Abstract

INTRODUCTION

There is growing evidence that glutamate, the principal central nervous system (CNS) excitatory neurotransmitter, plays a role in the pathology of the inflammatory demyelinating disease multiple sclerosis (MS) and experimental autoimmune encephalomyelitis (EAE), the model of MS. Elevated extracellular glutamate levels can result in the death of neurons and also oligodendrocytes (⁴³) through excitotoxic mechanisms. In vivo magnetic resonance spectroscopy has shown that glutamate is elevated in MS white matter and acute contrast‐enhancing lesions (⁷³). Cerebrospinal fluid glutamate levels are significantly higher in MS patients in relapse than in silent MS or controls and correlate with disease severity and course (⁶³, ⁷⁴). In active lesions high levels of glutaminase, a glutamate production marker, were seen in macrophages around dystrophic axons (⁷⁸), and damaged axons in MS white matter expressed metabotropic glutamate receptor (mGluR) 1α (¹⁹). Although glutamate can evoke T cell‐mediated protective mechanisms (⁶⁴), there is a significant decrease in glutamate‐induced inhibition of MS T cell proliferation in comparison to other neurological disease and normal control groups, suggesting an abnormal response to glutamate (³⁹).

In mice with EAE, motor transmission mediated by AMPA (α‐amino‐3‐hydroxy‐5‐methyl‐4‐isoxazole propionic acid) glutamate receptors is suppressed (²³), and a number of glutamate receptor antagonists reduce neurological deficits in EAE (⁸). Treatment at onset of neurological decline with AMPA/kainate receptor antagonists NBQX, MPQX, GYKI52466 and GYKI53773 resulted in marked reduction of neurological deficits in Lewis rat acute EAE independent of any effect on CNS lymphocyte infiltration, and inflammation‐mediated spinal cord motor neuron degeneration was reduced significantly (²², ⁷¹). In EAE mice NBQX treatment resulted in substantial amelioration of disease, increased oligodendrocyte survival and reduced dephosphorylation of neurofilament H, an indicator of axonal damage, without effect on lesion size or degree of inflammation (⁵⁸). When oligodendrocytes and neurons were protected against glutamate‐mediated damage using NBQX with a neuroprotector while inflammation was blocked, disease amelioration was associated with remyelination (³⁰). In addition, glutamate NMDA (N‐methyl‐D‐aspartate) receptor antagonists memantine, amantadine (⁷⁷) and MK‐801 (⁷) and also the glutamate neurotransmission inhibitor riluzole (²¹) used in amyotrophic lateral sclerosis (ALS) treatment, all reduce neurological deficits in EAE.

Glutamate acts on neuronal and glial ionotropic receptors (GluRs), coupled to specific ligand‐gated cationic channels, and mGluRs, coupled to second messengers (⁶⁹). AMPA receptor subunits GluR1 to GluR4 are primarily responsible for CNS fast excitatory neurotransmission and the subunit composition determines receptor electrophysiological and pharmacological properties (⁷²). These subunits assemble in vitro to form homomeric or heteromeric configurations permeable to Na⁺ and K⁺ ions, but receptors expressing only GluR1 and GluR3 form homomeric or heteromeric channels showing Ca⁺⁺ conductance that is blocked by inclusion of a GluR2 subunit (²⁷). Group I (mGluR1 and 5) are coupled to phospholipase C and K⁺ channels whilst Groups II (mGluR2–3) and III (mGlu4 and 6–8) are associated with ion channels and adenylate cyclase inhibition.

Under normal conditions astrocytes in gray matter maintain low extracellular glutamate levels by using transporters to rapidly take up glutamate, but very little is known about the ability of reactive astrocytes in white matter lesions to control extracellular glutamate levels. High extracellular glutamate levels are neurotoxic through an agonist effect on AMPA, kainate or group 1 mGluRs but the relative contribution of these receptor classes varies according to cell type, and excitotoxicity susceptibility may be under genetic control (⁴⁶). Much interest has focused on the role of glutamate in acute brain damage in ischemia, head trauma and epilepsy as glutamate is highly toxic to neurons with GluR1 Ca⁺⁺‐permeable receptors. High extracellular glutamate leads to receptor over‐activation resulting in Ca⁺⁺ entry which activates Ca⁺⁺‐dependent proteases, phospholipases, kinases, nitric oxide synthase and endonucleases. These enzymes mediate proteolysis, lipid peroxidation and free‐radical generation leading to neuronal death (²⁸). NMDA receptor activation can induce neuron death by a nitric oxide synthase dephosphorylation pathway (⁶⁰). Glutamate receptor activation may also contribute to the chronic neurodegenerative conditions Huntington's chorea, ALS and Alzheimer's disease (²⁹, ⁴⁶, ⁶⁶).

AMPA and kainate receptor agonists cause excitotoxic damage and death in oligodendrocytes in vitro (⁴¹, ⁴⁵, ⁸⁰) and in situ (³⁸, ⁴⁴). Lesions with oligodendrocyte death, demyelination, inflammation, astrogliosis and axon damage are seen in rat optic nerve after kainate infusion (⁴¹). Oligodendrocytes which lack the GluR2 subunit are susceptible to AMPA/kainate mediated excitotoxicity and in vivo delivery of excitotoxins can cause demyelinating lesions similar to those in MS (⁴², ⁴⁴). There is diversity in oligodendrocyte AMPA/kainate glutamate receptor expression between species and during development. KA2 is highly expressed in mouse oligodendrocytes while GluR4 and GluR2/3 predominate in rat (⁹). Low GluR2 and GluR3 expression in adult human oligodendrocytes is in contrast to high GluR2 and GluR2/3 in rat cells (⁷⁹). Healthy human oligodendrocytes in vitro were resistant to glutamate toxicity because of lack of AMPA/kainate receptors, and GluR1 immunostaining was not seen on oligodendrocytes in human brain but tissue sections from the CNS of MS patients were not examined (⁷⁹). Progenitor and immature oligodendrocytes are more sensitive to excitotoxic damage than mature cells (¹⁶) as expression of subunits, splicing isoforms and edited AMPA receptors change during development (¹⁷).

In MS lesions loss of oligodendrocytes, myelin and variable numbers of axons occurs during inflammatory attack. If glutamate excitotoxicity plays a role in the pathophysiology of MS, AMPA and kainate antagonists could be of therapeutic value in MS. The aim of this study was to investigate whether glutamate may be involved in the pathogenesis of MS using immunocytochemical staining to examine expression of AMPA GluR1–4, kainate GluR5–7, NMDA receptor subunit 1 (NMDAR1), mGluRs, glutamate transporters and metabolizing enzymes in MS plaques.

MATERIALS AND METHODS

Post‐mortem CNS samples from 19 cases diagnosed clinically as MS and subsequently confirmed by neuropathological examination were obtained from the NeuroResource tissue bank, UCL Institute of Neurology, London, UK. Serial cryostat sections (10µm) were cut onto Polysine coated slides (Merck, Lutterworth, UK) from a total of 42 snap‐frozen tissue blocks (approximately 1cm³) dissected from MS plaques located in brain and spinal cord white matter (⁵¹). MS cases had an average age of 46 years (range 22–65), a clinical disease duration of 13.8 years (range 9 months–27 years) and an interval between death and tissue snap‐freezing of 20.8h (range 7–34). Sections were cut from 18 normal control samples dissected from the CNS of 10 cases without neurological disease, with an average age of 59.3 years (range 28–80) and an interval between death and snap‐freezing of 18h (range 7–26).

Oil red O (ORO)‐hematoxylin staining to detect foamy macrophages containing neutral lipids resulting from ongoing or recent myelin breakdown was carried out on duplicate sections taken at the beginning, middle and end of each set of sections (³⁷). ORO macrophages and hematoxylin‐stained perivenular lymphocyte cuffing were scored on a scale of 0–5; 0 is what would be expected in normal control CNS tissues. Lesions defined as active had an ORO score of 3 to 5 with variable lymphocyte cuffing scores, whilst chronic lesions contained no or few ORO‐positive macrophages. Seventeen plaques were classified as actively demyelinating, 16 were subacute and nine were chronic.

For immunocytochemical staining sections were fixed with pre‐cooled acetone (4°C, 10 minutes), but for immunostaining with anti‐glutamate antibodies a modification of the fixation method of Moffett etal (⁵⁰) was used. Sections on 10 slides were incubated in a glass Coplin jar with 50mL of 100mM 1‐ethyl‐3‐(3‐dimethylaminopropyl)‐carbodiimide hydrochloride and 1mM N‐hydroxysuccinimide (Sigma, Poole, UK) in de‐ionized water (10 minutes, room temperature) followed by washing in phosphate‐buffered saline (PBS). After fixation all sections were incubated with blocking serum and then with 75µL of primary antibodies (45 minutes in a wet box, room temperature). Immunoperoxidase staining was carried out with a Vectastain avidin‐biotin kit (Vector Laboratories, Peterborough, UK) using diaminobenzidine and nickel (II) chloride to give black staining (⁵³) followed by dehydration and mounting in DPX (Merck).

Sections were immunostained with the following antibodies: rabbit polyclonal anti‐GluR1 subunit antibody (1:15; Upstate Biotechnology, Millipore, Temecula, CA, USA); anti‐GluR2 subunit protein (1:50), anti‐GluR 3 (1:25), anti‐NMDAR1 (1:10), and anti‐glutamate transporter EAAC1 (1:500), (all mouse monoclonals, Zymed Laboratories Inc., Invitrogen, Carlsbad, CA, USA); monoclonal anti‐glutamate receptors 5, 6, and 7 (GluR5–7; 1:300), rabbit anti‐mGluR2/3 (1:10) and anti‐glutamate dehydrogenase (GDH, 1:500), and guinea pig polyclonal anti‐glutamate transporter GLT‐1 (EAAT2, 1:500) (Chemicon International Inc., Millipore); rabbit anti‐mGluR1 and anti‐mGluR5 (1:500; Upstate); rabbit anti‐GDH (1:500; Biogenesis, AbD Serotec, Kidlington, UK); monoclonal anti‐glutamine synthetase (GS, 1:500; Transduction Laboratories, BD Biosciences, San Jose, CA, USA); monoclonal anti‐GluR4 (1:50) and a monoclonal and two rabbit anti‐glutamate antibodies (1:50–1:2000; Sigma).

Cells immunostained with these antibodies were identified on the basis of their characteristic morphology and immunostaining of adjacent serial sections with marker antibodies. Identification of cells with oligodendrocyte morphology was confirmed by immunostaining serial sections with 14E (1:10 culture supernatant; ⁵³), CNPase (rabbit, 1:400; a gift from Dr Terry Sprinkle) and carbonic anhydrase (rabbit, 1:400; a gift from Dr Norman Gregson) antibodies. These antibodies were also used to identify satellite glial cells which surround many neuronal perikarya in gray matter. The pan macrophage marker EBM11 (CD68, 1:25; Dako, Glostrup, Denmark) identified microglia and macrophages. Anti‐glial fibrillary acidic protein (GFAP) mouse monoclonal (1:1000; ⁵²) and rabbit (1:1000; ⁵² and Sigma) antibodies were used to demonstrate astrocytes. Axons were visualized with a monoclonal antibody against phosphorylated and non‐phosphorylated neurofilaments (1:100; Sigma). The T‐lymphocyte marker CD2 (1:100; Dako) and an anti‐galactocerebroside (1:100; ⁶¹) antibody to show myelin loss were also used. Plaques and their borders were localized in sections with reference to neighboring ORO and anti‐galactocerebroside stained sections.

All antibodies used in this study were titered on normal control CNS tissues and MS plaques to determine optimum antibody dilutions. For immunocytochemistry controls blocking serum was applied to the sections instead of primary or secondary antibodies. Antigen blocking experiments were carried out by pre‐incubating glutamate, gamma‐gamma glutamate and glutamine (Sigma) with antibodies against glutamate (2h at room temperature) before application to sections, in parallel with antibody dilutions which had been pre‐incubated with PBS only. A C‐terminal peptide KMSHSSGMPLGATGL, custom‐synthesized at the Advanced Biotechnology Centre, Imperial College School of Medicine, London, UK, was used in blocking experiments with an anti‐GluR1 antibody (Upstate) raised against this peptide. The peptide was confirmed to be part of the human GluR1 molecule using the NCBI Protein‐protein BLAST website.

RESULTS

Distribution of ionotropic receptors. Weak immunostaining of oligodendrocytes was seen in control and MS white matter with an antibody against the Ca⁺⁺‐permeable AMPA GluR1 subunit. However, increased numbers (Table1) of strongly stained enlarged and normal‐sized oligodendrocytes were GluR1‐immunopositive in the borders of MS plaques containing ORO‐positive macrophages, indicative of ongoing or recent demyelination (Figure1). Small numbers of oligodendrocytes were very weakly positive in chronic lesions. In and around active lesions GluR1 immunostaining of endothelial cells was elevated compared with control tissue. All oligodendrocyte and endothelial GluR1 immunostaining in control and MS samples was completely abolished by antibody pre‐incubation with GluR1 C‐terminal peptide, KMSHSSGMPLGATGL. Microglia and macrophages were still weakly immunopositive after antigen blocking, which suggested they were stained nonspecifically (Figure1). No cells, processes, axons or blood vessels were visualized when antibodies were omitted in control experiments.

Table 1

Summary of AMPA, kainate and NMDA receptor immunostaining. Abbreviations: NC=normal control; GM=cortical gray matter; WM=white matter. Cells with the morphology: OG=oligodendrocytes; Mac=macrophages; AS=astrocytes; ET=endothelial cells; AX=dystrophic axons.

Antibody and cells	GluR1		GluR2		GluR3	GluR4	GluR5–7		NMADR1
	OG†	ET	OG	Mac	AS	Mac	ET	AX	Mac
NC GM	+*	−/+	+*	−	−	−	−	−	−
NC WM	++	+	++	−	−	−	−	−	−
Active plaque rim	++++	++	+++	+++	+++	++++	+	+	++++
Active plaque	++	++	++	++	+++	++++	+	+	++++
Chronic plaque	−/+	++	−/+	−/+	−/+	−/+	−	−	−/+

Open in a separate window

The presence of immunoreactive cells is expressed as a range from − (no cells) to ++++ (high numbers of cells).

^* Perineuronal satellite glial cells.

^† Counts of GluR1‐positive cells with oligodendrocyte morphology made using a ×40 objective with a counting grid on captured images gave 786/mm²±31.4 (10 fields, n=10) in normal control brain white matter; 1560/mm²±62.4 and 869/mm²±34.8 (n=10) respectively for borders and within brain active plaques; and 84.4/mm²±3.38 (n=8) within chronic plaques.

Open in a separate window

Figure 1

Distribution of GluR1 and GluR2 immunostaining. A–C. Anti‐GluR1 immunostaining of (A) weakly immunolabeled oligodendrocytes in normal control brain white matter, (B) strongly stained oligodendrocytes in the border of an active plaque; (C) same area as (B) on an adjacent section; the antibody was antigen‐blocked by pre‐incubation with the peptide sequence used to raise the antibody; no oligodendrocytes were stained and only weakly immunolabeled macrophages seen. D. 14E immunostaining of oligodendrocytes in same area as B and C. E,F. Anti‐GluR2 staining of (E) oligodendrocytes in normal control brain white matter and (F) weak staining of oligodendrocytes and macrophages in a section serial to those shown in B–D. All acetone fixed, no hematoxylin counterstaining; scale bar 25µm.

In control white matter weak AMPA GluR2 subunit immunostaining was seen on oligodendrocytes (Figure1, Table1). Microglia were not GluR2 immunostained in control tissues but immunopositive macrophages, lymphocytes and background staining made counting the weakly GluR2‐positive oligodendrocytes difficult in lesions. Many astrocyte perikarya and processes in plaques expressed GluR3. The same normal and activated microglia and macrophage populations were immunostained with GluR4, NMDAR1 and EBM11 antibodies (Table1). In active lesions, kainate receptor GluR5–7 antibody gave strong granular immunostaining of endothelial cells in scattered capillaries and blood vessels, and of variable numbers of dystrophic axons around blood vessels (Figure2). These axons were difficult to visualize in the axon meshwork with anti‐neurofilament staining.

Open in a separate window

Figure 2

GluR5–7 immunostaining. A. Normal control white matter, longitudinal section of spinal cord; only very weak GluR5–7 staining of a blood vessel is seen. B. Punctate GluR5–7 immunostaining of endothelial cells in a spinal cord active plaque blood vessel. C. High power view of a GluR5–7‐immunopositive axon subpopulation around a blood vessel in a brain active plaque. D. Lower power view of GluR5–7‐immunopositive axons around another blood vessel in the same active plaque as in C. Scale bars 25µm.

Expression of mGluRs. The mGluR1 antibody immunostained reactive microglia and astrocytes in MS white matter and scattered dystrophic axons in active lesions, while the mGluR2/3 antibody visualized large reactive astrocytes, both intensely immunostained scattered groups of small round cells without processes or 2–4 short processes and also elongated bipolar cells in active plaques (Figure3, Table2). These cellular morphologies were not seen in any normal control tissues with mGluR1, mGluR2/3 or GFAP immunostaining. Dense astrocyte processes stained by anti‐GFAP antibodies may obscure these cells, which could be glial progenitors. By contrast, mGluR5 immuostaining was not selective as mGluR5 and GFAP antibodies gave the same staining patterns throughout lesions (Figure3).

Open in a separate window

Figure 3

Expression of mGluRs. A. Normal control white matter. B–F. Adjacent sections from the same active plaque. A–C. Anti‐mGluR1. A. Weakly immunolabeled oligodendrocytes. B. Active plaque, possible glial progenitor cells and small reactive astrocytes. C. Positive axons in another area of the same plaque. D. Anti‐mGR2/3: possible glial progenitor cells and immature bipolar astrocytes. E. Anti‐mGR5. F. Monoclonal anti‐GFAP antibody immunostaining of dense astrocyte processes. Scale bar 25µm.

Table 2

Summary of mGluR, transporters, enzymes and glutamate immunostaining. Abbreviations: NC=normal control; GM=cortical gray matter; WM=white matter. Cells with the morphology: AS=astrocytes; AX=dystrophic axons; Mac=macrophages.

Antibody and cells	mGluR1		mGluR2/3	mGluR5	GLT‐1	EAAC1		GDH		GS		Glutamate
	AS	AX	AS	AS	AS	AS	Mac	AS	Mac	AS	Mac	AS
NC GM	+	−	−	−	−/+‡	++*	−	−	−	−	−	−
NC WM	−	−	−	−	−/+‡	++	−	−	−	−	−	−
Active plaque rim	+	−/+	+	++++	++++	++	+	++	+	+++	+	++
Active plaque	−/+†	−/+	+†	++++	++++	++	+	++	+	+++	+	++
Chronic plaque	−/+	−	−/+	++++§	++++§	+++§	−/+	−/+	−/+	−/+	−/+	−/+

Open in a separate window

The presence of immunoreactive cells is expressed as a range from − (no cells) to ++++ (high numbers of cells).

^* Perineuronal satellite glial cells.

^† Possible glial progenitor cells.

^‡ Large cells with complex branching processes.

^§ Dense astrocyte processes with few visible perikarya.

Glutamate transporter and enzymes immunostaining. In normal control white and gray matter, the anti‐GLT‐1 (EAAT2) antibody only immunostained occasional unidentified large cells with complex fine branching processes (Figure4A) which were not visualized with any other antibodies. However, GLT‐1‐positive astrocytes and processes (Figure4B, Table2) had the same morphology and distribution as seen with GFAP immunostaining in active and chronic plaques. EAAC1 transporter immunostaining was seen on oligodendrocytes and astrocytes in normal white matter, and on astrocytes and in macrophages containing EAAC1‐positive granules in active plaques (Figure4C). Macrophages containing GFAP‐immunopositive granules were also seen among dense astrocyte processes in active plaques (not shown). The anti‐GDH antibodies visualized oligodendrocyte‐like cells and GS‐immunostained microglia in control white matter, and both stained hypertrophic astrocytes and foamy macrophage subpopulations in active plaques (Figure4D, Table2).

Open in a separate window

Figure 4

Anti‐GLT‐1, EAAC1 and GDH immunostaining. A,B. Anti‐GLT‐1 immunostaining of only occasional astrocytes which have complex branching processes in normal control white matter. However, in (B) dense astrocyte process staining is seen throughout an active multiple sclerosis lesion. C,D. Adjacent sections of a brain active plaque. C. EAAC1 expression in a reactive astrocyte subpopulation and granules in macrophages. D. GDH immunostaining of hypertrophic astrocytes. Scale bar 25µm.

Localization of glutamate immunostaining. The cytoplasm of a hypertrophic astrocyte subpopulation in active plaques was strongly immunostained with anti‐glutamate monoclonal or polyclonal antibodies although astrocytes were not visualized in normal white matter (Table2). Acetone or carbodiimide fixation gave the same staining patterns in all tissues but the sensitivity of monoclonal anti‐glutamate staining was increased from a dilution of 1:50 on acetone‐fixed to 1:2000 on carbodiimide‐fixed sections. GFAP antibodies also immunostained the cytoplasm of these large astrocytes in serial sections, but astrocyte processes throughout lesions were always more weakly stained with glutamate than GFAP antibodies (Figure5). Staining intensity was reduced by pre‐incubating the anti‐glutamate antibodies with glutamate.

Open in a separate window

Figure 5

Anti‐glutamate antibody immunostaining. A–C. Anti‐glutamate monoclonal antibody, 1:2000. A. Weakly immunolabeled glia in normal control brain white matter. B. Hypertrophic astrocytes surrounding a small blood vessel in an active MS plaque border. C. Top: white matter; bottom: active MS plaque. D. Anti‐GFAP, astrocyte cell bodies and processes in the same area as (B) on an adjacent section. All sections carbodiimide fixed and no hematoxylin staining. Scale bars 25µm in A, B and D; 250µm in C.

DISCUSSION

Glutamate excitotoxicity is implicated as a mechanism of neuronal death in a number of acute CNS diseases and chronic neurodegenerative disorders. As there is accumulating evidence that glutamate may also be involved in EAE and MS (²², ⁵⁸, ⁷¹, ⁷⁸) and AMPA/kainate receptor agonists cause oligodendrocyte excitotoxic death (³⁸, ⁴³, ⁴⁵), we have investigated whether glutamate may play a role in MS pathogenesis.

Immunostaining with an antibody against GluR1, a Ca⁺⁺‐permeable AMPA subunit, was elevated on oligodendrocytes in the borders and within active MS plaques in comparison with white matter. However, oligodendrocyte staining intensity with an antibody against the Ca⁺⁺‐impermeable GluR2 subunit was the same in plaque borders and white matter. GluR1 up‐regulation on oligodendrocytes around and within plaques may render these cells vulnerable to glutamate excitotoxic damage and death mediated by intracellular Ca⁺⁺ ions. Increased numbers of GluR1‐immunopositive oligodendrocytes were present in active plaque borders compared with white matter. Growth factors can induce robust and selective enhancement of GluR1 expression in functional channels on oligodendrocyte progenitors (¹⁰). Thus growth factor‐induced GluR1 up‐regulation on proliferating oligodendrocytes in MS lesions may increase their vulnerability to glutamate, and therefore limit remyelination in plaques. In chronic plaques GluR1 expression on remaining oligodendrocytes was lower than that on normal white matter oligodendrocytes.

Low Ca⁺⁺ permeability of AMPA receptors on microglia suggests they include GluR2 (²⁷), and GluR2 and also GluR3 and GluR4 are detected on rat microglia (⁵⁴). As we did not see GluR2‐positive microglia in normal control tissue, low constitutive GluR2 expression may be up‐regulated in activated microglia and macrophages by raised extracellular glutamate. Activated microglia mediate glutamate toxicity as glutamate activation of microglial AMPA receptors significantly enhances tumor necrosis factor (TNF) production (⁵⁴). This pro‐inflammatory cytokine kills oligodendrocytes (⁴⁷, ⁶⁵), enhances astrocyte glutamate release (⁶) and potentiates oligodendrocyte susceptibility to excitotoxicity (⁴⁹). Our NMDAR1 and GluR4 antibodies immunolabeled normal and activated microglia and macrophages, and it is interesting that MK‐801, the NMDA receptor antagonist which reduces EAE deficits (⁷), prevents mouse microglia activation in culture and striatum (⁷⁵). NMDA receptors are present on rodent oligodendrocyte processes and myelin (⁴⁸), but there are currently no literature reports of NMDAR1 receptors on human microglia/ macrophages or oligodendrocytes/myelin.

mGluRs also appear to play roles in MS pathology. In our study the mGluR1 and mGluR2/3 antibodies immunostained scattered reactive astrocytes and intensely labeled groups of small round and elongated bipolar cells which may be glial progenitors as normal adult human glial precursors express mGluR3 and also mGluR5‐α (⁴⁰). Our mGluR1 antibody also immunostained scattered dystrophic axons in active plaques, and mGluR1‐α has been seen in damaged MS white matter axons (¹⁹, ²⁰). In the present study mGluR5 was expressed by all astrocytes in plaques, and increased mGluR2/3 and mGluR5 staining has been reported on reactive astrocytes in MS plaques (¹⁹), in ALS (³) and in a rat epilepsy model (²). As growth factors regulate mGluR3 and mGluR5 in astrocyte cultures they may contribute to changes in glutamate receptor profiles of astrocytes as well as oligodendrocytes in plaques.

Glutamate transporters are essential for normal CNS function as they rapidly remove glutamate from extracellular fluid. GLT‐1 (EAAT2) and GLAST (EAAT1) are expressed by astrocytes at synapses and EAAC1 (EAAT3) is on the post‐synaptic neuronal element. In our study GLT‐1 was expressed only on rare cells with complex branching processes in normal tissues, although GLT‐1‐positive and GFAP‐positive astrocytes had the same distributions throughout plaques. GLT‐1 is located on astroglia in rat brain (⁶²), but all three transporters were reported on control and MS white matter oligodendrocytes (⁷⁸). GLT‐1 and GLAST increases in MS optic nerve were mimicked with excitotoxic glutamate levels in rat optic nerve (⁷⁶). TNF‐reduced human oligodendrocyte GLAST expression and glutamate uptake decreased by >75%. Loss of oligodendrocyte GLAST and EAAT‐2 around MS lesions suggested compromised glutamate uptake (⁵⁹, ⁷⁸).

EAAC1 was thought to be exclusively neuronal. In EAE GLT‐1 and GLAST were down‐regulated, but as a dramatic EAAC1 increase corresponded closely with neurological clinical signs EAAC1 may be expressed on non‐neuronal cells (⁵⁵). In our study reactive astrocytes and also granules in a macrophage subpopulation were strongly EAAC1 immunostained. Rare EAAC1‐positive astrocyte processes were seen in rabbit and rat (¹²). Granular EAAC1 staining of rat neurons suggested vesicle membrane localization (³⁵), and traumatic injury resulted in microglial de novo EAAC1 expression (³⁶). Thus EAAC1 may play a role in glutamate uptake by astrocytes and macrophages in MS lesions. We observed increased expression of glutamate metabolizing enzymes GDH and GS in active plaques, which was in contrast to their reduced expression in a non‐demyelinating EAE mouse model (²⁶). We unexpectedly detected GS, which converts glutamate to non‐toxic glutamine, in normal and activated microglia and macrophages. GS was also reported in these cells in simian immunodeficiency virus‐infected primates (¹¹).

The predominant axonal pathology observed in our study was kainate receptor GluR5–7 immunostaining of variable numbers of dystrophic axons located mostly around blood vessels in active plaques. In addition, GluR5–7 immunostaining of endothelial cells in active plaques, and elevated GluR1 in plaque and MS white matter endothelial cells, may also be related to blood–brain barrier (BBB) damage. Rat, pig and human cerebral endothelial cells express several glutamate receptor types, and high glutamate levels can disrupt the cerebral endothelial barrier via NMDA receptor activation (⁶⁸). Interestingly, BBB damage occurs in response to NMDA injection in rats (⁵⁶), and EAE treatment with NMDA antagonist MK‐801 significantly inhibited BBB disruption and restricted lesion formation and disease development (⁷)

Glutamate, the most abundant free amino acid in the CNS, is synthesized by most CNS cell types through many metabolic pathways and from protein recycling. Release of glutamate from astrocytes, activated microglia and lymphocytes (⁶, ³², ⁵⁷), or inhibition of astrocyte glutamate uptake, may result in elevation of extracellular glutamate to excitotoxic levels. Extracellular glutamate is increased up to 160‐fold in ischemia and up to 140‐fold in traumatic brain injury, but these increases are transitory and probably much lower in long‐term glutamate‐mediated diseases. Glutamate flux from plasma into CNS parenchyma can be mediated by a BBB high affinity transport system, but little glutamate is thought to pass through the intact BBB (¹³, ⁷⁰). The normal CNS interstitial glutamate concentration, 0.2–5.0µmol/L, is much lower than that of plasma. Extracellular glutamate is increased in and around rat gliomas as a result of the abnormal BBB in tumor microvasculature (⁵). Thus the leakage of plasma through a damaged BBB in MS plaques or white matter could result in local high extracellular glutamate levels. When BBB damage has been repaired, glutamate levels could be lowered to normal levels by astrocytes, and microglia may also play a role as antigen‐specific autoimmune T cells can increase the ability of microglia‐enriched cultures to remove glutamate (⁶⁷).

Although astrocytes respond to a wide variety of insults, they themselves may be damaged and lose their ability to restore homeostasis which includes glutamate uptake, K⁺ buffering and free radical elimination (²⁴). These cells were previously thought to be relatively resistant to excitatory amino acid toxins (³³), but they have been shown to express Ca⁺⁺‐permeable AMPA and kainate subunits (¹). Astrocyte damage and death result from AMPA receptor activation when desensitization is blocked pharmacologically, kainic acid injection in rat thalamus, and glutamate excitotoxicity in spinal cord (¹⁴, ¹⁵, ³⁸). Prolonged glutamate re‐uptake blockade induced the death of astrocytes on which Ca⁺⁺‐permeable GluR3 was the dominant subunit, and these cells were more vulnerable than oligodendrocytes or neurons to a gradual increase in extracellular glutamate (³⁴).

In our study active plaque astrocytes expressed GluR3, and high levels of glutamate appeared to be present in the cytoplasm of a subpopulation of large reactive astrocytes. These cells could arise as a result of glutamate‐induced swelling as elevated extracellular glutamate causes astrocyte swelling through intracellular glutamate accumulation coupled to K⁺ uptake with a secondary influx of water, which may lead to astrocyte death (⁴, ¹⁸, ³¹). Activation of mGluRs by agonists also induces swelling of rat astrocytes (²⁵). In hypocellular chronic plaques astrocyte perikarya are small and present in low numbers compared with active lesions, and the GFAP‐immunopositive granules we saw in macrophages could be remnants of dead astrocytes. Damage and loss of astrocytes in MS plaques may slow glutamate removal and enhance oligodendrocyte excitotoxic damage.

In conclusion, this study has shown that reactive astrocytes in demyelinating MS lesions in white matter appear to be well‐equipped for glutamate uptake. Therefore, these cells may have a protective role in removing extracellular glutamate from plaques. However, the uptake of local high levels of extracellular glutamate, perhaps resulting from BBB disruption, may not be rapid enough to prevent excitotoxic damage to oligodendrocytes rendered vulnerable by up‐regulation of the Ca⁺⁺‐permeable AMPA GluR1 subunit.

REFERENCES