Brain Inflammation in Epilepsy: Experimental and Clinical Evidence


Address correspondence and reprint requests to Dr. A. Vezzani at Lab Exp Neurol, Istituto di Ricerche Farmacologiche Mario Negri, Via Eritrea 62, 20157 Milano, Italy. E-mail


Summary:  Inflammatory reactions occur in the brain in various CNS diseases, including autoimmune, neurodegenerative, and epileptic disorders. Proinflammatory and antiinflammatory cytokines and related molecules have been described in CNS and plasma, in experimental models of seizures and in clinical cases of epilepsy. Inflammation involves both the innate and the adaptive immune systems and shares molecules and pathways also activated by systemic infection. Experimental studies in rodents show that inflammatory reactions in the brain can enhance neuronal excitability, impair cell survival, and increase the permeability of the blood–brain barrier to blood-borne molecules and cells. Moreover, some antiinflammatory treatments reduce seizures in experimental models and, in some instances, in clinical cases of epilepsy. However, inflammatory reactions in brain also can be beneficial, depending on the tissue microenvironment, the inflammatory mediators produced in injured tissue, the functional status of the target cells, and the length of time the tissue is exposed to inflammation. We provide an overview of the current knowledge in this field and attempt to bridge experimental and clinical evidence to discuss critically the possibility that inflammation may be a common factor contributing, or predisposing, to the occurrence of seizures and cell death, in various forms of epilepsy of different etiologies. The elucidation of this aspect may open new perspectives for the pharmacologic treatment of seizures.


The immune system and the associated inflammatory reactions play an important role in tissue protection and repair from a variety of infectious or noninfectious insults. To defend the host from the invasion of foreign organisms or pathogenic threats, the immune system has evolved into two parts: one responsible for an immediate action against external agents, called the innate immune system, and the other that allows antigen recognition via specific antigen-presenting cells (APCs) and antigen receptors, constituting the adaptive immune system. Whereas the innate immune system uses mainly phagocytic cells, including monocytes/macrophages and microglia, B and T lymphocytes are the pivotal cellular members of the adaptive immune system (1,2). Communication between cells of the immune system occurs either via direct cell-to-cell contact or via soluble factors called cytokines.

The central nervous system (CNS) is considered an immunoprivileged site because of the presence of a blood–brain barrier (BBB), graft acceptance, lack of a conventional lymphatic drainage, and an apparently low traffic of monocytes and lymphocytes. It should, however, be defined as an immunologically specialized site (3) because it is becoming clear that immune and inflammatory reactions do occur in the CNS, where they appear to originate either intrinsically, thus constituting part of the innate immunity, or in the peripheral tissues and imported by competent immune cells into the CNS (acquired immunity). The transition between innate and adaptive immunity is mediated by a large variety of inflammatory mediators, among which cytokines and Toll-like receptors (TLRs) play a key role (4).


Inflammation in the brain is a condition characterized by the presence of an array of molecules (cytokines and established mediators of inflammation) that are not detectable or are barely detectable in physiologic conditions. These molecules are classically produced by cells of the immune system in response to infection or various kinds of pathologic threats; however, it is well established that inflammatory mediators also are produced by brain parenchymal cells (microglia, astrocytes, and neurons) and by cells of the BBB and choroid plexus. One of the best characterized ways for inducing a rapid and robust inflammatory reaction in the CNS consists of the administration to rodents of lipopolysaccharide (LPS), a component of the outer membrane of gram-negative bacteria, which mimics systemic infection. LPS is one element of the “pathogen-associated molecular patterns,” which are produced by invading microorganisms and are recognized by specific receptors (TLRs) and cells of the immune system. TLRs are a family of type 1 transmembrane proteins evolutionarily conserved between insects and mammals and expressed by cells of the immune system (4). The stimulation of these receptors and costimulatory molecules (e.g., CD14) activates downstream events in APCs that are in part shared by the interleukin (IL)-1–receptor type I intracellular signaling. TLR activation results in the induction of transcriptional factors such as nuclear factor κB (NFκB), which has the ability to trigger various proinflammatory genes such as those encoding cytokines, chemokines, proteins of the complement system, cyclooxygenase-2, and inducible nitric oxide (2,5,6). Recently, TLR 2 and TLR 4 have been described in the CNS, where they are present in parenchymal microglia and macrophages, in the circumventricular organs, the choroid plexus, the microvasculature, and in the ependymal cells of the ventricles (6,7). These receptors are rapidly increased at these CNS sites after LPS, suggesting that a systemic immune challenge induces inflammation in the CNS also by a direct action on brain cells and not only by increasing cytokines and inflammatory mediators in blood (8). The activation of microglia by endotoxemia appears to be a crucial step in the CNS inflammatory response to infection and overlaps with the progressive activation of monocytes, neutrophils, and lymphocytes, as reflected by circulating levels of different proinflammatory molecules (5,6).

An immune response in the CNS may be triggered also by endogenous ligands that stimulate TLRs. For example, signals from damaged cells (heat-shock proteins, components of the extracellular matrix degradation) or arising from molecules entering the brain through a damaged BBB may initiate an inflammatory response. In this respect, it is noteworthy that the CNS shows a robust inflammatory response not only to infectious agents but also to a large spectrum of injuries, such as those occurring after ischemic, traumatic or excitotoxic brain damage, or during seizures (see Fig. 3) (9,10). The regional and cellular patterns of induction of inflammatory molecules and their time course of activation and persistence in brain tissue appear to depend on the nature of the CNS injury.

Figure 3.

Expression of genes encoding proinflammatory molecules in rodent forebrain areas after seizures. A: Intense hybridization signal of TLR2, IκBα, an index of NFκB activation, the cytokine tumor necrosis factor-α (TNF-α), the chemokine monocyte-chemoattractant protein 1 (MIP-1), and cyclooxygenase-2 (COX-2), in mouse forebrain 24 h after seizures induced by pilocarpine (37). Very low or nondetectable expression was found in basal conditions (saline-injected mice). The phenotype of the expressing cells was identified by using selective markers: the majority of cells expressing these inflammatory mediators were infiltrating monocytes and parenchymal microglia; in some instances, endothelial cells were positive. Neurons overexpressed COX-2 and IκB mRNA. B: Time-dependent increase of interleukin (IL)-1β, TNF-α, and IL-6 mRNA in the rat hippocampus after electrically induced status epilepticus (40). C: IL-1β immunoreactivity (a, c) is enhanced after seizures in microglia-like cells in the hippocampus (b, d). Micrographs in (a) and (b) depict sham-operated rats, whereas panels (c) and (d) depict IL-1β and B4-isolectine positive microglia, respectively (39). Ctx, cortex; CA1, CA3, pyramidal layer regions; DG, dentate gyrus; Hb, habenular nucleus; MD, mediodorsal thalamic nucleus; MeA, medial amygdala; Hyp, hypothalamus; Pir, Piriform cortex; Py, pyramidal cell layer; SO and SR, strata oriens and radiatum.


The cellular basis of the BBB is at the level of the CNS microvasculature and consists morphologically of nonfenestrated endothelial cells with interendothelial tight junctions. In addition, the maintenance of the BBB function is dependent on normal functioning of pericytes, perivascular microglia, astrocytes, and the basal lamina, which are annexed to the capillary and postcapillary venules in the CNS. In some areas in the CNS, the BBB is incomplete or absent, such as the circumventricular organs and the choroid plexus. A blood–cerebrospinal fluid (CSF) barrier is found at the level of apical tight junctions in the choroid plexus epithelial cells.

For more detailed information about the anatomic and physiologic characteristics of the BBB and blood–CSF barriers, see the comprehensive review articles (11,12).

Under normal physiologic conditions, the BBB provides protection of the CNS by strictly regulating the entry of plasma-born substances and immune cells into the nervous tissue. A variety of CNS injuries [including seizures (13), infections, traumatic and ischemic events] cause transient changes in the physiologic and structural features of the BBB. In particular, an impairment of BBB integrity and an inflammatory state are common features of several neurologic conditions (14) associated with the late onset of epilepsy.

During inflammation, the enhanced production of cytokines by the endothelial cells of the BBB, the circulating immune cells, and brain parenchymal microglia and astrocytes results in the upregulation of adhesion molecules, activation of metalloproteinases, and catabolism of arachidonic acid at the level of the brain microvasculature (12,15,16). These events contribute to increase the permeability of the BBB through mechanisms that are not yet fully elucidated (16).

Inflammatory events at the BBB and blood–CSF barrier are pivotal for promoting the entry of leukocytes into the CNS, thus determining intrathecal immune reactivity (see later; Fig. 1). Although immune cells are found to patrol the CSF in the subarachnoid space, the perivascular Virchow–Robin space, and the choroid plexus, relatively few leukocytes are present in the CNS unless inflammation occurs (3). One of the mechanisms by which cytokines contribute to the inflammatory response at the level of the BBB and blood–CSF barrier is by increasing the expression of selectins and adhesion molecules [e.g., intercellular adhesion molecule (ICAM)-1, vascular CAM (VCAM)-1, and platelet endothelial CAM (PECAM-1)], chemokines, and their receptors on endothelial and epithelial cells. These molecules, by interacting with integrin molecules on leukocyte membrane surface, are responsible for leukocyte recruitment from the bloodstream, promoting their adhesion and eventual entry into the perivascular space, CSF, and CNS parenchyma (3).

Figure 1.

Schematic representation of the cerebral vasculature and CSF circulation and their relation to inflammatory mediators and antigen transport. Antigenic products (Ag) can be drained from the brain parenchyma to the CSF. From this compartment, Ag may circulate along the ventricular system into the subarachnoid space; from here Ag can enter the venous sinuses through the arachnoid villi and reach the systemic circulation. Alternatively, they can drain into the cervical lymph nodes, traveling along the subarachnoid and perivascular space. Trafficking of leukocytes into the CNS may occur through at least three routes: (a) Leukocytes can migrate into the brain following the formation route of CSF. They can extravasate through the fenestrated endothelium of the chroid plexus, migrate across the stroma core, interact with the epithelial cells, and enter the CSF. This route may be one by which immune cells enter the brain in physiologic conditions (the CSF of healthy individuals contains ∼3,000 leukocytes/ml); (b) Leukocytes can have access to the brain via the arterial blood supply by extravasating at the pial surface of the brain into the subarachnoid space and across the Virchow–Robin perivascular spaces, which are in direct contact with the CSF compartment. In rodent, the cervical lymph nodes are connected by nasal lymphatics to subarachnoid space and perivascular space at the olfactory bulbs; (c) Leukocytes can enter the brain parenchyma by extravasation across the blood–brain barrier nonfenestrated endothelium and basal lamina. Antigen-presenting cells such as dendritic cells and macrophages are found in the subarachnoid space, in the perivascular space, and in the choroid plexus, whereas microglia are located in brain parenchyma; at these locations, these cells may encounter circulating CNS-borne antigens. (Adapted from Aloisi F, Ria F, Adorini L. Regulation of T-cell responses by CNS antigen-presenting cells: different roles for microglia and astrocytes. Immunol Today 2000;21:141–7, with permission.)

During inflammatory/immune reactions, activated T cells can enter the CNS by at least three different routes:

  • 1The first pathway follows the formation of CSF; leukocytes extravasate across the fenestrated endothelium in the choroid plexus stroma, cross the epithelium of the choroid plexus, and enter the CSF.
  • 2Leukocytes can extravasate across postcapillary venules at the pial surface of the brain into the subarachnoid space and the Virchow–Robin perivascular space, which is in direct communication with the CSF.
  • 3Leukocytes cross the BBB and the endothelial basal lamina of the postcapillary venules [(3); see Fig. 1].

Leukocytes can be restimulated in the CNS by APCs circulating in the subarachnoid or perivascular space or by APCs located in the brain parenchyma. Antigenic proteins in the CNS, which are mostly inacessible to T cells because of the restrictive nature of the BBB and the lack of a conventional lymphatic system, may become available to recognition by the action of APCs. Thus during inflammatory conditions, microglia and macrophages, endothelial cells of the BBB, and epithelial cells of the choroid plexus are all capable of presenting antigens to T cells (1,17–19). These APCs can express major histocompatibility complex (MHC) class I and II and costimulatory molecules during inflammation, which are essentials for T lymphocytes to recognize and respond to an antigenic peptide. Endothelial cells, unlike perivascular microglia, do not constitutively express MHC class II molecules; however, they can be induced to express these molecules by a variety of cytokines. The role of MHC class II on astroglia remains elusive, and a prevailing view is that these cells exert a negative immunoregulatory function by favoring the induction of a nonresponsive state by the T cells (17).

In addition to be expressed on the surface of APCs, antigenic proteins may drain from the CNS into the cervical lymph nodes along the CSF-draining pathways or enter the venous sinuses through arachnoid villi and reach the spleen (3,17) (see Fig. 1). CNS antigens also may be presented to naive T cells by dendritic cells, which are localized in the choroid plexus and the meninges. Evidence exists that inflammatory mediators produced in the CNS can stimulate dendritic cells to migrate into the brain parenchyma, where they can phagocytize antigens and subsequently migrate to cervical lymph nodes for antigen presentation to immunocompetent cells (20) (see Fig. 1).


Stress and the immune system appear to be strictly interconnected, and activation of the stress axis is demonstrated to occur during seizures (21–23). The stress–induced release of hypothalamic corticotropin-releasing hormone (CRH) leads ultimately to the systemic secretion of glucocorticoids (GCs), epinephrine, and norepinephrine, which in turn strongly influence immune and inflammatory reactions. Conversely, immune molecules, in particular IL-1, IL-6, and tumor necrosis factor (TNF)-α stimulate CRH secretion, thus activating the hypothalamic–pituitary–adrenal (HPA) axis and the sympathetic nervous system during inflammatory responses. CRH modulates immune/inflammatory reactions through receptor-mediated actions of GCs on target immune tissues and by directly acting on its own receptors on immune cells. However, the actions of CRH as a modulator of stress are not confined to the HPA system but appear to involve also CRH receptors in the limbic areas, such as the amygdala and the hippocampus (22).

GCs have immunosuppressant effects by preventing the migration of leukocytes from the circulation into extravascular spaces, reducing the accumulation of monocytes and granulocytes at inflammatory sites, and suppressing the production and actions of many cytokines and inflammatory mediators. Although stress has been in general regarded as immunosuppressive, evidence indicates that stress hormones differentially regulate the differention of CD4+ T cells into two distinct phenotypes: T helper (Th)1 and Th2 patterns and their consequent type 1 and type 2 cytokine secretion (24). Th1 cells regulate cellular immunity (CD8+ T-cytotoxic cells, natural killer cells, and activated macrophages) by primarily secreting type 1 cytokines: interferon (IFN)-γ, IL-2, and TNF-β, whereas Th2 cells secrete primarily type 2 cytokines, such as IL-4, IL-10, and IL-13, which promote humoral immunity (mast cells, eosinophils, B cells). Strong evidence suggests that stress hormones amplify the Th2 response while suppressing the Th1 response.

When entering the brain, Th cells can interact with microglia, which in turn can efficiently restimulate both Th1 and Th2 cells and release inflammatory mediators (Fig. 2). Astrocytes appear to have a limited APC capacity restricted to recognition by Th2 cells (17).

Figure 2.

Intracerebral T-cell responses, mediated by intrapenchymal antigen-presenting cells (APCs), APCs decide the differentiation of T-helper cells (Th) into Th1 and Th2 cells, which secrete type 1 and type 2 cytokines, rspectively. These cells are pivotal in the regulation of cellular and humoral immunity, respectively. T cells primed in peripheral lymphoid tissues, or in the perivascular space and CSF by circulating APCs (see also Fig. 1), when entering the CNS parenchyma can interact with microglia and astrocytes. Microglia can efficiently express major histocompatibility complex (MHC) class II and adhesion/costimulatory molecules [e.g., intercellular adhesion molecule (ICAM)-1, B7] when activated by inflammatory stimuli and can restimulate both Th1 and Th2 cells and release mediators [including chemokines, interleukins, tumor growth factor β (TGF-β) and prostaglandin E2 (PGE2)] that regulate the phenotype, recruitment, activation, and eventually apoptosis of Th cells. Astrocytes appear to have a poor ability to express MCH class II and adhesion/costimulatory molecules, and their APC capacity may be limited to Th2-cell restimulation and inhibition of the Th1 response. CD28, costimulatory signal; TCR, T-cell receptor. (Adapted from Aloisi F, Ria F, Adorini L. Regulation of T-cell responses by CNS antigen-presenting cells: different roles for microglia and astrocytes. Immunol Today 2000;21:141–7, with permission.)

In summary, an excessive immune/inflammatory response by activation of the stress system may inhibit cellular immunity while favoring humoral immunity. This mechanism may serve to protect the organism from excessive production of proinflammatory cytokines with tissue-damaging potential. As a consequence, excessive HPA activity may increase susceptibility to infectious agents but enhance resistance to autoimmune/inflammatory diseases dominated by cellular immune responses.


Several experimental findings have highlighted a dichotomous role for immune/inflammatory reactions in the CNS (2,9,25), showing that these reactions can either be protective, thus constituting an adaptive, beneficial, endogenous response, as occurring during the classic immune response to infection, or be a direct or indirect cause of neuronal dysfunctions.

Notable examples exist of the dual role of immunity and inflammation in the CNS: in particular, immunologic/inflammatory challenges may contribute to the etiopathogenesis of some CNS disorders such as Rasmussen encephalitis (RE), stiff-man syndrome, and paraneoplastic neurologic diseases (2). In these disorders, antibodies can be generated against specific neuronal targets, and in association with inflammatory mediators produced in the CNS or entering the brain from blood, they can induce neuronal dysfunction and damage. A robust inflammatory response in the brain is associated also with extracellular deposition of β-amyloid protein, and it has been assumed to contribute to neurodegeneration in Alzheimer disease. However, a beneficial effect of inflammation in this disease is indicated by the recent evidence that activation of microglia and the production of the cytokine transforming growth factor (TGF)-β1, may participate in the clearance of β-amyloid protein deposits, thus preventing their harmful effects (26).

Deleterious effects of cytokines on neuronal survival are likely to be mediated by their ability to provoke an extracellular increase of glutamate by acting on the mechanisms of its release and reuptake, potentiate the function of ionotropic glutamate receptors (27,28) and enhance the production of mediators of oxidative stress (i.e., arachidonic acid and nitric oxide) (9,29). However, microglia-derived IL-1β can induce also the synthesis of nerve growth factor, ciliary neurotrophic factor, and insulin-like growth factor from astrocytes that can promote repair of the CNS. Other potential mechanisms of neuroprotection induced by cytokines include stimulation of antioxidant pathways (30) and enhanced expression of manganese superoxide dismutase or calbindin, leading to attenuation of the elevation of intracellular Ca2+ induced by cell injuries (31,32).

It is increasingly clear that the final outcome of inflammation on cell function and survival is highly dependent on the extent to which cytokines are produced, the length of time the tissue is exposed to inflammation, and the balance between the neurotrophic and inflammatory factors produced by the competent cells.


Seizures provoked experimentally in rodents induce a pattern of inflammatory mediators in the brain similar to that occurring after endotoxemia. However, notable differences exist because a massive inflammatory response is induced by seizures in microglia, astrocytes, and neurons, whereas these cells except for microglia, are only marginally involved or not involved by endotoxemia. Moreover, changes in inflammatory mediators induced by endotoxemia are relatively brief when compared with those observed after seizures.

Experimentally induced seizures in rodents trigger a prominent inflammatory response in brain areas recruited in the onset and propagation of epileptic activity (10,33–40). The inflammatory response occurring during the first 3 days after pilocarpine-induced seizures does not progress from the periventricular structures and blood vessels to the surrounding parenchymal elements, such as after LPS (or ischemia/hypoxia), but it directly involves microglia, astrocytes, and, in some instances, neurons (5,6,37). Members of the TLR family are significantly upregulated by seizures in microglia throughout the forebrain (37), and their engagement may lead to transcriptional activation of cytokines, chemokines [macrophage inflammatory protein-1α (MIP-1α), monocyte chemoattractant protein-1 (MCP-1), and regulated on activation in normal T cells, expressed, probably secreted (RANTES)], MHC class I and II and costimulatory molecules in brain parenchymal cells, and in endothelial and epithelial cells of the BBB and choroid plexus, respectively. This process may enable these cells to present antigenic peptides to infiltrating T lymphocytes. However, markers of adaptive immunity have not been detected within 3 days after pilocarpine-induced seizures in mouse brain, as assessed by immunostaining for T or B cells or production of IL-12 and IFN-γ (37,41). No evidence yet supports the secondary recruitment of the cellular element of the adaptive immune system by seizure activity per se, at least in the time frame and in the epilepsy models investigated so far.

Cytokines and related molecules: expression studies in experimental models of seizures

Seizures induced either chemically or electrically increase cytokines in rodent brain (10,33–40,42) (Fig. 3 and Table 1). Although IL-1β, TNF-α and IL-6 are expressed at very low levels in normal brain, their messenger RNA (mRNA) and protein levels are rapidly (≤30 min) increased after the induction of seizures, declining to basal levels within 48–72 h from the onset of seizures (40). Notably, IL-1β is still upregulated in the brain 60 days after status epilepticus in rats with spontaneous seizures (40). Proinflammatory cytokines are induced in brain also by audiogenic and kindled seizures (36,38).

Table 1. Inflammation in experimental models of seizuresThumbnail image of

Cytokine synthesis in the brain has been detected in microglia and astrocytes, although subsets of hippocampal neurons can express IL-1β in spontaneously epileptic rats.

Production of proinflammatory molecules [cytokines inducing a portfolio of genes that are established mediators of inflammation (43)] is typically accompanied by the concomitant synthesis of antiinflammatory mediators and binding proteins apt to modulate the inflammatory response, thus avoiding the occurrence of deleterious effects. In this respect, upregulation of IL-1–receptor antagonist (Ra), a naturally occurring antagonist of IL-1β, has been described after acute seizures, status epilepticus, and in kindling (38–40,42). However, IL-1Ra and IL-1β are induced by seizures to a similar extent, and IL-1Ra is produced with a delayed time course, differing from classic inflammatory reactions in which IL-1Ra is produced 100- to 1,000-fold in excess, and concomitant with IL-1β (44). Thus the brain is less effective than the periphery in inducing a crucial mechanism for rapidly terminating the actions of a sustained increase in endogenous IL-1β.

Cytokine receptors in the CNS are expressed by neurons, microglia, and astrocytes. Evidence exists for cell type–specific IL-1β signaling in the CNS through the IL-1 type 1 receptor: IL-1β activates the p38 mitogen-activated protein kinase (MAPK) pathway in neurons, leading to induction of cyclic adenosine monophosphate (cAMP) response element–binding protein, whereas NF-κB is activated predominantly in astrocytes, suggesting that this cytokine may have distinct functional effects on neurons and glia (45).

IL-1 and TNF-α receptors are rapidly upregulated in neurons during seizures (46,47), suggesting that they mediate the effects of cytokines on neuronal excitability (see later paragraphs). Cytokine receptors are induced in astrocytes several hours after the onset of seizures in brain areas where degenerating neurons are found. Cytokines acting on glial-cell receptors can induce the production and release of either cytotoxic or neurotrophic molecules, which may contribute to determining whether cells survive or degenerate in hostile conditions. In this respect, IL-1β and TNF-α can either exacerbate or reduce α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA)-receptor–mediated excitotoxicity in organotypic slice cultures, depending on their extracellular concentrations and the length of time the tissue is exposed to these cytokines during the injury (48).

TNF-α type 1 (p55) receptors contain an intracellular “death domain,” and they appear to contribute predominantly to cell damage (49,50). Seizures promote the formation of a complex between the “death domain” of this receptor and activating factors that result in the induction of proapoptotic signals. Accordingly, neutralizing antibodies to TNF-α reduced the number of DNA-damaged cells in the hippocampus after kainate seizures (50). Differently, p75 receptor can mediate trophic, neuroprotective, and anticonvulsant effects of TNF-α in mice (see later paragraph).

IL-6 receptor and its signaling transducer protein, Gp130, are increased in rat forebrain and in the meninges after seizures; however, the specific cell types expressing this receptor have not been identified (51).

In summary, the functional status of cytokine receptors and their expression in healthy and injured tissue, and the production of antiinflammatory molecules during seizures, represent two important components of the inflammatory reactions that contribute to determine the final physiopathologic outcome of an increase in tissue cytokines.

Other inflammatory mediators

Proinflammatory cytokines trigger the transcription of various inflammatory genes including cyclooxygenase-2 (COX-2), the limiting enzyme for the formation of prostaglandins (PGs). Whereas endotoxemia induces COX-2 predominantly in endothelial cells of blood vessels, seizures rapidly upregulate the same gene mainly in hippocampal neurons and in glia (see Fig. 3) (6,37).

COX-2 induction in neurons after seizures is transient and resolves within 72 h from the onset of seizures, but its expression in glial cells (presumably in astrocytes) is long-lasting and persists for several weeks. The late expression of COX-2 in glia, similar to that of IL-1β and TNF-α receptors, typically occurs in brain areas involved in neuronal damage, suggesting a link between PGs production and neurodegeneration. Indeed, treatments with COX-2 inhibitors protect from seizure-induced damage and N-methyl-d-aspartate (NMDA)-induced excitotoxicity (see later). The consequences of COX-2 activation on neuronal excitability and excitoxicity depend on the types of PGs that are produced in the various phases of epileptogenesis, and the involvement of EP1 or EP2 prostanoid receptors or both. In particular, PG E2 and PG F are significantly increased in the brain during the acute phases of status epilepticus, whereas PG D2 is enhanced mostly during the latency phase, and interictally, during chronic seizures (52).

Other cytokines have been studied in epilepsy models including fibroblast growth factor (bFGF) and its receptors subtypes, TGF-β1, and leukemia inhibitory factor (LIF) (for review, see ref. 10).

Functional and pharmacologic studies

To understand the functional consequences of the presence of inflammatory mediators in the brain, two main experimental approaches have been taken: on the one hand, proinflammatory molecules have been injected into the rodent brain before, or together with, the convulsive stimulus to exacerbate the early endogenous production of these molecules during seizure activity; on the other hand, a chronic inflammatory state has been induced in brain either pharmacologically or by using trangenic mice overexpressing specific cytokines, and seizure susceptibility was then assessed.

Effects of proinflammatory molecules on seizures: the role of IL-1β and TNF

Seizures of focal onset: The preapplication of IL-1β in rodent brain, by using concentrations within the range of those endogenously produced by seizures, prolongs the duration of electrographic and behavioral seizures induced by intracerebral application of kainate or bicuculline methiodide (39). Importantly, the intracerebral injection of IL-1Ra has powerful anticonvulsant effects (53), and transgenic mice overexpressing IL-1Ra in astrocytes have a reduced susceptibility to seizures (42). Because the only action of IL-1Ra is to inhibit the effects of IL-1β, these data indicate that an endogenous increase in brain IL-1β contributes to the maintenance of seizures in these models.

This finding is reinforced by the evidence that impairment of the endogenous production of IL-1β by using selective blockade, or gene knockout, of caspase-1, the enzyme producing the biologically active form of IL-1β, significantly reduces seizures (54).

One exception to the evidence that IL-1β plays a proconvulsive role in the brain is provided by a study showing that daily intraventricular injections of IL-1β doses 100 times lower than those shown to exacerbate seizures significantly retarded the acquisition of amygdala kindling in rats (55). This evidence is in accordance with the inhibition of long-term potentiation (LTP), a form of neuronal plasticity that shares common mechanisms with kindling, by low doses of IL-1β (56).

The effect of TNF-α on seizures depends on its endogenous brain levels and the receptor subtypes predominantly stimulated by this cytokine. Injection of nanomolar amounts of mouse recombinant TNF-α into the mouse hippocampus reduces seizures, and this action is mediated by neuronal p75 receptors (47). Transgenic mice with low to moderate overexpression of TNF-α in astrocytes also show decreased susceptibility to seizures (47), whereas mice with high overexpression of TNF-α develop signs of neurologic dysfunction (57,58).

Fever-related seizures: Fever is a systemic response to infection, inflammation or stress and it can evoke febrile seizures in infants and children (see later). Long and repetitive febrile seizures (FSs) are closely linked to the development of mesial temporal lobe epilepsy (MTLE). However, the mechanisms by which fever induces seizures are still unknown. Several proinflammatory cytokines, including IL-1β, act as pyrogens after central or systemic administration. Two recent studies have addressed the possibility that the increase in brain IL-1β during fever may directly evoke seizures in immature rodent brain (59,60): intracerebroventricular injection of IL-1β reduces the seizure threshold in 14-day-old mice subjected to hyperthermia (59) or in 14-day-old rats exposed to LPS-induced fever (60); IL-1β receptor–deficient mice or IL-1Ra–injected rats were resistant to induction of this kind of seizure. In addition, relatively high doses of IL-1β evoked seizures in immature mice (59). These data suggest that IL-1β signaling contributes critically to fever-induced hyperexcitability underlying FSs.

Infection-related seizures: Systemic infection caused by Shigella dysenteriae in mice significantly increases their response to pentylenetetrazol (PTZ)-induced seizures (61,62). IL-1β and TNF-α play an important role in the sensitization of the CNS to infection-related seizures because systemic preinjection of inactivating antibodies against these two cytokines in the infected mice prevented the increase in seizure susceptibility (61). A subsequent study found either exacerbating or protective effects of TNF-α in this model, depending on its circulating levels (62).

In a neonatal rat model of pneumococcal meningitis, a large-spectrum inhibitor of TNF-α–converting enzyme reduced CSF-soluble TNF-α and decreased the incidence of spontaneous seizures in mice without affecting the bacterial growth in the CSF (63).

Mouse models of cerebral malaria (64) or cysticercosis (65) show various neurologic dysfunctions including seizures.

Effects of other cytokines

Limited information exists on the role of other cytokines in seizures. In brief, prolonged intraventricular infusion of relatively low doses of bFGF did not modify kainate-induced seizures in rats but effectively protected from the subsequent neuronal cell loss and raised the rat threshold to flurothyl-induced seizures (66). However, an acute intrahippocampal bolus injection of bFGF can induce seizures in rats (67). Acidic FGF-1 given systemically, both before and after kainate in rats, significantly attenuated both seizure activity and hippocampal damage (68).

IL-2 was shown to facilitate sound- and chemoconvulsant-induced seizures in DBA2 mice. In addition, relatively high doses of IL-2 cause recurrent, although transient, seizure activity (69,70).

Spike-and-wave discharges were observed after intraventricular infusions of IL-3, and epilepsy-like activity was described after in vitro application of interferon to neuronal cultures or hippocampal slices (70).

These findings indicate that the various proinflammatory cytokines studied in experimental models of acute seizures either decrease the threshold for seizure induction or increase the duration and severity of epileptiform activity. FGF, TNF-α, and IL-1Ra inhibit seizures and afford neuroprotection, depending on their concentration, their short- or long-term increase in the brain, or, for FGF and TNF-α, the receptor subtypes predominantly involved.

Chronic inflammation and seizure predisposition

The use of transgenic mice overexpressing IL-6 or TNF-α indicates that a chronic inflammatory state in the brain can predispose to the occurrence of seizures and neuronal cell loss. Thus transgenic mice overexpressing IL-6 in astrocytes showed an increased sensitivity to seizures induced by glutamatergic agonists and a constitutive loss of γ-aminobutyric acid (GABA)- and parvalbumin-positive neurons in the hippocampus, which may be implicated in their propensity to develop seizures (71). High overexpression of brain TNF-α and IL-6 in transgenic mice was associated with the occurrence of age-dependent neurodegenerative changes and sporadic spontaneous seizures (57,58,72).

Seizure susceptibility also was enhanced in mice by preadministration of LPS, and this phenomenon was blocked by antiinflammatory drugs (73).

Finally, seizures and neuronal cell loss were induced in rats by intrahippocampal sequential infusion of individual proteins of the complement cascade, leading to the formation of the complement membrane attack complex (MAC) (74).

Inflammation, glutamate, and GABA

It is well established that glutamate and GABA are crucially involved, in concert with other neurotransmitter and neuromodulator systems, in the initiation and propagation of seizures. In this respect, specific functional interactions exist between cytokines and PGs with glutamate- and GABA-mediated neurotransmission, which may constitute the key mechanism linking inflammation to epilepsy. For example, NMDA-receptors activation is involved in COX-2 induction in the brain (75) and in the proconvulsant effect of IL-1β on kainate seizures in rats (39). In addition, IL-1β inhibits glutamate reuptake by astrocytes (76) and increases glutamate release (77), a feature shared by TNF-α also (78), interacts at the molecular level with the NMDA receptors resulting in their enhanced function (28). IL-1β has been reported to inhibit GABAergic transmission (79,80). IL-1β, IL-6, and TNF-α modulate ionic currents both in neurons and in glia, and relatively low concentrations inhibit the induction of hippocampal LTP (56). However, IL-1β also was found to contribute to the maintenance of LTP in CA1 (81).

A direct interaction between TNF-α and AMPA receptors was recently demonstrated in hippocampal neurons. This cytokine, acting on p55 receptors, regulates the cellular trafficking of AMPA receptors by inducing their membrane expression in a molecular conformation that amplifies the glutamate responses. Conversely, TNF-α causes endocytosis of GABAA receptors, thus resulting in decreased inhibitory strength (27,82).

Because of the functional interactions between cytokines and classic neurotransmitters (83), an increase in their extracellular concentration may influence neuronal excitability and the brain response to injury.

Antiinflammatory drugs: effects on seizures and their sequelae

Table 1 summarizes the pharmacologic findings obtained by using antiinflammatory drugs in experimental models of seizures.

Nonsteroidal antiinflammatory drugs attenuate seizures induced by cortical application of penicillin in rats (84). Antiinflammatory drugs with anti-COX activities were shown either to reduce or to exacerbate seizures induced by kainic acid. For example, phenidone and the association of lipoxygenase inhibition and aspirin decreased kainate seizures, and the latter reinforced the anticonvulsant activity of sodium valproate in this model (85,86). However, in different studies, pretreatment of rats with indomethacin, aspirin, nimesulide, or selective COX-2 inhibitors augmented kainate-induced seizures (87,88). Indomethacin afforded protection against seizures triggered by tacrine, an anticholinesterase agent (89) but lowered the threshold and accelerated the onset of seizures induced by PTZ (90). These apparent dual effects of COX inhibitors likely depend on their specific actions on the basal production of the various PGs (87) and on the different profiles of PGs produced during seizures in the various experimental models (see earlier paragraph). For example, PG D2 has anticonvulsant properties, and a reduction of its brain levels may increase excitability and cell death (91,92). It is possible also that COX inhibitors augment seizures by other mechanisms than inhibition of PGs synthesis, such as uncoupling of the oxidative phosphorylation (93) or vasoconstriction (94).

Postseizure administration of COX-2 inhibitors or prostanoid-receptor ligands (95) protects hippocampal neurons from damage induced by pilocarpine seizures and enhances functional recovery of cognitive functions after kainate seizures, supporting a relevant role of PGs in the mechanisms of neurodegeneration (96).

GCs are potent inhibitors of the transcription of genes encoding most of the proinflammatory molecules, thus representing a critical endogenous negative-feedback system with antiinflammatory properties. In support of this evidence, the use of a GC-receptor inhibitor increased the inflammatory reactions induced by LPS in brain and enabled IL-1β and TNF-α to cause neurotoxic effects that were otherwise undetectable (97). However, prolonged elevation of GCs in the high physiologic range may exacerbate the damage induced by excitotoxins, and this effect is ascribed to their ability to induce a catabolically vulnerable state in neurons (i.e., inhibition of glucose uptake, glycogen synthesis, and protein synthesis) (41,98). This dichotomy likely explains the paradoxical proconvulsant effects of GCs on kainate seizures and their associated long-term events (99,100).

Finally, intravenously injected human globulin-N protected for 8 days against generalized seizures induced by electrical stimulation of the amygdala in fully kindled cats (101). Immunoglobulins have been shown to inhibit NF-κB activation induced by TNF-α in endothelial cells and macrophages, suggesting that their anticonvulsant effect may be at least in part mediated by antiinflammatory mechanisms (102,103).


From a clinical standpoint, a role of inflammation in the pathophysiology of human epilepsy is still hypothetical, although this possibility is supported by abundant evidence.

The first insights into a role of inflammation in epilepsy originate from the demonstrated antiepileptic activity of selected powerful antiinflammatory drugs, including steroids. Moreover, several reports showed increased markers of inflammation in serum, CSF, and brain resident cells in patients with epilepsy. For example, recent tonic–clonic seizures in epilepsy patients induce a proinflammatory profile of cytokines in plasma and CSF, consisting of higher IL-6 levels and lower IL-1Ra–to–IL-1α ratio (104–107). Because CSF IL-6 concentration is much higher than that measured in plasma (104–106), and the contribution of peripheral blood mononuclear cells (PBMCs) to increased plasma levels of cytokines is still unclear (104,108), the most likely origin of CSF cytokines appears to be the brain.

An increased expression of proinflammatory molecules has been demonstrated in neurons and glia in brain tissue obtained from patients surgically treated for drug-resistant epilepsies (109–113). In particular, the demonstration that inflammatory reactions occur also in epilepsy disorders that do not feature an inflammatory pathophysiology, such as temporal lobe epilepsy (TLE) or tuberous sclerosis (TS), raised the possibility that inflammation in the brain may be a common factor contributing or predisposing to the occurrence of seizures and cell death, in various forms of epilepsy of different etiologies.

Different genetic programs tightly regulate the inflammatory response, and among these, it has been reported that homozygosity in allele variants of the IL-1β gene promotes enhanced cytokine production (114). The analysis of IL-1β, IL-1α, and IL-1Ra gene polymorphisms (115) in a cohort of drug-resistant epilepsy patients versus healthy controls suggested an association between cytokine gene haplotypes and development of intractable focal seizures, but this observation was not confirmed by others (116,117). These discrepancies may be at least in part due to ethnic differences between these studies and highlight the need for further investigations to confirm this evidence unequivocally.

Clinical studies highlight the role of stress hormones in epilepsy, and it is well known that stress is not simply linked to immunosuppressive and antiinflammatory effects, but it rather results in profound changes in immunocompetence [see details earlier (23)]. Increased adrenocorticotropic hormone (ACTH) and cortisol levels, end products of the stress response, have been reported in human epilepsies and in status epilepticus (21,118,119). Activation of the ACTH–GC axis in response to antecedent injury or stress, leading to hyperfunction of CRH–neuronal pathways, has been suggested to play a role in the pathogenesis of West syndrome (120).

Another important aspect linking inflammation to epilepsy concerns the effects of both seizures and inflammatory mediators on the BBB function (see earlier paragraph). BBB dysfunction in epilepsy has been demonstrated in experimental models (13) and in human epilepsy tissue (121). In clinical practice, however, BBB disruption is more often suspected than documented: postcontrast signal abnormalities and focal single-photon emission computed tomography (SPECT) changes consistent with the location of slow-wave or epileptic EEG activity have been only occasionally reported in patients with partial status epilepticus (122,123) or focal epilepsy (124,125). CSF albumin, as an index of increased BBB permeability (126) and pleiocytosis (107,127), has been seldom reported in epilepsy patients, although one study showed a correlation between enhanced CSF albumin levels and EEG slow-wave activity in areas of robust BBB dysfunction (128). In most cases, routine examination does not provide evidence of BBB disruption: CSF analysis is usually normal in epilepsy patients (129,130), and postcontrast CT and MRI do not reveal BBB damage, even in Rasmussen encephalitis (RE), where inflammation is demonstrated and BBB damage strongly suggested (131). The lack of systematic studies relating imaging techniques and CSF measurements of serum components in epilepsy patients are needed to evaluate the sensitivity of these approaches in detecting even subtle BBB abnormalities.

Increased BBB permeability and changes in endothelial transporters function induced by seizures or by inflammation or by both have several clinical implications, which include (a) BBB leakage allows the entry of compounds with immunogenic or inflammatory potentials, as predicted for example in interferon-induced seizures (125) or in RE; (b) upregulation in the expression of multidrug transport proteins may limit antiepileptic drug (AED) access to the brain, thus contributing to drug resistance (132), but it also may facilitate the entry of compounds with therapeutic potential with limited or no access to the CNS, such as immunoglobulins.

The following sections of this review focus first on the description of various forms of human epilepsy in which CNS inflammation and markers of adaptive immunity have been described; second, they offer an overview of the clinical evidence showing the efficacy of antiinflammatory approaches in seizure control (Table 2).

Table 2. Inflammation in human epilepsies and convulsive disorders
Epileptic syndrome
Convulsive disorder
Inflammatory markersAntiinflammatory treatments
Plasma or CSFBrain tissue
  1. *Only in patients with MTLE and HS. n.d., not determined; PEX; plasma exchange; PAI; protein A immunoabsortion; IVIg, intravenous immunoglobulin; MAPK, mitogen activated protein kinase.

Rasmussen encephalitisGluR3 Ab, Munc-18 AbGluR3 AbACTH, steroids, IVIg
CD-8+ lymphocytes
GrB; MAC; cytokines
PEX, PAI, immunosuppressant
West syndromeIFN-alpha, TNF-alpha, IL2n.d.ACTH, steroids, IVIg
Lennox–Gastaut syndromen.d.n.d.ACTH, steroids, IVIg
Landau Kleffner syndromen.dn.d.ACTH, steroids, IVIg
Febrile seizuresIL-1beta, IL-1Ra, IL-6,
IL-10, TNF-alpha
TLEIL-6, IL-1beta, IL-1RaIL-1, NFkB*n.d.
Tonic-clonic seizuresIL-6, IL-1alpha, IL-1betan.d.n.d.
Tuberous Sclerosisn.d.CD-68 macrophages
ICAM-1, TNF-alpha,

Our attempt is to bridge the experimental work in rodent models of seizures with clinical evidence to discuss the possibility that inflammatory reactions in the CNS may be involved in some step in the etiopathogenesis of seizures and their long-term consequences.

Rasmussen encephalitis

RE is an acquired progressive disease characterized by intractable focal seizures and neurologic and cognitive deterioration, resulting from dysfunction of one hemisphere (131). The etiopathogenesis of RE is still not fully understood; nevertheless, it has been considered a chronic inflammatory disease since its original description in 1958. This definition stems from the features of its clinical course, histopathology, and the reported efficacy of immunomodulatory treatments in delaying the disease progression (133–136).

Histopathology demonstrates the progression of an inflammatory process from an early phase consisting of astrocytic and microglia reactivity and mild lymphocyte infiltration, without significant evidence of neuronal or cortical cell loss, to progressively severe stages in which neuroglial cell reactivity and the magnitude of lymphocytic infiltration are greatly increased, and a significant decrease in neuronal populations occurs. In the end stages of cortical disease, extensive destruction of the cerebral cortex and cortical vacuolation dominate, residual astrogliosis and foci of activated microglia are observed, while minimal or absent lymphocyte infiltration remains (136). Different stages of inflammation coexist in the same patient with a multifocal distribution, which is consistent with an ongoing and progressive immune-mediated process (136,137).

CD8+ T cells are located in close apposition to degenerating neurons, and they may contribute to neuronal cell death via the secretion of granzyme B, a strong activator of caspase-mediated apoptosis (138–140). Activated CD4+ T cells may also prime B cells to produce autoantibodies. After the original findings of antibodies against glutamate-receptor subunit 3 (GLUR3), at least two other autoantibodies have been measured in the plasma of RE patients (141–144). The generated autoantibodies may destroy neurons either directly by excess stimulation of receptor-mediated ion channels, or indirectly by binding complement factors and leading to the formation of the MAC (145,146), which can induce neuronal loss and seizures (74).

Although the pathogenic effects and specificity of these antibodies are still a matter of investigation, the efficacy of selective removal of circulating immunoglobulins in RE patients, with or without measurable antiGluR3 antibodies (147), points to a role of an autoimmune response in the disease pathogenesis.

The role of seizures in sustaining the disease progression must be emphasized, because this aspects confers on RE the clinical features of an epileptic encephalopathy. How seizures can contribute to the disease progression is still a matter of debate. Epileptic activity may induce inflammatory mediators in microglia, astrocytes, neurons, and endothelial cells (see earlier), and alter the properties and permeability of the BBB, thus facilitating the entry of components of the adaptative immune system and molecules usually excluded from the brain parenchyma. These phenomena may consolidate and perpetuate inflammatory reactions in the brains of the affected individuals and exacerbate brain damage (148), thus contributing to brain atrophy.

It is still unknown, however, which is the initial trigger in this hypothetical pathologic cascade of events: viral infection (149), head trauma, or even a first seizure per se, as suggested by findings in experimental models of seizures (see earlier), can be envisaged as possible precipitating conditions. In this respect, structural or acquired brain lesions also may play a role, because these conditions are known to induce inflammation (2,9,150). Patients affected by a preexisting brain lesion account for ∼10% of RE cases (151). This “dual pathology” has been documented in low-grade tumors, cortical dysplasia, tuberous sclerosis, vascular malformations, or old ischemic lesions (131,152–154).

An epileptic encephalopathy with pathologic features resembling RE has been also reported in two siblings showing rapidly progressive alternating epilepsia partialis continua. Although the precise etiology of this condition was unknown, microglial activation and perivascular lymphocytes were described in the brains of these children, suggesting a possible role of inflammation in the epileptic process (155).

A similar clinical picture of bilateral multifocal seizures, progressive neurologic deterioration, and diffuse brain atrophy can be observed in children in whom a precise diagnosis is lacking despite extensive laboratory investigations (156,157) and in patients with subtle diffuse abnormality of cortical architecture, undetectable by MRI, and identified on neuropathologic grounds (personal observation). Interestingly, high titers of anti-GLUR3 antibodies were measured in infants with catastrophic epilepsies, either symptomatic or cryptogenic (158). Although these antibodies may merely represent an epiphenomenon of diffuse cerebral damage, one cannot exclude their putative role in sustaining epileptic activity and contributing to neurodegeneration. Further studies are warranted to better define the role of inflammation in epileptic encephalopathies other than RE.

Temporal lobe epilepsy

The term temporal lobe epilepsy (TLE) refers to both lesional and nonlesional epilepsies characterized by focal seizures arising from either the neocortex or the mesial temporal structures. One of these syndromes is MTLE associated with hippocampal sclerosis (HS), a histopathologic picture characterized by neuronal loss and gliosis, predominantly involving the CA1 region and the dentate gyrus. An increased IL-1α expression in microglia-like cells has been documented by immunohistochemistry in brain specimens obtained from patients surgically treated for TLE (109).

A subsequent study described the expression of NF-κB in reactive astrocytes and surviving neurons in hippocampal specimens surgically removed from patients with MTLE and HS who had a history of febrile convulsions (110). This evidence indicates that the activation of an inflammatory cascade in brain parenchymal cells, as predicted by increased expression of NF-κB, specifically occurs in lesioned hippocampi. The presence of NF-κB in surviving neurons raises the possibility that this factor plays a neuroprotective role in these cells.

Indeed, the inactivation of NF-κB before administration of kainate results in a significant increase in the extent of pyramidal neuron death in the rodent hippocampus (159). Moreover, seizure-induced neuronal degeneration is increased in mice with a defective NF-κB system (160). In general, it appears that the activation of NF-κB in neurons protects them against degeneration, whereas its activation in glia may promote neuronal degeneration (160).

In summary, the studies in human epilepsy specimens from TLE patients support the existence of a chronic inflammatory state sustained by microglia, astrocytes, and neurons in the epileptic focus, although it remains to be established whether inflammation is strictly associated with cell death, is induced by seizures, or both.

Insights into these aspects are given by the findings obtained in experimental models of seizures indicating that (a) epileptic activity per se is sufficient to induce a prominent inflammatory response in the brain; (b) seizure-induced inflammation, although outlasting seizure duration, is reversible, thus anticipating that the time elapsed from the last seizure may determine whether inflammatory markers are detectable in the CNS; and (c) inflammation occurs to a larger extent when seizures are associated with cell damage.

A relevant but still unresolved issue in epileptic disorders that do not feature a typical inflammatory pathophysiology, such as TLE, is whether inflammation plays a role in epileptogenesis and in the progression of the disease. Because chronic inflammation in rodent brain can predispose to seizures and cell death, alter the BBB permeability, affect synaptic plasticity and neurogenesis, a potential epileptogenic role of inflammation is certainly conceivable also in humans.

Febrile seizures

FSs are the most common convulsive disorder in childhood and are to be considered, in most cases, a benign age-related disorder. Nevertheless, the risk of developing epilepsy is higher in infants with FSs than in the general population, and long and repetitive FSs are thought to be involved in the genesis of MTLE with HS (161).

Although the pathophysiology of FSs is unknown, cytokines have been proposed to play a role. Particular attention has been focused on IL-1β because this cytokine can induce fever by stimulating the production of PGs in the hypothalamus (44), and fever of any cause induces IL-1β in brain microglia (162). Moreover, exogenous application of IL-1β reduces seizure threshold in immature mice (59,60).

Plasma and CSF levels of cytokines have been studied in patients with FSs, but variable results have been reported. IL-6 and IL-1Ra–to–IL-1β ratio were increased in plasma of children with FSs compared with those in children with febrile illness without convulsions (163). In a different study, IL-1β and TNF-α were increased in serum and CSF, respectively (164). CSF levels of IL-1β were found to increase after FSs by Haspolat et al. (165), whereas no changes in TNF-α were detected. Other groups were unable to detect significant changes in the serum or CSF levels of IL-1β, even when measured shortly after the occurrence of FSs (166–168). The reasons for these variable results may depend on various factors, including the time elapsed between the last seizure and the sampling because of the rapid half-life of the cytokines, or the control population used for comparison.

Only one study reported cytokine gene polymorphisms in FSs (169): these authors showed an increased frequency of biallelic polymorphism in the promoter region of the IL-1β gene at the –511 position (IL-1β-511) in children with FSs as compared with healthy control subjects. An increased frequency of the same polymorphisms was reported in patients with TLE and HS versus TLE patients without HS (170). In this study, TLE with HS was far more likely to be preceded by prolonged FSs than was TLE without hippocampal damage. Because polymorphisms in the IL-1β gene can influence the cytokine production both under physiologic conditions and after appropriate stimuli (114,171), it is conceivable that minor events during brain development, such as FSs, may trigger in predisposed individuals a cascade of inflammatory reactions resulting in epilepsy and the associated neuronal damage.

Tuberous sclerosis

Tuberous sclerosis (TS) is a genetic disease clinically characterized by mental retardation, autism, and, in >75% of cases, by seizures. Epilepsy is related to the presence of cortical tubers (i.e., focal cortical malformations characterized by disorganized lamination, dysplastic neurons, and giant cells). The expression of a large array of proinflammatory molecules has been investigated in tubers both at the mRNA and protein levels (111). This study demonstrates an increased number of CD68-immunoreactive macrophages adjacent to giant cells and an overexpression of ICAM-1, TNF-α, mitogen-activated protein kinase (MAPK), and NF-κB in cellular components of tubers, particularly in giant cells and astrocytes. Proinflammatory molecules and CD68 immunoreactivity were not observed in perituberal cortex or control cortex without histologic abnormalities. An increase in ICAM-1 also was shown in reactive astrocytes and cytomegalic cells in neocortex and hippocampus of Tsc1 knockout mice, an experimental model of TS that reproduces some of the cellular features of tubers. Importantly, Tsc1 knockout mice exhibit spontaneous seizures by age 2 months, suggesting the possibility that proinflammatory signals in tubers contribute to epileptogenesis.

Finally, a recent study (172) reports increased expression of macrophage chemoattractant protein (MCP-1) and TGF-1β in brain specimens from autistic children. These findings suggest a potential role of inflammatory response in the pathogenesis of behavioral disorders in TS patients and increase the possibility that immunomodulatory treatments may be efficacious also in this respect.

West syndrome

West syndrome (WS) is an age-related epileptic encephalopathy with onset in the first year of life, featuring clustered spasms and hypsarrhythmia. It may occur in previously healthy children (cryptogenic WS) but more frequently is a symptom of different congenital or acquired diseases (symptomatic WS). Independent of its etiology, WS patients mostly benefit from steroid treatment. Steroid efficacy, together with the possibility of spasm disappearance after viral infections (173), has long been considered an index for an inflammatory or immunomediated pathogenesis. The recent report of increased serum levels of IL-2, TNF-α, and IFN-α in both cryptogenic and symptomatic WS reinforces this hypothesis. These cytokines are produced by monocytes and lymphocytes, and TNF-α also by brain glial cells, and they all have effects that may contribute to seizures and neuronal cell damage.

The presence of proinflammatory molecules in both cryptogenic and symptomatic WS patients suggests that cytokine changes are likely to be related to epilepsy rather than to the underlying etiology. However, symptomatic patients displayed a greater elevation of IL-2 levels, which varied depending on the underlying disorder (174). If the extent of the inflammatory reaction depends also on the underlying disease, this could in part explain the different efficacies of steroid treatment in selected etiologic subgroups of symptomatic WS patients.

Anticonvulsant activity of antiinflammatory drugs

The anticonvulsant efficacy of immunomodulatory drugs with antiinflammatory actions has been well documented. In particular, ACTH and steroids are the most effective treatments for seizure control in WS and other epileptic encephalopathies featuring a high seizure frequency (135,175–178). The mechanism of action of these drugs is complex and incompletely defined, thus impairing the possibility of understanding fully which are the crucial pathways involved in their anticonvulsant effects. Acute seizure inhibition may involve in part the ability of steroids to modulate various neurotransmitters, including GABA (179,180). However, the anticonvulsant action of steroids and ACTH often lasts after drug discontinuation, thus suggesting that besides acute seizure control (180,181), steroids may reset a deranged homeostatic mechanism in the brain, increasing its refractoriness to seizure recurrence (182). Different mechanisms of actions have been proposed to explain this phenomenon: (a) ACTH and steroids may accelerate brain maturation, and this effect would explain their efficacy in age-related syndrome such as infantile spasms; and (b) they reduce the production and release of CRH in specific brain regions, a peptide that is known to cause severe limbic seizures in developing rodents and promotes the late onset of chronic hyperexcitability (183,184). Their immunosuppressant and antiinflammatory activities may play a major role in epilepsies with an immune component, such as RE, or in WS.

Intravenous administration of high doses of immunoglobulins has an anticonvulsive effect in selected human epilepsies (135,178,185–187) and in experimental models (101). The anticonvulsive properties of immunoglobulins suggest that the presence of circulating or brain antigens may contribute to seizure generation and recurrence in these epileptic syndromes.

Reports on the anticonvulsant effects of nonsteroidal antiinflammatory drugs in human epilepsy are still anedoctal, although a transient reduction of seizure frequency during the antiinflammatory treatment is occasionally observed in current practice.

Finally, evidence exists of antiinflammatory actions of some AEDs, such as valproate (188) and carbamazepine (189). In rat glial cells or human glioma cells, these AEDs decrease the LPS-induced activation of NF-κB and the LPS-induced production of nitric oxide and PGs. It remains to be shown whether these nonconventional effects of AEDs significantly contribute to their anticonvulsant activity.


Experimental and clinical studies showed that various mediators of inflammation are present in brain, CSF, and blood in epileptic conditions. In particular, the histologic analysis of human brain from individuals with epilepsy of various etiologies strongly suggests the existence of a chronic inflammatory state in the brain almost invariably associated with neuronal loss, reactive gliosis, or malformations of cortical architecture. This observation, together with reports that antiinflammatory drugs have anticonvulsant efficacy in some cases of drug-refractory epilepsies, suggests the possibility that chronic inflammation in the brain may be implicated in the etiopathogenesis of seizures and the associated long-term events.

This hypothesis is supported by functional studies in experimental models of seizures, showing that some proinflammatory molecules exacerbate seizures, decrease the threshold for inducing convulsions, or cause seizures per se.

Increased susceptibility to seizures and sporadic spontaneous seizures in mice overexpressing relatively high amounts of cytokines highlights the possibility that a chronic inflammatory state can be epileptogenic per se.

The consensus is that a transient and strictly controlled inflammation in the brain, as occurring during endotoxemia, represents an adaptive beneficial response aimed at protecting the brain from noxious events; however, if inflammation in the brain is chronic or inappropriately controlled, it may become detrimental to neurons, thus representing one of the various maladaptive changes induced in the CNS by epileptic activity or by a preexisting brain pathology.

Initiating events in the inflammatory cascade

What remains unknown is the initial trigger of an inflammatory response in the brain when the original description of the disease does not include the presence of specific pathogens, such as viral particles or bacteria.

Experimental evidence in rodents clearly shows that a large variety of insults (Fig. 4) can induce inflammation in the brain (2,9), suggesting that an injury, even if subtle, occurring at birth or during the lifetime may initiate a cascade of chronic inflammatory events in the CNS that contributes to setting the basis for the late onset of epilepsy.

Figure 4.

Schematic representation of the hypothetical cascade of inflammatory events that may be triggered by brain injury, eventually resulting in epilepsy. A large variety of injuries (2,9), even if subtle, occurring at birth or during a lifetime may initiate a cascade of chronic inflammatory events in the brain that contribute to setting the basis for the late onset of epilepsy. We envisage that both innate and adaptive immune responses can play a role in initiating and consolidating inflammation in the brain. The blood–brain barrier is crucially involved in mediating the recruitment of cell components of the adaptive immune system, and alterations of the barrier properties may predispose the brain to the occurrence of seizures (190). The dichotomous role of inflammation is depicted by the scale, which is tilted toward deleterious effects in pathologic conditions (lightning). The beneficial role of inflammation may be restored by antiinflammatory strategies (sun).

Two main possibilities may explain the initiation of inflammatory responses within the CNS: (a) the immune/inflammatory challenge is initiated within the CNS, and the infiltration of blood-borne immune cells or circulating inflammatory mediators is a consequent response to this intrinsic event; or (b) the CNS is the target of an immune/inflammatory response that originates within peripheral lymphoid tissues.

The first scenario can be envisaged when structural abnormalities of brain development due, for example, to gene mutations in TS (111), or to cortical dysplasia (112), are associated with pronounced inflammatory reactions specifically localized in the brain tissue affected by the pathology. These inflammatory molecules may be induced by the cellular constituents of the malformed cortical region or reflect downstream effects of the gene mutations and may constitute an important component of hyperexcitabilty related to epileptogenesis in these lesions.

An intrinsic immune response also may be triggered in brain tissue against components of neurons or glia exposed by injuries, or in response to molecules normally excluded by the brain, which enter because of a leaky BBB; this may occur, for example, in some cases of symptomatic TLE.

The second scenario may occur when epilepsy evolves after systemic infectious diseases, in chronic inflammatory diseases such as RE, or in seizures associated with fever.

A genetic predisposition to develop sustained inflammatory reactions in response to otherwise ineffective stimuli also should be taken into consideration (i.e., gene polymorphisms).

The use of experimental models in which brain inflammation is triggered by intrinsic CNS events or by a systemic challenge will be helpful to study the sequelae of inflammatory events involving innate and adaptive immunity and their respective roles in epileptogenesis.

In conclusion, the involvement of inflammatory signals in the etiopathogenesis and sequelae of seizures is still conjectural, and its verification in humans is limited because the data available concern epileptic disorders of different etiology with varying clinical features. Moreover, pathologic data are available only from patients who underwent surgical treatment for focal, long-lasting drug-resistant epilepsy. For these reasons, with the exception of RE, no definite conclusions may be drawn on the role of inflammation in the early stages of epileptogenesis.

We hope that some insights into this issue will come in the future from the systematic study of (a) the expression of proinflammatory molecules in subsets of patients with focal epilepsy, either symptomatic or cryptogenic, which do not show an overt inflammatory pathology; (b) the profiles of various inflammatory markers in different body fluids evaluated at the onset of seizures and during the follow-up; (c) the definition of the genetic background regulating the expression of the inflammatory response in epilepsy patients; or (d) a thorough characterization of inflammatory markers in those cases in which trivial febrile illnesses or antiinflammatory drugs seem to modify seizure frequency.

If a relation between inflammation and epilepsy were proven, it might open new perspectives to the pharmacologic treatment of seizures and possibly to retarding epileptogenesis or delaying the progression of the disease.


Acknowledgment:  We apologize to the many authors whose contributions were not acknowledged because of space limitations. A.V. was supported by CURE, Telethon Onlus, Fondazione Mariani Onlus, Vertex Pharmac Inc.