Epilepsy: A Comprehensive Textbook
2nd Edition

Anatomists have settled on a general six-layer nomenclature for the neocortex, although some layers can be further subdivided in many cortical areas and species.^24,104 Layering is easy to see, so the different lamination patterns of neurons, myelin, or various other structural features (glia, blood vessels, biochemical markers) have served as the basis for many schemes for neocortical parcellation. The most common approach to neocortical cytoarchitecture is Nissl staining, which highlights the cell nuclei and darkly staining clumps of material around the nuclei of neurons (Fig. 1). The Nissl-stained neocortical layers are distinguished by the density and size of their constituent neurons. In general, layer I (the “molecular” layer) is just below the pia and is thin and virtually neuron free. Layers II and III are hard to separate in many areas and contain medium-sized neurons in moderate density. Layer IV is the “granular” layer because of its densely packed, small neurons. Layer V contains the largest neurons of the neocortex and includes the output cells. Layer VI (the “multiform” layer) has a wide range of neuronal sizes and shapes. This general scheme has myriad variations that early anatomists, notably Brodmann, used to distinguish and define the areas of the neocortex. For example, Brodmann’s area 17 (primary visual cortex) has a thick layer IV with a “striated” appearance, whereas area 4 (primary motor cortex) lacks an obvious layer IV, but has exceptionally large, pyramidal-shaped neurons in layer V.

Modern anatomic methods show that the axonal connections of the neocortex are also distinctively laminated (Fig. 2). As a rule, specific neural information enters an area of the neocortex only via axons from the thalamus or other areas of cerebral cortex. Additional sources of extrinsic input include a variety of nuclei in the brainstem that project diffusely across the neocortex and deliver modulatory substances, such as norepinephrine, serotonin, acetylcholine, and dopamine.¹²¹ Thalamic afferents from specific nuclei terminate primarily within layers III, VI, and (especially) IV. However, all other cortical layers may also receive thalamic input, depending on the area and thalamic nucleus in question.⁸¹ Although thalamic axons are the source of input to the neocortex from the external world, they provide only a minority of all cortical synapses; even in layer IV, only 5% to 20% of synapses are from the thalamus.^142,194 Most of the rest of neocortical synapses are from the neocortex itself. Neurons in layers II and III project densely upon layers III and IV (primarily) in other areas, whereas cells in layers V and VI project to distant areas and terminate above and below layer IV.

Nissl stains give the illusion that the neurons of the neocortex are like so many stratified grains of sand and gravel. Despite variable lamination, there is surprising uniformity of dimension. Across cortical areas and even across species, the number of neurons within a vertical column of gray matter with a surface area of 1 mm² is nearly identical.¹⁴⁸ Although some variations exist among areas,¹³ the only major exception to this rule of uniformity is the primary visual cortex of primates, which has twice the neuron density of most other areas.

Nissl stains illuminate only somata, but the vast majority of each neuron’s volume and surface area is in its dendrites and axons, which may stretch across layers and between areas. In the neocortex, a tendency exists for these cellular processes to extend strongly in the vertical dimension and for cells in vertical arrays to be densely interconnected. Pyramidal cells, in particular, seem to have evolved an efficient shape to collect synaptic input from a vertically oriented zone of tissue that may span several layers.^56,174 Physiologic studies show that the sensory or motor response properties of neurons in vertical “columns” of neocortex are often quite similar, whereas neighboring neurons along the horizontal dimension are relatively more different.^9,86,127,128 The widespread tendency for neocortical neurons to organize into vertical groups has prompted speculations that the neocortex is basically modular¹²⁸; small columns form the irreducible units of neocortical organization, and new areas and functions are built up by simply adding more columnar modules. This scheme is appealing but unproven.

Systematic anatomic and physiologic measurements have suggested a basic vertical circuit for neocortex.^7,8,72 It begins in layers III and IV, where the main thalamic input arrives. Local axons of the spiny stellate cells in layer IV then excite pyramidal neurons in the upper layers, which in turn strongly excite neurons of layer V. Layer V cells synapse upon cells of layer VI, and these return excitation both to the thalamus and to layer IV. The neocortex is primarily feeding information back upon itself, which may be the single most important thing we can say about its susceptibility to seizures. If excitation were the only story, however, the cortex would seize interminably. In parallel with the loops of excitatory circuitry, γ-aminobutyric acid (GABA)-releasing neurons also receive excitation and, in turn, inhibit pyramidal neurons, both ahead of and behind them in the circuit flow.^42,96,162

Topologic maps of the vertical organization of neocortex have been described for many different areas and species; these reveal that, in many cases, “columns” have been misnamed. For example, in visual cortex, the zones specific for ocular dominance look, from the surface, like alternating bands or slabs, whereas the organization of neurons with similar stimulus orientation preferences resembles radiating pinwheels.^11,21 Furthermore, these two maps of visual function are coextensive—they use the same sheet of cortex simultaneously but divide it quite differently. For the most part, the anatomic underpinnings of functional columns have not yet been demonstrated in detail.^28,83 In a general way, vertical connectivity must link neurons across layers, but with a high degree of specificity.

Although vertical connections received early¹¹⁴ and sustained publicity, the horizontal connections of the neocortex are also widespread, complex, and essential. Their existence provides a ready set of pathways for the rampant propagation of seizures. A few generalizations about horizontal connections can be made: They vary from very short to very long, they are reciprocal, and they tend to terminate in discontinuous, patchy patterns.

In a single brain, the neocortex consists of two continuous sheets, one in each hemisphere. Horizontal connections exist between points on this sheet at very local scales (within single microcolumns), at intermediate scales between columns, within areas, between adjacent areas, and at long distances between far-flung ipsilateral areas, as well as between homologous contralateral areas via connections through the corpus callosum. In most cases, the horizontal interconnections at all scales are reciprocal; a set of connections from cell group A to cell group B is matched by a set of connections from B to A. The patchiness of interconnections is nearly ubiquitous, although there are exceptions,¹⁸⁵ and both the regular patterns of connections and the scale of their periodicities tend to be strikingly similar across cortical areas.^{73,97,110,111}

Horizontal interconnections, especially over intermediate and long distances, are overwhelmingly excitatory, reinforcing their relevance for the spread of seizure activity. Experimental work has shown that local variations of horizontal excitatory connectivity are an important determinant of the speed and route of seizure propagation.^35,143,183 At shorter distances, on the order of adjacent intercolumn distances within one area, there may be a liberal mix of horizontal inhibitory and excitatory connections. The spatial details and relative strengths of these connections are variable with area and are probably critical in helping to determine the unique functions of each area.

Growing evidence also suggests that horizontal connections mediate some forms of experience-dependent plasticity in sensory and motor areas of the neocortex.^71,153 Cortical synapses can either strengthen or weaken, depending on their recent history of activity,¹² axons and synapses remodel dynamically¹⁶⁵ and, over weeks and months, cortical axons can also sprout new branches in response to chronic alterations of thalamic input.⁵³ It seems very likely that the grossly abnormal activity of chronic seizures could induce significant physical and physiologic changes in cortical circuitry.

At first glance, the neurons of the neocortex are dazzling in their diversity. By examining cellular morphology, connections, biochemistry, and physiology, however, order emerges.^4,195 The neurons can be divided into two major groups based on structure—spiny neurons (about 70%–85% of the total cells) and aspiny (“sparsely spiny” or “smooth”) nonpyramidal neurons.^56,141,162 Spines are small (about 1–0.2 μm on average), membranous excrescences along the shafts of dendrites.

Within the cortex, spines are notable as the major site of excitatory synaptic input onto spiny cells. The spiny and aspiny morphologic categories also correspond to basic cellular functions. Spiny cells make excitatory synapses that use the excitatory amino acid glutamate; aspiny nonpyramidal neurons make inhibitory GABA-utilizing synapses. Neurotransmitters of the neocortex will not be specifically covered in this chapter, but numerous reviews are available (see Chapters 22 and 23).

The two major classes of cortical neurons, in turn, are subdivided. The excitatory, spiny cells include pyramidal cells and spiny nonpyramidal cells (often called spiny stellate cells). Pyramidal cells are the majority class of neurons in the neocortex. They have a high density of dendritic spines (rare exceptions to this rule exist), prominent apical dendrites, and an axon that projects out of the cortex, as well as locally. Pyramidal cells vary widely in size, shape, axonal structure, and their somata appear in all layers except layer I.⁵⁶ Spiny stellate (nonpyramidal) cells are relatively small neurons of layer IV, with high spine density and axons that terminate in excitatory synapses. Spiny stellate cells are distinguished from pyramidal cells by the
absence of an obvious apical dendrite, and their axons usually do not leave the cortex (although a minority may send axons out of the cortex).¹⁸⁸ Because of their excitatory function and spiny dendritic structure, spiny stellate cells may be considered a simple variant of the pyramidal cells. This view is supported by the development of spiny stellate cells; apparently, they begin life as pyramidal neurons, and during postnatal maturation, they gradually retract their apical dendrites.¹⁸⁷

The inhibitory, sparsely spiny, nonpyramidal neurons are a very diverse group.^102,118,195 Whereas their dendritic patterns vary systematically,³⁷ a more useful characteristic is the pattern of their axon arbors, which do not leave the local area of cortex but which may be distinctive and specific.⁶¹ As schematized in FIGURE 3, common morphologies for inhibitory interneurons include double bouquet cells, whose axons and dendrites project within a long, narrow, vertical column; bitufted cells, whose collaterals form two highly branched protrusions from the soma; bipolar cells, which have long collaterals projecting vertically above and below the soma; multipolar cells, which have round somata with multiple dendrites projecting radially; neurogliaform cells, whose dendrites form relatively tight, symmetrical spheres around their somas; basket cells, whose axons tend to form basket-like shapes around the somata of their target cells; Martinotti cells, which can have their soma in layers II to V and project their axon up to layer I, where they ramify up to several millimeters; and axo-axonic cells, whose dendrites terminate on the axon initial segments of target pyramidal cells, forming structures that resemble candlesticks (thus the alternative name, chandelier cells). Note that this is not an exhaustive list of inhibitory neuron morphologies, nor do neurons always fit clearly into one morphologic subtype.

It has been shown that somatic and perisomatic targeting cells can readily suppress action potentials,¹²⁴ suggesting that these inhibitory neurons could closely control the output of their target neurons. Because of their potentially strong control over pyramidal cell output, loss of axo-axonic cells has been implicated in epilepsy.⁵⁵ It has also been proposed, paradoxically, that axo-axonic cells actually excite pyramidal neurons because of a relatively high internal chloride ions [Cl^-] in axon initial segments.¹⁷³

Other inhibitory neurons terminate on the dendrites of their target cells. Examples include double bouquet cells, bitufted cells, bipolar cells, neurogliaform cells, and Martinotti cells (Fig. 3A–C, E, G). The nature of the inhibition from these cells probably differs from interneurons that synapse in somatic or perisomatic regions. Dendrite-targeting inhibitory cells can alter the way their target cell integrates synaptic input within the dendrites, and it is thought that these cells can suppress dendritic calcium ion (Ca²⁺) spikes,^106,124 decreasing the ability of excitatory inputs to propagate along the dendrites.

Classes of inhibitory cells may also be defined by interesting variations in peptide cotransmitters (in addition to the GABA that they all contain),⁸⁰ by different types of Ca²⁺ binding proteins they express,^100,184 and by their expression patterns of a wide variety of genes.^22,131,170 Three main Ca-binding proteins have been useful as markers of interneuron types: Calbindin (CB), calretinin (CR), and parvalbumin (PV). The most common neuropeptides used to identify inhibitory interneuron subtypes include cholecystokinin (CCK), somatostatin (SOM), and vasoactive intestinal peptide (VIP). Double labeling studies have shown that several of these markers tend to co-exist within cells (e.g., CR and VIP), but others never or very rarely co-exist (e.g., PV and SOM). Kawaguchi and Kubota¹⁰³ divided prefrontal neocortical inhibitory neurons in layers II through VI into four main groups based on these markers. Group I contains PV; group II contains SOM and neuropeptide Y (NPY); group III contains CR, VIP, and CCK; and group IV contains large CCK cells. However, it is not clear that these groups correlate well with specific morphologic or physiologic subcategories of inhibitory neurons. One pattern that does emerge is that PV
appears to be associated with somatic/perisomatic or axo-axonic targeting cells, whereas CR is largely associated with dendritic targeting cells (reviewed in¹¹⁸).

Just as neurons vary in their shape, transmitter, and synaptic connections, they may also differ in their intrinsic physiologic properties.^46,49,112 “Intrinsic physiology” is defined as a neuron’s inherent ability to respond to stimuli, apart from the effects of the neural network surrounding it; intrinsic physiologic properties derive from the cell’s membrane (and its ion channels, primarily) and metabolic properties (such as ion pumps and control systems, second-messenger systems, and energy management). Takahashi¹⁷⁶ was the first to show that the intrinsic properties of neocortical cells had diverse properties. Recording from the pyramidal tract neurons of the cat, he found that axonal conduction velocities correlated with somatic spike duration, time constant, input resistance, and duration of the afterhyperpolarization. Since then, numerous studies have used methods of intracellular recording, single-cell staining, and isolated brain slices in vitro to demonstrate that the intrinsic physiology and morphology of neurons in the neocortex are related. The impact of a neuron’s intrinsic properties can be appreciated from its response to simple injected current stimuli, as shown in FIGURE 4. Some cells produce a tonic, uniform spike frequency; others display strong adaptation; and still others generate periodic bursts. Clearly, when faced with the same input, each of these neurons provides a very different message to its postsynaptic targets. Myriad variations on these basic themes exist in the neocortex and in the rest of the brain.

Ironically, some of the characteristics of neocortical seizures that make them so devastating clinically have also made them a popular subject of experiments: Seizures are readily induced by various means, their electrical correlates are large and easily measured, and neocortical activity during seizures is very simple, compared to activity during normal functions. Detailed studies of experimental animal models of partial seizures have suggested numerous hypotheses of neocortical seizure generation.^10,58,92,144 This section focuses on one explicit proposal for the cellular origins of seizure initiation and propagation.⁴³

During a seizure, the neocortex generates hypersynchronous activity that may encompass almost every neuron across a wide area. Where does the trigger for these synchronous events arise? How does that trigger spread to engulf the activity of other neurons? How does a very localized seizure propagate into adjacent areas and beyond? One simple way to induce seizures is to reduce the level of synaptic inhibition. Experiments on slices of rodent neocortex in vitro have provided a variety of evidence that certain pyramidal neurons in layer V may act as both the initiators and propagators of hypersynchronous activity. Seizure-like activity is readily induced when synaptic inhibition is depressed by applying antagonists of GABA_A receptors. Even a modest reduction of inhibition, perhaps by 20%, leads to sharply synchronized, often rhythmic activity that can propagate for many millimeters across a cortical slice.³³ Under these conditions, synchronized activity is labile, changing its shape from trial to trial. When GABA_A receptors are more strongly blocked, synchronized events are more stereotyped, often spontaneous, and propagate unimpeded across the cortex.^{35,41,74,76,78,143} In each case, layer V is implicated as the initiation site for synchronized activity. If inhibition is eliminated, it is in layer V that epileptiform currents first appear during each event. Layers IV and V are the most sensitive to GABA_A antagonists—they have the lowest threshold for event initiation, and synchronous events can be most easily blocked there by applying GABA itself.⁴¹ With GABAergic inhibition only slightly reduced, the IB neurons of layer V are the only class of pyramidal cells that consistently generate synchronized excitatory events and spike firing; other pyramidal cells tend to be synchronously inhibited and only weakly excited.³³

Moreover, when slices are further dissected with horizontal cuts and bathed in low doses of GABA_A antagonists, those microslices containing layer V (but not other layers) are able to support both the initiation and propagation of
synchrony.^178,179 Small injections of convulsant drugs into neocortex in vivo are most effective at producing epileptiform activity when they are placed in the vicinity of layer IV.^60,84

Synchronous activity can also be generated by boosting synaptic excitation. For example, the function of one type of glutamate receptor, the N-methyl-D-aspartate (NMDA) receptor, is facilitated by lowering the magnesium ion [Mg²⁺] in the bathing solution. This reduces the voltage dependence of the NMDA receptor–operated channel¹³⁴ and, in the neocortex, leads to spontaneous, highly synchronous events that occur in rhythmic epochs with 4- to 7-Hz discharges.^158,171 In horizontally dissected slices in low [Mg²⁺], a small fragment of layer V is all that is necessary to generate the same type of rhythmic synchrony as the intact slice.^62,158 Microslices without layer V resemble intact control slices and are incapable of supporting rhythmic, synchronous activity.

Once synchronous events are initiated locally by neurons of layer V, they may propagate both vertically into other layers and horizontally into adjacent cortex.^32,35 During moderate disinhibition, the pathway of preference for horizontal movement is within layer V.¹⁷⁸ When inhibition is strongly suppressed, alternative pathways can be found either above or below layer V after removal of layer V. However, it is likely that the pathway through layer V remains the primary mediator of horizontal movement when the cortex is intact.¹⁷⁹ Layer V is also implicated in other forms of cortical synchrony, including chronic models of epilepsy (see later discussion).

Novel electrical stimulation–based techniques that target the excitation thresholds of layer V pyramidal cells may prove to be an effective method of seizure control. Electric field modulation of neuronal excitability has been shown to successfully slow and even halt propagating neocortical epileptiform activity in acute preparations.¹⁴⁷

In summary, layer V has neurons with the appropriate intrinsic membrane properties, intralaminar connections, neurotransmitter systems, and axonal outputs to generate highly synchronized activity and to impose it on the other neurons of the cortex. It may not be the sole synchronizing network in the cortex, but it is a major one. We do not know which properties of layer V are most responsible for its capabilities. A case can be made that intrinsically bursting neurons are likely instigators of synchrony: They are prevalent in layer V, they are exceptionally excitable, and they have the requisite local connections.^32,41,76,158 However, bursting may be only part of the answer or none of it. Unusually strong or dense interconnections between bursting cells, relatively weak inhibition, a unique complement of neurotransmitter receptors, or especially excitable dendrites could be more important. It is likely that several characteristics conspire to make layer V a uniquely excitable network in neocortex.

Seizure disorders have many causes. Sometimes a clear genetic basis is present (see Chapters 17, 18 and 37). Many developmental anomalies are accompanied by severe epilepsy (see Chapters 14 and 259).^3,189 Neocortical insults from tumors, trauma, stroke, and infection are also common instigators (see Section X, Epilepsy Syndromes). The origins of most seizure disorders undoubtedly lie in abnormalities of the basic biology of the cerebral cortex. Unfortunately, the large majority of human seizure syndromes cannot yet be described in cellular or molecular terms, and many cannot be described in any terms beyond their signs and symptoms.

While it has been possible to examine epileptogenic tissue resected from epileptic human patients, recording from normal, relevant control tissue for comparison is rarely feasible. Various chronic animal models of different epilepsies have therefore been developed to examine the differences between normal and epileptogenic tissue (Chapters 36,37,38,39). In general, hyperexcitable or epileptogenic neural tissue from both animal models and human epilepsy patients have demonstrable abnormalities of structure and function. These involve growth of dendritic or axonal arbors, changes in synaptic densities or transmitter receptor numbers, and either increases or decreases in specific types of neurons (Fig. 6). We highlight here a few studies of the changes in anatomy and physiology that are associated with seizure-like activity and hyperexcitability in neocortex.