Background
Rasmussen encephalitis (RE) is an inflammatory neurodegenerative disease primarily seen in young children. It is clinically characterized by intense focal and generalized seizures with inflammation almost invariably confined to one cerebral hemisphere [
1‐
3]. Seizure frequency may decrease over time, but patients are left with unilateral hemiparesis and significant cognitive deficits [
4,
5]. Histopathologic examination of RE brain tissue, usually after years of seizures, reveals the presence of activated microglia and T lymphocytes [
6‐
8]. CD8
+ T cells containing granzymes are observed in close apposition to neurons and astrocytes, but not oligodendrocytes [
6,
9]. Auto-antibodies against the glutamate receptor GluR3, once considered a possible cause of RE [
10], are not present in all RE cases and are not specific to the disease [
11‐
13]. Circulating antibodies to other neuronal proteins are found in other RE cases [
14‐
16], leading to the view that a humoral immune response may be a secondary event to T cell immunity [
6,
17]. Thus, RE resembles an autoimmune disease although the histopathology is also consistent with an immune response to a virus or other infectious agent [
1,
17]. Cytomegalovirus, herpes simplex virus, and Epstein Barr virus sequences have been detected in some RE brain specimens, but have not been reproducibly identified in all cases [
18‐
22]. Neither autoimmunity nor infection can easily explain the unilateral hemispheric involvement.
To gain further insight into the immunopathology of RE, we measured the relative expression of 84 mRNA transcripts associated with inflammation and autoimmunity in brain tissue from 12 RE and 12 cortical dysplasia (CD) epilepsy patients. The CD cases constituted a reference group against which to compare gene expression levels. Inflammation is associated with CD, but is less severe than in RE, and T cell involvement is limited [
23,
24]. Quantitative differences in transcript levels between RE and CD specimens were found for several genes involved in the activation and recruitment of effector T cells. The highest levels of expression were found in early stage RE cases.
Methods
Cohort recruitment and clinical variables
Under University of California, Los Angeles, (UCLA) Institutional Review Board (IRB) approval, brain tissue was collected at surgery as part of UCLA’s Pediatric Epilepsy Surgery program. Informed consent to use the surgically resected tissue for research was obtained through the parents or legal guardians. The 12 RE and 12 CD cases used in the study were selected with similar ages at surgery. Seven CD cases were classified as CDI, and five as CDII [
25]. The clinical protocols for patient evaluation and surgical procedures for collection and processing of cortical specimens have been previously published [
26,
27]. Clinical variables included, age at seizure onset, age at surgery, and disease progression (age at seizure onset to age at surgery). Magnetic resonance imaging (MRI) scans were assessed by one investigator (GWM). A semiquantitative score was assigned to the T2 and FLAIR signal changes (0 = none, 1 = slight; 2 = mild; 3 = moderate; 4 = extensive) to estimate the degree of tissue destruction and inflammation, respectively.
Real-time PCR and analysis
Total RNA was purified from flash frozen blocks of involved tissue consisting of mostly cortical gray matter (approximately 50 mg) using Trizol™ (Life Technologies, Carlsbad, CA, USA) and reverse transcribed (Qiagen, Valencia, CA, USA). PCR reactions were carried out in an ABI 7300 thermocycler using SYBR™ green chemistry (SABiosciences, Valencia, CA, USA). The 96-well format qPCR array contained primers for 84 genes of interest and 5 reference genes (SABiosciences inflammation and autoimmunity qPCR array, cat no. PAHS-0077Z). Standard cycling parameters were used as follows: 1 cycle: 95°C 10 min, 40 cycles: 95°C 15 sec, 60°C 1 min. As part of the cycling program, all PCR products were thermally denatured and dissociation curves were obtained. Two of the primer sets resulted in dissociation curves with multiple peaks (CCL16 and CCL24) and were eliminated from the analysis. To calculate relative transcript levels, baseline-subtracted fluorescence values per cycle for each primer set were entered into LinRegPCR [
28] and PCR efficiencies (E) and Ct values were determined. The relative expression of each gene (X
0) [
29] in each array was calculated from:
NormalFinder [
30] was used to identify the least variable housekeeping gene across the 24 arrays.
ACTB encoding β-actin was found to be the most stable reference gene, and was used to normalize all of the data including the four other reference genes (
HPRT1,
RPL13A,
GAPDH,
B2M). Samples were grouped by consensus clustering [
31] using the non-negative matrix factorization algorithm [
32] found in the GENE-E bioinformatics package (
http://www.broadinstitute.org). Statistical analyses utilized R-project programs (
http://www.r-project.org). Plots were generated by the Deducer graphical user interface and exported into CorelDRAW X6 (Corel Corporation, Ottawa, ON, Canada). Significant differences (
P <0.05) in gene expression between RE and CD cases were determined by pairwise comparison of the distribution of transcript levels for each gene using Harrell-Davis quantile estimators [
33,
34].
Immunocytochemistry and image analysis
Paraffin-embedded blocks of involved tissue were serially sectioned (5 μm), deparaffinized, and microwaved for 20 minutes in buffered citrate (10 mM, pH 6.0) for antigen retrieval. After one hour in blocking solution (Impress Kit, Vector Laboratories, Burlingame, CA, USA) sections were incubated overnight at 4°C with rabbit anti-human CD4 (1:250, Novus Biologicals, Littleton, CO, USA) or mouse anti-human CD8 (1:100, Dako, Carpinteria, CA, USA). Sections were immunostained for one hour at 25°C with peroxidase-conjugated anti-rabbit or anti-mouse secondary antibodies (Impress Kit, Vector Laboratories) followed by incubation with 3,3′-diaminobenzidine (DAB) substrate (MP Biomedicals, Santa Ana, CA, USA), then counterstained with hematoxylin. Sections of tonsil tissue were used as positive controls, and omission of primary antibodies served as negative controls. Images of entire sections were acquired with an Aperio ScanScope XT scanner (Aperio, Vista, CA, USA) and transferred to CorelDRAWX6. Strong DAB staining of CD4 and CD8 immunoreactive cells was quantified using the positive pixel count algorithm, part of the Aperio ImageScope software package.
Western blotting
Blocks of flash frozen involved tissue were homogenized in RIPA buffer containing protease and phosphatase inhibitors (Sigma-Aldrich, St. Louis, MO, USA). Lysates were separated on precast 10% polyacrylamide gels (Biorad, Hercules, CA, USA) and transferred to PVDF membranes (Biorad). Prestained molecular standards were used (Biorad). The membrane was blocked in Tris-buffered saline (pH 7.4) containing 5% nonfat dried milk and 0.1% Tween™ 20, and probed with a monoclonal antibody to hypoxanthine-guanine phosphoribosyltransferase (anti-HPRT 1:1000, Proteintech Group Inc., Chicago, IL, USA). Proteins were visualized with a secondary antibody conjugated to horseradish peroxidase (1:2500, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) using a chemiluminescent substrate (Thermo-Scientific, Waltham, MA, USA). The blot was stripped and re-probed with a monoclonal antibody to glyceraldehyde 3-phosphate dehydrogenase (anti-GAPDH 1:1000, Stressgen, Victoria, BC, Canada). X-ray films were scanned and processed (background subtraction and enhancement using default settings) in Image J, and exported to CorelDRAWX6.
Discussion
We used a qPCR array to measure inflammatory gene expression in 12 RE specimens and 12 CD specimens, and then correlated the expression of genes that differed between the two groups with clinical parameters. Selection of the CD specimens from the UCLA Pediatric Surgery Program’s tissue bank was based on choosing patients whose ages at surgery were approximately the same as the much rarer RE cases that took approximately ten years to accrue. There were no other exclusion criteria. The array included genes encoding pro-inflammatory T helper 1 (Th1) and anti-inflammatory T helper 2 (Th2) cell cytokines, chemokines, Toll-like receptors and other factors involved in downstream signaling. The expression levels of only eight transcripts significantly differed between the RE and CD specimens. With the exception of HPRT mRNA, the transcripts were expressed at higher levels in the RE specimens. Reduced HPRT, a housekeeping gene on the array, was unexpected and may reflect disease-associated atrophy. Alternatively, a reduced level of HPRT in brain tissue may be significant since a decrease in this enzyme would be expected to affect purine metabolism, with possible effects on brain function [
36]. Analysis of HPRT protein levels in more RE samples will be necessary to establish the significance of this finding to the disease process.
In RE there is clear histopathologic evidence for the involvement of CD8
+ T cells in the disease process [
6‐
8,
23]. In agreement with these data, four of the seven genes whose mRNA levels were higher in RE compared with CD cases encode proteins involved in Th1-driven immune responses, namely IFN-γ, CCL5, CXCL9 and CXCL10. Activated CD8
+ cytotoxic T cells (Tc1) and CD4
+ Th1 cells produce IFN- γ [
37,
38], thus infiltrating CD8
+ T cells in neuropil and CD4
+ T cells in Virchow-Robin spaces and leptomeninges (Figure
7) could be sources of IFN-γ transcripts in the RE brain specimens. Although the immunopathology in RE appears to be driven by Tc1 cells, CD4
+ cells, in the perivascular space may play a role in sustaining Tc1 activity [
39‐
41]. Natural killer (NK) cells, NKT cells, and γδ T cells produce IFN-γ [
37,
42], which may also account for the IFN-γ transcripts in RE tissue. To date, we and others have found no evidence for significant numbers of NK cells in RE brain tissue [
7]. However, we have recently identified γδ T cells in brain infiltrating lymphocytes isolated from fresh RE brain tissue (unpublished data).
IFN-γ can induce major histocompatibility complex (MHC) class I molecules on the surface of neurons, rendering them vulnerable to attack by autoantigen-sensitized MHC class I-restricted Tc1 cells [
43]. Further, IFN-γ has been shown to induce bursting of hippocampal pyramidal neurons
in vitro[
44], providing a possible link between T cells and epileptogenesis.
It has been reported that IFN-γ induces the production of CXCL9 by microglia and CXCL10 by microglia and astrocytes [
45], and promotes IL-1-induced synthesis of CCL5 by astrocytes [
46]. The positive correlation between the relative amounts of CXCL9, CXCL10, CCL5 and IFN-γ mRNA that we observed among the RE specimens is consistent with these reports. All three chemokines have been implicated in attracting Th1, Tc1, γδ T cells and NK cells to sites of inflammation [
47]. The presence of CXCR3, the receptor for CXCL9 and CXCL10 [
48], and CCR5 (CCL5 receptor) on infiltrating T cells in brain sections from a single RE case has been documented [
49]. Other single patient studies have provided evidence for the expression of CCL5 and CXCL10 in RE brain tissue [
50,
51].
Quantitative PCR allowed correlations to be made between the amount of inflammatory gene transcripts and clinical variables. Notably, much higher levels of IFN-γ and CXCL9 mRNAs were detected in specimens from patients that had undergone surgery within shorter times from disease onset compared to later times. This suggests that there is a pronounced Th1 immune response in the early phase of the disease that declines after 1 to 2 years. The observation of large numbers of CD4
+ and CD8
+ lymphocytes in sections from specimens in which high levels of IFN-γ mRNA were detected is consonant with an initial Th1 polarized response. In support of these data, it was previously reported that cerebrospinal fluid levels of IFN-γ were higher in the earlier stages of RE [
52].
A role for the Fas/FasL-mediated cell death pathway in RE is indicated from the qPCR data. Based on the observation of granzyme B immunoreactivity in T cells in close apposition to neurons and astrocytes, it has been suggested that MHC class I-restricted killing occurs by the perforin lytic pathway [
6,
9]. However MHC class I-restricted killing of neurons can also occur by Fas ligand-induced apoptosis [
53]. On the other hand, the detection of FasL transcripts in the present study could be explained by activation-induced cell death of activated T cells in the brain [
54]. In experimental autoimmune encephalitis, astrocytes expressing FasL have been implicated in T cell homeostasis [
55].
In contrast to the chemokines associated with a Th1 response, CCL22, which binds CCR4 [
56], is associated with a Th2 polarized response, which can facilitate B cell activation [
38]. The finding of higher levels of this chemokine in RE brain tissue is therefore consistent with the presence of circulating antibodies to neuronal proteins in some RE patients [
10,
11,
14‐
16]. Unlike the Th1 cytokines, CCL22 expression did not strongly correlate with time from disease onset to surgery, supporting the notion that a B cell response may be a secondary consequence of tissue destruction mediated by Tc1 cells [
17]. CCL22 may also be involved in recruiting immunosuppressive T regulatory cells into the brain to modulate the Tc1 response [
57].
A positive correlation was observed between the extent of tissue destruction as measured by MRI and the level of CCL23 mRNA. CCL23 can act as a chemoattractant for monocytes [
58]. A role for monocytes in RE pathogenesis is therefore possible as is the case in viral encephalitis [
59]. Since monocytes also produce vascular endothelial growth factor [
60], this may explain the vascular changes that have been observed in resected RE but not CD brain tissue [
61].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
GCO designed study, performed qPCR, analyzed the data and drafted the manuscript; MNH performed immunocytochemistry; JWC performed western blot analysis; DLMcA carried out statistical analysis; MJH assisted with data interpretation and helped draft the manuscript; HVV provided tissue sections and helped draft the manuscript, GWM provided surgical specimens, evaluated MRI scans, and helped draft the manuscript, CAK, provided project oversight and helped draft the manuscript. All authors read and approved the final manuscript.