCT is rate limiting because xCT overexpression alone (in astrocytes or fibroblasts) was sufficient to increase GSH synthesis/release and confer neuroprotection.
Role of xCT in astrocyte–neuron GSH coupling and neuroprotection
Our data also suggest that xCT protein is widely expressed in other brain structures including the cortex, hippocampus, striatum, and cerebellum, albeit at much lower levels than meningeal tissue. For this work, we used Western blot because we could confidently distinguish bands representing xCT from those that were nonspecific using standards derived from cultures overexpressing xCT. Although our antibody exhibited nonspecific reactivity in Western blots (95 kDa band), there is no evidence that this is a problem for in situ immunostaining performed in other studies. With caveats in mind, we had previously detected xCT expression primarily in neurons of the prefrontal cortex (Melendez et al., 2005). However, in a more detailed analyses with another peptide-directed xCT antibody, Burdo et al. (2006) also found xCT expression in both MAP2 (microtubule-associated protein-2)-positive neurons and GFAP-positive astrocytes of the adult mouse cortex. A large body of in vitro data show that system xc− activity can exist in mature astrocytes and immature neurons (Murphy et al., 1990; Sagara et al., 1993a,b; Kranich et al., 1998; Schubert and Piasecki, 2001; Lewerenz et al., 2003). We also detected significant xCT-dependent l-[35S]cystine uptake in cultured astrocytes (Fig. 2C). However, cyst(e)ine supply remains a limiting factor for GSH synthesis in astrocytes because xCT overexpression continues to increase GSH synthesis/release. Similarly, because cystine (and not cysteine) is most readily available in culture media, it is logical to expect that increased xCT activity in mature neurons would promote neuronal GSH synthesis and neuroprotection. Unfortunately, we could not perform this experiment using Ad-xCT because the adenovirus has poor tropism for mature neurons.
Given that astrocyte-specific enhancement of xCT activity promotes neuroprotective GSH coupling between astrocyte and immature neurons in a coculture system, targeted and controlled xCT expression/induction may be a strategy to achieve neuroprotection in vivo. Although xCT is only one of many Phase 2 genes regulated by Nrf2, our in vitro results show that xCT overexpression alone is sufficient to partially recapitulate the neuroprotective potential of Nrf2, at least with the toxicity paradigms we have examined (Ishii et al., 1999; Sasaki et al., 2002; Shih et al., 2003). To dissociate the role of xCT-mediated neuroprotection in vivo, future studies could examine whether loss of xCT function in sut/sut mice blunts the effect of Nrf2-inducing electrophilic agents during in vivo toxicity and injury models (Shih et al., 2005a,b; Satoh et al., 2006).
Controlling glutamate release with enhanced xCT activity
Increasing cystine–glutamate exchange can potentially lead to release of excitotoxic amounts of glutamate because extracellular cystine is exchanged for intracellular glutamate (Warr et al., 1999). For example, glioma cells exhibit particularly high amounts of system xc− activity but low Na+-dependent glutamate transport activity (Ye et al., 1999; Chung et al., 2005). In these extreme circumstances, release of glutamate by xCT and its excitotoxic effect on surrounding neurons is thought to facilitate growth of gliomas in vivo. It is also conceivable that gliomas require high xCT activity because they are under continuous oxidative stress as a result of their rapid growth. xCT activity in microglia and macrophages may also increase cytotoxicity through release of glutamate (Piani and Fontana, 1994; Barger and Basile, 2001). Thus, induction of xCT activity as a neuroprotective strategy should be targeted to enhance the neuro-supportive role of astrocyte/meningeal cells, in which glutamate efflux may be better controlled. There is evidence suggesting that xCT expression is normally coregulated with high-affinity Na+-dependent glutamate transporters that reuptake glutamate released by xCT to both minimize extracellular accumulation and maintain adequate glutamate driving force for xCT activity (Lewerenz et al., 2006). Indeed, control mechanisms may exist in vivo because astrocytes are well known to express high-affinity glutamate transporters GLAST and GLT-1 (Danbolt, 2001).
Role of xCT in brain cell growth and proliferation
The rate of cell proliferation is strongly associated with cyst(e)ine availability and intracellular GSH content (Godwin et al., 1992; Noda et al., 2002). For example, continuous proliferation of cancer cells is associated with high oxidative load, and pharmacological blockade of xCT inhibits growth of various cancer cell types by depleting GSH (Uren and Lazarus, 1979; Gout et al., 2001; Chung et al., 2005). Our results show that astrocytes and meningeal cells derived from sut/sut mice were unable to grow in vitro without the presence of β-ME (Chintala et al., 2005). However, unlike other cell types derived from xCT-deficient mice (i.e., melanocytes and fibroblasts) (Chintala et al., 2005; Sato et al., 2005), subsequent removal of β-ME after 7 d of growth did not decrease GSH content or cell viability. Thus, astrocytes may adapt to have complementary cyst(e)ine transport mechanisms during maturation (i.e., system xAG and ASC) (Cho and Bannai, 1990; Bender et al., 2000; Allen et al., 2002; McBean, 2002). It is also possible that mature astrocytes are quiescent and do not require high levels of GSH. Accordingly, GSH content and viability of mature astrocytes are primarily unaffected by treatment with millimolar levels of glutamate or sulfasalazine, two treatments known to block xCT transport (Shih et al., 2003; Chung et al., 2005).
Interestingly, immature neurons and oligodendrocyte precursors are also highly sensitive to cystine deprivation in vitro (Murphy et al., 1990; Sagara et al., 1993a; Back et al., 1998), suggesting a critical developmental role for xCT that may apply to multiple brain cell types. We previously hypothesized that high xCT expression and GSH content in the lateral ventricular ependyma may be uniquely positioned to support proliferation of neuronal progenitors in the subventricular zone in vivo (Sun et al., 2006). Studies using the transplantation of neuronal stem cells may benefit from xCT-dependent manipulation of high-affinity cystine uptake for GSH synthesis.
We also observed brain atrophy in sut/sut mice by early adulthood. Future studies will examine when these morphological changes begin and what brain regions/cells are most dependent on xCT in vivo. Because immature brain cells appear to be most sensitive to xCT blockade in vitro, we speculate that atrophy begins at the embryonic to perinatal stage (Murphy et al., 1990). Loss of xCT function may impede GSH synthesis necessary for detoxifying metabolic byproducts and oxidative stressors produced during cell proliferation and maturation, possibly leading to cell death during development (Lowndes et al., 1994). Notably, the morphological changes observed in sut/sut mice brains bear remarkable similarities to aged EAAC1 knock-out mice, where brain atrophy is caused by age-dependent thiol depletion and toxicity of mature neurons. The fact that we see obvious differences at 3.5 months, a relatively young age compared with the aged EAAC1 knock-out mice examined in their study (11 months), further suggests that disturbances occur early in development (Aoyama et al., 2005).
A.Y.S. was supported by studentships from the Canadian Institutes of Health Research and the Michael Smith Foundation for Health Research (MSFHR). T.H.M. was supported by a grant in aid from the Heart and Stroke Foundation of British Columbia and Yukon and is an MSFHR senior scholar and member of the Canadian Stroke Network. We thank Lei Jiang for technical assistance and Dr. Richard Swank (Roswell Park Cancer Institute, Buffalo, NY) for sending us the sut/sut mice. We also thank the staff at the University of British Columbia Animal Research Unit and Steve Callaghan and Dominique Vaillant at the Adenovirus Core Facility, University of Ottawa (Ottawa, Ontario, Canada), for adenovirus production.
- Correspondence should be addressed to Dr. Timothy H. Murphy, 4N1-2255 Wesbrook Mall, University of British Columbia, Vancouver, British Columbia, Canada V6T 1Z3.