Gene Expression Analysis of Tuberous Sclerosis Complex Cortical Tubers Reveals Increased Expression of Adhesion and Inflammatory Factors

Abstract

Cortical tubers in patients with tuberous sclerosis complex are associated with disabling neurological manifestations, including intractable epilepsy. While these malformations are believed to result from the effects of TSC1 or TSC2 gene mutations, the molecular mechanisms leading to tuber formation, as well as the onset of seizures, remain largely unknown. We used the Affymetrix Gene Chip platform to provide the first genome‐wide investigation of gene expression in surgically resected tubers, compared with histological normal perituberal tissue from the same patients or autopsy control tissue. We identified 2501 differentially expressed genes in cortical tubers compared with autopsy controls. Expression of genes associated with cell adhesion, for example, VCAM1, integrins and CD44, or with the inflammatory response, including complement factors, serpinA3, CCL2 and several cytokines, was increased in cortical tubers, whereas genes related to synaptic transmission, for example, the glial glutamate transporter GLT‐1, and voltage‐gated channel activity, exhibited lower expression. Gene expression in perituberal cortex was distinct from autopsy control cortex suggesting that even in the absence of tissue pathology the transcriptome is altered in TSC. Changes in gene expression yield insights into new candidate genes that may contribute to tuber formation or seizure onset, representing new targets for potential therapeutic development.

INTRODUCTION

Tuberous sclerosis complex (TSC) is an autosomal dominant disorder resulting from mutations in either the TSC1 (encoding TSC1) or TSC2 (TSC2) gene 22, 77. In addition to autism and cognitive disabilities, epilepsy is the most common neurological symptom of TSC, and is present in 70%–80% of individuals with TSC [reviewed in Crino et al (18)]. The neurological features of TSC are highly associated with cortical tubers, which are developmental cortical malformations histologically characterized by disordered cortical lamination, astrogliosis, dysplastic neurons and the presence of so‐called “giant cells”(56). Epilepsy is, in most cases, difficult to manage with antiepileptic drugs and often requires surgical resection of one or more tubers (35).

The formation of cortical tubers during brain development has been intensively studied and is likely linked to activation of the phosphatidylinositol‐3 kinase‐mammalian target of rapamycin (Pi3K‐mTOR) signaling pathway (41) in the setting of TSC1 or TSC2 mutations. The Pi3K‐mTOR signaling pathway is critically involved in cell growth and proliferation. Activated downstream mediators of Pi3K‐mTOR, including phosphorylated (p) ribosomal protein S6, p‐4E‐BP1, and p‐eIF4G (55), likely explains cytomegaly in tubers. Additionally, via interaction with ezrin‐radixin‐moesin (ERM), regulatory proteins of neuronal migration (37), TSC proteins might also be involved in the aberrant positioning of neurons in tubers.

Epileptogenesis in cortical tubers remains poorly understood. Altered expression of specific GABA_A receptor subunits and glutamate receptors 13, 73, 86 may contribute to hyperexcitability and seizure generation in cortical tubers. Altered expression of proinflammatory cytokines and components of both the innate and adaptive immune system has been identified in tuber specimens, suggesting a possible role for the inflammatory response in generating seizures (11), as postulated for other epilepsy subtypes.

Previous gene array studies in tuber specimens focused either on specific cell types [nestin‐immunoreactive cells; (19)] or on targeted selected cDNA sequences 42, 86, but did not approach the entire transcriptome. Thus, we analyzed transcriptional changes across the genome using oligonucleotide arrays in tuber homogenates from genotyped TSC patients. Gene sets were classified using the Gene Ontology (GO) terms (6) to understand the biological meaning of changes in gene expression levels. The major aim of our study was to identify differentially regulated genes and pathways in cortical tubers compared to control cortex to improve our understanding of the formation of cortical tubers and the underlying epileptogenic mechanisms. Gene expression analysis was also performed in perituberal tissue to gain insight in the still debated contribution of this area to the epileptogenicity of TSC cortical lesions.

MATERIALS AND METHODS

Human specimens

For the microarray analysis, cortical tuber specimens were obtained during epilepsy surgery from four TSC patients with intractable epilepsy fulfilling the diagnostic criteria for TSC (27). A TSC1 mutation was defined in two cases (cortical tuber (CT)2 and CT3) and a TSC2 mutation in the other two cases (CT1 and CT4; Table 1). In two patients, a significant amount of perituberal tissue (PT; non‐lesional tissue adjacent to the cortical tuber; histologically normal, not containing dysplastic neurons or giant cells; Table 1) was also resected and frozen material was available. Resection was guided by intra‐operative ECoG and the perituberal tissue was part of the epileptogenic region. This tissue provides an important reference tissue, because it is exposed to the same antiepileptic medications as tubers, and age and gender are the same. Histologically normal cortex obtained at autopsy from four controls without a history of seizures or other neurological diseases was also analyzed (Table 1; samples AC1‐AC4; cause of death was acute cardiorespiratory failure for each). All the specimens used for the array were carefully inspected by microscopy prior to mRNA extraction using both histological and immunocytochemical stainings (Hematoxylin/Eosin (HE), luxol‐PAS, glial fibrillary acidic protein (GFAP), vimentin, neurofilament, neuronal nuclear protein (NeuN) and phosphorylated ribosomal S6 potein) and matched for equal amounts of gray and white matter, using a microdissection approach. Representative examples of cortical tuber, perituberal and control cortical specimens are shown in Figure 1. For the validation of the microarray results, we included six additional surgically resected cortical tuber specimens (CT5‐CT10), two surgically resected perituberal specimens (PT3‐PT4) and six autopsy control specimens (AC5‐AC10; Table 1). Tissue was obtained from the Department of Neuropathology of the Academic Medical Center (AMC), University of Amsterdam, Amsterdam, The Netherlands; the Department of Pathology of the University Medical Center, Utrecht, The Netherlands; and the PENN Epilepsy Center, Department of Neurology of the University of Pennsylvania Medical Center, Philadelphia, PA, USA. Informed consent was obtained for the use of brain tissue and for access to medical records for research purposes. All samples were obtained and used in a manner compliant with the Declaration of Helsinki and the University of Pennsylvania Institutional Review Board and Committee on Human Research.

Table 1.

Summary of clinical characteristics of human specimens. Abbreviations: CT = cortical tuber; PT = perituberal; AC = autopsy control; CPS = complex partial seizures; GTCS = generalized tonic‐clonic seizures; MS = multiple seizures.

Patient/Sex/Age (years)	Sample ID	Mutation	Location	Duration of epilepsy (years)	Seizure type
1/M/13 a , b	CT1	TSC2	Temporal	13	CPS
2/F/17 a , b	CT2	TSC1	Temporal	17	CPS
3/F/18 a , b	CT3	TSC1	Temporal	18	CPS
4/F/14 a , b	CT4	TSC2	Temporal	14	CPS
5/M/3 b	CT5	TSC2	Parietal	3	CPS, GTCS
6/M/11 b	CT6	TSC1	Frontal	11	CPS, MS
7/M/0.6 c	CT7	TSC2	Occipital	0.6	CPS, GTCS
8/F/22 c	CT8	TSC1	Frontal	21	CPS, GTCS
9/M/36 c	CT9	TSC2	Temporal	20	CPS
10/F/1 c	CT10	TSC2	Temporal	1	CPS
11/M/13 a , b	PT1	TSC2	Temporal	13	CPS
12/F/17 a , b	PT2	TSC1	Temporal	17	CPS
13/F/6 b	PT3	TSC1	Temporal	4	CPS
14/F/4 b	PT4	TSC2	Temporal	3	CPS, GTCS
15/F/18 a , b	AC1		Temporal
16/M/48 a , b , c	AC2		Temporal
17/M/31 a , b	AC3		Temporal
18/M/57 a , b	AC4		Temporal
19/M/62 b	AC5		Temporal
20/F/8 b	AC6		Frontal
21/M/16 b	AC7		Parietal
22/F/43 c	AC8		Temporal
23/M/67 c	AC9		Temporal
24/M/9 c	AC10		Temporal

^aSpecimens used for : microarray analysis,

^b b: real‐time PCR and

^c c: immunocytochemistry.

Figure 1.

Histopathological features of TSC—cortical tubers. A–B. Hematoxylin/Eosin (HE) staining; representative photomicrographs of cortical tubers (TSC) showing an area of cortical dislamination, containing different cell types, such as dysplastic neurons (arrows in B), reactive astrocytes (arrowheads in B) and giant cells (asterisks in B). C–E. NeuN staining showing the disorganization of the neuronal component within the cortical tuber (C) compared to perituberal (D) and control (E) cortical specimens. Scale bar in A: A, C–E: 250 µm; B, 27 µm.

RNA isolation and Affymetrix GeneChip processing

Equal amounts of gray and white matter from tissue specimens was used for RNA isolation. The sample quality was evaluated according to several recently suggested tissue quality markers, including the RNA integrity number (RIN) (71). We only included samples with RIN around 7 (TSC: mean 7.3, range 6.6–7.9; perituberal: 6.9 and 7.3; autopsy control: mean: 8.0, range 7.0–8.6). As expected on the basis of previous studies 53, 71, some variability in the RIN value was present in both control autopsy and surgical specimens.

Brain tissue was homogenized and total RNA was isolated using TRIzol LS Reagent (Invitrogen–Life Technologies, Breda, The Netherlands). The RNA concentration was measured on a nanodrop ND‐100 (NanoDrop Technologies Inc., Wilmington, DE, USA) and RNA integrity was checked on an Agilent 2100 BioAnalyzer (Agilent Technologies, Palo Alto, CA, USA). Total RNA was labeled with the GeneChip Expression 3′ amplification one‐cycle target labeling system. Each labeled sample (10 µg) was hybridized to an Affymetrix GeneChip (U133 plus 2.0 array) for 16 h. After hybridization, the GeneChips were washed and stained on a fluidics station (Affymetrix) and scanned in a confocal scanner (Affymetrix GeneArray Scanner) according to the Affymetrix GeneChip Expression Analysis Manual. The U133 plus 2.0 array (Affymetrix) comprises over 54 000 probe sets and 38 500 well‐characterized human genes.

Microarray data and statistical analysis

Data transformation (log2 conversion), selection and statistical analyses were performed with either Microsoft Excel 9.0 or MATLAB 7.5. All statistical tests were performed on the log‐transformed intensities, using a combination of Excel and SigmaStat (SPSS, v2). The microarray data were normalized using Robust Multi‐Array Average analysis. The Affymetrix analysis (MAS 5.0) provides a P‐value for the presence of each gene on each chip, and genes that were considered absent were eliminated (26). A gene was considered present when P < 0.05 held for all samples in one group (cortical tuber or autopsy control) or no more than one sample had a P > 0.07 or no more than two samples had 0.05 < P < 0.07. In the perituberal group the gene was considered present when P < 0.05 held for both samples, or only one sample had 0.05 < P < 0.07. Genes present in one group but not in another group were included in the further analysis. In this way, inducible or strongly repressed genes were not overlooked. We linked the different genes to GO terms with R and statistical packages obtained at http://www.bioconductor.org (annotation data for Affymetrix Human Expression Set: hgu133a2(cdf); GO, which is a data package containing annotation data for GO; and GOstats which are tools for manipulating GO and microarray data). Fold changes in gene expression were calculated by dividing the mean signal intensity from the four cortical tuber samples by the mean signal intensity from the four autopsy control samples. Perituberal material was only available from two patients, and for these two patients, we calculated the fold changes in gene expression by dividing the mean signal intensity from the perituberal specimens by the mean signal intensity from the autopsy control specimens. In addition, we compared the mean signal intensity from the TSC2 mutated cortical tuber specimens with the signal intensity from the TSC1 mutated cortical tuber specimens. Changes in gene expression required at least P < 0.01 and a twofold change to be considered as statistically significant (Student's t‐test).

To obtain a global impression of the significantly altered genes, an unsupervised hierarchical clustering analysis was performed using Spotfire® Decision Site for Functional Genomics program, MATLAB 7.5 and Kaleidograph 4.0. All significantly altered genes were entered in Database for Annotation, Visualization and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov/), which provides several tools to help interpret the biological meaning of genome‐scale data sets. With the Functional Annotation tool, all genes were classified using the GO system as biological process, molecular function and/or cellular component (6). In addition, DAVID provides an overview of pathways containing the differentially expressed genes, including the number of differentially expressed genes per pathway and a P‐value representing gene enrichment. This gene enrichment score determines whether the selected genes are associated with a signaling pathway specifically or by random chance. P < 0.05 is, in general, used to represent strong enrichment.

Microarray results were validated with quantitative real‐time polymerase chain reaction (PCR) and immunocytochemistry (see Supplement). We selected genes that belong to the most prominently over‐ or under‐expressed processes.

RESULTS

Of the 54 675 probe sets, 29 132 (53%) were assigned as “present” on the array. There were 2501 differentially regulated genes, of which 1147 had an increased expression level and 1354 had a decreased expression level in tubers, compared to autopsy controls (P < 0.01 and a minimal twofold change). Hierarchical clustering on this gene set revealed a differential expression profile for the cortical tuber specimens compared to the control specimens (Figure 2A). Gene expression profiles in the tuber specimens clustered together according to genotype (TSC1 or TSC2) (CT2 and CT3 carry a TSC1 mutation, and the other two specimens (CT1 and CT4), a TSC2 mutation; Figure 2A; Table 1). Hierarchical clustering on the set of genes (111 probes) differentially expressed in the perituberal specimens compared to autopsy controls demonstrated a clear and distinct separation between the two groups (Figure 2B), indicating that the perituberal tissue cannot be included in the control group. The red dots in the scatter diagram of Figure 2C represent the signal intensities (²log) of the 2501 genes that were significantly different in the TSC profile compared to the autopsy control profile (twofold difference and P < 0.01). All other genes that were present are depicted in blue. Figure 2D shows the scatter of genes as function of the P‐value that have at least twofold change in the cortical tubers compared to the autopsy control specimens. For comparison, we show that there is a considerable smaller number of significantly altered genes present in perituberal specimens (compared to the autopsy controls) than in TSC specimens (Figure 2E).

Figure 2.

Gene expression profiles. A. Hierarchical clustering on the set of genes (2501) that are significantly (and twofold) different between control and TSC specimens demonstrates a nice separation between the cortical tuber (CT1‐CT4) specimens and autopsy controls (AC). The dendrogram at the top displays the relationship of the samples based on their pattern of gene expression. Each column represents the gene expression profile on one specific array with each row representing the expression level of a specific gene. The expression levels are depicted in three colors; blue: low expression, white: medium expression and red: high expression; deeper shade indicates larger difference. B. Hierarchical clustering on the set of genes (111) that were significantly (and twofold) different between controls and perituberal tissue. The hierarchical clustering on the set of genes that were significantly (and twofold) different shows nice separation between the two groups, indicating that the samples are more similar within each condition. C. Scatter plot showing the correlation between average signal intensities in TSC and autopsy control samples. Data points represent average signal intensity (²log; n = 4 for each condition) calculated from Affymetrix signal intensities for all present genes expressed in the cortical tuber (TSC) and the autopsy control specimens. Red dots represent the significantly changed genes with at least a twofold change in gene expression and a P‐value < 0.01. Blue dots represent the genes that do not fulfill this criterion. D. Scatter diagram of the fold change in the cortical tuber specimens (compared to the autopsy controls) as a function of the P‐value. E. Scatter diagram of the fold change in the perituberal specimens (compared to the autopsy control specimens) as a function of P‐value. The pink dots represent the genes associated with cell adhesion or the immune/inflammatory response.

Altered gene expression was defined in three comparison groups: cortical tubers vs. autopsy control tissue, perituberal tissue vs. autopsy control tissue and tubers removed from patients with TSC1 vs. TSC2 genotypes. All microarray data will be available at the GEO website (accession GSE16969; http://www.ncbi.nlm.nih.gov/geo/).

Gene ontology, pathway and expression analysis

Cortical tuber specimens (TSC) vs. autopsy control specimens

First, we selected the “present” genes with at least a fourfold change in expression level, classified them according to their P‐value and selected the top 100 significant hits. These 100 genes were classified according to their biological process consistent with the GO system (Table 2). For example, a considerable and disproportionate number of genes related to cell adhesion and the immune/inflammatory response were altered in tubers compared with autopsy control specimens (depicted as pink dots in Figure 2D).

Table 2.

The 100 most significantly differentially expressed genes between cortical tubers (TSC) and autopsy controls with at least a fourfold change in expression level.

Gene symbol	Description	Fold change	P‐value
Cell adhesion
CD44	CD44 molecule (Indian blood group)	94.05	0.00001
TNC	Tenascin C (hexabrachion)	75.22	0.00000
AEBP1	AE binding protein 1	19.16	0.00000
ECM2	Extracellular matrix protein 2, female organ and adipocyte specific	12.89	0.00008
VCAM1	Vascular cell adhesion molecule 1	10.30	0.00002
ITGB4	Integrin, beta 4	9.19	0.00003
RAB13	RAB13, member RAS oncogene family	8.44	0.00002
SSPN	Sarcospan (Kras oncogene‐associated gene)	6.06	0.00007
SRPX	Sushi‐repeat‐containing protein, X‐linked	5.38	0.00017
ERBB2IP	Erbb2 interacting protein	4.03	0.00008
Cell cycle, cell growth and differentiation
ANGPT1	Angiopoietin 1	19.75	0.00000
EMP1	Epithelial membrane protein 1	13.63	0.00016
RARRES3	Retinoic acid receptor responder (tazarotene induced) 3	11.92	0.00002
PLCE1	Phospholipase C, epsilon 1	6.12	0.00004
PARD3B	Par‐3 partitioning defective 3 homolog B (C. elegans)	5.65	0.00015
FH	Fumarate hydratase	0.20	0.00009
Cell death
ATG5	ATG5 autophagy related 5 homolog (S. cerevisiae)	0.25	0.00002
API5	Apoptosis inhibitor 5	0.24	0.00003
NCKAP1	NCK‐associated protein 1	0.15	0.00001
Intermediate filament‐based process
GFAP	Glial fibrillary acidic protein	12.87	0.00020
VIM	Vimentin	5.71	0.00003
Immune/Inflammatory response
SERPINA3	Serpin peptidase inhibitor, clade A (alpha‐1 antiproteinase, antitrypsin), member 3	163.03	0.00000
CFI	Complement factor I	29.72	0.00002
C4A	Complement component 4A (Rodgers blood group)	29.49	0.00001
C3	Complement component 3	14.85	0.00001
ANXA1	Annexin A1	14.29	0.00020
COLEC12	Collectin sub‐family member 12	10.42	0.00004
GBP1	Guanylate binding protein 1, interferon‐inducible, 67 kDa	5.81	0.00017
EDG3	Endothelial differentiation, sphingolipid G‐protein‐coupled receptor, 3	5.72	0.00004
BCL6	B‐cell CLL/lymphoma 6 (zinc finger protein 51)	5.53	0.00004
MASP1	Mannan‐binding lectin serine peptidase 1 (C4/C2 activating component of Ra‐reactive factor)	5.18	0.00006
Nervous system and skeletal development
DCAMKL1	Doublecortin and CaM kinase‐like 1	9.93	0.00002
ANXA2	Annexin A2	9.58	0.00012
AHNAK	AHNAK nucleoprotein (desmoyokin)	4.05	0.00006
PPP1R9A	Protein phosphatase 1, regulatory (inhibitor) subunit 9A	0.24	0.00010
CHRM3	Cholinergic receptor, muscarinic 3	0.12	0.00019
Protein processing
CHI3L2	Chitinase 3‐like 2	60.55	0.00001
CHST6	Carbohydrate (N‐acetylglucosamine 6‐O) sulfotransferase 6	11.61	0.00015
GMPR	Guanosine monophosphate reductase	8.81	0.00003
GYG2	Glycogenin 2	8.58	0.00013
LIMK2	LIM domain kinase 2	7.80	0.00017
SPARC	Secreted protein, acidic, cysteine‐rich (osteonectin)	6.05	0.00001
PALLD	Palladin, cytoskeletal associated protein	4.99	0.00003
HSPB8	Heat shock 22 kDa protein 8	4.02	0.00011
DNAJB4	DnaJ (Hsp40) homolog, subfamily B, member 4	0.25	0.00011
ETNK1	ethanolamine kinase 1	0.24	0.00001
DNAJC12	DnaJ (Hsp40) homolog, subfamily C, member 12	0.22	0.00013
KLK7	Kallikrein‐related peptidase 7	0.20	0.00004
PTPN5	Protein tyrosine phosphatase, non‐receptor type 5 (striatum‐enriched)	0.16	0.00018
DNAJA4	DnaJ (Hsp40) homolog, subfamily A, member 4	0.08	0.00003
Replication, transcription and translation
WWTR1	WW domain containing transcription regulator 1	20.23	0.00000
RNASE4	Ribonuclease, RNase A family, 4	9.16	0.00001
PRRX1	Paired related homeobox 1	8.50	0.00008
MAML2	Mastermind‐like 2 (Drosophila)	5.30	0.00020
RFX4	Regulatory factor X, 4 (influences HLA class II expression)	5.02	0.00001
ZNF702	Zinc finger protein 702	0.25	0.00001
ORC5L	Origin recognition complex, subunit 5‐like (yeast)	0.24	0.00000
TSC22D2	TSC22 domain family, member 2	0.20	0.00020
GFM2	G elongation factor, mitochondrial 2	0.17	0.00002
Signal transduction
AGTRL1	Angiotensin II receptor‐like 1	32.22	0.00000
S100A10	S100 calcium binding protein A10	9.99	0.00013
FGG	Fibrinogen gamma chain	9.60	0.00009
DTNA	Dystrobrevin, alpha	8.37	0.00007
GNG12	Guanine nucleotide binding protein (G protein), gamma 12	4.40	0.00003
NMB	Neuromedin B	4.26	0.00018
SKAP2	Src kinase associated phosphoprotein 2	0.24	0.00008
ARL6	ADP‐ribosylation factor‐like 6	0.24	0.00016
NPY2R	Neuropeptide Y receptor Y2	0.23	0.00004
PTHR2	Parathyroid hormone receptor 2	0.22	0.00014
CNKSR2	Connector enhancer of kinase suppressor of Ras 2	0.12	0.00015
RXFP1	Relaxin/insulin‐like family peptide receptor 1	0.11	0.00018
Transport
CP	Ceruloplasmin (ferroxidase)	53.51	0.00000
CYP21A2	Cytochrome P450, family 21, subfamily A, polypeptide 2	16.93	0.00015
MATE2	H+/organic cation antiporter	12.38	0.00018
PDPN	Podoplanin	10.15	0.00010
MAOB	Monoamine oxidase B	5.86	0.00002
SLC9A9	Solute carrier family 9 (sodium/hydrogen exchanger), member 9	5.09	0.00007
BBOX1	Butyrobetaine (gamma), 2‐oxoglutarate dioxygenase (gamma‐butyrobetaine hydroxylase) 1	4.12	0.00013
FABP5	Fatty acid binding protein 5 (psoriasis‐associated)	4.02	0.00020
SLC4A4	Solute carrier family 4, sodium bicarbonate cotransporter, member 4	0.24	0.00003
SLC25A27	Solute carrier family 25, member 27	0.22	0.00003
KCNIP2	Kv channel interacting protein 2	0.18	0.00017
SLC1A2	Solute carrier family 1 (glial high affinity glutamate transporter), member 2	0.05	0.00002
Unknown biological process
GALNTL2	UDP‐N‐acetyl‐alpha‐D‐galactosamine : polypeptide N‐acetylgalactosaminyltransferase‐like 2	20.05	0.00002
hCG_1776018	hCG1776018	13.46	0.00006
ARRDC4	Arrestin domain containing 4	13.03	0.00000
CCDC80	Coiled‐coil domain containing 80	11.25	0.00011
CD109	CD109 molecule	6.91	0.00010
DKFZP434P211	POM121‐like protein	5.77	0.00003
ANTXR1	Anthrax toxin receptor 1	5.47	0.00001
TMEM16F	Transmembrane protein 16F	5.10	0.00018
SAMD4A	Sterile alpha motif domain containing 4A	4.24	0.00013
RELL1	Receptor expressed in lymphoid tissues like 1	4.03	0.00016
GHITM	Growth hormone inducible transmembrane protein	0.25	0.00003
C12orf29	Chromosome 12 open reading frame 29	0.22	0.00002
LGALS8	Lectin, galactoside‐binding, soluble, 8 (galectin 8)	0.22	0.00009
MUM1L1	Melanoma associated antigen (mutated) 1‐like 1	0.20	0.00002
LMBR1	Limb region 1 homolog (mouse)	0.20	0.00018
RY1	Putative nucleic acid binding protein RY‐1	0.20	0.00020
PPP3R1	Protein phosphatase 3 (formerly 2B), regulatory subunit B, alpha isoform	0.19	0.00002

Next, we used DAVID to determine groups of functionally related genes, again based on the GO system (P‐value < 0.01 and at least a twofold change in gene expression level). Of the 1147 genes with a higher expression level, 886 corresponding DAVID IDs were recognized, and of the 1354 genes with a lower expression level, 1151 DAVID IDs were recognized. One DAVID gene ID represents one unique gene cluster belonging to one single gene entry. The genes were classified in three different groups based on GO terms (individual genes can belong to different GO terms): (i) biological process; (ii) molecular function; and (iii) cellular components. Significantly changed GO terms (P < 0.01) relating to genes with increased expression levels are presented in Table S‐II, with a decreased expression in Table S‐III, and summarized as follows:

• (i)  Biological process: Genes with higher expression levels were associated with immune/inflammatory response (eg complement activation, innate and adaptive immune response, and regulation of immune response), cell adhesion, development and cell death (eg regulation of apoptosis). Lower expressed genes were associated with ubiquitination (eg ubiquitin cycle and protein ubiquitination), synaptic plasticity and glutamate transport.
• (ii)  Molecular function: genes associated with major histocompatibility complex (MHC) class II receptor activity were higher expressed and genes associated with voltage‐gated channel activity and ubiquitin‐protein ligase activity were lower expressed.
• (iii)  Cellular components: higher expressed genes were associated with cytoskeleton, extracellular matrix and plasma membrane, while the GO terms intracellular, synapse and organelle were linked to genes with a lower expression level.

In addition, genes were analyzed according to cell signaling pathways (BioCarta database), and 17 signaling pathways with strong gene enrichment (score < 0.05; Table 3) were found.

Table 3.

Differentially regulated signaling pathways in cortical tuber specimens compared to autopsy controls.

Pathway	No	Enrichment score
p38 MAPK Signaling Pathway	13	0.0022
FAS signaling pathway (CD95)	11	0.0056
Cell Cycle: G2/M Checkpoint	9	0.0085
MAPKinase Signaling Pathway	20	0.0122
Integrin Signaling Pathway	11	0.0138
EGF Signaling Pathway	9	0.0143
Links between Pyk2 and Map Kinases	9	0.0182
Ras‐Independent pathway in NK cell‐mediated cytotoxicity	8	0.0231
Keratinocyte Differentiation	11	0.0242
Signaling of Hepatocyte Growth Factor Receptor	10	0.0262
Rab GTPases Mark Targets In The Endocytotic Machinery	6	0.0271
Angiotensin II mediated activation of JNK Pathway via Pyk2 dependent signaling	9	0.0279
mTOR Signaling Pathway	8	0.0291
TNFR1 Signaling Pathway	9	0.0339
PDGF Signaling Pathway	8	0.0360
Classical Complement Pathway	5	0.0447
Complement Pathway	6	0.0476

Using DAVID we determined which signaling pathways (BioCarta database) were differentially regulated using the dataset of significantly changed genes (P < 0.01; > twofold change) in cortical tuber specimens compared to autopsy controls. The first column shows the signaling pathway, the second column shows the number of significantly changed genes per pathway and the last column shows the enrichment score (P‐value).

Immune and inflammatory response

Increased expression of immune system genes (68 genes) and inflammatory response genes (33 genes) was observed in tubers compared with control tissue, of which 19 genes were associated with both processes (Table S‐II; individual genes in Table S‐VIII). Genes with a higher expression level encoded complement components, chemokines, and MHC elements. Altered complement components, including the complement inhibitors serpinG1, clusterin and complement factor I (CFI) supported the finding that the complement pathway was differentially regulated (Table 3). Expression of serpinA3 and lactotransferrin (LTF) was highly increased, 163‐fold and 128–fold, respectively.

The increased expression levels of serpinA3 and chemokine (C‐C motif) ligand 2 (CCL2) were confirmed with quantitative real‐time PCR in the cortical tuber specimens compared to autopsy control specimens (Figure 3A; P < 0.05). CCL2 and serpinA3 proteins were detected immunocytochemically in all cortical tuber specimens tested (Figure 4A–F). Both dysplastic neurons and giant cells strongly expressed CCL2 (Figure 4C–D), while moderate neuronal immunoreactivity was detected in histologically normal cortex (Figure 4A). Strong expression of serpinA3 within the lesion was observed in giant cells, as well as in reactive glial cells and other inflammatory cells (microglia/macrophages) surrounding blood vessels (Figure 4F).

Figure 3.

Validation of gene expression data with quantitative real‐time PCR. Increased expression levels of serpinA3, CCL2, CX3CR1, ECM2, VCAM1 and Integrin β4 in the cortical tubers were confirmed with quantitative real‐time PCR (A), while lower expression levels were observed for GAD67, GLT1, GABRA5 and Kir3.1 (B). Increased expression levels of genes associated with either the immune/inflammatory response or cell adhesion were confirmed in the perituberal regions compared to autopsy control specimens (C). Expression levels in cortical tuber specimens (n = 6) or perituberal specimens (n = 4) were compared to levels in autopsy control specimens (n = 7). Expression levels were corrected for the expression levels of TBP and normalized to control expression levels. The error bars represent SEM and * represents a P‐value < 0.05 (Student's t‐test).

Figure 4.

CCL2, SerpinA3 and Integrin β1 immunoreactivity in cortical tubers (TSC). A–D. CCL2 immunoreactivity (IR). A. Moderate neuronal IR is observed in histologically normal cortex (CTX; insert), without detectable glial staining. B. Strong IR is observed in cortical tubers (TSC). CCL2 IR is detected in dysplastic neurons of different size and shape (arrows in C), in reactive glial cells (arrow‐heads in C) and in giant cells (arrows in D). E–F. SerpinA3 IR in CTX (E) and TSC (F). E. SerpinA3 IR is not detected in histological normal cortex and white matter (Wm, insert). F. Strong SerpinA3 IR is observed in the cortical tuber, with prominent expression in giant cells (arrows in F and insert a; and insert c, positive giant cell in Wm) as well as in reactive glial cells (insert b: arrow‐heads indicated reactive astrocytes surrounding a negative dysplastic neuron) and inflammatory cells (microglia/macrophages) surrounding blood vessels (arrow‐heads in insert a and insert a'). G–J. Integrin β1 IR in CTX (G) and TSC (H–J). G. Integrin β1 is only detected in endothelial cells in histological normal cortex (arrows and insert). H. Strong integrin β1 IR is detected in TSC; IR is mainly observed in giant cells (arrows in H–J), while dysplastic neurons have low or undetectable IR (arrow‐heads in H,I). Scale bar in J: A and G: 200 µm, B, E, F, H: 150 µm; C, I, J and inserts in F: 40 µm; D: 80 µm.

Cell adhesion

All individual genes associated with cell adhesion were extracted from the DAVID analysis and are presented in Table S‐IX (total 58 genes; Table S‐II). Increased expression levels were observed for laminins, integrins, collagens and several CD antigens, of which CD44 was increased 94‐fold in the tuber specimens. A significant increase in expression level was also observed for the adhesion molecules ezrin and moesin, which interact with TSC1 (hamartin) in the mTOR signaling pathway (43). In addition to the individual genes associated with cell adhesion, we identified the integrin signaling pathway as differentially regulated in the cortical tuber specimens (Table 3; gene enrichment score 0.014).

The increased expression levels of extracellular matrix protein 2 (ECM2), vascular cell adhesion molecule 1 (VCAM1), chemokine (C‐X3‐C motif) receptor 1 (CX3CR1) and Integrin β4 in tubers were confirmed with quantitative real‐time PCR (P < 0.05; Figure 3A). By immunocytochemistry, we observed robust expression of integrin β1 in giant cells in tubers (Figure 4H–J). Low expression levels of integrin β1 were observed in the dysplastic neurons (arrows in Figure 4H, I).

Development

The GO term developmental process, encompassing other GO terms like nervous system development, neurogenesis and neuron differentiation, was most highly represented in the tuber specimens compared to the autopsy control specimens (Table S‐II). Increased expression levels were detected for doublecortin and cam kinase like 1 (DCAMKL1), critically involved in migration during development (82) and filamin A, G protein‐coupled receptor 56 (GPR56) and paired box gene 6 (pax6), in which mutations have been identified in other cortical malformations associated with impaired neuronal migration and organization (31). Other highly expressed genes related to development included spondin2, meteorin, annexinA2 and AHNAK nucleoprotein. Lower expression levels (Table S‐III) were observed for the neuronal adhesion molecules NCAM1 and NCAM2 and for the transcription factors Neurod1 and T box brain 1 (TBR1) (33) in the cortical tuber specimens compared to autopsy controls.

Apoptosis

Increased expression levels for genes defined by GO terms associated with apoptosis were detected in tubers vs. control specimens (Table S‐II), including the initiator of apoptosis caspase 8 and the effector caspases 6 and 7. The tumor necrosis factor receptor 1 (TNFR1) and FAS (CD95) signaling pathway, both associated with the induction of apoptosis and the regulation of the inflammatory responses (30), were differentially expressed in the cortical tuber specimens (Table 3). Other genes with increased expression were the annexins A1, A4 and A5.

Ion channels and synaptic transmission

Of the genes with a lower expression level in cortical tuber specimens compared to autopsy control specimens, 41 were linked to the biological process synaptic transmission and 32 to the molecular function gated channel activity (Table S‐III; individual genes in Table S‐X). Lower expression levels were observed for several GABA_A receptor subunits (α2, α5, β3 and γ2), but also for several glutamate receptors [metabotropic glutamate receptor (mGluR) 7, mGluR8, N‐methyl‐D‐aspartate receptor subunit 2A (NMDAR2A or GRIN2A), glutamate receptor (GluR) 6 (GRIK2) and GluR4 (GRIA4)]. Of particular interest was the prominent lower expression level of the glial glutamate transporter SLC1A2 (GLT1 or EAAT2). Several ion channel genes were lower expressed in the cortical tuber specimens compared to autopsy controls, particularly potassium channels. Significant lower expression levels for SLC1A2 (GLT1) and the inward rectifier potassium channel 3.1 (Kir3.1 or KCNJ3) were detected with quantitative real‐time PCR, and a strong tendency toward reduced expression was detected for glutamic acid decarboxylase 67 (GAD67) and GABA_A receptor subunit α5 (GABRA5) in the comparison to autopsy control specimens (Figure 3B).

Ubiquitination

The most significantly changed biological process associated with genes with lower expression levels in the cortical tuber specimens compared to the autopsy control specimens was ubiquitin cycle, including protein ubiquitination. Ubiquitination labels proteins for degradation and is involved in almost all cellular processes, including apoptosis, immune response and inflammation (48). We observed significant lower expression levels of genes encoding ubiquitin protein ligase E3A and SUMO2, as well as several ubiquitin‐conjugating enzymes.

mTOR signaling pathway

The TSC1 and TSC2 genes are pivotal modulators of the mTOR (mammalian target of rampamycin) signaling pathway. We identified this signaling pathway to be differentially expressed in the cortical tuber specimens with a gene enrichment score of 0.0291 (Table 3). Seven of the eight differentially expressed genes had a lower expression level in the cortical tuber specimens compared to the autopsy controls (Pi3K, PDK1, PDK2, RHEB, FKBP12, PP2A and eIF4E). One gene, MNK1, had a twofold increased expression level. Prominent differences (> twofold) were not observed in the expression levels of the individual TSC genes in the cortical tuber specimens compared to the autopsy specimens. Any functional consequences at the protein level in the regulation of the Pi3K‐mTOR signaling cascade cannot be measured by this microarray analysis as the signaling proteins are activated by phosphorylation and not directly by differences in expression levels.

Perituberal specimens vs. autopsy control specimens

Because the perituberal tissue of only two TSC patients was available for the microarray analysis, a meaningful statistical comparison for gene expression between tubers and autopsy control tissue is limited. In comparison to the gene expression in autopsy control specimens, we identified 46 genes with a higher expression level in the perituberal specimens and 65 with a lower expression level (P < 0.01; minimal twofold change). The GO terms associated with these differentially expressed genes are shown in Table S‐IV and S‐V. Only five, rather general, GO terms were associated with the genes with a lower expression level in the perituberal tissue compared with the autopsy controls. Biological processes associated with genes with a higher expression level in the perituberal tissue were inflammatory response, chemotaxis, developmental process and cell adhesion. Chemokine activity was the most significant changed molecular function (Table S‐IV). Differentially expressed genes associated with cell adhesion (eg VCAM1, tenascin C and selectin E) and the immune/inflammatory response (eg several chemokines: CXCL2, CXCL14, CCL2, −3, −4 and −8) were depicted with pink dots at different P‐values in Figure 2E. We selected eight genes associated with either the immune/inflammatory response or cell adhesion for validation with real‐time PCR. Significant higher expression levels were detected for CCL2, CCL4, CX3CR1 and VCAM1 (P < 0.05, Figure 3C), whereas a strong tendency of increased expression was observed for serpinA3, CCL3 and ECM2. Integrin β4 was not significantly higher expressed in the perituberal specimens compared with the autopsy controls (1.2‐fold).

TSC2 vs. TSC1 mutated cortical tuber specimens

Hierarchical clustering (Figure 2A) revealed a distinction between gene expression in the specimens containing a TSC2 vs. a TSC1 mutation. There were 271 differentially expressed genes, of which 222 had a higher expression level and 49 genes had a lower expression level in the TSC2 associated tubers compared to the TSC1 associated specimens. The GO terms associated with these differentially expressed genes are presented in Table S‐VI and S‐VII. The majority of terms linked to genes with a higher expression level in TSC2 associated tubers were associated with development (eg nervous system development, neuron development and developmental process). Other GO terms were cell adhesion, channel activity, GABA receptor activity and synapse. None of the genes related to cell adhesion or immune/inflammatory response observed in cortical tuber specimens compared to control specimens were found in the comparison of TSC2 mutated to TSC1 mutated cortical tuber specimens. Interestingly, of all genes differentially expressed in the comparison of TSC2 associated tubers with TSC1 associated tubers only 9% (n = 25) were differentially expressed in the tuber specimens compared to autopsy controls, indicating a genotype specific expression profile. Very few and rather general GO terms were found for the genes with a lower expression level in the TSC2 associated cortical tuber specimens.

DISCUSSION

Gene array analysis has been widely used to investigate differences in gene expression in human brain tissue resected during epilepsy surgery 2, 3, 16, 17, 19, 38, 51, 60. We report the first genome wide microarray analysis comparing gene expression in cortical tubers with perituberal tissue from the same patients or with autopsy control specimens. The most important finding was the significantly increased expression of numerous genes related to inflammatory/immune responses and cell adhesion and the diminished expression of neurotransmitter uptake site and receptor subunit genes and voltage‐gated ion channel genes. Much of our data provide new insights into molecular mechanisms that may contribute to tuber formation, and potentially to the genesis of seizures in TSC. In addition, some of our results in human tissue corroborate data obtained from existing animal models of TSC.

We acknowledge several limitations to the interpretation of our results. First, our sample size was small. An ideal experimental design with a large number of well‐matched samples is difficult to achieve in human studies, particularly in the case of rare genetic disorders such as TSC. Not all patients with TSC undergo surgical resection of their epileptogenic tubers. Moreover, frozen representative material is not available in all cases. However, in our study, the tissue specimens were matched in terms of age and pathology, and both TSC1 and TSC2 genotypes were analyzed. Perituberal material that can be frozen is even more rarely collected, and because of our collaborative efforts, we were fortunate to obtain the samples included in the study. Second, changes in mRNA expression do not necessarily imply similar changes in protein levels. However, specific differentially regulated genes and pathways were confirmed with quantitative real‐time PCR and immunocytochemistry, indicating that our mRNA expression changes were consistent across samples and may indeed reflect changes in protein expression. Third, the choice of the control material is extremely difficult in human studies, particularly in the case of pathologies affecting young patients, which limits the number of cases suitable for gene expression studies. Finally, understanding whether changes in gene expression represent the cause or the consequence of chronic pharmacoresistant seizure activity is extremely difficult because epilepsy surgery targeting cortical tubers is performed several years after the start of seizures. Although the TSC patients were well matched in terms of duration of epilepsy, comparison with non‐epileptogenic tubers or with histologically normal (although epileptogenic) tissue from other epilepsy related conditions (eg temporal lobe epilepsy (TLE) with hippocampal sclerosis) could not be performed. The collection of representative tissue from epileptogenic and non‐epileptogenic tubers at autopsy requires a multidisciplinary effort of several years. Moreover, patients undergoing large neocortical resections for TLE are often adult patients (older than TSC patients) and the resected cortex often displays histological changes, including gliosis and/or dysplastic features 10, 74.

Immune and inflammatory molecules in cortical tubers

Enhanced expression of numerous genes associated with the immune and inflammatory response was observed, supporting previous observations in TSC 11, 52. The activation of inflammatory molecules is extensively described in both experimental models of epilepsy and human epilepsy tissue 28, 49, 80, 81, including epilepsy‐associated malformations of cortical development 2, 4, 12, 61, suggesting a pathogenic role for inflammation in human epilepsy. Enhanced expression of genes encoding proinflammatory proteins in TSC may reflect a direct effect of altered mTOR pathway signaling in TSC 47, 68. Indeed, altered expression of inflammatory genes was more prominent in tubers where the mTOR cascade is highly activated, than in the perituberal regions where there is minimal change on mTOR activity. For example, the cytokine CCL2, also known as monocyte chemoattractant protein‐1 is directly dependent on TSC2 levels in vitro and is highly expressed in fibroblasts taken from TSC patients and in TSC2 null fibroblasts derived from the Eker rat model of TSC (46). CCL2 is critically involved in the regulation of brain inflammation by activating microglial cells, the production of proinflammatory cytokines (75) and increasing blood‐brain barrier (BBB) permeability (70). CCL2 protein was prominently expressed in glial cells, dysplastic neurons and giant cells in our tuber specimens.

The inflammatory mediators produced in the epileptic brain, including cytokines (eg IL‐1β) and complement factors, can induce BBB permeability (57). The exact mechanisms how cytokines increase the BBB permeability needs to be elucidated. However, the increased production of selectins, adhesion molecules, cytokines and their receptors on the brain endothelium has been implicated in increased BBB permeability 59, 81. This results in the leukocyte recruitment from the blood stream enhancing the inflammatory response. Accordingly, the increased expression levels of cell adhesion and inflammatory molecules detected in the present and a previous study (52) suggest a possible modulation of the BBB integrity in TSC. This is supported by our previous observation in which we detected extravasation of albumin accompanied by an increase in the number of lymphocytes in TSC (11). Accumulation of lymphocytes is not detected in both human and experimental temporal lobe epilepsy associated with hippocampal sclerosis (62), despite evidence of BBB leakage and the production of proinflammatory cytokines. Given the fact the Pi3K‐mTOR signaling pathway has been implicated in the regulation of the innate and adaptive immune response (84), we might speculate that aberrant mTOR signaling in TSC contribute to the observed lymphocyte infiltrations. Changes in the BBB integrity have been shown to induce seizure progression in epileptic rats 23, 63, 79.

Cell adhesion

Numerous cell adhesion genes exhibited enhanced expression in tubers, including the integrins and their ligands tenascin C, VCAM1 and laminins (72). Altered expression of cell adhesion molecules in TSC during brain development could be linked to the apparent disorganization of cortical lamination characteristic of tubers. An example is given by integrin β1 which is critically involved in proper layering of the developing cortex by establishing adhesive contacts between the glial endfeet of the radial glial cells and the basement membrane. Loss of integrin β1 results in excessive neuronal migration and a disorganized marginal zone 29, 54. The significance of higher expression levels of integrin β1 are not known, but abnormal layering of the cerebral cortex in the cortical tubers might be a consequence. Integrin β1 was mainly observed in giant cells, and given the fact that integrin β1 is expressed in neural stem cells (32), this supports the previously suggested hypothesis that giant cells in TSC [or so‐called balloon cells in focal cortical dysplasia (FCD)] retain an immature and possibly stem cell‐related phenotype 44, 45. Increased CD44 expression, supporting previous observations (1), might play a crucial role in enhancing inflammation in the cortical tubers by increasing the blood brain barrier permeability (59).

Apoptosis

Various apoptotic genes and two cell death signaling pathways (TNFR1 and FAS (CD95) signaling) were differentially expressed in the cortical tuber specimens compared to the autopsy controls. Expression of several apoptotic markers, including caspase 3, 8 and FAS, has been reported previously in cortical tuber specimens 11, 52. The activation of cell death programs may relate to aberrant mTOR signaling in tubers and is supported by prior studies demonstrating TdT‐mediated dUTP‐biotin nick‐end labeling (TUNEL)‐positive cells in human tuber specimens (52).

Synaptic transmission and ion transport in cortical tubers

Alterations in glutamatergic and GABAergic synaptic transmission is critically involved in the initiation of epileptiform activity 7, 69. A pivotal finding was the reduced expression of the glial glutamate transporter (GLT‐1 or EAAT2) in tubers which is in line with the reported reduced expression of the glutamate transporters GLT‐1 and GLAST in the TSC1 ^GFAP conditional knock‐out mice 88, 89. These knock‐out mice show no focal abnormalities resembling cortical tubers 76, 87; however, their phenotype includes progressive epilepsy and astrogliosis. These observations support a link between altered TSC1 or TSC2 function, reduced GLT‐1 expression and seizures, representing a possible therapeutic target.

Reduced expression levels of GABA_A receptor subunits α1 and α2 are reported for dysplastic neurons and giant cells in cortical tubers (86). In the present study, we observed lower expression levels for the GABA_A receptor subunits α2, α5, β3 and γ2. Lower expression levels are reported for various GABA_A receptor subunits, especially GABA_Aα5 in both experimental and human epilepsy 5, 21, 28, 34, emphasizing the important role of GABAergic inhibition in preventing hyperexcitability and epileptogenesis.

A fascinating finding was the reduced expression of several ion channels, especially potassium channels. Buffering of the extracellular K⁺ levels, mainly by glial cells, is in addition to the regulation of glutamate levels important to prevent excessive neuronal excitation 15, 39. Reduced potassium current is measured in the TSC1 ^GFAP CKO mice, indicating that loss of function of TSC1 modulates K⁺ buffering of glial cells (36). Impaired function of various potassium channels have been associated with epilepsy in both humans and animal models of epilepsy 9, 25.

Genotype

The hierarchical clustering revealed a distinct genetic profile related to a mutation in either of the two genes. In addition to processes regulated by both proteins, as for example the activation of the Pi3K‐mTOR signaling, growing evidence indicates individual functions for both proteins as well [reviewed in Rosner et al (64)]. These genotype specific expression profiles might contribute to the observed clinical differences related to either of the two genes 20, 65. Prominently changed genes associated with immune/inflammatory response and cell adhesion observed in tubers compared to autopsy controls were not differentially expressed in the TSC2 mutated tubers compared with the TSC1 mutated ones, indicating that these processes are altered independently of the mutation and generally associated with TSC. Moreover, this suggests that in addition to altered mTOR signaling, other genes or mechanisms, for example, seizures, regulate the expression levels of these genes.

Perituberal tissue

In comparison with the autopsy control specimens, 111 genes were differentially regulated in the perituberal specimens. We observed increased expression levels of genes associated with inflammation and cell adhesion, which was confirmed with real‐time PCR. At this point, it is impossible to dissect to what extent the observed differences in gene expression in the perituberal regions contribute to the global changes in brain function or epileptogenesis in TSC; however, the higher expression levels of cytokines and cell adhesion molecules in perituberal tissue is likely to be relevant. The role of the perituberal regions in epileptogenesis is still debated 50, 87, and increasing evidence supports the importance of the perituberal cortex in TSC, as well as in other malformations of cortical development [for review, see Wong (87)]. Although epilepsy surgery in TSC patients often results in seizure freedom 40, 85, some patients continue to have seizures after targeted tuber resection. Moreover, the surgical resections contain often a small margin of perituberal, histological normal cortex and a few cases of TSC patients who become seizure free after resection of non‐tuberal cortex have been reported (83). Thus, we cannot exclude that network abnormalities or molecular or cellular defects may occur in the perituberal regions that become part of the epileptogenic network and can be even the primary source of seizures.

Cortical tuber gene expression profile compared to changes in gene expression reported in epilepsy‐associated glioneuronal developmental lesions

An important question is whether the microarray data highlight epileptogenic mechanisms common to other glioneuronal developmental lesions [such as FCD and ganglioglioma (GG)], or whether distinct molecular mechanisms can be detected in TSC cortical tubers. Comparison of microarray data obtained from different studies requires caution, because several technical and data analytical discrepancies may influence the results, including the type of platform used, as well as the age of patients and the type control tissue used.

Comparison with the gene expression profiles performed in cortical dysplasias (38) suggests differences in the transcriptional profiles. Although in both studies, genes associated with apoptosis showed altered expression, the FCD gene expression profile (38) indicates induction of anti‐apoptotic mechanisms, whereas our study shows more prominent activation of cell death programs. However, the FCD specimens analyzed by Kim et al (38) were heterogeneous [possibly including both type I and type II FCD (58)] and the control material used was obtained from a 2‐year‐old patient with intractable epilepsy and therefore not comparable with our study.

The comparison of the TSC data with the gene expression profiles performed in GG 2, 24 shows intriguing similarities with prominent elevated expression of genes involved in the immune and inflammatory responses as well as in the regulation of cell adhesion. The data indicate activation of innate and adaptive immunity in both pathologies and point to the role of complement and IL‐1β‐mediated pathways, as well as increased expression of serpinA3. The similarities between cortical tubers and GG in terms of major regulated processes (such as adhesion and inflammatory processes) are not surprising, considering that GG share with TSC lesions not only the history of chronic epilepsy but also common pathogenetic mechanisms. In particular, recent studies suggest a role for the mTOR‐cascade signaling pathway in the molecular pathogenesis of these glioneuronal lesions 8, 14, 66, 67. Interestingly, the mTOR pathway not only plays a role in regulating cell growth and size, but also regulates both the innate and adaptive immune response 47, 68, 84.

Expression of genes related to synaptic transmission and voltage‐gated channel activity was reduced compared to autopsy controls in both TSC and GG specimens. In addition, microarray analysis in both GG and TSC specimens shows upregulation of genes related to the plasminogen system, as well as high expression of several factors involved in the regulation of angiogenesis. Interestingly, alterations of the blood brain barrier permeability and angiogenesis have been recently observed in both human and experimental models of TLE with positive correlation between the increased vascular permeability and the occurrence of spontaneous seizures in chronic epileptic rats 63, 78. Thus, these similarities support the concept of common pathways contributing to the epileptogenicity and intractability of focal chronic epilepsies.

CONCLUSIONS

In this study, we demonstrate that cell adhesion and inflammatory genes are most highly expressed, while genes related to synaptic transmission exhibited reduced expression in tubers compared with autopsy controls. Furthermore, we demonstrate that various genes and processes were differentially expressed in the perituberal regions, indicating that the area of functional abnormality is more extensive and perituberal cortex might be critical in seizure generation. Better understanding of the underlying mechanism in cortical tubers and the possible contribution of the perituberal regions is crucial for the development of novel treatment strategies.

Supporting information

Table S‐I. List of primers used for quantitative real‐time PCR.

Table S‐II. Significantly changed GO terms (P < 0.01) associated with genes with a higher expression level in cortical tuber specimens compared with autopsy controls.

Table S‐III. Significantly changed GO terms (P < 0.01) associated with genes with a lower expression level in cortical tuber specimens compared with autopsy controls.

Table S‐IV. Significantly changed GO terms (P < 0.05) associated with genes with a higher expression level in perituberal specimens compared to autopsy control specimens.

Table S‐V. Significantly changed GO terms (P < 0.05) associated with genes with a lower expression level in perituberal specimens compared to autopsy control specimens.

Table S‐VI. Significantly changed GO terms (P < 0.05) associated with genes with a higher expression level in TSC2 mutated cortical tuber specimens compared to TSC1 mutated cortical tuber specimens.

Table S‐VII. Significantly changed GO terms (P < 0.05) associated with genes with a lower expression level in TSC2 mutated cortical tuber specimens compared to TSC1 mutated cortical tuber specimens.

Table S‐VIII. Significantly differentially expressed genes (P < 0.01) linked to inflammatory response and immune system process in cortical tuber specimens compared to autopsy controls.

Table S‐IX. Significantly differentially expressed genes (P < 0.01) linked to cell adhesion in cortical tuber specimens compared to autopsy controls.

Table S‐X. Significantly differentially expressed genes (P < 0.01) linked to synaptic transmission and gated channel activity in cortical tuber specimens compared to autopsy controls.

Please note: Wiley‐Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

Supporting info item

Supporting info item

Supporting info item

Supporting info item