New insights into a spectrum of developmental malformations related to mTOR dysregulations: challenges and perspectives

Abstract

In recent years the role of the mammalian target of rapamycin (mTOR) pathway has emerged as crucial for normal cortical development. Therefore, it is not surprising that aberrant activation of mTOR is associated with developmental malformations and epileptogenesis. A broad spectrum of malformations of cortical development, such as focal cortical dysplasia (FCD) and tuberous sclerosis complex (TSC), have been linked to either germline or somatic mutations in mTOR pathway‐related genes, commonly summarised under the umbrella term ‘mTORopathies’. However, there are still a number of unanswered questions regarding the involvement of mTOR in the pathophysiology of these abnormalities. Therefore, a monogenetic disease, such as TSC, can be more easily applied as a model to study the mechanisms of epileptogenesis and identify potential new targets of therapy. Developmental neuropathology and genetics demonstrate that FCD IIb and hemimegalencephaly are the same diseases. Constitutive activation of mTOR signalling represents a shared pathogenic mechanism in a group of developmental malformations that have histopathological and clinical features in common, such as epilepsy, autism and other comorbidities. We seek to understand the effect of mTOR dysregulation in a developing cortex with the propensity to generate seizures as well as the aftermath of the surrounding environment, including the white matter.

Introduction

Distinction between aetiology and pathogenesis is primordial for understanding central nervous system (CNS) malformations. Aetiology implies the most fundamental causative factor, such as an infectious agent, a specific genetic mutation or an epigenetic teratogenic agent to which the fetus is exposed. But identifying the aetiology does not explain the mechanism by which tissue morphogenesis is altered, particularly during development and maturation. In the fetal CNS, pathogenesis is the altering effect of aetiology on developmental processes that lead to abnormal tissue architecture (e.g. lissencephaly; polymicrogyria) and at times to dysmorphic individual cells (dysmorphic megalocytic neurones and glial cells).

Neuroembryology and developmental neuropathology are the basis of pathogenesis revealed by morphological examination of tissues. At the macroscopic (‘gross anatomical’) level, abnormal organisation of brain tissue can often be discerned by modern imaging, as well as by neuropathological inspection of surgically resected brain tissue and at autopsy. Modern microscopic neuropathology consists not only of histological stains used for more than a century, but also the application of specific histochemical and especially immunocytochemical markers of cellular lineage, maturation, metabolic products including intermediate filaments, enzymes of the mitochondrial respiratory chain and neurotransmitters, including their enzymes of biosynthesis and degradation. In the case of neurones, there are markers of late maturation (e.g. neuronal nuclear antigen or NeuN and synaptophysin) and markers of early differentiating neurones such as calcium‐binding peptides (e.g. calretinin; parvalbumin; calbindin D28k) that are specific in identifying GABAergic neurones (Ulfig, 2002; Sarnat et al. 2015). Among astrocytes, transitory filaments such as nestin are demonstrable in their earliest differentiation; later, vimentin filaments including in radial glial cells of the cerebral mantle and Bergmann glia of the cerebellar cortex, and finally the vimentin filaments coexist with and eventually are replaced by the more mature glial fibrillary acidic protein (GFAP). Understanding normal developmental morphogenesis and cellular maturation is key to revealing pathogenetic mechanisms that may be altered by specific genetic mutations or epigenetic events such as fetal exposure to teratogenic neurotoxins, congenital infections and transitory or chronic ischaemic events. Again, timing is a key determinant factor in understanding cerebral dysgeneses.

Genetic expression is not always uniform and predictable for individual patients: genotype : phenotype correlation may vary greatly among individuals with the same fundamental genetic defect. Recent neuroembryological and genetic revelations have formed a basis for updating the International League Against Epilepsy (ILAE) scheme of Neuropathological Classification of the focal cortical dysplasias, together with considerable new experience gained since 2011 when the original scheme was first published (Najm et al. 2018).

Developmental processes in ontogenesis of the human nervous system

Many individual processes of nervous system morphogenesis and maturation can be identified. Some, such as the formation of the neural placode at gastrulation followed by neurulation, and migration of neural crest cells are early processes. Others, such as synaptogenesis and myelination, are late processes, but most developmental processes occur between these extremes and are simultaneous rather than sequential events: in particular, neuroblast migration, axonal pathfinding, organisation and rearrangement of the cortical plate from micro‐columnar to horizontal laminar architecture.

Organisation of the cortical plate

The histological architecture of the cortex in the first half of gestation is initially radial micro‐columnar, followed from about 22 weeks’ gestation with a superimposed horizontal laminar architecture, which eventually predominates in the 3rd trimester (Sarnat & Flores‐Sarnat, 2013). Nevertheless, radial columnar synaptic pathways continue in the cortex after maturation, despite the additional development of horizontal pathways between columns (Hubel & Wiesel, 1963; Mountcastle, 1997). The Reelin gene (RELN) expressed by Cajal‐Retzius neurones of the pre‐cortical plate plexus and later by the transitory subplate neurones is a key element in cortical architectural rearrangements (Nishikawa et al. 2002; Sarnat & Flores‐Sarnat, 2002). Thalamo‐cortical projections to the cortical plate also play a role in cortical plate organisation (Allendoerfer & Shatz, 1994). Synaptogenesis is not initiated within the cortical plate until 22 weeks’ gestation (Sarnat et al. 2010). Shortly thereafter, lamina‐specific genetic markers identify neuronal types (Hevner, 2007; Beaumont et al. 2012; Dachet et al. 2015). Furthermore, micro‐RNAs show that there also are lamina‐specific differences in astrocytes within the mature cortex and less such expression in the fetus, except for some scattered white matter and germinal matrix astrocytes that show greater maturity by single cell examination (Rao et al. 2016).

Defining mature crenated shapes of some nuclei, such as the inferior olivary and dentate nuclei that initially appear as rather amorphous globular structures, is another aspect of morphogenic development, but synapse formation may precede the mature crenated shape (Sarnat et al. 2013). The bending of the primitive telencephalic hemisphere follows prosencephalic ‘cleavage’ into two hemispheres by formation of the sagittal fissure at 4–5 weeks’ gestation. This bending, the telencephalic flexure, eventually forms the operculum and lateral cerebral (Sylvian) fissure, but changes anatomical relations so that the posterior pole of the primitive telencephalon becomes the temporal, not the occipital, lobe, and both the frontal and temporal lips of the Sylvian fissure as well as the insula or ‘third lip’ all derive from the ventral surface of the early telencephalon and thus may be influenced by ventro‐dorsal genetic gradients in the vertical axis of the neural tube. These developmental relationships help explain malformations such as schizencephaly and perisylvian polymicrogyria and why all three lips of the Sylvian fissure are involved (Sarnat & Flores‐Sarnat, 2016). Gyration or convolution formation of the cerebral (and cerebellar) cortices occurs in the second half of gestation in humans and is related to growth of the neuropil, cytoplasmic volume of individual neurones and glioblast migration into the cortex. It is Nature's way of creating a greater surface area of cortex without concomitant increase in fetal brain volume and size, which would not be good for survival of either fetus or mother at delivery.

Maturation of pluripotential progenitor neuroepithelial cells to neuroblast to differentiation of specific types of neurones

Most studies of embryonic and fetal development of the brain focus on tissue architecture and morphogenesis, at both macroscopic (e.g. gyration) and microscopic (e.g. lamination of the cerebral cortex) levels. The differentiation and maturation of individual neurones is another aspect that must be considered in brain maturation, without which processes such as synaptogenesis and the formation of synaptic circuits and networks that generate normal brain function and epileptic states cannot be fully appreciated (Bystron et al. 2006, 2008; Clowry et al. 2010). Neuronal maturation can be conveniently divided for purposes of understanding into three periods: (1) transformation of uncommitted neuroepithelial progenitor cells into neuroblasts; (2) maturation of primitive neuroblasts already committed to neuronal lineage into secretory cells with a resting membrane potential (definition of a neurone); (3) final neuronal maturation, including differentiation of type, size or cytoplasmic volume, morphology and length of neurites, type of neurotransmitter synthesised and its influence on other neurones with which it makes axonal contact, as functionally excitatory or inhibitory (Rakic & Lombroso, 1998; Rakic et al. 2007).

Five initial ‘stages’ of human neuroepithelial cell transition to neuroblast may be identified, based upon weighted gene co‐expression or ‘transcriptome analysis’ of cell cultures (Li et al. 2017). Stage 1 cells are pluripotential or uncommitted to lineage and still capable of mitosis; stage 2 cells exhibit initial commitment to neural lineage; stage 3 cells are irreversibly committed to neuronal lineage; and stage 4 cells are neuronal progenitors that generate neuronal proteins. Stage 4 cells are differentiating neuroblasts that also show early indications of the type of neurone they will become (Lalli, 2014). The mTOR pathway is critical to this early cellular lineage and maturation (Sandsmark et al. 2007).

The second period of neuroblast maturation to become a neurone involves several processes that occur more or less simultaneously: the development of the Na⁺/K⁺ ATP energy pump that enables cellular membrane polarisation and a stable resting membrane potential; the synthesis of neurone‐specific proteins and intermediate filaments (initially transitory filaments such as vimentin, later replaced by mature neurofilaments); formation of ion channels and receptors in the electrically polarised cellular membrane; axonal extension followed by dendritic sprouting (Watanabe & Fukuda, 2015; Rich & Terman, 2018). Axonal extension may begin during neuroblast migration, even before the cell is in its final anatomical in the mature brain. The third period of neuroblast maturation involves the biosynthesis and secretion of neurotransmitter molecules; initial synaptogenesis followed by the formation of local synaptic circuits and long‐distance networks; myelination of axons (Sarnat, 2013). The role of mTOR in these later stages of neuronal maturation is still incompletely defined. Figure 1 summarises the major components of the mTOR signalling.

Figure 1.

A schematic overview of the mTOR pathway. A schematic overview of the mTORC1 signalling pathway showing the proteins that are affected by mutations of different mTORpathies (FCD, TSC, megalencephaly and hemimegencephaly) as summarised in Table 1. Mutations related to FCD are indicated with a red star, mutations related to TSC with a blue star, mutations related to megalencephaly with a light green star, and mutations related to hemimegencephaly with a dark green star. IRS1, insulin receptor substrate 1; PI3K, PI3kinase; PDK1, phosphoinositide‐dependent kinase‐1; PTEN, phosphatase and tensin homologue; AKT, protein kinase B; BRAF, v‐raf murine sarcoma viral oncogene homolog B1; MEK, mitogen activated protein kinase; ERK, extracellular signal‐regulated kinase; LKB1, tumor suppressor liver kinase B1; STRADα, STE20‐related kinase adaptor alpha; AMPK, AMP‐activated protein kinase; TBC1D5, TBC1 Domain Family Member 5; RHEB, ras homolog enriched in brain; mTORC1, mammalian target of rapamycin complex 1; DEPDC5, DEP Domain Containing 5; NPRL2, NPR2 Like, GATOR1 Complex Subunit; NPRL3, NPR3 Like, GATOR1 Complex Subunit; GATOR1, Gap Activity TOward Rags 1; S6K1, p70S6kinase; S6, ribosomal S6 protein; 4EBP1, eIF4E‐binding protein 1; eIF4E, binding of eukaryotic translation.

mTOR pathway‐related malformations

Tuberous sclerosis complex (TSC) represents the prototypic monogenic disorder of mTOR pathway dysregulation (Fig. 1) which provides the rational mechanistic basis of a direct link between gene mutation and brain pathology (structural and functional abnormalities) associated with a complex clinical phenotype including epilepsy, autism and intellectual disability (Curatolo, 2015; Curatolo et al. 2016, 2018b). Neuropathological examination of TSC brain specimens reveals three major lesions: subependymal nodules (SENs), subependymal giant cell tumours and cortical tubers (Mizuguchi & Takashima, 2001; DiMario, 2004; Aronica & Crino, 2014; Muhlebner et al. 2016b; Aronica & Muhlebner, 2017). Over the past decade, increasing evidence suggests that constitutive activation of the mTOR signalling cascade is considered to be a common feature in a subset of malformations of cortical development (MCDs) sharing histopathological and clinical characteristics (Galanopoulou et al. 2012; Wong & Crino et al. 2012a; Barkovich et al. 2015).

Since the first reports of MCDs, ranging from focal cortical dysplasia (FCD; Crome, 1957; Taylor et al. 1971) to hemimegalencephaly (HME) and megalencephaly (Sims, 1835; Flores‐Sarnat, 2002; Flores‐Sarnat et al. 2003), it has become clear that a shared mechanism of pathogenesis might be a possibility, as the same morphological alterations may be seen in any of these pathologies, suggesting a disturbance of cellular lineage and cellular growth (Aronica & Crino, 2014; Aronica & Muhlebner, 2017). Barkovich et al. (2005) have developed a classification scheme, subsequently updated (Barkovich et al. 2012, 2015; Desikan & Barkovich, 2016), based on the earliest developmental stage at which cortical development is first affected. The categories integrate pathological and neuroimaging features with genetics (when information is available). HME is included among the non‐neoplastic malformations secondary to cortical dysgenesis with abnormal cell proliferation, together with TSC and specific FCD subtypes, suggesting a deregulation the mTOR signalling as common link between these MCDs (Barkovich et al. 2012). Heterotopic dysplastic neurones within the U‐fibre layer beneath cortical FCD (‘neuronal dispersion’) that form intricate synaptic plexi for integration with cortical circuits as part of the epileptic network, are similar to dysplastic neurones within the cortex in the context of mTOR activation (Sarnat et al. 2018).

Therefore, the term ‘mTORopathies’ (mTOR pathway‐related malformations) has been introduced to define a spectrum of MCDs characterised by altered cortical architecture, abnormal neuronal/glial morphology and intractable seizures as a consequence of a deregulation of the mTOR signalling (Fig. 1; Crino, 2007, 2009, 2011; Wong, 2009; Aronica et al. 2012a; Aronica & Crino, 2014; Barkovich et al. 2015; Desikan & Barkovich, 2016).

In recent years, novel unbiased techniques such as whole exome or whole genome sequencing have been applied to discover novel genetic causes of MCDs, resulting in the identification of a number of genetic alterations affecting the mTOR pathway in a quite extensive spectrum of MCDs sharing the same features of abnormal cell proliferation and abnormal neuronal/glial morphology, including HME and FCD (Barkovich et al. 2015; Crino, 2015a; Desikan & Barkovich, 2016; Table 1).

Table 1.

Modified from Curatolo et al. (2018b). Mutations in different genes acting as regulators of the mTOR signalling are associated with malformations of cortical development

Gene	Mutation type	Effect of mutation	CNS pathological manifestations
TSC1/TSC2	Germline and somatic	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	TSC: cortical tubers, subependymal nodules, SEGA Hemimegalencephaly Focal cortical dysplasia
MTOR	Somatic	Activation of mTOR signalling	Focal cortical dysplasia, hemimegalencephaly
DEPDC5	Somatic and germline	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	Hemimegalencephaly, megalencephaly Focal cortical dysplasia
NPRL2	Germline	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	Focal cortical dysplasia
NPRL3	Germline	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	Focal cortical dysplasia
AKT1	Somatic	Activation mutation, increased mTOR signalling	Hemimegalencephaly
AKT3	Somatic and germline	Activation mutation, increased mTOR signalling	Hemimegalencephaly, megalencephaly
PIK3CA	Somatic and germline	Activation of mTOR signalling	Hemimegalencephaly, megalencephaly
PI3KR	Germline	Activation of mTOR signalling	Megalencephaly
PTEN	Germline	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	Focal cortical dysplasia, hemimegalencephaly, megalencephaly
STRADα	Germline	Inactivating mutation, releasing mTOR inhibition inducing enhanced mTOR signalling	Megalencephaly
TBC1D7	Germline	Inactivating mutation, inducing enhanced mTOR signalling	Megalencephaly

See text for definition of mTORopathies. TSC, tuberous sclerosis complex; SEGA, subependymal giant cell astrocytoma; mTOR, mammalian target of rapamycin; DEPDC5, DEP Domain Containing 5; NPRL3, NPR3 Like, GATOR1 Complex Subunit; AKT3, serine/threonine kinase 3; PIK3CA, phosphatidylinositol‐4,5‐bisphosphate 3‐kinase catalytic subunit alpha; PI3KR, regulatory subunit of PI3K; PTEN, phosphatase and tensin homologue; STRADα, STE20‐related kinase adaptor α.

Hemimegalencephaly (HME, previously referred to as ‘dysplastic megalencephaly’, DMEG)

HME (previously referred to as ‘dysplastic megalencephaly’, DMEG) is a rare malformation of cortical development characterised by overgrowth of one cerebral hemisphere and is associated with developmental delay and frequently severe epilepsy, typically developing within the first few months of life (Trounce et al. 1991; Janszky et al. 2003; Sanghvi et al. 2004; Sasaki et al. 2005; Tinkle et al. 2005). HME was first described by Sims (1835), who reported one autopsy case with hypertrophy of a single hemisphere. Since then, several small series have been published, suggesting that HME can occur isolated or associated with syndromic disease (Trounce et al. 1991; Janszky et al. 2003; Sanghvi et al. 2004; Sasaki et al. 2005; Tinkle et al. 2005). HME has been reported in patients with Proteus syndrome (DeLone et al. 1999), neurofibromatosis (Cusmai et al. 1990b), hypomelanosis of Ito (Tagawa et al. 1997), Klippel–Weber–Trenauny syndrome (Torregrosa et al. 2000), TSC (Cartwright et al. 2005) and linear sebaceous nevus syndromes (Solomon & Esterly, 1975; Sakuta et al. 1991; Flores‐Sarnat, 2002). On macroscopy, the surface of the brain often shows an abnormal gyral pattern, such as pachygyria, polygyria or polymicrogyria, or increased thickness of the grey matter within the abnormal hemisphere (for review see Flores‐Sarnat, 2002; Flores‐Sarnat et al. 2003; Manoranjan & Provias, 2010; Aronica et al. 2012a). Epilepsy is often resistant to drugs, requiring surgery to remove or functionally disconnect the epileptogenic area within the affected hemisphere (Jonas et al. 2004; Luders & Schuele, 2006; Bulteau et al. 2013; Ramey et al. 2013; Cuddapah et al. 2015). In both isolated and syndromic forms, a large spectrum of morphological alterations in combination with the different features of TSC cortical lesions and FCD have been reported (for review see Flores‐Sarnat, 2002; Flores‐Sarnat et al. 2003; Manoranjan & Provias, 2010; Aronica et al. 2012a). In a large number of cases, cortical dyslamination with the presence of hypertrophic and dysmorphic neurones, as well as heterotopic neurones in the white matter and in some cases leptomeningeal glioneuronal heterotopia, can be found upon microscopic inspection (Fig. 3T–W; Robain et al. 1988; De Rosa et al. 1992; Yasha et al. 1997; Adamsbaum et al. 1998; Flores‐Sarnat et al. 2003; Antonelli et al. 2004; Salamon et al. 2006; Boer et al. 2007); for review see (Flores‐Sarnat, 2002; Flores‐Sarnat et al. 2003; Manoranjan & Provias, 2010); balloon cells have also been detected in some, but not all, HME specimens (Flores‐Sarnat et al. 2003; Barkovich et al. 2005; Salamon et al. 2006; Boer et al. 2007).

Figure 3.

Macroscopy and histology in mTORopathies. (A–K) Tuberous sclerosis complex (TSC). (A) MRI showing a TSC lesion (histologically proven; arrow). (B–D) Coronal sections of the brain (32‐year‐old patient with a germline mutation in the TSC2 gene; Boer et al. 2008a; Aronica et al. 2012a) showing several regions with blurring of the cortex/white matter junction (arrows in B and C) and subependymal nodules (arrowheads in B and arrows in D). (B,C) A high magnification of abnormal brain regions with loss of a distinct cortex/white matter junction and mushroom‐like appearance of the gyri, indicating the presence of tubers. (D) A high magnification of the SENs which appear as firm oval‐shaped structures projecting into the ventricles. (E,F) Histological stains: Luxol fast blue‐PAS‐staining (E) and haematoxylin and eosin (H&E; panel F), showing cortical tubers (arrows) and subependymal nodules (arrowheads; Boer et al. 2008a; Aronica et al. 2012a). (G) phospho‐S6 ribosomal protein (pS6) staining positive in giant cells of a tuber (Prabowo et al. 2013). (H,I) NeuN staining showing difference in the architecture between the perituberal cortex (H) and the tuberal cortex (I) with dyslamination within the cortical tuber. (J,K) H&E staining of a tuber (J) and SEGA (K). Tuber showing large dysmorphic neurones (arrows), calcification (arrowheads) and a giant cell in insert (Boer et al. 2008a; Aronica et al. 2012a). SEGA showing giant cells (arrowheads) with a mixed glial background and blood vessels (Bongaarts et al. 2017). (L–S) Focal cortical dysplasia (FCD). (L) Coronal MRI image revealing a focal lesion at the left frontal lobe (arrow). (M) Coronal section of the brain showing a region with blurring of the cortex/white matter junction (indicated with arrow). (N) SMI32 staining showing accumulation of non‐phosphorylated neurofilaments (indicated with arrow). (O) H&E staining of FCD showing dysmorphic neurones (arrow) and balloon cells (asterisk). (P) Vimentin staining showing balloon cells (arrow and insert). (Q) Higher magnification of (N) showing accumulation of non‐phosphorylated neurofilaments (SMI32 antibody; arrows indicate dysmorphic neurones). (R,S) NeuN staining showing difference in the architecture between the control cortex (R) and FCDIIb cortex (S) with cortical dyslamination. (T–W) Hemimegalencephaly (HME; Boer et al. 2008a; Aronica et al. 2012a). (T) Coronal section showing diffuse and severe hypertrophy of the left hemisphere with lateral ventricular dilation, periventricular cysts, pachygyria and thickened cortex. (U) Loss of lamination with irregular distribution of neuronal cells and clusters of NeuN‐positive elements in Layer I. (V,W) Nissl staining showing the difference in the architecture of the control cortex (‘normal’ side; V) and the affected left side (HME; W). HME (W and insert) show severe cortical disorganisation with loss of lamination and presence of cytomegalic neurones (insert in W). Scale bars: (G) 60 μm, (H/I) 250 μm, (J/K) 40 μm, (O–Q) 50 μm, (R,S) 200 μm, (U) 320 μm, (W) 200 μm. (M,L) Courtesy of T. Veersema and K. Braun, University Medical Centre Utrecht, Utrecht, The Netherlands.

HME is sometimes detected prenatally (Sarnat et al. 2012; Lang et al. 2014), as early as 22 weeks fetal age (Manoranjan & Provias, 2010), and this supports the hypothesis of an underlying primary genetic defect that disturbs the early stages of cortical development.

HME and the cortical lesions in TSC show striking morphological similarities together with the expression of the mTOR effector pS6, which has led to the suggestion of a pathogenic link involving an hyperactivation of the mTOR signalling (Barkovich et al. 2015; Crino, 2015a; Desikan & Barkovich, 2016). Accordingly, de novo somatic mutations in PIK3CA, AKT3 and MTOR, encoding well‐known regulators of the mTOR signalling pathway, have been reported (Lee et al. 2012). Additional studies confirm the occurrence of pathogenic germline‐ and mosaic mutations in multiple phosphatidylinositol 3‐kinase (PI3K)‐AKT3‐mTOR signalling genes in HME and syndromes associated with megalencephaly (such as PTEN, DEPDC5, PIK3CA, PIK3R2 and MTOR; Poduri et al. 2012; Riviere et al. 2012; Mirzaa & Poduri, 2014; D'Gama et al. 2015; Jansen et al. 2015; Mirzaa et al. 2016; Table 1). Inherited mutations of PTEN can be associated to different hamartomatous syndromes, including Cowden syndrome, Bannayan‐Riley‐Ruvalcaba syndrome and Lhermitte‐Duclos disease; AKT3, PIK3CA and PIK3R2 mutations have been identified in HME, megalencephaly – polymicrogyriapolydactyly‐hydrocephalus syndrome, megalencephaly‐capillary malformation syndrome, congenital lipomatous overgrowth, vascular malformations and epidermal naevi (CLOVE) syndrome and FCD (Poduri et al. 2012; Riviere et al. 2012; Mirzaa & Poduri, 2014; D'Gama et al. 2015; Jansen et al. 2015; Desikan & Barkovich, 2016; Mirzaa et al. 2016; Table 1). These findings show large phenotypic variability that could reflect the involvement of other modifier genes or mechanisms of epigenetic regulation that is still incompletely understood (Mirzaa & Poduri, 2014; Jansen et al. 2015; Crino, 2016).

Within a short time, a number of genetic mutations affecting the PI3K/Akt/mTOR pathway have been published, providing additional opportunities for the development of more specific animal models to investigate other downstream potential therapy targets (Baulac et al. 2015; Mirzaa et al. 2016; Sim et al. 2016; Lim et al. 2017; Ribierre et al. 2018; Sarnat, 2018). Mouse models of Pik3ca mutations recapitulate the key human pathological features, including brain enlargement, cortical malformation and epilepsy. Additionally, treatment with PI3K inhibitors shows an anti‐epileptic effect in these animals (Roy et al. 2015).

From a neuropathological perspective, the most important distinction between hemimegalencephaly and FCD is their relative extent or the size/volume of the lesion. This difference is related to timing of onset of the post‐zygotic somatic mutation with activation of the mTOR pathway, and can be explained by neuroembryology. Only 33 mitotic cycles of the progenitor cells of the periventricular neuroepithelium are required to produce all of the neurones of the human cerebral cortex (10–11 cycles in the rodent; Caviness, 1981). If the somatic mutation is first expressed in a late mitotic cycle, the number of abnormal clones is smaller and the resulting lesion is more focal, involving a single gyrus, such as FCDIIb; if the mutation is expressed in a somewhat earlier cycle, the lesion is more extensive and may involve several adjacent gyri or part of one hemisphere as more extensive FCDIIb or limited HME; the involvement of an earlier mitotic cycle can produce HME of the entire hemisphere and expression in the earliest cycles generates ‘total HME’, which includes the ipsilateral cerebellar hemisphere and ipsilateral brainstem because the neuroepithelium in early development is continuous between future supratentorial and infratentorial compartments. Figure 2 summarises this concept of timing as being the determining factor in the extent of the dysplasia. Recent genetic data have confirmed by another technique that FCDIIb and HME represent a spectrum of the same disorder (Poduri et al. 2012; D'Gama et al. 2015, 2017; Ottman et al. 2018).

Figure 2.

Lesion differences in mTORopathies. (A) A schematic overview showing the concept of mosaicism in the brain where somatic mutations can occur during development with or without an underlying germline mutation. Somatic mutations that appear during early development (in blue) can lead to hemimegalencephaly (in blue). Somatic mutations that occur later in development (in purple) or second‐hit mutations (in green) that occur when a germline mutation is present (in orange) can lead to focal lesions (indicated in purple and green). Both time and location of the somatic mutations can influence the affected region and size of the brain. (B) A schematic overview showing both the location and the histological characteristics of TSC tubers, FCD and SEGAs.

FCD type II

FCD type II represents the most frequent substrate in paediatric epilepsy surgery patients (Harvey et al. 2008; Blumcke et al. 2009). FCD was first described by Taylor et al. (1971) as a distinctive anomaly of cortical lamination and organisation with balloon‐shaped cells (Fig. 3L–S). Their description included ‘disruption of the normal cortical lamination’ and ‘aberrant nerve cells’ with unusual numbers, sizes and orientation. The authors suggest that these aberrant (‘exotic’) populations of nerve cells may underlie the clinical manifestations of certain forms of focal epilepsy. The authors also discuss the similarities with the neuropathological features of cortical lesions in patients with TSC, emphasising the importance of the differential diagnosis with ‘forme fruste’ of TSC (Taylor et al. 1971). Since the first reports it has become clear that the morphological spectrum of cortical dysplasias is broad and different histopathological features can be encountered in clinical practice (Aronica et al. 2012a). Cortical dysplasia may represent an isolated lesion, but distorted cortical lamination can also be detected adjacent to hippocampal sclerosis, glio‐neuronal tumours and vascular malformations, as well as adjacent to a large variety of lesions acquired early during brain development (Blumcke et al. 2009). Over the past decades, several systems for the classification of cortical dysplasias have been proposed. While some of them are based on the histopathological features (Mischel et al. 1995), others rely on imaging and genetic findings (Barkovich et al. 2012), or combine clinical and histopathological aspects (Palmini et al. 2004).

The Task Force of the ILAE on Diagnostic Methods re‐defined the FCD subtypes in 2011 (Blumcke & Spreafico, 2011). The new classification scheme is based on histopathological findings but at the same time tries to incorporate the pathological neurodevelopmental processes leading to the formation of lesions. The 2011 system distinguishes three types of FCD: FCD type I describes isolated focal lesions with architectural abnormalities; type II includes isolated focal lesions with architectural and dysmorphic abnormalities; and the new group named FCD type III refers to cortical disorganisation ‘associated with or adjacent to other principal lesions’. Knowledge about the molecular pathology of FCD type I is still rather limited, whereas molecular alterations of FCD type II are better understood (Aronica et al. 2012a; Aronica & Muhlebner, 2017). FCD type II refers to an isolated malformation characterised by disrupted cortical lamination and cytological abnormalities and includes two subtypes, FCDIIa (with dysmorphic neurones, but without balloon cells) and FCDIIb (with dysmorphic neurones and balloon cells; Blumcke & Spreafico, 2011). These cellular features, resembling those observed in TSC brain lesions, have been critical in the effort to understand the pathogenesis of FCD type II, providing clues to the cell signalling abnormalities that now link FCDIIb to mTOR signalling deregulation. Several immunohistochemical studies provide evidence of enhanced mTOR signalling in FCDII with strong expression of the phosphorylated form of the downstream target S6 of the mTORC1 (Crino, 2015b). These observations suggested that these MCDs could potentially represent the result of a somatic gene mutation of PI3K/Akt/mTOR pathway regulatory genes. This has been confirmed recently by the detection in FCD of somatic mutations in different genes acting as regulators of the mTOR signalling, including PTEN, PI3KCA, AKT (Schick et al. 2006; Conti et al. 2015; Jansen et al. 2015), DEPDC5 and NPRL3 (both components of the mTOR regulatory GATOR1 complex; Baulac et al. 2015; Scerri et al. 2015; Sim et al. 2016; Ribierre et al. 2018), as well as MTOR itself (D'Gama et al. 2015; Lim et al. 2015; Nakashima et al. 2015; Mirzaa et al. 2016); Table 1. Thus, these recent findings support the inclusion of FCDIIb within the spectrum of mTOR pathway‐related malformations (‘mTORopathies’; Barkovich et al. 2015; Crino, 2015b; Desikan & Barkovich, 2016). However, an explanation of the wide phenotypic variability is still lacking. For example, mutations in DEPDC5 have been identified in HME, FCD and also in non‐lesional epilepsy (Baulac et al. 2015; Scerri et al. 2015; Sim et al. 2016). Moreover, not only somatic but also germline mutations have been detected in patients with HME and FCD, and in one individual both germline and somatic mutations of DEPDC5 have been detected (Baulac et al. 2015; Ribierre et al. 2018), suggesting, as discussed below for TSC, the possibility of a mutational ‘two‐hit’ model. However, the variable mutation level within the lesion, in some cases with low level mosaic mutations (below the detection limit of our techniques), may result in negative sequencing results, even in FCD with clear evidence of increased mTOR pathway activity. The rapidly expanding knowledge of the pathogenesis of FCDs underlies the need to revise the classification of FCDs. The development of a more pathway‐based classification system may guide the design of preventative trials in specific patient populations. Interestingly, focal cortical expression of mutant Mtor by in utero electroporation in mice results in activation of mTOR kinase, leading to neuronal migration defects with dysmorphic neurones and spontaneous seizures, which were rescued by the mTOR inhibitor rapamycin (Lim et al. 2015). This represents an in vivo disease modelling of FCD that directly proves the casual relationship between the identified mutation and the mTOR‐dependent cellular abnormalities and epilepsy.

Glioneuronal tumours (GNTs)

Enhanced mTOR signalling pathway activation has also been detected in glioneuronal tumours (GNT; Samadani et al. 2007; Boer et al. 2010b; Prabowo et al. 2014).

GNT (such as gangliogliomas) are low‐grade neuroepithelial tumours that present with early‐onset focal epilepsy and are mostly seen in children and young adults (previously designated as long‐term epilepsy‐associated neuroepithelial tumours, LEATs, Blumcke et al. 2016). Gangliogliomas (GGs) consist of a mixture of neurones and glial tumour cells mainly represented by a large spectrum of astroglial cells. The neuronal component, which varies in amount, is represented by dysplastic neurones with abnormal shapes and sizes, and lack uniform orientation and often expression of the phosphorylated form of the downstream target S6 of the mTORC1. Mutational analysis of TSC1 and TSC2 failed to identify mutations in GGs (Parry et al. 2000) and only a somatic mutation in intron 32 of the TSC2 gene was reported in one GG patient in glial cells, but not in dysplastic neurones (Becker et al. 2001). However, more recent studies have reported a mutation in the BRAF oncogene (a member of the RAF family of serine/threonine protein kinases involved in the RAS‐RAF‐MEK‐ERK‐MAP kinase signalling pathway) in a large majority of GNTs (Dougherty et al. 2010; Schindler et al. 2011; Chappe et al. 2013; Dahiya et al. 2013; Koelsche et al. 2013; Prabowo et al. 2014). Interestingly, BRAF V600E mutation has been shown to be significantly associated with the expression of pS6 in GNT (Prabowo et al. 2014). Accordingly, it has been shown that BRAF V600E mutant cells have a dysfunctional tumour suppressor liver kinase B1 (LKB1)‐AMP‐activated protein kinase (AMPK)‐mTOR signalling (Esteve‐Puig et al. 2009; Zheng et al. 2009). Thus, BRAF‐induced phosphorylation of LKB1 may represent a possible mechanism contributing to mTOR activation in BRAF V600E mutated GNTs, possibly through uncoupling of the LKB1‐AMPK‐mTOR signalling. However, seizure activity itself and the inflammatory environment (i.e. via interleukin‐1β) could also contribute to the activation of the PI3K–AKT3–mTOR signalling pathway (Liu et al. 2014; Rossini et al. 2017). Interestingly, using a mouse model harbouring the Braf v600e somatic mutation during early brain development, it has recently been shown that Braf v600e‐induced epileptogenesis is mediated by RE1‐silencing transcription factor (REST) mediated by MYC through activation of the RAS–RAF–MAPK signalling (Koh et al. 2018). Furthermore, the recently reported genetic alterations detected in LEATs provide evidence of a functional connection between two major signalling pathways: RAS–RAF–MAPK and PI3K–AKT–mTOR (Blumcke et al. 2016).

Mutations in the STE20‐related kinase adaptor alpha (STRADalpha; Table 1), a negative regulator of mTOR via AMPK, are associated with polyhydramnios, megalencephaly, symptomatic epilepsy syndrome (PMSE), a rare neurodevelopmental disorder characterised by macrocephaly, craniofacial dysmorphism, severe, infantile‐onset intractable epilepsy and intellectual disability (Puffenberger et al. 2007; Orlova et al. 2010; Bi et al. 2016; Evers et al. 2017). Neuropathological evaluation of the PMSE brain tissue revealed cytomegaly, heterotopic neurones in subcortical white matter and aberrant activation of mTORC1 signalling (Puffenberger et al. 2007; Orlova et al. 2010). Rapamycin has been shown to rescue aberrant cortical lamination and heterotopia associated with STRADA depletion in the mouse cerebral cortex; treatment of five PMSE patients with sirolimus (rapamycin) resulted in a reduction in seizure frequency and an improvement in receptive language (Parker et al. 2013). These observations together support the functional link between STRADA loss and mTORC1 dysfunction, and the classification of PMSE within the spectrum of mTOR‐associated neurodevelopmental disorders.

Tau upregulation in activated mTOR developmental disorders

Tau is a microtubule‐associated protein that is overexpressed in an abnormally hyperphosphorylated form in neurones in several neurodegenerative diseases of late adult life, including Alzheimer and Parkinson diseases and fronto‐temporal lobar degeneration (Kovacs 2015). It is a major cause of dementia but not of severe epilepsy in adults.

The microtubule is one of the earliest subcellular organelles to develop in the progenitor neuroepithelial cell with neuronal lineage and is influenced by the many microtubule‐associated protein products for which the mutation of their gene causes generalised disorders of cortical development, such as the Doublecortin (DCX) gene that is the aetiology of subcortical laminar ‘band’ heterotopia (Gleeson et al. 1998, 1999a,1999b). Myotubular dysregulation by an abnormal genetic expression of an abnormally phosphorylated tau protein early in cellular maturation disturbs programmed growth and morphology of that neurone, a major factor in the genesis of megalocytic dysplastic neurones. The hippocampus of infants with HME shows abnormal tau immunoreactivity not only in the mature granular neurones of the dentate gyrus but also in the undifferentiated progenitor cells of the sub‐dentate polymorphic layer (Sarnat et al. 2012). The early timing of this expression in neuronal maturation is consistent with abnormal growth and morphogenesis of neurones in disorders of mTOR pathway activation. There are four ‘infantile tauopathies’ described: FCDII, HME, TSC and the cortical developmental tumour ganglioglioma (Sarnat & Flores‐Sarnat, 2015). By contrast with adult tauopathies, all infantile tauopathies are highly epileptogenic but do not cause progressive dementia.

TSC brain pathology

Cortical tuber

Cortical tubers are focal developmental malformations, characterised by disturbed neuronal architecture and layering, representing the pathological hallmark of TSC (Fig. 3A–K). Several studies have provided evidence of the link between cortical tubers, the generation of seizures and associated comorbidities in individuals with TSC (Cusmai et al. 1990a; Chugani et al. 1998; Koh et al. 2000; Ridler et al. 2004; Jansen et al. 2006; Wong & Khong, 2006). Cortical tubers are detected as single or multiple lesions in individuals with TSC (for reviews see Crino et al. 2006; Orlova & Crino, 2010; Aronica & Crino, 2014; Aronica & Muhlebner, 2017; Curatolo et al. 2018b). They consist in areas of cortical dyslamination that contain different cell types, including dysmorphic neurones, giant cells and reactive astrocytes (Mizuguchi & Takashima, 2001; Boer et al. 2008a; Grajkowska et al. 2010; Zurolo et al. 2011; Aronica et al. 2012a; Aronica & Crino, 2014; Aronica & Muhlebner, 2017; Curatolo et al. 2018b). Analysis of cortical layer markers supports the severe disturbance of cortical structure, indicating a dysmaturation affecting early and late migratory patterns, with a more severe impairment of the late stage of maturation (Muhlebner et al. 2016a). Dysmorphic neurones (DNs) are characterised by abnormal morphology, abnormal orientation and abnormally large sizes. DNs express different cortical layer markers, regardless of their laminar location, and their immunophenotype resembles that of cortical projection neurones and suggests an alteration of a selected population of intermediate progenitor cells (Muhlebner et al. 2016a; Aronica & Muhlebner, 2017). Giant cells have been shown to express both neuronal and immature glial markers, indicating a failure to differentiate prior to migration into the cortex (Crino et al. 1996; Lee et al. 2003; Talos et al. 2008; Boer et al. 2009); for review see Orlova & Crino, 2010). Enhanced activation of mTOR signalling is evidenced by enhanced phospho‐activation of the mTOR effectors p70S6kinase and pS6 in both dysmorphic neurones and giant cells, providing the molecular explanation for their aberrant morphological features (Orlova & Crino, 2010; Curatolo et al. 2016, 2018b). Recently, three distinct histological cortical tuber subtypes (designated as types A, B and C) have been described based on the proportion of calcifications, dysmorphic neurones and giant cells (Muhlebner et al. 2016b). Interestingly, Type C (characterised by calcification, prominent inflammation, increased gliosis and prominent myelin loss) occurs more often in very young children and possibly reflects the time point of tuber formation leading to such severe pathological changes already in the early stages of cerebral development (Muhlebner et al. 2016b). White matter pathology, with depleted myelin and oligodendroglia, represents another major feature of the TSC pathology (Scholl et al. 2017; see sub‐session ‘white matter pathology’). Cortical tubers can be detected prenatally (Park et al. 1997; Chen et al. 2006; Glenn & Barkovich, 2006; Wortmann et al. 2008; Prabowo et al. 2013), indicating that tubers form during embryonic brain development with evidence of increased mTORC1 activation in fetal tuber specimens, before seizure development (Prabowo et al. 2013; Tsai et al. 2014). Seizures in the fetus are not physiologically possible before about 24–25 weeks’ gestational age because synaptogenesis in the cortical plate is not initiated until 22 weeks (Sarnat et al. 2010). These observations support the establishment of a pathological neural network during brain development, with widespread cellular and molecular abnormalities, driven by mTOR dysfunction (Curatolo et al. 2016, 2018b). Accordingly, seizure onset is observed in the first year of life in about 67% of affected children, with focal seizures and spasms (Curatolo et al. 2016, 2018b).

Epilepsy surgery represents a promising treatment option in TSC patients in which the epileptogenic focus is easily recognizable (Koh et al. 2000; Weiner et al. 2006; Bollo et al. 2008) and may also lead to improvements in quality of life and IQ, especially in postoperative patients who remained seizure‐free (Liang et al. 2017). Although epilepsy surgery often results in freedom from seizures, increasing evidence supports the importance of the perituberal cortex in TSC (Madhavan et al. 2007; Wang et al. 2007; Major et al. 2009; Moshel et al. 2010; Mohamed et al. 2012; Fallah et al. 2015; Fujiwara et al. 2016; Liang et al. 2017). The presence of a complex and widespread epileptic network is also supported by the detection of multiple extensive zones with a high occurrence rate of interictal high frequency oscillations (HFO, Okanishi et al. 2014). Moreover, focal seizures and interictal epileptiform discharges detected in the centre of epileptogenic tubers have been shown to propagate to the tuber rim, perituberal cortex and other epileptogenic tubers (Kannan et al. 2016). Structural abnormalities with evidence of enhanced activation of mTOR signalling have also been detected in the surgically resected epileptogenic, perituberal tissue (Sosunov et al. 2015; Muhlebner et al. 2016b). Moreover, neuropathological examination of postmortem TSC brain has provided evidence of multiple subtle structural abnormalities, present throughout the brain (i.e. ‘microtubers’, focal dyslamination, isolated giant cells), which may contribute to the complex and variable neurological phenotype encountered in TSC patients (Dimario, 2004; Luat et al. 2007; Griffiths et al. 2011; Marcotte et al. 2012).

Patients with TSC have an increased risk of developing different types of tumours (e.g. subependymal giant cell astrocytomas which lack neurones; see below), which are commonly associated with mutations in the second allele of TSC1 or TSC2, known as loss of heterozygosity (LOH), leading to complete loss of the protein. The same mechanism (‘two‐hit’ model with somatic inactivation of TSC1/TSC2) has been suggested to explain tuber formation in TSC (Crino et al. 2010; Qin et al. 2010). However, second‐hit mutations are not always detected in cortical tubers (Caban et al. 2017). A recent study indicates that only one‐third of cortical tubers are driven by somatic TSC1/TSC2 inactivation, suggesting that either that only a small portion of cells within the tuber is affected by a second hits, limiting their identification (mutational burden below detection limits), or that a mono‐allelic mutation (perhaps in association with detrimental non‐cell‐autonomous function of mutant cells during corticogenesis) could be sufficient for tuber development (Martin et al. 2017).

Subependymal giant cell astrocytomas (SEGAs)

According to the updated diagnostic criteria for TSC, SEGAs are one of the major pathological features of TSC (Northrup et al. 2013). SEGAs are low‐grade glioneuronal tumours classified as WHO grade I and represent 1–2% of all paediatric brain tumours (Jozwiak et al. 2015; Louis et al. 2016). Although SEGAs are benign and slow‐growing tumours, extensive growth can cause obstruction of the cerebrospinal fluid tract, leading to hydrocephalus and in some cases even sudden death (Cuccia et al. 2003; Amin et al. 2013). The prevalence of SEGAs in patients with TSC ranges from 5 to 25% (Cuccia et al. 2003; Goh et al. 2004; Adriaensen et al. 2009; Kothare et al. 2014; Kingswood et al. 2017). SEGAs can already be detected in fetal and neonatal periods, but generally develop during the first two decades of life and are rarely detected de novo during adulthood (Oikawa et al. 1994; Torres et al. 1998; Raju et al. 2007; Phi et al. 2008; Adriaensen et al. 2009, 2014; Prabowo et al. 2013). They are thought to arise from SENs along the ependymal lining of the lateral ventricles at the height of the foramen of Monro (Morimoto & Mogami, 1986; Fujiwara et al. 1989; Nabbout et al. 1999). SEGAs consist of spindle cells, gemistocytic‐like cells and giant cells, and are histologically indistinguishable from SEN. At the molecular level, little is known about the establishment and progression of SEGAs. Evidence of second‐hit inactivation of TSC1 or TSC2 has been reported in approximately 80% of SEGAs analyzed, the majority of which are copy‐neutral loss of heterozygosity (CN‐LOH) events (Chan et al. 2004; Bongaarts et al. 2017; Martin et al. 2017). However, second‐hit mutations are not always seen in SEGAs and may already be present in SEN, suggesting that additional molecular mechanisms may play a role in their progression and growth. Recently, it was shown that SEGAs do not have other somatic mutations in mTOR pathway‐related genes and that the overall mutation burden is low, which is consistent with their slow growing character (Bongaarts et al. 2017; Martin et al. 2017). In the same study by Martin et al. (2017) it was shown that gene ontologies related to cell systems such as inflammation, extracellular matrix organisation and synaptic transmission were mainly affected in SEGAs (Martin et al. 2017). However, it is still unknown how the deregulation of these cell systems could contribute to the SEGA pathology and this needs to be investigated further.

Commonality and distinction

Tuberous sclerosis is genetically more complex than FCDIIb/HME because it is a similar post‐zygotic somatic mutation with mTOR activation but also a germline mutation with mutation of either gene TSC1 (locus 9q34) or TSC2 (locus 16p13; Crino et al. 2006; Curatolo, 2015; Curatolo et al. 2018c). These genes are also tumour‐suppressors. Although some authors have reported that these two genes are not mutated in FCDII (Fauser et al. 2009), other authors have found find somatic mutations associated with them in FCDII (Lim et al. 2017). Whole exome sequencing demonstrates that mosaic parental germline mutations cause other more generalised malformations of cortical development (Zillhardt et al. 2016). Cortical tubers become epileptogenic at least in part because of decreased GABAergic inhibition and increased glutamatergic excitation (Curatolo et al. 2018c). Tubers can be first recognised in fetal MRI at 27–28 weeks’ gestation and are demonstrated at fetal autopsy even earlier (Prabowo et al. 2013).

SENs are composed of pure abnormal glial cells, but cortical tubers are characteristic of FCDIIb and HME, including the presence of balloon cells and dysmorphic megalocytic neurones and glia, and disorganised cortical architecture. Transmantle dysplasias with displaced grey matter extending from around the ventricular wall to the bottom of a sulcus or base of a gyrus are common to all three entities. TSC cortical tubers can sometimes be distinguished histologically by an additional presence of small calcifications that are rare in FCDIIb and HME. Subtle differences in topographic distribution of abnormal cells can also be demonstrated between TSC and FCDII (Cepeda et al. 2010).

Balloon cells are found in FCDIIb, HME and TSC. They are primitive globoid cells of variable size, often large, and histologically showing amorphous, featureless eosinophilic cytoplasm and one or more nuclei displaced to the periphery of the cell. They are strongly immunoreactive with immature intermediate filament proteins such as nestin and vimentin and also with the heat‐shock protein α‐B‐crystallin that is overexpressed in glial cells near epileptogenic foci (Sarnat & Flores‐Sarnat, 2009). Cultures of isolated balloon cells show similar reactivities as in tissue sections, and a subpopulation is also labelled with β1 integrin, a marker of progenitor stem cells, thus indicating that balloon cells are non‐neoplastic stem cells that have failed to differentiate. However, we have found that many balloon cells in HME and TSC are of mixed lineage and express neuronal proteins such as synaptophysin and MAP2 (but not NeuN or neurofilament protein) and also express glial proteins such as GFAP, in addition to the nestin, vimentin, and the α‐B‐protein already mentioned (Sarnat & Flores‐Sarnat, 2009; Blumcke et al. 2017).

Inflammation is recently recognised as an important component in the brains of patients with tuberous sclerosis, including fetuses (Prabowo et al. 2013; Curatolo, 2015; Curatolo et al. 2018c). Whether FCDIIb/HME cases also include an inflammatory component is not certain because studies of the latter are sparse. The heat‐shock protein α‐B‐crystallin is overexpressed in glial cells in both cortical grey and subcortical white matter at or near to electrographic epileptic foci (Sarnat & Flores‐Sarnat, 2009), and may be regarded as another form of inflammation. Other similarities between TSC and FCDIIb are represented by the presence of white matter pathology (as discussed below).

What we have learned so far

Mechanisms of epileptogenesis in mTORopathies

Surgery has become an important option in epilepsy‐associated focal MCD (Blumcke et al. 2017; Liang et al. 2017) with a parallel increase in the availability of clinically well‐characterised surgical specimens that have provided the opportunity to study the cellular mechanism(s) underlying their epileptogenicity by combining traditional routine histopathological methods with molecular and functional analysis of the resected tissue (reviewed in Crino, 2013; Aronica & Crino, 2014). Mounting evidence indicates that focal MCD (i.e. cortical tubers in TSC patients and FCDs) display an abnormal electrical activity that is associated with epilepsy (Palmini et al. 1995; Matsumoto et al. 2005; Otsubo et al. 2005; Okanishi et al. 2014; Ueda et al. 2015; Kannan et al. 2016; Cuello‐Oderiz et al. 2018) and their removal often results in seizure freedom or reduction of seizure frequency (Blumcke et al. 2017; Liang et al. 2017). Several studies using in vitro electrophysiological recording in surgical specimens of patients with cortical tubers and FCDs support that giant/balloon cells were essentially electrically silent, whereas neurones display hyper‐excitable intrinsic membrane properties after depolarisation (Najm et al. 2004; Avoli et al. 2005; Cepeda et al. 2005, 2006; Wang et al. 2007). Several studies support the role of developmental alterations of the balance between excitation and inhibition in the pathogenesis of epileptic focal discharges in both cortical tubers and FCDs (reviewed in Holmes et al. 2007; Wong, 2008; Crino, 2013; Aronica & Crino, 2014; Curatolo et al. 2018b). The mechanistic link between genetic mutations, dysregulation of mTOR pathway and excitatory/inhibitory synaptic imbalance leading to hyperexcitability has also been supported by different experimental studies (Tavazoie et al. 2005; Bateup et al. 2013; Lozovaya et al. 2014; Santos et al. 2017).

Attention has first been focused on the excitatory amino acid synaptic transmission in the dysplastic cortex through the activation of ionotropic glutamate receptors (iGluRs; e.g. NMDA, GluN and AMPA, GluA receptors). Alteration in GluN subunit expression has been reported in cortical tubers, HME and FCD type II, with increased expression of the subunit NR2B (Ying et al. 1999; Crino et al. 2001; White et al. 2001; Moddel et al. 2005; Finardi et al. 2006). Changes in the expression of components of the membrane‐associated guanylate kinase (MAGUK) protein family, interacting with GluN, have been also reported (Finardi et al. 2006; Qu et al. 2009). In a TSC murine model, Tsc1^+/− mice (with minimal pathological abnormalities) mTOR activation increased NMDA‐mediated excitatory activity due to an upregulation of a specific subunit, GluN2C. These findings were confirmed in human TSC surgical resection samples (Lozovaya et al. 2014). Additional observations also indicate a deregulation of AMPA (Crino et al. 2001; White et al. 2001; Wang et al. 2007; for reviews see Najm et al. 2004; Avoli et al. 2005; Talos et al. 2008; Aronica & Crino, 2014; Curatolo et al. 2018b). Moreover, the microtransplantation method of injecting Xenopus oocytes with membranes from TSC cortical tubers and control brain tissues, has provided additional functional evidence of an immaturity of AMPAR‐mediated responses (with high calcium‐permeable AMPARs), reflecting the absence of, or a delay in, the developmental switch of the receptor subunit composition (Ruffolo et al. 2016).

In addition to the deregulation of iGluR, the cellular distribution of mGluR subtypes, with high expression of mGluR1 and mGluR5 in dysmorphic neurones, suggests a possible contribution of group I mGluRs to the intrinsic and high epileptogenicity of dysplastic cortical regions (TSC, HME, FCD; Aronica et al. 2017; Boer et al. 2007, 2008b). Alterations of the glutamatergic transmission in focal MCD may also involve an abnormal glutamate homeostasis due to a reduced expression of the glial glutamate transporters (i.e. GLT‐1; Boer et al. 2010a; Ulu et al. 2010) and neuronal glutamate transporters (i.e. EAAC‐1; Baybis et al. 2004).

There is also a substantial amount of data demonstrating a deregulation of inhibitory synaptic transmission [in particular of gamma‐aminobutyric acid type A receptor (GABAAR) signalling] in mTOR pathway‐related malformations (Cepeda et al. 2006; Wang et al. 2007; Talos et al. 2008, 2012; Bateup et al. 2011; Abdijadid et al. 2015; reviewed in Aronica & Crino, 2014; Curatolo et al. 2018b). The impairment of GABAergic inhibition in human mTOR pathway‐related malformations is also supported by evidence, from tissue obtained from individuals with TSC and FCD type II, of an aberrant GABAA reversal potential, paralleling a differential expression of cation‐chloride (NKCC1 and KCC2) cotransporters (CCTs) and an altered expression of receptor subunits (Ruffolo et al. 2016).

Dysfunction of neurone‐glia interactions and modification of the extracellular space, induced by astrogliosis or by modified extracellular matrix (ECM) composition, might be involved in the epileptogenic mechanisms of mTOR pathway‐related malformations (for reviews see Binder & Steinhauser, 2006; Wong, 2008; Seifert & Steinhauser, 2013; Aronica & Crino, 2014; Jovanov Milosevic et al. 2014; Curatolo et al. 2018b).

Reactive astrogliosis has been shown to contribute critically to the mechanisms underlying the development of epilepsy (Coulter & Steinhauser, 2015; Steinhauser et al. 2016) and mounting evidence support its contribution to the epileptogenicity of mTOR pathway‐related malformations through different mechanisms involving impaired astrocytic gap junction coupling, glutamate and potassium buffering, as well as overexpression of adenosine kinase (which induces a deficiency in the homeostatic tone of adenosine; reviewed in Wong & Crino, 2012b; Aronica & Crino, 2014; Boison & Aronica, 2015; Curatolo et al. 2018b).

Astrocytes, together with microglia, are also key players in the induction of major proinflammatory pathways which have been shown to contribute to the pathophysiology of human epilepsy. Accordingly, inflammatory mediators released by the brain's resident cells during epileptic activity (i.e. cytokines, chemokines, prostaglandins, complement system factors), not only are effector molecules of the immune system, promoting local inflammation, but also function as neuromodulators, directly affecting neuronal function and excitability and contributing to the mechanisms of seizure generation, epileptogenesis and cognitive dysfunction (for review see Aronica & Crino, 2011; Vezzani et al. 2011, 2012, 2015; Aronica et al. 2012b, 2017; van Vliet et al. 2018). In particular, a growing body of evidence indicates that the classical complement system components, C1q and C3, may function as important regulators of synaptic pruning and plasticity (Stevens et al. 2007; Perry & O'Connor, 2008; Stephan et al. 2012; Mastellos, 2014).

Activation of both innate and the adaptive immune response with concomitant induction of various proinflammatory proictogenic inflammatory pathways has been reported in different mTOR pathway‐related malformations (i.e. FCD and TSC; Mills et al. 2017; reviewed in Aronica & Crino, 2014; Curatolo et al. 2018b). The detection of activated microglia/macrophages even during brain development in TSC fetal brain (Prabowo et al. 2013), supports the role of immune‐inflammatory responses in the early epileptogenic processes and suggests a link to the mTOR pathway dysregulation (Curatolo et al. 2018b). Interestingly, over‐activation of pro‐inflammatory signalling pathways in astrocytes has been observed before epilepsy onset in a mouse model of TSC (Zhang et al. 2015). More recent studies in TSC mouse models point to a possible role for microglia abnormalities (resulting from Tsc1 inactivation) in contributing to epileptogenesis; however, whether microglial cells are primary contributors to epilepsy and associated comorbidities is still matter of discussion (Zhang et al. 2018; Zhao et al. 2018).

Evidence from experimental and human studies points to the role of mTOR‐dependent deregulation of autophagy, suggesting that impaired autophagy could also contribute to epileptogenesis in mTOR pathway‐related malformations (McMahon et al. 2012; Giorgi et al. 2015). Interestingly, abnormalities in autophagy have been reported in mTOR pathway‐related malformations [i.e. FCD and TSC; Yasin et al. 2013; Iyer et al. 2014). Moreover, as discussed above, the occurrence of tau pathology in different mTOR pathway‐related malformations (Sen et al. 2007; Iyer et al. 2014; Sarnat & Flores‐Sarnat, 2015) provides additional potential mechanisms contributing to the intrinsic neuronal network hyper‐excitability characteristic of these MCD and may suggest new targets for therapeutic intervention (Xi et al. 2011; Jones et al. 2012; Kearney, 2013; Gheyara et al. 2014; Liu et al. 2016).

The identification of a specific molecular pathway underlying seizure development in a spectrum of MCD paved the road to personalised medicine with clinical trials that evaluated the efficacy and safety of mTOR inhibitors (such as everolimus) in TSC‐related epilepsy (French et al. 2016; Curatolo et al. 2018a, 2018b, 2018c). These studies support the use of everolimus in drug‐resistant epilepsy associated with TSC, but also suggest the possibility to explore mTOR inhibitors in other mTOR pathway‐related malformations (Curatolo et al. 2018b; Jeong & Wong, 2018). A recent study comparing ex vivo brain slices from TSC, HME and non‐mTOR pathology demonstrated robust anti‐seizure effects of everolimus (through reduction in glutamate receptor‐mediated excitatory activity after sustained exposure over time) in mTOR pathway‐related malformations (TSC and HME; Cepeda et al. 2018).

White matter pathology

Selective loss of myelin, directly associated with the affected brain region in TSC and FCDIIb brain lesions depicts a unique phenomenon which is unknown in other epilepsy‐associated diseases (Muhlebner et al. 2012; Scholl et al. 2017). Only a small number of studies so far have focused on myelin loss within the spectrum of FCD and TSC (Muhlebner et al. 2016b; Scholl et al. 2017). In recent decades, efforts have been made to uncover the pathogenesis of both diseases. Interestingly, the increased mTOR activation in FCDIIb and TSC lesions correlates with myelin loss (Scholl et al. 2017). A murine model of TSC also shows this hypomyelination phenotype. Herein, the secretion of the neuronal connective tissue growth factor was identified to be key player in the inhibition of the myelination process (Ercan et al. 2017). On the other hand, Shepard et al. (2013) explored the possibility of a direct correlation between the duration of epilepsy and the loss of myelin (Shepherd et al. 2013). However, little is known about whether seizures and their abnormal electrical activity are responsible for the severe decrease in white matter, or whether the white matter pathology itself is part of the malformative process.

Whereas the first brain abnormalities in TSC are seen prenatally (Boer et al. 2008a), it might be arguable that an adequate myelination was never initiated; in this condition termed as dysmyelination. On the other hand, de‐ and remyelination are processes that only occur after a normal myelin development. This pathology has been described and studied very thoroughly in patients suffering from multiple sclerosis (MS) as a model disease for white matter deficiencies (Deshmukh et al. 2013). It is known that the myelin damage causes physical, mental and psychiatric problems (Compston & Coles, 2008). Additionally, studies of MS cohorts suggest an up to 3‐fold increase in seizure incidence above the general age‐matched population (Kelley & Rodriguez, 2009). The inflammatory process and the oedema associated with MS lesions, may be responsible for seizures (Thompson et al. 1993); however studies remain controversial. Other diseases, which cause damage to the brain parenchyma, such as viral encephalitis (Misra et al. 2008), stroke (Slapo et al. 2006) and Alzheimer's disease (Amatniek et al. 2006), are also linked to an increased incidence of developing seizures (Kelley & Rodriguez, 2009), and a large proportion of these patients suffer from psychiatric problems as well. White matter abnormalities were described in conditions such as autism and schizophrenia (Haroutunian et al. 2014). Furthermore, a neuroimaging study showed that abnormal white matter microstructures were mainly found in TSC patients with autism, whereas TSC patients with normal white matter do not have this psychiatric condition (Lewis et al. 2013). Combining these studies, white matter pathology, mTOR and psychiatric problems would appear to be connected, directly or indirectly, to each other and depict a complex disease network.

Perspective

One of the major advances in the last few years has been the increased possibility of detecting these types of malformations prenatally or early after birth, before symptoms appear. A prompt diagnosis allows close clinical and electroencephalogram (EEG) monitoring in infants and children even before seizure onset. Parental education also helps with the recognition of clinical seizures (Curatolo et al. 2018a,2018b,2018c). In a cohort of 74 children with TSC, Chung et al. (2017) observed that children who are diagnosed with TSC prenatally or postnatally before the seizure onset have less severe epilepsy and better development outcomes. A major issue for the development of an antiepileptogenic or disease‐modifying therapy is the identification of optimal timing for the initiation of treatment. The mTOR inhibitors proved to be a safe and effective therapeutic option for TSC‐related epilepsy, and were also able to address other systemic manifestations of the disease, potentially even benefiting the neuropsychiatric comorbidities of seizures. However, data on the safety profile of children under the age of 18 months are still limited. Furthermore, little is as yet known about the potential long‐term complications for neurodevelopment when continuous mTOR inhibition is used from early life stages. Data available up to now suggest that targeted treatment should be started as soon as possible after the failure of the first two appropriate anti‐epileptic drugs and after excluding surgery as a possible curative option. The overall risk–benefit assessment supports the use of everolimus in drug‐resistant epilepsy associated with TSC, and possibly in other ‘mTORopathies’. A major aim of future clinical trials will be to clarify whether mTOR inhibition produces genuine antiepileptogenic effects. As discussed above, TSC represents a model disease to evaluate the antiepileptogenic properties of mTOR inhibition.

Conflict of interest

None of the authors has any conflict of interest to disclose.

Authors’ contributions

A.M. supervised and designed the content of this review. A.B., H.B.S. and T.S. contributed specific topics. E.A. conceived the idea and was invited to participate at the symposium.