The basic science of memory as it applies to epilepsy


  • Presented as part of the AES Annual Course: Problems for People with Epilepsy Beyond Seizures, December 2, 2006, San Diego, CA.

Address correspondence to Kimford J. Meador, M.D., Department of Neurology, University of Florida, McKnight Brain Institute (L3–100), 100 South Newell Drive, PO Box 100236, Gainesville, FL 32610, U.S.A. E-mail:


SummaryThe mechanisms of memory delineated by the model of long-term potentiation (LTP) are similar to those underlying epileptogenesis by kindling. Memory is impaired by seizures and epilepsy. High frequency neural activity is important in both memory formation and seizures. Both kindling and LTP are most effectively induced by high-frequency stimuli, involve synaptic facilitation, and share overlapping molecular mechanisms, such as N-methyl-d-aspartate (NMDA) receptor-induced calcium cascade and protein synthesis. The hippocampus contributes to both through its role in memory formation and its low seizure threshold.

Long-Term Potentiation (LTP): A Model of Memory Formation

   LTP is a form of synaptic plasticity, the most widely studied physiological model of mammalian memory formation (Cooke and Bliss, 2006). It results in facilitation of synaptic efficacy lasting hours to months depending on the stimulus parameters and repetition of the inducing stimulation. LTP is most readily produced in the hippocampus by high-frequency stimulation and coincidence activity between pre- and postsynaptic neurons is essential. The induction process is highly input-specific. For example, a single pathway can be potentiated without effect on neighboring inputs to the same cell. In contrast, long-term depression (LTD) is a long-lasting decrease in synaptic efficacy that is induced by trains of low frequency (e.g., 1 Hz) stimulation or by mismatching pre- and postsynaptic potentials. LTD may mediate forgetting, behavioral extinction, or reverse learning.

Molecular Mechanisms of LTP

LTP mechanisms vary across species, brain regions and even synapses within the same neuron (Barco et al., 2006; Cooke & Bliss, 2006). A summary of some common mechanisms follows. LTP requires input-specific pathways and temporal relationship of the stimuli. LTP involves activation of N-methyl-d-aspartate (NMDA) receptors in many neuronal models. Presynaptic release of glutamate will not active the NMDA receptor when the neuron is at resting potential because it is blocked by magnesium. Depolarization of the postsynaptic neuron expels the magnesium from the NMDA receptor and then glutamate may activate the NMDA receptor resulting in an influx of calcium and sodium ions. Calcium then triggers an enzymatic cascade leading to early and late phases of LTP.

Calcium can activate calcium/calmodulin-dependent kinase II (CaMKII), which converts adenosine triphosphate (ATP) to cyclic adenosine monophosphate (cAMP). Then cAMP activates cAMP-dependent protein kinases (PKs), which can phosphorylate alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptors leading to early phase LTP. Unlike late phase LTP, this early phase LTP is not dependent on protein synthesis. After activation by cAMP, protein kinase A (PKA) can increase activity of CREB (cAMP-responsive element binding protein) by a positive effect on a CREB activator (ApCREB1) and by reduction of CREB repression as a result of PKA activation of MAPK (mitogen-activated protein kinase), which deactivates the CREB repressor (ApCREB2). CREB is a transcription factor, which is transported to the cell nucleus leading to synthesis of messenger ribonucleic acid (mRNA) and transcription of proteins. CREB can also activate other transcription factors (e.g., c-fos, EGR-1, C/EPP, protein kinase M zeta). The transcribed and activated proteins are transported back to the synapse leading to structural changes and the long-lasting late phase LTP.

Ltp and Synaptic Structure

Structural synaptic changes have been associated with learning. Dendritic spine efficacy appears to be increased by increased numbers of synapses, shorter spine neck, larger synapse head, perforated synapse head, and bifurcated synapse head. LTP effects on dendritic spine synapses appear to be dynamic and unstable. Variable findings on the number of synapses after LTP have been reported, but LTP can increase synaptic density and the number of shafts (Barco et al., 2006).

Possible Self-Perpetuating Mechanisms Underlying LTP and Memory

Since proteins have a relatively short half life compared to the duration of some memories, it unclear how the synaptic changes are sustained. Several mechanisms have been postulated (Barco et al., 2006). CaMKII has autophosphorylation and self-activation properties. CPEB (cytoplasmic polyadenylation element binding protein) can be converted to an activated state, which has self-perpetuating prion-like properties. AMPAR has a maintainance pathway. LTP and other forms of memory formation result in structural changes in the synapse (e.g., increase size of the synaptic head), which make the synapse resistant to further alterations.

Role of High-Frequency Activity in Normal Cerebral Processes

High frequency neuroelectric activity in the gamma range (30–100 Hz) is a ubiquitous brain phenomenon. In animals, gamma-range coherence has been observed in olfactory bulb and cortex of rabbits during inhalation, in the primary and secondary auditory cortex of rats during auditory stimulation, in the primary and secondary somatosensory cortex of rats during tactile stimuli, in the visual cortical areas of cats during visual stimuli, and in various cortical regions of monkeys during behavioral tasks. Cortical gamma waves are influenced by the thalamus, mesencephalic reticular formation, and cholinergic modulation. Gamma oscillations are associated with neuronal discharges and may result from intrinsic neuronal properties and synaptic summation.

In humans, gamma activity has been recorded from intracranial electrodes (iEEG) and from the scalp using EEG and magnetoencephalogram (MEG) (Kaiser & Lutzenberger, 2005). Gamma activity increases during sensory stimulation, movement, and cognitive tasks in humans. Under different behavioral conditions, gamma activity dissociates from other bandwidths including beta, and bands within gamma may respond differentially. Topographic specificity of induced gamma activity is seen for both sensory and motor tasks, and the topographic distributions of gamma are more discrete and more consistent with traditional functional anatomy than similar maps for other bandwidths. Coherent activity in the gamma range (40–80 Hz) has been postulated to underlie conscious awareness by integrating neural activity across different cerebral areas. However, the relationship of gamma coherence to cognitive tasks in humans requiring conscious perception remains controversial.

Electrophysiology of Memory

LTP is most easily induced by high frequency stimulation (Cooke & Bliss, 2006). Spontaneous gamma activity (i.e., >30 Hz) is phase locked to theta activity (i.e., >4 to <8 Hz) in the human brain. Stimuli in the gamma range occurring in theta bursts are highly effective for LTP induction. High-frequency synchronous activity in the pyramidal cells has been postulated to influence synaptic connectivity in the hippocampus. Thus, the interrelation of attention and memory may be due to the dual role that gamma coherence plays in perception and synaptic linkages. In humans, coherent gamma EEG activity is present during associative-learning (Miltner et al., 1999), and increases in gamma and theta power occurs during encoding predicts subsequent recall (Sederberg et al., 2003). Memory formation in humans has been associated with elevated rhinal-hippocampal gamma synchronization (Axmacher et al., 2006) although the coupling may be broadband and not specific to memory.

Role of High-Frequency Activity in Seizures

Multiple investigators have noted a relationship of high frequency EEG activity to seizures (Rampp & Stefan, 2006). Fast EEG rhythms have been seen at the onset of epileptic discharges in humans and animal models of epilepsy. Intracranial recordings in patients have found high-frequency ictal activity and emphasized the localizing value of this activity at seizure onset in prediction of epileptic focus and good postoperative outcome following resection of this region.

Frequencies faster than gamma have also been emphasized in regards to seizures and brain function. High-frequency oscillations within the range 100–200 Hz have been recorded in the hippocampus and entorhinal cortex of animals and humans and termed “ripples.” The functional role of ripples is uncertain, but it has been suggested that they are involved in memory consolidation processes. Higher-frequency oscillations called “fast ripples” (250–500 Hz) have been described in seizure-generating limbic areas of rodents after intrahippocampal injection of kainic acid where their strong association with regions of seizure initiation has been demonstrated (Staba et al., 2002).

Gamma range EEG activity and synchrony increase progressively until the onset of seizures at which point the increase is abolished (Medvedev, 2002). These findings have been interpreted as suggesting that an increased and highly synchronous gamma activity is one of the major pathophysiological determinants of the preseizure state and that seizures may temporarily suppress a pathological increased gamma coherence activity, thus “resetting” the functional state of the brain.

Seizures, LTP and Memory

Neural circuits undergo long-term progressive structural and functional alterations in response to seizures. Kindling involves repeated administration of brief low-intensity trains of electrical stimuli that result in a permanent state of increased susceptibility and even spontaneous seizures. Kindling was first described from studies originally designed to examine the effects of electrical stimulation of the amygdale on learning. The stimulus parameters which best induce kindling are similar to LTP (e.g., high frequencies) and both produce synaptic facilitation, depend on protein synthesis, and result in structural synaptic changes, suggesting that a continuum of neural modification, some such as LTP leading to normal functions like learning and others pathological in nature, such as in kindling and excitotoxicitic effects (McEachern & Shaw, 1999).

Repeated electroconvulsive seizures (ECS) reduce the ability to induce LTP and impair spatial learning in animals (Reid & Stewart, 1997). Resolution of the learning deficits is similar in duration to the memory deficits seen in humans receiving ECS. Further, the effects of ECS on LTP can be blocked by the NMDA antagonist ketamine. This suggests that the seizures “saturate” the synapses with long-term facilitation that decreases the capacity for plasticity including LTP and memory. Kindling also suppresses LTP (Leung & Wu, 2003), and repeated seizure like events can erase LTP (Hu et al., 2005). Neonatal seizures in animals can induce long term loss of LTP, impair spatial learning, and alter NMDA protein expression (Bo et al., 2004). LTP is markedly reduced in the hippocampus of the epileptic focus in humans with temporal lobe epilepsy (Beck et al., 2000). However, LTP and LTD are present in human hippocampus if not the primary seizure focus. This LTP is NMDA dependent and sensitive to the adeylyl cyclase agonist forskolin and phosphodieseterase inhibitor IBMX, which are know to affect LTP in animals.


One important contributing factor to the memory complaints of people with epilepsy may be the overlap in the mechanisms underlying memory and epilepsy. The hippocampus is critical for formation of episodic memory, and the lowest seizure thresholds in the brain are in the hippocampus. Seizures and epilepsy impair memory. Both seizures and cognitive processes including memory formation involve high-frequency neural activity. Kindled seizures and LTP have overlapping molecular mechanisms and both are induced by gamma or theta-burst gamma activity. These similarities are unlikely to be coincidental.

Disclosure of Conflicts of Interest

The contributing author to this article has declared the following conflicts of interest: Dr. Meador has received research funding from BlaxoSmithKline, Marius, McKnight Brain Institute, Myriad, Neuropace, SAM Technology, and UCB Pharma.