Epilepsy: A Comprehensive Textbook
2nd Edition

Hippocampal sclerosis is the most frequent pathologic finding in patients with temporal lobe epilepsy undergoing resective neurosurgery, and its association with epilepsy has been recognized since the early 1800s. This chapter will focus on the following: The first section reviews the important historical literature to introduce the pathology and highlight some of the clinical controversies that often first arose years ago that are still argued about today.¹⁸⁹ The middle section describes in vivo electrophysiologic findings associated with hippocampal sclerosis, with special emphasis on newly identified fast ripples that may be a surrogate marker of epileptogenesis. The final section highlights recent molecular studies that are beginning to identify possible epileptogenic mechanisms in hippocampal sclerosis related to changes in synaptic circuits, postsynaptic receptors, and intrinsic membrane properties. Because of the extensive literature, this chapter emphasizes findings mainly from human studies. Relevant studies related to animal models of temporal lobe epilepsy can be found in Chapters 36 and 40 and in other texts.¹⁶³

The literature describing the clinicopathologic relationship between seizures and hippocampal pathology spans more than 150 years. Early studies were limited to autopsy material and concentrated on determining if individuals with seizures of any type showed cerebral and hippocampal pathology. It was not until autopsy studies in the 1930s and correlative surgical-pathology studies in the 1960s that the mesial temporal lobe epilepsy syndrome was linked to severe neuronal loss in a characteristic pattern, referred to in this chapter as hippocampal sclerosis. This clinicopathologic association has been complicated by a literature that uses numerous names for the hippocampal pathology found in individuals with epilepsy. Understanding the historical use of these terms in the context of the different clinicopathologic studies is perhaps the best way to comprehend what constitutes hippocampal sclerosis and the relationship of this pathologic substrate with temporal lobe epilepsy.

Early autopsy studies frequently observed hippocampal damage in patients with different types of epilepsy. At first, the term sclerosis referred to the gross macroscopic features of a hard shrunken hippocampus, first described by Bouchet and Cazauvielh in 1825,³¹ and in other autopsy studies of that era of individuals with chronic epilepsy.^141,160 Probably the first microscopic description of hippocampal sclerosis was by Sommer in 1880,¹⁷⁸ and his case study illustrates several important clinicopathologic features of temporal lobe epilepsy that are still relevant today. The patient was a 25-year-old man with what were described as two to six “petit mal” attacks each day, and several “complete” seizures each week. As part of his epileptic syndrome, he had vivid hallucinations in which God told him he could fly, and once, as proof of his belief, he jumped from a roof. He survived the fall only to die some years later of a systemic infection. Sommer observed that the only cerebral pathology found at autopsy involved the hippocampus, and using a microscope he noted a unique pattern of neuron loss. Specifically, the pyramidal neurons of the Ammon horn were largely destroyed, especially in the portion of the hippocampus that Lorente de Nó¹⁰⁶ would later label as CA1 and prosubiculum. Neuron loss in CA1 is such a consistent finding in hippocampal sclerosis that this region is often referred to as the Sommer sector (see SS in Fig. 1B). Furthermore, Sommer described other hippocampal damage involving the granule cells and hilar neurons of the fascia dentate.

Sommer’s other historical contribution, in addition to the earliest microscopic description, was his interpretation that there must be a pathologic relationship between hippocampal damage and clinical seizure symptoms. He reasoned that the hippocampus was probably the initial site for seizure generation involving a prodrome of abnormal sensory phenomena or illusions. This is probably the first time that hippocampal pathology was associated with clinical features of what would probably be recognized today as mesial temporal lobe epilepsy.

The other important historical figure of the 1800s was Bratz.³⁶ His contribution was a detailed histologic description of hippocampal sclerosis and the observation that not all seizure types were associated with hippocampal pathology. He reported pathologic findings in the brain from 50 autopsy specimens of patients with antemortem chronic seizures associated with various etiologies common in his day, such as syphilis and cysticercosis. Bratz found hippocampal sclerosis in 25 (50%) specimens. His 1899 microscopic observations were remarkably accurate, and neuropathologists today would use the same histopathologic criteria to define hippocampal sclerosis (Fig. 1). Bratz noted that there was severe pyramidal neuron loss and gliosis throughout the hippocampus, especially in the Sommer sector of the Ammon horn. In addition, he noted that subicular neurons were not as depleted and that a fairly sharp boundary separated the profound prosubiculum neuron loss from the relatively preserved subiculum (Fig. 1B, dashed line). Within the remainder of the hippocampus, there was a second area of severe damage involving neurons between the granule cell blades. This area, later termed the end folium by Margerison and Corsellis,¹¹¹ included the CA4 pyramids and hilar neurons as described by Lorente de Nó.¹⁰⁶ By comparison, fascia dentate granule cells were only partially destroyed. In contrast to the severe neuron loss in the Sommer sector and the end folium, pyramidal neurons in CA3 and, especially, CA2 seemed to be more “resistant” to injury (therefore termed the resistant sector).

Bratz’s histopathologic description of hippocampal sclerosis is different from the hippocampal neuronal injury associated with other cerebral diseases. For example, neuron loss
related to chronic liver disease or hypoxia-ischemia involves the entire hippocampus, including CA2, along with the subiculum and parahippocampal gyrus. Hence, neuron loss and gliosis by themselves are insufficient histopathologic criteria for the diagnosis of hippocampal sclerosis. This explains the confusion that has arisen from claims by some authors that hippocampal sclerosis can be found in conditions without seizures, such as Alzheimer disease.^1,84 Thus, the histopathologic diagnosis of hippocampal sclerosis should be restricted to specimens that display the microscopic pattern of selective neuron loss as originally described by Bratz.

Bratz also noted that the neuronal loss in hippocampal sclerosis appeared old and chronic. Because hippocampal sclerosis was found in only half of the autopsy specimens of patients with epilepsy, he reasoned that this pattern of damage was not the result of repeated seizures. Instead, he suggested that hippocampal sclerosis probably generated certain types of seizures, similar to the conclusion of Sommer.¹⁷⁸ Bratz’s final contribution was his observation that many of the patients with hippocampal sclerosis had clinical histories involving early childhood convulsions. This finding would resurface about 60 years later in clinicopathologic studies of surgical patients with mesial temporal lobe epilepsy, and the nearly continuous debate since then about the pathogenesis of hippocampal sclerosis.

The studies of the 1800s confirmed an association between seizures and hippocampal pathology, but the link between hippocampal sclerosis with psychomotor and complex partial temporal lobe seizures took several more decades to confirm. The clinical autopsy study of Stauder¹⁸⁷ was probably the first. He studied 53 patients with chronic epilepsy to determine whether hippocampal sclerosis was associated with those ictal symptoms and signs that he was convinced could only be attributed to temporal lobe seizures, such as olfactory and gustatory auras. Autopsy cases were separated into three clinical groups with definite, probable, or no ictal temporal lobe symptoms by clinical description. Of the 36 hippocampal sclerosis cases at autopsy, 33 (92%) had a history of definite or probable temporal lobe seizures. By contrast, of the 17 cases without sclerosis, only two (12%) showed only probable (not definite) temporal lobe seizure symptoms and 15 (88%) had none of his defined clinical signs. These results linked antemortem temporal lobe seizure symptoms with hippocampal sclerosis at autopsy.

In another famous autopsy study from the 1960s, Margerison and Corsellis¹¹¹ examined pathology results in 55 patients with epilepsy and found a clinicopathologic association between antemortem temporal lobe seizures using clinical and electroencephalogram (EEG) criteria and postmortem hippocampal sclerosis. This study is often cited in the literature, and the reader should be aware of this study’s design, findings, and limitations. For example, their patient population is somewhat different than contemporary surgical case series of temporal lobe epilepsy patients. The patients resided in two long-term care hospitals in London because of severe mental and physical handicaps. Fifteen (27%) patients had chronic motor paralysis; in 17 (31%) the first habitual seizure was before the age of 1 year, and in 20 (36%) there were other cerebral cortical abnormalities such as congenital brain malformations, evidence of cerebral trauma, or old infections. Surgical patients with temporal lobe epilepsy typically have a lower incidence of paralysis and cerebral malformations, and the first habitual seizure usually begins around the age of 10 years.¹¹⁴ Margerison and Corsellis found that most of their patients regularly experienced generalized convulsions and only a proportion of the time had temporal lobe seizures. Margerison used clinical characteristics (n = 26; 47%) or interictal scalp EEG (n = 33; 60%) to identify those cases that in addition to generalized seizures, also probably had temporal lobe convul-sions.

At autopsy, Corsellis defined two types of hippocampal damage. The first consisted of classic “Ammon horn sclerosis”
and the second was a new entity characterized by neuron loss limited only to the end folium. Corsellis introduced the term hippocampal sclerosis and defined it as an inclusive term that included both Ammon horn sclerosis and end folium sclerosis. (As a reminder, this chapter uses the term hippocampal sclerosis only to refer to what Corsellis considered classic Ammon horn sclerosis.) As shown in Table 1, hippocampal sclerosis (Ammon horn sclerosis) was found in 58% of autopsy specimens of patients with clinical or EEG criteria for temporal lobe seizures. By contrast, 60% of patients without clinical or EEG criteria for temporal lobe epilepsy had no hippocampal pathology (Chi-square; p <0.005). The incidence of end folium sclerosis averaged 25% for all patients with epilepsy, and was not associated with temporal lobe seizures (p <0.25). Hence, as in Stauder’s 1936 study,¹⁸⁷ hippocampal sclerosis at autopsy was linked specifically to clinical and EEG characteristics of temporal lobe epilepsy. End folium sclerosis, however, was associated with repeated generalized seizures and was not a marker of temporal lobe epilepsy, as has been suggested by some authors.¹⁷⁷

In addition to hippocampal neuron injury, Margerison and Corsellis found that there could be damage to other cerebral brain areas. For example, in their 22 cases of hippocampal sclerosis, additional damage was noted in the amygdala (64%), thalamus (50%), and neocortex (27%). Such findings are not limited to autopsy studies of patients with temporal lobe epilepsy. Falconer et al.⁶⁷ found pathologic evidence for injury to the amygdala and white matter in an unspecified number of en bloc temporal lobe surgical specimens from patients with temporal lobe epilepsy and hippocampal sclerosis. These authors proposed the term mesial temporal sclerosis to indicate damage to the hippocampus and other mesial temporal sites. Recent magnetic resonance imaging (MRI) studies confirm the original autopsy pathology studies by finding evidence for extrahippocampal signal changes in a proportion of patients with temporal lobe epilepsy.^{19,48,145,208} Thus, while the traditional focus has been on hippocampal sclerosis as the probable site that generates chronic seizures, many patients will also demonstrate extrahippocampal injury. Such findings are germane to understand the pathogenesis of hippocampal sclerosis and how this pathology contributes to the development of seizures, and in deciding the extent of surgical resection in order to best achieve seizure control.

Autopsy and surgical studies support the idea that patients with intractable temporal lobe epilepsy frequently have bilateral hippocampal damage. However, the amount of damage is usually asymmetric, with one side showing hippocampal sclerosis and the other side milder forms of neuron loss. The best evidence comes from the study of Margerison and Corsellis,¹¹¹ where of the 22 patients with EEG criteria for temporal lobe epilepsy and hippocampal sclerosis, the sclerosis was unilateral in 90% and bilateral in 10%. Similar results were reported by Sano and Malamud¹⁶⁸ in another autopsy study of 29 patients with antemortem ictal “psychic phenomena.” Bilateral hippocampal damage was noted in 11 cases (39%). Meencke and Veith¹³⁸ reported on results from 650 autopsy cases of patients with chronic epilepsy. They found hippocampal sclerosis in 198 (30.5%) and in 56% of these, the findings were bilateral but asymmetric. Similar results have been reported in the limited number of surgically treated temporal lobe epilepsy patients who later died and in whom the other hippocampus became available for study.^6,126 In agreement with the pathology studies, brain MRI findings have shown that many patients with temporal lobe epilepsy have abnormal signal changes contralateral to the atrophic epileptogenic hippocampus.^174,198 Thus, the available data support the concept that while hippocampal damage is often bilateral in patients with temporal lobe epilepsy, most of the time hippocampal sclerosis is unilateral and coincides with the epileptogenic focus.

The pathologic origins of hippocampal sclerosis have been debated, often contentiously, for over 60 years, and center on whether neuron loss is the “cause” or “consequence” of
repeated temporal lobe seizures. Despite the remarkable astute clinicopathologic observations of Sommer¹⁷⁸ and Bratz³⁶ supporting the hypothesis that hippocampal sclerosis represented an area of chronic damage and gliosis that probably generated seizures, Spielmeyer,¹⁸³ Scholz,¹⁷² Peiffer,¹⁵⁹ and more recently Sutula and Pitkanen¹⁸⁹ have argued that hippocampal injury is the consequence of repeated seizures. As will be detailed below, the correct answer probably lies somewhere in between.

Earnest debate of the pathogenesis of hippocampal sclerosis began in the 1950s when surgical specimens became available from patients with temporal lobe epilepsy. Probably the first “modern” concept regarding pathogenesis was that of Earle et al.⁶⁰ They examined 157 of Penfield’s temporal lobe resections and found macroscopic pathologic abnormalities in 100, ranging from focal gyral toughness to atrophy of the entire lobe. It should be noted that for many years Penfield did not routinely remove the hippocampus en bloc for histopathologic examination. Earl et al.⁶⁰ suggested that the most likely explanation as to the “cause” of hippocampal sclerosis was transtentorial herniation of the mesial temporal lobe during a difficult delivery or as a result of birth anoxia with secondary brain swelling. They proposed that in herniating across the tentorium, the mesial temporal lobe compressed the adjacent posterior cerebral and anterior choroidal arteries to generate an ischemic lesion they termed incisural sclerosis. With time, this brain lesion “ripened” into an epileptic focus. Of interest, however, is that only 10% of their patients had a difficult birth history.

For the next 10 years, birth injury was considered to be the pathogenic etiology for hippocampal sclerosis, and this concept was initially supported by Falconer et al. in the United Kingdom.^43,140 However, after reviewing their first 100 surgical cases in the early 1960s, Falconer began to realize that there was more than one possible clinical factor associated with hippocampal sclerosis. Of 47 cases of hippocampal sclerosis, a history of difficult birth, early childhood convulsions, or head injury was noted in 42 (89%). Of these, childhood seizures were the most common predisposing factor. Falconer et al. concluded that there was a strong association between a clinical history of childhood seizures, especially febrile seizures, and the finding of hippocampal sclerosis at surgery,^39,67,68 and hypothesized that early seizures are a cause of hippocampal sclerosis. This idea has subsequently been referred to as Meyer’s hypothesis.¹³⁹

This concept, however, has been challenged by epidemiologic studies showing that the risk of temporal lobe epilepsy after febrile convulsions is very low.^3,42,149 Beginning in the mid-1990s, the UCLA group readdressed Falconer’s hypothesis in clinicopathologic studies of a large series of surgical patients with temporal lobe epilepsy.^118,121,125 By expanding the concept of potential brain insults to include any significant medical event likely to injure the brain that occurred before the onset of epilepsy, the authors found that initial precipitating injuries were strongly linked to hippocampal sclerosis in surgical cases (Fig. 2A). However, initial precipitating injuries, while most likely to occur under age 5 years, were not restricted to a young age or to febrile convulsions. In fact, studies of hippocampi removed from pediatric patients with frequent nontemporal lobe seizures generally find only limited hippocampal neuron loss, an observation that supports the notion that childhood seizures, by themselves, do not lead to hippocampal sclerosis.^121,123 Thus, the UCLA data supported previous arguments that hippocampal sclerosis predates the onset of epilepsy, is associated with some significant brain injury not necessarily linked with seizures or an early age, and is probably a cause of temporal lobe epilepsy. However, the UCLA group also found that seizure durations of 15 years or more were associated with progressive hippocampal neuron loss in all subfields of the hippocampus, and this occurred in temporal lobe epilepsy patients with or without hippocampal sclerosis (Fig. 2B). In other words, repeated seizures over time are associated with hippocampal neuron loss, but the damage is throughout the hippocampus and not in the selective pattern consistent with sclerosis. Recent neuroimaging and other studies have supported this conclusion.¹⁸⁹ Thus, the pathogenesis of most cases of hippocampal sclerosis appears to be from some antecedent brain injury (thus acquired), but there is also progressive neuron loss with longer seizure durations. The latter finding may impact the decision as to when to refer patients with temporal lobe epilepsy for surgery.

Surgical patients with temporal lobe epilepsy sometimes have more than one lesion or area of injury within the temporal lobe. This is termed dual pathology, and it appears to be more common in younger patients with temporal lobe epilepsy.¹⁴⁴ It has been difficult to interpret these studies because the definition of what constitutes a “second” pathology is sometimes vague, and it may also be unclear if both abnormalities are epileptogenic. For example, Babb and Brown⁷ found a 13% incidence of dual pathology when the other pathology was defined as a macroscopic mass lesion. Levesque et al.,⁹⁶ using the same UCLA database, included microscopic lesions, such as heterotopias, and found dual pathology in 30% of temporal lobe resections. Other authors have reported rates of dual pathology ranging from 3% to 95% depending on the definition of a second pathology.^5,88,190 Thus, a variable percentage of temporal lobe epilepsy patients will have hippocampal sclerosis plus some other histopathologic finding in the surgical specimen, such as an increase in heterotopic neurons in the subcortical white matter, but whether that second pathology contributes to seizure generation remains unclear despite recent attempts using intracranial EEG recordings.^69,86

Intraoperative electrophysiologic recordings in the form of electrocorticography (ECoG) were the original source of functional data used to localize interictal spikes (see also Chapter 172). Histologic changes such as gliosis and neuronal loss in the resected tissue often correlate with interictal spikes and other abnormalities in the ECoG. A limitation of ECoG is that the activity comes from the cortical surface, not from deep structures like the hippocampus. Early attempts to examine the properties of the “epileptic neuron” were based on single unit recordings from the lateral temporal or frontal cortex carried out in the late 1960s and early 1970s.^41,165 In a slightly later study, Wyler et al.²⁰⁷ sought to evaluate cellular discharges during intracranial recordings by identifying single unit burst discharge patterns that accompanied interictal spikes, and described synchronization of single unit discharges with one another and with local and surface field potentials during occasional intraoperative ictal events.

Although McKhann et al.¹³⁵ described use of ECoG re-corded directly from the hippocampus for determining the extent of hippocampal resection, intraoperative single neuron recordings from hippocampus are technically challenging and difficult to perform, particularly in terms of correlating such activity with hippocampal sclerosis. Perhaps the most direct intraoperative electrophysiologic correlates of hippocampal sclerosis were published by Rutecki et al.¹⁶⁶ Prior to resection of hippocampal tissue, they stimulated the entorhinal cortex or alveus and recorded either from the surface of or from within the hippocampus. They compared evoked potentials recorded

in 16 patients with hippocampal sclerosis and eight without sclerosis. They found that the patients with hippocampal sclerosis showed simple monophasic or biphasic responses with long-onset latencies (mean, 21.9 msec), whereas those without sclerosis responded with complex, multiphasic potentials that had much shorter-onset latencies (11.8 msec). This is consistent with expectations for hippocampi with severe neuronal damage and gliosis characteristic of hippocampal sclerosis. However, such findings do not explain the epileptogenicity of hippocampal sclerosis, nor do they provide histologic identification of the position of the recording electrode in relation to hippocampal laminae. Of more promise are recent investigations using multiple contact microelectrodes with vertical spacing small enough to study the voltage depth profile of tissue in hippocampal sclerosis.¹⁹²

In some surgical centers, electrophysiologic recording from depth electrodes has been accompanied by in vivo micro-dialysis.^59,204 This has provided an opportunity to evaluate the release of glutamate and other neurotransmitters associated with hippocampal seizure activity. During and Spencer⁵⁹ showed that glutamate release occurred not only during seizures, but also preceding seizure onset. Furthermore, the glutamate release was much greater on the side of seizure onset in patients with hippocampal sclerosis. They suggested that the high levels of glutamate occurring during seizures could reach neurotoxic levels and play a role in the progressive neuronal loss associated with hippocampal sclerosis (Fig. 2B). They also speculated that reuptake transporters were not functioning properly. This study also showed prominent γ-aminobutyric acid (GABA) release during seizures in the nonsclerotic hippocampus and less release on the side of hippocampal sclerosis. In another paper, During et al.⁵⁸ showed that GABA release during K⁺depolarization was increased in the hippocampus of the epileptogenic temporal lobe, but there was no difference from baseline when the microdialysis perfusate was Ca⁺² free. During et al. provided additional evidence suggesting that this GABA release was mediated by reverse transport, not synaptic release. Such findings are consistent with anatomic data showing changes in glutamate and GABA transporters.¹²⁴ A more recent study by Cavus et al.,⁴⁴ using zero flow measures of baseline amino acids, supports the conclusion that hippocampal sclerosis is associated with high lactate levels, a general reduction in the glutamate-glutamine cycle and glutamate uptake, which leads to increased glutamate levels and neurotoxicity (see Chapter 87).

The well-established association between histologic damage and seizure propensity has been clarified over the years by correlating the electrographic changes recorded using depth electrodes with various measures of neuronal loss. These include the presence of hippocampal sclerosis with focal ictal onsets or interictal spikes,^64,102,197 the pattern of hippocampal pyramidal and granule cell loss correlated with the area of hippocampal ictal onset,^8,12,180,181 cell density in sclerotic or atrophic hippocampus correlated with interhemispheric propagation time,^{99,101,179,182} or hippocampal thiopental EEG activation.¹⁰⁰ The correlation of interictal spikes with hippocampal sclerosis is clearly state dependent, because interictal spikes are widespread during slow-wave sleep but may be focal during rapid eye movement (REM) sleep.^{103,109,110,167} Sensory-evoked or event-related potentials also have been considered a means of assessing hippocampal or mesial temporal damage associated with temporal lobe epilepsy.^{46,75,133,134,136,137}

One of the most studied depth electrode EEG correlates of hippocampal sclerosis has been the morphology of ictal onsets.^{63,108,157,171,180,181,194} In these studies, a clear marker of hippocampal sclerosis has been seizure onsets characterized by repetitive high-amplitude sharp waves just before or at ictal onset. This has been variously called “hypersynchronous spiking,” “periodic spiking,” “repetitive spike pattern,” or “rhythmic sharp waves” (Fig. 3A), although some difference of opinion still exists over whether the spiking precedes or is part of the onset.¹⁹⁸ A second pattern that is commonly seen in depth EEG recordings is a low-voltage fast discharge (Fig. 3B), which occurs more commonly in nonsclerotic mesial temporal lobe, in sclerotic hippocampus after periodic or repetitive spiking, or in sclerotic hippocampus at the point of propagation to the contralateral hemisphere. Quantification of hippocampal cell loss is different between these two ictal patterns in all hippocampal fields except CA2, with cell loss greater in patients with the hypersynchronous spiking type of onset (Fig. 3C).

As was the case with some of the first studies using chronic depth electrodes, early microelectrode studies obtained chronically from patients outside the operating room often assumed the presence of hippocampal sclerosis without providing quantitative evidence based on cell counts. Thus, studies by Babb et al.⁹ and Babb and Crandall¹⁰ focused on descriptions of the bursting patterns of action potential discharge in the hippocampus of patients with mesial temporal lobe epilepsy or changes in firing rates that occurred at seizure onset.¹⁶ Later studies compared hippocampal unit activity recorded from the epileptogenic temporal lobe with that of the contralateral nonepileptic side and quantified differences in a number of measures of neuronal discharge.²⁰¹ For example, the duration of firing suppression following single pulse electrical stimulation was significantly greater on the side of seizure onset, and cells showing this suppression were those that fired synchronously with adjacent neurons.⁸³ Early quantification of firing rate, burst discharge, and degree of synchronous discharge was difficult to interpret until later studies showed that these variables were highly state dependent.¹⁸⁶ Although there were no significant differences during the waking state, recordings during polysomnographically staged non-REM sleep and REM sleep showed that neurons on the side of seizure onset had significantly greater firing rates, burst propensity, and synchronous firing as might be expected for epileptogenic tissue.

High frequency oscillations (HFOs) in the range of 80 to 500 Hz were discovered by Bragin et al.^33,34 during wide band field recordings from depth microelectrodes in the human hippocampus. Oscillations in the range of 100 to 200 Hz had previously been described in the normal rat hippocampus as “ripple” oscillations,²⁰⁹ but in the unilateral intrahippocampal kainic acid–injected rat, oscillatory activity up to 500 or even 600 Hz can be recorded.³⁵ Therefore, high-frequency oscillations in both epileptic rat and epileptic human hippocampus are called “fast ripple” oscillations to distinguish them from endogenous ripples. In patients with temporal lobe epilepsy, these oscillations are often associated with interictal spikes in which the use of wide band microelectrode recordings reveals high-frequency oscillations (Fig. 4). Quantitative studies by Staba et al.¹⁸⁵ showed that the distribution of high-frequency activity recorded in patients was bimodal, with ripples falling in the range of 80 to 150 Hz and fast ripples from 150 to 500 Hz. When oscillations were separated on the basis of the temporal lobe of seizure onset versus the contralateral temporal lobe,
fast ripple activity was found most frequently in the former and ripple activity in the latter (Fig. 4B). Of patients with unilateral seizure onsets, those with hippocampal atrophy have a higher rate of fast ripple oscillations and a lower rate of endogenous ripple oscillations.

Subsequent analysis has shown that the degree of atrophy and percent of cell loss correlates with the ratio of ripple to fast ripple discharges, indicating that hippocampal sclerosis is associated with a decrease in the rate of normal endogenous ripples and an increase in the rate of pathologic fast ripple oscillations.¹⁸⁴ In addition, the presence of high-frequency discharges during seizure onset has been demonstrated in patients with temporal lobe epilepsy during hypersynchronous spiking seizure onsets in the epileptic rat³² and during spiking and low-voltage fast seizure onsets in human hippocampus.⁸⁵ Because of the association between fast ripple oscillations and hippocampal epileptogenesis, the presence of these oscillations has been suggested as a surrogate marker for localization sites for surgical treatment of mesial temporal lobe epilepsy.⁶⁵ In FIGURE 5, an example of both ripple and fast ripple oscillations during the onset of a low-voltage fast seizure onset is shown during a wide band recording (C. Wilson, unpublished data). In FIGURE 5A, with a time base of 200 msec in this wide band recording, fast oscillations and details of the unit activity in entorhinal cortex are visible, and with a time base of 25 msec, oscillations of 400 Hz are present on a hippocampal microelectrode (Fig. 5B). In the study by Jirsch et al.,⁸⁵ similar high-frequency oscillations were recorded using wide band amplifiers, but without microelectrodes, indicating that this marker of potential pathologic tissue may be visible in wide band recordings from depth electrodes without microelectrodes.

Multiple cellular and molecular changes have now been described in human hippocampal sclerosis. One of the challenges for epilepsy researchers is to determine which of these are related to the neuron loss associated with hippocampal sclerosis and which are from repeated seizures. To sort this out, most studies compare findings from temporal lobe epilepsy patients with hippocampal sclerosis with findings from patients whose seizures arise from macroscopic mass lesions where the hippocampus is less damaged.^116,117 The advantage of this experimental design is that tissue from both groups can be identically processed after surgical removal, which is especially suitable for experiments that require living tissue (i.e., molecular biology, electrophysiology). While hippocampi from patients with lesions are less severely affected than those with hippocampal sclerosis, they still must be considered abnormal because of frequent seizures. Thus, parallel studies in animal models of temporal lobe epilepsy have proved to be a fruitful approach to delineate mechanisms underlying the development of neuropathologic and functional changes. Conversely, experiments on human tissue have been important to evaluate which aspects of human temporal lobe epilepsy are best replicated by animal models.

From experiments using tissue from humans with hippocampal sclerosis and experimental models of temporal lobe epilepsy, several molecular and cellular changes have emerged that likely play an important role in the propensity of the human hippocampus to generate seizures. Possible mechanisms of epileptogenesis can be broadly categorized into three groups: Changes in synaptic properties, neuronal connectivity, and alterations in the intrinsic properties of neurons. In additional to changes in neurons, glial cells have emerged as important players in mediating hyperexcitability in hippocampal sclerosis.

In association with neuronal loss, aberrant axon and synaptic reorganization is a characteristic feature of human hippocampal sclerosis.^11,188 The best characterized type of aberrant axon sprouting, both in human and experimental epilepsy, is the synaptic reorganization of the mossy fiber system (the axons of dentate granule neurons) because mossy fibers can be easily identified using the Timm staining procedure (Fig. 6). Mossy fiber sprouting is characterized by the formation of

novel, aberrant synaptic contacts of mossy fibers onto the proximal dendrites of hippocampal dentate granule neurons.^40,79,82 The trigger that initiates synaptic reorganization probably involves multiple mechanisms. The traditional concept is that loss of axon afferents onto the proximal dendrites of granule cells from death of hilar neurons initiates this process.⁴⁹ This idea is in line with neuropathologic studies demonstrating that the amount of hippocampal pyramidal cell loss in the CA3 region correlates with the extent of mossy fiber sprouting.⁶² Another hypothesis has suggested that granule cell neurogenesis, which is increased following status epilepticus in rats, may lead to outgrowth of new mossy fiber axons, which are then misrouted. Evidence from animal models argues against a profound impact of neurogenesis in synaptic reorganization, because disrupting status epilepticus–induced granule cell neurogenesis by irradiation does not inhibit mossy fiber sprouting.¹⁵⁵ Given the prominence of axon reorganization in human hippocampal sclerosis (Fig. 6), studies have focused on determining the distribution of proteins that are known to influence axonal outgrowth and targeting. In hippocampal sclerosis, changes in some of these proteins have been reported, such as the extracellular matrix component tenascin and the reticulin protein Nogo-A.^18,170

The formation of recurrent collaterals of granule cell axons onto other granule cells has been proposed as a major epileptogenic mechanism that may compromise the normally inhibitory function of the dentate gyrus promoting spontaneous seizures.^11,188 The properties and functional importance of this aberrant circuitry has been partially demonstrated in human hippocampal sclerosis patients.^70,79,80 However, in animal models of hippocampal sclerosis, recordings from such aberrant synaptic connections support the existence of monosyn-aptic excitatory recurrent synapses consistent with mossy fibers.^70,80,81,82 It should be noted that under normal circumstances, an excitatory disynaptic feedback circuit via mossy fiber activation of mossy cells, which project back to the dentate gyrus, also exists. Due to the loss of mossy cells and the presence of recurrent mossy fiber sprouting in hippocampal sclerosis, this normal disynaptic circuit is replaced by an abnormal more local monosynaptic one.⁴⁰ Apart from creating a monosynaptic recurrent excitatory loop, a mechanism that may further compromise inhibition is the synaptic-mediated release of zinc by mossy fibers onto the proximal dendrites of granule neurons. Because granule cells in chronic human and experimental epilepsy express GABA receptors with increased zinc sensitivity, this has been thought to cause a collapse of GABAergic inhibition, especially at the start of seizure activity.^37,38

Synaptic reorganization in hippocampal sclerosis is not restricted to the fascia dentata and excitatory mossy fiber system. Aberrant axon collaterals of CA1 pyramidal cells are increased in human and experimental temporal lobe epilepsy, and project to the stratum pyramidale and the stratum radiatum of area CA1.⁹⁵ In addition, it is likely that the projections of GABAergic interneurons also undergo significant axon reorganization^15,115 as noted for Chandelier cells.⁴ It is likely that newly formed synapses have properties that distinguish them from pre-existing synapses, and it will be important to extend our understanding of the elementary properties of these synapses in determining mechanisms of epileptogenesis in hippocampal sclerosis.¹⁶⁹

Considerable effort has been expended on characterizing inhibitory and excitatory neurotransmission in the epileptic human hippocampus.^{13,14,25,26,30,51,52,53,54,55,112,113,119,122,127,128,129,130,131,132} Excitatory glutamatergic synaptic transmission is altered in hippocampal sclerosis.⁸¹ N-methyl-D-aspartate (NMDA) receptors, one type of inotropic glutamate receptor, are assembled from different subunits termed NR1 and NR2 A–D.¹⁴⁶ The subunit composition of NMDA receptors and their alternative splicing determine the functional properties of the receptor channel. Hyperexcitability in human epilepsy has been attributed in part to modified NMDA receptor function, presumably due to prolongation of channel opening, increase in agonist sensitivity, and/or decrease in Mg²⁺ sensitivity.^90,104 Changes in the expression of NMDA receptor subunits have been described on the protein and mRNA levels, as well as in ligand binding studies in human hippocampal sclerosis.^{51,53,80,81,128,132} In addition to changes in the expression of NMDA receptor subunits, their posttranscriptional regulation appears to play an important role in chronic epilepsy. The increased open times observed in human epileptic tissues are likely due to a persistent decrease in calcineurin activity, an intracellular phosphatase that normally curtails NMDA receptor openings.¹⁹¹

α-Amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors, another class of inotropic glutamate receptor, are composed of the GluR1-4 subunits, each of which exists in two alternatively spliced forms. The functional properties of AMPA receptors are critically dependent on their subunit composition and the process of alternative RNA splicing.^148,195 Studies in human surgical tissue have demonstrated region-specific changes in AMPA receptor expression on the mRNA and protein level.^13,26 Similar findings have been reported for kainate receptor subunits, another type of inotropic glutamate receptor subtype.¹²⁷

Metabotropic glutamate receptors (mGluRs) are also altered in human epilepsy with hippocampal sclerosis,^28,97,151 and these may profoundly modulate the generation and propagation of epileptiform activity.²⁰⁵ The mGluR family consists of at least eight different subtypes.¹⁵⁰ Activation of class I mGluRs (i.e., mGluR1 and mGluR5) leads to an excitatory membrane depolarization followed by release of Ca²⁺ from intracellular stores, which appears to be mediated by inositol phosphate hydrolysis. Class II (mGluR2 and mGluR3) and class III (mGluR4 and mGluR6-8) mGluRs operate mainly via a G-protein–mediated inhibition of adenylate cyclase. mGLuR receptors are predominantly presynaptic, and they reduce transmitter release in rodent and human hippocampus,⁵⁶ probably via inhibition of voltage-gated Ca²⁺ channels.¹⁷³

GABA is the predominant inhibitory neurotransmitter in the adult brain and plays a critical role in the regulation of excitability of neuronal networks.¹⁴² GABA binding to inotropic GABA_A receptors opens the receptor ionophore, which is permeable to Cl^- and, to a lesser extent, to HCO₃. In the presence of a normal adult transmembranous Cl^- gradient, this results in expression of an inhibitory postsynaptic current that hyperpolarizes the postsynaptic neuronal membrane. GABA_A receptor–mediated synaptic currents have been studied in human hippocampal neurons from epilepsy patients.^37,176 These studies have revealed a significantly higher sensitivity of human GABA_A receptors to zinc, a finding also observed in chronic epilepsy animal models. Increased zinc sensitivity of GABA_A receptors is thought to render them susceptible to blockade by zinc released from recurrent mossy fibers during seizures, potentially causing a breakdown of dentate inhibition.¹⁴² Studies of mRNA expression for different GABA_A receptor subunits and correlation with physiologic and pharmacologic properties suggest that GABA_A receptors are regulated in a coordinate fashion in human hippocampal neurons.³⁸ The epilepsy-associated regulation of GABA_A receptor subunits has been addressed using immunohistochemistry and in situ hybridization in hippocampal sclerosis.^107,161

Apart from changes in GABA receptors, GABAergic neurotransmission is also influenced by changes in the transmembrane chloride gradient. Indeed, GABA_A receptor activation becomes depolarizing because of altered chloride homeostasis in a subset of subicular neurons in human epilepsy.^47,206 GABAergic inhibitory postsynaptic potentials are governed by the time course with which GABA is removed from the synaptic cleft. This task is subserved by GABA transporters, which may be impaired in the human dentate gyrus.¹⁵⁸ The molecular substrate for this phenomenon may be down-regulation of the underlying transporter proteins.^93,124 Impaired uptake of GABA may underlie prolonged GABA_A-mediated responses in hippocampal sclerosis.¹⁹⁹ Metabotropic GABA_B receptors inhibit neurotransmitter release from presynaptic terminals and mediate the late inhibitory postsynaptic potential. The expression of GABA_B receptors has been examined in human temporal lobe epilepsy using various approaches, including in situ hybridization, immunohistochemistry, and ligand labeling.¹⁶⁴ Given the significant changes in synaptic properties as well as the profound changes in Ca²⁺ dynamics and handling, it is not surprising that synaptic plasticity is greatly impaired in hippocampal sclerosis.²⁰

It was long held as an axiom of neurobiology that new neurons are not produced in the adult brain. It has now become clear that niches in the adult brain of rodents contain neuronal precursors that continue to produce new neurons throughout life. Whether this occurs in the normal adult human brain with hippocampal sclerosis is still unclear. One such area is the subgranular zone in the hippocampus, in which the generation of new dentate granule cells persists throughout life in rodents.⁹¹ In humans, it has been possible to demonstrate generation of new granule cells in terminally ill cancer patients receiving bromodeoxyuridine (BrDU), which can be used to label dividing cells in vivo.⁶⁶ It has also been possible to isolate multipotent precursor cells from human dentate gyrus that can give rise to functional neurons in culture.¹⁴³ It is well established that neurogenesis in the dentate gyrus is increased in animal models of mesial temporal lobe epilepsy.^154,155,156 In human epilepsy, demonstrating increased neurogenesis unequivocally has been more difficult, primarily because labeling dividing cells in vivo is not possible. Nevertheless, a few studies have shown increased expression of division markers, as well as increased expression of markers labeling neural progenitors and immature neurons in patients with temporal lobe epilepsy. These include, among others, Musashi-1, a marker of neural progenitors, nestin, doublecortin, and the polysialated neural cell adhesion molecule PSA-NCAM.^29,50 These findings suggest the possibility that neural progenitors proliferate in hippocampal sclerosis. However, hippocampal sclerosis patients do not demonstrate increased granule cell neuronal densities arguing against neurogenesis, and children with severe epilepsy demonstrate evidence for decreased granule cell neurogenesis.^115,121

The generation of new neurons in the epileptic dentate gyrus raises the question of where and how these neurons integrate into the pre-existing neuronal network. A systematic analysis of where newly generated neurons incorporate has revealed that progenitors migrate aberrantly to the hilus and molecular layer after prolonged seizures and differentiate into ectopic dentate granule cells (DGCs) in rats.¹⁵³ In human hippocampal sclerosis, ectopic putative DGCs were also found in the hilus and molecular layer of epileptic human dentate gyrus. Furthermore, hippocampal sclerosis is frequently accompanied by granulecell dispersion, a broadening of the granule cell layer suggestive of aberrant migration.⁷⁸

It has been speculated that granule cell dispersion results from abnormal positioning of newly generated granule cells. The positioning of granule neurons in the dentate gyrus is controlled by the reelin protein, which acts as a stop signal for migrating neurons. In human hippocampal sclerosis, numbers of calretinin-containing putative Cajal-Retzius cells were increased.²⁷ Interestingly, expression of reelin mRNA by hippocampal Cajal-Retzius cells is decreased, and correlates with the extent of granule cell dispersion, raising the possibility that decreased reelin production may contribute to granule cell dispersion in human temporal lobe epilepsy.⁷⁴ It is also possible that other proteins such as, for instance, cystatin C, a protease inhibitor linked to both neurodegeneration and neurogenesis, may play a role in granule cell dispersion.¹⁶² However, it should be noted that granule cell dispersion is observed in only 50% of patients with hippocampal sclerosis (not all cases), has not been observed in children with early-onset epilepsy, and has been found in circumstances of hippocampal damage without seizures.^76,120,123 Thus, the etiology of granule cell dispersion, its relationship with neurogenesis, and the role these cells might play in epileptogenesis remains unclear.

Since its discovery in the 19th century, much has been learned about hippocampal sclerosis and its relation to temporal lobe epilepsy. The pace of discovery has accelerated, particularly with the increased availability of human tissue following temporal lobe resections and the application of the tools of cellular and molecular neurobiology. There are still many unanswered questions, however, and we still do not know how to treat mesial temporal lobe epilepsy without surgery. There are also a number of clinically important questions that remain to be answered. For example, it is not clear why only a minority of patients exposed to initial precipitating brain injuries, like childhood febrile convulsions, develop hippocampal sclerosis and temporal lobe epilepsy. What are the genetic or clinical factors that predispose an individual to developing sclerosis after being exposed to injuries that otherwise do not harm others? Are there surrogate markers that can be used to identify individuals at risk for developing hippocampal sclerosis after an initial precipitating injury? Can the processes, like synaptic reorganization, that convert a damaged hippocampus into one capable of producing seizures be identified and, more critically, be augmented or prevented? What are the molecular markers and processes that produce seizures in hippocampal sclerosis?¹⁵² Can we begin to take what appear to be diverse findings from molecular and electrophysiologic studies and convert them into a unified global hypothesis of epileptogenesis in hippocampal sclerosis? Can these ideas lead to treatments without surgery, or at least prevent late recurrent seizures after surgical treatment? These and other questions pose future challenges to the study and treatment of temporal lobe epilepsy patients with hippocampal sclerosis.