Astrocytes Are a Specific Immunological Target in Rasmussen’s Encephalitis Jan Bauer, PhD,1 Christian E. Elger, MD,2 Volkmar H. Hans, MD,3 Johannes Schramm, MD,4 Horst Urbach, MD,5 Hans Lassmann, MD,1 and Christian G. Bien, MD2 Objective: The current histopathological criteria of Rasmussen’s encephalitis (RE) include the presence of T-cell–dominated inflammation, microglial activation, neuronal loss, and astrocytic activation. An in vitro study, however, suggested glutamate receptor 3 (GluR3) antibody–mediated astrocytic loss. Therefore, we investigated astrocytic apoptosis and loss in situ. Methods: Histochemical, immunohistochemical, terminal deoxynucleotidyltransferase–mediated biotin-dUTP nick end labeling and in situ hybridization techniques were applied to paraffin sections of 20 RE cases, 6 healthy control subjects, and 6 paraneoplastic encephalomyelitis, 10 Ammon’s horn sclerosis, and 5 focal cortical dysplasia cases. Results: Astrocytic apoptosis and subsequent loss of these cells is a specific feature of RE. Such lesions are not found in the control groups. In RE, astrocytic apoptosis and loss was present both in cortical and in white matter areas. Astrocytes in these tissues showed major histocompatibility complex class I expression. Furthermore, granzyme-B⫹ lymphocytes were found in close apposition to astrocytes bordering astrocyte-deficient lesions. Granzyme-B⫹ granules in these lymphocytes were polarized and faced the astrocytic membranes. No evidence was found for an antibody-mediated destruction. Interpretation: We suggest a specific attack by cytotoxic T lymphocytes as a possible mechanism responsible for astrocytic degeneration in RE. The loss of astrocytes might play a role in neuronal dysfunction, seizure induction, and enhancement of neuronal cell death. Ann Neurol 2007;62:67– 80 Rasmussen’s encephalitis (RE) is an inflammatory, unihemispheric brain disorder that mainly affects children,1,2 although adult cases have been described.3–5 The clinical course of RE is characterized by intractable focal onset seizures, namely, epilepsia partialis continua6 and progressive deterioration of functions associated with the affected hemisphere. The patients finally reach a residual stage with a decrease in seizure frequency and a stable neurological deficit.7,8 Inflammation is one of the characteristic features that accompanies neuropathological changes such as microglia activation and the presence of microglial nodules, neuronal loss, and astrogliosis.9 Since the first description of RE, several pathophysiological mechanisms have been suggested. Early studies described viral infections with enterovirus, Epstein– Barr virus, cytomegalovirus, or herpes simplex virus.10 –15 However, none of the studies could conclu- sively link a specific virus to RE. Studies from McNamara’s group demonstrated that immunization of rabbits with the glutamate receptor 3 (GluR3) produces a disease resembling RE and serum samples of patients contain anti-GluR3 antibodies.16,17 Furthermore, these anti-GluR3 antibodies could activate the glutamate receptor and might trigger the epileptic seizures.18 In addition, these antibodies can destroy neurons and astrocytes either directly, by excess stimulation of the receptor ion channel, or indirectly, by complement-mediated cell death.19 –21 Recent studies, however, demonstrated that anti-GluR3 antibodies are not specific for RE, and numerous RE cases were found to be GluR3 antibody–negative.22–24 Recently, we provided evidence for another mechanism of cell death in RE by showing that cytotoxic T cells may destroy neurons by the release of granzyme-B (GrB).25 There are several reasons to investigate whether, besides From the 1Division of Neuroimmunology, Center for Brain Research, Medical University of Vienna, Vienna, Austria; and Departments of 2Epileptology, 3Neuropathology, 4Neurosurgery, and 5Radiology/Neuroradiology, University of Bonn, Bonn, Germany. Received May 20, 2006, and in revised form Mar 7, 2007. Accepted for publication Mar 13, 2007. This article includes supplementary materials available via the Internet at http://www.interscience.wiley.com/jpages/0364-5134/suppmat Address correspondence to Dr Bauer, Division of Neuroimmunology, Center for Brain Research, Medical University of Vienna, Spitalgasse 4, A-1090 Vienna, Austria. E-mail: firstname.lastname@example.org Published online May 14, 2007 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ana.21148 © 2007 American Neurological Association Published by Wiley-Liss, Inc., through Wiley Subscription Services 67 neurons, astrocytes are damaged in RE. First, antiGluR3 antibodies destroyed astrocytes in vitro, suggesting that these cells might be affected in RE lesions as well. In addition, astrocytic loss recently also has been shown in animal models for epileptogenesis.26 –28 This study demonstrates that astrocytic loss is a feature of RE and suggests that its extent might affect physiological functions of neurons in the compromised areas. Subjects and Methods Brain Specimen Collection The study was performed on resective epilepsy surgery tissue or diagnostic brain biopsies from 20 RE patients studied at the Epilepsy Center of the University of Bonn8 and, for comparison, 6 control brains (autopsies from the Center of Brain Research, Vienna, Austria), 10 noninflammatory epilepsy surgery cases of Ammon’s horn sclerosis in pharmacoresistent epilepsy patients (AHS, from Bonn, Germany), 5 epilepsy surgical cases of focal cortical dysplasia (FCD, from Bonn), and 6 cases of paraneoplastic encephalitis (PE; autopsies, Bonn/Vienna). Diagnosis of RE was based on recently published diagnostic criteria.29 See Table 1 for further demographic data. Histochemistry and Immunohistochemistry For basic classification of inflammation, demyelination, and diffuse white matter injury, sections were stained with hematoxylin and eosin, Luxol fast blue myelin stain, and Bielschowsky silver impregnation. Immunohistochemical stainings (primary antibodies used are depicted in Table 2) were performed on 3 to 5m paraffin sections. Before staining, endogenous peroxidase was blocked by 30-minute incubation in methanol with 0.02% H2O2. This was followed with antigen retrieval by heating the sections for 90 minutes in EDTA (0.05M) in tris(hydroxymethyl)aminomethane (Tris) buffer (0.01M, pH 8.5) in a household food steamer device (MultiGourmet FS 20; Braun, Kronberg/Taunus, Germany). To detect IgG and complement C9, we incubated sections with 0.03% protease from Streptomyces griseus (Sigma, St. Louis, MO) for 15 minutes at 37°C. Sections were then incubated with 10% fetal calf serum (FCS) in 0.1M phosphate-buffered saline (FCS/PBS). Next, primary antibodies were applied in FCS/PBS at 4°C overnight. After washing with PBS, secondary antibodies in PBS/FCS with 3% normal human serum were applied for 1 hour at room temperature. We used biotinylated secondary antibodies at a concentration of 1:200 (donkey anti–rabbit, sheep anti– mouse; Amersham Pharmacia Biotech, Uppsala, Sweden). As a third step, avidin peroxidase (1:100; Sigma) was used. For the CD3, CD8, major histocompatibility complex (MHC) class I, and GrB stainings, biotinylated tyramine enhancement was used as described previously.25 Immunoglobulin staining was done with biotinylated sheep anti–human antibody (Amersham). Labeling was visualized with 3,3⬘ diaminobenzidine-tetrahydrochloride (Sigma) or aminoethyl carbazole (Sigma). In case of double labeling for caspase-3 with glial fibrillary acidic protein (GFAP), 2⬘3⬘cyclic nucleotide 3⬘ phosphohydrolase (CNPase), or microtubule-associated protein 2 68 Annals of Neurology Vol 62 No 1 July 2007 (MAP2) or neuronal nuclei (NeuN), both primary antibodies were diluted in FCS/PBS and sections were incubated at 4°C overnight. After washing in PBS, a mixture of alkaline phosphatase–conjugated goat anti–mouse antibodies (Jackson ImmunoResearch, West Grove, PA) and biotinylated donkey anti–rabbit (1:100; Amersham) antibodies were applied in FCS/PBS with 3% normal human serum for 1 hour at room temperature. As a third step, avidin peroxidase (1:100; Sigma) was applied for 1 hour at room temperature. Next, the alkaline phosphatase label was visualized with fast blue B base (Sigma) as substrate. This was followed by visualization of peroxidase with diaminobenzidine-tetrahydrochloride or aminoethyl carbazole. Finally, sections were counterstained with hematoxylin. In Situ Hybridization Detection of messenger RNA was performed as described previously.30 In brief, paraffin sections were dewaxed, pretreated with 10g proteinase K (Sigma) in Tris-buffered saline (pH 7.2) and incubated with digoxigenin-labeled probes specific for GFAP (gift from Melitta Schachner, Hamburg, Germany) or proteolipid protein (PLP).30 Sections then were incubated with alkaline phosphatase–labeled anti-digoxigenin antibody (Boehringer-Mannheim, Mannheim, Germany) and developed using Niro blue tetrazolium chloride/5-Bromo4-chloro-3-indolyl phosphate (NBT/BCIP), (BoehringerMannheim) as substrate. Sections stained for GFAP or PLP mRNA were double stained with anti-GFAP antibodies with alkaline phosphatase–conjugated secondary antibodies using fast red as substrate. As a result, mRNA appears black, whereas GFAP protein is stained red. Sections were counterstained with hematoxylin. Terminal Deoxynucleotidyltransferase–Mediated Biotin-dUTP Nick End Labeling To detect cells with DNA fragmentation, we performed terminal deoxynucleotidyltransferase–mediated biotin-dUTP nick end labeling (TUNEL) with the In Situ Cell Death Detection Kit (Alkaline Phosphatase) from Roche (Mannheim, Germany). Paraffin sections were deparaffinized, treated with chloroform, and air-dried. This was followed by incubation with labeled dUTP in the presence of terminal transferase according to the manufacturer’s guidelines. Sections were developed with NBT/BCIP. Subsequently, the sections were stained for GFAP as described earlier. As a result, DNA fragmentation in the nucleus appears black, whereas GFAP protein appears red. Confocal Laser Fluorescence Microscopy Fluorescence immunohistochemistry was performed on paraffin sections as described earlier with few modifications. For confocal fluorescent double labeling with primary antibodies from different species (eg, rabbit anti-GFAP and mouse anti-GrB), antibodies were applied simultaneously at 4°C overnight. After washing with PBS, secondary antibodies consisting of goat anti–mouse Cy3 (1:200; Jackson ImmunoResearch) and biotinylated anti–rabbit (1:200; Amersham) were applied simultaneously for 1 hour at room temperature. The staining was finished by application of streptavidin-Cy2 (1:75; Jackson ImmunoResearch) for 1 hour at room temperature. In Table 1. Patient Data and Semiquantitative Quantification of Astrocytic and Neuronal Degeneration and Loss Patient No. and Sex Age at Brain Specimen Collection (yr) Disease Duration after Onset of Acute Stage (mo) Site of Specimen Collection Number of Sections Stagea Astrocytic Apoptosisb Astrocytic Lossc (range) Neuronal Apoptosisb Neuronal Lossc (range) Rasmussen encephalitis (n ⫽ 20) 1, M 2, F 3.7 39.0–40.8d 0.6 Frontal neocortex 1 2 Y 1 Y 1 1.8/2.3 Superior temporal gyrus Precentral gyrus 3 2–3 Y/N 1 N 0–3 1 2 Y 0 ND ND 3, F 1.9 2.0 Temporal neocortex 4, F 13.6 3.6 Superior frontal gyrus 1 2 Y 2 N 2 5, M 7.9 5.5 Insula 1 2 N 3 N 1 6, M 5.8 6.7 Superior frontal gyrus 1 2 N 3 Y 2 7, M 22.2 7.1 Frontal neocortex 1 2 Y 2 Y 2 8, F 15.0 7.5 Temporal neocortex and parahippocampal gyrus 3 3 Y 1–3 Y 3 9, M Medial frontal gyrus 2 2 Y 1–2 Y 1–3 10, F 3.8/3.9d 4.9 15.6/17.2 8.8 Temporal neocortex/ rolandic region 5 3 Y 1–2 Y 2–3 11, F 27.4 27.3 Temporal and parietal neocortex parahippocampal gyrus 3 3 Y/N 2–3 Y/N 1–3 12, M 8.2 28.1 Frontal and temporal neocortex 5 2 Y 2 Y 2–3 13, F 45.3 33.9 Inferior temporal gyrus 1 3 N 1 N 3 14, F 7.7 44.6 Temporal neocortex 1 3 Y 1 Y 2 15, M 23.8 58.8 Temporal neocortex 1 3 Y 1 Y 2 16, M 15.9 111.8 Amygdala parahippocampal gyrus hippocampus temporal neocortex 3 3 Y/N 1 Y 3 17, F 6.7 11.2 Superior frontal gyrus 1 2 Y 2 N 2 18, M 6.8 2.7 Superior frontal gyrus 1 2 Y 2 Y 2 19, F 3.0 8.8 Parietal neocortex 1 2 Y 0 N 0 20, M 7.6 18.3 Hippocampus with parahippocampal gyrus 1 3 Y 1 N 0 Control subjects (n ⫽ 6; 3 F, 3 M) 67.0 ⫾ 4.9 ⫺ Neocortex 1/case ⫺ N N N N Ammon’s horn sclerosis (n ⫽ 10, 5 M, 5 F) 17.8 ⫾ 9.9 ⫺ Hippocampi 1/case ⫺ N N N Y Focal cortical dysplasia (n ⫽ 5; 3 M, 2 F) 30.6 ⫾ 18.6 ⫺ Neocortex 1/case ⫺ N N N Y Paraneoplastic encephalitis (n ⫽ 6, 3 F, 3 M) 65.8 ⫾ 13.6 4.6 ⫾ 3.6 Diverse 1/case ⫺ Ye N N Y 1 ⫽ prodromal stage; 2 ⫽ acute stage; 3 ⫽ residual stage (see Bien and colleagues8). Presence of caspase-3–positive astrocytes or neurons, respectively. c Semiquantitative assessment, categories 1–3; for details, see Subjects and Methods. d For Patients 2 and 10, specimens from two surgical procedures were evaluated. e Some single scattered apoptotic astrocytes in one case, in the absence of large loss of astrocytes. ND ⫽ not determined. a b Bauer et al: Astrocytic Degeneration in RE 69 Table 2. Primary Antibodies Antigen Pretreatment Dilution Antibody Type Target Source CNPase Steamer 1:2,000 mAb, mouse Oligodendrocytes Sternberger Monoclonals, Lutherville, MD GFAP Steamer 1:6,000 1:200 polyAb, rabbit mAb, mouse Astrocytes Dakopatts, Hamburg, Germany Labvision, Fremont, CA MAP2 Steamer 1:200 mAb, mouse Neurons Sigma, St. Louis, MO NeuN Steamer 1:100 mAb, mouse Neurons Chemicon, Temecula, CA S100␤ Steamer 1:200 mAb, mouse Astrocytes Labvision, Fremont, CA APP Steamer 1:1,000 mAb Axons Boehringer-Mannheim, Germany MAG Steamer 1:4,000 mAb Myelin Gift from Dr C. Linington, Aberdeen, United Kingdom53 MHC class I (HC10) Steamer 1:2,000 (C) 1:250 (F) mAb, mouse ␣-Chain MHC class I Gift from Dr Ploegh, Harvard Medical School, Boston, MA54 CD3 Steamer 1:1000 (C) 1:50 (F) polyAb, rabbit T cells Dakopatts, Hamburg, Germany CD8 Steamer 1:250 (C) mAb, mouse T cells Labvision, Fremont, CA Caspase-3 (CM-1) Steamer 1:3,000 polyAb, rabbit Activated caspase-3 in apoptotic cells Becton Dickinson, San Diego, CA Granzyme-B (clone GZB01) Steamer 1:1,000 (C) 1:50 (F) mAb, mouse Granzyme-B Labvision, Fremont, CA CD68 Protease 1:100 mAb, mouse Macrophages, microglial cells Dakopatts, Hamburg, Germany C9neo Protease 1:20 polyAb, rabbit Lytic complement complex S. Piddlesden, University of Cardiff, United Kingdom IgG Protease 1:200 polyAb, goat IgG, plasma cells Amersham Pharmacia Biotech, Uppsala, Sweden CNPase ⫽ 2⬘3⬘cyclic nucleotide 3⬘ phosphohydrolase; mAb ⫽ monoclonal antibody; GFAP ⫽ glial fibrillary acidic protein; polyAb ⫽ polyclonal antibody; MAP2 ⫽ microtubule-associated protein 2; NeuN ⫽ neuronal nuclei; APP ⫽ amyloid precursor protein; MAG ⫽ myelin-associated glycoprotein; C ⫽ Catalysed System Amplification (see Subjects and Methods); F ⫽ fluorescence; MHC ⫽ major histocompatibility complex. case of triple labeling for mouse anti-GFAP/mouse anti-GrB/ rabbit anti-CD3, the three primary antibodies were applied simultaneously and the staining was finished as described earlier. This results in the presence of red GrB⫹ granules inside of a green CD3⫹ T lymphocyte and red GFAP⫹ astrocytes. Fluorescent preparations were examined using a confocal laser scanning microscope (LSM 410; Carl Zeiss, Jena, Germany). Analysis of Astrocyte, Neuronal, and Oligodendrocyte Degeneration and Loss Sections double stained for light microscopy (combinations of caspase-3 with GFAP, S100␤, NeuN, MAP2, or CNPase) 70 Annals of Neurology Vol 62 No 1 July 2007 were analyzed for the presence of apoptotic astrocytes, neurons, and oligodendrocytes, and for loss of these cells. First, whole sections from RE cases and control subjects (no neurological disease, AHS, FCD, PE) were semiquantitatively studied for loss of GFAP⫹ astrocytes and MAP2⫹ neurons by two investigators (J.B., C.G.B.). Loss was scored by division in three categories: (1) loss of some cells (mostly around blood vessels), often found together with the presence of caspase-3⫹ astrocytes or neurons; (2) small- to middle-sized areas (lesions) with loss of cells; and (3) large (cortical or white matter) areas with loss of cells. Second, we determined the magnitude of astrocytic and neuronal loss. We assessed the lesional size, that is, area without GFAP or combined NeuN/MAP2 signal within the cortex of each sample. As a first step, we analyzed the size of gray and white matter of each sample as follows: The largest available Luxol fast blue–stained section of each patient containing cortex and white matter was scanned at 1,000dpi. In these images, the cortical gray matter and underlying white matter were outlined and measured using Scion Image (freeware, http://www.scioncorp.com), an image processing and analysis program based on National Institutes of Health Image. Next, at 10⫻ objective magnification, a 100-point morphometric ocular grid was sequentially superimposed over the whole section area. The size of GFAP- or NeuN/MAP2-deficient areas was assessed by counting the number of grid points (intersections), which were not on all sides surrounded by GFAP⫹ or NeuN/MAP2⫹ cells. The total GFAP- or NeuN/MAP2deficient area, indicated by the total number of “lesional” grid points, was divided by the total cortical area of the specimen and the result given as percentage of area devoid of GFAP⫹ or NeuN/MAP2⫹ cells. In addition, this procedure was performed for GFAP⫹ cells in white matter. Third, we analyzed loss of GFAP⫹ or S100⫹ astrocytes and NeuN⫹ neurons in specific RE areas and compared the numbers of cells with the numbers of astrocytes and neurons in healthy control subjects. Areas in RE were divided into three categories: (I) “RE GFAP normal,” no apparent loss of GFAP reactivity; (II) “RE GFAP loss,” areas with lack of GFAP (absence of GFAP-reactive cells); and (III) “RE GFAP gliosis,” areas with dense fibrillary gliotic scar tissue. The numbers of GFAP⫹, S100␤⫹, and MAP⫹ cells were measured in control and RE brain by superimposing a morphometric grid at 20⫻ objective magnification. Cell counting started at the molecular layer (cortical layer 1) and was continued perpendicular to the meningeal lining until the white matter was reached. Cortical areas were taken in the middle of sulcal banks to standardize the procedure as much as possible, thereby avoiding the thinner sulcal bases and thicker gyral crowns. The assessed areas are 0.5mm broad strips of cortex of variable length (normally about 4mm) perpendicular to the cortical surface. The results of this quantification procedure are given in cells per square millimeter of cortex. Statistical Analysis Two-sided Pearson’s or two-sided t tests were performed, as appropriate; p ⬍ 0.05 was considered significant. Results Qualitative Assessment of Apoptotic Astrocytic and Neuronal Loss in Active Rasmussen’s Encephalitis Lesions Neuronal destruction and loss is regarded as the hallmark of lesional pathology in RE.9 A further characteristic feature of RE is the presence of astrogliosis, that is, the presence of hypertrophic, strongly stained, GFAP⫹ reactive astrocytes. Here, however, we investigated astrocytic loss in RE. In all RE patients, large parts of RE cortex showed GFAP⫹ astrocytes with normal or activated morphology. Many patients, however, also showed some loss of GFAP⫹ astrocytes around blood vessels or loss of astrocytes close to the meningeal lining (Figs 1B and 4Q). In about one third of the patients, loss of GFAP⫹ astrocytes was even more pronounced. Here, multiple small- to middle-sized lesions with loss of GFAP reactivity were found (see Figs 1G and 4M). Finally, in some patients, GFAP reactivity was absent in large parts of the cortex. Often these lesions were bordered by hypertrophic activated astrocytes strongly stained for GFAP (see Figs 1I, 2D, and 4A). Because these lesions were not completely devoid of nuclei, we also examined these lesions for the presence of other cells. Stainings for neurons (MAP2 or combination of NeuN with MAP2) showed that cell bodies were largely missing (see Figs 1F, H). Bielschowsky staining, in addition, showed that, in cortical areas, the axonal density often was largely reduced (not shown), whereas in white matter lesions, GFAP loss was not always combined with axonal injury (see underneath). Staining for oligodendrocytes cell bodies by CNPase and microglial cells by CD68, however, showed that the oligodendrocytes and microglial cells were present in normal numbers. The range of astrocytic loss in all sections from all RE patients was determined semiquantitatively and presented in Table 1. Lack of staining for GFAP does not necessarily mean that astrocytes are in fact lost. To confirm that there was actual astrocytic loss, we performed additional immunohistochemical stainings for another marker for astrocytes (always S100␤), as well as in situ hybridization (ISH) for GFAP mRNA (on selected cases that were graded with severe loss in Table 3). In areas with cortical loss of GFAP⫹ astrocytes, we also found a loss of S100␤⫹ cells (see Fig 1J) and of GFAP mRNA⫹ cells (see Fig 1L). In contrast, ISH for PLP mRNA demonstrated numbers of oligodendrocytes comparable with those in areas without loss of astrocytes (see Fig 1M). This indicates that loss of GFAP mRNA is not due to a general downregulation of mRNA, but rather to a selective astrocytic loss. Taken together, these findings confirm that loss of GFAP⫹ cells does not merely reflect a loss of GFAP immunoreactivity but is a real loss of astrocytes. Staining for activated caspase-3 demonstrated the presence of single caspase-3⫹ cells within otherwise normal-appearing cortices (see Fig 1A). Double-labeling studies for caspase-3 and GFAP demonstrated that these caspase-3⫹ cells exhibited astrocytic morphology and were either (weakly) GFAP⫹ (see Figs 1B, 1C, 4C, and 4M), had GFAP reactivity only in part of the cytoplasm (see Figs 4E, F), or were GFAP⫺ (see Figs 1A, 1B, 4D, and 4M). Besides the presence of activated caspase-3 immunoreactivity, these caspase-3⫹ or GFAP⫹ astrocytes showed nuclear condensation (see Figs 1D and 4C) or nuclear fragmentation (see Figs 1E and 4D). Moreover, although in much lower numbers than caspase-3⫹ astrocytes, TUNEL staining in combination with GFAP Bauer et al: Astrocytic Degeneration in RE 71 Fig 1. Astrocytic pathology in Rasmussen’s encephalitis (RE). (A–C) Immunohistochemistry (IHC) for glial fibrillary acidic protein (GFAP) (blue) and caspase-3 (brown). (A) Two cells with morphology of astrocytes. Whereas the lower cell (arrowhead) is double stained, the other one shows staining only for caspase-3 (brown), indicating loss of only GFAP reactivity in this stage. Original magnification ⫻388. (B) An area without GFAP reactivity around a small blood vessel with caspase-3⫹ astrocyte-like cells in the center, whereas on the edge a double-stained astrocyte (arrowhead) is shown. Original magnification ⫻308. (C) A greater magnification of the latter cell shows condensation of the nucleus (arrowhead), another criterion for apoptosis. Original magnification ⫻792. (D, E) Two examples of GFAP-positive astrocytes with condensed (D, arrowhead) or fragmented (E, arrowhead) nuclei indicative of apoptosis. Original magnification ⫻990 (D, E). (F) Staining for microtubule-associated protein 2 (MAP2) shows a cortical lesion with loss of MAP2⫹ neurons. Original magnification ⫻40. (G) The same area as in (F) stained for GFAP (blue) and caspase-3 (brown). Original magnification ⫻40. The border of this lesion is graphically delineated and was projected in (F). Inside the delineation this lesion is almost devoid of GFAP⫹ astrocytes. (H) Greater magnification of the rectangle in (F) showing that left, above, and on the right side of the lesion MAP2-positive neuronal cell bodies (arrowheads) are still present. Original magnification ⫻60. (I) Enlargement from the edge of the lesion (rectangle) from (G) showing the presence of caspase-3–stained (brown, arrowheads) apoptotic glial cells. Original magnification ⫻161. (J) Double staining for S100␤ (blue) and caspase-3 (brown) in the middle of a cortical area shows some double-stained S100␤⫹ astrocytes (arrowheads). Original magnification ⫻308. (K) Double staining for CD68 (blue, microglial cells) and caspase-3 (red) shows that microglial cells are not apoptotic. Original magnification ⫻244. (L) ISH for GFAP messenger RNA (mRNA; black) in combination with IHC for GFAP protein (red) shows a small white matter area in which both GFAP mRNA and GFAP protein are lost. Original magnification ⫻96. (M) ISH for proteolipid protein (PLP) mRNA in combination with GFAP IHC (red) does not show any downregulation of oligodendrocyte mRNA (black dots). Thus, no loss of oligodendrocytes is found. Original magnification ⫻96. 72 Annals of Neurology Vol 62 No 1 July 2007 Fig 2. Neurons and astrocytes in controls and various epileptic diseases. Staining for neuronal nuclei (NeuN) (A, C, E, G; original magnification ⫻80) and glial fibrillary acidic protein (GFAP) (B, D, F, H; original magnification ⫻80) in the cortex (layers I [left] to VI [right]) of control subjects (con; A, B), Rasmussen’s encephalitis patients (RE; C, D), cortical dysplasia patients (FCD; E, F), and in the CA1 hippocampal region of Ammon’s horn sclerosis (AHS; G, H) patients. In RE, both neuronal loss and astrocyte loss is seen. In FCD, the NeuN staining shows a lower density of neurons and the presence of balloon cells (arrowheads). GFAP reactivity is seen in hypertrophic astrocytes. In AHS (G, H; rectangle in the inset shows localization of G), the right side of the CA1 region shows loss of neurons, whereas in the same area with GFAP, a dense gliosis is observed. demonstrated the presence of astrocytes with DNA fragmentation (see Fig 4G). These three combined characteristics (activated caspase-3 reactivity, nuclear condensation, and DNA fragmentation) imply that astrocytes die by apoptosis. The presence of caspase-3⫹/GFAP⫺ cells raises the question whether some of these cells could be microglial cells. Double staining for caspase-3 and CD68, however, showed no CD68/caspase-3 coexpression (see Fig 1K), thus suggesting that these caspase-3⫹ cells are indeed astrocytes. Astrocytic loss and presence of caspase-3⫹ astrocytes in the different sections from the various RE patients are depicted in Table 1. Although, in most cases, the caspase-3⫹ cells had astrocytic morphology, some of these cells were much larger and had only one or two short processes, suggesting that they were apoptotic neurons. However, these caspase-3⫹ neurons were found only rarely. Two of 20 patients showed astrocytic apoptosis without apparent loss of astrocytes (see Table 1; Patients 3 and 19). Caspase-3⫹ cells with astrocytic morphology were found in and on the borders of all lesions with astrocytic loss (see Figs 1A, B). Bauer et al: Astrocytic Degeneration in RE 73 Table 3. Areas with Astrocytic and Neuronal Loss Group Cases (n) Mean NeuN/MAP2 Loss ⴞ SEM (%) Mean Area ⴞ SEM (mm2) Mean GFAP Loss ⴞ SEM (%) Mean S100␤ Loss ⴞ SEM (%) 6 144.3 ⫾ 38.5 0.2 ⫾ 0.4 0.0 ⫾ 0.1 0.1 ⫾ 0.1 RE 20 75.7 ⫾ 51.7 7.6 ⫾ 9.8 6.0 ⫾ 6.5 3.4 ⫾ 4.9 AHS 10 31.6 ⫾ 12.3 21.3 ⫾ 13.5 0.0 ⫾ 0.0 0.0 ⫾ 0.0 FCD 5 140.8 ⫾ 66.3 8.6 ⫾ 3.7 0.0 ⫾ 0.0 0.0 ⫾ 0.0 PE 6 213.9 ⫾ 73.1 0.0 ⫾ 0.0 0.0 ⫾ 0.0 0.0 ⫾ 0.0 CON In control subjects (CON), Rasmussen’s encephalitis (RE) patients, and focal cortical dysplasia (FCD) patients, neocortical areas were quantified. In Ammon’s horn sclerosis (AHS) patients, quantification was done only in the pyramidal layer of the cornu ammonis and the granular layer of the dentate gyrus. In paraneoplastic encephalomyelitis (PE) patients, quantification was performed in inflammatory regions. SEM ⫽ standard error of the mean; NeuN ⫽ neuronal nuclei; MAP2 ⫽ microtubule-associated protein 2; GFAP ⫽ glial fibrillary acidic protein. The extent of loss of GFAP-reactive astrocytes and of apoptotic (caspase-3⫹) astrocytes not only varied between patients but also between different biopsy regions (sections) from one patient and even within one section. These differences in the degree of apoptosis and loss of GFAP⫹ cells between the various sections, which are summarized in Table 1, probably reflect different stages of astrocytic loss. Astrocytic Destruction and Loss Is a Specific Feature of Rasmussen’s Encephalitis Pathology To investigate the specificity of astrocytic loss, we quantified the degree of astrocytic and neuronal loss in the cortex of RE patients and compared these with healthy control subjects, brains of PE patients, and brains of epilepsy patients (AHS and FCD). The degree of astrocytic loss (GFAP or S100␤) or neuronal loss (NeuN/MAP2) in these control subjects and various patient groups is presented in Table 3. In FCD and AHS cases, a considerable degree of neuronal loss was found. In FCD, astrocytes in these areas with reduced neuronal density showed a hypertrophic morphology. In AHS, neuronal loss was colocalized with fibrillary gliosis. Areas with loss of astrocytes, as seen in RE, were found neither in FCD nor in AHS patients. Representative pictures from stainings for GFAP and NeuN from control subjects and RE, FCD, and AHS cases are depicted in Figure 2. Evidence for apoptosis of astrocytes in these brains (see Table 1), except for some single apoptotic cells with astrocytic morphology in one PE case, was absent. Taken together, these results suggest that astrocytic apoptosis and loss are specific features of RE. Rasmussen’s Encephalitis Cortex with Astrocytic Fibrillary Gliosis Besides normal-appearing areas and areas with (combined) astrocytic and neuronal loss in RE, we also 74 Annals of Neurology Vol 62 No 1 July 2007 found areas with profound fibrillary astrocytic scar formation. These areas were present in a small number of patients (N ⫽ 4; Patients 2, 9, 11, and 12 in Table 1). All of these patients were in the residual clinical stage (stage 3). NeuN or MAP2 immunohistochemistry, as well as Bielschowsky’s silver impregnation in these lesions, demonstrated severe cortical atrophy with almost complete loss of neurons and axons. A combined GFAP/caspase-3 staining did not show apoptotic astrocytes in these areas. Relation of Astrocytic and Neuronal Loss Figure 3 shows the correlation of cortical areas devoid of neurons versus the cortical areas devoid of astrocytes. In 13 of 20 RE patients, the area of astrocytic and neuronal loss was less than 10% of the total cortex investigated. The other seven patients showed more extensive areas of astrocytic (up to 26%) and neuronal loss (up to 36%). Astrocytic and neuronal loss were significantly correlated ( p ⫽ 0.01; see Fig 3), although the extent of neuronal loss appears larger than that of astrocytic loss. A representative pair of figures illustrates a similar degree of loss of GFAP⫹ astrocytes and MAP2⫹ or NeuN⫹ neurons (see Figs 1F, 1G, 2C, and 2D) in a cortical area. Taken together, we were able to separate cortical areas into three categories: (I) areas without apparent astrocytic or neuronal loss, but in some cases, in the presence of single apoptotic cells (underneath referred to as RE “GFAP normal”); (II) areas with loss of GFAP-reactive astrocytes and neurons in the presence of apoptotic astrocytes (RE “GFAP loss”); and (III) areas with dense GFAP⫹ fibrillary gliotic scar formation and severe loss of neurons, in the absence of apoptotic astrocytes (RE “GFAP gliosis”). We quantified the densities of neurons and astrocytes in the three cortical areas (areas without GFAP loss [I], with GFAP loss [II], and with GFAP gliosis reactive hypertrophic astrocytes surrounded the larger lesions. Immunohistochemical staining for CNPase (see Fig 4B) and ISH for PLP (see Fig 1L) demonstrated that oligodendrocytes and myelin in these white matter areas were unaffected. In addition, Bielschowsky silver stain for axons (see Fig 4I) and staining for ␤-amyloid precursor protein (see Fig 4J) showed acute axonal injury in some areas. However, astrocytic loss was not coupled to axonal injury of traversing fiber tracts. The percentage of areas with loss of astrocytes in subcortical white matter ranged from 0 to 22.0% (average ⫾ standard error: 2.0 ⫾ 1.8%). Possible Mechanisms Responsible for Astrocytic Loss ASTROCYTIC LOSS AND IMMUNOGLOBULIN AND COMPLEMENT DEPOSITION Because complement-mediated Fig 3. Quantification of astrocytic and neuronal loss in cortical areas from Rasmussen’s encephalitis (RE) patients. In 20 cortical sections from 20 RE cases, the areas (average area ⫾ standard error ⫽ 75.7 ⫾ 4.5; minimal area ⫽ 20.2mm2; maximal area ⫽ 223.3mm2) with loss of glial fibrillary acidic protein (GFAP; astrocytes) or microtubule-associated protein 2/neuronal nuclei (MAP2/NeuN; neurons) was quantified. Areas with loss of astrocytes ranged from 0 to 37%, whereas areas with loss of neurons ranged from 0 to 26%.The correlation between astrocyte and neuronal loss was significant (p ⫽ 0.01). [III]) and compared those with the respective cells in control cortices. These results are shown in Figure 5. Densities of GFAP⫹ and S100␤⫹ cells in “GFAP normal” areas did not differ from those in control brain. However, the numbers of neurons in these areas were decreased as compared with the cortex in control subjects. In areas with GFAP loss, the numbers of all cell types were significantly diminished. In gliotic areas (RE “GFAP gliosis”), the numbers of GFAP⫹ and S100␤⫹ cells were increased in comparison with the control subjects, as well as with the other RE specimens. Staining for NeuN, however, demonstrated a loss of neurons as compared with the control subjects and the other RE subgroups. astrocyte death has been noted after treatment with anti-GluR3 antibodies,21 we investigated whether immunoglobulin or complement deposition on astrocytes was present in our material. In most cases, a diffuse staining for immunoglobulin and complement C9neo was observed around blood vessels (see Figs 4K, L). In areas with astrocytic loss, IgG and C9neo deposition on astrocytes, however, was absent. This suggests that a complement-mediated killing of these cells can be ruled out. Furthermore, 4 of our 20 patients have been tested for serum anti-GluR3 antibodies. None of them was found to be positive.25 HYPOXIA-LIKE TISSUE DAMAGE AS CAUSE OF CELL LOSS. Severe brain inflammation may result in hypoxia-like tissue injury that shows specific loss of myelinassociated glycoprotein immunoreactivity31 and could be responsible for damage to both neurons and astrocytes. To check whether the loss of GFAP immunoreactivity could also be the result of ischemia, we stained for PLP mRNA and myelin-associated glycoprotein immunoreactivity in areas showing loss of GFAP⫹ cells. ISH demonstrated a loss of GFAP mRNA⫹ cells (see Fig 1L), but no loss of PLP mRNA (see Fig 1M) or myelin-associated glycoprotein (not shown) in the same areas. These results suggest that loss of astrocytes or neurons does not result from hypoxia-like tissue injury. ASTROCYTE AND CYTOTOXIC T-CELL INTERACTIONS. Astrocytic Apoptosis and Loss Also Affect the Subcortical White Matter Astrocytic apoptosis and loss in RE cases are not restricted to cortical areas. To our surprise, we discovered small and large areas devoid of GFAP⫹ astrocytes deeply in the subcortical white matter (see Fig 4A). Many of these white matter lesions showed caspase-3⫹ or TUNEL⫹ astrocytes within as well as on the edge of these lesions (see Figs 4C–H). Similar to cortical areas, Previously, we showed that cytotoxic T lymphocytes interacting with neurons are a considerable component of the infiltrating inflammatory cells in RE.25 To investigate whether astrocytic cell death could be mediated by cytotoxic T lymphocytes, we performed double (GFAP and GrB) and triple (GFAP, CD3, and GrB) stainings. Cytotoxic T cells were present around blood vessels and on the border of lesions in which astrocytes were dying or already lost (see Figs 4N, Q). Often, Bauer et al: Astrocytic Degeneration in RE 75 . Figure 4 single or multiple cytotoxic T cells were seen in close apposition to astrocytes. GrB⫹ granules in several lymphocytes were polarized and facing the astrocyte surface (see Figs 4N–P) similar to that described in in vitro GrB studies.32 Because MHC class I expression is a prerequisite for antigen-specific cytotoxicity mediated by CD8⫹ T cells, we performed stainings with for MHC class I. Although weaker than on endothelial cells, lymphocytes, neurons, and microglial cells, astrocytes indeed were MHC class I⫹ (see Fig 4R). The online supplementary data show to what extent loss of astrocytes, neurons, and gliosis may be reflected in MRI recordings. Discussion The current histopathological criteria for the diagnosis of RE include the presence of T-cell–dominated inflammation, microglial activation, and microglial nod- 76 Annals of Neurology Vol 62 No 1 July 2007 ules, as well as neuronal loss and astrocytic activation. This study suggests that astrocytic apoptosis and loss, most probably induced by a cytotoxic T-cell response toward astrocytes, is a common finding in RE. The major arguments for the presence of astrocytedeficient lesions are the observations of areas that specifically lack GFAP or S100␤ protein, GFAP mRNA, as well as the presence of apoptotic astrocytes at the borders of those regions. Oligodendrocytes, stained for PLP mRNA or CNPase, and CD68⫹ microglial cells were present in normal numbers, indicating that these lesions are not necrotic or hypoxic in nature. Astrocyte death has been shown in several studies of animal models for epilepsy.26 –28,33,34 Therefore, the possibility exists that astrocytic loss is not specific for RE but is a consequence of epileptic seizures per se. Alternatively, astrocytes may die from an unspecific reaction to inflammation. Astrocytic loss, however, was not present in noninflammatory focal epilepsy controls (AHS and FCD) or in PE cases (ie, inflammatory control). In AHS cases, however, severe gliosis was seen in the hippocampal CA1 region where maximal neuronal loss is observed. This scarlike gliosis is comparable with gliotic areas in the severely affected RE specimens. It could still be possible that, during the early stages of AHS, hippocampal astrocytes are lost. On the other hand, epileptic seizures in these AHS cases were present at the time of epilepsy surgery, and astrocyte death, if generated by the seizures, could be expected. Therefore, astrocytic loss does not appear to be a consequence of epileptic seizures or an unspecific inflammatory reaction and should be considered specific for RE. Double labeling of GFAP with caspase-3 showed that only a fraction of the caspase-3⫹ cells contain GFAP reactivity in parts of the cytoplasm, whereas others, although showing an astrocytic morphology, fail to express GFAP completely. Furthermore, we found much lower numbers of GFAP⫹/TUNEL⫹ cells than cells double labeled for GFAP and caspase-3. These findings suggest that, in astrocytes, upregulation of caspase-3 is followed by a degradation of GFAP, suggesting that GFAP itself is a substrate for caspase-3. Š Comparable results were found in neurons in glutamate excitotoxity studies in vitro.35 An important question is the reason for the specific astrocyte death described here. A previous study21 provided evidence that treatment of mixed neural cell cultures with anti-GluR3 antibodies induces astrocyte cell death in these cultures. However, although we cannot completely rule out a role of anti-GluR3 or other antibodies,36 it is unlikely that antibodies are largely involved in astrocytic damage in vivo. Our results show that none of our patients, despite a disturbed blood– brain barrier and severe inflammatory response, shows IgG or complement C9neo deposition on astrocytes. Moreover, all 4 of the 20 patients tested previously for the presence of serum GluR3 antibodies were negative for these antibodies25 but showed unequivocal astrocytic loss. This is corroborated by publications24,37 that question the role of anti-GluR3 antibodies in RE because these antibodies were detected in only some of the RE cases and have been found at an equal frequency in noninflammatory epileptic disorders. Another possible explanation for astrocytic damage could be an indirect phenomenon, namely, the lack of production of “astrotrophic” factors by neurons. We have tried to find evidence for this by correlating neu- Fig 4. Astrocytic degeneration and loss in white matter. (A) Double staining for glial fibrillary acidic protein (GFAP; blue) and caspase-3 (brown) showing an area with loss of GFAP-reactive astrocytes (arrowhead indicates small vessel). Original magnification ⫻75. (B) Same area as in (A) stained for CNPase, showing only a mildly reduced staining intensity, indicative of well-preserved oligodendrocytes and myelin. Original magnification ⫻75. (C–F) Double staining for caspase-3 (brown/red) and GFAP (blue). Two examples of caspase-3⫹ cells in and on the border of the lesion as seen in (A). (C) This caspase-3–positive cell still shows some GFAP immunoreactivity in the cytoplasm (arrowhead). Nuclear counterstain (hematoxylin) shows a condensed nucleus (arrow), another indication for apoptosis. Original magnification ⫻990. (D) In this caspase-3–positive cell, GFAP reactivity is absent. Nuclear counterstain (arrow) shows fragmentation of the nucleus. Original magnification ⫻990. (E, F). Two examples of caspase-3 and GFAP double-labeled apoptotic astrocytes. GFAP reactivity (arrowhead, blue) is seen only in part of the cytoplasm of the astrocytes. Original magnification ⫻792 (E), ⫻673 (F). (G) Terminal deoxynucleotidyltransferase–mediated biotin-dUTP nick end labeling (TUNEL) staining (black) in combination with GFAP (red) shows two GFAP-positive astrocytes with DNA fragmentation in the nucleus (arrows), another indication of apoptotic cell death. Original magnification ⫻330 (G). (H–L) Various stainings in the white matter of a lesion with GFAP loss. (H) Double staining for GFAP (blue) and caspase-3 (brown) shows the presence of multiple apoptotic astrocytes (arrowheads) in this white matter lesion. Original magnification ⫻890. (I) Bielschowsky stain for axons shows the absence of acute axonal damage. Original magnification ⫻200. (J) Double labeling for GFAP (brown, arrows point out astrocytic processes) and amyloid precursor protein (blue) shows the loss of astrocytic cell bodies. Except for a single axonal spheroid (arrowhead), no axonal pathology is evident. Original magnification ⫻277. (K) Staining for immunoglobulins shows some leakage around blood vessels without binding to astrocytes. Original magnification ⫻ 185. (L) Staining for C9neo shows a diffuse reactivity around the same blood vessel as in (K), but no deposition on astrocytes. Original magnification ⫻185. (M–R) Inflammation and astrocyte damage. (M) Staining for GFAP (red) and caspase-3 (green). An astrocyte in the center shows staining for caspase-3 in the absence of GFAP. Two other GFAP⫹ astrocytes show some caspase-3 reactivity in the cytoplasm. Original magnification ⫻675. (N–P) Triple staining for GFAP (red), CD3 (green), and granzyme-B (GrB; red granules inside of lymphocytes). (N) A lesion with two astrocytes (arrowheads) showing swollen cell bodies and degeneration of processes. Original magnification ⫻720. In the surrounding area, several lymphocytes are seen in close contact with astrocyte processes. A lymphocyte in apposition to an astrocyte (arrow), which is further enlarged in (O) (original magnification ⫻3,600), showing GrB⫹-granules polarized toward the astrocyte process. (P) Another example of a lymphocyte with polarized GrB⫹ granules facing an astrocyte process. Original magnification ⫻900. (Q) Multiple lymphocytes (green) are attached to a degenerating perivascular astrocyte (red). In the top left corner a normal astrocyte is seen. Original magnification ⫻720. (R) Double staining for major histocompatibility complex (MHC) class I (green) and GFAP (red) shows a neuron (left from the astrocyte) and blood vessels strongly stained for MHC class I. Astrocytic cell bodies and processes (arrowheads) also stain positive for MHC class I, although less strongly than neurons and endothelial cells. Original magnification ⫻1,350. Bauer et al: Astrocytic Degeneration in RE 77 Fig 5. Numbers of glial fibrillary acidic protein–positive (GFAP⫹; hatched bars), S100␤⫹ (black bars), and neuronal nuclei– positive (NeuN⫹; white bars) cells in defined cortical areas. GFAP⫹, S100␤⫹, and NeuN⫹ cells were counted in the cortex of healthy control subjects and in the three different regions (I: normal numbers of GFAP⫹ cells; II: areas with loss of GFAP⫹ cells; III: areas with GFAP⫹ gliotic cells) found in Rasmussen’s encephalitis (RE) patients. In the RE “GFAP normal” group, the numbers of GFAP⫹ and S100␤⫹ cells do not differ from the GFAP⫹ and S100␤⫹ cells in control subjects. The number of NeuN⫹ neurons, however, is significantly lower (*1p ⫽ 0.004). In the RE “GFAP loss” group, all three parameters are lower than in the control group (*2p ⫽ 0.01; *3p ⫽ 0.037; *4p ⫽ 0.004). The number of neurons do not differ from those in the RE “GFAP normal” group. In the RE “GFAP gliosis” group, the numbers of GFAP⫹ (*5p ⫽ 0.011) and S100␤⫹ (*6p ⫽ 0.011) cells are significantly greater than in control subjects, whereas the NeuN⫹ cells are significantly lower than in control subjects (*7p ⫽ 0.011), lower than in the RE “GFAP normal” group (p ⫽ 0.006), and lower than in the RE “GFAP loss” group (p ⫽ 0.036). The pictures on top show representative GFAP stainings for these different groups (from left to right: control subjects, RE “GFAP normal,” RE “GFAP loss,” and RE “GFAP gliosis”). ronal and astrocytic loss in cortical areas. Indeed, our results show that astrocytic loss and neuronal loss can be found in identical areas and to a similar degree. In addition, however, there are areas with sole loss of neurons and, more important, also areas of astrocytic apoptosis and loss in white matter. Axonal stainings indicated that in the white matter astrocytic lesions, axonal damage was absent, and thus loss of “astrotrophic” support by neurons may not be responsible for astrocytic apoptosis and loss. Another option for astrocytic loss can be specific damage by cytotoxic T lymphocytes. Recent findings by us25 show that cytotoxic T lymphocytes filled with GrB⫹-positive granules are found in close apposition to MHC class I⫹ neurons. Here we show that in areas with degradation of astrocytes, such interactions between T lymphocytes and astrocytes are readily found. As in the interaction of lymphocytes with neurons and as shown in in vitro studies,38 we found polarization of cytotoxic granules toward the astrocytic surface. In our samples, MHC class I molecules were expressed in and on astrocytes. Furthermore, MHC class I localization on astrocytes in RE has 78 Annals of Neurology Vol 62 No 1 July 2007 been described previously by immunohistochemical staining for the associated membrane protein ␤2microglobulin.39 As Christinck and colleagues40 have shown, only minimal expression of MHC class I antigens is needed for an interaction between a cytotoxic T cell and its target cell. Thus, induction of apoptosis in astrocytes probably is generated by a GrB-mediated, MHC class I–restricted T-cell response. A further argument for a role of cytotoxic T cells in the pathogenesis of RE comes from our tacrolimus (a T-cell–specific immunosuppressant) studies.41 Eight patients included in this study have received long-term treatment with tacrolimus. Six of eight patients have not or only mildly progressed by motor deterioration and hemispheric cerebral tissue loss. The other two patients progressed and finally underwent hemispherectomy. Five of these eight patients have been presented previously with a shorter follow-up.41 The underlying reason for T-cell cytotoxicity toward astrocytes and neurons, however, remains to be elucidated. Although it may be an autoimmune condition, cytotoxicity against a slowly spreading local viral infection might even better ex- plain the unilateral, centrifugal degeneration present in RE.42 Many different viruses can cause alterations of cytoskeletal structures.43 Moreover, in vitro infection of astrocytoma cell lines with measles virus specifically disrupts the GFAP cytoskeleton.44 Therefore, it is not inconceivable that an as yet unknown virus may contribute to the loss of GFAP in RE. Destruction could be explained as an effect of the infection itself, as well as by a cytotoxic T-cell response to the infected cells (ie, astrocytes and neurons). Similar conclusions were drawn in an earlier immunohistochemical study by Farrell and colleagues,39 who also mentioned loss of GFAP immunoreactivity in cortical areas of RE but did not further pursue this issue. Astrocytes conduct a large number of functions.45– 48 Pathophysiologically important in RE may be the roles of astrocytes in energy metabolism, in maintaining potassium homeostasis, and in the catabolism of GABA and glutamate, neurotransmitters critically involved in epileptic processes. Hansson and coworkers’ studies,49 for instance, suggest that glial glutamate transporters are essential for maintaining low extracellular glutamate levels, as well as for preventing chronic glutamate neurotoxicity. In some forms of human epilepsy, impaired potassium buffering by astrocytes may contribute to seizure generation or perpetuation.50 Finally, several reports indicate that hypertrophic astrocytes can aggravate or induce seizures by glutamate release51 or by upregulation of adenosine kinase, a key regulator of the anticonvulsant adenosine.52 Because both dying and activated hypertrophic astrocytes are present in RE, it is impossible to define in which state these astrocytes influence neuronal activity most. Either way, astrocytic pathology in RE may be considered to play a role in neuronal dysfunction and may contribute both to the induction of seizures and to the progressive deterioration of functions associated with the affected hemisphere. This study was supported by the Austrian “Fonds zur Förderung der wissenschaftlichen Forschung” (P16063-B02). We thank H. Breitschopf, U. Köck, A. Kury, and M. Leisser for expert technical assistance. References 1. Rasmussen T, Olszewski J, Lloyd-Smith D. Focal seizures due to chronic localized encephalitis. Neurology 1958;8:435– 445. 2. Bien CG, Granata T, Antozzi C, et al. Pathogenesis, diagnosis and treatment of Rasmussen encephalitis: a European consensus statement. Brain 2005;128:454 – 471. 3. Leach JP, Chadwick DW, Miles JB, et al. Improvement in adult-onset Rasmussen’s encephalitis with long-term immunomodulatory therapy. Neurology 1999;52:738 –742. 4. McLachlan RS, Girvin JP, Blume WT, et al. Rasmussen’s chronic encephalitis in adults. Arch Neurol 1993;50:269 –274. 5. Vadlamudi L, Galton CJ, Jeavons SJ, et al. Rasmussen’s syndrome in a 54 year old female: more support for an adult variant. J Clin Neurosci 2000;7:154 –156. 6. Koshewnikow AJ. Eine besondere Form von corticaler Epilepsie. Neurol Centralbl 1895;14:47– 48. 7. Oguni H, Andermann F, Rasmussen TB. The natural history of the syndrome of chronic encephalitis and epilepsy: a study of the MNI series of forty-eight cases. In: Andermann F, ed. Chronic encephalitis and epilepsy. Rasmussen’s syndrome. Boston: Butterworth-Heinemann, 1991:7–35. 8. Bien CG, Widman G, Urbach H, et al. The natural history of Rasmussen’s encephalitis. Brain 2002;125:1751–1759. 9. Robitaille Y. Neuropathologic aspects of chronic encephalitis. In: Andermann F, ed. Chronic encephalitis and epilepsy. Rasmussen’s syndrome. Boston: Butterworth-Heinemann, 1991: 79 –110. 10. Vinters HV, Wang R, Wiley CA. Herpesviruses in chronic encephalitis associated with intractable childhood epilepsy. Hum Pathol 1993;24:871– 879. 11. Power C, Poland SD, Blume WT, et al. Cytomegalovirus and Rasmussen’s encephalitis. Lancet 1990;336:1282–1284. 12. Jay V, Becker LE, Blaser S, et al. Pathology of chronic herpes infection associated with seizure disorder: a report of two cases with tissue detection of herpes simplex virus 1 by the polymerase chain reaction. Pediatr Pathol Lab Med 1995;15: 131–146. 13. McLachlan RS, Levin S, Blume WT. Treatment of Rasmussen’s syndrome with ganciclovir. Neurology 1996;47:925–928. 14. Iannetti P, Nigro G, Imperato C. Cytomegalovirus encephalitis and ganciclovir. Lancet 1991;337:373. 15. Walter GF, Renella RR, Hori A, et al. Nachweis von EpsteinBarr-Viren bei Rasmussen’s Enzephalitis. Bericht ueber zwei Faelle [Detection of Epstein-Barr viruses in Rasmussen’s encephalitis. Report of 2 cases]. Nervenarzt 1989;60:168 –170. 16. Rogers SW, Andrews PI, Gahring LC, et al. Autoantibodies to glutamate receptor GluR3 in Rasmussen’s encephalitis. Science 1994;265:648 – 651. 17. Andrews PI, Dichter MA, Berkovic SF, et al. Plasmapheresis in Rasmussen’s encephalitis. Neurology 1996;46:242–246. 18. Twyman RE, Gahring LC, Spiess J, et al. Glutamate receptor antibodies activate a subset of receptors and reveal an agonist binding site. Neuron 1995;14:755–762. 19. Levite M, Fleidervish IA, Schwarz A, et al. Autoantibodies to the glutamate receptor kill neurons via activation of the receptor ion channel. J Autoimmun 1999;13:61–72. 20. Whitney KD, Andrews JM, McNamara JO. Immunoglobulin G and complement immunoreactivity in the cerebral cortex of patients with Rasmussen’s encephalitis. Neurology 1999;53: 699 –708. 21. Whitney KD, McNamara JO. GluR3 autoantibodies destroy neural cells in a complement-dependent manner modulated by complement regulatory proteins. J Neurosci 2000;20: 7307–7316. 22. Wiendl H, Bien CG, Bernasconi P, et al. GluR3 antibodies: prevalence in focal epilepsy but no specificity for Rasmussen’s encephalitis. Neurology 2001;57:1511–1514. 23. Watson R, Jiang Y, Bermudez I, et al. Absence of antibodies to glutamate receptor type 3 (GluR3) in Rasmussen encephalitis. Neurology 2004;63:43–50. 24. Mantegazza R, Bernasconi P, Baggi F, et al. Antibodies against GluR3 peptides are not specific for Rasmussen’s encephalitis but are also present in epilepsy patients with severe, early onset disease and intractable seizures. J Neuroimmunol 2002;131: 179 –185. 25. Bien CG, Bauer J, Deckwerth TL, et al. Destruction of neurons by cytotoxic T cells: a new pathogenic mechanism in Rasmussen’s encephalitis. Ann Neurol 2002;51:311–318. Bauer et al: Astrocytic Degeneration in RE 79 26. Kang TC, Kim DS, Kwak SE, et al. Epileptogenic roles of astroglial death and regeneration in the dentate gyrus of experimental temporal lobe epilepsy. Glia 2006;54:258 –271. 27. Borges K, McDermott D, Irier H, et al. Degeneration and proliferation of astrocytes in the mouse dentate gyrus after pilocarpine-induced status epilepticus. Exp Neurol 2006;201: 416 – 427. 28. Revuelta M, Castano A, Machado A, et al. Kainate-induced zinc translocation from presynaptic terminals causes neuronal and astroglial cell death and mRNA loss of BDNF receptors in the hippocampal formation and amygdala. J Neurosci Res 2005;82:184 –195. 29. Bien CG, Granata T, Antozzi C, et al. Pathogenesis, diagnosis and treatment of Rasmussen encephalitis: a European consensus statement. Brain 2005;128:454 – 471. 30. Breitschopf H, Suchanek G, Gould RM, et al. In situ hybridization with digoxigenin-labeled probes: sensitive and reliable detection method applied to myelinating rat brain. Acta Neuropathol (Berl) 1992;84:6 –7. 31. Aboul-Enein F, Rauschka H, Kornek B, et al. Preferential loss of myelin-associated glycoprotein reflects hypoxia-like white matter damage in stroke and inflammatory brain diseases. J Neuropathol Exp Neurol 2003;62:25–33. 32. Hahn K, DeBiasio R, Tishon A, et al. Antigen presentation and cytotoxic T lymphocyte killing studied in individual, living cells. Virology 1994;201:330 –340. 33. Schmidt-Kastner R, Ingvar M. Loss of immunoreactivity for glial fibrillary acidic protein (GFAP) in astrocytes as a marker for profound tissue damage in substantia nigra and basal cortical areas after status epilepticus induced by pilocarpine in rat. Glia 1994;12:165–172. 34. Narkilahti S, Pirttila TJ, Lukasiuk K, et al. Expression and activation of caspase 3 following status epilepticus in the rat. Eur J Neurosci 2003;18:1486 –1496. 35. Brecht S, Gelderblom M, Srinivasan A, et al. Caspase-3 activation and DNA fragmentation in primary hippocampal neurons following glutamate excitotoxicity. Brain Res Mol Brain Res 2001;94:25–34. 36. Watson R, Jepson JE, Bermudez I, et al. Alpha7-acetylcholine receptor antibodies in two patients with Rasmussen encephalitis. Neurology 2005;65:1802–1804. 37. Wiendl H, Bien CG, Elger CE, et al. Anti-GluR3autoantibodies: prevalence in focal epilepsy but no specificity for Rasmussen’s encephalitis. Akt Neurol 2001;28(suppl 2): S135 (Abstract). 38. Hahn K, DeBiasio R, Tishon A, et al. Antigen presentation and cytotoxic T lymphocyte killing studied in individual, living cells. Virology 1994;201:330 –340. 80 Annals of Neurology Vol 62 No 1 July 2007 39. Farrell MA, Droogan O, Secor DL, et al. Chronic encephalitis associated with epilepsy: immunohistochemical and ultrastructural studies. Acta Neuropathol Berl 1995;89:313–321. 40. Christinck ER, Luscher MA, Barber BH, et al. Peptide binding to class I MHC on living cells and quantitation of complexes required for CTL lysis. Nature 1991;352:67–70. 41. Bien CG, Gleissner U, Sassen R, et al. An open study of tacrolimus therapy in Rasmussen encephalitis. Neurology 2004; 62:2106 –2109. 42. Bien CG, Urbach H, Deckert M, et al. Diagnosis and staging of Rasmussen’s encephalitis by serial MRI and histopathology. Neurology 2002;58:250 –257. 43. Cudmore S, Reckmann I, Way M. Viral manipulations of the actin cytoskeleton. Trends Microbiol 1997;5:142–148. 44. Duprex WP, McQuaid S, Rima BK. Measles virus-induced disruption of the glial-fibrillary-acidic protein cytoskeleton in an astrocytoma cell line (U-251). J Virol 2000;74: 3874 –3880. 45. Vernadakis A. Glia-neuron intercommunications and synaptic plasticity. Prog Neurobiol 1996;49:185–214. 46. Hertz L, Dringen R, Schousboe A, et al. Astrocytes: glutamate producers for neurons. J Neurosci Res 1999;57:417– 428. 47. Anderson CM, Swanson RA. Astrocyte glutamate transport: review of properties, regulation, and physiological functions. Glia 2000;32:1–14. 48. Dong Y, Benveniste EN. Immune function of astrocytes. Glia 2001;36:180 –190. 49. Hansson E, Muyderman H, Leonova J, et al. Astroglia and glutamate in physiology and pathology: aspects on glutamate transport, glutamate-induced cell swelling and gap-junction communication. Neurochem Int 2000;37:317–329. 50. Hinterkeuser S, Schroder W, Hager G, et al. Astrocytes in the hippocampus of patients with temporal lobe epilepsy display changes in potassium conductances. Eur J Neurosci 2000;12: 2087–2096. 51. Tian GF, Azmi H, Takano T, et al. An astrocytic basis of epilepsy. Nat Med 2005;11:973–981. 52. Fedele DE, Gouder N, Guttinger M, et al. Astrogliosis in epilepsy leads to overexpression of adenosine kinase, resulting in seizure aggravation. Brain 2005;128:2383–2395. 53. Dobersen MJ, Hammer JA, Noronha AB, et al. Generation and characterization of mouse monoclonal antibodies to the myelinassociated glycoprotein (MAG). Neurochem Res 1985;10: 499 –513. 54. Stam NJ, Vroom TM, Peters PJ, et al. HLA-A- and HLA-Bspecific monoclonal antibodies reactive with free heavy chains in Western blots, in formalin-fixed, paraffin-embedded tissue sections and in cryo-immuno-electron microscopy. Int Immunol 1990;2:113–125.