Address correspondence and reprint requests to Albert J. Becker, M.D., Department of Neuropathology, University of Bonn Medical Center, Sigmund-Freud Str. 25, D-53105 Bonn, Germany. E-mail: email@example.com
Summary: Low-grade glioneuronal lesions involving tumors such as gangliogliomas and focal cortical dysplasias (FCD) predispose individuals to pharmacoresistant epilepsy. A frequent variant of FCD is composed of dysplastic cytomegalic neurons and Taylor-type balloon cells (FCDIIb). Those are similar to cellular elements, which are present in cortical tubers in the autosomal dominant inherited tuberous sclerosis complex (TSC). This phacomatosis is caused by mutations in the TSC1 or TSC2 genes. Recent data have indicated accumulation of distinct allelic variants of TSC1 also in FCDIIb. TSC1 represents a key factor in the phosphatidylinositol 3-kinase (PI3K) pathway. A variety of alterations in the PI3K-pathway have been recently reported in epilepsy-associated glioneuronal malformations. Here, we discuss pathogenetic similarities and differences between cortical dysplasias as well epilepsy-associated glioneuronal tumors and TSC-associated cortical tubers with a focus on PI3K-pathway components including ezrin, radixin and moesin (ERM), which represent downstream effectors involved in cytoskeleton-membrane interference. No evidence has been found for mutational events of ERM genes to play a major pathogenetic role in epilepsy-associated glioneuronal malformations. In contrast, aberrant expression of ERM proteins in FCDs and gangliogliomas was observed. These alterations may relate to compromised interactions of dysplastic cellular components in epilepsy-associated glioneuronal lesions and be involved in aberrant PI3K-pathway signaling in epilepsy-associated malformations. However, the underlying cause of PI3K-pathway activation and the functional relationship of PI3K-pathway activity to generation of seizures in epilepsy-associated glioneuronal lesions will need to be determined in the future.
Epilepsies represent frequent neurological disorders that affect approximately 1% of the population worldwide (Wiebe et al., 2001; Elger, 2002; Engel, 2003; Fisher et al., 2005). Focal lesions include glioneuronal malformations that contain focal cortical dysplasias (FCD) as well as low-grade glial and glioneuronal neoplasms such as gangliogliomas (Blümcke et al., 1999a; Becker et al., 2001; Urbach et al., 2003; Luyken et al., 2004; Palmini et al., 2004). These lesions share a variety of characteristics. They constitute highly differentiated malformations with generally benign biological behavior. The lesions reveal a distinct histological appearance, i.e., a biphasic composition of highly differentiated glial and often dysplastic neuronal components. They are generally associated with focal epilepsies.
Although epilepsy-associated malformations share a variety of aspects, which may suggest them as a clinico-pathological family, their classification and pathogenetic relationship is still under debate (Mischel et al., 1995; Barkovich et al., 2001; Tassi et al., 2002; Palmini et al., 2004). Current molecular analyses in glioneuronal lesions are based on their differentiation according to distinct neuropathological characteristics.
NEUROPATHOLOGICAL FINDINGS IN GLIONEURONAL MALFORMATIONS
Malformative lesions of the CNS comprise a wide spectrum of neuroradiological and histomorphological alterations (Mischel et al., 1995; Barkovich et al., 2001; Patel and Barkovich, 2002; Tassi et al., 2002; Urbach et al., 2002; Palmini et al., 2004), ranging from subtle architectural aberrations to substantial dysplastic lesions. Present classifications on cortical malformations are based on histological characteristics including loss of cortical lamination, glioneuronal and/or neuronal heterotopias, the occurence of dysplastic or cytomegalic neuronal elements and so-called “Taylor-type” balloon cells (Taylor et al., 1971; Mischel et al., 1995; Palmini et al., 2004). A general distinction is made between (1) pure malformations, i.e., lesions with highly differentiated, non-neoplastic glial and often dysplastic or immature neuronal components and (2) tumors with highly differentiated, nevertheless neoplastic glial components and often dysplastic neuronal elements.
Gangliogliomas represent the most frequent tumors in patients with focal epilepsy. These tumors account for 5% of brain tumors in childhood, but are rare in adults (Blümcke et al., 1999b). The composition of dysplastic neurons combined with neoplastic glial cells constitutes a histopathological hallmark of gangliogliomas. The neoplastic nature is given by the proliferative activity of the glial cell component. Nuclear labeling for the proliferating cell nuclear antigen Ki-67 is generally observed exclusively in the astrocytic component (Wolf et al., 1994). The focal nature of gangliogliomas, their differentiated glioneuronal phenotype and the benign clinical character point to an origin from developmentally compromized or dysplastic precursor lesions (Blümcke et al., 1999b). The stem cell epitope CD34 is highly expressed in gangliogliomas (Blümcke et al., 1999b). By in situ reverse transcription combined with laser microdissection and PCR detection, CD34 expression is mainly detected in dysplastic neuronal components in gangliogliomas (Blümcke et al., 1999b; Fassunke et al., 2004b). However, CD34 expression is also found in a subpopulation of balloon cells of FCDIIb (Fauser et al., 2004). A second major entity within the group of glioneuronal tumors is constituted by the dysembryoplastic neuroepithelial tumors (DNTs), composed of “floating neurons” and oligodendroglia-like glial elements (Daumas-Duport et al., 1988; Daumas-Duport, 1993; Wolf et al., 1997). Furthermore, a recently described subtype of epilepsy-associated astrocytoma with a striking monomorphic appearance has been termed as “isomorphic variant” and is characterized by a particularly long patient survival (Luyken et al., 2003).
In contrast to the tumors characterized above, FCDs have no proliferative cellular component. A frequent subtype has been characterized as FCD of Taylor's balloon cell type (FCDIIb) (Taylor et al., 1971), with striking similarities to cortical tubers in patients with tuberous sclerosis (Wolf et al., 1995b). Cortical tubers show a derangement of the cortical laminar structure (Fig. 1), i.e., they manifest as multiple nodules and are microscopically characterized by disruption of the hexalaminar cortical organization, blurring of grey-white matter borders, calcification as well as occurrence of dysplastic neurons and giant cells similar to balloon cells present in FCDIIb. Due to the recent progress in magnetic resonance imaging (MRI), malformations of the cerebral cortex can increasingly be recognized during the presurgical evaluation of pharmacoresistant epilepsy patients. High resolution MRI allows the topographical characterization of the lesion with respect to size, localization and extension (Palmini and Lüders, 2002; Urbach et al., 2002).
In contrast to FCDIIb, FCDIIa consist of dysplastic neuronal components characterized by aberrant orientation and enlarged cell size but lack balloon cells (Fig. 1). In contrast to FCDIIb, FCDIIa, and gangliogliomas share a preference for temporal localization (Blümcke et al., 1999a; Palmini et al., 2004).
MOLECULAR PATHOLOGICAL ALTERATIONS IN GLIONEURONAL MALFORMATIONS: THE ROLE OF PHOSPHATIDYLINOSITOL 3-KINASE (PI3K) PATHWAY COMPONENTS
Similar histologies between epilepsy-associated circumscribed lesions and CNS abnormalities in patients with familial syndromes may argue for a potential overlap with respect to molecular pathways involved in their pathogenesis. Genes with pathogenetic relevance in other brain tumors, e.g., p53, EGFR, do not play substantial roles in the molecular pathology of epilepsy-associated malformations (Duerr et al., 1998; von Deimling et al., 2000).
With respect to a potential pathogenetic role of TSC1/TSC2 in epilepsy-associated glioneuronal lesions we have analyzed whether aberrant patterns of allelic variants in these genes are observed in sporadic FCDIIb (Becker et al., 2002), gangliogliomas (Becker et al., 2001) as well as FCDIIa (Majores et al., 2005a). In a cohort of 48 FCDIIb, two-third of the specimens shows sequence alterations in TSC1. Two sequence alterations affecting exons 5 and 17 result in amino acid exchange of the TSC1 gene. Intriguingly, the base transition in exon 17 (2415C>T; His732Tyr), a complex exon 14/intron 13 polymorphism and a silent polymorphism in exon 22 of TSC1 (3050C>T; 943Ala) are all significantly increased in FCDIIb compared to controls (Fig. 2) (Jones et al., 1997; Dabora et al., 1998; van Slegtenhorst et al., 1999). The base exchange in exon 17 is localized in the interaction domain of hamartin with tuberin, suggesting a potentially compromised tumor suppressor function of the resulting protein (van Slegtenhorst et al., 1998). Moreover, 11 of 15 FCDIIb specimens analyzed in a loss of heterozygosity (LOH) analysis at the TSC1 locus reveal LOH in the chromosomal region 9q34 of TSC1 in multiple microsatellite markers. Chromosomal instability in FCDIIb is observed mainly at the 9q area but not at other localizations of the genome (Fassunke et al., 2004a). With respect to the two-hit hypothesis for the inactivation of tumor suppressor genes, i.e., LOH and associated mutation in the second allele (Knudson, 1996), the observed combinations of LOH at the TSC1 locus and sequence polymorphisms in the second allele may suggest that the latter can act as a predisposing germ-line variant with low penetrance and a rather restricted manifestation pattern. In the light of the increasing information on hamartin and tuberin as cell cycle regulating complex (Potter et al., 2001; Tapon et al., 2001), such variant alleles can induce proliferation activity only at a narrow time span in brain development.
In contrast to FCDIIb, mutational analysis in 20 ganglioglioma specimens demonstrates abundant sequence alterations of the TSC2 gene (Becker et al., 2001). An intronic polymorphism in intron 4 and a silent polymorphism located in exon 40 of TSC2 are substantially abundant in gangliogliomas compared to controls (Platten et al., 1997; Becker et al., 2001). In contrast to FCDIIb, in gangliogliomas no polymorphisms in TSC1 are substantially increased (Becker et al., 2001). In particular, the exon 17 polymorphism frequently encountered in FCDIIb is not observed in gangliogliomas. In a ganglioglioma, a sequence alteration observed in intron 32 represents a somatic mutation. Subsequent laser-capture microdissection of glial and dysplastic neuronal components shows the mutation to be present only in the glial cell component. This observation points towards a clonal origin of the glial tumor component in this case. The finding underlines the hypothesis that gangliogliomas develop from a dysplastic precursor lesion by neoplastic transformation of the glial cellular fraction.
Recent data have suggested that hamartin and tuberin constitute a tumor suppressor mechanism (Dan et al., 2002), which plays a central role in the insulin/PI3K-signaling pathway (Ito and Rubin, 1999; Kwiatkowski, 2003). The PI3K-pathway is critically for cell size- and growth-control, cortical development and neuronal migration (Potter et al., 2001; Tapon et al., 2001). Binding of insulin to its membrane receptor activates the cascade components PI3K, Akt, TSC1/TSC2, mTOR (mammalian target of rapamycin), and the transcription factors p70S6kinase (S6K) and ribosomal S6 protein (S6) (Fig. 3) (Gao et al., 2002; El-Hashemite et al., 2003). Inactivation of the TSC1/TSC2 complex by phosphorylation through Akt results in phosphorylation of mTOR and subsequent activation of the outlined transcription factors interfering with cell size control (Inoki et al., 2002; Kenerson et al., 2002; McManus and Alessi, 2002). Hamartin and tuberin regulate cellular differentiation, migration, cell cycle and size by interaction with numerous different molecules (Ito and Rubin, 1999; Kwiatkowski, 2003). Tuberin includes a region homologous to the GTPase-activating protein (GAP) for the small-molecular-weight-GTPase Rap1 (The European Chromosome 16 Tuberous Sclerosis Consortium, 1993). A rabaptin-5 binding domain has been reported in the vicinity of the GAP-related domain at the C-terminal region of tuberin (Xiao et al., 1997). A tuberin-rabaptin-5 complex is involved in the regulation of endocytosis (Xiao et al., 1997). Furthermore, interaction of tuberin and hamartin with CDK1 and cyclin B1 as well as of tuberin with cyclin B1 have been reported compatible with an important role of the hamartin/tuberin complex in cell cycle control (Fig. 3) (Catania et al., 2001). Recent findings have also pointed towards a link of the TSC complex with the band-4.1 superfamily of membrane-cytoskeleton-linking proteins ERM (Tsukita and Yonemura, 1999).
Are there pathogenetic similarities between TSC and FCD that may contribute to explain the pathogenesis of these conditions? Overlap in allelic distribution patterns between entities may suggest common pathogenetic mechanisms and support recent clinicopathological classification systems (Mischel et al., 1995; Tassi et al., 2002; Palmini et al., 2004). Whereas significantly increased genomic polymorphisms for TSC1 are present in FCDIIb (Becker et al., 2002), accumulation of allelic variants only of TSC2 is found in FCDIIa samples (Fig. 2). The coding polymorphism in exon 17 of TSC1 constitutes a highly pronounced classifier between FCDIIb and FCDIIa patients. The data argue in favor of the concept that different pathogenetic events are present at least in FCDIIb and FCDIIa. Also clinical observations with respect to a favorable postsurgical outcome in FCDIIb patients may support this concept (Urbach et al., 2002). Certainly, these results in epilepsy-associated glioneuronal lesions are in contrast to TSC-associated lesions such as cortical tubers, where mutations in TSC1 or TSC2 are frequent. A further limitation of these genetic studies is given by the fact that correlations of allelic variants and potentially aberrant protein expression of hamartin and tuberin in epilepsy-associated glioneuronal lesions are not yet available. Distinct similarities between TSC-related cortical tubers and epilepsy-associated FCDIIb are observed with respect to alterations of the PI3K pathway. In giant cells in cortical tubers with mutations in TSC1 or TSC2, inhibition of the PI3K-pathway is compromised and results in extensive pathway activation downstream of the ablated tumor suppressor mechanism, namely increased presence of phospho-S6, phospho-S6K and its targets phospho-STAT3 and phospho-4EBP1 (Baybis et al., 2004; Miyata et al., 2004). Individual PI3K-pathway downstream components are, however, activated also in FCDIIb, i.e., the eukaryotic translation initiation factor (eIF) 4G and phospho-S6 (Baybis et al., 2004; Miyata et al., 2004). Although different components of the PI3K-pathway have been shown to be activated in cortical tubers and FCDIIb, involvement of the same pathway may reflect pathogenetic similarities between FCD and cortical tubers. However, the underlying mechanisms of PI3K-pathway activation in FCDIIb have not yet been determined. Furthermore, it has to be considered that contaminating effects, e.g., of seizure activity on protein expression and/or phosphorylation of distinct PI3K-pathway components in glioneuronal lesions cannot be definitely ruled out. Nevertheless, analysis of additional PI3K-pathway components in epilepsy-associated glioneuronal malformations may contribute to elucidate further pathological alterations in these lesions.
ALTERATIONS OF EZRIN, RADIXIN AND MOESIN IN GLIONEURONAL MALFORMATIONS
ERM are part of the band 4.1-protein superfamily and share cross-linking activities between actin filaments and plasma membranes. ERM proteins have a role in physiological and pathophysiological cell growth control (McClatchey, 2003). ERM proteins are involved in the downstream compartment of the PI3K-pathway (Fig. 3), i.e., it has been shown that hamartin interacts with the ERM family of actin-binding proteins (Lamb et al., 2000). Inhibition of hamartin in cells with focal adhesions results in loss of cell-matrix adhesion whereas overexpression results in activation of Rho, assembly of actin stress fibers and the formation of focal adhesions. In the distal part of the growth cone, hamartin overlaps with ERM proteins and interacts with moesin (Haddad et al., 2002). Moreover, stably expressed full-length human tuberin increases cell adhesion in MDCK and ELT3 cell types, and decreases chemotactic cell migration in ELT3 cells (Astrinidis et al., 2002). In cortical tubers, ezrin and moesin are upregulated and colocalize with tuberin and hamartin within dysplastic neurons and abnormal giant neuroglial cells (Johnson et al., 2002). At the functional level, ERMs are essential for the generation of cell-surface structures such as microvilli (Takeuchi et al., 1994; Paglini et al., 1998). Ezrin and radixin have been previously demonstrated to be present in the adult CNS and specifically localized to astrocytes (Derouiche and Frotscher, 2001). During development, however, ERMs are expressed in neurons and have been shown to affect morphology, motility, and process formation of growth cones (Paglini et al., 1998).
Human ezrin (cytovillin) on chromosome 6q is highly similar, both in protein sequence and in functional activity, to merlin/schwannomin, a neurofibromatosis-2 (NF2)-associated tumor-suppressor protein (Majander-Nordenswan et al., 1998). This holds true also for radixin, which is located on chromosome 11q, and which has a strong homology to merlin (Wilgenbus et al., 1993; Hoeflich and Ikura, 2004). One unique case of NF2 and ganglioglioma has been described so far, arguing against a substantial contribution of NF2 in the pathogenesis of gangliogliomas (Sawin et al., 1999). Human moesin (membrane-organizing extension spike protein) is localized on chromosome Xq (Wilgenbus et al., 1994). The lack of gender predilection in gangliogliomas does not favor moesin as a potential candidate gene in gangliogliomas (Blümcke and Wiestler, 2002).
We have recently determined potential accumulation of allelic variants and altered protein expression and distribution of ERM in FCDs and gangliogliomas (Majores et al., 2005b). Sequence analysis of the ezrin and radixin genes showed only occasional polymorphisms (Majores et al., 2005b). These results strongly argue against mutational events of these genes to play a substantial role in epilepsy-associated tumors and FCDs. Cellular distribution patterns of ERM have been studied in FCDIIa, FCDIib, and gangliogliomas (Fig. 4). Compared to neocortical tissue without histological alterations as controls, aberrant labeling of ERM proteins is present in a high percentage of dysplastic elements in the different glioneuronal lesions. Immunoreactivity patterns are observed as granular cytoplasmatic staining compatible with a cytoskeleton-related cellular distribution of ERMs. We find no phenotypic differences between histological specimens of same patient cohorts (FCDIIa vs. FCDIIb vs. gangliogliomas), nor between dysplastic neurons obtained from different diagnostic entities. Apart from a diffuse and variable background staining previously described for ERM immunoreactivity (Derouiche and Frotscher, 2001), a reproducible specific expression of ERM proteins in normal CNS tissue adjacent to respective lesions or in the neoplastic glial component of gangliogliomas is not present. Balloon cells, which can be clearly distinguished from dysplastic neuronal components by opaque cytoplasm reveal a considerable range of ezrin, moesin and radixin immunoreactivity (Fig. 4). However, variation of staining intensities do not correlate with cortical or subcortical localization. The relative number of ERM-expressing balloon cells and dysplastic neurons is variable between individual FCDIIb specimens. Similar findings with pronounced expression of ERM proteins in dysplastic neurons are found in the present series of FCDIIa and gangliogliomas (Fig. 4).
Aberrant expression of the ERM proteins in dysplastic neuronal elements of gangliogliomas, FCDIIa and FCDIIb as well as balloon cells of FCDIIb point to common pathogenetic mechanisms, namely activation of the PI3K-signaling cascade, to be present in epilepsy-associated glioneuronal lesions. ERM expression has been critically related to migration and differentiation of neural precursors during brain development (Takeuchi et al., 1994; Paglini et al., 1998). Ezrin is abundantly expressed and developmentally regulated within radial glia and migration streams of the intermediate zone (Johnson et al., 2002). Axonal outgrowth of immature neurons is guided by highly dynamic growth cones and ERMs obviously play a crucial role in modulating membrane protrusion (Takeuchi et al., 1994; Paglini et al., 1998). Ezrin and moesin are co-expressed in dysplastic neurons and TSC-associated giant (balloon) cells (Johnson et al., 2002). Aberrant expression of ERM proteins constitutes a striking feature in balloon cells obtained from FCDIIb and in dysplastic neurons of other glioneuronal lesions (Majores et al., 2005b). However, toxic effects by upregulation of ERM can neither be completely ruled out nor be proven. With respect to the important role of ERM proteins for differentiation and migration of neural precursor cells, aberrant upregulation may contribute to impaired shape of cells, inadequate interaction with adjacent cells as well as aberrant positioning within and architectural disruption of the neuronal network. A dysplastic neuronal network may contribute to pathological signal transmission and predispose to seizures. However, it has been demonstrated that glioneuronal lesions and the perilesional zone highly express neurotransmitter-producing enzymes, neurotransmitter receptors as well as neuropeptides and calcium-binding proteins (Wolf et al., 1995a; Wolf et al., 1996). Therefore, the lesions as well as perilesional reorganization areas may actively contribute to the onset of epileptic seizures and the constitution of a hyperexcitable focus in the brain (Wolf et al., 1995a; Wolf et al., 1996; Cepeda et al., 2003).
With respect to substantial neuropathological similarities between epilepsy-associated glioneuronal malformations and cortical tubers in TSC, it has been postulated that FCDs and gangliogliomas may also pathogenetically be related to TSC. Although recent data suggest PI3K-pathway activation in FCDs, gangliogliomas and TSC-associated lesions, distinct patterns of activated elements within the PI3K-signal transduction cascade are observed between FCDs and cortical tubers in TSC. Future efforts will have to unravel the origin of PI3K-pathway activation in cortical dysplasias as well as a potential functional relevance of increased PI3K-signaling with respect to generation of epileptic seizures.
Acknowledgment: Our work is supported by DFG (SFB TR3), BMBF, BONFOR and Deutsche Krebshilfe.