Main path and byways: non-vesicular glutamate release by system x_c⁻ as an important modifier of glutamatergic neurotransmission

Abstract

System x_c⁻ is a cystine/glutamate antiporter that exchanges extracellular cystine for intracellular glutamate. Cystine is intracellularly reduced to cysteine, a building block of GSH. As such, system x_c⁻ can regulate the antioxidant capacity of cells. Moreover, in several brain regions, system x_c⁻ is the major source of extracellular glutamate. As such this antiporter is able to fulfill key physiological functions in the CNS, while evidence indicates it also plays a role in certain brain pathologies. Since the transcription of xCT, the specific subunit of system x_c⁻, is enhanced by the presence of reactive oxygen species and inflammatory cytokines, system x_c⁻ could be involved in toxic extracellular glutamate release in neurological disorders that are associated with increased oxidative stress and neuroinflammation. System x_c⁻ has also been reported to contribute to the invasiveness of brain tumors and, as a source of extracellular glutamate, could participate in the induction of peritumoral seizures.

Two independent reviews (Lewerenz et al. 2013, Bridges et al. 2012), approached from a different perspective, have recently been published on the functions of system x_c⁻ in the central nervous system. In this review, we highlight novel achievements and insights covering the regulation of system x_c⁻ as well as its involvement in emotional behavior, cognition, addiction, neurological disorders and glioblastomas, acquired in the past few years.

Graphical abstract

Introduction

System x_c⁻ or the cystine/glutamate antiporter, composed of a heavy chain subunit 4F2hc (encoded by the slc3a2 gene) and a light chain specific subunit xCT (encoded by the slc7a11 gene), exchanges glutamate for cystine in a 1:1 ratio and according to the respective concentration gradients (figure 1). Under physiological conditions, cystine is imported and intracellularly reduced to cysteine, a building block of the antioxidant GSH. In vitro, cystine supply via system x_c⁻ is crucial for survival of certain cell types as they can only survive in the absence of system x_c⁻ when the medium is supplemented with reducing agents (Sato et al. 2005). In vivo, however, it has been shown that genetic deletion of system x_c⁻ does not necessarily lead to any gross abnormality in the CNS nor are there signs of increased oxidative stress since other sources of GSH can be supplied by different cell-types to sensitive cells (Massie et al. 2011, De Bundel et al. 2011, Sato et al. 2005). While cystine is imported, glutamate is obligatorily exported and system x_c⁻ has been identified as the major source of extracellular glutamate in several rodent brain regions (De Bundel et al. 2011, Massie et al. 2011, Baker et al. 2002b). Glutamate released via system x_c⁻ physiologically modulates synaptic transmission via activation of pre- and postsynaptic metabotropic glutamate receptors located in the vicinity of the synaptic cleft (Baker et al. 2002b). Moreover, it was recently shown that glutamate released via system x_c⁻ regulates glutamatergic synapse strength by reducing the number of postsynaptic alpha-amino-3-hydroxy-5-methylisoxazole-4-propionate (AMPA) receptors (Williams & Featherstone 2014). Additionally, this glutamate could also activate extrasynaptic NMDA receptors, and as such in high concentrations may induce excitotoxicity (Hardingham & Bading 2010).

Figure 1. System x_c⁻.

System x_c⁻ is composed of the 4F2 heavy chain (4F2hc) and the light chain, xCT, which are linked by a disulfide bond (-S-S-). System x_c⁻ imports cystine in exchange for glutamate. Cystathionine represents another transported substrate for system x_c⁻ of which it is not known whether it is preferentially imported or exported from neuronal cells. Homocysteic acid and α-amino adipate are other potentially relevant substrates in vivo. System x_c⁻ activity can be modulated by phosphorylation of xCT (-P).

System x_c⁻ is expressed in the brain parenchyma and at high levels in the meninges and the ependyma (Shih et al. 2006, Sato et al. 2002). The CNS cell types that contribute most to system x_c⁻ activity (measured by cystine uptake in acute brain slices (Xi et al. 2002) or by microdialysis in living animals (Massie et al. 2011, De Bundel et al. 2011, Baker et al. 2002a) and as xCT protein levels assessed by western blotting (Shih et al. 2006)) is not known in detail as high-sensitivity in situ hybridization data and reliable antibodies for immunohistochemistry are missing. Importantly, using antibodies that are supposed to bind xCT often leads to confusing and contradicting data. Whereas in western blot analyses many antibodies label a band of 50kDa, corresponding to the theoretical molecular weight of xCT (Sato et al. 1999), it has been shown that xCT – when proper controls are applied and depending on the source and the gel system used - migrates at a considerable lower molecular weight ranging from 35 to 45kDa (Massie et al. 2008, Shih et al. 2006, Lewerenz et al. 2012). Moreover, xCT antibodies are often used for immunohistochemistry. Of note, with most antibodies the staining is merely identical in xCT wildtype compared to knock-out tissues, indicating prominent unspecific binding of these antibodies (Mesci et al. 2015, Massie et al. 2008). In a recent study, only a single xCT antibody of a total of 53 antibodies tested, resulted in specific xCT staining and exclusively on acetone-fixed sections. The poor morphology of acetone-fixed sections did not allow analyzing the detailed cellular distribution of xCT-specific immunoreactivity. Yet, in certain brain regions neurons could be clearly distinguished. In addition, in line with previous findings obtained by non-radioactive in situ hybridization (Sato et al. 2002), strong labeling was observed in meninges and the ependymal cells of the choroid plexus (Van Liefferinge et al., in revision at J. Comp. Neurol.).

Also, the use of primary cell cultures to test for the relative expression of xCT and system x_c⁻ activity is inherently prone to give results that are likely to be very different from the in vivo situation as xCT is heavily induced under regular cell culture conditions (reviewed in (Lewerenz et al. 2013)). Still, an increased activity of system x_c⁻ in meningeal cells compared to astrocytes in vitro has been reported, suggesting that some fundamental differences in xCT expression, as described above, are preserved ex vivo (Shih et al. 2006). Whereas Jackman et al. observed highest system x_c⁻-dependent cystine uptake in microglia, compared to astrocytes and neurons (i.e. microglia>astrocytes>neurons) (Jackman et al. 2010), Resch et al. compared either cortical astrocyte- or neuron-enriched to microglial primary cell cultures and concluded that system x_c⁻ activity in astrocytes is higher relative to neurons and microglia (Resch et al. 2014). Recent evidence using RT-PCR from laser-microdissected motor neurons in comparison to whole spinal cord indicated that at least this type of neuron does not express (or only traces of) xCT mRNA in vivo (Mesci et al. 2015). In contrast, microglia purified from spinal cord showed much higher xCT mRNA levels than whole spinal cord, indicating that an important contribution of the system x_c⁻ activity or xCT protein detected in the brain could be of microglial origin.

Given the widespread distribution of system x_c⁻ in the CNS and the contribution of glutamate released via system x_c⁻ to glutamatergic neurotransmission, it is evident that manipulation of this antiporter can significantly affect brain function under physiological as well as pathological conditions. Due to the absence of specific, blood-brain barrier-permeable inhibitors of system x_c⁻, most of the information about the function of system x_c⁻ is obtained from xCT knock-out (xCT^−/−) mice with a C57BL/6J background, descendants of the strain originally described by Sato and co-workers (Sato et al. 2005), or ‘subtle grey (sut)’ mice with a C3H/HeSnJ background that carry a spontaneous mutation resulting in a large 480kb deletion, including the last exon of slc7a11 (Chintala et al. 2005). Of note, conflicting data are often reported by using both system x_c⁻-deficient mouse strains. Whereas in the xCT^−/− mice, no anatomical or neurochemical changes are seen in the brain, besides the strong decrease in extracellular glutamate levels (De Bundel et al. 2011), sut/sut mice have been reported to develop brain atrophy after 15 weeks of age (Shih et al. 2006). Of note, Hewett and colleagues – who obtain sut/sut mice via heterozygous breeding - do not replicate these latter findings (SJH, unpublished observations). Whether the differences between xCT^−/− mice and sut mice are thus background-specific, mutation-specific or due to genetic drift, has not been clarified yet.

For an extensive review on system x_c⁻ in health and disease, we refer to our previous review (Lewerenz et al. 2013). In the current review, we will discuss recent breakthroughs and novel findings concerning the involvement of system x_c⁻ in normal and pathological brain functioning with a focus on the transcriptional regulation of xCT expression, the role of system x_c⁻ in mood disorders and cognition, the potential of system x_c⁻ as a target in the development of new treatments for neurological disorders, the role of system in drug addiction and the link between system x_c⁻, brain tumor development and peritumoral seizures.

New insights in the regulation of xCT expression and system x_c⁻ activity

As reviewed in detail previously (Lewerenz et al. 2013), system x_c⁻ is regulated on the levels of transcription (figure 2), protein insertion into the membrane and via the availability of counter-transported substrates (figure 1). Despite the fact that in cell culture, system x_c⁻ is prominently induced (probably because the ambient oxygen pressure is much higher in cell culture than in vivo) cell culture systems including primary cells as well as neuronal cell lines have been successfully employed to elucidate the pathways that regulate xCT expression and system x_c⁻ function. The two transcription factors that have a predominant role in the regulation of xCT expression are nuclear factor (erythroid-derived 2)-like 2 (Nrf2) and activating transcription factor 4 (ATF4). Both of these explain some, maybe most, of the inducibility of system x_c⁻ activity by a plethora of cellular insults (figure 2).

Figure 2. Transcriptional regulation of xCT expression.

A variety of stimuli, e.g. electrophiles, heavy metals, and reactive oxygen species (ROS), lead to activation of the nuclear factor NF-E2-related factor 2 (Nrf2), which binds to the electrophile response element (EpRE) within the xCT promoter region and activates transcription. Nrf1 inhibits the action of Nrf2. Amino acid (AA) starvation leads to phosphorylation of eIF2a (peIF2α), which via GCN2 leads to the translational up-regulation of the transcription factor activating transcription factor 4 (ATF4). ATF4 activates the transcription of xCT by binding to the amino acid response element (AARE) contained in the xCT promoter. Alternatively, growth factors and neuronal activity can induce xCT promoter activity via phosphoinositol-3 kinase (PI3K), glycogen synthase-3β (GSK-3β) and the GCN2/eIF2α/ATF4 module. STAT3/STAT5A negatively regulate the xCT promoter activity via a GAS element. Interleukin-1β (IL-1β) activates the xCT promoter by binding to the interleukin 1 receptor (Il-1R) and NF-κB. Tumor necrosis factor α (TNFα) and erythropoietin (EPO) increase the transcription of xCT through unknown signaling pathways. Bacterial lipopolysaccharides (LPS) and polyinosinic-polycytidylic acid (Poly I:C) activate the xCT promoter presumably via TLR4 and TLR3, respectively. PACAP and VIP increase cAMP and thereby activate xCT transcription. MicroRNA-26b directly targets xCT mRNA.

Induction of xCT through Nrf2

Nrf2 is induced by oxidative stress and other stimuli (reviewed in (Kensler et al. 2007)) and binds to an antioxidant response element (ARE) in the xCT promoter region (Sasaki et al. 2002). It has been documented that Nrf2 regulates system x_c⁻ activity in astrocytes (Shih et al. 2003). In addition, there is evidence that the neuroprotective β-lactam antibiotic ceftriaxone induces xCT expression via Nrf2 in astrocytes and motor neurons (Lewerenz et al. 2009). Before discovery of the effect of ceftriaxone on xCT, neuroprotective properties of this compound were ascribed to the enhancement of glutamate reuptake by the excitatory amino acid transporter 2 (EAAT2/GLT-1) (Rothstein et al. 2005). In murine retina, the sigma receptor 1 regulates xCT protein expression via Nrf2 in Müller cells, the radial glia of the retina (Wang et al. 2015, Ha et al. 2014). In contrast, Nrf1 constitutively binds to the ARE in the xCT promoter acting as a transcriptional suppressor of Nrf2 action, at least in the liver and in fibroblasts (Tsujita et al. 2014). It seems possible that Nrf1 is also involved in the regulation of xCT expression in the brain although experimental evidence is lacking.

Induction of xCT through ATF4

Via the four different kinases of the eukaryotic initiation factor 2α (eIF2α), i.e. protein kinase R (PKR), heme-regulated eIF2α kinase (HRI), PKR-like endoplasmic reticulum kinase (PERK) and general control non-depressible-2 (GCN2), signals as diverse as endoplasmic reticulum stress, amino acid starvation, infection and radiation are relayed via eIF2α phosphorylation and subsequent translational upregulation of ATF4 (reviewed in (Wek et al. 2006)). ATF4 plays a major role in the resistance of neuronal cell lines against oxidative stress via direct transcriptional upregulation of xCT (Lewerenz et al. 2012, Lewerenz & Maher 2009), ATF4 being able to bind an amino acid response element (AARE) in the xCT promotor (Sato et al. 2004). However, xCT expression and system x_c⁻ activity can also be upregulated by growth factor signaling. Recently, Lewerenz et al. could demonstrate that growth factor signaling induces xCT expression and system x_c⁻ activity via glycogen synthase kinase 3β, GCN2 and subsequently ATF4 (figure 2). This pathway upregulates system x_c⁻ in response to robust neuronal activity in vitro. In addition, markers for activation of this pathway suggested that it is involved in the upregulation of xCT in epileptic hippocampi in humans with temporal lobe epilepsy (Lewerenz et al. 2014). In breast cancer cells, insulin-like growth factor 1 (IGF-1) upregulates xCT in an IGF-1 receptor substrate 1 dependent manner (Yang & Yee 2014). In astrocytes, fibroblast growth factor 2 upregulates system x_c⁻ by a transcriptional mechanism whereas IGF-1 is ineffective (Liu et al. 2012, Liu et al. 2014). Thus, some of the responses to growth factors might be cell type-specific. To make the role of eIF2α regarding xCT expression even more complex, it was recently reported that in cancer cells and fibroblasts, eIF2α phosphorylation might directly regulate the stability of xCT mRNA via inhibition of the nonsense-mediated RNA decay (Martin & Gardner 2014). Whether this pathway contributes to the regulation of xCT expression in neuronal or glial cells remains to be demonstrated.

Regulation of system x_c⁻ activity and xCT expression by kinases, specifically cAMP-dependent pathways

Upon treatment with the NMDA receptor antagonist phencyclidine, McClatchy et al. observed that a phosphorylated peptide of xCT is highly upregulated in the prefrontal cortex in rats by an unknown kinase (McClatchy et al. 2015). In vitro, mutated xCT mimicking dephosphorylated xCT showed a lower activity than wildtype and mutated xCT mimicking a constitutively phosphorylated xCT. In acute rat striatal slices, the activation of metabotropic glutamate receptors 2/3 (mGluR 2/3), which lowers cellular cAMP levels, rapidly decreases system x_c⁻ activity (Baker et al. 2002b). However, Resch et al. could not detect any rapid upregulation of system x_c⁻ in astrocyte-enriched primary mixed cortical cell cultures when treated with the cAMP-increasing neuropeptide PACAP (Resch et al. 2014). Thus, system x_c⁻ activity might be rapidly regulated by direct phosphorylation, including cAMP-dependent pathways, under some circumstances but not others. In contrast to the lack of effect of short-term PACAP and VIP exposure on system x_c⁻ activity, Resch et al. observed that both neuropeptides induced system x_c⁻ in astrocytes via a pathway that includes the activation of the VPAC1 receptor and cAMP upon prolonged activation times of six hours and more. As xCT mRNA was increased, this pathway presumably upregulates system x_c⁻ via transcriptional mechanisms (Resch et al. 2014). This is in line with an early report by Gochenauer and Robinson, showing that prolonged treatment with the cAMP analogue dibutyryl-cAMP for 10 days induced system x_c⁻ in astrocytes (Gochenauer & Robinson 2001).

Inflammatory stimuli

An early report regarding the role of inflammatory stimuli in the regulation of xCT and system x_c⁻ indicated that cytokines like tumor necrosis factor α (TNFα) and the toll-like receptor 4 (TLR4) ligand bacterial lipopolysaccharide (LPS) upregulate system x_c⁻ activity in peritoneal macrophages (Sato et al. 1995). The mechanism for this upregulation remained elusive. Although a putative binding site for nuclear factor-κB (NF-κB) was identified in the murine xCT gene 5’ flanking region, LPS concentrations too low to activate NF-κB strongly stimulated xCT expression in macrophages (Sato et al. 2001), indicating that TLR4 activation acts independently of NF-κB. The LPS-induced upregulation of system x_c⁻ activity is preserved in macrophages derived from Nrf2 knock-out mice (Ishii et al. 2000). Thus, both Nrf2 and NF-κB can be excluded as responsible for the upregulation of system x_c⁻ activity via TLR4 activation, at least in macrophages.

Microglia are the resident macrophage population of the brain. It could be demonstrated that both LPS and TNFα strongly induce glutamate release via system x_c⁻ in primary microglia (Piani & Fontana 1994, Figuera-Losada et al. 2014) and by using xCT^−/− microglia it was confirmed that system x_c⁻ was the major source of microglial glutamate (Mesci et al. 2015). Not only TLR4 activation but also TLR3 activation via the viral double-stranded RNA mimetic polyinosinic–polycytidylic acid (poly I:C) induced system x_c⁻ activity in mixed microglial/astrocytic cultures (Scumpia et al. 2014). In vivo, intraspinal injection of LPS robustly induced xCT mRNA levels, most possibly, as indicated by immunofluorescence, in microglia and macrophages (Kigerl et al. 2012). The authors observed that co-injection of cystine along with LPS amplified LPS-induced neurotoxicity (Kigerl et al. 2012). This effect is compatible with the idea that glutamate released by system x_c⁻ is, at least in part, responsible for the observed neurotoxicity of neuroinflammation in vivo. However, the exact mechanism through which inflammatory stimuli upregulate system x_c⁻ in microglia remains enigmatic. Interestingly, compounds like fisetin, minocycline and apigenine that inhibit microglial activation, also inhibit LPS-induced glutamate release via system x_c⁻ in this cell type (Figuera-Losada et al. 2014). The effect of the TLR3 agonist poly I:C on system x_c⁻ activity was found to be partially inhibited by the antioxidant α-lipoic acid (Scumpia et al. 2014). The role of microglia and neuroinflammation in the regulation of system x_c⁻ activity in the brain does not only rely on the expression of xCT in microglia. Upon inflammatory activation, these cells produce many compounds including the pro-inflammatory cytokine interleukin-1β (IL-1β). IL-1β specifically upregulated system x_c⁻ in astrocytes but not in neurons or microglia (Jackman et al. 2010). In contrast to the data that exclude NF-κB as a mediator of LPS-induced system x_c⁻ upregulation in macrophages (Sato et al. 2001), the effect of IL-1β on system x_c⁻ expression was found to be mediated via NF-κB (He et al. 2015). Most recently, it was shown that the janus kinase (JAK)/ signal transducer and activator of transcription (STAT) pathway negatively regulates xCT transcription by binding of STAT3 and/or STAT5A to a gamma-activated site (GAS) motif in the xCT promoter region in breast cancer cells (Linher-Melville et al. 2015). STATs are activated in response to cytokine and growth factor signaling (O'Shea et al. 2013). Thus, the JAK/STAT pathway might alleviate the upregulation of xCT via these pathways (figure 2). Interestingly, deficiency in xCT reduced the release of TNFα and, in part, of IL-1β in microglia, in vitro and in vivo, respectively, indicating a positive feedback loop from cytokine-induced xCT expression to cytokine release (Mesci et al. 2015).

In summary, multiple stressors induce system x_c⁻ in brain-derived cells, some of those aimed at neutralizing oxidative stress or amino acid deficiency. In addition, neuronal activity, second messengers like cAMP, growth factor and cytokine signaling regulate system x_c⁻. Which of these pathways importantly regulates system x_c⁻ in the brain in vivo, remains to be determined.

System x_c⁻ substrates

As system x_c⁻ is a mandatory antiporter (Bannai 1986), its activity is not only regulated on the level of transcription and protein expression but also by the availability of the counter-transported substrates (figure 1). Intracellular glutamate, as a driving force for cystine import, is most probably not a rate-limiting substrate in the brain as the mean intracellular concentration is in the millimolar range (Kvamme et al. 1985) whilst the extracellular glutamate concentration is about 2–20 µM (Baker et al. 2002b, De Bundel et al. 2011). Using microdialysis, Baker et al. could demonstrate that extracellular cystine concentrations in the brain are extremely low (130–190 nM) (Baker et al. 2003). In general, system x_c⁻ is assumed to act as a cystine/glutamate antiporter. However, as deletion of xCT massively decreases cerebral extracellular glutamate concentration by >50% (De Bundel et al. 2011, Massie et al. 2011), other imported substrates more abundant than cystine have to be assumed. Candidates include the system x_c⁻ substrates L-α-aminoadipate (AA) and L-homocysteate (HCA) (figure 1). HCA was identified in astrocytes and is released from these cells (Do et al. 1997). As glutamate, HCA is an endogenous agonist of ionotropic glutamate receptors, specifically the NMDA receptor (Do et al. 1988). AA is a product of lysine metabolism and is also present in the brain (Chang 1982). In addition, cystathionine has been recently identified as a substrate of system x_c⁻ (Kobayashi et al. 2015) (figure 1). Interestingly, although cystathionine was absent in thymus and spleen in xCT knock-out mice, indicating that system x_c⁻ plays an essential role in importing cystathionine into these tissues, cystathionine levels were increased in the CNS in the absence of system x_c⁻. Thus, it can be hypothesized that system x_c⁻ mediates a net efflux of cystathionine from the brain. In addition, cystathionine can replace cystine for GSH synthesis (Kobayashi et al. 2015). In summary, the action of system x_c⁻ in the brain might be much more complex than assumed previously. Further research is needed taking into account all possible transported substrates, either glutamate and glutamate receptor agonists like HCA, or GSH precursors like cystine and cystathionine or other substrates like AA.

The effect of inhibition or loss of system x_c⁻ on emotional and cognitive features of behavior

Depressive- and anxiety-like behavior

Changes in glutamatergic signaling may be involved in the development of mood disorders. It has been shown in animal models that interference with glutamate reuptake as well as glutamate receptor signaling can lead to changes in depressive – and/or anxiety-like behavior. Antagonists of NMDA receptors (Sanacora et al. 2008) as well as agonists of mGluRs 2/3 (Fell et al. 2011, Matrisciano et al. 2008) exert anxiolytic and anti-depressive-like effects. Anti-depressive-like effects of ceftriaxone have been attributed to enhancement of glutamate reuptake (Mineur et al. 2007). At that time however, it was not known yet that ceftriaxone is also an inducer of xCT expression (Lewerenz et al. 2009, Knackstedt et al. 2010a) and as such the potential involvement of system x_c⁻ in inducing this effect had not been considered. N-acetylcysteine, a cysteine prodrug and known activator of system x_c⁻, has also been reported to possess anti-depressive-like activity in rodents. Although the authors do not exclude the possible involvement of glutamate released via system x_c⁻ and acting on mGluR 2/3, the effects of N-acetylcysteine were shown to be mediated through reduction of oxidative stress (Smaga et al. 2012).

Recently, the possible involvement of system x_c⁻ in depression and anxiety was further investigated by submitting xCT^−/− mice (McCullagh & Featherstone 2014, Bentea et al. 2015a), sut/sut mice or a heteroallelic xCT^+/−/sut mutant to a battery of behavioral paradigms commonly used in rodents as measures of such (McCullagh & Featherstone 2014). Moreover, the effect of inhibiting system x_c⁻ using sulfasalazine (SSZ; systemically and acutely) in male rats was investigated by Lutgen and colleagues (Lutgen et al. 2014). Of note, xCT^−/− and sut/sut mice as well as rats treated with SSZ do not show any motor deficits in a plethora of different tests (Bentea et al. 2015a, McCullagh & Featherstone 2014, Lutgen et al. 2014). Also visual deficits have been excluded in xCT^−/− mice (Bentea et al. 2015a) as they might, similar to motor deficits, represent important confounders for behavioral testing.

When it comes to emotional behavior, contradicting findings were reported. Male xCT^−/− mice clearly demonstrated reduced anxiety (manifested as increased time-in-the-center of the open field during a 60-min recording period, increased time-outside-the-shelter in the light/dark paradigm and decreased latency-to-feed in the novelty suppressed feeding) as well as reduced depressive-like behavior (manifested as decreased immobility in both the tail suspension and forced swim test), compared to their wildtype littermates (Bentea et al. 2015a). Similar to the observations of Bentea et al. in xCT^−/− mice after a five-min recording period, no effects were seen by McCullagh and Featherstone in the time-in-the-center of the open field during a eight-min recording period for the xCT^−/− mice, sut/sut mice or xCT^+/−/sut mice (McCullagh & Featherstone 2014). In contrast, an anxiogenic effect of SSZ was reported in rats as they spent less time in the open arm of the elevated plus-maze and less time-in-the-center of the open field (15-min recording) after SSZ administration. Moreover, SSZ had no effect on immobility in the forced swim test, a measure for depressive-like behavior (Lutgen et al. 2014).

The differences described above can of course result from the fact that inhibition of system x_c⁻ by SSZ is an acute effect whereas system x_c⁻-deficient mice have a chronic loss of system x_c⁻. Moreover, it is tempting to speculate that the effects of SSZ might be mediated via a mechanism independent of system x_c⁻. SSZ has been demonstrated to inhibit NF-κB activation (Wahl et al. 1998). In addition, NF-κB-independent actions of SSZ were shown to be mediated by stimulation of adenosine release (Cronstein et al. 1999). Moreover, it is unknown how much of SSZ crosses the blood-brain barrier (for a critical review on the potential of SSZ to be used as a drug to inhibit system x_c⁻ in CNS see: (Sontheimer & Bridges 2012)). SSZ has been shown to decrease brain tumor growth and tumor-associated seizures in animal models of glioblastoma (GBM) and as such it has to reach the brain (Buckingham et al. 2011, Chung et al. 2005). Yet, in the case of brain tumors, the blood-brain barrier might be compromised. Finally, the half-life of SSZ in rodents is very short (80–180 minutes) (Zheng et al. 1993) and anticonvulsive effects were reported to last for only two-three hours (Buckingham et al. 2011), meaning that a behavioral characterization two hours after SSZ administration might be misleading. However, in the study of Lutgen and colleagues SSZ did affect extracellular glutamate levels in the brain and the effects of SSZ could be reversed by N-acetylcysteine (Lutgen et al. 2014). As such, the changes in behavior observed in rats after SSZ indeed are compatible with the hypothesized action of SSZ via inhibition of system x_c⁻.

Cognition

As the major excitatory neurotransmitter in the mammalian CNS, glutamate is undeniably involved in most aspects of normal brain functioning, including higher brain functions such as cognition, memory and learning. A total loss or partial inhibition of system x_c⁻ can as such affect these functions. Although McCullagh and Featherstone describe deficits in spontaneous alternation tasks for the sut/sut mice when compared to wildtype C3H/HeSnJ mice, consistent with the impaired spatial working memory reported before in the xCT^−/− mice (De Bundel et al. 2011), they could not confirm these effects in xCT^−/− mice or xCT^+/−/sut mutants when compared to C57BL/6J or C3H/HeSnJ/C57BL/6 first generation offspring, respectively (McCullagh & Featherstone 2014). SSZ was reported to induce some cognitive impairment as it negatively affected attentional set shifting in rats (Lutgen et al. 2014). Moreover, Li and co-workers described reduced long-term memory in sut/sut mice as evaluated by investigation of their fear memory in a fear conditioning as well as a passive avoidance paradigm (Li et al. 2012). It should be mentioned however, that fear memory might be affected by anxiolytic effects that have been observed in the xCT^−/− animals (Bentea et al. 2015a). Still, adult xCT^−/− mice do show impaired spatial working memory (De Bundel et al. 2011) and rodents receiving pharmacological drugs now known to inhibit system x_c⁻, have impaired long term memory consolidation (Rickard & Ng 1995, Bianchin et al. 2000) as well as deficits in spatial working memory (Bordi et al. 1996, Wetzel et al. 1995), indicating that reductions in the physiological function of system x_c⁻ basally result in negative effects on cognition. Notably, enhanced levels of system x_c⁻ could also be ultimately deleterious to cognitive function via its propensity to contribute to excitotoxic neuronal injury, possibly explaining why xCT^−/− mice are protected against age-dependent decline in spatial working memory (De Bundel et al. 2011).

It is clear that available data on cognition and memory are very diverse and covering many different aspects of this complex behavior. Therefore, more experimental evidence is certainly needed to provide a clear-cut view on the possible consequences of inhibiting system x_c⁻ (which might be an attractive new strategy for the development of new treatments for several neurological disorders – see below) on these brain functions.

System x_c⁻ and drug addiction

The nucleus accumbens (NAc) is a key region mediating the long-term behavioral pathologies produced by the self-administration of addictive drugs. In the NAc core, system x_c⁻ is a significant contributor to basal extracellular glutamate levels which are decreased by 60% when applying inhibitors of this system (Baker et al. 2002b). Basal levels of glutamate in this brain region were found to be altered by cocaine (Baker et al. 2003), methamphetamine (Parsegian & See 2014, Lominac et al. 2012) and ethanol (Griffin et al. 2014, Griffin et al. 2015). In the case of cocaine, these changes in basal glutamate have been shown to produce alterations in the function of both pre- and post-synaptic glutamate receptors and to contribute directly to the risk of relapse to cocaine-seeking in rodents (figure 3).

Figure 3. Four neuroadaptations in the NAc core promote the reinstatement of cocaine seeking.

The top panel depicts a NAc core synapse of a drug-naïve rat while the bottom panel depicts that of a rat post-cocaine. Relative to drug naïve animals, cocaine animals display: 1) reduced expression and function of xCT/system x_c⁻, leading to reduced basal non-vesicular glutamate; 2) reduced expression and function of EAAT2, leading to increased glutamate overflow during reinstatement (3); and 4) increased expression of GluR2-lacking AMPA receptors and the AMPA-receptor clustering protein Narp. As discussed below, compounds which restore both xCT and GLT-1 restore basal levels of glutamate and attenuate glutamate spillover during reinstatement.

Drug seeking is studied in rodents using the extinction-reinstatement model of relapse, in which animals self-administer drug in the operant chamber and then undergo extinction training for one-three weeks. During extinction, the operant response that previously delivered drug no longer does so and responding typically declines. The drug-seeking response is reinstated with stimuli known to induce relapse in humans, namely stress, cues and the drug itself (Epstein et al. 2006). Relapse to cocaine seeking in the extinction-reinstatement model is associated with decreased basal extrasynaptic glutamate in the NAc core. The decrease in basal glutamate at this time stems from a decrease in NAc core expression and function of xCT/system x_c⁻ (Baker et al. 2003, Knackstedt et al. 2010a). Decreased basal glutamate in turn causes a loss of tone on release-regulating glutamate autoreceptors, namely mGluR 2/3 (Moran et al. 2005). This adaptation is likely a major contributing factor to the enhanced synaptically released (vesicular) glutamate that drives the drug-primed reinstatement of cocaine seeking (McFarland et al. 2003), in addition to the observed reduction in glutamate re-uptake (Knackstedt et al. 2010a). Synaptically-released glutamate promotes relapse via actions at post-synaptic AMPA and mGluR5 receptors, as the intra-NAc infusion of AMPA and mGluR5 antagonists attenuates cocaine relapse (Cornish & Kalivas 2000, Wang et al. 2013).

Chronic reductions in basal extracellular glutamate contribute to the development of post-synaptic adaptations. Following cocaine self-administration and extinction, we have observed increased expression of the protein Narp, responsible for clustering AMPA receptors and promoting long-term potentiation (Knackstedt et al. 2010b). Cocaine self-administration followed by both extinction (Moussawi et al. 2011) and withdrawal without extinction training (Conrad et al. 2008) produces enduring synaptic potentiation in the NAc core, accompanied by a change in AMPA receptor subunit composition towards the promotion of higher conductance GluR2-lacking AMPA receptors (Conrad et al. 2008). Taken together, these results suggest that cocaine self-administration induces a potentiation of glutamatergic transmission from the prefrontal cortex to the NAc core. It has been proposed that the decrease in basal glutamate, via system x_c⁻ downregulation, is the causative factor in the post-synaptic adaptations observed after cocaine (Wolf 2010) and restoring basal glutamate levels normalizes some of these observations (Trantham-Davidson et al. 2012). Figure 3 depicts these alterations and the two pools of glutamate impacted by cocaine: extrasynaptic (non-vesicular) and synaptic (vesicular).

Basal extracellular glutamate levels in the NAc are also reduced two weeks after methamphetamine self-administration but only when animals undergo extinction training (Parsegian & See 2014). Interestingly, following three weeks of abstinence without extinction training, methamphetamine self-administration produces an increase in basal glutamate in the NAc (Lominac et al. 2012). Differences in glutamatergic adapations following extinction vs. abstinence have been reported following cocaine as well (e.g.(Knackstedt et al. 2010b)). The contribution of system x_c⁻ to these changes in basal glutamate has not yet been investigated.

Non-contingent ethanol alone or in combination with ethanol self-administration has been repeatedly shown to increase basal glutamate in the NAc core and shell (e.g. (Griffin et al. 2014, Griffin et al. 2015). It was recently demonstrated that this increase in basal glutamate does not stem from increased release via system x_c⁻, as its function was not altered in animals exposed to a non-contingent ethanol administration paradigm that increased basal glutamate (Griffin et al. 2015).

While ceftriaxone treatment (200 mg/kg for five-seven days) restores EAAT2/GLT-1 expression following cocaine self-administration, it also restores expression of xCT in the NAc (Knackstedt et al. 2010a). As discussed above, ceftriaxone induces a known transcriptional regulator of xCT, Nrf2, and thereby increases xCT expression and system x_c⁻ activity (Lewerenz et al. 2009). Later, it was demonstrated that the ceftriaxone-mediated increase in xCT expression leads to enhanced system x_c⁻ activity and increased basal glutamate (Trantham-Davidson et al. 2012). The chronic administration of ceftriaxone prevents the reinstatement of cocaine seeking in response to drug-associated cues and cocaine itself (Knackstedt et al. 2010a). Reinstatement is attenuated as late as two weeks after the cessation of ceftriaxone administration (Sondheimer & Knackstedt 2011), indicating a long-lasting reversal of cocaine-induced changes in brain physiology. Ceftriaxone reduces ethanol consumption in ethanol-preferring rats, while increasing xCT expression in the NAc and prefrontal cortex (Alhaddad et al. 2014). This finding is interesting in light of the consistently observed increases in basal glutamate in the NAc after ethanol and in theory increasing xCT may increase these levels even further.

In conclusion, the importance of xCT in addiction has been clearly demonstrated for cocaine and it remains a strong treatment target for reducing cocaine relapse. More work is needed to clarify the role of xCT in mediating the decrease in basal glutamate observed following methamphetamine self-administration and the reduction in ethanol consumption following ceftriaxone administration.

Contribution of system x_c⁻ to neurological disease

The bimodal actions (sustaining GSH synthesis and releasing glutamate) of system x_c⁻ intriguingly suggest that this antiporter can either protect from or contribute to neural injury in different affections of the CNS. Its expression and function have indeed been studied in different in vitro and in vivo models pertinent to CNS diseases/disorders, as described below.

Alzheimer’s disease

Alzheimer’s disease (AD) is the leading cause of dementia in the population and is caused by neurodegeneration predominantly of brain regions involved in mediating memory and cognition, including the hippocampus and cerebral cortex. Neurofibrillary tangles of intraneuronal phosphorylated Tau accumulation and extracellular amyloid-β (Aβ) plaques consisting of aggregated Aβ_1–40/42 peptides proteolytically derived from the amyloid precursor protein (APP), are the hallmarks of AD brain pathology. An important field of research on AD focuses on the role of microglial cells, the macrophages of the CNS, as potent players in plaque removal as well as in the production of neurotoxic and/or neurotrophic factors. Microglial cells — depending on the origin, strength and length of stimuli —are able to release different cytokines, chemokines, reactive oxygen species (ROS) and amino acids that could positively or negatively influence AD brain pathology (Tang & Le 2015).

Initial studies that potentially link system x_c⁻ to AD pathology were performed in cell culture. Primary rodent microglial cell cultures treated with soluble APP (sAPP) or aggregated Aβ_1–40 peptides increased cellular release of glutamate through system x_c⁻ (Qin et al. 2006, Barger & Basile 2001). Such activation of system x_c⁻ in microglia with sAPP led to compromised synaptic density in hippocampal neurons in co-culture (Barger & Basile 2001) whereas activation with Aβ_1–40 caused frank excitotoxic cortical neuronal death (Qin et al. 2006). In addition to sAPP and Aβ, several inflammatory factors that are either associated with AD or recapitulate cellular responses in AD, including LPS, oxidative stress, and TNFα can also activate or increase the expression of macrophage/microglial system x_c⁻, as discussed above (Piani & Fontana 1994, Sato et al. 2001, Mesci et al. 2015, Sato et al. 1995, Albano et al. 2013). Therefore, the possibility that several factors linked to chronic AD could together cooperate to increase glutamate release through system x_c⁻ and as such adding to ongoing neurodegeneration cannot be excluded.

In vivo, increased xCT expression has been demonstrated in the cerebral cortex of an animal model of AD, aged AβPP23 (18 month-old) mice, when gliosis and Aβ plaques are prominent (Schallier et al. 2011). Due to the non-specificity of antibodies against xCT for immunostaining on mouse tissues (Massie et al. 2008, Mesci et al. 2015, Van Liefferinge et al., in revision at J. Comp. Neurol.), the cell-types overexpressing system x_c⁻ protein in AD mouse brains have been difficult to identify. However, microglial cells activated by Aβ prominently express xCT mRNA as visualized by in situ hybridization in two different mouse models of AD: wildtype mice injected with Aβ into the hippocampus and in the vicinity of Aβ plaques in Thy1-APP₇₅₁ mice (Qin et al. 2006). Interestingly, the glutamate transporter EAAT2/GLT1, which takes up glutamate from the extracellular space, was found to be downregulated in AβPP23 mice (Schallier et al. 2011). Transgenic overexpression of EAAT2 ameliorates AD pathology and memory loss in AD mice (Takahashi et al. 2015), suggesting that glutamate dysregulation is pathophysiologically important in AD (figure 4). However, studies involving inhibition of system x_c⁻ or its genetic deletion (Sato et al. 2005) in AD mouse models are needed to fully analyze the role of xCT in AD brain. Of note, when considering system x_c⁻ as a possible target in the treatment of AD, the effects of inhibition of system x_c⁻ on cognition should be taken into account (see above).

Figure 4. The hypothetical role of system x_c⁻ in excitotoxicity in neurodegenerative diseases.

(Left panel) In the healthy brain, glutamate release by system x_c⁻ (here only depicted in microglia but also present on astrocytes; conflicting data about neuronal expression have been published) is balanced by glutamate uptake by EAAT2, which leads to negligible activation of extrasynaptic ionotropic glutamate receptors (GluR). (Right panel) In many disease states of the brain, system x_c⁻ is upregulated, in ALS and AD especially in microglia and subsequently glutamate release is increased. Astrocytes may also contribute to the increase in glutamate release via system x_c⁻ (not shown). Simultaneously, EAAT2 is down-regulated. The increased extrasynaptic glutamate concentration activates extrasynaptic ionotropic glutamate receptors and induces excitotoxicity.

Parkinson’s disease

Parkinson’s disease (PD) is a slowly progressive neurodegenerative disorder affecting 2% of individuals over 65 years of age. The clinical motor symptoms are linked to a profound loss of striatal dopaminergic innervation caused by a progressive degeneration of dopaminergic neurons within the midbrain's substantia nigra pars compacta. To model the dopaminergic neurodegeneration seen in PD, toxins, including 6-hydroxydopamine (6-OHDA) and MPTP can be injected either into the rodent nigrostriatal pathway directly (6-OHDA) or systemically (MPTP) (Dauer & Przedborski 2003).

Of note, an ipsilateral increase in striatal xCT protein levels occurred in hemi-Parkinson rats three weeks after 6-OHDA injection into the medial forebrain bundle (Massie et al. 2008). A follow-up study in xCT^−/− mice demonstrated that expression of system x_c⁻ can impact dopamine neuron survival. Specifically, dopamine neurons in the substantia nigra pars compacta of xCT^−/− mice striatally injected with 6-OHDA were significantly protected as compared to those from 6-OHDA-injected wildtype mice (Massie et al. 2011). Reduced extracellular glutamate levels in xCT^−/− mice, coupled with an increased survival of dopaminergic neurons in the 6-OHDA model, suggest that system x_c⁻ might be a promising pharmacological target for PD therapy. However, a more complex response is seen using the MPTP model (Bentea et al. 2015b). Mice lacking xCT are equally susceptible to MPTP-induced parkinsonism, both pathologically and behaviorally, as wildtype littermates, despite the latter having an increased expression of xCT in the striatum. Interestingly, the same treatment resulted in a down-regulation of xCT in the substantia nigra pars compacta (Bentea et al. 2015b). Regional control of xCT changes might represent a plausible reason for the differences seen between models, as do obvious differences in the experimental paradigms employed. Hence, additional studies are necessary to better understand the involvement of system x_c⁻ in the chronic neurodegenerative processes associated with PD.

Amyotrophic Lateral Sclerosis

Amyotrophic Lateral Sclerosis (ALS) is the most common motor neuron disease of the adult leading to progressive paralysis and death of the patients within two-five years after disease onset, which commonly occurs in the fifth decade of life (Johnston et al. 2006). While mainly sporadic, 10% of ALS cases have a familial history of which in two thirds of these cases the genetic cause has been discovered. Besides in the C9ORF72 gene (DeJesus-Hernandez et al. 2011), mutations occur in Cu/Zn superoxide dismutase (SOD1) —the first ALS gene discovered — which led to the design of ALS mouse models that express SOD1 with different mutations (Joyce et al. 2011). Studies in mutant SOD1 ALS mice have shown that ALS is a non-cell autonomous disease with glial cells participating in motor neuron degeneration (Boillée et al. 2006a, Boillée et al. 2006b, Beers et al. 2006).

Recently, the regulation of system x_c⁻ in acute spinal cord slices was assessed through measurement of cystine uptake in the rapidly-progressing ALS hSOD1^G93A mice at different ages (Albano et al. 2013). In presymptomatic 70-day-old ALS hSOD1^G93A mice (but not 55-day-old mice), cystine uptake was significantly increased as compared to age-matched non-transgenic controls. This change in system x_c⁻ function was, in this experimental context, transitory, since at later ages including at symptomatic stages, no difference was measured (Albano et al. 2013). The authors speculate that this increased expression might represent a compensatory change precipitated by oxidative stress that is measureable by 60 days in this mouse model (Liu et al. 1998).

In the more slowly progressing ALS hSOD1^G37R mouse model, however, expression of xCT mRNA was demonstrated to be increased over the course of the disease. Cellular analysis determined that this increase occurred in microglia, but not in motor neurons (see above) (Mesci et al. 2015). Relevantly, post-mortem spinal cord tissue from ALS patients also demonstrated expression of xCT in spinal cord where levels are correlated with a macrophage marker of inflammation (CD68) (Mesci et al. 2015). In line with the view that system x_c⁻ is involved in the antioxidant defense, deleting xCT in ALS hSOD1^G37R mice led to earlier symptom appearance. However, these mice demonstrated a prolonged symptomatic phase where xCT expression was found to be upregulated. Grip strength was also preserved at late symptomatic stages and overall motor neuron survival was increased at disease end stage (Mesci et al. 2015). Hence, a reduction in glutamate release from system x_c⁻ could be enough to slow the rise of extracellular glutamate levels that result from a reduction in its astrocytic clearance, which is also pathologically affected in ALS patients (Rothstein et al. 1995, Rothstein et al. 1992) and tardively in ALS mutant mice (Canton et al. 1998, Warita et al. 2002) due to loss of EAAT2/GLT-1 expression (figure 4).

Finally, another interesting link between system x_c⁻ and ALS stems from the finding that the excitotoxin β-N-methylamino-L-alanine is a transportable inhibitor of system x_c⁻ (Warren et al. 2004), which simultaneously competes with cystine uptake whilst eliciting glutamate release (Liu et al. 2009). β-N-methylamino-L-alanine has been described as the putative causative agent underlying the motor neuron degeneration and spastic paraparesis that affected the inhabitants of Guam in the late 1950s. In this context, enhanced oxidative stress as well as an increase in glutamate excitotoxicity affected neuronal survival (Liu et al. 2009).

All together, these data indicate that the contribution of enhanced system x_c⁻ function to ALS progression and pathology may be complex. It is intriguing to speculate that in an attempt to mitigate oxidative stress, a presymptomatic upregulation provides some protection initially, but eventually contributes to injury during the later stages of the disease via increased glutamate release and modulation of microglial functions (figure 4).

Huntington’s disease

Huntington’s disease (HD) is an autosomal dominant fatal neurodegenerative disease caused by polyglutamine expansion in the gene huntingtin (htt) that leads to the death of GABAergic medium-sized spiny neurons in the striatum, although other brain regions are affected as the disease progresses. HD presents as a movement disorder, with co-morbid psychiatric and cognitive symptomology (Nance 1997). Mechanistically, it has been proposed that transcriptional impairment resulting from the mutation of htt leads to mitochondrial dysfunction, ultimately rendering neurons more sensitive to excitotoxicity and oxidative stress (Johri & Beal 2012).

Recent studies demonstrate a dysregulation of system x_c⁻ in R6/2 mutant htt exon1 transgenic mice, a model for HD, and in immortalized striatal cells derived from htt Q111 knock-in mice (STHdh^Q111/^Q111) (Trettel et al. 2000). Specifically, xCT mRNA and protein levels were demonstrated to be decreased in the striatum of six-week-old R6/2 mice as well as in the STHdh^Q111/^Q111 cells. The activity of system x_c⁻ (only assessed in the striatal cell line) was also significantly decreased (Frederick et al. 2014). This change in the STHdh^Q111/^Q111 cells was accompanied by a decrease in cellular GSH levels and a corresponding increase in basal ROS expression (Ribeiro et al. 2012). Accordingly, cells showed enhanced sensitivity to oxidative stressors (Frederick et al. 2014). Interestingly, it has been discovered recently that cystathionine is an alternative substrate for system x_c⁻ and cystathionine γ-lyase, the enzyme that catalyzes the break-down of cystathionine into α-ketobutyrate, ammonia and the GSH precursor cysteine, is also downregulated in cellular and animal models of HD (Paul et al. 2014). Both cystine and cystathionine imported via system x_c⁻ can sustain intracellular GSH levels (Kobayashi et al. 2015). Of note, cysteine supplementation prolongs life span in R6/2 mice (Paul et al. 2014). Thus, the oxidative stress that has been found in HD (Browne et al. 1999) might be in part explained by the dysregulation of the cystine/cysteine metabolism. Whether a decrease in system x_c⁻ expression in HD mice and in STHdh^Q111/^Q111 striatal cells is due to its transcriptional dysregulation (as has been proposed, (Frederick et al. 2014)) remains to be determined. Interestingly, glutamate uptake via EAAT2 is also impaired in HD (reviewed in (Sheldon & Robinson 2007) and thus enhancing endogenous system x_c⁻ levels might lead to the risk of excitotoxocity. As such, dual EAAT2/xCT inducers like ceftriaxone (Knackstedt et al. 2010a, Lewerenz et al. 2009), might potentially be effective strategies for treatment in HD. Indeed, ceftriaxone attenuated the clinical phenotype in R6/2 mice (Miller et al. 2008).

Cerebral Ischemia

Interruption of blood flow to the brain, i.e. cerebral ischemia or stroke, is the leading cause of adult disability in the world. Overwhelming evidence indicates that glutamate-mediated excitotoxicity is a major contributor of neuronal death. Yet, inhibitors of post-synaptic glutamate receptors at doses not leading to intolerable side effects, have not proven to be an effective therapeutic option in stroke (Ikonomidou & Turski 2002, Muir & Lees 2003, Ginsberg 2009, Ginsberg 2008). Thus, understanding the source(s) of glutamate contributing to ischemic injury is of paramount importance. Evidence suggests system x_c⁻ is one such source.

Positron emission tomography imaging of rat brains after focal experimental cerebral ischemia (transient middle cerebral artery occlusion or MCAO) demonstrated a rapid increase in activity of system x_c⁻ that peaked five hours later (Soria et al. 2014). Interestingly, pharmacological inhibition of system x_c⁻ reduced oxygen-glucose deprivation-induced neuronal currents (i.e. anoxic depolarizations) as well as cell death in slices and slice cultures, respectively (Soria et al. 2014). In addition, using a mixed cortical cell culture system, it was found that enhanced astrocytic system x_c⁻ activity ― while not deleterious in and of itself ― becomes a major contributor of excitotoxic glutamate under conditions of energy deprivation: hypoxia or hypoglycemia (Fogal et al. 2007, Jackman et al. 2012, Jackman et al. 2010).

At present, there are no commercially available selective pharmacological reagents with which to directly demonstrate that inhibiting system x_c⁻ activity contributes to cerebral ischemic injury. Nevertheless, carboxyphenylglycine compounds that antagonize both system x_c⁻ and mGluR1α (Brabet et al. 1995, Kingston et al. 2002) — which have proven useful to teasing out the contributions of system x_c⁻ in vitro — have been utilized in animal experiment ischemia models. Interestingly, mice who were administered one such compound (LY367385) via tail vein three hours after ischemia showed less infarct volume when evaluated one day later (Li et al. 2013). Similar protective effects were found in rats (Moroni et al. 2002) and gerbils (Bruno et al. 1999). Curiously, neuronal injury following MCAO is not reduced in mGluR1α-deficient mice (Ferraguti et al. 1997). Thus, the notion that inhibition of system x_c⁻ activity underlies the neuroprotective effects of the carboxyphenylglycine derivatives under the conditions of cerebral ischemia cannot be excluded and should be tested directly. In support of this idea, Hewett and colleagues demonstrate that sut/sut mice were found to be less susceptible to injury that follows transient cerebral ischemia induced by MCAO (SJH, unpublished observations). Hence, in amalgamation, evidence suggests a potential role for system x_c⁻ in cerebral ischemic damage.

Multiple Sclerosis

Multiple Sclerosis (MS) is best considered a chronic primary demyelinating disease of autoimmune origin with more than two million people affected worldwide. One of the best studied animal models of autoimmune inflammatory demyelination as found in MS, is experimental autoimmune encephalomyelitis (EAE) elicited by immunization of animals with peptides derived from myelin proteins (reviewed in (Rangachari & Kuchroo 2013)). Alterations in glutamate homeostasis contribute to oligodendrocyte cell death and neurological deficits in animal models of MS (reviewed in (Gonsette 2008)). The source of this glutamate has been a subject of much investigation. Pertinently, cells of the monocyte-macrophage-microglia lineage taken from animals with EAE have higher xCT expression than those taken from control animals (Pampliega et al. 2011). Although these findings were not recapitulated in tissue from mice infected with Theiler’s murine encephalomyelitis virus, another common animal model of MS (Merckx et al. 2015), enhanced xCT expression was found in leukocytes taken from human MS patients as well as in MS post-mortem optic nerve (Pampliega et al. 2011). The potential significance of these findings was demonstrated in vitro. In oligodendrocyte-microglia co-culture, activated microglia increased extracellular glutamate release, which resulted in oligodendrocyte excitotoxicity that was prevented by inhibition of system x_c⁻ (Domercq et al. 2007).

Treatment with the unspecific system x_c⁻ inhibitor SSZ, reduced the clinical severity of EAE in mice, an effect associated with a reduction in T cell infiltration, reactive gliosis and myelin damage (Evonuk et al. 2015). To further confirm the involvement of system x_c⁻ in this effect, sut/sut mice and wildtype littermates were subjected to EAE, to which mice deficient in xCT proved to be largely resistant. Finally, the authors found that myelin-specific CD4+ T helper type 1 cells provoked microglia to release glutamate via the system x_c⁻ transporter causing excitotoxic death to mature myelin-producing oligodendrocytes (Evonuk et al. 2015).

Taken together, the findings suggest that microglial system x_c⁻ upregulation can produce excitotoxic damage to myelin in the setting of EAE. The additional intriguing observation that pharmacological inhibition of system x_c⁻ can also suppress immune cell infiltration suggests that modulation of system x_c⁻ either alone or as an add-on therapy might be therapeutically useful in MS.

Epilepsy

Epilepsy is one of the most common neurological disorders, characterized by the occurrence of epileptic seizures as a result of perturbations in the excitation/inhibition balance. The most common and sometimes devastating form, temporal lobe epilepsy, is initiated by an initial insult, often resulting in a status epilepticus. Next, during a latent phase of variable duration, epileptogenesis takes place, a form of synaptic maladaptation finally leading to a reduced threshold for seizures in the epileptic brain. It was recently discussed that glutamate release from astrocytes, although not necessary for the generation of seizures, can modulate the threshold for seizure generation (Steinhäuser et al. 2015). After all, extrasynaptic glutamate cannot only modulate synaptic transmission via activation of pre- and postsynaptic mGluRs, but it can also act on extrasynaptic NMDA receptors and as such increase neuronal excitability. Moreover, several agonists and antagonists of respectively group II and group I mGluRs as well as antagonists of ionotropic glutamate receptors have been shown to exert anticonvulsive properties (for review (Tang 2005, Casillas-Espinosa et al. 2012)).

Increased xCT expression levels were reported in EL mice, a model for generalized seizures (Takaki et al. 2008). Moreover, it was shown that xCT^−/− mice were less susceptible for chemically- as well as electrically-induced acute limbic seizures compared to their wildtype littermates (De Bundel et al. 2011). The increased threshold for eliciting limbic seizures in xCT^−/− mice was attributed to the reduction in extracellular hippocampal glutamate levels with about 60% in these mice. Recently, Lewerenz et al. reported increased hippocampal xCT expression levels in temporal lobe epilepsy patients and unveiled the pathway that resulted in this increased xCT expression (see above) (Lewerenz et al. 2014). Increased xCT expression in the epileptic brain can as such result in a decreased threshold for seizures, in accordance with the pro-convulsive effect that has been described for N-acetylcysteine (De Bundel et al. 2011). Finally, as described below, increased expression of system x_c⁻ in brain tumor cells has been suggested to contribute to peritumoral seizures (Buckingham et al. 2011).

To conclude, whereas the involvement of system x_c⁻ in seizure development seems to be straightforward, the potential role of system x_c⁻ in epileptogenesis remains elusive.

xCT and glioblastoma

GBM are brain tumors derived from cells of glial lineage. GBMs grow rapidly and relentlessly and invade normal brain tissue. In 1999, Ye et al. reported that the activity of EAATs which take up glutamate and aspartate, is downregulated in human GBM cell lines compared to astrocytes while the activity of system x_c⁻ is upregulated (Ye et al. 1999). The combination of impaired glutamate uptake and increased system x_c⁻ activity explained the continuous glutamate release observed in these tumor cells in vitro (Ye & Sontheimer 1999). Glutamate released via system x_c⁻ in vitro was shown to positively regulate GBM cell migration, most possibly via AMPA receptor-induced calcium signaling (Lyons et al. 2007). Thus, glutamate release via system x_c⁻ from GBM cells might contribute to GBM brain tissue invasion.

In a follow up study, the same group could demonstrate that inhibition of system x_c⁻ activity prominently decreases cellular GSH levels, slows proliferation and induces cell death in human GBM cell lines in vitro suggesting that in addition to glutamate release cystine import via system x_c⁻ might be pathophysiologically important for the growth of GBMs. In addition, the authors observed that high-dose SSZ (16 mg/mouse daily i.p.) also significantly suppressed growth of GBM cells implanted in mouse brains suggesting that this effect is active in vivo (Chung et al. 2005). In contrast to the studies by the Sontheimer lab, Savaskan et al. neither observed that siRNA-mediated knock-down of system x_c⁻ nor inhibition using the non-substrate inhibitor (S)-4-carboxyphenylglycine decreased GBM cell proliferation in vitro nor tumor growth in intracerebral implants in vivo. However, the decreased glutamate release induced by knock-down of system x_c⁻ was associated with decreased neurodegeneration when tumor cells were co-cultured with brain slices (Savaskan et al. 2008). In addition, both (S)-4- carboxyphenylglycine and xCT knock-down increased survival of mice with intracranially implanted tumors. This effect was associated by a prominently decreased peritumoral edema. Thus, these findings indicate that the detrimental role of system x_c⁻ in GBM pathophysiology might rather consist of glutamate-mediated neurotoxicity and edema formation of the peritumoral brain tissue. Surprisingly, in humans high xCT expression was found associated with a rather invasive phenotype consistent with the idea that system x_c⁻ fosters GBM cell migration, whereas an edematous phenotype was associated with low xCT expression (Takeuchi et al. 2013), inconsistent with the idea that system x_c⁻ activity induces brain edema. However, in this study tumor xCT expression was judged based upon xCT immunoreactivity using paraffin embedded formaldehyde-fixed tumor samples. Tumors were grouped dichotomically in high and low xCT expressing tumors. There was a correlation of a 57kDa band judged to be xCT upon western blotting with immunohistochemistry signals in a subgroup of samples. However, semiquantitative xCT expression, assessed by western blotting, was in average only two-fold higher in samples with high xCT expression when judged by immunohistochemistry when compared to samples with low expression. In addition, values obtained by semiquantiative western blotting for the high and low xCT group were partially overlapping. Astonishingly, the authors observed two-fold longer progression-free survival and a seven months longer overall survival in patients with GBM with low xCT-like tissue immunoreactivity.

Up to 50% of patients with GBM suffer from sometimes debilitating epileptic seizures (Paillas 1991). Buckingham et al. demonstrated that the human GBM cells when implanted into the brain of scid mice induce epileptiform neuronal activity upon electrophysiological recordings in acute slices as well as epileptic seizures detected by EEG/video monitoring (Buckingham et al. 2011). The authors concluded that glutamate released by system x_c⁻ in the tumor tissue was involved in the epileptic activity, as SSZ-inhibitable glutamate release was observed in the tumor-containing acute brain slices when 100µM cystine was added. In addition, SSZ decreased epileptiform activity induced in these slice preparation by NMDA receptor disinhibition due to magnesium removal, also in the presence of 100µM cystine. Finally, spontaneous epileptiform EEG activity was prominently decreased within the first hour of intraperitoneal injection of high-dose SSZ (8 mg/mouse) exceeding the maximal orally administered single dose in humans more than 10-fold.

There is no cure for patients suffering from the rapidly growing tumor, which with time becomes resistant against radio- and chemotherapy. Temozolomide is the standard chemotherapeutic compound administered in GBM patients. In vitro, exposure to temozolomide leads to increased GSH levels in GBM cell lines, probably via ATF4- and Nrf2-mediated upregulation of system x_c⁻ activity (Chen et al. 2015). Inhibition of system x_c⁻ prominently exacerbated the sensitivity of GBM cell lines to temozolomide suggesting that upregulation of xCT is part of the acquired temozolomide resistance in GBMs. Similar observations were made using the prooxidant experimental anticancer drug cannabidiol (Singer et al. 2015). However, in a small clinical trial, adjuvant SSZ given in combination with temozolomide and radiotherapy did not result in increased antitumor activity of radiochemotherapy and was poorly tolerated (Takeuchi et al. 2014). Another phase 1/2 clinical trial using SSZ monotherapy at doses from 1.5 g to 6 g/day had to be terminated prematurely as severe adverse events were encountered frequently including increase in brain edema, increased seizure frequency associated with decreased plasma levels of anticonvulsant drugs without a relevant effect on tumor growth (Robe et al. 2009).

In summary, preclinical models of GBM suggest that system x_c⁻ might play a role in GBM pathophysiology by (1) supporting tumor cell growth, (2) fostering GBM cell migration, (3) by inducing peritumoral neurotoxicity, (4) by inducing epileptic seizures and/or (5) by inducing chemotherapy resistance. However, the results obtained using different preclinical models are in part inconsistent. In addition, more rigorous approaches to quantify xCT expression in human tumor samples to unequivocally characterize the association of the clinical course of the disease with xCT expression are highly warranted. Finally, the only approved drug that inhibits system x_c⁻, SSZ, is poorly tolerated in patients receiving GBM standard therapy and progressive GBM at doses much lower than found effective in preclinical models.

Conclusion

Taken together, with exception of HD (see above), there is accumulating evidence from animal models of neurological diseases, which is in part supported by the analysis of human brain tissues, that xCT and thereby system x_c⁻ are upregulated during neurodegeneration and contribute to damage to the CNS. Especially, studies in ALS and AD mouse models hint a predominant expression of system x_c⁻ in microglia. However, astrocytic system x_c⁻ might also play a role. In theory, glutamate release via system x_c⁻ will increase extrasynaptic glutamate, especially as glutamate reuptake via EAATs is generally assumed to be perturbed in many neurological diseases (Sheldon & Robinson 2007). This might lead to the activation of extrasynaptic NMDA receptors, which mediate excitotoxicity (figure 4). The modulation of extrasynaptic glutamate and thereby the activity of metabotropic and ionotropic glutamate receptors might also modulate neuronal excitability in epilepsy and behavior.

Inducible and cell type-specific knock-out models of xCT might help to unequivocally identify the cell types responsible for the above described role of system x_c⁻ in neurological diseases. In addition, the development of pharmacologically tolerable and highly selective inhibitors of system x_c⁻ (Patel et al. 2010, Newell et al. 2014) will represent an important step toward our ability to analyze whether inhibition of system x_c⁻ represents a potent therapeutic strategy for the slowing of neural injury when applied during the symptomatic phase of disease in humans.

Abbreviations

6-OHDA

6-hydroxydopamine

Aβ

amyloid-β

AA

L-α-aminoadipate

AARE

amino acid response element

AD

Alzheimer’s disease

ALS

amyotrophic lateral sclerosis

AMPA

alpha-amino-3-hydroxy-5-methylisoxazole-4-propionate

APP

amyloid precursor protein

ARE

antioxidant response element

ATF4

activating transcription factor 4

EAAT

excitatory amino acid transporter

EAE

experimental autoimmune encephalomyelitis

eIF2α

eukaryotic initiation factor 2 α

GAS

gamma-activated site

GBM

glioblastoma

GCN2

general control non-depressible-2

HCA

L-homocysteate

HD

Huntington’s disease

HRI

heme-regulated eIF2α kinase

Htt

huntingtin

IGF-1

insulin-like growth factor-1

IL-1β

Interleukin-1β

JAK

janus kinase

LPS

lipopolysaccharide

MCAO

middle cerebral artery occlusion

mGluR

metabotropic glutamate receptor

MS

multiple sclerosis

NAc

nucleus accumbens

NF-κB

nuclear factor-κB

Nrf2

nuclear factor (erythroid-derived 2)-like 2

PD

Parkinson’s disease

PERK

PKR-like endoplasmic reticulum kinase

PKR

protein kinase R

Poly I:C

polyinosinic-polycytidylic acid

ROS

reactive oxygen species

sAPP

soluble amyloid precursor protein

SOD-1

Cu/Zn superoxide dismutase

SSZ

sulfasalazine

STAT

signal transducer and activator of transcription

TLR

Toll-like receptor

TNFα

tumor necrosis factor α