Excitotoxic neuronal cell death during an oligodendrocyte-directed CD8+ T cell attack in the CNS gray matter

Background Neural-antigen reactive cytotoxic CD8+ T cells contribute to neuronal dysfunction and degeneration in a variety of inflammatory CNS disorders. Facing excess numbers of target cells, CNS-invading CD8+ T cells cause neuronal cell death either via confined release of cytotoxic effector molecules towards neurons, or via spillover of cytotoxic effector molecules from 'leaky’ immunological synapses and non-confined release by CD8+ T cells themselves during serial and simultaneous killing of oligodendrocytes or astrocytes. Methods Wild-type and T cell receptor transgenic CD8+ T cells were stimulated in vitro, their activation status was assessed by flow cytometry, and supernatant glutamate levels were determined using an enzymatic assay. Expression regulation of molecules involved in vesicular glutamate release was examined by quantitative real-time PCR, and mechanisms of non-vesicular glutamate release were studied by pharmacological blocking experiments. The impact of CD8+ T cell-mediated glutamate liberation on neuronal viability was studied in acute brain slice preparations. Results Following T cell receptor stimulation, CD8+ T cells acquire the molecular repertoire for vesicular glutamate release: (i) they upregulate expression of glutaminase required to generate glutamate via deamination of glutamine and (ii) they upregulate expression of vesicular proton-ATPase and vesicular glutamate transporters required for filling of vesicles with glutamate. Subsequently, CD8+ T cells release glutamate in a strictly stimulus-dependent manner. Upon repetitive T cell receptor stimulation, CD25high CD8+ T effector cells exhibit higher estimated single cell glutamate release rates than CD25low CD8+ T memory cells. Moreover, glutamate liberation by oligodendrocyte-reactive CD25high CD8+ T effector cells is capable of eliciting collateral excitotoxic cell death of neurons (despite glutamate re-uptake by glia cells and neurons) in intact CNS gray matter. Conclusion Glutamate release may represent a crucial effector pathway of neural-antigen reactive CD8+ T cells, contributing to excitotoxicity in CNS inflammation.


Introduction
Cytotoxic CD8 + T cells directed against neuronal or glial antigens contribute to neuronal cell death in a variety of inflammatory CNS disorders [1][2][3]. They are capable of harming neuronal function and integrity either by directly targeting neurons and their neurites or by exerting their action via 'collateral' mechanisms of neuronal damage that might follow destruction of oligodendrocytes (ODCs), myelin sheaths or astrocytes in the CNS gray and white matter [4].
Recently, we have shown that activated T cell receptor (TCR)-transgenic, ovalbumin (OVA)-reactive CD8 + T cells (OT-I cells [5]) are capable of initiating 'collateral' neuronal apoptosis in the gray matter of otherwise non-inflamed acute brain slices from ODC-OVA mice selectively expressing OVA in ODCs under the control of a truncated myelin basic protein (MBP)-promotor [6]. 'Collateral' killing of neurons by OT-I cells was in part due to release of perforin (and granzymes) whereas Fas-ligand (FasL) -Fas-receptor (Fas) interactions were not relevant [7][8][9]. However, a certain part of the collateral neuronal cell death induced by CD8 + T cells remained unexplained [7,10].
During culture of polyclonally activated CD3 + T cells, high levels of extracellular glutamate have been shown to accumulate, which are efficiently cleared if T cells are co-cultured with astrocytes [11]. In turn, the T cell-derived glutamate elicits the release of thiols (cysteine, glutathione and cysteinyl-glycine) and lactate from astrocytes, which reduce neuronal apoptosis induced by oxidative stress [11]. However, whether glutamate derived from T cells migrating through the intact CNS parenchyma and scanning for their cognate antigen may also exert excitotoxic neuronal cell death, is unclear at present.

Mice
Wild-type (WT), ODC-OVA [6] and OT-I [5] mice were kept at the animal facility of the University of Münster, Germany, and had access to food and water ad libitum. All experiments were conducted according to the German law of animal protection and were approved by local authorities. All transgenic mice were created on or backcrossed to a C57BL/6 genetic background for at least ten generations.
For in vitro T cell stimulation experiments, CD8 + T cells were isolated from WT splenocytes using the respective MACS® T cell isolation kit (Miltenyi, Bergisch Gladbach, Germany) according to the manufacturer's instructions. This yielded a purity of about 90% of CD3 + CD8 + T cells (Additional file 1: Figure S1A, B). CD8 + T cells from WT mice were cultured for different periods of time at a density of 1 × 10 6 /well on 24-well plates with anti-CD3/CD28 beads (Dynal Biotech, Oslo, Norway) at different bead-to-cell ratios in the absence or presence of certain glutamate release blockers (all from Sigma-Aldrich, München, Germany) or were left unstimulated. In a subset of experiments, CD8 + T cells from WT mice were stimulated in glutamine-free DMEM-based medium (BioWhittaker, Verviers, Belgium). After three days of stimulation, supernatants were removed and analyzed for IFNγ using a mouse-IFNγ-ELISA kit according to the manufacturer's protocol (Duoset, R&D Systems, Wiesbaden, Germany).
For in vitro T cell stimulation experiments, splenocytes from OT-I mice were plated at a density of 1 × 10 7 /well on 12-well plates and primed by incubation for five days with OVA-peptide 257-264 (SIINFEKL; 1 nM) and IL-2 (500 U/ml). On day three, supernatants were removed, analyzed for IFNγ using a mouse-IFNγ-ELISA kit according to the manufacturer's protocol (Duoset, R&D Systems, Wiesbaden, Germany) and substituted by fresh medium. On day four, again 500 U/ml of IL-2 were added to the medium. In a subset of experiments, OT-I T cells were repetitively stimulated by incubation with OVApeptide 257-264 (SIINFEKL; 1 nM) and IL-2 (500 U/ml) every three days for a total of fifteen days.

Flow cytometry
Before and after various time periods of in vitro culture, flow cytometry of stimulated WT CD8 + T cells and OT-I cells was performed using standard methods. For analysis of T cell subtype distribution, cells were stained for 20 minutes with allophycocyanin (APC)-labeled anti-mouse CD3, phycoerythrin (PE)-labeled anti-mouse CD8 and fluorescein isothiocyanate (FITC)-labeled anti-mouse CD4 (all BD Bioscience, Heidelberg, Germany). As isotype controls, cells were stained with APC-labeled anti-mouse IgG1, PE-labeled anti-mouse IgG1 or FITC-labeled antimouse IgG1 (all by BD Bioscience, Heidelberg, Germany). For analysis of T cell activation markers, cells were stained for 20 minutes with FITC-labeled anti-mouse CD25 or FITC-labeled anti-mouse CD69 (all BD Bioscience, Heidelberg, Germany). For analysis of T cell viability, cells were stained for 20 minutes with FITC-labeled anti-mouse Annexin V (BD Bioscience, Heidelberg, Germany) and propidium iodide (PI). All antibodies were titrated for optimal staining. Flow cytometry analysis was performed using a FACSCalibur® system (BD Biosciences, Heidelberg, Germany) and results were analyzed using CellQuest Pro Software (BD Bioscience, Heidelberg, Germany).

Determination of supernatant glutamate levels
Supernatant glutamate levels were determined at different time points of in vitro stimulation of WT CD8 + T cells and OT-I cells using a commercially available enzymatic determination kit (Glutamine-Glutamate Kit, Sigma-Aldrich, München, Germany) following manufacturer's instructions. Background glutamate levels were determined in the absence of cells and subtracted from those obtained during incubation and stimulation of cells. All pharmacological blockers used did not interfere with measurements of fixed concentrations of glutamate.
For estimation of single cell glutamate secretion rates, splenocytes from OT-I mice were plated at a density of 1 × 10 7 /well on 12-well plates and iteratively stimulated by incubation for time intervals of 72 hours with OVApeptide 257-264 (SIINFEKL; 1 nM) and IL-2 (500 IU/ml) for a total of 15 days. Numbers (N) of CD8 + T cells were determined by resuspending and counting total cell numbers/well at the beginning (b) and the end (e) of each interval and multiplying each by the percentage of CD3 + CD8 + T cells as determined by standard flow cytometry. Cumulative glutamate concentrations ([Glu]) were determined at the end (e) of each time interval as described. Volumes (V) of the culture medium were determined at the beginning (b) and the end (e) of the individual time intervals. Single cell glutamate secretion rates (ΔGlu/Δt/CD8 T cell) during repetitive stimulation were calculated using the following formula:

Quantitative real-time PCR
For analysis of glutaminase-, vesicular (V-type) proton ATPase-, and vesicular glutamate transporter (VGluT) 1-3-mRNA expression regulation, RNA was purified using Trizol reagent (Invitrogen, Carlsbad, CA, USA) and complementary DNA (cDNA) synthesis was performed using a standard protocol with random hexamer primers (all reagents were purchased from Applied Biosystems, Foster City, CA, USA) from WT CD8 + T cells before and after three days of anti-CD3/CD28 bead-stimulation. cDNA was used in quantitative real-time PCR (qRT-PCR) assays with specific TaqMan® primers and probes for murine glutaminase (Mm01257297_m1), V-type proton ATPase (Mm01222963_m1), VGluT1 (Mm00812886_m1), VGluT2 (Mm00499876_m1) and VGluT3 (Mm008054 13_m1, all Applied Biosystems, Foster City, CA, USA) as well as primers for 18sRNA (Hs_4319413E; Applied Biosystems, Foster City, CA, USA) as endogenous control. qRT-PCR was performed according to the manufacturer's protocol.
Neuronal cell culture and glutamate excitotoxicity assays Neuronal cell cultures were obtained from WT C57BL/6 mice embryos (E18) following previously described protocols [12]. Pregnant mice were killed by cervical dislocation and embryos were removed and transferred into warmed Hank's balanced salt solution (HBSS, Gibco, Darmstadt, Germany). After preparation of hippocampi, tissue was collected in 5 ml of 0.25% trypsin in HBSS. After five minutes of incubation at 37°C, tissue was washed twice with HBSS and dissociated in 1 ml of neuronal medium (10% 10× modified Earl's medium (MEM), 0.2205% sodium bicarbonate, 1 mM sodium pyruvate, 2 mM L-glutamine, 2% B27 supplement (all Gibco, Darmstadt, Germany), 3.8 mM glucose (Merck, Darmstadt, Germany), 1% penicillin/streptomycin (Biochrom AG, Berlin, Germany)) by triturating with fire polished Pasteur pipettes of decreasing tip diameter. Neurons were diluted in neuronal medium and plated at a density of 60,000 cells/cm 2 on poly-D-lysine (Sigma-Aldrich, München, Germany) coated cover-slips placed in fourwell plates (Nunc, Roskilde, Denmark). All neuronal cell cultures were maintained at 37°C and 5% CO 2 for up to five to seven days prior to experiments. The neuronal cell culture system exhibited a purity of about 80% [12].
Analysis of excitotoxic neuronal cell damage was performed via two distinct experimental approaches: (i) cultured neurons were incubated for 30 minutes with standard artificial cerebrospinal fluid (ACSF, for composition see below) supplemented with different concentrations of L-glutamate (Sigma-Aldrich, München, Germany) followed by a washout for 30 minutes with standard ACSF. The proportion of neurons showing focal swellings (beading [13]) was calculated as a percentage of all MAP2 + cells; (ii) cultured neurons were incubated for six hours with standard ACSF (for composition see below) supplemented with or without L-glutamate (Sigma-Aldrich, München, Germany) at a concentration of 100 μM. The proportion of neurons staining positive for activated caspase-3 was calculated as a percentage of all NeuN + cells.

Immunofluorescence-staining and -microscopy
Immunohistochemical staining was performed on 10 μm coronary sections. For double labeling, slices were postfixated in 4% PFA for ten minutes and afterwards incubated in blocking solution (PBS containing 5% BSA, 1% normal goat serum and 0.2% Triton X-100). Slices were then incubated simultaneously with antibodies to NeuN (1:1,000, Chemicon, Billerica, MA, USA) and activated caspase-3 (1:400, Cell Signaling, Boston, MA, USA) overnight at 4°C. Secondary antibodies were Alexa Fluor 488-coupled goat anti-mouse IgG and Cy3-coupled goat anti-rabbit IgG. Negative controls were obtained by either omitting the primary or secondary antibody and gave no signal (data not shown). For quantification of cell densities, sections were examined with an Axiophot2 microscope (Zeiss, Oberkochen, Germany) with a CCD camera (Visitron Systems, Tuchheim, Germany). Cell densities were determined in preselected fields within the neocortex.

Statistical analysis
All results are presented as mean ± standard error of the mean (SEM). Statistical analysis was performed using student's t-test for normally distributed samples or Mann-Whitney test for not-normally distributed datasets [16]. A Bonferroni-corrected ANOVA was used in case of multiple comparisons. Analysis was performed using the Sigma Plot 11.0® software (Systat Software Inc., Erkrath, Germany). P-values < 0.05 were considered statistically significant (indicated as *) and P-values < 0.01 were considered highly statistically significant (indicated as **).

CD8 + T cells release glutamate by a TCR-stimulus-dependent mechanism
In a first set of experiments, we isolated naive CD8 + T cells from splenocytes of WT C57BL/6 mice, stimulated them polyclonally with and without anti-CD3/CD28-coated beads at a bead-to-cell ratio of 1:1 in medium containing 2 mM glutamine and determined supernatant glutamate concentrations at distinct time points. Supernatant glutamate concentrations increased exponentially with time and reached levels of about 0.5 mM after 72 hours of bead-stimulation (n = 3 mice, experiments performed in triplicates; Figure 1A). In contrast, in the absence of bead-stimulation, supernatant glutamate concentrations after 72 hours were significantly lower than in its presence (relative supernatant glutamate: bead-stimulation for 72 hours: 1.0 ± 0.01, no bead-stimulation for 72 hours: 0.17 ± 0.04, P = 0.002, n = 3 mice, experiments performed in triplicates; Figure 1C). Moreover, supernatant glutamate concentrations determined after 72 hours of stimulation clearly depended on the bead-to-cell ratio as a measure of stimulation intensity ( Figure 1B).
As glutamate has important roles in the intermediary metabolism, it is present at millimolar concentrations in the cytosol of most cell types [17]. Cytosolic glutamate may stem from transamination of α-ketoglutarate by glutamate dehydrogenase or from deamination of glutamine by glutaminase [17]. Consistent with predominant synthesis of intracellular glutamate from deamination of glutamine, supernatant glutamate levels were significantly reduced following stimulation of CD8 + T cells for 72 hours in glutamine-free medium (relative supernatant glutamate: bead-stimulation for 72 hours in glutamine-containing medium at a bead-to-cell ratio of 1:1: 1.0 ± 0.01, beadstimulation for 72 hours in glutamine-free medium at a bead-to-cell ratio of 1:1: 0.20 ± 0.05, P = 0.003, n = 3 mice, experiments performed in triplicates; Figure 1C).
As we were unable to unambiguously identify a non-vesicular mechanism of glutamate release from stimulated CD8 + T cells, we studied the expression regulation of genes essential for vesicular glutamate release during anti-CD3/CD28-bead stimulation of WT CD8 + T cells ( Figure 1D). Quantitative real-time qRT-PCR analysis revealed an up-regulation of expression levels of glutaminase (3.3 ± 0.9 fold) responsible for the cytosolic conversion of glutamate from glutamine [17]. Moreover, vesicular proton ATPase (3.6 ± 0.1 fold), required for vesicle acidification and vesicular glutamate transporters (VGluT 1: 77.0 ± 12.0 fold; VGluT2: 12.6 ± 7.2 fold; VGluT3: 5.2 ± 1.2 fold; n = 3 mice, all experiments performed in triplicates) required for uptake of cytosolic glutamate into vesicles [21] showed significant expression regulation ( Figure 1D). These findings point towards vesicular release as the predominant mechanism of glutamate liberation by polyclonally activated CD 8 + T cells.
During repetitive antigen-stimulation, CD8 + T cells with strong IL-2 receptor α (CD25) expression represent a population of rather short-lived effector cells, whereas those with low CD25 expression form a population of rather long-lived memory cells [22]. To examine an effect of the CD8 + T cell activation status on the rate of glutamate release, splenocytes from OT-I mice were repetitively stimulated by incubation for time intervals of 72 hours with OVA-peptide 257-264 (SIINFEKL; 1 nM) and IL-2 (500 IU/ml) for a total of 15 days. Single CD8 + T cell glutamate secretion rates were determined from the supernatant glutamate concentrations and cell numbers at the end of each stimulation interval (see Methods; Figure 2D). In parallel, flow cytometry was repetitively performed for activation markers CD69 and CD25 ( Figure 2C). CD69 expression levels peaked after 12 hours of stimulation and went back to baseline levels afterwards (CD69: mean fluorescence intensity (MFI): day 0: 5.6 ± 4.2, day 0.5: 20.6 ± 3.2, P = 0.015). In contrast, both CD25 expression levels and single CD8 + T cell glutamate secretion rates increased during the two stimulation intervals, peaked at day 6, and subsequently decreased back towards baseline levels (CD25: mean fluorescence intensity (MFI): day 0: 2.5 ± 1.2, day 6: 43.3 ± 12.5, P = 0.002; glutamate secretion rate: day 3: 3.8 ± 1.8 fmol/h/ cell, day 6: 41.1 ± 16.0 fmol/h/cell, P = 0.034; Figure 2C,D) suggesting glutamate liberation as a possible effector mechanism of antigen-activated CD8 + T cells.
Antigen-dependent glutamate release from CD8 + T cells causes excitotoxic neuronal cell death during an oligodendrocyte-directed CD8 + T cell attack in the CNS gray matter Glutamate release from activated CD8 + T cell might represent an effector mechanism in autoimmune CNS inflammation due to the ubiquitous expression of synaptic and extrasynaptic ionotropic glutamate receptors in neurons [23], which renders them highly sensitive to excitotoxic cell death. Moreover, encephalitogenic CD8 + T cells are known to migrate through the CNS parenchyma making short-lasting direct contacts with parenchymal cells in search of their cognate antigen. Upon antigen-recognition, CD8 + T cells make long-lasting contacts with their target cells and exert direct and collateral [4,7,12] cytotoxic effects. In the intact CNS parenchyma, glutamate is efficiently cleared from the extracellular space through (re-)uptake by high-affinity glial and neuronal excitatory amino acid transporters (EAATs; [24]). Thus, relevant excitotoxic effects of CD8 + T cell-mediated glutamate liberation can only be studied in the intact CNS parenchyma.
First, we tested the impact of MK801 and NBQX at the respective concentrations on the killing behavior of activated OT-I T cells using glutamate-insensitive murine EL-4 lymphoma cells [30] loaded with OVA-peptide 257-264 (SIINFEKL; 1 nM) as target cells (effector-to-target cell ratio 1:1; [10]; Figure 5). The amount of ATP in the supernatant following cell lysis was assessed as a parameter of target cell viability. As a control, supernatant ATP levels were determined from 0, 50 000 and 100 000 EL-4 cells and activated OT-I T cells cultured separately for six hours (n = 3 mice, experiments performed in triplicates). Under all conditions, supernatant ATP levels showed a linear dependence on the number of cultured cells ( Figure 5A    impact of CD8 + T cell-mediated glutamate liberation on excitotoxic neuronal cell death within the intact CNS parenchyma. To this end, we transferred activated OT-I T cells [5] into acute brain slices from mice selectively expressing OVA in oligodendrocytes (ODC-OVA mice [6]) and tracked collateral neuronal cell death in the cortical gray matter via double-staining for NeuN and activated caspase-3 [7,10]. After six hours exposure to activated OT-I T cells, the fraction of activated caspase-3 + neurons was significantly increased compared to control conditions without OT-I T cell incubation (OT-I T cells: 32.1 ± 4.0%; control: 5.2 ± 0.8%, P = 0.008). Moreover, glutamate receptor blockade with NBQX and MK801 significantly reduced neuronal cell death (OT-I T cells: 32.1 ± 4.0%; OT-I T cells + MK801/NBQX: 13.6 ± 1.9%, P < 0.001). Importantly, incubation with NBQX and MK801 did not reduce background levels of neuronal cell death in the absence of incubation with OT-I T cells (MK801/NBQX: 6.4 ± 0.9%, P = 0.236) ruling out unspecific inhibition of excitotoxicity in the absence of T cells. Furthermore, incubation for six hours with L-glutamate at a concentration of 100 μM significantly increased the fraction of activated caspase-3 + neurons compared to control conditions (glutamate: 11.2 ± 1.8%, P = 0.041, n = 3 mice, each providing three brain slices for all experimental groups; Figure 6) corroborating caspase-3 activation in neurons undergoing excitotoxic cell death within the intact CNS parenchyma. Figure 6 Antigen-dependent glutamate release from CD8 + T cells causes collateral excitotoxic neuronal cell death during an oligodendrocyte-directed CD8 + T cell attack in the CNS gray matter. (A-F) Activated OT-I T cells were incubated for six hours [5] in acute brain slices from ODC-OVA mice in the absence (A, OT-I) or presence (B, OT-I NBQX MK801) of MK801 (10 μM) and NBQX (30 μM), and collateral neuronal cell death (white arrows) in the cortical gray matter was tracked via staining for NeuN (green), DAPI (blue) and activated caspase-3 (red). As controls, acute brain slices from ODC-OVA mice were left untreated (C, control), or incubated either with L-glutamate (D, Glu, 300 μM) or with MK801 and NBQX (E, NBQX MK801, 10 μM and 30 μM) for six hours. (F) Fractions of activated caspase-3 + NeuN + neurons detected in the cortical gray matter of acute brain slices from ODC-OVA mice under the distinct experimental conditions (n = 3 mice, each providing three brain slices for all experimental groups).

Discussion
CD8 + T cell-mediated neuronal cell death is a relevant pathogenic mechanism in a variety of inflammatory CNS disorders [4]. CD8 + T cell cytotoxicity within the CNS is predominantly mediated via two independent pathways depending on the strength of the antigen-signal (that is, the number of antigen-peptide loaded major histocompatibility complex class I (pMHC I) molecules and the affinity of the CD8 + T cell receptor (TCR) complex to the pMHC I complex) delivered by the glial or neuronal target cell [31,32]. Strong antigen-signals promote killing via CD8 + T cell-mediated perforin and granzyme exocytosis and perforin-dependent delivery of granzymes into the target cell, whereas weak antigen-signals favor killing via interaction of CD8 + T cell-derived Fas-ligand (FasL) with the Fas-receptor (Fas) on the target cell surface [31][32][33]. Moreover, CD8 + T cells are able to contribute to inflammatory tissue damage by release of proinflammatory cytokines such as IFNγ and tumor necrosis factor-(TNF)α [34]. These CD8 + T cell effector mechanisms impair electrical signaling [12] and cause cell death of neurons as direct and 'collateral' target cells [4,7].
Glutamate excitotoxicity is known to contribute to neuronal damage in autoimmune CNS inflammation [35][36][37]. Under inflammatory conditions, glutamate is known to be released from microglia and astrocytes [13,20].
We provide evidence of glutamate release as another crucial effector pathway of CD8 + T cells leading to 'collateral' excitotoxic neuronal degeneration during an oligodendrocyte-directed CD8 + T cell attack in the CNS gray matter. Following TCR stimulation, CD8 + T cells acquire the molecular repertoire for vesicular glutamate release: (i) they upregulate expression of glutaminase required to generate glutamate via deamination of glutamine and (ii) they upregulate expression of vesicular proton-ATPase and vesicular glutamate transporters required for filling of vesicles with glutamate. In contrast, there was no evidence for nonvesicular mechanisms of glutamate liberation. Furthermore, CD25 high CD8 + T effector cells exhibit higher estimated single cell glutamate release rates than CD25 low CD8 + T memory cells underpinning glutamate liberation as a possible effector mechanism of antigen-activated CD8 + T cells entering the CNS. Moreover, activated oligodendrocyte-reactive CD25 high CD8 + T effector cells cause 'collateral' neuronal cell death within intact cortical gray matter of acute brain slices which can partially be blocked by a combination of ionotropic glutamate receptor antagonists. Hence, glutamate liberation by oligodendrocyte-reactive CD8 + T cells is capable of eliciting excitotoxic cell death of neurons despite glutamate reuptake by glia and neurons in the intact CNS gray matter [24]. Encephalitogenic CD8 + T cells (in contrast to CD4 + T cells) make close contacts with their target cells in the CNS [12] and release proinflammatory cytokines which may reinforce local glutamate excitotoxicity: TNFα is known to reduce the expression of EAATs and detoxifying enzymes in glia cells limiting their glutamate uptake capacity [38,39]. IFNγ has been shown to exert neurotoxicity by gating a calcium-permeable complex of the IFNγ receptor and AMPA glutamate receptor in neurons [40]. Moreover, neuron-reactive CD8 + T cells which directly contact neurites induce cytoskeleton breaks with adjacent neuritic spheroids [41] closely resembling neuronal beading upon exposition to glutamate [13]. Importantly, it has been demonstrated that cell death in mature but not immature cultured neurons treated with cytotoxic secretory granules isolated from CD8 + T cells can be inhibited by blockers of ionotropic (NMDA-but not AMPA-) but not metabotropic glutamate receptors [30]. As in our case, this effect was neuron-specific as glutamate receptor blocker failed to protect glutamate-insensitive lymphoma cells from secretory granule-induced cytotoxicity [30]. Moreover, blocking of neuronal glutamate reuptake did not significantly enhance killing of neurons elicited by secretory granules, and relevant glutamate concentrations could not be detected in the culture supernatant. This refutes the possibility that significant neuronal (auto-)toxicity was induced by vesicular or nonvesicular release of glutamate from neurons following their perforin-mediated depolarization [12] upon treatment with secretory granules [30]. However, these findings are consistent with storage of glutamate in secretory granules which together with perforin and granzymes elicits neuronal calcium overload and cell death.

Conclusions
Facing excess numbers of target cells, CNS-invading CD8 + T cells may kill neurons either via confined release of cytotoxic effector molecules towards the neuron or via spillover of cytotoxic effector molecules from 'leaky' immunological synapses and non-confined release by CD8 + T cells themselves during serial and simultaneous killing of oligodendrocytes or astrocytes. In both scenarios glutamate-mediated excitotoxicity may contribute to neuronal cell death.

Additional file
Additional file 1: Figure S1. (A, B) WT CD8 + T cells were isolated from splenocytes (see Methods) yielding a purity of about 90% of CD8 + cells (A) of which about 90% were CD3 + T cells (B) as revealed by standard flow cytometry analysis. (C) The presence of a variety of glutamate release blockers (see text) did not affect the activation status of CD8 + T cells after 72 hours (d 3) of CD3/28 bead-stimulation as under all conditions the fraction of activated CD8 + CD25 + T cells was > 95%, whereas it was < 1% before CD3/28 bead-stimulation (d 0). (D) The presence of a variety of glutamate release blockers (see text) did not impact CD8 + T cell viability after 72 hours (d 3) of CD3/28 bead-stimulation as the fraction of viable annexin V -PIcells was always > 90%.