Epilepsy is one of the most common neurological disorders, characterized by the repeated occurrence of spontaneous bursts of neuronal overactivity, known as seizures. Seizures typically arise in restricted regions of the brain and may remain confined to these areas or spread to the whole cerebral hemispheres. The hippocampal formation and cerebral cortex are considered the most epileptogenic regions of the brain (Pitkanen and Sutula, 2002; Avanzini and Franceschetti, 2003). The behavioral manifestations of seizures, as well as the severity of the epileptic condition, strictly relate to the brain regions that are affected by overactivity. Epilepsy comprises a large group of syndromes with different etiologies. A large series of recent studies demonstrated that several developmental factors (including congenital brain malformations, altered neuronal signaling during embryonic life, and defects in postnatal maturation of neuronal networks) contribute to epileptogenesis, leading to the concept of epilepsy as a neurodevelopmental disorder.

In the first part of the review, we will focus on those alterations in embryonic development of the cerebral cortex that lead to an hyperexcitability phenotype in postnatal life. Taking examples from both animal and human studies, we will describe the role of a number of key developmental genes controlling the migration of projection (excitatory) and local circuit (inhibitory) neurons, and describe how their altered function may result in epileptogenesis. In the second part of the review, we will describe the defects in the developmental assembly of excitatory and inhibitory synapses that are also involved in epileptogenesis. Specifically, we will focus on defects in synapse elimination and remodeling during early “critical periods” that may trigger hyperexcitability later in life. The deep understanding of the complex developmental processes involved in epileptogenesis may have important implications for disease prevention and therapy.

The development of the mammalian cerebral cortex can be subdivided into three partially overlapping phases (Rubenstein, 2000, 2011; Lui et al., 2011). During the first phase, stem cells located in the ventricular and subventricular zones of the telencephalon proliferate and differentiate into neuronal precursors or glial cells. During the second phase, neurons migrate from their place of origin and reach their final destination in the cerebral cortex (Marin and Rubenstein, 2003; Kriegstein et al., 2006; Fishell, 2007; Fishell and Hanashima, 2008). The mature cerebral cortex is organized in six layers (Molyneaux et al., 2007) and contains two major types of projection neurons. The vast majority (∼80%) are glutamatergic (excitatory) neurons extending their long axon into the ipsilateral or contralateral cortex (cortico-cortical neurons, located in layers 2/3) or toward subcortical regions (cortico-fugal neurons, located in layers 5/6). The remainder (∼20%) are GABAergic local circuit neurons (inhibitory interneurons) that establish synaptic contacts with excitatory neurons located in their proximity.

Migration of glutamatergic and GABAergic cortical neurons occurs in two different ways. Glutamatergic projection neurons are generated from neuronal precursors located in the neocortical neuroepithelium via asymmetric cell divisions of cortical primary progenitors (radial glia) located in the ventricular and subventricular zones (Malatesta et al., 2003; Kriegstein et al., 2006; Hansen et al., 2010). Asymmetric divisions generate immature projection neurons, that migrate toward the cortical plate along radial glial processes (Rakic, 2007) and hence reach their final destination in specific cortical layers through the interaction with local molecular cues, such as Reelin (Frotscher, 2010). Cortical neurons are generated in an inside-out pattern, layer 6 neurons being the first to be born. The majority of GABAergic interneurons are instead generated outside of the cerebral cortex, in the ganglionic eminences of the basal forebrain (Wonders and Anderson, 2006). The immature interneurons initially migrate tangentially along the subventricular zone of the basal forebrain, and then change direction by following a radial or an oblique path to enter the cortical plate, from where they reach their final destination into the layers of the cerebral cortex (Kriegstein and Noctor, 2004).

After completion of neuronal migration, the third, last phase of cortical development involves a complex series of apoptotic and synaptogenic events aimed at finely regulating the number of mature neurons and their connections, ultimately leading to the shaping of cortical circuits.

The embryonic development of the cerebral cortex is a complex process, tightly controlled by a series of gene expression cascades (Guillemot et al., 2006). Alterations of these gene regulatory pathways during development may lead to cortical malformations, resulting in malfunction during postnatal life. Accordingly, several cortical abnormalities have been identified that are caused by mutations in key genes involved in the different phases of cortical development. Cortical malformations may result from abnormal neuronal proliferation, migration defects of both excitatory and inhibitory neurons, or altered synaptogenesis/circuit formation, and are usually epileptogenic (Barkovich et al., 2005; Guerrini and Parrini, 2010; Manzini and Walsh, 2011).

Defects of neuronal and glial proliferation during embryonic development result in epileptogenic cortical lesions. Tuberous sclerosis complex (TSC) is a genetic disorder characterized by the widespread development of benign tumors (hamartomas) in multiple organ systems including the brain. Cortical tubers, subependymal nodules, and subependymal giant cell astrocytomas represent the typical lesions observed in TSC. These malformations result in early-onset seizures, that are often accompanied by intellectual disability and autism. TSC results from mutations of TSC1 (hamartin) or TSC2 (tuberin) genes that lead to hamartomatous growths of neuronal and glial cells (Holmes and Stafstrom, 2007).

Neuronal migration disorders are a heterogeneous group of neurological conditions characterized by abnormal neuronal positioning in the cerebral cortex. Smooth cortex/lissencephaly (absent or reduced convolutions resulting in cortical thickening and smooth cerebral surface) and heterotopias (typically, ectopic nodules of gray matter located in a periventricular or subcortical position) have been associated to mutations in a number of genes regulating cortical neuron migration and are characterized by severe neurological impairment and epilepsy (Guerrini and Parrini, 2010). Migration defects of both excitatory and inhibitory neurons contribute to these conditions and lead to an altered excitation/inhibition balance and aberrant network hyperexcitability.

Following the initial generation and migration of excitatory and inhibitory neurons, immature neural networks are transformed into organized circuits through a process of refinement that is largely controlled by electrical activity (Katz and Shatz, 1996). During sensitive phases of the early postnatal life, called “critical periods,” initially exuberant connections are eliminated and the remaining synapses undergo a functional maturation (Katz and Shatz, 1996; Berardi et al., 2000). Perturbations in this developmental refinement of neuronal circuitry during critical periods may trigger hyperexcitability and epilepsy later in life. In the following sections, we will detail some of the most significant molecular mechanisms involved in neuronal proliferation, neuronal migration, and synaptic refinement (schematically illustrated in Figure 1 and summarized in Table 1), whose perturbation during embryonic or postnatal development results in epileptogenesis.

Tuberous sclerosis complex is a neurocutaneous syndrome characterized by benign tumors, early-onset epilepsy, intellectual disability, and autism (Holmes and Stafstrom, 2007). TSC results from loss-of-function mutations of TSC1 or TSC2 genes, which are crucially involved in the control of neuronal and glial cell proliferation during embryonic development. TSC1 encodes a protein (hamartin) containing two coiled-coil domains, while TSC2 encodes a GTPase activating protein (tuberin) that inhibits small G-proteins belonging to the Ras-related super-family. Hamartin and tuberin are both expressed in neurons and astrocytes of specific central nervous system (CNS) regions such as forebrain, cerebellum, and brainstem, where they form a protein–protein complex that constitutively inhibits mammalian target of rapamycin (mTOR), a serine–threonine kinase positively regulating protein synthesis, cell proliferation, and survival. The function of the TSC1/2 complex is controlled by multiple intracellular signaling pathways converging on AKT, a pro-survival and pro-oncogenic kinase that directly phosphorylates TSC2 inhibiting its function. Loss of TSC1/2 function leads to activation of the mTOR cascade and results in increased cell proliferation; conversely, inhibition of mTOR function (e.g., by rapamycin) results in growth suppression and reduced cell size (Jozwiak, 2006; Holmes and Stafstrom, 2007). The occurrence of tubers in animal models of TSC has been a point of controversy in the field. In contrast with the main feature of the human disease, Tsc1^+/− or Tsc2^+/− mice do not develop tubers (Kobayashi et al., 2001; Uhlmann et al., 2002; Goorden et al., 2007; Ehninger et al., 2008; Bonnet et al., 2009). However, conditional inactivation of Tsc1 or Tsc2 in astrocytes leads to tuber-like lesions and severe seizures in mice (Zeng et al., 2008, 2011; Way et al., 2009). More recently, in utero electroporation was used to induce homozygous (Tsc1^−/−) cell clones on a TSC1 heterozygous (Tsc1^+/−) background. This strategy resulted in cortical tuber-like lesions and a lower seizure threshold, suggesting that cells inside the tubers might have an additional somatic mutation that might contribute to the pathological phenotype (Feliciano et al., 2011). These studies clearly indicate that altering TSC1/2 signaling in specific CNS cell types is at the origin of TSC, but also point out the difficulty of modeling TSC in mice.

The histopathological features of cortical tubers in TSC reflect the antiproliferative role normally exerted by the TSC1/2 signaling complex. Cortical tubers in TSC patients are characterized by giant, dysplastic, and heterotopic neurons with aberrant dendrites and axons, as well as by proliferating astrocytes (Holmes and Stafstrom, 2007). Cortical hyperexcitability arises in the proximity of tubers, but its causes remain largely unknown (Major et al., 2009). Several hypotheses have been proposed to explain epileptogenesis in TSC. Increased expression of NMDA glutamate receptors and decreased expression of the GABA synthetic enzyme glutamic acid decarboxylase (GAD65), GABA vesicular transporter (vGAT), and GABA receptor subunits have been described in cortical tubers from human TSC patients (White et al., 2001), suggesting that excitation/inhibition imbalance in cortical circuits may contribute to epileptogenesis. Another intriguing hypothesis is that cortical tubers may alter thalamocortical connectivity during early brain development, thus resulting in hyperexcitable cortical circuits (Holmes and Stafstrom, 2007).

Seizure suppression in TSC remains a difficult task to be achieved. Infantile spasms in TSC often respond to vigabatrin (a GABA-transaminase inhibitor), but not to other antiepileptic drugs (AEDs), and the surgical removal of tubers remains in many cases the only therapeutic option (Holmes and Stafstrom, 2007). For these reasons, much emphasis has been put to the potential beneficial effects of the mTOR inhibitor rapamycin as a novel anticonvulsant and antiepileptogenic drug. Indeed, rapamycin is able to reduce seizures and prevent epileptogenesis in various animal models. For example, rapamycin treatment blocked epilepsy progression in conditional mutant mice lacking TSC1 or the TSC-positive regulator PTEN (Zeng et al., 2008; Sunnen et al., 2011), and reversed learning deficits in a Tsc2^+/− mice (Ehninger et al., 2008). Importantly, rapamycin administered to chronically epileptic rats following kainic acid (KA; Zeng et al., 2009) or pilocarpine (Huang et al., 2010) treatment has been shown to suppress acquired epilepsy, even though these results have not been replicated in mice (Buckmaster and Lew, 2011). However, some authors pointed out that rapamycin treatment in animal models is still far to be optimal, since seizures may reappear after treatment cessation, and continuous rapamycin exposure might severely affect animal growth and health (Sunnen et al., 2011). In this respect, valid alternative strategies might be represented by a high-dose pulse treatment (Raffo et al., 2011) or even prenatal exposure (to be applied in cases of familial TSC predisposition; Anderl et al., 2011), that have been successfully tested in rodents. According to these results obtained in animal models, preliminary findings in human patients are encouraging: rapamycin has been shown to induce the regression of astrocytomas in a small group (n = 5) of TSC cases (Franz et al., 2006), and to reduce seizure frequency in a single young TSC patient (Muncy et al., 2009). However, it is important to point out that while the effects of rapamycin on tumor growth in TSC patients are now well documented and reproduced in many cases, its efficacy on seizure control and other neurological deficits needs to be further investigated.

The classification and neuropathological features of genetic neuronal migration disorders have been described in other reviews (Guerrini and Marini, 2006; Guerrini and Parrini, 2010). Different forms of lissencephaly (“smooth brain”) and heterotopias have been associated to mutations in genes involved in cortical neuron migration (LIS1, DCX, ARX, TUBA1A, RELN, FLNA, and ARFGEF2). Here we will focus on two of these genes, doublecortin (DCX) and reelin (RELN), whose function in neuronal migration and epilepsy has been investigated in more detail.

Cortical GABAergic interneurons are generated in ganglionic eminences of the basal forebrain, from where they migrate to reach their final location (Wonders and Anderson, 2006). Several genes have been implicated in interneuron differentiation and migration; the role of specific interneuron types in epilepsy is a developing field, which may help us to better understand the complex neurodevelopmental processes underlying seizure onset and control. Here we will focus on Dlx, Arx, Dcx, and Reelin genes, whose role in interneuron migration and possible implications in epileptogenesis has been extensively studied in mouse models.

Studies performed on corticohippocampal organotypic co-cultures from EGFP-expressing mouse embryos demonstrated that GABA and glutamate may modulate neuronal migration in the developing forebrain. Specifically, GABA and glutamate modulate the migration of hippocampal pyramidal neurons respectively acting on GABA_A and NMDA receptors (Manent et al., 2005), while glutamate controls the migration of GABAergic interneurons via AMPA receptor activation (Manent et al., 2006). These results led the authors to postulate that the migrations of glutamatergic and GABAergic interneurons are inter-dependent: glutamate released from pioneer glutamatergic neurons controls the migration of GABAergic interneurons, which in turn would facilitate glutamate neuron migration via GABA release (Manent and Represa, 2007). In keeping with these observation, the same authors showed that prenatal exposure to some AEDs acting on GABA signaling (such as vigabatrin, valproate, and lamotrigine) results in hippocampal and cortical dysplasias in the developing embryos (Manent et al., 2007, 2008), thus raising serious concerns about the possible consequences of AEDs use during pregnancy.

In the previous sections we have considered how defects in the early stages of brain development (i.e., neuronal proliferation and migration) can induce an epileptic phenotype. Following the initial assembly of excitatory and inhibitory neurons, immature neural networks are transformed into organized circuits that subserve adult brain function (Katz and Shatz, 1996). This process of network refinement is largely controlled by electrical activity. In particular, work in the sensory cortices has clearly established the existence of so-called “critical periods,” during which patterns of activity generated by sensory experience play a critical role for maturation of cortical function (Berardi et al., 2000). During these sensitive phases of development, initially exuberant connections are eliminated and neuronal dendrites are correspondingly pruned and narrowed. The remaining synapses functionally mature (Katz and Shatz, 1996). In particular, in the visual system, synapse elimination and remodeling depends upon the amount and patterning of neural activity within the visual pathway (Fagiolini et al., 1994; Caleo et al., 2007).

Perturbations in this developmental refinement of neuronal circuitry during critical periods may trigger hyperexcitability and epilepsy later in life. One clear example comes from a study on blockade of hippocampal activity during early development (Galvan et al., 2000). In the rodent hippocampus, CA3 pyramidal neurons display an exuberant growth of axon collaterals during postnatal weeks 2 and 3; this developmental time period corresponds to a critical phase, when CA3 networks have a marked propensity to generate epileptic seizures (Swann and Brady, 1984). Exuberant CA3 axons are then remodeled with maturation, so that about half of the branches are eliminated (Gomez-Di Cesare et al., 1997). To prevent this axonal remodeling, Swann and colleagues infused tetrodotoxin (a blocker of voltage-gated sodium channels) into the rat hippocampus for about 10 days starting from P12 (Galvan et al., 2000). This transient blockade of neuronal activity resulted in the establishment of a chronic epileptic focus in adulthood, with prolonged electrographic seizures originating from the infused hippocampus (Galvan et al., 2000). These results are consistent with the idea that blockade of neuronal activity during early critical phases can enhance seizure susceptibility later in life by preventing developmental remodeling of neural circuits. Indeed, the immature brain is more prone to seizures than the adult brain (Ben-Ari and Holmes, 2006).

Recently, a persistent immaturity of glutamatergic circuitries in the hippocampus was found to underlie seizure susceptibility in autosomal dominant lateral temporal lobe epilepsy (ADLTE; Zhou et al., 2009). This study investigated the function of LGI1 (leucine-rich, glioma-inactivated 1), a gene that is implicated in about half the cases of ADLTE (Kalachikov et al., 2002; Striano et al., 2011). LGI1 encodes a protein that localizes to glutamatergic synapses, binds its receptors ADAM22 and ADAM23 (disintegrin and metalloproteinase domain 22 and 23), and copurifies with presynaptic and postsynaptic regulatory molecules (Fukata et al., 2006; Schulte et al., 2006). LGI1 expression increases in the rodent hippocampus exactly during the third postnatal week, when glutamatergic synapses are pruned and mature (Fukata et al., 2006; Zhou et al., 2009). In particular, excitatory neurons downregulate their presynaptic vesicular release probability and reduce their postsynaptic NMDA-receptor subunit NR2B; during this same period, dendritic arbors and spines are pruned and remodeled. Anderson and colleagues generated mice expressing a truncated form of the LGI1 gene, found in ADLTE patients, and showed that hippocampal glutamatergic synapses did not properly mature in these mice (Zhou et al., 2009). Specifically, (i) the normal, age-dependent decrease of probability of glutamate release was blocked in mutant mice; (ii) developmental changes in glutamate receptor composition were impaired by mutated LGI1; and (iii) dendrites of adult mutant mice remained immature, with a high density of branches and spine protrusions (Zhou et al., 2009). Importantly, mutant LGI1 mice displayed a lower threshold for the development of pharmacologically induced seizures. There was no significant effect of mutated LGI1 on GABAergic neurotransmission (Zhou et al., 2009; Yu et al., 2010). All together, these data suggest that mutated LGI1 blocks the normal functional maturation of both presynaptic and postsynaptic compartments and halts structural pruning, thus maintaining a high density of excitatory synaptic inputs converging onto hippocampal neurons and leading to a hyperexcitability phenotype (Caleo, 2009; Zhou et al., 2009). Other reports are consistent with the idea that LGI1 regulates the development of glutamatergic synapses (Fukata et al., 2010; Yu et al., 2010). In keeping with this findings, recent data show that heterozygous mice lacking LGI1 (LGI1^+/− mice) display a lower threshold to auditory stimuli induced seizures, and homozygous LGI1^−/− mice develop spontaneous recurrent seizures followed hippocampal neuronal loss, mossy fiber sprouting, astrocyte reactivity, and GCD (Chabrol et al., 2010).

In the last years, a number of epileptogenic mutations have been identified in different genes (Noebels, 2003). The majority of these genes code for voltage- or ligand-gated ion channels. Mutations in different subunits of sodium, potassium, and calcium channels underlie different forms of genetic epilepsies (“channelopathies”; Kullmann, 2010; Mantegazza et al., 2010). Mutations of the nicotinic acetylcholine receptor subunits are associated with autosomal dominant nocturnal frontal lobe epilepsy (De Fusco et al., 2000). Mutations in the voltage-gated potassium channel subunit genes KCNQ2 and KCNQ3 are associated with benign familial neonatal seizures (Leppert and Singh, 1999). Mutations in the GABA_A receptor, which is the primary mediator of synaptic inhibition, have also been found to contribute to several idiopathic epilepsies and febrile seizures (Galanopoulou, 2010). The functional consequences of these epileptogenic mutations have been amply studied, and the data have provided significant knowledge on the pathogenic mechanisms that lead to epilepsy. For more details on these findings, the reader is referred to a series of excellent reviews (Noebels, 2003; Kullmann, 2010; Mantegazza et al., 2010).

Mutations in voltage-gated sodium channel genes (Nav) are the most common genetic cause of familial epilepsy. Specifically, mutations in the Nav1.1 alpha subunit gene (SCN1A) are responsible for “generalized epilepsy with febrile seizures plus” (GEFS+; Scheffer and Berkovic, 1997) and Dravet’s syndrome (Mantegazza et al., 2010; Meisler et al., 2010). Missense mutations in Nav1.2 alpha subunit (SCN2A) are found in patients with benign familial neonatal-infantile convulsions (Avanzini et al., 2007). Functional studies demonstrate that Nav1.2 mutations result in modifications of the gating properties of the channel, with a net amplification of the sodium current and resulting greater depolarization, consistent with the hyperexcitability phenotype (Scalmani et al., 2006).

It is important to point out that it is often difficult to predict the epileptic phenotype based solely from the change in the behavior of the mutated channel. First, it is crucial to determine cell-specific patterns of expression of the mutated subunit, as expression in excitatory vs. inhibitory neurons can lead to completely opposite effects on excitability status of the neuronal network (Yu et al., 2006). Second, the cellular background (i.e., the modifier effect of other channels expressed in the same neuron) has to be considered (Glasscock et al., 2007). “Humanized” mouse models (in which human mutated channel sequences are inserted in genetically engineered mice) offer an unique opportunity to study the modulatory role of the genetic background and of the interactions between different genes. One example is represented by transgenic mice (Q54) carrying a human, epileptogenic missense mutation of the sodium channel SCN2A (Nav1.2; Kearney et al., 2001). Mice carrying the Q54 transgene on a SJL strain background exhibit severe spontaneous seizures originating in the hippocampus with onset at about 4 weeks, and progressive hippocampal sclerosis with extensive cell loss and gliosis in areas CA1, CA3, and hilus (Kearney et al., 2001; Manno et al., 2011). However, on a pure C57BL/6J background, onset of seizures is delayed and the epileptic phenotype is mild. Two modifier loci responsible for the difference in severity between strains C57BL/6J and SJL have been mapped, and the evidence points to the voltage-gated potassium channel gene Kcnv2 as one modifier (Bergren et al., 2005, 2009). These data indicate that severity of the epileptic condition may be significantly impacted by gene interactions.

Finally, in the study of channelopathies it is important to underlie the possible effects of the complex interaction between mutated genes and the environment. As we have described in the previous section, early environmental conditions shape neuronal connectivity, and the resulting changes in network organization can potently affect clinical expression of a mutated channel. In keeping with this notion, Q54 transgenic mice (carrying an epileptogenic mutated SCN2A; see above) exposed to environmental enrichment from birth show a dramatic reduction of spontaneous seizures and hippocampal cell loss (Manno et al., 2011). The data indicate that an enriched housing from birth may have profound antiepileptic and neuroprotective effects. Environmental enrichment may exert these actions by up-regulation of neurotrophic factors, plastic rearrangements in excitatory/inhibitory circuits, and stimulation of neurogenesis (Dhanushkodi and Shetty, 2008; Sale et al., 2009). Thus, changes in network organization due to environmental influences can halt epileptogenic changes and dampen hyperexcitability of neural networks.

In addition to channelopathies, defects in the control of neurotransmitter release account for a wide variety of epileptic syndromes. Presynaptic proteins that have been found to be involved in epilepsy include the synaptic vesicle (SV) protein SV2A (Crowder et al., 1999; Janz et al., 1999), the SV-anchoring phosphoproteins synapsins (Rosahl et al., 1995; Baldelli et al., 2007) and the plasma membrane fusion protein SNAP-25 (Zhang et al., 2004). Not all disruptions of the neurotransmitter machinery are equally epileptogenic, as shown by the lack of seizures in mice lacking other presynaptic proteins, such as synaptotagmin (Fernandez-Chacon et al., 2001) or synaptobrevin/VAMP (Schoch et al., 2001). Data obtained in mice mutant for synapsins and SNAP-25 are particularly interesting, since they suggest that hyperexcitability might result from developmental alterations causing an unbalance in the activity of excitatory and inhibitory neurons.

In this review, we have highlighted some of the neurodevelopmental pathways that lead to a chronic epileptic condition in adulthood. Abnormal development of the cerebral cortex (due to perturbations of neuronal proliferation and/or migration) is a frequent cause of epilepsy. In addition, more subtle alterations in the assembly and fine-tuning of neuronal networks may also lead to an hyperexcitability phenotype later in life. A better understanding of the biology of these epileptogenic mechanisms has important implications for the development of novel therapeutic approaches.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.