2.1. Seizures in patients with AD
The occurrence of seizures in patients with AD has been thoroughly reviewed recently [4
]. The incidence of unprovoked seizures is 5–10-fold greater in sporadic AD than in reference populations, and as much as 87-fold greater in patients with “early” disease onset (before 60 yrs of age [5
]). In autosomal dominant forms of AD, the relationship between seizures and AD is most remarkable. In these forms of AD, there are either mutations in amyloid precursor protein (APP), the precursor to Aβ, or one of the presenilins (presenilin-1, PS1; presenilin-2, PS2; critical components of the γ-secretase complex that cleaves APP to produce Aβ), or other genetic causes such as duplication of the APP gene. Approximately 50–80% of patients with these forms of AD present with overt (convulsive) seizures [11
]. All of the genetic alterations that cause autosomal dominant AD increase Aβ production or aggregation, which suggests that Aβ accumulation contributes to the seizures. This idea is supported by studies of Down’s syndrome, in which there is an extra copy of the APP gene, and in which a high percentage of individuals develop seizures [6
]. Recent studies in mouse models of AD suggest that neurofibrillary tangles and specifically tau, which is a major component of tangles, also contribute to seizures [17
Despite the significant seizure incidence in patients with AD relative to reference populations, episodes of convulsive seizures are relatively infrequent [6
]. However, non-convulsive seizures may be underestimated because of their subtle behavioral manifestations, which could easily be missed by someone without an understanding of epilepsy. Retrospective studies and clinical reports have demonstrated that patients with AD do have an increased incidence of non-convulsive seizures [8
]. This idea is also supported by the fact that mouse models of AD also exhibit non-convulsive seizures [21
]. Thus, convulsive seizures may be fairly well documented, but non-convulsive seizures are likely to be underestimated [8
]. Non-convulsive seizures are similar in both humans and mice: they can include behavior such as staring with a frozen posture for a few seconds, making them difficult to detect without EEG recordings [23
]. Two examples of non-convulsive seizures are absence seizures and partial seizures. Absence seizures are associated with a blank expression, even in the middle of a conversation, which makes the person appear to be ‘staring off into space’; these ‘absences’ are accompanied by generalized 3-Hz spike-and-wave discharge [23
]. Partial seizures may be accompanied by “automatisms,” which are perseverative movements such as licking the lips or chewing without food in the mouth. For both absence and partial seizures, the episodes are usually not remembered by the patient. One of the potential reasons for a patient’s inability to recall seizures is that seizure activity associated with these non-convulsive behaviors interrupts normal thalamocortical sensory processing (in absence seizures) or memory that is dependent on structures such as the hippocampus (in partial seizures). Although not recognized by the patient, the behaviors associated with these types of non-convulsive seizures are commonly described by caregivers of patients with AD [26
2.2. Epilepsy and TLE in AD
As mentioned above, patients with AD can exhibit seizures. But do these patients have epilepsy? The definition of epilepsy only requires that an individual exhibit more than one spontaneous seizure, so in fact, the answer may be yes.
Do individuals with AD have temporal lobe epilepsy (TLE)? Here, the definition of TLE is important. Some researchers consider TLE to depend on a particular pattern of pathology called mesial temporal lobe sclerosis (MTS), which is not clear in AD. When one examines hippocampal pathology in AD and TLE, some similarities are observed, such as sparing of dentate gyrus granule cells when other neurons in the hippocampus are damaged. In addition, there is circuit reorganization and other changes in the dentate gyrus that are similar in some patients with AD and TLE [27
]. Therefore, some pathological characteristics are shared between some patients with AD and some patients with TLE.
The type of seizure that characterizes TLE is also important to consider: seizures in TLE are typically simple partial or complex partial. However, this is not always true. For example, generalized tonic-clonic seizures can develop. A variety of seizure types also occur in patients with AD [8
]. Therefore, one cannot necessarily infer that all patients with AD who have recurrent seizures have a TLE-like syndrome, but some may have, and in others, some overlap in pathology and seizure type can occur.
2.3. Seizures in transgenic mice used to study AD pathophysiology
One of the difficulties in resolving clinical questions about seizures in AD is practical — lengthy or invasive (and hence more accurate) EEG monitoring is prohibitive. In these cases, animal models can provide an alternative approach to gain more insight. In this regard, mouse models of AD pathophysiology, particularly those that express high levels of Aβ, have provided a great resource. Recent studies in transgenic mouse models of AD with high levels of Aβ suggest that seizures occur much earlier than had been expected relative to the progression of Aβ accumulation and amyloid pathology [21
]. These new data have suggested a different time course in the disease, where seizures and cognitive impairment occur earlier than the development of amyloid plaques. In addition, the data suggest that seizures could contribute to the progressive decline in cognition and progressive neurodegeneration in AD. Although these hypotheses have to be tested directly in human patients, the studies in mouse models have led to a greater understanding of how seizures might arise and affect cognition in AD.
The mouse models in which seizures have been primarily studied to date are transgenic mice that express human amyloid precursor protein (hAPP) carrying one or more mutations associated with familial forms of AD. Mutant APP is expressed in neurons in a relatively widespread manner in these mice because the promoter used to drive neuronal expression is normally located throughout the brain, such as platelet-derived growth factor (PDGF). In some mouse models, hAPP transgenic mice are crossed with transgenic mice overexpressing mutant human PS1 to simulate another genetic, autosomal dominant form of AD. These mutations increase the production and accumulation of Aβ within the first months of life, which continues as the animals age. Electroencephalographic recordings from these mice often demonstrate spontaneous seizures early in adulthood, and the seizures can occur with and without convulsive behaviors [17
]. The seizures are robust electrographic events, even if they are not accompanied by convulsions, with large voltage excursions, lasting tens of seconds or longer [21
]. The typical rhythmicity of seizures and fast frequencies are also observed. Some spike-wave discharges are found. When multiple electrodes are used, many seizures appear to be generalized. The mouse models of AD that exhibit seizures also exhibit numerous impairments in behavioral tests of cognitive (e.g., spatial memory) and emotional (e.g., anxiety) functions [21
]. A causal relationship between the seizures and the cognitive/behavioral impairments is suggested by the reduction in deficits when seizures are reduced [17
Importantly, spontaneous seizures develop with age and rising Aβ levels and are generally not evident in transgenic mice that do not have high levels of Aβ [17
]. Thus, epileptiform activity is unlikely to be a nonspecific effect of transgene overexpression. The role of Aβ may be important even before plaques form, presumably due to the actions of oligomeric Aβ species [21
]. It is notable that these AD mouse models exhibit little or no neuronal loss [36
], indicating that the general assumption that seizures result only at the end stage of AD, after extensive neurodegeneration, is not true, at least in the mouse models. It would be valuable to know whether this occurs also in patients with AD, but systematic long-term quantitative EEG evaluation (with neuropathology) is not available. Epileptiform activity in hAPP mice induces gene expression and circuit reorganization that is potentially neuroprotective, such as upregulation of neuropeptide Y levels in GABAergic neurons and the mossy fiber pathway of the hippocampus [21
]; upregulation of neuropeptide Y also occurs in epileptic rodents and in the same hippocampal pathways — the mossy fiber pathway and GABAergic neurons called HIPP cells [17
]. Therefore, the “neuroprotection” in hAPP mice does not necessarily (effectively) protect against seizures. It is important to keep in mind, however, that the magnitude of induction of the “neuroprotective” alterations may provide a reliable measure of the severity of underlying seizure activity even if their functional effects are not clear. The reason is that neuropeptide Y expression in mossy fibers appears to occur whenever there are recurrent seizures in rodents [21
]. These “neuroprotective” alterations may also be relevant to cognitive deficits, because the magnitude of some of the changes in circuitry in hAPP mice that are considered protective is correlated tightly with hippocampal-dependent memory deficits [34
]. The remarkable similarities between the hAPP or hAPP/PS1 mice and rodent models of epilepsy [27
] suggest that seizures play a potential role in cognitive deficits and underscore the need to identify the underlying mechanisms.
2.4. Epilepsy in AD mouse models: is it TLE?
Changes in hippocampal circuitry in the same mice are similar to what has been found in animal models of TLE (described further below). Therefore, one might assume that these animal models of AD have TLE or a TLE-like syndrome. However, it is hard to judge from the literature whether this is true or not. Hippocampal discharges and seizures involving temporal lobe areas have been shown in mouse models of AD Ref. [21
], but this does not prove that a TLE-like syndrome exists. One needs to record from multiple sites in the brain to be sure that the hippocampal discharges/seizures are initiated in that structure, rather than caused by a seizure elsewhere.
There are also a few additional reasons to be cautious about concluding that mouse models of AD have a TLE-like syndrome. First, the rodent models of TLE are commonly criticized for having characteristics unlike human TLE, such as excessive hippocampal degeneration. In addition, there is a diversity of animal models of TLE, not just one, and they are quite different in seizure phenotype and pathology. Therefore, the fact that mouse models of AD and rodent models of TLE share some characteristics is interesting and may lead to important insights, but may not be particularly helpful to address the question of clinical similarity between AD and TLE.
2.5. Insight into the similarities and differences between epilepsy and AD from studying mechanisms of hyperexcitability and seizures in experimental preparations
Several studies in transgenic mouse models of AD suggest that mechanisms underlying the abnormalities in excitability in both epilepsy and AD may be shared. The mechanistic information suggests that widespread brain damage is not necessarily what AD and TLE have in common and is not what causes seizures — instead, there are specific defects in the molecular mechanisms that regulate excitability that are shared. Seizures may be worse in epilepsy than they are in AD because of slight differences in the molecular defects, even if they are similar defects. For example, a mutation in an ion channel that causes complete loss of function would be likely to cause a more severe disturbance than a mutation in the ion channel that causes partial loss of function.
In AD, it seems likely that a source of increased excitability, particularly in the early stages of the disease, is related to the abnormal metabolism of APP and not to degenerating neurons. In other words, the seizures are due to a “peptidopathy” [42
]. Peptide products of APP influence several aspects of neuronal function and increased excitability results. On the other hand, abnormal metabolism of APP in epilepsy is not considered relevant to the underlying mechanisms of seizures.
The peptides that result from APP cleavage influence excitability in AD but may do so in a complex manner. For example, Aβ increases hippocampal glutamatergic synaptic transmission and long-term potentiation (LTP) at levels that are just above normal, but an excess of the same peptide reduces excitability [43
]. These data suggest that APP metabolites need to be regulated within a tight window or hyperexcitability may result.
In addition to the peptides that result from abnormal metabolism of APP, there are potential abnormalities in excitability that arise from mutations in other proteins. Presenilin-1 is an example. Mutations or deletion of PS1 increases excitability, decrease seizure threshold, and are associated with seizures in patients [45
]. Tau also appears to affect excitability and seizure threshold in rodents; reduction of tau reduces seizures [17
Amyloid β-independent mechanisms of abnormal excitability may be shared in AD and epilepsy. For example, in hAPP mice, expression of the voltage-gated sodium channel subunit Nav
1.1 is reduced in GABAergic interneurons, and it has been suggested that the consequence is a loss of inhibition [46
]. Thus, increased excitability may not be a direct effect of Aβ per se, but an indirect effect of altered Nav
1.1 levels. The hAPP mice with deficits in Nav
1.1 expression are interesting to consider because the hAPP mice are similar to two syndromes — generalized epilepsy with febrile seizures+ (GEFS+) and severe myoclonic epilepsy of infancy (SMEI), two genetic epilepsy syndromes in which Nav
1.1 is mutated. The comparison suggests that at least one molecular target in hAPP mice and mouse models of epilepsy can be the same (Nav
1.1) and located similarly (GABAergic neurons) but the type of deficit is different: decreased levels in hAPP mice and mutation in the epileptic animals.
Modulation of voltage-gated sodium channel subunits may also contribute to hyperexcitability due to the actions of BACE1, the rate-limiting enzyme that cleaves APP to produce Aβ. Another substrate of BACE1 is the β2 subunit of voltage-gated sodium channels [47
], which is an accessory subunit that is responsible for proper membrane localization of pore-forming α subunits. Cleavage of the β2 subunit by BACE1 produces a C-terminal fragment that is subsequently cleaved by γ-secretase to release an intracellular domain (ICD). The β2-ICD translocates to the nucleus and triggers the expression of Nav
1.1, which gets trafficked to the cell surface in part via binding to β2 [47
]. However, overexpression of BACE1 in transgenic mice or cell lines results in excessive cleavage of β2 subunits and surplus expression of Nav
1.1 that is intracellularly retained, leading to reduced surface levels of Nav
1.1, reduced sodium currents, and impairment in action potentials [48
]. The Nav
1.1 channel is highly expressed in the subtype of GABAergic interneuron that exerts a powerful inhibition on principal cells. Thus, a loss of functional Nav
1.1 subunits in these interneurons may contribute to disinhibition of cortical networks [46
]. The β2 subunit cleavage and reduced levels of functional Nav
1.1 subunits might also occur in neuronal populations that control activity of other brain regions, which could also lead to disinhibition. Thus, regulation of Nav
1.1 expression may have an important effect on network activity. Indeed, mouse models in which β subunits of voltage-gated sodium channels are ablated or mutated exhibit a predisposition to seizures [48
Notably, BACE1 levels are increased in the brains of patients with AD and hAPP mouse models of AD [48
], and the magnitude of the increase is correlated with increased β2 subunit cleavage in the brains of patients with AD [48
]. These studies suggest that BACE1-mediated cleavage of the β2 subunit may contribute to AD-related increase in principal cell activity, although this has yet to be determined. Other lines of research suggest that BACE1 has diverse roles in modulating sodium channel function and seizures: for example, ablation of BACE1 can produce opposing changes in sodium channels but still increase the susceptibility to seizures [61
]. Such studies indicate that a balance of BACE1 activity must be maintained to preserve normal neuronal and network activity.
Such comparisons show that AD and epilepsy indeed may have similarities that are so much alike that they are unlikely to be coincidental. Moreover, the few differences that occur may help explain why there are more seizures in epilepsy than in AD. For example, in the discussion above, the amino acid sequence of Nav1.1 is normal in AD but its levels of surface expression are reduced. In the types of epilepsy where Nav1.1 is involved (GEFS+ and SMEI), there is a mutation in the Nav1.1 sequence which is a potentially more severe defect. The fact that there is a normal amino acid sequence in Nav1.1 in AD, but it is mutated in epilepsy, may lead to less severe seizures in AD than in GEFS+ or SMEI.