The Glial Glutamate Transporter 1 (GLT1) Is Expressed by Pancreatic β-Cells and Prevents Glutamate-induced β-Cell Death*

Glutamate is the major excitatory neurotransmitter of the central nervous system (CNS) and may induce cytotoxicity through persistent activation of glutamate receptors and oxidative stress. Its extracellular concentration is maintained at physiological concentrations by high affinity glutamate transporters of the solute carrier 1 family (SLC1). Glutamate is also present in islet of Langerhans where it is secreted by the α-cells and acts as a signaling molecule to modulate hormone secretion. Whether glutamate plays a role in islet cell viability is presently unknown. We demonstrate that chronic exposure to glutamate exerts a cytotoxic effect in clonal β-cell lines and human islet β-cells but not in α-cells. In human islets, glutamate-induced β-cell cytotoxicity was associated with increased oxidative stress and led to apoptosis and autophagy. We also provide evidence that the key regulator of extracellular islet glutamate concentration is the glial glutamate transporter 1 (GLT1). GLT1 localizes to the plasma membrane of β-cells, modulates hormone secretion, and prevents glutamate-induced cytotoxicity as shown by the fact that its down-regulation induced β-cell death, whereas GLT1 up-regulation promoted β-cell survival. In conclusion, the present study identifies GLT1 as a new player in glutamate homeostasis and signaling in the islet of Langerhans and demonstrates that β-cells critically depend on its activity to control extracellular glutamate levels and cellular integrity.


Glutamate is the major excitatory neurotransmitter of the central nervous system (CNS) and may induce cytotoxicity through persistent activation of glutamate receptors and oxidative stress. Its extracellular concentration is maintained at physiological concentrations by high affinity glutamate transporters of the solute carrier 1 family (SLC1). Glutamate is also present in islet of Langerhans where it is secreted by the ␣-cells and acts as a signaling molecule to modulate hormone secretion. Whether glutamate plays a role in islet cell viability is presently unknown.
We demonstrate that chronic exposure to glutamate exerts a cytotoxic effect in clonal ␤-cell lines and human islet ␤-cells but not in ␣-cells. In human islets, glutamate-induced ␤-cell cytotoxicity was associated with increased oxidative stress and led to apoptosis and autophagy. We also provide evidence that the key regulator of extracellular islet glutamate concentration is the glial glutamate transporter 1 (GLT1). GLT1 localizes to the plasma membrane of ␤-cells, modulates hormone secretion, and prevents glutamate-induced cytotoxicity as shown by the fact that its down-regulation induced ␤-cell death, whereas GLT1 up-regulation promoted ␤-cell survival. In conclusion, the present study identifies GLT1 as a new player in glutamate homeostasis and signaling in the islet of Langerhans and demonstrates that ␤-cells critically depend on its activity to control extracellular glutamate levels and cellular integrity.
Diabetes mellitus is characterized by absolute or relative insulin deficiency due to an inadequate mass of insulin-produc-ing ␤-cells and concomitant insulin resistance (1)(2)(3). In both autoimmune type 1 diabetes mellitus (T1DM) 4 and non-autoimmune type 2 diabetes mellitus (T2DM), the loss of ␤-cells starts years before the onset of overt hyperglycemia. ␤-Cell autoimmunity, ␤-cell overstimulation, glucotoxicity, and lipotoxicity are recognized causes of ␤-cell death in diabetes, and research is underway to identify specific therapeutic protocols (4,5). The discovery of new factors that might harm the ␤-cell would increase our understanding of diabetes pathogenesis and our chances to design novel cytoprotective treatments capable to arrest ␤-cell death.
Pancreatic ␤-cells show numerous common features with neurons including shared transcriptional activators (6), the expression of proteins specialized in synaptic transmission, and the presence of neurotransmitters such as ␥-aminobutyric acid (GABA) and glutamate (7,8). Islet cells express functional glutamate receptors (9 -12) and vesicular glutamate transporters (13), suggesting that glutamate is a crucial intercellular signal mediator. Glutamate is co-secreted with glucagon, and acute stimulation of its receptors modulates the secretion of glucagon, insulin, and somatostatin (8, 14 -19). Nevertheless, our knowledge regarding the effects of glutamate on islet cell physiology is still incomplete.
In the CNS, glutamate is an excitatory neurotransmitter that binds to ionotropic and metabotropic glutamate receptors. Extracellular glutamate must be rapidly removed from the synaptic cleft to control synaptic events and to prevent sustained activation of ionotropic receptors, which has a potent neurotoxic effect (20). The clearance of extracellular glutamate is performed by the glutamate transporters of the SLC1A family located on the plasma membrane of astroglial and neuronal cells. Five Na ϩ -dependent high affinity glutamate transporters * This work was supported by University Research Program 2008 (to C. P.) and National Institutes of Health Grant RO1 DK080148 (to F. F.). E. S. D. C., A. M. D., F. F., and C. P. are inventors in a Patent Cooperation Treaty application. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental materials and methods and Figs. S1-S5. 1  (EAAT1-5) have been described (21), but GLT1/EAAT2 exhibits the highest level of expression in the CNS (22). Mice deficient in GLT1 show increased susceptibility to acute cortical injuries (23), and inhibition of GLT1 activity increases the glutamate concentration to toxic levels (22). A glutamate uptake activity has been described in pancreatic islets, and a high affinity glutamate/aspartate transporter has been cloned from the human pancreas (24,25); however, its precise identity and physiological role have not been characterized. In the present study, we demonstrate that ␤-cells are selectively vulnerable to glutamate-induced toxicity and that GLT1 is present and functional in the islet where it influences glucagon release and promotes ␤-cell survival.

Cell Lines
Mouse ␤TC3 and ␣TC1 cells were kindly provided by Prof. Douglas Hanahan (Department of Biochemistry and Biophysics, University of California, San Francisco, CA). The INS1 rat cells were originally provided by Prof. Claes B. Wollheim (Department of Internal Medicine, University Medical Centre of Geneva, Geneva, Switzerland).

Human Islet Isolation and Culture
The islets used in this study were isolated from seven cadaveric multiorgan donors by using the procedure already described by Ricordi et al. (26) in conformity to the ethical requirements approved by the Niguarda Cà Granda Ethics Board. Cell lines and human islets were cultured in RPMI 1640 medium (Sigma) containing 0.1 mM glutamic acid and supplemented with 0.7 mM glutamine as described (27). The glucose concentration was 11 mM for cell lines and 5.5 mM for islets.
Cell Death Detection ELISA-Batches of 40 human isolated islets were cultured for 3 days in different conditions, and apoptosis was quantified with the Cell Death Detection ELISA assay (Roche Applied Science) following the manufacturer's protocols. Data were expressed as -fold increase over control samples.
Terminal Deoxynucleotidyltransferase-mediated dUTP-biotin Nick End Labeling (TUNEL) Assay-A TUNEL (Promega) assay was used to estimate apoptosis in ␤TC3 cells. TUNEL-positive cells were counted by two independent observers using a 40ϫ objective from at least 40 randomly selected fields per coverslip. Data were plotted as the number of TUNEL-positive cells per field.

Quantification of Total SH Content
␤TC3 cells were incubated for 5 days in 5 mM glutamate and then lysed using a standard method. Cleared lysate was then incubated in reaction buffer (0.1 M sodium phosphate, pH 8, 1 mM EDTA) and 5Ј,5Ј-dithiobis(2-nitrobenzoic acid (Sigma; stock solution, 4 mg/ml) at room temperature. The absorbance at 412 nm was measured after 30 min. A standard curve was generated using 0 -0.5 mM reduced GSH.

Glutamate Determination
␤TC3 cells were plated on 6-cm Petri dishes, grown in standard medium for 24 h, and then treated for 5 days with 0.3 mM DHK (Sigma-Aldrich). Cell media from treated and control plates were collected and used to assess the glutamate concentration by enzymatic assay (Glutamate-Glutamine Determination kit, Sigma-Aldrich) following the manufacturer's protocols.

RNA Isolation and RT-PCR Analysis
Total RNA from 9 ϫ 10 6 ␤TC3 and ␣TC1 cells or 1500 isolated human islets was extracted with an RNA fast isolation system (Molecular Systems, San Diego, CA). For cDNA synthesis, 2 g of digested RNA was reverse transcribed using random oligonucleotides (final concentration, 12.5 ng/l) as primers and 200 units of Moloney murine leukemia virus reverse transcriptase in the presence of RNaseOUT, 0,1 M DTT, and dNTPs. Reagents were from Invitrogen. Primers used for PCR amplification of the reverse transcribed RNA and PCR conditions are reported in the supplemental Materials and Methods. To confirm the absence of genomic contamination in RNA samples, reverse transcriptase-negative controls were introduced in each experiment (no Moloney murine leukemia virus reverse transcriptase).

Cell Lysis, Immunoprecipitation, and Western Blotting Analysis
9 ϫ 10 6 ␤TC3 or ␣TC1 cells or 1500 human islets were harvested and lysed in 100 l of lysis buffer (150 mM NaCl, 30 mM Tris-HCl, 1 mM MgCl 2 , 1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and 1 g/ml aprotinin and leupeptin) as described (28). For immunoprecipitation experiments, lysates were incubated overnight with the indicated antibodies and protein A-Sepharose. Immunocomplexes were resolved by 9% SDS-PAGE. The blots were probed with the indicated primary antibodies followed by anti-rabbit HRP-conjugated IgG (80 ng/ml; Amersham Biosciences/GE Healthcare) and visualized by ECL (PerkinElmer Life Sciences). For the P2 extract, total homogenates of rat brain tissues were prepared as described previously (28).

Immunofluorescence
Cell Cultures-␤TC3 cells were plated onto sterile glass coverslips, fixed in ice-cold methanol, and permeabilized with 0.5% Triton X-100 in PBS. Immunostaining with primary antibodies was followed by incubation with the appropriate secondary antibodies.
Co-localization Analysis-Single channel images obtained with a Bio-Rad confocal microscope (GLT1, channel 1/red; hormones, channel 2/green) were analyzed for pixel intensity, and co-localization between the two channels was calculated with ImageJ software. The product of the differences from the mean intensity (PDM value) was calculated in each location of the image as follows: PDM ϭ (red intensity Ϫ mean red intensity) ϫ (green intensity Ϫ mean green intensity).

Ultrastructural Analysis
For ultrastructural analysis, samples of human islets cultured in different conditions (0, 0.5, and 5 mM glutamate and 0.5 mM glutamate plus 0.025 mM CNQX) were fixed in a mixture of 2% paraformaldehyde and 2% glutaraldehyde, postfixed in 1% osmium tetroxide, and embedded in Epon-Araldite (Sigma-Aldrich). After counterstaining with uranyl acetate and lead citrate, thin sections were examined with a Morgagni electron microscope (Philips, Eindhoven, Netherlands). The quantification of apoptotic/degenerated cells was performed in electron microscopy samples by counting at least two different series of pancreatic cultured islets maintained in different conditions. At least 100 ␤-cells and 100 ␣-cells were observed in each experiment. Results were statistically evaluated using the 2 test.

Insulin, Proinsulin, and Glucagon Secretion
The total insulin, proinsulin, and glucagon contents were measured by static incubations in lysates and media of batches of 25 human isolated islets after acute (1 h) or chronic (3 days) exposure to glutamate, DHK, or both. Islets were lysed in lysis buffer, and the medium was collected. Hormone secretion and content were determined by means of an ELISA immunoassay (ALPCO Diagnostic).

Statistical Analysis
Data are presented as means Ϯ S.E. of at least three independent experiments. The number of replicates for each experiment is reported. Statistical comparisons were performed by

. Chronic incubation with glutamate induces ␤-cell apoptosis via ionotropic receptor activation and increased oxidative stress.
A, dose-dependent effect of glutamate on ␤-cell viability. ␣TC1 and ␤TC3 cells were incubated for 5 days with the indicated glutamate concentrations, and viability was measured by the MTT assay. The relative growth rate (RGR) is presented as percentage of 0 mM glutamate (n ϭ 7 with eight replicates). *, p Ͻ 0.05; **, p Ͻ 0.01 versus 0 mM glutamate. B, time-dependent effect of glutamate (Glu) on ␤-cell viability. ␤TC3 cells were incubated with 5 mM glutamate for the indicated times, and viability was measured by the MTT assay. Data are presented as percentage of each relative control (CTR) (without glutamate) (n ϭ 4 with eight replicates). **, p Ͻ 0.01; ***, p Ͻ 0.001 versus relative control. C, glutamate increases ␤-cell apoptosis. ␤TC3 cell apoptosis was quantified by TUNEL assay after incubation with 0.5 mM glutamate for 5 days. Cell nuclei were labeled with propidium iodide, and data are presented as the number of TUNEL-positive cell per field (n ϭ 3 in duplicate). Bar, 50 m. *, p Ͻ 0.05. D, CNQX protects ␤-cells from glutamate toxicity. ␤TC3 cells were incubated for 5 days with or without 5 mM glutamate in the presence of ionotropic receptor antagonist CNQX (0.025 mM) or D-2-amino-5-phosphonovaleric acid (APV) (0.1 mM), and the viability was measured by MTT assay (n ϭ 3 with eight replicates). *, p Ͻ 0.05; **, p Ͻ 0.01 versus control 0 mM glutamate; #, p Ͻ 0.05; ##, p Ͻ 0.01 versus control 5 mM glutamate. E, glutamate causes oxidative stress in ␤-cells. SH content in ␤TC3 lysates exposed to 5 mM glutamate for 5 days was quantified. *, p Ͻ 0.05. Inset, RT-PCR analysis on cDNA from ␤TC3 cells showing the transcript for the glutamate/cysteine exchanger subunit xCT. Ϫ, RT-PCR without enzymes. The error bars indicate standard errors.
unpaired Student's t test or analysis of variance followed by multiple post hoc comparison analysis performed with Tukey's test. The difference was considered statistically significant when the p value was Յ0.05.

RESULTS
␤TC3 Cells Are Vulnerable to Glutamate-induced Cytotoxicity-We first determined whether ␤and ␣-cell lines were vulnerable to extracellular glutamate. ␤TC3 and ␣TC1 were cultured in the presence of different glutamate concentrations (from 0.1 to 5 mM), and cell viability was assayed by the MTT test. Exposure to glutamate induced a dose-and time-dependent decrease of ␤TC3 viability (effective dose (ED 50 ), 0.5 mM for a 5-day exposure to glutamate) (Fig. 1, A and B). As previously shown in MIN6 cells (30), we did not observe glutamate-induced cytotoxicity after a 20-min incubation at 0.5 or 5 mM as usually occurs in neurons (28). Notably, ␣TC1 cells were resistant to glutamate-induced toxicity, showing a small 22.2 Ϯ 3.3% decrease in viability only after a 5-day exposure to 5 mM glutamate (Fig. 1A).
Glutamate caused ␤TC3 death by apoptosis as shown by the fact that 5-day incubation in 0.5 mM glutamate, a physiological concentration, significantly increased the number of TUNELpositive ␤TC3 cells by 54 Ϯ 2% (Fig. 1C). These data indicate that ␤-cells are vulnerable to glutamate-induced cytotoxicity, whereas ␣-cells are resistant.
Excitotoxicity and Oxidative Stress Are Responsible for Glutamate-induced ␤TC3 Apoptosis-A well characterized mechanism of glutamate toxicity is excessive activation of ionotropic glutamate receptors and cytosolic Ca 2ϩ increase (so-called excitotoxicity) (20). To verify whether ␤TC3 cells express functional glutamate receptors and respond to glutamate with an increase in intracellular Ca 2ϩ , they were loaded with the Ca 2ϩ indicator Fluo3 (supplemental Fig. S1). 88 Ϯ 3% of monitored cells responded to 1 mM glutamate with an increase in the fluorescence signal. The effect was abolished by co-application of 0.025 mM CNQX, an ␣amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and kainate receptor antagonist, thus indicating that ␤TC3 cells express functional AMPA/kainate glutamate receptors whose activation induces an increase in intracellular Ca 2ϩ .
To test the involvement of these receptors in glutamate toxicity, cells were co-incubated with glutamate and CNQX for 5 days (Fig. 1D). We demonstrated that glutamate toxicity was partially prevented by the co-administration of 0.025 mM CNQX (46 Ϯ 3% increase in relative growth rate relative to 5 mM glutamate; p Ͻ 0.05) but not by 0.1 mM D-2-amino-5-phosphonovaleric acid, a selective N-methyl-D-aspartate receptor antagonist. In the absence of glutamate supplementation, ␤TC3 viability was not affected by the presence of both glutamate receptor antagonists. These data indicate that glutamateinduced ␤TC3 cytotoxicity is due in part to activation of AMPA and/or kainate ionotropic glutamate receptors as occurs in the CNS (31). Increased extracellular glutamate may also lead to cell death by oxidative stress (oxidative glutamate toxicity), which is mediated by the glutamate/cysteine exchanger (x Ϫc ) (32). We found that chronic incubation with glutamate caused a significant reduction in the total SH (Fig. 1E) content in ␤TC3 cells that expressed the xCT subunit of the x Ϫc exchanger (Fig.  1E, inset), thus suggesting that glutamate-induced ␤-cell death might be also mediated by oxidative stress.
Na ϩ -dependent Glutamate Transporter GLT1 Is Selectively Expressed by ␤-Cells-We then explored whether ␤-cells express a glutamate clearance system that controls the extracellular glutamate concentration. RT-PCR performed in clonal ␤and ␣-cells (␤TC3, INS1, and ␣TC1 cells) demonstrated that the GLT1a subtype was the only glutamate transporter expressed by the ␤-cells ( Fig. 2A). None of the canonical high affinity glutamate transporters (SLC1A1-3) were detected in ␣TC1 cells. Conversely, both cell lines expressed the neutral amino acid transporters ASCT1 (SLC1A4) and ASCT2 (SLC1A5). Western blot confirmed the expression of the GLT1 protein in ␤TC3 cells (Fig. 2B). Two bands running with an apparent electrophoretic mobility of 60 and 120 kDa were detected in immunoblots of both ␤TC3 and brain lysates. Conversely, no bands were detected using a control serum. Immunofluorescence analysis (Fig. 2C) revealed that GLT1 was mainly localized to the plasma membrane of ␤TC3 cells. Accordingly, a Na ϩ -dependent D-[ 3 H]aspartic acid uptake was measured in ␤TC3 cells (Fig. 2D). In contrast, the uptake of D-[ 3 H]aspartic acid was prevalently Na ϩ -independent in ␣TC1 cells (Fig. 2D). Subtype-specific expression of glutamate transporters was characterized pharmacologically using DHK, a selective GLT1 inhibitor (33), and HIP-A, a non-selective high affinity glutamate transporter blocker (34). In ␤TC3, the Na ϩ -dependent uptake of D-[ 3 H]aspartic acid was completely inhibited by both 0.025 mM HIP-A and 0.1 mM DHK, indicating that the uptake of glutamate in these cells is exclusively driven by GLT1 (Fig. 2D).
GLT1 Protects ␤TC3 from Glutamate Toxicity-Having established the selective expression of GLT1 in ␤-cells, we investigated its physiological relevance. Exposure to DHK induced a dose-dependent inhibition of Na ϩ -dependent D-[ 3 H]aspartic acid uptake in ␤TC3 (IC 50 ϭ 0.05 mM DHK) (supplemental Fig. S2A) and significantly increased the concentration of glutamate in the medium (Fig. 3A). In addition, DHK induced a decrease in ␤TC3 viability that was already maximal at 0.05 mM DHK (supplemental Fig. S2B). Cytotoxicity was due to apoptosis as revealed by a TUNEL assay performed on ␤TC3 incubated with 0.1 mM DHK for 5 days (Fig. 3B).
Similar results were obtained by down-regulating the expression of the GLT1 gene in ␤TC3 by means of short hairpin RNA (shRNA). Two different shRNAs (SH1 and SH3) were effective in reducing the total expression of GLT1 as shown by Western blotting and uptake experiments (supplemental Fig. S2, C and D). A 35% down-regulation of GLT1 activity was sufficient to increase ␤TC3 cell apoptosis by 2-4-fold after a 24-h incubation in 0.5 mM glutamate (Fig. 3C). Interestingly, the SH1 shRNA construct increased ␤TC3 apoptosis also in the absence of glutamate supplementation, suggesting that impaired GLT1 activity "per se" is sufficient to induce ␤TC3 cell death (Fig. 3C). No effect was observed in the presence of a control shRNA. Taken together, these data suggest that the function of GLT1 is to control the extracellular glutamate concentration and preserve ␤-cell survival.
To further support this role of GLT1, we up-regulated its expression by pharmacological treatment with CEF. CEF is a ␤-lactam antibiotic that has been shown to increase brain GLT1/EAAT2 expression and activity and to induce neuroprotection from glutamate toxicity in models of ischemic injury and motor neuron degeneration (35,36). A 5-day incubation with 10 and 100 M CEF induced a 2-fold increase in GLT1 expression also in ␤TC3 cells (Fig. 3D) and led to a dose-dependent protection from glutamate-induced toxicity (Fig. 3E). These data confirm that GLT1 is a key player in the control of glutamate homeostasis and in the maintenance of ␤-cell integrity.
Human ␤-Cells Express a GLT1 System-Given the possible implications in ␤-cell physiology and pathology, we validated our findings on glutamate toxicity and GLT1 cytoprotection in human islets. We first confirmed the expression of GLT1 in human and monkey pancreatic sections by immunohistochemistry. As shown in Fig. 4A, in both human and monkey pancreases, anti-GLT1 reactivity was limited to the islet, whereas anti-GLT1 reactivity was absent in the exocrine tissue. In both species but even more clearly in human, GLT1 staining was almost exclusively localized to the cell membrane. Islets did not show GLT1 staining in the presence of GLT1 preimmune serum or a blocking peptide (supplemental Fig. S3).
To more precisely localize GLT1 expression in human islets, we performed double immunofluorescence experiments using hormones (red) as markers of endocrine cell ␤TC3 cells were incubated with 0.1 mM DHK for 5 days, and cell apoptosis was determined by the TUNEL assay. Cell nuclei were labeled with propidium iodide (n ϭ 3 in duplicate). Bar, 50 m. *, p Ͻ 0.05. C, GLT1 knockdown increases ␤-cell apoptosis. ␤TC3 cells were transfected with control shRNA (SHC) or GLT1 shRNA (SH1 and SH3) to down-regulate GLT1 and incubated with or without 0.5 mM glutamate (Glu). Three days later, apoptosis was determined by the TUNEL assay. Cell nuclei were labeled with propidium iodide (n ϭ 3 in duplicate). Bar, 50 m. *, p Ͻ 0.05 versus control shRNA at 0 mM glutamate; **, p Ͻ 0.01 versus control shRNA at 0.5 mM glutamate. D, incubation of ␤TC3 cells with 10 or 100 M CEF for 5 days increases the GLT1 expression. After incubation with the indicated CEF and glutamate concentrations for 5 days, 100 g of each total cell lysate was immunoblotted (IB) with anti-GLT1 and anti-actin antibodies. A representative blot is shown. GLT1 expression was quantified by densitometry and normalized to actin content (n ϭ 3). *, p Ͻ 0.05. E, ceftriaxone protects ␤TC3 cells from glutamate toxicity. Cells were incubated for 5 days with the indicated ceftriaxone and glutamate concentrations, and ␤-cell viability was determined by MTT assay (n ϭ 3 with eight replicates). *, p Ͻ 0.05; ***, p Ͻ 0.001 versus 0 M CEF; ##, p Ͻ 0.01 versus 0 M CEF at 5 mM glutamate. CTR, control. The error bars indicate standard errors. types (Fig. 4B). Immunostaining with chromogranin, a marker of endocrine cells, confirmed that GLT1 was expressed only in the islet. GLT1 signal was concentrated in the plasma membrane where it co-localized with chromogranin granules fused to the cell membrane as shown by the yellow/orange staining at the cell periphery. Similar yellow/orange staining at cell-cell boundaries was detected in insulin-positive cells, indicating GLT1 expression in ␤-cells. In contrast, we did not observe co-localization of GLT1 with both glucagon and somatostatin, suggesting that ␣and ␦-cells did not express GLT1 or that if expressed it was under the level of detection. This conclusion was confirmed by confocal image analysis (Fig. 4, C and D). Taken together, these data indicate that, in humans, GLT1 is mainly expressed in ␤-cells where it localizes to the cell membrane.
GLT1 Expressed by ␤-Cells Regulates Clearance of Glutamate in Isolated Human Islets-We next sought to verify the physiological relevance of GLT1 in human isolated islets, an in vitro model system proven to be useful to study the mechanisms involved in ␤-cell function and death (27). We therefore confirmed that GLT1 is expressed by ␤-cells of isolated human islets (Fig. 5, A, B, and C). Moreover, we demonstrated that GLT1 was the main regulator of the glutamate clearance in isolated islets. In fact, selective GLT1 inhibition with DHK almost completely blocked the uptake of glutamate in isolated human islets (Fig. 5D).
Although data are somewhat controversial, a recognized effect of glutamate is to modulate the release of insulin, glucagon, and GABA induced by changes in glucose concentration (15)(16)(17)(18)(19). In our experiments, glutamate decreased the physiological release of glucagon in response to an acute fall in glucose concentrations (Fig. 5E), whereas glutamate did not potentiate insulin secretion in response to an acute glucose change as previously reported by others (8,16) (Fig. 5F). Similar results were obtained in the presence of DHK, suggesting that GLT1 may be involved in the regulation of glucagon secretion but not in that of insulin. The effects of glutamate and DHK on both glucagon and insulin release were not additive, suggesting a common mechanism of action. In the inset, a particular region is shown at higher magnification (2ϫ). The yellow/orange staining indicates co-localization between the transporter and hormones. C, scatter plot analysis of GLT1 (red; channel 1) and hormone (green; channel 2) staining generated from B with ImageJ software reveals partial co-localization of GLT1 with chromogranin and insulin. D, intensity correlation analysis of single channel images from B supports co-localization between GLT1 and insulin or chromogranin at the plasma membrane. The PDM value was calculated in each location of the image. The analysis was performed on the entire islet for chromogranin and insulin and on the region shown at higher magnification for glucagon and somatostatin. Results are presented in pseudocolored images. A PDM scale bar is inserted. A positive PDM value is indicative of dependent staining (co-localization), and a negative value is indicative of segregated staining. Bar, 10 m.
Human ␤-Cells Are Also Vulnerable to Glutamate-induced Toxicity-We next explored whether altering islet glutamate homeostasis by chronic incubation with glutamate or inhibi-tion of GLT1 could affect human islet cell viability. When measured by a quantitative apoptosis assay, a 3-day exposure to glutamate induced a dose-dependent increase in human islet apoptosis that was statistically significant at 5 mM glutamate (Fig. 6A) and quantitatively similar to that observed at high glucose concentrations (16.7 mM). A similar dose-dependent increase in apoptosis was detected in the presence of DHK and was potentiated by co-application of glutamate and DHK (Fig.  6A). These data indicate that GLT1, by regulating the extracellular glutamate concentration, participates in the control of cell survival also in human islets. Apoptosis was restricted to ␤-cells as shown by double staining for insulin and TUNEL in dispersed islet cells (Fig. 6B). Data were confirmed by a quantitative electron microscopy analysis performed on human islets (Fig. 6, C and D, and supplemental Fig. S4). In the absence of glutamate, the majority of ␣-(97 Ϯ 3%) and ␤-cells (78 Ϯ 1%) lacked nuclear and cytoplasmic degenerations (Fig. 6C, panels  a-c). Conversely, 75% of ␤-cells of islets exposed to 5 mM glutamate showed severe degenerative features including condensed apoptotic nuclei and numerous cytoplasmic vacuoles, some of which contained dark bodies (Fig. 6C, panels d and e, and supplemental Fig. S4A). Interestingly, ␣-cells of glutamateexposed islets were well preserved (Fig. 6C, panel f, and supplemental Fig. S4B).
To further support a role of glutamate in the selective ␤-cell dysfunction, basal (3.3 mM glucose) and stimulated (16.7 mM glucose) insulin and proinsulin secretions were measured after 3 days of chronic incubation with glutamate, DHK, or both (Fig.  7). We found that chronic exposure to glutamate induced a dose-dependent increase in insulin secretion under basal conditions (Fig. 7A) that was paradoxically higher than that under high glucose stimulation. Chronic exposure to glutamate also induced an increase in the proinsulin to insulin ratio both under basal and stimulated conditions (Fig. 7B). All these secretory defects (increased basal insulin secretion, reduced stimulated insulin release, and increased proinsulin to insulin ratio) have already been observed in "stressed" human islets after chronic exposure to high glucose (27,37,38). Effects similar to those observed after chronic glutamate exposure were also detected in the presence of DHK alone or supplemented with 0.5 mM glutamate.
Glutamate Induces Increased Oxidative Stress in Human Islets-Finally, we explored the mechanisms involved in glutamate toxicity in human islets. CNQX co-application did not prevent ␤-cell apoptosis induced by chronic incubation with glutamate as measured by both ELISA and electron microscopy (Fig. 8, A and B). However, differently from ␤TC3 cells and consistent with previous reports (14,15), the majority of glucose-responsive insulin-positive cells dispersed from isolated human islets did not respond to glutamate with an increase in intracellular Ca 2 (supplemental Fig. S5A). Only a small fraction of insulin-positive cells (15 Ϯ 5%; n ϭ 10 of 65 insulin-positive cells) exhibited an increase in intracellular Ca 2ϩ after incubation with glutamate (supplemental Fig. S5B).
Oxidative stress is an alternative mechanism possibly involved in glutamate-induced human ␤-cell toxicity. In support of this hypothesis, we found that human islets expressed the xCT subunit of the x Ϫc exchanger (Fig. 8C). As a marker of . C, control. B, immunoprecipitation (IP) of total human islet extracts with a rabbit serum (IgG) or the GLT1 antibody. 50 g of brain and 100 g of human islet extracts (Lys) were loaded in the same gel. Markers on the right indicate kDa. **, oligomer; *, monomer; E, nonspecific band. C, immunolocalization of GLT1 (green) and insulin (red) on dispersed human ␤-cells seeded on glass coverslips. Bar, 10 m. D, D-[ 3 H]aspartic acid uptake measurements in batches of 40 human islets. 0.3 mM DHK was added to the uptake solution. Data are expressed as percentage of the Na ϩ -dependent uptake (NaCl) (three different islet preparations in triplicate). **, p Ͻ 0.01. E, acute glutamate (Glu) and DHK application modulates the stimulated glucagon secretion. Human islets were preconditioned in 3.3 or 16.7 mM glucose for 1 h (Preconditioning) and then stimulated with 3.3 or 16.7 mM glucose for 1 h (Stimulus) in the presence of 0.5 mM glutamate, 0.1 mM DHK, or both glutamate and DHK as specified. The glucagon released over the 1-h stimulus was measured. Data were normalized to glucagon content in the islet and are expressed as -fold increase over 3.3-3.3 mM glucose (islets were kept in 3.3 mM glucose and then transferred to a solution containing 3.3 mM glucose) (three different islet preparations in triplicate). **, p Ͻ 0.01 versus 3.3-3.3 mM glucose; #, p Ͻ 0.05 versus 16.7-3.3 mM glucose, 0 mM glutamate, and 0 mM DHK. F, acute glutamate and DHK application does not affect the stimulated insulin release. Human islets were first preconditioned in 3.3 or 16.7 mM glucose (Preconditioning) and then stimulated with 3.3 or 16.7 mM glucose for 1 h (Stimulus) in the presence of 0.5 mM glutamate, 0.1 mM DHK, or both glutamate and DHK as specified. The insulin released over the 1-h stimulus was measured. Data were normalized to insulin content, and data are expressed as -fold increase over 3.3-3.3 mM glucose (three different islet preparations in triplicate). **, p Ͻ 0.01 versus 3.3-3.3 mM glucose. ChCl, choline chloride. The error bars indicate standard errors.

GLT1/EAAT2 Prevents Glutamate-induced ␤-Cell Death
increased oxidative stress we measured the amount of protein adducts of 4-HNE (Fig. 8D) (39). Compared with control samples, the culture in 5 mM glutamate for 3 days showed an increase up to 1.8 Ϯ 0.2-fold of the total amount of adducts of 4-HNE with discernible new bands of about 50 and 80 kDa (Fig.  8D, arrowhead). A similar increase in adducts of 4-HNE was detected after a 3-day culture in 16.7 mM glucose, a recognized condition of oxidative stress in ␤-cells. Glutamate-induced oxi-

DISCUSSION
The main finding of this work is that pancreatic ␤-cells express the glutamate transporter GLT1 whose function is to control the extracellular glutamate concentration and to preserve ␤-cell integrity. The precise concentration of extracellular glutamate in the islet microenvironment is presently unknown, but it is likely remarkably high. The blood glutamate concentration is relatively high (50 -100 M) as compared with the CNS and can increase up to ϳ500 M under a glutamateenriched diet (40). Moreover, islets are highly vascularized structures (41), and glutamate may reach very high concentrations in islet capillaries especially after a meal. In addition, glu-tamate is a major secretory product of the ␣-cell and is physiologically co-secreted with glucagon (8). Extracellular glutamate concentrations may increase further in the islets of human and non-human primates with T2DM, which are characterized by   an inverted ␤-cell/␣-cell ratio and glucagon hypersecretion (30,42,43) and also during "insulitis" in T1DM islets because dendritic and T cells can release glutamate (44,45). Of note, changes in glutamate concentration in the islet have been reported in response to high glucose (46), and increased glutamate levels have been found in sera of subjects with both T1DM (47) and T2DM (48).
We reasoned that if ␤-cells are exposed to high extracellular glutamate levels they might be vulnerable to glutamate toxicity unless they express a glutamate clearance system. We demonstrate that clonal ␤-cells, but not ␣-cells, are indeed susceptible to glutamate cytotoxicity (Fig. 1). Isolated human islets are also vulnerable to glutamate-induced toxicity (Fig. 6), which is restricted to ␤-cells as shown by TUNEL assay and electron microscopy analysis. The ultrastructural features observed in human ␤-cells exposed to glutamate are characterized by cytoplasmic vacuolization and nuclear apoptosis and are reminiscent of the degenerative features observed in motor neurons undergoing autophagic degeneration due to glutamate toxicity (49). Interestingly, autophagy has been recently described as an additional mechanism of ␤-cell death in human T2DM (50), thus supporting a role of glutamate in diabetes pathogenesis.
Excessive activation of AMPA/kainate glutamate receptors contributes to cell death in clonal ␤-cells but not in primary human ␤-cells as co-administration of CNQX could not prevent/reduce glutamate-induced toxicity in these cells. Accordingly, we found that the majority of human ␤-cells do not express functional glutamate receptors as also reported by others (14,15). Our data suggest that glutamate-induced oxidative stress is probably the prevalent mechanism of ␤-cell death in human islets. Originally described in astrocytes, excess extra-cellular glutamate reverts the activity of the glutamate/cysteine antiporter system x Ϫc , thus depleting the cells of cysteine, a building block of the antioxidant glutathione (51). We show that both clonal and human ␤-cells express the x Ϫc exchanger (Figs. 1E and 8C), and they are particularly vulnerable to oxidative stress (52). Accordingly, we found a reduced free SH content in ␤TC3 cells exposed to elevated glutamate or DHK concentrations ( Fig. 1E and data not shown) and an increased amount of adducts of 4-HNE in human ␤-cells (Fig. 8, D and E). 4-HNE is an irreversible protein modification that alters the conformation and function of proteins involved in important cellular pathways and is a recognized marker of oxidative stress (39).
Regardless of the mechanisms involved in glutamate toxicity both in neurons and in ␤-cells, glutamate-induced toxicity is physiologically prevented by GLT1 activity. By using molecular, biochemical, pharmacological, and physiological approaches, we showed that GLT1 is selectively expressed by the ␤-cell, whereas it is virtually absent in all the other islet cell types. Immunohistochemistry and confocal microscopy of human pancreatic sections showed that GLT1 is prevalently localized to the plasma membrane. None of the other known high affinity glutamate transporter subtypes (EAAT1/GLAST and EAAT3/ EAAC1) were detected in human islets (data not shown). Our observation that GLT1 was selectively expressed in ␤-cells contrasts with previous data showing a Na ϩ -dependent glutamate/ aspartate transport activity confined to the non-␤ islet cell mantle (25). This discrepancy may be explained by considering that the localization of the transporter substrate (glutamate) and not the transporter itself was studied in the latter report. In agreement with a lack of GLT1 expression, we did not detect FIGURE 9. Proposed events leading to progressive ␤-cell death by glutamate toxicity in diabetes mellitus. The extracellular glutamate concentration in the islet of Langerhans is controlled by the activity of GLT1. In diabetes mellitus, a combination of insults may trigger ␤-cell death and consequently modify the extracellular glutamate concentration by increasing its release (via ␣-cells) and decreasing its clearance (via ␤-cells). Increased glutamate levels may cause ␤-cell death, thus triggering a vicious cycle that further increases the glutamate levels in the islet of Langerhans and ␤-cell death.
measurable Na ϩ -dependent glutamate uptake in ␣TC1 cells or the presence of any other high affinity glutamate transporter. Our findings and recent reports indicate that ␣-cells express a functional glutamine uptake system (ASCT2 and SAT2) (53) and a phosphate-dependent glutaminase activity; thus, consequently intracellular glutamate in the ␣-cell may be prevalently derived from glutamine transamination (54).
In this study, we also demonstrated that GLT1 is the main regulator of the extracellular glutamate clearance in the islet and that its normal function is critical for glucagon release and ␤-cell survival. Indeed, GLT1 down-regulation by pharmacological blockade decreased glucagon release and induced ␤-cell death. The most important function of GLT1 in the CNS is to regulate the extracellular glutamate concentration, thus preserving cell signaling and preventing glutamate-mediated toxicity (22,23). Accordingly, our data show that the pharmacological blockade of GLT1 by the selective inhibitor DHK induced an increase in extracellular glutamate concentration to a degree that is toxic to ␤-cells. Vice versa, GLT1 up-regulation obtained with CEF, which is known to increase GLT1 expression in the CNS, thereby providing neuroprotection against glutamate toxicity, also reduced glutamate-induced ␤-cell death (Fig. 3). These data suggest that CEF and other compounds capable to increase GLT1 expression and/or activity may represent novel therapeutic tools to achieve ␤-cell cytoprotection.
In conclusion, glutamate homeostasis is critical to preserve islet cell signaling and ␤-cell integrity, and GLT1, by controlling extracellular glutamate levels, may play an important role in these processes. Excessive extracellular glutamate and abnormal GLT1 function may contribute to the pathogenesis of ␤-cell loss in both T1DM and T2DM. Early on, in the natural history of both these diseases, the islet ␤to ␣-cell ratio changes in favor of the ␣-cell (55)(56)(57), and this may perturb islet glutamate homeostasis as glutamate released by the redundant ␣-cells cannot be efficiently cleared by the neighboring ␤-cells. In turn, increased extracellular glutamate may induce ␤-cell death, thus triggering a vicious cycle that further increases extracellular glutamate and ␤-cell death (Fig. 9).