Excitation/Inhibition Balance

Although there are differences in the mechanisms that underlie partial and generalized seizures, it is useful to view any seizure activity as a perturbation in the normal balance between inhibition and excitation. This imbalance can involve a localized region, multiple brain areas, or the entire brain, and often reflects a combination of increased excitation and decreased inhibition. However, the traditional concept of a seizure arising when the excitation/inhibition balance is altered is useful but oversimplified. It is now appreciated that factors once thought to be solely inhibitory (e.g., gamma-aminobutyric acid (GABA) synaptic transmission) can actually be excitatory in some circumstances (see Increased Seizure Susceptibility of the Immature Brain, below). Furthermore, synaptic inhibition is critical for certain normal brain rhythms and probably also plays a role in abnormal rhythmic activity, such as spike-wave discharges [Fritschy, 2008].

Structural Correlates of Epilepsy: Hippocampus and Neocortex

Brain regions differ in their propensity to generate seizures, based on intrinsic membrane properties, synaptic organization, cell density, and pattern of cellular inteconnectivity. Even within the same brain region, cell types differ with regard to their excitability [Steriade, 2004]. The hippocampus and neocortex are particularly prone to seizure generation. The hippocampus, with its orderly laminar organization and trisynaptic excitatory circuitry, has been used extensively in electrophysiologic studies of seizure mechanisms [Schwartzkroin and Mueller, 1987].

Familiarity with hippocampal anatomy will facilitate understanding of the physiological concepts discussed below (Figure 51-2). The hippocampal formation consists of the dentate gyrus, hippocampus proper (i.e., Ammon’s horn, with subregions CA1, CA2, and CA3), subiculum, and entorhinal cortex. These four regions are linked by prominent excitatory, largely unidirectional, feed-forward connections. The forward-projecting trisynaptic circuit begins with neurons in layer II of the entorhinal cortex that project axons to the dentate gyrus along the perforant pathway, where they synapse on granule cell and interneuron dendrites. Granule cells, the principal cell type of the dentate gyrus, send their axons, called mossy fibers, to synapse on cells in the hilus and in the CA3 field of Ammon’s horn. Several classes of inhibitory interneurons within the dentate hilus modulate on-going excitatory neuronal activity [Lawrence and McBain, 2003]. CA3 pyramidal cells project to other CA3 pyramidal cells through local collaterals, to the CA1 field of Ammon’s horn through Schaffer collaterals, and to the contralateral hippocampus. CA1 pyramidal cells send their axons into the subicular complex. Neurons of the subicular complex project to the entorhinal cortex and other cortical and subcortical targets.

Fig. 51-2 Schematic of major pathways of excitatory synaptic transmission in the hippocampal formation.

The hippocampal trisynaptic pathway begins with neurons in layer II of the entorhinal cortex that project axons to the dentate gyrus along the perforant path (1), where they synapse on granule cell dendrites. Dentate granule cells send their axons (called mossy fibers) to synapse on cells in the hilus and in the CA3 field of Ammon’s horn (2). CA3 pyramidal cells project to the CA1 field of Ammon’s horn by means of Schaffer collaterals (3). CA1 neurons send projections outward through the fornix to other brain regions and back to the subiculum. For simplicity, only the classic feed-forward projections of the trisynaptic pathway are shown. Omitted are the known backward projections and local circuit interactions.

Overview of Ion Channels

The key channels and receptors involved in normal and epileptic firing are summarized in Figure 51-3 and Table 51-2. Two major types of ion channels – voltage-gated and ligand-gated – are responsible for inhibitory and excitatory activity. Voltage-gated channels include sodium and calcium channels that function to depolarize the cell membrane toward the action potential threshold, while voltage-gated potassium channels largely dampen neural excitation. Voltage-gated channels are activated by membrane potential changes that alter the conformational state of the channel and allow selective passage of charged ions through a pore. Ligand-gated channels constitute the second type. In ligand-gated channels, a neurotransmitter (e.g., glutamate, GABA) is released from a presynaptic terminal (after presynaptic calcium influx) into the synaptic cleft and then binds with selective affinity to a membrane-bound receptor on the postsynaptic membrane. This binding activates a cascade of events, including a conformational shift to reveal an ion-permeant pore. Passage of ions across these channels results in depolarization (i.e., inward flux of cations) or hyperpolarization (i.e., inward flux of anions or outward flux of cations). Numerous ion channel mutations underlie epilepsy syndromes, giving rise to the concept of “epilepsy channelopathies” [Reid et al., 2009].

Fig. 51-3 Normal neuronal firing is illustrated by the schematic of a neuron with one excitatory (E) and one inhibitory (I) input.

Right trace shows membrane potential (in millivolts [mV]), beginning at a typical resting potential (−70 mV). Activation of E leads to graded excitatory postsynaptic potentials (EPSPs), the larger of which reaches threshold (approximately −40 mV) for an action potential. The action potential is followed by an after-hyperpolarization (AHP), the magnitude and duration of which determine when the next action potential can occur. Activation of I causes an inhibitory postsynaptic potential (IPSP). The inset shows a magnified portion of the neuronal membrane as a lipid bilayer with interposed voltage-gated Na⁺ and K⁺ channels; the direction of ion fluxes during excitatory activation is indicated. After firing, the membrane-bound Na⁺-K⁺ pump and star-shaped astroglial cells restore ionic balance.

(From Stafstrom CE. An introduction to seizures and epilepsy: Cellular mechanisms underlying classification and treatment. In: Stafstrom CE, Rho JM, eds. Epilepsy and the ketogenic diet. Totowa, NJ: Humana Press, 2004;3–29. Reprinted with kind permission of Springer Science+Business Media.)

Table 51-2 Roles of Channels and Receptors in Normal and Epileptic Firing

AHP, after-hyperpolarization; AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; EPSP, excitatory postsynaptic potential; GABA, γ-aminobutyric acid; IPSP, inhibitory postsynaptic potential; NMDA, N-methyl-d-aspartate; PDS, paroxysmal depolarization shift.

Voltage-Dependent Membrane Conductances

Synaptic Physiology

Abnormal Neuronal Firing

Specific pathophysiological mechanisms mediate each stage of seizure evolution, including transitions from a normal neuronal firing pattern to interictal epileptiform bursts, from interictal firing to seizure activity, and from the seizure to the postictal state [Stafstrom, 2004; Lado and Moshe, 2008]. Figure 51-4 depicts EEG and intracellular changes that can be seen in normal, interictal, and ictal states. In the normal situation, action potentials are generated in neuron 1 when the membrane potential reaches threshold for firing. These discharges may influence the activity of an adjacent neuron (neuron 2) synaptically, resulting in an excitatory postsynaptic potential. An adjacent interneuron (neuron 3, which is inhibitory) may also be activated by a discharge from neuron 1 after a brief delay, giving rise to an inhibitory postsynaptic potential. The activity recorded in neuron 2 reflects the temporal and spatial summation of excitatory and inhibitory postsynaptic potentials.

Fig. 51-4 Abnormal neuronal firing at the levels of the brain (A) and a simplified neuronal network (B), consisting of two excitatory neurons (1 and 2) and an inhibitory interneuron (solid circle, 3).

EEG (top set of traces) and intracellular recordings (bottom set of traces) are shown for normal (left column), interictal (middle column), and ictal (right column) conditions. Numbered traces refer to like-numbered recording sites. Notice the time scale differences in different traces. A, Three EEG electrodes record activity from superficial neocortical neurons. In the normal case, activity is low-voltage and desynchronized (i.e., neurons are not firing together in synchrony). In the interictal condition, large spikes are seen focally at electrode 2 (and to a lesser extent at electrode 1, where they may be called sharp waves), representing synchronized firing of a large population of hyperexcitable neurons (expanded in time below). The ictal state is characterized by a long run of spikes. B, At the neuronal network level, the intracellular correlate of the interictal EEG spike is called the paroxysmal depolarization shift (PDS). The PDS is initiated by a non-NMDA-mediated, fast excitatory postsynaptic potential (EPSP) (shaded area), but it is maintained by a longer, larger, NMDA-mediated EPSP. The post-PDS hyperpolarization (asterisk) temporarily stabilizes the neuron. If this post-PDS hyperpolarization fails (right column, arrow), ictal discharge can occur. The lowest traces, recordings from neuron 2, show activity similar to that recorded in neuron 1, with some delay (double-headed arrow). Activation of inhibitory neuron 3 by firing of neuron 1 prevents neuron 2 from generating an action potential (i.e., the inhibitory postsynaptic potential [IPSP] counters the depolarization caused by the EPSP). If neuron 2 does reach firing threshold, additional neurons will be recruited, leading to an entire network firing in synchrony (i.e., a seizure). NMDA, N-methyl-d-aspartate.

(From Stafstrom CE. An introduction to seizures and epilepsy: Cellular mechanisms underlying classification and treatment. In: Stafstrom CE, Rho JM, eds. Epilepsy and the ketogenic diet. Totowa, NJ: Humana Press, 2004;3–29. Reprinted with kind permission of Springer Science+Business Media.)

If this integrative concept is extrapolated to thousands of synaptic contacts, it is easy to envision the “sculpting” or grading of individual cellular responses by degrees of inhibition. For discharges of a discrete group of hyperexcitable neurons to spread to adjacent areas, the epileptic firing must overcome the powerful inhibitory influences (inhibitory surround) that normally keep aberrant excitability in check. Therefore, even a single cell can influence the output of a network and contribute to hypersynchronous firing.

Paroxysmal Depolarization Shift

The intracellular correlate of the focal interictal epileptiform discharge on the EEG is known as the paroxysmal depolarization shift (PDS) [Gorji and Speckmann, 2009]. Initially, there is a rapid shift in the membrane potential in a depolarizing direction, followed by a burst of repetitive action potentials lasting several hundred milliseconds (Figure 51-5). The initial depolarization is mediated by AMPA receptors, whereas the sustained depolarization is a consequence of NMDA receptor activation. Afterward, the PDS terminates with a repolarization phase, primarily a consequence of inhibitory potassium and chloride conductances carried by voltage-gated potassium channels and GABA receptors. The prolonged period of hyperpolarization after the PDS is mediated by inhibitory conductances and constitutes a refractory period. PDS activity in several adjacent neurons would be expected to facilitate synchronous firing [LeDuigou et al., 2009].

Fig. 51-5 The paroxysmal depolarization shift (PDS) is the intracellular correlate of an interictal spike on the surface EEG.

The temporal correlations are illustrated for an interictal epileptiform discharge (top, spike and slow-wave complex), intracellular recording (middle), and sequence of ionic conductance changes. The PDS is initiated by excitatory conductances (top four bars) and terminated by inhibitory conductances (bottom four bars). The initial depolarization (phase 1), which is an excitatory postsynaptic potential (EPSP), is mediated by α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors and sustained by N-methyl-d-aspartate (NMDA) receptors, whereas the overriding action potentials (phase 2) are generated by voltage-gated sodium channels and calcium channels. Repolarization and after-hyperpolarization (phase 3) are mediated by inhibitory conductances (i.e., multiple voltage-gated potassium channels and chloride flux, such as through gamma-aminobutyric acid [GABA] receptors).

Synchronizing Mechanisms

The hippocampal formation (see Figure 51-2) normally displays robust neuronal synchronization. Sharp waves, dentate spikes, theta activity, 40-Hz oscillations, and 200-Hz oscillations are all forms of neuronal synchronization that can be recorded in various regions of the hippocampal formation [Buzsaki and Draguhn, 2004]. Synchronization of neuronal activity appears to be intrinsic to the mechanism by which the hippocampus performs its normal functions. However, neuronal synchronization is also a hallmark of epilepsy and exaggerated synchrony among hippocampal neurons may directly generate seizures [Engel et al., 2009; Hughes, 2008]. Forms of synchronized activity that do not trigger seizures in a normal hippocampus may trigger epileptiform discharges in a hippocampus that has undergone selective neuronal loss, synaptic reorganization, or changes in receptor expression.

In the hippocampus, synchronizing mechanisms include inputs from subcortical nuclei and intrinsic interneuron-mediated synchronization. For example, high-amplitude theta activity is a salient feature of the normal hippocampus. The theta rhythm represents synchronized activity of hippocampal neurons and largely depends on input from the septum [Buzsaki and Draguhn, 2004]. Subcortical nuclei, such as the septum, have divergent inputs that target hippocampal interneurons. The divergent axon projections of interneurons and the powerful effect of the GABA_A receptor-mediated conductances that they produce enable interneurons to entrain the activity of large populations of principal neurons. These characteristics make interneurons a very effective target for subcortical modulation of hippocampal principal cell activity.

Recurrent excitatory circuits are another substrate for neuronal synchronization. Recurrent excitatory collaterals are a normal feature of the hippocampal CA3 region. CA3 pyramidal cells form direct, monosynaptic connections with other CA3 pyramidal cells, contributing to the synchronized bursts that characterize this region [Traub et al., 2004a]. In the epileptic temporal lobe, synaptic reorganization and axonal sprouting may lead to aberrant recurrent excitation, providing a synchronizing mechanism in other parts of the hippocampal formation, including CA1, subiculum, entorhinal cortex, and dentate gyrus. In the dentate gyrus of the normal hippocampus, for example, granule cells form few or no monosynaptic contacts with neighboring granule cells. However, in the epileptic hippocampus, mossy fiber sprouting results in direct excitatory interactions among granule cells [Nadler, 2003] (Figure 51-6).

Fig. 51-6 Hippocampal circuitry and seizure-induced circuit reorganization.

Granule cells (GC) receive their major input by the perforant path. The perforant path also stimulates hilar interneurons (e.g., mossy cells, basket cells) to provide feed-forward inhibition of the granule cells. Granule cell axons, the mossy fibers (MFs), make synaptic contact with CA3 pyramidal cells. Mossy fiber collaterals innervate the hilar interneurons, such as the mossy cell shown in the diagram. Mossy cells are excitatory to GABAergic basket cells that provide feedback inhibition to the granule cell. Sprouting of mossy fibers (in response to seizure-induced loss of CA3 pyramidal cells and hilar mossy cells) can result in enhanced excitation by forming autapses (i.e., an axon sprout synapsing with the dendrites of the same cell) and can augment synchronization by stimulating neighboring granule cells (not shown), contributing to epileptogenicity. It has been suggested that the sprouted mossy fibers may restore inhibition lost after seizure-induced death of hilar mossy cells by direct stimulation of deafferented (dormant) basket cells.

(From Sankar R et al. Paroxysmal disorders. In: Menkes JH et al., eds. Child Neurology, 7th ed. Philadelphia: Lippincott Williams & Wilkins, 2006;857–942.)

Mechanisms independent of chemical synaptic transmission may synchronize neuronal firing under some circumstances. Such nonsynaptic mechanisms include gap junctions, electrical field (i.e., ephaptic) effects, and changes in extracellular ion concentrations. Gap junctions allow electrical signals to pass directly between cells. Gap junctions are upregulated in epileptic brain tissue and blockade of gap junctions significantly affects the duration of seizure activity [Traub et al., 2004b]. Ephaptic effects, generated by current flow through the extracellular space, can enhance epileptogenic neuronal synchronization [Dudek et al., 1998]. Increases in extracellular potassium concentration have long been known to affect epileptic excitability and synchronization, and experiments have demonstrated clear epileptogenic effects of blocking potassium regulation (e.g., through inwardly rectifying potassium channels) [Traynelis and Dingledine, 1988; McBain et al., 1993].

Glial Mechanisms for Modulating Epileptogenicity

Glial contributions to epileptic mechanisms have been relatively neglected until recently [Wetherington et al., 2008]. Because the ionic balance between the intracellular and extracellular compartments is altered after neuronal activity (especially after repetitive discharges seen with seizures), there must exist mechanisms to restore ion homeostasis. Astrocytes are perhaps most closely associated with regulation of extracellular potassium levels (i.e., potassium buffering) because glial cell membranes are preferentially permeable to potassium. A variety of inwardly rectifying potassium channels mediate potassium uptake, and the association of glial endfeet with brain microvasculature provides a convenient “sink” for potassium release. Glial cell membrane potential changes are directly correlated with changes in extracellular potassium concentration, and blockade of potassium channels selective for glial cells results in neuronal hyperexcitability. Therefore, certain glial cells can modulate neuronal discharge by regulating extracellular potassium.

Another important role of glia is transporting synaptically released glutamate out of the extracellular space. Glia cell membranes are uniquely equipped for this role, possessing powerful glutamate transporters [Shigeri et al., 2004]. Rapid and efficient removal of extracellular glutamate characterizes normal, healthy brain tissue, and this is essential because residual glutamate would continue to excite surrounding neurons. Blockade of glutamate transporters (or knockout of the genes for these transport proteins) results in epilepsy or excitotoxicity [Meldrum et al., 1999].

Glial cells modulate excitability in other ways, as well. First, glia play a critical role in regulating extracellular pH by a proton exchanger and by bicarbonate transporter mechanisms. Even low levels of neuronal activity create significant pH transients. NMDA receptor function is very sensitive to pH. Second, glial cells are thought to release powerful neuroactive agents into the extracellular space. Glial glutamate release can excite neighboring neurons, and some investigations have suggested that other glia-related factors – such as the cytokine interleukin-1 – can have profound proconvulsant effects [Vezzani et al., 2008].

Physiology of Absence Epilepsy

The generalized spike-wave discharges accompanying absence seizures reflect a widespread, phase-locked oscillation between excitation (i.e., spike) and inhibition (i.e., slow wave) in mutually connected thalamocortical networks [McCormick and Contreras, 2001; Huguenard and McCormick, 2007]. Pyramidal neurons from layer VI of the neocortex send excitatory projections to thalamic relay neurons and inhibitory GABAergic neurons comprising the nucleus reticularis thalami (Figure 51-7). Excitatory outputs of the thalamic relay neurons impinge on the apical dendrites of layer VI pyramidal neurons in the neocortex. This thalamocortical relay is a critical substrate for the generation of cortical rhythms and is influenced by sensory inputs (e.g., from the retina) and by several brainstem nuclei that constitute the origin of cholinergic, noradrenergic, serotonergic, and other projections. This reciprocal circuitry, which is responsible in large part for normal EEG oscillations during wake and sleep states, can become hyperactive and generate generalized spike-wave discharges or can be dampened to reduce spontaneous cortical rhythms. The anatomy implies that spike-wave discharges are interrupted at cortical or thalamic levels.

Fig. 51-7 Thalamocortical circuitry is thought to form the basis of normal oscillatory rhythms, which, when perturbed, can produce generalized spike-wave discharges seen with absence seizures.

This circuitry involves excitatory projections from layer VI neocortical pyramidal neurons on to thalamic relay (TR) neurons and inhibitory neurons comprising the nucleus reticularis thalami (NRT). TR neurons send excitatory axons back to the neocortex. Activation of NRT neurons results in recurrent inhibition of adjacent neurons and TR neurons. The neurotransmitters at the excitatory (+) and inhibitory (−) synapses are thought to be glutamate and gamma-aminobutyric acid (GABA), respectively. Low-threshold (T-type) calcium channels and hyperpolarization-activated cation (HCN) channels in TR and NRT neurons help regulate intrinsic rhythmicity. Extrinsic modulatory influences on this circuitry include inputs from the forebrain and brainstem nuclei and the sensory inputs from other thalamic nuclei.

Although multiple ionic conductances are involved in rhythmic pacemaking activity, two channels are believed to play a key role in regulating thalamocortical activity. The first is a subtype of voltage-gated calcium channel known as the low-threshold (T-type) calcium channel. These channels are activated by small depolarizations of the plasma membrane. In many neurons, calcium influx through these channels triggers low-threshold spikes and activates a burst of action potentials [McCormick and Contreras, 2001; Perez-Reyes, 2003]. Such an excitatory burst is thought to underlie the spike portion of a generalized spike-wave oscillation. AEDs known to be clinically effective against absence seizures (e.g., ethosuximide, valproic acid) block T-type calcium currents.

The second important ion channel involved in the regulation of thalamocortical rhythmicity is the hyperpolarization-activated cation channel, responsible for the so-called h-current. These channels, densely expressed in the thalamus and hippocampus, are activated by hyperpolarization and produce a depolarizing current carried by an inward flux of sodium and potassium ions. This depolarization helps to bring the resting membrane potential toward threshold for activation of T-type calcium channels, which produces a calcium spike and a burst of action potentials. Hyperpolarization-activated cation channels are highly expressed in dendrites and are expressed to a lesser extent in the soma. They are also critically involved in developmental plasticity [Bender and Baram, 2008].

Unlike other voltage-gated conductances that can be labeled either inhibitory or excitatory, h-currents are both inhibitory and excitatory [Poolos, 2004; Dyhrfjeld-Johnsen et al., 2009]. Hyperpolarization-activated cation channels possess an inherent negative-feedback property; hyperpolarization activates certain cation channels, which then leads to depolarization that then deactivates these channels. The net effect of hyperpolarization-activated cation channel activation is a decrease in the input resistance of the membrane, which is the voltage change produced by a given synaptic current. H-currents tend to stabilize membrane potential around the resting potential, countering hyperpolarizing and depolarizing inputs. The relevance of hyperpolarization-activated cation channels in the pathogenesis of absence seizures was underscored by the demonstration that lamotrigine, an anti-absence agent, enhances dendritic h-currents in hippocampal pyramidal neurons and by the experimental finding that deletion of a specific hyperpolarization-activated cation isoform results in absence epilepsy [Ludwig et al., 2003].

With respect to antiepileptic action, it is uncertain whether known T-type calcium channel blockers, such as ethosuximide, in fact block absence seizures through interactions with this type of calcium channel [Huc et al., 2009]. Antagonists of GABA_B receptors and dopaminergic agonists can also interrupt abnormal thalamocortical discharges in experimental absence epilepsy models [Snead, 1996]. GABA_B receptors are involved in mediating long-lasting thalamic inhibitory postsynaptic potentials involved in the generation of normal thalamocortical rhythms, whereas brainstem monoaminergic projections disrupt these rhythms. Although it is appealing to think of absence seizures as a byproduct of dysfunction of the T-type calcium channel or h-channel, or both, the actual pathophysiologic mechanisms are probably much more complex [Hughes, 2009].

Increased Seizure Susceptibility of the Immature Brain

Seizure incidence is highest during the first decade and especially during the first year of life. Multiple physiologic factors (Table 51-3 and Figure 51-8) contribute to the increased susceptibility of the developing brain to seizures [Sanchez and Jensen, 2001; Velísková et al., 2004; Wong, 2005; Ben-Ari and Holmes, 2006; Silverstein and Jensen, 2007; Stafstrom, 2007a]. This information is mainly derived from experimental epilepsy models in rodents, in which a window of heightened excitability in the second postnatal week gives rise to a lowered seizure threshold. This period of hyperexcitability in the rodent is approximately analogous to the first year of life in the human infant [Avishai-Eliner et al., 2002].

Table 51-3 Factors Promoting Increased Seizure Susceptibility in the Developing Brain

GABA, gamma-aminobutyric acid; NR, N-methyl-d-aspartate (NMDA) receptor.

(Adapted from Stafstrom CE et al. Consequences of epilepsy in the developing brain: Implications for surgical management. Semin Pediatr Neurol 2000;7:147.)

Fig. 51-8 Schematic depiction of changes in receptor expression in the developing brain.

Relative changes in the expression of GABA receptors and glutamate receptor subtypes are shown relative to adult levels (100 percent). Human and rodent data are compared in the panel below the graph. Between about postnatal (P) day 4 in the rat (corresponding to postconceptual week [PCW] 28 in the human) until about P15 in the rodent (1–2 years of age in the human), there is a window of over-abundance of excitation. Excitatory glutamate receptor subtypes NMDA and AMPA (light blue and red lines) expression peaks during this time interval and GABA-induced depolarization (purple line) is high relative to GABA-induced inhibition (green line). These factors contribute to enhanced seizure predisposition in this age range. AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; GABA, gamma-aminobutyric acid; NMDA, N-methy-d-aspartic acid.

(From Silverstein FS, Jensen FE. Neonatal seizures. Ann Neurol 2007;62:112–120. Wiley–Liss, Inc.)

Each seizure susceptibility factor alters the net brain excitatory–inhibitory balance in favor of enhanced excitation, and each involves a multiplicity of substrates and mediators, including ion channels, neurotransmitters and their receptors, structural changes in the maturing brain, and ionic gradients. Seizure propensity in the very young involves a complex interplay between the timing of these cellular and molecular changes [Rakhade and Jensen, 2009]. Although there are disadvantages to the physiologic adaptations that make the immature brain especially vulnerable to hyperexcitability and seizure generation, such idiosyncrasies of early brain development also provide the opportunity to develop novel, age-specific therapies.

Development of Ionic Channels and Membrane Properties

The relative timing of ion channel development plays a major role in the enhanced excitability of the immature brain [Spitzer, 2006]. Sodium and calcium ion channels, which mediate neuronal excitation, develop relatively early. Action potentials persist longer in early development, and this prolonged excitation mediates a greater calcium current, particularly at the presynaptic terminal; these factors increase excitability by enhancing neurotransmitter release. However, the brain must achieve a critical balance; activity-driven calcium channel activation is necessary for normal developmental processes such as cellular differentiation, migration, and synaptogenesis, but excessive inward calcium current can cause seizures and neuronal damage. Mutations of the α-1 subunit of the sodium channel are responsible for a spectrum of developmental epilepsies, including generalized epilepsy with febrile seizures plus and Dravet syndrome (also known as severe myoclonic epilepsy of infancy) [Meisler and Kearney, 2005; Harkin et al., 2007; Stafstrom, 2009].

Through a large variety of potassium channels outward currents counter depolarizing influences and tend to stabilize membrane excitability. Potassium channels develop slightly later than sodium and calcium channels. During a window of development, the net balance of ion fluxes favors excitation [Rakhade and Jensen, 2009].

Intrinsic membrane properties, such as membrane resistance and time constant, contribute to the neuron’s response to a synaptic input. During development, neurons have relatively higher membrane resistances and shorter time constants, and a depolarizing stimulus of a given amplitude and duration can have a greater effect in an immature neuron than in a mature one [McCormick and Prince, 1987]. Such electrical factors act in concert with ionic conductances to determine the magnitude of the membrane’s response to a synaptic input.

Development of Neurotransmitters, Receptors, and Transporters

Excitatory synapses tend to form before inhibitory ones. Each glutamate receptor type, and each receptor subunit within the receptor type, possesses a distinct ontogenetic profile. For example, excitatory NMDA receptors are transiently overexpressed early in postnatal development, when they are needed for critical developmental processes [Simeone et al., 2004]. The early developmental stoichiometry of NMDA receptor subunits favors prolonged depolarizing responses. Early in development, NMDA receptor subunits NR2A, NR2B, and NR3A are elevated, favoring longer depolarizations, decreased senstivity to blockade by magnesium ions, and enhanced calcium influx. All of these physiological adaptations favor depolarization, seizures, and excitotoxicity. Similarly, the developmental profile of AMPA receptors favors subunit stoichiometries (fewer combinations with GluR2 subunits) with enhanced calcium influx. Metabotropic glutamate receptors may also be expressed developmentally to favor seizure generation [Avallone et al., 2006].

Electrical synapses appear to be more prevalent in the developing brain than in the mature brain. Electrical synapses have been reported in several brain regions, including neonatal neocortex and hippocampus. Fast-acting electrical transmission can facilitate rapid synchrony of the neuronal network and precipitate seizures and excitotoxicity [Velazquez and Carlen, 2000; Talhouka et al., 2007].

GABA receptors also exhibit a developmental profile, with different expression of GABA_A receptor subunits at different stages of development [Brooks-Kayal, 2005; Huang, 2009]. GABA’s physiologic action also varies during development. Early in development, GABA exerts an excitatory action, rather than the inhibitory effect seen later. Emerging evidence suggests that GABA is depolarizing early in development because of the particular distribution of chloride ions (Figure 51-9) [Staley, 2006b]. In the developing brain (up to about the second week of life in rats, which corresponds roughly to a few months of age in humans), the intracellular chloride concentration is much higher than in mature neurons. This distribution is due to the presence of a membrane pump, called sodium, potassium, chloride co-transporter 1 (NKCC1), that actively influxes chloride ions, utilizing the energy stored in the transmembrane sodium gradient to import a potassium ion, along with one sodium ion and two chloride ions. NKCC1’s activity results in an intracellular chloride concentration about 3–4 times greater than that found in mature neurons. Thus, when GABA_A receptors are activated by GABA release from interneurons, the postsynaptic chloride-permeable channels open and chloride effluxes out of the neuron down its concentration gradient. Chloride efflux results in a loss of intracellular negativity and a depolarizing current; the membrane potential deviates toward the chloride equilibrium potential, which is several tens of millivolts less negative than resting potential (i.e., depolarizing). The resulting depolarizing current can lead to action potential generation. This membrane depolarization is sufficient to activate NMDA receptors and allow entry of calcium ions into the cell.

Fig. 51-9 The role of intracellular chloride concentration as a function of age on neuronal responses to GABA (orange dots).

At early postnatal ages (left), intracellular chloride concentration is high due to the activity of a membrane transporter (NKCC1, green oval) that causes chloride influx; activation of GABA_A receptors at this age (lower left) causes chloride efflux and depolarization. In the mature brain (right), NKCC1 expression is much reduced, whereas another chloride transporter (KCC2, pink oval) extrudes chloride, keeping the intracellular chloride concentration relatively low. Therefore, in the mature brain, activation of GABA_A receptors (lower right) results in chloride influx and hyperpolarization of the neuron. Relative expression of the chloride transporters NKCC1 and KCC2 at the two ages is indicated by the sizes of the symbols, and relative intracellular chloride concentrations at the two ages is indicated by the sizes of the Cl⁻ lettering. Bumetanide (upper left), a loop diuretic, blocks NKCC1 action and protects against neonatal seizures, probably by reducing the depolarizing action of GABA at immature ages. GABA_A-R, gamma-aminobutyric acid type A receptor; KCC2, potassium-chloride co-transporter 2; NKCC1, sodium-potassium-chloride co-transporter 1.

(Modified from Stafstrom CE. Neurobiological mechanisms of developmental epilepsy: translating experimental findings into clinical application. Semin Pediatr Neurol 2007;14:164–172. Reprinted with permission of Elsevier, Inc.)

Even when GABA is depolarizing, another mechanism – shunting – can effectively produce inhibition by increasing membrane conductance. Shunting dissipates depolarizing synaptic currents without causing membrane potential change [Isaev et al., 2007]. That is, when chloride is passively distributed across the membrane, increasing chloride conductance will not generate transmembrane chloride currents. Therefore, the neonatal brain retains some inhibition, even in the face of depolarizing GABA action.

As development proceeds, the expression and activity of the NKCC1 transporter declines and another membrane pump is expressed, potassium chloride co-transporter 2 (KCC2) [Rivera et al., 1999]. KCC2 has the opposite action – it actively extrudes chloride to the outside of the neuron, leaving the intracellular compartment with a lower basal chloride concentration (see Figure 51-9). Binding of GABA to the GABA_A receptor of a mature neuron that expresses KCC2 will cause channel opening and influx of chloride into the neuron, down its concentration gradient. Increased intracellular negativity will hyperpolarize the neuron, moving its membrane potential closer to the chloride reversal potential, which is more negative than resting potential. These actions will keep the neuron further from the threshold for action potential firing, comprising its inhibitory action.

The transition from GABAergic excitation to inhibition is an important physiological milestone. As glutamatergic synapses develop and begin to mediate their characteristic excitatory action, GABA switches to its mature inhibitory action (via change in expression of the chloride pumps described above; also see Figure 51-8). This time point also corresponds to other developmental milestones: namely, the appearance of GABA_B inhibitory responses and disappearance of large oscillatory currents known as giant depolarizing potentials (GDPs) [Mohajerani and Cherubini, 2006]. GDPs are seen only at early ages, are dependent upon both NMDA and depolarizing GABA action, and function to facilitate the formation of synapses and neuronal circuits. The peak point of seizure susceptibility resides during this transition, when GABA is still excitatory, excitation via glutamate receptors has not yet matured, and GABA_B inhibition is not yet complete. This maximum susceptibility occurs around postnatal (P) day 10–14 in rats, corresponding to approximately term in humans, an observation of major relevance for neonatal seizures and their treatment.

Phenobarbital, which blocks GABA_A receptors, is typically used to suppress seizures in the neonatal period, but its efficacy against neonatal seizures is imperfect [Painter et al., 1999]. GABA_A receptor ontogeny is also associated with varying sensitivity to benzodiazepines, an observation relevant to seizure treatment. Therefore, novel approaches to the management of neonatal seizures are needed. In this regard, bumetanide, a diuretic similar to furosemide that blocks NKCC1 (see Figure 51-9), has been shown to suppress epileptic activity in brain slices and animal models [Dzhala et al., 2005], an effect that is enhanced with concomitant phenobarbital use [Dzhala et al., 2008].

Procedures for Antiepileptic Drug Testing

Since the introduction of phenobarbital in 1912, dozens of AEDs have been approved for medical use by the U.S. Food and Drug Administration after extensive testing in a variety of animal seizure models and in human clinical trials. Until recently, the paradigms of drug discovery have been biased toward the development of drugs sharing properties exhibited by older agents, such as phenytoin and phenobarbital. Now, AED screening programs incorporate advances in molecular and cellular neurobiology, resulting in more rational screening programs, with new molecular targets being discovered beyond neuronal membrane-bound ion channels and enzymes involved in neurotransmitter metabolism and reuptake [Rogawski, 2006].

The two most commonly used models for the routine screening and identification of new AEDs are the maximal electroshock (MES) and subcutaneous pentylenetetrazol (PTZ) tests conducted with mice or rats [White et al., 2007]. MES tests the ability of a compound to eliminate the tonic extensor component of an electrically evoked seizure, whereas PTZ (a GABA_A receptor antagonist) tests the ability of a drug to inhibit a generalized clonic seizure. MES seizures can be blocked by agents such as phenytoin and carbamazepine, which are effective in treating seizures with partial onset. In contrast, drugs effective in treating generalized absence seizures (e.g., ethosuximide) effectively inhibit PTZ-induced seizures. Valproic acid, a broad-spectrum agent, is effective in MES and PTZ tests, and is clinically effective against most seizure types. Alternative screeing tests are currently being used to assess AED effects on limbic seizures (e.g., 6-Hz test, kindling) and other seizure types [Barton et al., 2001; Bertram, 2007].

Despite the prevailing view that acute animal seizure models can predict efficacy against partial or generalized seizures in humans, there is a significant limitation of this strategy. These screening tests may fail to identify substances acting through novel mechanisms. Such is the case with levetiracetam, which is effective in the treatment of temporal lobe epilepsy but is inactive in the MES and PTZ models. A further confound is that potential AEDs are typically screened in adult animals, yet effects may be different at young ages.

Since 1993, many new AEDs have been approved for use in the United States – felbamate, gabapentin, lamotrigine, topiramate, tiagabine, oxcarbazepine, levetiracetam, zonisamide, pregabalin, lacosamide, rufinamide, and vigabatrin. Many of these agents were successful products of an extensive drug discovery campaign carried out in collaborative fashion by the National Institutes of Health, academic medical centers, and the pharmaceutical industry. Although the newer drugs appeared promising in early studies, particularly with regard to their side-effect profile, they have not yielded significantly higher clinical efficacy rates than older medications. The real impact of these newer drugs has been in diminished adverse effects and fewer drug interactions. Many children with epilepsy continue to experience intractable seizures and long-term cognitive impairment. Although AEDs are emphasized here, there is an urgent need to explore the mechanisms of action of alternative epilepsy treatments, such as dietary, surgical, and brain stimulation therapies.

Antiepileptic Drug Mechanisms of Action

Most AEDs exert their principal effects on the following molecular targets: GABA_A receptors, glutamate receptors (both NMDA and non-NMDA), voltage-dependent sodium channels, and voltage-dependent calcium channels [Rogawski and Loscher, 2004]. Each AED is likely to have multiple mechanisms of action because there is no single action of an AED that can completely account for all of its clinical effects. Several AEDs block voltage-gated sodium channels, but the interactions with other molecular targets help to define a drug’s clinical spectrum. Although this discussion focuses primarily on studies of hippocampal neurons and circuits, insights into antiepileptic mechanisms have emerged from studies of many neural tissues (e.g., spinal cord, peripheral nerve, neocortex) and experimental preparations [Rho and Sankar, 1999]. Putative molecular targets of clinically approved antiepileptic drugs are shown in Figure 51-10 and Figure 51-11, which depict key elements of central nervous system excitatory and inhibitory synapses, respectively.

Fig. 51-10 Schematic diagram of an excitatory synapse in the central nervous system and the putative major sites of action of various antiepileptic drugs.

Levetiracetam has been found to bind to a synaptic vesicle protein, SV2A, which is involved in exocytosis. AMPA, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid; HCN, hyperpolarization-activated cation; NMDA, N-methyl-d-aspartate.

(Adapted from Rho JM, Sankar R. The pharmacological basis of antiepileptic drug action. Epilepsia 1999;40:1471–1483. Reprinted with the permission of Blackwell Publishing Ltd.)

Fig. 51-11 Schematic diagram of an inhibitory synapse in the central nervous system, and the putative major sites of actions of various antiepileptic drugs.

GABA, gamma-aminobutyric acid; GABA-T, GABA transaminase; GAD, glutamic acid decarboxylase.

(Adapted from Rho JM, Sankar R. The pharmacological basis of antiepileptic drug action. Epilepsia 1999;40:1471–1483. Reprinted with the permission of Blackwell Publishing Ltd.)

For many years, the primary drugs of choice for the treatment of partial epilepsy have been phenytoin and carbamazepine. Both drugs cause voltage-, frequency-, and use-dependent block of sodium channels in a wide variety of neuronal preparations. Sustained, high-frequency, repetitive firing of neurons is limited by both agents at free plasma concentrations found in patients. Oxcarbazepine, a structural analog of carbamazepine that is reduced to a monohydroxy derivative, also blocks sustained repetitive firing and is effective only against partial seizures.

Although structurally unrelated to classic agents such as phenytoin and carbamazepine, lamotrigine blocks sodium channels in a voltage-, frequency-, and use-dependent manner. However, lamotrigine has a much broader spectrum of activity than phenytoin or carbamazepine, demonstrating efficacy against several forms of generalized seizures in addition to partial seizures. The mechanistic basis for this difference may be related to the drug’s enhancement of dendritic hyperpolarization-activated cation channels [Rogawski and Loscher, 2004].

The principal molecular targets of benzodiazepine and barbiturate antiepileptics are postsynaptic GABA_A receptors. Binding of benzodiazepines or barbiturates to their respective recognition sites results in enhancement of GABA_A receptor current. Two newer agents affecting the GABA system are vigabatrin, an irreversible inhibitor of the major GABA degradative enzyme, GABA transaminase, and tiagabine, a selective blocker of presynaptic GABA reuptake. Both vigabatrin and tiagabine elevate synaptic GABA levels.

Valproic acid is a broad-spectrum AED that induces a wide variety of biochemical and neurophysiologic changes in multiple neurotransmitter systems. Despite numerous studies, valproic acid’s precise mechanisms of action remain a mystery. Investigation has focused on its multiple effects on the GABAergic system, resulting in elevated brain GABA levels. At therapeutic levels, valproic acid also diminishes sustained repetitive firing of neocortical neurons, implicating actions on voltage-gated sodium channels as well.

Contrary to the prediction that gabapentin, a GABA analog, may act on GABAergic systems, gabapentin does not interact with GABA_A or GABA_B receptors, and does not affect GABA reuptake, synthesis, or metabolism. However, the existence of a specific binding site for gabapentin has been reported. Unlike other pharmacologic agents (e.g., flunarizine, dihydropyridines) that interact with the α₁ subunit of voltage-dependent calcium channels, gabapentin binds to the α₂δ subunit of the L-type calcium channel [Rogawski and Bazil, 2008]. Gabapentin (much like lamotrigine) enhances h-currents in hippocampal neurons, but the clinical significance of this remains unclear. Recent data confirms an interaction of gabapentin with GABA_B receptor coupling to G-proteins and modulation of an inwardly rectifying potassium channel and N-type voltage-gated calcium channels. Whatever the relevant mechanisms of antiepileptic action, elevated brain GABA levels are found in patients taking gabapentin.

Felbamate is the first pharmacologic agent to potentiate GABA_A receptor-mediated responses and inhibit NMDA receptor-mediated responses within the same drug concentration range [Rho and Sankar, 1999]. These dual actions may result in synergism with respect to seizure protection. Similarly, topiramate inhibits the AMPA or kainate subtype of glutamate receptor (specifically the GluR5 kainate receptor) and augments GABA_A receptor-mediated chloride currents [Rogawski and Loscher, 2004].

Zonisamide is a broad-spectrum agent that has a unique mechanistic profile [Rogawski and Loscher, 2004]. In cultured spinal cord neurons, zonisamide reduces sustained repetitive firing of action potentials, consistent with actions on voltage-gated sodium channels; in cultured neurons from rat cerebral cortex, zonisamide blocks low-threshold T-type calcium currents, which predicts efficacy against generalized spike-wave epilepsies.

Levetiracetam has challenged our conceptual understanding of relevant mechanisms of action. It failed the MES and PTZ tests in animals, but had a profound effect in retarding amygdala kindling in rats, and has proven clinical efficacy against partial and perhaps some generalized seizures. The strongest clue to identifying levetiracetam’s principal molecular target is the demonstration of a specific high-affinity neuronal binding site [Kaminski et al., 2008]. Levetiracetam binds to a specific synaptic vesicle protein, SV2A, which is involved in neurotransmitter release. The exact mechanism by which levetiracetam is dependent on SV2A for its antiepileptic action remains unknown [van Vliet et al., 2009].

The most recently approved AEDs in the United States are pregabalin, lacosamide, and rufinamide. Pregabalin, like gabapentin, seems to work by binding to the α2δ subunit of voltage-activated calcium channels, which reduces synaptic release of glutamate [Rogawski and Bazil, 2008]. Lacosamide reportedly possesses a novel mechanism of action – enhancing slow inactivation of sodium channels (older AEDs that block sodium channels do so by enhancing fast inactivation, which is mechanistically distinct) [Errington et al., 2008]. Whether this molecular mechanism will lead to enhanced seizure suppression is currently unkown, but it does represent an attempt to explore novel therapeutic approaches using physiologically relevant targets. Further preliminary evidence links lacosamide to modulation of collapsin-response mediator protein-2, which is involved in neuronal growth and plasticity; whether this action affects seizure suppression is unknown [Beydoun et al., 2009]. Rufinamide’s mechanism of action is currently unknown, but preliminary results point to an effect on fast sodium channel inactivation [Arroyo, 2007].

Despite their diverse molecular actions, current AEDs fail to cure the disease; rather, they attempt to eliminate the major symptom – seizures – by dampening neuronal excitation, synchronization, and spread of seizure activity. There is little experimental evidence to support the idea that these drugs are truly antiepileptogenic. As epilepsy mechanisms are better understood, it may be possible to design more specific and rational interventions.