System xc− in microglia is a novel therapeutic target for post-septic neurological and psychiatric illness

Post-septic neurological and psychiatric illness (PSNPI) including dementia and depression may be observed after sepsis. However, the etiology of PSNPI and therapeutic treatment of PSNPI are unclear. We show that glutamate produced from microglia through the activity of system xc− plays a role in PSNPI. We established a mouse model of PSNPI by lipopolysaccharide (LPS) treatment that shows a disturbance of short/working memory and depression-like hypoactivity. Glutamate receptor antagonists (MK801 and DNQX) reduced these phenotypes, and isolated microglia from LPS-treated mice released abundant glutamate. We identified system xc− as a source of the extracellular glutamate. xCT, a component of system xc−, was induced and expressed in microglia after LPS treatment. In xCT knockout mice, PSNPI were decreased compared to those in wildtype mice. Moreover, TNF-α and IL-1β expression in wildtype mice was increased after LPS treatment, but inhibited in xCT knockout mice. Thus, system xc− in microglia may be a therapeutic target for PSNPI. The administration of sulfasalazine, an inhibitor of xCT, in symptomatic and post-symptomatic mice improved PSNPI. Our results suggest that glutamate released from microglia through system xc− plays a critical role in the manifestations of PSNPI and that system xc− may be a therapeutic target for PSNPI.

receptor, are expressed abundantly in the brain 17,18 . Although the primary origin of released glutamate in the pathogenesis of neurological and psychiatric diseases is thought to be from neuronal cells, other type of cells such as glial cells can also release glutamate. Therefore, the all sources of extracellular glutamate in the brain have not been established. System x c − is transmembrane sodium-independent and chloride-dependent antiporter of cystine and glutamate. It consists of the 4F2 heavy chain (4F2hc; CD98; Slc3a2) and xCT (Slc7a11), a specific molecule for system x c −19-21 . xCT is expressed constitutively in immune-related organs, such as spleen, thymus, and bone marrow. However, in the CNS, the expression level of xCT is quite low in normal conditions 22 . xCT is an inducible molecule that is dependent on the transcriptional activation of NF-E2-related factor-2 (Nrf2) or hypoxia inducible factor-1α (Hif-1α) 23,24 . For instance, under oxidative stress, increased xCT can play an important role in the synthesis of the antioxidant molecule glutathione (GSH). System x c − transports extracellular cystine into the cells under the condition of oxidative stress and is the cysteine, prepared by the reduction of cysteine, is required for the major component of antioxidant GSH [19][20][21] . Although system x c − provides the resistance against oxidative stress 25 , it may affect adversely the surrounding neurons because of glutamate released in the CNS 26,27 .
In order to clarify the occurrence of post-septic neurological and psychiatric illness (PSNPI), we investigated the role and origin of glutamate in neurological and psychiatric symptoms using mice treated with a systemic administration of low dose LPS. Here, we report the importance of system x c − in microglia as a source of glutamate in the pathogenesis of post-septic memory disorder and depression-like hypoactivity. Furthermore, we also examined the therapeutic effect of xCT inhibitor and evaluated the inhibition of system x c − as a novel therapeutic target of these neurological and psychiatric symptoms.

Results systemic administration of Lps activates microglia and causes depression-like hypoactivity and memory disorder. First, we established the mouse model with LPS-induced depression and memory
disturbance. Four types of behavioral tests were used: (i) the change of body weight (BW) as a parameter of appetite; (ii) the Y-maze spontaneous alteration test (Y-maze) as a parameter of recent memory and working memory; (iii) the wheel running activity (WRA) as a parameter of hyper/hypoactivity and anxiety; and (iv) the rotarod test (RR) as a parameter of coordination and skill learning. BW was reduced transiently at the point of day 3 in LPSadministered mice, however, the difference of BW between sham and LPS-administered mice disappeared gradually (Fig. 1a). The result of the Y-maze showed a significant disturbance in recent memory and working memory in the LPS-administered mice (Fig. 1b). The WRA showed hypoactivity and mice treated with LPS showed a tendency to remain in the dark box, suggesting depression-like hypoactivity (Fig. 1c). The RR score of LPS-treated mice is significantly reduced at 15 days after LPS treatment (Fig. 1d). Together, these results demonstrate that the mice administered LPS show depression-like hypoactivity and memory disturbance.
Histological analyses by hematoxylin-eosin (HE) staining did not show obvious neuronal death in the hippocampus, cerebral cortex, and cerebellum ( Supplementary Fig. S1). We also confirmed the activation of microglia in the brain after the administration of low dose LPS. Iba-1 staining, that indicates the activation of microglia, was enhanced after LPS treatment (Fig. 1e,g,i). As reported previously, the shape of microglia is observed as ramified-type in resting state (Fig. 1f), and an amoeboid-type in mice administered with LPS (Fig. 1h,j). In the hippocampus from the mouse 15 days after LPS treatment, both ramified-type and amoeboid-type microglia were observed (Fig. 1j).
The proinflammatory cytokines TNF-α (Fig. 1k) increased in the hippocampus and cerebral cortex at 2 days after LPS administration and decreased gradually to the base line at day 9. Similarly, IL-1β (Fig. 1l) increased in the hippocampus and cerebral cortex, and cerebellum at 2 days after LPS administration and decreased gradually to the base line, while the anti-inflammatory cytokine IL-10 ( Fig. 1m) did not change significantly in our experimental conditions. extracellular glutamate is associated with the behavioral disturbance of mice treated with Lps. Glutamate has been reported to play an important role in neurological/psychiatric behavioral abnormalities. Thus, we investigated the association of glutamate with behavioral disturbance induced by LPS administration ( Fig. 2a-d). We used an antagonist for NMDA and AMPA receptors, MK801 and 6,7-dinitroquinoxaline-2,3 -dione (DNQX), respectively, to examine the association between glutamate and depression-like hypoactivity and memory disturbance. The administration of MK801 and DNQX showed a tendency to improve the loss of BW, but it was not statistically significant (Fig. 2a). In the Y-maze, MK801 significantly inhibited the memory disturbance induced by LPS (Fig. 2b), while DNQX tended to inhibit the memory disturbance, but the change was not significant. In WRA, DNQX improved the hypoactivity induced by LPS significantly, while MK801 tended to inhibit the hypoactivity, but the change was not significant (Fig. 2c). In RR, no significant difference was observed with MK801 or DNQX administration (Fig. 2d).
shaped microglia are observed at high magnification (arrowheads). (i,j) In the hippocampus from mice 15 days after LPS treatment, the Iba-1 signal is dense, and both amoeboid-shaped (arrowheads) and ramified-shaped microglia (arrows) are observed at high magnification. Time course analyses for (k) TNF-α, (l) IL-1β, and (m) IL-10. *p < 0.01 vs day 0 mice (sham-treated mice). TNF-α and IL-1β expression in the hippocampus and cerebral cortex are increased at 2 days after LPS treatment, followed by a gradual decrease. TNF-α and IL-1β expression in the cerebellum are mildly increased, but the increase is not statistically significant. IL-10 is mildly increased in the hippocampus, cerebral cortex, and cerebellum at 2 days after LPS treatment, but the increase is not statistically significant. (k-m; sham-treated; n = 7, LPS-treated 2 days; n = 7, LPS-treated 9 days; n = 3).
In order to confirm whether glutamate was actually released to extracellular space after LPS administration, we measured extracellular glutamate using in vivo microdialysis system. Although the increase of extracellular glutamate was not observed until 5 h after LPS treatment, mild increase of extracellular glutamate observed at 8 h after LPS administration, and glutamate level was reached 4-fold at 24 h after LPS treatment (Fig. 2e). Furthermore, we quantitated the release of glutamate from isolated microglia. The level of glutamate released by microglia isolated from LPS-treated mice was significantly higher than that released by microglia isolated from www.nature.com/scientificreports www.nature.com/scientificreports/ sham-treated mice (Fig. 2f). It was similar result even in the examination which divided the cerebral cortex and hippocampus ( Supplementary Fig. S2).

system x c − expressed in microglia is an important source of extracellular glutamate. Previous
reports have suggested that system x c − and gap junction hemichannel play roles in the release of glutamate 27,28 . We investigated the expression of xCT in the brain using immunofluorescent staining. In sham-treated mouse brain, quite low level of xCT was observed in cells expressing Iba-1, a major marker of microglia ( Fig. 3a-c). Immunohistochemistry indicated that inducible expression of xCT was observed mainly in Iba-1positive microglia ( Fig. 3d-l), while weak induction of xCT was observed in GFAP-positive astrocytes ( Fig. 3m-r). This induction of xCT in astrocytes was observed in later phase (day 15, Fig. 3p-r), The expression of xCT in neurons and oligodendrocytes was not detected.
We measured inducible expression of xCT in microglia isolated from sham-and LPS-treated mice using western blotting. Although the expression of xCT in the resting state (sham-treatment) was quite low, a 4-fold increase in the expression of xCT was detected in microglia from LPS-treated mice (Fig. 3s). www.nature.com/scientificreports www.nature.com/scientificreports/ Deficiency of xCT reduces LPS-induced behavioral disturbance. We investigated of the effect of a deficiency of xCT expression on LPS-induced behavioral disturbance using xCT knockout mice. We did not observed a significant difference in BW and RR between wildtype and xCT deficient mice treated with LPS (Fig. 4a,d), but a reduced disturbance in the test of Y-maze (Fig. 4b) and WRA (Fig. 4c) in xCT knockout mice was observed. www.nature.com/scientificreports www.nature.com/scientificreports/ We examined the effect of genetic depletion of xCT on glutamate release from microglia. We measured both extracellular glutamate in the hippocampus using in vivo microdialysis system and released glutamate from isolated microglia. LPS-treated xCT knockout mice did not show an increase in glutamate release (Fig. 4e,f). All columns indicate mean ± S.D., n = 5. In wildtype mice, TNF-α and IL-1β in the hippocampus and cerebral cortex are upregulated at 2 days after LPS treatment (*p < 0.01). The level of TNF-α and IL-1β in the cerebellum rises mildly, but the change is not significant. The level of IL-10 in each part of the brain shows a tendency for mild increase, but the differences are not statistically significant. In xCT knockout mice, each cytokine is upregulated mildly, but the change is not significant.
www.nature.com/scientificreports www.nature.com/scientificreports/ Deficiency of xCT inhibits the activation of microglia and the release of inflammatory cytokines. Proinflammatory cytokines TNF-α and IL-1β, as well as the anti-inflammatory cytokine IL-10 in the hippocampus, cerebral cortex, and cerebellum were measured in mice at 2 days after LPS treatment. Western blot analyses showed an increase of TNF-α and IL-1β in the hippocampus and cerebral cortex from LPS-treated wildtype mice (Fig. 5a,b,d,e). A significant increase of these cytokines in the hippocampus and cerebral cortex from LPS-treated xCT knockout mice was not detected (Fig. 5a,b,d,e). A tendency of an increase in TNF-α and IL-1β in the cerebellum was observed (Fig. 5g,h), but the increase was not statistically significant. A significant difference in IL-10 expression in these brain regions was not detected between sham-treated and LPS-treated groups both in wildtype and xCT knockout mice (Fig. 5c,f,i).
The xCT inhibitor sulfasalazine improves the memory disturbance and depression-like hypoactivity induced by Lps administration. We hypothesized that xCT may be a novel therapeutic target for memory disorder and depression-like phenotype, and investigated the effect of sulfasalazine (SSZ) on PSNPI (Fig. 6a-d). The simultaneous administration of SSZ with LPS inhibited the deterioration of recent and working memory in the Y-maze at 15 days after LPS treatment (Fig. 6b), and inhibited the hypoactivity in WRA at 8 days after LPS treatment (Fig. 6c). Moreover, the administration of SSZ at 1 h after LPS treatment, a time point when the symptoms have already appeared, also inhibited memory disturbance and hypoactivity induced by LPS administration (Fig. 6b,c). We did not observe a significant difference in BW and RR among LPS-treated mice with/without simultaneous or post-symptomatic administration of SSZ (Fig. 6a,d).
We investigated the inhibitory effect of SSZ on glutamate release by microglia from mice administered LPS. We carried out the quantification of extracellular glutamate in vivo and glutamate release from isolated microglia. SSZ administration significantly inhibited glutamate release from microglia (Fig. 6e,f).
We also investigated the effect of INI0602, an inhibitor of gap junction hemichannel, because it has been reported that the gap junction hemichannels are involved in glutamate release from microglia 28 . However, in our experimental conditions, neither an induction of connexin 43, one of the components of gap junction hemichannel, nor an inhibitory effect of INI0602 was detected ( Supplementary Fig. S3).

Discussion
Neurological and psychiatric illness may be observed in the delayed phase after sepsis 1,2 . These symptoms may occur even if the sepsis is not severe with no obvious necrosis of the brain tissue 7 . When the brain shows an inflammatory status including that from exposure to LPS, activated microglia are observed 12 . The importance of activation of microglia in neurological and psychiatric disorders has been received much attention in recent years, and has been reported in the brain due to infectious disease such as encephalitis, autoimmune diseases such as multiple sclerosis, and cerebrovascular diseases such as cerebral infarction or brain hemorrhage 12,29 . Similarly, the activation of microglia is observed in neurodegenerative and/or psychiatric diseases, including Parkinson's disease, Alzheimer's disease, amyotrophic lateral sclerosis, depression, and schizophrenia 15,16,[30][31][32] .
Most of the studies describing microglial activation in neurological and/or psychiatric diseases report that proinflammatory cytokines or nitric oxide are critical triggers for neurological and/or psychiatric symptoms 15,16,[30][31][32] . However, neuronal excitotoxicity induced by the excitatory amino acid glutamate is reported to be important in several neurological and psychiatric diseases 18 . Indeed, the inhibition of excitotoxicity by glutamate is used as a therapy in Alzheimer's disease and amyotrophic lateral sclerosis 33,34 .
Using our established mouse model of memory disorder and depression-like hypoactivity after the treatment with LPS (Fig. 1), we investigated the role of glutamate in neurological and psychiatric symptoms and the sources of released glutamate. As shown in Fig. 2, the NMDA receptor antagonist MK801 and AMPA receptor antagonist DNQX reduced the symptoms induced by systemic administration of LPS. This result suggests that the recent/ working memory disorder and depression-like hypoactivity are associated with extracellular glutamate. Since half-life time of these antagonists was quite short, about 2 h, MK801 or DNQX was injected not only on day 0 but also on day 1. Although it is known that NMDA antagonist can inhibit both cognitive decline and depressive symptoms, stimulation of AMPA receptor can bring an antidepressant effect 35 . Therefore, we hypothesized that AMPA antagonist DNQX could not improve the WRA score. However, DNQX improved the WRA score significantly in our experimental condition, suggesting that not only depression or anxiety but also motor function may affect this WRA score. Since microglia are activated in the brain with systemic LPS treatment, we focused on microglia as an origin of extracellular glutamate. We measured extracellular glutamate in the hippocampus using in vivo microdialysis technique, and we also measured glutamate released from microglia isolated from the mice with/without systemic administration of LPS. As shown in Fig. 2e,f, released glutamate from LPS-treated mice was significantly higher than that from sham-treated mice, suggesting that microglia may be an important source of extracellular glutamate in our experimental model.
Next, we focused on microglia as an origin of glutamate release in the pathogenesis of LPS-induced neurological and psychiatric disorders. Since microglia from LPS-treated mice released higher levels of glutamate than sham-treated mice, we hypothesized that an inducible mechanism for glutamate release is active in the symptomatic condition. Previous studies have identified system x c − and gap junction hemichannel as inducible systems 27,28 . A gap junction hemichannel in microglia has been reported as an important source of glutamate release in some neurodegenerative disease. Furthermore, Takeuchi et al. reported that INI0602, an inhibitor of gap junction hemichannel, suppresses disease progression of Alzheimer's disease and amyotrophic lateral sclerosis 28 . In the present study, neither inducible expression of connexin 43, one of the components of gap junction hemichannel, nor an inhibitory effect of INI0602 on glutamate release from microglia was observed (Supplementary Fig. S3). On the other hand, we detected inducible expression of xCT, a special molecule for system x c − , in microglia by both western blot analysis using microglia lysate and by immunohistochemistry (Fig. 3). Since the expression of xCT in astrocytes in LPS or 6-hydroxydopamine-treated animal model has been reported 36,37 , we also investigated www.nature.com/scientificreports www.nature.com/scientificreports/ the expression of xCT in astrocytes. The expression of xCT in astrocytes was also detected in later phase (Fig. 3q), but the expression level of xCT in astrocytes was lower than that in microglia. It suggests that system x c − in microglia may play an important role in induction of PSNPI in acute or subacute phase after LPS administration but system x c − in astrocytes may play another role especially in delayed phase. www.nature.com/scientificreports www.nature.com/scientificreports/ In order to confirm the importance of system x c − in the increase of extracellular glutamate in response to systemic administration of LPS, we examined released glutamate from isolated microglia and the behavioral analyses using xCT knockout mice. Isolated microglia derived from xCT knockout mice, even if treated with LPS, did not increase the level of glutamate release, while those derived from wildtype mice increased glutamate release. These results strongly suggest that microglial system x c − may play a crucial role in glutamate release in the brain in response to systemic administration of LPS.
Since system x c − is known as a transporter that induces an influx of cystine, the oxidized dimer form of amino acid cysteine, to the cells, it is very important to maintain cellular conditions under oxidative stress by supplying antioxidant GSH, which includes cysteine residue in its tripeptide 12,20 . Therefore, the absence of xCT may induce a vulnerability in microglia. To assess this point, we measured and compared the level of proinflammatory or anti-inflammatory cytokines in the brain of wildtype and xCT knockout mice. Recent studies suggest that microglia adopt two functionally distinct phenotypes, composed of phenotypes of "classical" activation (M1) and "alternative" activation (M2), and activated M1 microglia produce neurotoxic cytokines such as TNF-α and IL-1β 38,39 . Although the proinflammatory cytokines TNF-α and IL-1β were increased in the hippocampus and cerebral cortex collected from wildtype mice at 2 days after LPS administration, these increases were not observed in xCT knockout mice (Fig. 5). The expression level of the anti-inflammatory cytokine IL-10 was similar in wildtype and xCT knockout mice (Fig. 5). These expression patterns of cytokines are consistent with the neurological and psychiatric phenotype. For example, recent memory disturbance is believed to be associated with the neuronal dysfunction of the hippocampus where the increased cytokines were detected.
Our results suggest that system x c − may be a novel therapeutic target of memory disorder and depression-like hypoactivity. SSZ is used widely in the treatment of autoimmune diseases, such as ulcerative colitis and rheumatoid arthritis 40,41 . A detailed mechanism of this compound's effects on these diseases has not been clarified. However, a recent report described SSZ has inhibitory effect to xCT 42 . Moreover, it has been reported that SSZ can pass through the BBB via the transporter Oatp2b1 43,44 . Therefore, we suggest that SSZ administration may improve the neurological and psychiatric symptoms observed after LPS administration. Simultaneous administration of SSZ with LPS reduced memory disturbance and depression-like hypoactivity almost to the sham-treated level. Surprisingly, post-symptomatic treatment with SSZ improved those symptoms to similar levels as simultaneous treatment with SSZ (Fig. 6). These results suggest that xCT may be novel therapeutic target of neurological and psychiatric disorders in the pathogenesis of post-septic illness. Moreover, the inhibition of system x c − may be useful for the medication of other neurological and psychiatric diseases.

Methods
Animals and housing. C57BL/6 mice were obtained from CLEA Japan (Tokyo, Japan). The xCT knockout mice used in this study were established and reported previously by H. Sato 45 . Since it has been reported that the morbidity of dementia and depression in women is more frequent than that in men 46 , we used female mice in this study. Wildtype (150 mice) and xCT knockout mice (50 mice) (7-8 weeks old) were used in this study. The mice were housed under standardized conditions of light (06:00-18:00), temperature (25 °C), and humidity (approximately 50%), and were allowed free access to food and water. Experiments were performed with age-and weight-matched animals. The Institutional Animal Care Committees of the Tottori University, Yonago, Japan approved all procedures (No. 16-Y-6, 18-Y-3, 19-Y-4), and all animal experiments were performed in accordance with relavant guidelines and regulations. experimental design. Experimental procedures are described in Fig. 7. The mice were divided randomly into 2-4 groups, and intraperitoneal injection (i.p.) of 0.5 mg/kg BW of LPS (Sigma-Aldrich, Saint Louis, MO) was carried out [LPS group] with/without several inhibitors. For the control group, mice were administrated an equal volume of PBS (i.p.) [Sham group]. MK801, DNQX, and SSZ were obtained from Sigma-Aldrich, Tocris (Bristol, UK), and Tokyo Chemical Industry (Tokyo, Japan). The dose and timing of these chemical compounds is shown in Fig. 7.
Behavioral test. Body weight (BW). As an index of appetite, the body weight of experimental mice was measured at day 0 (before LPS administration), 3, 8, and 15 (Fig. 7a).

Y-maze test (Y-maze).
The Y-maze apparatus (YM-03M; Muromachi, Tokyo, Japan) was composed of three arms (10 cm wide, 13 cm high, and 35 cm long) allowing the mice to see distal spatial features 47 . The insides of the arms were identical to each other, and no interior cues were provided. Continuous spontaneous alternation testing was performed by placing the mice in the Y-maze for 8 min with all three arms. The number and sequence of arms entered were recorded manually. One alternation was counted when mice visited three different arms consecutively. Immediate re-entries were discounted. The percentage of alternation was indicated as spatial working memory. The spontaneous alternative rate (%) was calculated as the number of alternations/(the number of total arm entries − 2) × 100 (Fig. 7a).
Wheel running activity (WRA). WRA was investigated using the MK-750PC (Muromachi) 47 . Locomotor and running activity was converted to revolutions per day. The mice were allowed free exercise, rest, food, and water for 24 hours (Fig. 7a).
Rotarod test (RR). The motor coordination of the mice was investigated using an accelerating rotarod apparatus (MK-630B, Muromachi) 47 . The mice were placed on the rod for four successive trials. The two highest scores of the four trials were used to calculate an average score, which indicates the latency to fall. During each trial, the www.nature.com/scientificreports www.nature.com/scientificreports/ rotating rod accelerates gradually from 0 to 40 rpm over a period of 5 min, and, thereafter, is maintained at the 40 rpm speed (Fig. 7a).
In vivo microdialysis. We measured glutamate released in extracellular space of the brain in freely moving mice using in vivo microdialysis system. Mice was anesthetized with mixture of medetomidine (0.3 mg/kg), midazolam (4 mg/kg), and butorphanol (5 mg/kg) (i.p.), and guide cannula (S-size: Arcrize Japan, Kurume, Japan) was implanted stereotaxically into the hippocampus (anterior: −3.8 mm, lateral: 2.0 mm, from the bregma, ventral:1.6 mm from the skull). A dialysis probe (S-size: Arcrize Japan) was inserted though the guide cannula and perfused with artificial cerebrospinal fluid (aCSF; 147 mM NaCl, 4 mM KCl, 2.3 mM CaCl 2 ) at a flow rate of 1.0 μl/min. Outflow fractions (60 μl) were collected at the point of 0, 1, 3, 5, 8, and 24 h. The level of extracellular glutamate was measured by the method described below, and the results are expressed as the relative increase of glutamate levels based on the start time (0 h).
Isolation of microglia and glutamate assay. In this study, microglia were isolated from whole brain as described previously 48 with some modifications. Briefly, following decapitation, the whole brain except the cerebellum was readily extracted and chopped finely with a fine sharp scissor in serum-free Dulbecco's modified Eagle's medium (DMEM: Sigma-Aldrich) containing trypsin-EDTA (final concentrate 0.05%, Thermo Fisher Scientific, Waltham, MA), dispase II (final concentrate 2 units/ml, Roche, Basel, Switzerland), and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA). The brain pieces prepared were incubated at 37 °C for 30 min. Enzymatic digestion was terminated by adding DMEM containing 20% fetal bovine serum and 1% penicillin/streptomycin. The brain pieces were further triturated by gently pipetting and passing the tissue through a 100-μm cell strainer (Corning, Corning, NY) to remove cell debris and undigested tissue pieces. The filtered cell suspension was centrifuged at 1,000 × g for 5 min at 4 °C, and the supernatant was decanted. The cell pellet was then re-suspended by slow pipetting with 30% isotonic Percoll (GE Healthcare, Buckinghamshire, UK) in Hank's balanced salt solution (HBSS) (Thermo Fisher Scientific) and centrifuged at 500 × g for 20 min at 4 °C. After centrifugation, the supernatant was aspirated, and the cell pellet was re-suspended in HBSS containing 0.1 μM L-cysteine (L-cysteine converts to cystine easily by autoxidation) with/without SSZ or INI0602 (Sigma-Aldrich). The eliminative process using red blood cell lysis buffer was omitted to avoid cytotoxicity.
The Glutamate-Glo TM assay kit (Promega, Madison, WI) was used to measure glutamate efflux levels according to the manufacturer's protocol. Primary microglia were suspended at 1 × 10 5 cells/ml in HBSS containing 0.1 μM L-cysteine, then placed into 24-well culture plates. After cells were cultured at 37 °C under 5% CO 2 in air for 30 min, the medium was collected for measurement.