Pharmacogenetics and Pharmacogenomics

Chapter 141

Nicole Walley

Sanjay M. Sisodiya

David B. Goldstein

Introduction

Although pharmacologic treatment successfully controls seizures for a majority of patients with epilepsy, current treatment options are far from satisfactory for many patients. Recent progress emerging from the genome project and related activities suggests that it may finally be possible to systematically search for genetic differences among patients that influence how they respond to treatment. Here we review the potential of pharmacogenetics to improve the treatment of epilepsy and address some of the barriers to progress.

Limitations of Current Epilepsy Treatment

Perhaps the most important weakness of current treatment options is that they fail to control seizures for a substantial minority of patients. Nearly one-third of patients do not achieve adequate seizure control upon treatment with any of the currently available antiepileptic drugs (AEDs).⁶³ Furthermore, those who fail even one drug trial have a poor prognosis: Only 11% of patients who withdraw from their first AED due to inefficacy will ever go on to achieve seizure freedom.¹⁷^,⁴⁰^,⁴¹ The end result is a significant number of patients who are forced to (a) live with recurrent seizures or (b) resort to invasive surgery, which is not available to all refractory patients, is not guaranteed to eradicate seizures, and can have devastating consequences in some cases.

Although nine new AEDs have been approved during the last 15 years, these have made only a modest impact on the proportion of patients who respond well to treatment. There are currently more than 15 pharmacologic agents now available for clinical use, but approximately 30% of epilepsy patients remain refractory to treatment.⁵^,⁶³ The chances of achieving seizure freedom with new AEDs has not been significantly improved, although levetiracetam (Keppra) is reported to reduce seizures in 40% of refractory patients, with 6% to 13% achieving seizure freedom.⁸^,³⁹^,⁶⁸

The other key limitation is that AEDs can cause both serious and life-threatening adverse reactions, as well as less serious reactions that may nonetheless have important effects on patient quality of life. It has been estimated that 17% of emergency room visits due to adverse drug reactions (ADRs),⁵⁸ and 10% of ADRs leading to hospitalization⁸⁷ are caused by AEDs. ADRs were cited as the cause of discontinuation in nearly 40% of decisions to terminate treatment with a particular medication⁴⁰ and, although patients exposed to high doses of AEDs are more likely to develop ADRs, side effects are not necessarily dose related and also can occur during the course of efficacious drug trials.

The newer AEDs, although offering little improvement for refractory cases, do tend to be tolerated better by patients. Withdrawal rates due to adverse reactions are significantly lower for the newer AEDs.⁴¹ However, older ADRs have not been eradicated and some new ADRs have been introduced, such as the behavioral and cognitive ADRs associated with levetiracetam and topiramate respectively. Tolerability remains a major limitation in epilepsy treatment.

In addition to these key aspects, there are secondary challenges in the use of AEDs, notably the trial-and-error process usually required to identify appropriate doses or drug combinations for individuals. Patients require dramatically different doses to control seizures (Fig. 1), however efficacious dose is impossible to predict. Current routine, nonemergency methods of dosing stipulate, “start low, go slow.” A patient is slowly titrated upward until he becomes seizure-free or he experiences dose-related ADRs; at this point, use of a new drug is often considered. This can take months for those who require uncharacteristically high doses of a medication, or years for those who fail one or more trials and must repeat the process—during which time patients continue to suffer seizures.

Pharmacogenetics as a Probe into Biological Processes

In our view, the improvement of seizure control and the minimization of side effects are the primary motivations for pharmacogenetics studies in epilepsy, and projects should be prioritized on this basis. It is also worth noting, however, that pharmacogenetics may open a window into biologic process that are poorly understood in humans and otherwise not amenable to study.

Levetiracetam and Human Behavior

The approval of levetiracetam has brought seizure relief to some previously refractory patients. It is generally well tolerated, but has been known to cause behavioral abnormalities. Behavioral ADRs have been cited as the most common reason for withdrawal from levetiracetam treatment, and between 5.9% and 30.7% of patients report experiencing behavioral changes.³⁵^,⁸⁶ Levetiracetam-induced behavioral abnormalities can manifest as irritability, aggression, depression, or psychosis. The identification of gene variants relevant to these ADRs therefore could identify pathways relevant to human behavior, and perhaps more specifically relevant to neuropsychiatric conditions such as schizophreniform psychosis.

FIGURE 1. Distribution of maintenance dose of lamotrigine. Maintenance dose among patients from the same clinic is highly variable. All patients were treated with Lamo-trigine at the National Hospital for Neurology and Neurosurgery, Queen Square, London. Additional (maximum) dose distributions can be found in Tate et al.⁷⁶

Topiramate and Human Cognition

Topiramate is a new AED commonly used as adjunctive and monotherapy. The most common reason for discontinuation of topiramate therapy is cognitive side effects, which can manifest as mental slowing, language difficulties, and confusion.

P.1498

Nearly one-half of patients exposed to topiramate report cognitive ADRs (unpublished data), and 27% of patients who discontinue topiramate treatment cite cognitive adverse events as the cause of cessation.⁷⁷ Although topiramate is most commonly used as add-on therapy, and the incidence of ADRs is increased with polytherapy, healthy volunteers challenged with topiramate alone also experience cognitive difficulties, as determined by neuropsychological tests.⁴⁷^,⁴⁹ Although many AEDs have negative cognitive side effects, topiramate’s profile is distinctive. For example, verbal fluency is more affected by topiramate than by other AEDs²⁴ and appears to affect only a subset of susceptible patients.⁵¹ If genetic differences that mediate to-piramate sensitivity exist among people, their identification may inform clinicians about the underlying brain systems involved in verbal fluency tasks.

Epilepsy Pharmacogenetics: What is Needed?

Genetic Methods

It is possible (although still not certain) that systematic pharmacogenetic research can identify gene variants that will considerably improve the use of AEDs. Here we describe the key elements of contemporary pharmacogenetics research.

Candidate Gene Approach

Until recently, it was not feasible to consider in a single study more than a handful of genes; even then, the genetic information considered for each gene was very limited. Very recently, however, it became feasible to carry out large-scale candidate gene studies in which the genetic variation in hundreds of genes is systematically analyzed. Moreover, whereas whole-genome association studies remain costly, reasonable coverage of most polymorphisms in the full human genome is available in standardized sets of polymorphisms, such as those supplied by Illumina.

In pharmacogenetics, a particularly strong case can be made for candidate gene approaches because the likely modes drug action are usually known or suspected, at least partially. Therefore drug targets make obvious candidate genes and clearly deserve careful evaluation. Consistent with this notion, recent work on the targets of warfarin and phenytoin/carbamazepine identified polymorphisms associated with dosing.¹⁴^,⁷⁶ Moreover, nearly 80% of genetic variants identified by pharmacogenetics reside in the three major categories for pharmacogenetic candidates: Drug targets, drug metabolizing enzymes, and drug transporters (although this reflects in part the bias of where researchers have chosen to look; for a review of this subject see the work by Goldstein, Tate, and Sisodiya²⁸ and their unpublished update.

For epilepsy pharmacogenetics, a set of high-priority candidate genes are readily identified on the basis of both pharmacokinetics and major modes of action.

Pharmacokinetic candidates

Drug-metabolizing enzymes (DMEs). The most studied genes to date in pharmacogenetics are those encoding DMEs. For most AEDs, the key enzymes that metabolize the parent compound and active (or toxic) metabolites are well known (Table 1), and many of them have well described functional variants. It is already clear that these variants have some impact on patient response to AEDs: For example, CYP2C9 variation affects phenytoin dosing (see discussion in next section). It does not appear likely, however, that DME variation will prove of substantial clinical relevance in epilepsy.
Transporters. Drug transporters may influence AEDs at a variety of points. They can influence uptake in the gut, the amount of metabolism in the liver and, perhaps most important, the properties of the blood–brain barrier. In the latter case, the effect of transporters may either be general or affected by seizures. As a class of proteins, however, drug transporters are not well studied. It is not clearly known what transporters move which drugs at therapeutically relevant concentrations, and few examples of functional variation have been characterized, making it unclear what role transporters play in pharmacogenetics.
However, drug transporters are known to be overexpressed in resection tissue from refractory epilepsy patients.⁷^,¹⁸^,⁷²^,⁷⁸ The apparent upregulation of transporter expression at the seizure focus indicates a potential causative role for transporters in refractory epilepsy.

Table 1 Drug-Metabolizing Enzymes of Commonly Used AEDs

AED	Major DMEs	Minor DMEs
Carbamazepine	CYP3A4 epoxide hydrolase	CYP2C8 GST
Ethosuximide	CYP3A4	CYP2B
Ethosuximide	CYP3A5	CYP2C isoenzymes CYP2E
100% Renal clearance
Gabapentin
Renal clearance
Levetiracetam
Lamotrigine	UGT1A4
Renal clearance
Pregabalin
Phenytoin	CYP2C9	CYP2C19
Topiramate	Renal clearance	CYPs (specific enzymes unknown)
Valproate	B-oxidization	CYP2C9 CYP2A6

Table 2 Known and Suspected Mechanisms of Action of Commonly Used AEDs

Drug	Major drug target	Other suspected actions
CBZ	Sodium (Na) channel α-subunit	Inhibition of voltage-gated calcium (Ca²⁺) channels and potentiation of potassium (K⁺) channels Antagonism of adenosine receptors Inhibition of glutamate release Increase of extracellular serotonin and dopamine transmission Decrease of basal and stimulated cAMP levels³
ETX	T-type calcium channels	Inhibition of Na⁺/K⁺ ATPase²⁷
GBP	Ca channel α2δ-subunits 1 and 2	Inhibition of GABA transaminase⁴³ and GABA reuptake²¹ Decreased free content of glutamine/glutamate and reduced glutamate release¹⁹^,⁵⁰
LEV	Synaptic vesicle protein 2A	N-type Ca²⁺ channels⁴⁶ Inhibition of intracellular Ca²⁺ release⁴
LTG	Na channel α-subunit	Suppressed presynaptic Ca²⁺ influx¹⁰ Vesicular release, independent of Na⁺ and Ca²⁺ currents¹³ Enhanced/inhibited hyperpolarizing K current³² Downregulation of cortical 5-HT1A receptors⁸²
PGB	Ca channel α2δ-subunit
PHY	Na channel α-subunit	Inhibition of Ca²⁺ channels⁶⁶^,⁸⁰ Inhibition of Ca²⁺ calmodulin-mediated protein phosphorylations¹⁵^,¹⁶
TPM	Voltage-gated Na channels L-type Ca channels AMPA/Kainate receptors GABA_A receptor Carbonic anhydrase
VPA	Glutamic acid decarboxylase GABA transaminase Succinic semialdehyde dehydrogenase Voltage-gated Na channels
CBZ, carbamazepine; ETX, ethosuximide; GBP, gabapentin; LEV, levetiracetam; LTG, lamotrigine; PGB, pregabalin; PHY, phenytoin; TPM, topiramate; VPA, valproate.

FIGURE 2. Chloride transporter expression and GABAergic neurotransmission. A: In fetal neurons, NKCC1 is the predominantly expressed chloride transporter, resulting in a high intracellular concentration of Cl^- ions and depolarization upon GABAergic opening of chloride channels. In fetal tissue, GABA is excitatory. B: In mature neurons, KCC2 is the predominant chloride transporter, resulting in a high extracellular concentration of Cl^- and hyperpolarization upon GABAergic opening of chloride channels. In adult tissue, GABA is inhibitory.

Pharmacodynamic candidates

Drug targets. The major modes of action for most AEDs are known, although in some cases the most important among several candidate modes of action remains unclear (Table 2). The major modes of action fall into one of three broad categories: Modulation of voltage-gated ion channel function, enhancement of γ-aminobutyric acid (GABA)-mediated inhibition, and attenuation of excitatory (glutamate-mediated) transmission.⁴² In addition to
P.1499

P.1500

the precise target itself, it should also be appreciated that molecules downstream of the drug targets upon which an AED acts may contain variation that influences response. Thus, pharmacogenetic studies of GABA-acting drugs should appropriately consider chloride ion homeostasis broadly and consider relevant genes. The importance of this perspective is clearly illustrated by chloride ion transporters.
Chloride ion homeostasis is primarily regulated by a balance of oppositely acting chloride transporters: NKCC1 (also known as SLC12A2), which transports chloride ions into the neuron and KCC2 (also known as SCL12A5), which extrudes chloride ions from the neuron. In adult neurons, KCC2 is the dominantly expressed transporter, resulting in a high extracellular concentration of chloride ions. In fetal neurons, the balance is reversed, such that NKCC1 is dominantly expressed, and there is a high intracellular chloride ion concentration.⁵⁴^,⁶¹ The change in expression results in GABA having an inhibitory effect in mature neurons and an excitatory effect in fetal neurons (Fig. 2). This is known to have pharmacologic implications, because GABA-acting drugs can exacerbate seizure activity in infantile seizures.²⁵ In addition, expression levels of these chloride channels are altered in seizure-affected brain tissue.⁵²^,⁶⁷ From a pharmacogenetic standpoint, genetic alteration of chloride transporter protein expression or function most certainly has the potential to affect patient response to GABA-acting drugs.

Direct and Indirect Genetic Methods

Association studies to identify gene variants that influence a particular phenotype can be divided into direct and indirect approaches. In the direct approach, all variants that are good candidates for being functional are identified and checked for association with response. In the indirect approach, a set of genetic markers is identified that is sufficient to represent common variation in the region of interest through haplotype tagging, relying on linkage disequilibrium (LD) among polymorphisms.

To date, pharmacogenetics has followed essentially a direct approach. Many of the important DMEs have been resequenced, and all variants that looked to be potentially functional have been studied. In this way, it is possible, for example, to carry out an association study for phenytoin simply by relating the two low-activity variants of its major metabolizing enzyme, CYP2C9, and patient response.

Beyond DMEs, however, knowledge of functional variation in genes is very limited. One could resequence exons, but this could still miss important regulatory variation. These reasons argue for a tagging-based approach.

Tagging, however, will never perfectly capture all genetic variation and, in particular, it is known that variants with low minor allele frequency may not be well represented.¹^,⁸⁸ For this reason, we strongly favor a hybrid approach that combines elements of direct and indirect association. This model utilizes bioinformatic criteria to identify variants that are more likely to cause functional consequences than a randomly chosen polymorphism (putatively functional single-nucleotide polymorphisms [SNPs]). In our own work we have used the following criteria:

SNPs that cause amino acid changes are predicted to cause functional differences (i.e., mRNA stability) or lie within splice junctions³⁶
SNPs in promoter regions⁷⁹
SNPs that occur in regions that are known to be highly conserved across species⁹^,⁷⁰

Today tagging SNPs (tSNPs) are usually selected on the basis of the genotype data provided by the HapMap project. This project has genotyped over 1 million polymorphisms in 269 DNA samples from four ethnic backgrounds (central

P.1501

European, Han Chinese, Japanese, and Yoruban Africans).² Critically, the project preferentially typed variants that are more likely to be functional. For this reason, the majority of SNPs selected on the basis of the above criteria have been typed in HapMap. It is therefore considerably more efficient to take into account the possible functional variants that are to be typed when selecting tSNPs. As an example, we applied this hybrid approach to 450 candidate genes for response to AEDs (Fig. 3). We found that, on average, 10 SNPs are required to tag a gene. It is estimated that humans have 10 million polymorphisms,¹² and there are currently over 9 million SNPs dbSNP (with some redundancy).⁴⁸ Thus, it is possible to screen most polymorphisms within a set of candidate genes.

FIGURE 3. Scheme of tag selection using hybrid tagging approach. Using the genomic sequence of all candidate genes, putatively functional variants are identified according to (a) potential coding/splicing function, (b) location in promoter region, and (c) location in regions of high evolutionary conservation. In addition, the genes are tagged using available HapMap data and force, including as tags HapMap SNPs that appear in the functional set. A final set of tags is produced that is predicted to represent all common variants up to a specified threshold.

High Throughput Genotyping

The comprehensive approach just described results in an average of about 10 tSNPs per gene. This means that approximately 5,000 SNPs for 500 genes would be reasonable to cover the most attractive candidate genes in epilepsy pharmacogenetics, a number that was previously cost-prohibitive. The genotyping costs for a high-throughput platform such as Illumina are now about US$0.03 to US$0.04 per sample per genotype, meaning that sample sizes in the thousands can now be analyzed in the context of project grant support.

Analyses

Several issues must be addressed when undertaking large-scale case-control studies. Of particular importance is population stratification, which has been known to cause false-positive associations, especially when allele frequencies differ across populations. Study groups, therefore, must be assessed for ethnicity using ancestry-informative markers. Results and replication attempts can also be confounded by genetic heterogeneity, where a different gene/allele contributes to disease based on population/ethnic differences. Most important, though, is power. To maintain power, especially when testing many genes/polymorphisms for the same phenotype, large numbers of cases and ethnically matched controls must be used.

In addition to these classical pitfalls of genetic association studies, there are novel challenges that currently have no solution. Particular to pharmacogenetics is the unmonitored exploration of complex phenotypic space. There are many ways to consider response phenotypes and, as discussed earlier, no consensus exists on how it should be done.

Also, phenotype depends on a number of factors, none of which relies solely on one gene or allele. Common disease and response phenotypes occur as a result of interactions between genes and environment, and genes and other genes. Analyses can match cases and controls to try to control for as much of the phenotypic variation as possible, however gene–gene interactions cannot be controlled for and neither does there exist a way to identify them.

Clinical Cohorts

In both disease genetics studies and treatment response studies, many of the key challenges are not on the genetic side, but rather on the clinical side. It is very difficult and time-consuming to build well-phenotyped clinical cohorts, and it is often not clear how phenotypes should be defined. These challenges are acute in epilepsy, where clinical diagnosis is inexact, and response phenotypes are often arbitrary.

Retrospective and Prospective Cohorts

Retrospective cohorts are valuable resources because they are immediately accessible; however, they also come with several limitations: Phenotypic data from retrospective cohorts are collected from a clinical perspective and can overlook certain points of interest for genetic studies (for example, serum levels are not always recorded in routine clinical practice but are important when studying dose from a pharmacogenetic standpoint). Additionally, retrieving response information from patient records can be difficult and time-consuming, and the phenotypes necessary for pharmacogenetic study must be tailored to the available clinical data, which can compromise its validity and accuracy.

Prospective cohorts can solve many of these problems. Phenotypic data are ascertained by the clinician with the specific aims of the genetic study in mind to ensure that the appropriate phenotypes are specific and consistent. Patient recruitment can also be tailored to the aims of the study. However, prospective cohorts take time to build, and doing so is a costly endeavor. For these reasons, few pharmacogenetic studies in epilepsy utilize prospective cohorts, and therefore these studies reflect the limitations that retrospective cohorts necessarily entail.

Drug Response Phenotypes

Efficacy phenotypes.

“Refractory,” as it is currently used is a qualitative clinical term applied to patients who have failed a specified number of syndrome-appropriate drug trials. However, this definition implies nothing about the biologic cause of drug failure, and does not necessarily specify the trialed drugs. However, patients who fail only sodium (Na) channel–acting drugs are not biologically equivalent to patients who fail all drugs regardless of drug mechanism. Furthermore, it can only be said that a patient is resistant to Drug A if Drug A has failed. Therefore, simply classifying a patient as “refractory” is an inadequate phenotype. Instead, refractory patients should be subgrouped according to the drug(s) or drug class(es) to which they are resistant.

“Responder” likewise is not an objectively measured phenotype. The responder phenotype can be confused with spontaneous seizure remission, and patients with infrequent seizures prior to drug treatment are frequently too hastily qualified as responders. Additionally, a patient who achieves seizure freedom after failing the first two AEDs is not necessarily biologically equivalent to a patient who responds to the first drug trial.

P.1502

Many factors must be considered. For example, a patient who becomes seizure-free from a starting point of one seizure per year clearly has a different response phenotype from a patient who becomes seizure-free from a starting point of hundreds of seizures per year. In epilepsy pharmacogenetics, currently no guidelines exist that specify how to take into account seizure frequency before and after initiation of treatment, and over what time frame, to arrive at a biologically (and clinically) meaningful definition of response.

Adverse events phenotypes.

Serious adverse events are probably of most interest for pharmacogenetic study. However, these are rare and are unlikely to be captured in sufficient numbers for study by assembled cohorts. Additionally, many mild adverse events can go unreported. For example, many patients, while exposed to topiramate, never complain of any cognitive side effects. It has been observed, however, that upon withdrawal of the drug, patients experience measurable cognitive improvement, indicating that these adverse events are easily overlooked and often underestimated.³⁸

Also, adverse events can be difficult to quantify. The behavioral effects of levetiracetam, for example, can be reported as irritability or feelings of aggression, effects that cannot be measured, thus making phenotype classification heavily reliant upon the description of the patient.

Dosing phenotypes.

During the course of treatment, patients are exposed to many doses during the titration period, and dose remains subject to change throughout the course of treatment. There is currently no way to integrate a patient’s dosing history into a single maintenance dose. Moreover, at a biologic level, clearly a number of different definable phenotypes relate to the dosing decisions made by a clinician. For example, the maintenance dose on polytherapy may be influenced by drug–drug interactions or the induction of relevant DMEs by concomitant medications in addition to any pharmacodynamic variants, whereas under monotherapy, variants relevant to drug interactions and enzyme induction may be less important.

Similarly, increases in dose may be constrained in some patients by tolerability, whereas other patients do not increase dose due to the achievement of effective seizure control.

These points alone make clear that dose, however defined, is likely to be a “heterogeneous trait,” to use the analogous terminology from the study of common disease. This sort of heterogeneity has made dose a surprisingly controversial phenotype in the epilepsy community, especially given the clear example from other therapeutic areas (e.g., warfarin).

We should therefore make clear that, from the perspective of both complex trait genetics and pharmacogenetics, dosing is quite a standard choice as a phenotype. Many things will influence it, and there are various subphenotypes masquerading underneath any specific “dosing” phenotype (e.g., maximum dose, maintenance dose defined one way or another, etc.). But two clear and related motivations exist for investigating dose as a phenotype in pharmacogenetic studies. First, although genetics will never explain all the variation among patients in what dose clinician ultimately settle on, in may well explain some of that variation. It is therefore possible that gene variants that predict dose requirements will eventually have clinical utility. How do you find such variants? You use dose as a phenotype, of course, in a complex trait study of the genetic determinants of dose. Second, dose as a phenotype may be an effective phenotype for identifying gene variants that influence how patients respond to AEDs, which may be variants that are well worth knowing about, regardless of whether they offer clinically useful predictions. The SCN1A intronic polymorphism was identified in exactly this way⁷⁶ and, if it turns out to be a real functional polymorphism that affects splicing, as appears very likely, it illustrates the value of dose as a phenotype.

Pharmacogenetic Studies to Date

Past pharmacogenetic associations in epilepsy have been limited by the technical issues discussed earlier and therefore have focused mainly on pharmacokinetics and have been limited to one or a few candidate genes.

CYP2C9 and Phenytoin Dosing

CYP2C9 is the major metabolizing enzyme for phenytoin and has two common variants, CYP2C9*2 and CYP2C9*3, with reduced enzymatic activity⁵⁹^,⁷⁴ making it an obvious candidate for dosing studies. The *3 allele was first associated to the phenytoin poor-metabolizer phenotype in a study of 12 individuals,³⁷ and several independent studies have since confirmed that CYP2C9 allelic variants are associated with a maximum tolerated dose of phenytoin.⁴⁵^,⁷⁶^,⁸¹ An additional study found that slow-metabolizing alleles are a risk factor for phenytoin-induced cutaneous ADRs (cADRs; rash).⁴⁴ These findings are among the first in epilepsy pharmacogenetics; however, they fail to account for the majority of dose variation in patients, thus suggesting that an additional, more important factor is responsible for dose. Indeed, more recent studies have found a pharmacodynamic factor that is more important than CYP2C9 in determining phenytoin dose.⁷⁶

ABCB1 and Drug Transport

ABCB1 (also known as multidrug resistance protein 1 [MDR1], P-glycoprotein) has been shown to have a weak affinity for AEDs.⁵⁶^,⁵⁷^,⁶⁴ It shows overexpression in resected human epileptiform brain tissue,¹⁸^,⁷²^,⁷⁸ and has characterized functional variation: A synonymous substitution in exon 26 that causes decreased transporter expression²⁹^,³⁰ due to its effects on mRNA stability.⁸⁴ This suggests that it may play an important role in drug-resistant epilepsy due to increased reduced concentrations of AEDs at their target site, the brain. Pharmacogenetic studies have shown this variant⁶⁹ and haplotypes containing this variant³³^,⁸⁹ to be associated with drug-resistant epilepsy. It should be emphasized, though, that the only true attempt to replicate the initial study found no association.⁷⁵

Recently, the transport of AEDs at clinically relevant concentrations by ABCB1 has come into question,⁷^,⁸⁵ suggesting that these associations are false positives or that the exon 26 variant is in high linkage disequilibrium with an as yet unidentified causal variant.⁷³

Although these pharmacokinetic pharmacogenetic associations have proved (a) to be without clinical application or (b) to be potentially false, more recent work in pharmacogenetics have identified pharmacodynamic variants that appear to be replicable and to have great potential for future clinical applicability.

FIGURE 4. SCN1A splicing. The proportion of the two splice variants is genotype-dependent. The major allele (A) disrupts a splicing sequence for exon 5N, whereas the GG genotype restores the site and allows for efficient splicing of the exon. Patients with the GG genotype have higher proportions of the 5N exon in resected brain tissue. This genotype also associates with lower maximum doses of phenytoin and carbamazepine, perhaps attributable to higher proportions of a more drug-sensitive exon being expressed in Na channels. Differential sensitivity of the 5A and 5N exons has yet to be assessed.

HLA and SJS/TEN in Carbamazepine-exposed Patients

Many different classes of drugs are known to cause severe cutaneous adverse reactions manifesting in the form of Stevens-Johnson syndrome (SJS) or toxic epidermal necrolysis (TEN). These conditions are rare but are acutely life threatening, with a mortality rate as high as 40%⁶²; they have been recorded as idiosyncratic reactions upon exposure to

P.1503

the aromatic AEDs, mainly carbamazepine, lamotrigine, and phenytoin.

cADRs have been hypothesized to result from the inability to metabolize drug compounds or certain active metabolites, however no genetic defects altering the structure of AED DMEs have been associated with cADRs. Instead, immune reactions are likely to be responsible for severe cADRs, and evidence suggests that major histocompatibility complex (MHC)-dependent presentation of a drug or its metabolites activates T cells and causes a severe immune response and cell death.⁵³^,⁶⁵ Furthermore, carbamazepine and/or its metabolites may be capable of binding peptides that are presented to the MHC and are recognized by T cells, which elicits the severe reaction,³¹^,⁵⁵ thus suggesting that HLA is a strong candidate region for these severe cADRs.

It was recently reported that the *1502 allele of the MHC B gene (HLA-B) is strongly associated with carbamazepine-induced SJS/TEN.¹¹ In this small study, this allele was present in 100% of SJS cases, but only 3% of carbamazepine-exposed controls.

An expansion of this study to include larger numbers of patients confirmed these findings. Of 60 SJS/TEN cases, 59 patients had the *1502 allele. The remaining patient had an additional HLA-B minor allele, HLA-B*1558, present at low frequency in the study population and absent from carbamazepine-exposed tolerant controls.³⁴

The strong association of the HLA-B*1502 variant suggests a potential for the clinical applicability of the SJS/TEN marker, although replication and prospective evaluation must precede large-scale clinical application.³⁴

SCN1A and Dosing

As previously mentioned, different patients require highly variable doses of the drugs phenytoin and carbamazepine to control seizures. Tate et al.⁷⁶ used a candidate gene approach to identify gene variants that might affect the dose of phenytoin and carbamazepine, namely CYP2C9, ABCB1, and SCN1A. SCN1A is the major target of both phenytoin and carbamazepine,²⁰^,⁴²^,⁷¹ and also harbors rare mutations that are responsible for some Mendelian forms of epilepsy.²²^,²³^,⁶⁰^,⁸³ CYP2C9 and ABCB1 both have known functional variation (see earlier sections), but no functional variants were known to exist in the SCN1A gene, so a haplotype tagging strategy was employed to represent common variation. A significant association was found with one of the tagging SNPs and the maximum dose of both drugs.

The Function of “Tag 7”

Upon association of genotype at Tag 7 with a maximum dose of phenytoin and carbamazepine, the SNP was assessed for functional implications that might explain how it affects dose. This SNP, SCN1A IVS5–91 G>A (rs3812718), lies in a donor splice site of an alternatively spliced exon. This alternative exon, 5N, is the predominant form of exon 5 in fetal Na channels, whereas 5A predominates in the adult form of Na channels. The major allele disrupts the splice sequence for 5N, whereas the minor allele (G) restores an intact splicing sequence (Fig. 4). The hypothesis, then, is that the polymorphism affects dosing requirements by changing the relative amounts of 5A and 5N forms of the α-subunit, and these forms are differentially sensitive to AEDs, although this is unconfirmed.

5N and 5A in the Epileptic Brain

To further investigate this hypothesis, Tate et al.⁷⁶ compared the relative proportions of 5A and 5N in resected tissue from patients who had undergone surgery for refractory temporal lobe epilepsy. Results showed that, among epilepsy patients, those with the GG genotype have a significantly higher proportion of 5N in temporal lobe tissue than do the wild-type (AA) and heterozygous (AG) genotypes. Surprisingly, there were no significant differences between genotype and proportion of 5N in the seizure focus: The hippocampus. This is possibly due to a high level of neuronal loss in the hippocampus; also, the lack of correlation between the G >A polymorphism and 5N proportions may reflect the relative paucity of neurons compared with other cell types. Future investigations will necessarily assess the proportions of 5N in individually identified neurons from the seizure focus.

In addition, comparisons of patient resection tissue with control brain tissue from a PD brain bank revealed that, regardless of genotype, the proportion of the 5N form of the α-subunit is significantly higher in patients than in controls. This is true for multiple Na channel subunits, including those encoded by the genes SCN1A and SCN8A (Fig. 5). The upregulation of the fetal form of the α-subunit in multiple Na channel genes is likewise seen in animal models following seizure induction,⁶^,²⁶ indicating that 5N upregulation is the result, rather than the cause, of seizure activity.

FIGURE 5. Upregulation of 5N in multiple Na channel genes. The proportion of 5N transcripts is significantly upregulated in two of four Na channel genes in resected seizure tissue when compared with postmortem control tissue.

P.1504

Implications of 5N in Epilepsy

Prior to this pharmacogenetic work in SCN1A, the upregulation of the fetal form of the α-subunit had never been observed in human epileptic brain tissue. Although the exact cause and effect of this phenomenon remain unknown, the observation that patients with the GG genotype (and consequently higher proportions of 5N) require lower doses of AEDs, and that 5N-containing forms of the α-subunit are upregulated in these patients’ brains suggests that this form of the exon may be somehow protective. The molecular characterization of a cause and effect relationship between seizures and 5N upregulation could will provide insight into the pathophysiology of epilepsy and perhaps suggest new molecular targets for the development of new AEDs.

Summary and Conclusions

The potential of epilepsy pharmacogenetics may be well illustrated by the examples of carbamazepine, phenytoin, and warfarin dosing. In these studies, one of the most obvious possible places to look for gene variants that influence response, the target of the gene, was very carefully studied. In both cases, intronic polymorphisms were discovered that associate with dosing decisions. If these are true associations, they suggest the possibility that pharmacodynamic variants that influence response to treatment might be common.

As studies that are sufficiently large to identify multiple effects are finally getting under way, the next several years will prove hugely informative. In our view, it is likely that, for certain genetic diagnostics, it may be necessary to consider multiple polymorphisms. Thus, although neither the SCN1A or CYP2C9 variants appear to explain enough variation in dosing decisions to be clinically important (although the role of SCN1A remains to be carefully studied in control settings), it may be the case that a number of such polymorphisms in combination will provide better dosing predictions.

It seems to us unlikely that treatment will be impacted over that time frame, but we should at least have an idea of how important genetic variation is in patient response. If many new variants are identified, it will warrant the establishment of translational research facilities that are sufficient to work out how to apply the genetic findings to improve care.

References

1. Ahmadi KR, Weale ME, Xue ZY, et al. A single-nucleotide polymorphism tagging set for human drug metabolism and transport. Nat Genet. 2005;37(1):84–89.

2. Altshuler D, Brooks LD, Chakravarti A, et al. A haplotype map of the human genome. Nature. 2005;437(7063):1299–1320.

3. Ambrosio AF, Soares-Da-Silva P, Carvalho CM, et al. Mechanisms of action of carbamazepine and its derivatives, oxcarbazepine, BIA 2-093, and BIA 2-024. Neurochem Res. 2002;27(1–2):121–130.

4. Angehagen M, Margineanu DG, Ben-Menachem E, et al. Levetiracetam reduces caffeine-induced Ca²⁺ transients and epileptiform potentials in hippocampal neurons. Neuroreport. 2003;14(3):471–475.

5. Annegers JF, Hauser WA, Elveback LR. Remission of seizures and relapse in patients with epilepsy. Epilepsia. 1979;20(6):729–737.

6. Aronica E, Yankaya B, Troost D, et al. Induction of neonatal sodium channel II and III alpha-isoform mRNAs in neurons and microglia after status epilepticus in the rat hippocampus. Eur J Neurosci. 2001;13(6):1261–1266.

7. Awasthi S, Hallene KL, Fazio V, et al. RLIP76, a non-ABC transporter, and drug resistance in epilepsy. BMC Neurosci. 2005;6:61.

8. Ben-Menachem E. Preliminary efficacy of levetiracetam in monotherapy. Epileptic Disord. 2003;5(Suppl 1):S51–S55.

9. Blanchette M, Kent WJ, Riemer C, et al. Aligning multiple genomic sequences with the threaded blockset aligner. Genome Res. 2004;14(4):708–715.

10. Braga MF, Aroniadou-Anderjaska V, Post RM, et al. Lamotrigine reduces spontaneous and evoked GABAA receptor-mediated synaptic transmission in the basolateral amygdala: Implications for its effects in seizure and affective disorders. Neuropharmacology. 2002;42(4):522–529.

11. Chung WH, Hung SI, Hong HS, et al. Medical genetics: A marker for Stevens-Johnson syndrome. Nature. 2004;428(6982):486.

12. Crawford DC, Akey DT, Nickerson DA. The patterns of natural variation in human genes. Annu Rev Genomics Hum Genet. 2005;6:287–312.

13. Cunningham MO, Jones RS. The anticonvulsant, lamotrigine decreases spontaneous glutamate release but increases spontaneous GABA release in the rat entorhinal cortex in vitro. Neuropharmacology. 2000;39(11):2139–2146.

14. D’Andrea G, D’Ambrosio RL, Di Perna P, et al. A polymorphism in the VKORC1 gene is associated with an interindividual variability in the dose-anticoagulant effect of warfarin. Blood. 2005;105(2):645–649.

15. DeLorenzo RJ. Calcium-calmodulin protein phosphorylation in neuronal transmission: A molecular approach to neuronal excitability and anticonvulsant drug action. Adv Neurol. 1983;34:325–338.

16. DeLorenzo RJ. Calmodulin systems in neuronal excitability: A molecular approach to epilepsy. Ann Neurol. 1984;16(Suppl):S104–S114.

17. Dlugos DJ, Sammel MD, Strom BL, et al. Response to first drug trial predicts outcome in childhood temporal lobe epilepsy. Neurology. 2001;57(12):2259–2264.

18. Dombrowski SM, Desai SY, Marroni M, et al. Overexpression of multiple drug resistance genes in endothelial cells from patients with refractory epilepsy. Epilepsia. 2001;42(12):1501–1506.

19. Dooley DJ, Donovan CM, Pugsley TA. Stimulus-dependent modulation of [(3)H]norepinephrine release from rat neocortical slices by gabapentin and pregabalin. J Pharmacol Exp Ther. 2000;295(3):1086–1093.

20. Eadie M. Phenytoin, 2nd ed. Oxford, UK: Blackwell Publishing; 2004.

21. Eckstein-Ludwig U, Fei J, Schwarz W. Inhibition of uptake, steady-state currents, and transient charge movements generated by the neuronal GABA transporter by various anticonvulsant drugs. Br J Pharmacol. 1999;128(1):92–102.

22. Escayg A, Heils A, MacDonald BT, et al. A novel SCN1A mutation associated with generalized epilepsy with febrile seizures plus—and prevalence of variants in patients with epilepsy. Am J Hum Genet. 2001;68(4):866–873.

23. Escayg A, MacDonald BT, Meisler MH, et al. Mutations of SCN1A, encoding a neuronal sodium channel, in two families with GEFS⁺2. Nat Genet. 2000;24(4):343–345.

24. Fritz N, Glogau S, Hoffmann J, et al. Efficacy and cognitive side effects of tiagabine and topiramate in patients with epilepsy. Epilepsy Behav. 2005;6(3):373–381.

25. Fukuda A. Diuretic soothes seizures in newborns. Nat Med. 2005;11(11):1153–1154.

26. Gastaldi M, Bartolomei F, Massacrier A, et al. Increase in mRNAs encoding neonatal II and III sodium channel alpha-isoforms during kainate-induced seizures in adult rat hippocampus. Brain Res Mol Brain Res. 1997;44(2):179–190.

27. Gilbert JC, Wyllie MG. Proceedings: The effects of the anticonvulsant ethosuximide on adenosine triphosphatase activities of synaptosomes prepared from rat cerebral cortex. Br J Pharmacol. 1974;52(1):P139—P140.

28. Goldstein DB, Tate SK, Sisodiya SM. Pharmacogenetics goes genomic. Nat Rev Genet. 2003;4(12):937–947.

29. Hitzl M, Drescher S, van der Kuip H, et al. The C3435T mutation in the human MDR1 gene is associated with altered efflux of the P-glycoprotein substrate rhodamine 123 from CD56⁺ natural killer cells. Pharmacogenetics. 2001;11(4):293–298.

30. Hoffmeyer S, Burk O, von Richter O, et al. Functional polymorphisms of the human multidrug-resistance gene: Multiple sequence variations and correlation of one allele with P-glycoprotein expression and activity in vivo. Proc Natl Acad Sci U S A. 2000;97(7):3473–3478.

31. Horton R, Wilming L, Rand V, et al. Gene map of the extended human MHC. Nat Rev Genet. 2004;5(12):889–899.

32. Huang CW, Huang CC, Liu YC, et al. Inhibitory effect of lamotrigine on A-type potassium current in hippocampal neuron-derived H19-7 cells. Epilepsia. 2004;45(7):729–736.

P.1505

33. Hung CC, Tai JJ, Lin CJ, et al. Complex haplotypic effects of the ABCB1 gene on epilepsy treatment response. Pharmacogenomics. 2005;6(4):411–417.

34. Hung SI, Chung WH, Jee SH, et al. Genetic susceptibility to carbamazepine-induced cutaneous adverse drug reactions. Pharmacogenet Genomics. 2006;16(4):297–306.

35. Hurtado B, Koepp MJ, Sander JW, et al. The impact of levetiracetam on challenging behavior. Epilepsy Behav. 2006;8(3):588–592.

36. Karchin R, Diekhans M, Kelly L, et al. LS-SNP: Large-scale annotation of coding non-synonymous SNPs based on multiple information sources. Bioinformatics. 2005;21(12):2814–2820.

37. Kidd RS, Straughn AB, Meyer MC, et al. Pharmacokinetics of chlorpheniramine, phenytoin, glipizide, and nifedipine in an individual homozygous for the CYP2C9*3 allele. Pharmacogenetics. 1999;9(1):71–80.

38. Kockelmann E, Elger CE, Helmstaedter C. Significant improvement in frontal lobe associated neuropsychological functions after withdrawal of topiramate in epilepsy patients. Epilepsy Res. 2003;54(2-3):171–178.

39. Krakow K, Walker M, Otoul C, et al. Long-term continuation of levetiracetam in patients with refractory epilepsy. Neurology. 2001;56(12):1772–1774.

40. Kwan P, Brodie MJ. Early identification of refractory epilepsy. N Engl J Med. 2000;342(5):314–319.

41. Kwan P, Brodie MJ. Effectiveness of first antiepileptic drug. Epilepsia. 2001;42(10):1255–1260.

42. Kwan P, Sills GJ, Brodie MJ. The mechanisms of action of commonly used antiepileptic drugs. Pharmacol Ther. 2001;90(1):21–34.

43. Leach JP, Sills GJ, Butler E, et al. Neurochemical actions of gabapentin in mouse brain. Epilepsy Res. 1997;27(3):175–180.

44. Lee AY, Kim MJ, Chey WY, et al. Genetic polymorphism of cytochrome P450 2C9 in diphenylhydantoin-induced cutaneous adverse drug reactions. Eur J Clin Pharmacol. 2004;60(3):155–159.

45. Lee CR, Goldstein JA, Pieper JA. Cytochrome P450 2C9 polymorphisms: A comprehensive review of the in-vitro and human data. Pharmacogenetics. 2002;12(3):251–263.

46. Lukyanetz EA, Shkryl VM, Kostyuk PG. Selective blockade of N-type calcium channels by levetiracetam. Epilepsia. 2002;43(1):9–18.

47. Martin R, Kuzniecky R, Ho S, et al. Cognitive effects of topiramate, gabapentin, and lamotrigine in healthy young adults. Neurology. 1999;52(2):321–327.

48. McLeod HL. Pharmacogenetic analysis of clinically relevant genetic polymorphisms. Clin Infect Dis. 2005;41(Suppl 7):S449–S452.

49. Meador KJ, Loring DW, Vahle VJ, et al. Cognitive and behavioral effects of lamotrigine and topiramate in healthy volunteers. Neurology. 2005;64(12):2108–2114.

50. Meder WP, Dooley DJ. Modulation of K(⁺)-induced synaptosomal calcium influx by gabapentin. Brain Res. 2000;875(1–2):157–159.

51. Mula M, Trimble MR, Thompson P, et al. Topiramate and word-finding difficulties in patients with epilepsy. Neurology. 2003;60(7):1104–1107.

52. Munoz A, Mendez P, Alvarez-Leefmans F, et al. Altered expression of NKCC and KCC2 Transporters and GABAergic innervation in human epileptic hippocampus. Soc Neurosci Abstr. 2004;567(6).

53. Nassif A, Bensussan A, Boumsell L, et al. Toxic epidermal necrolysis: Effector cells are drug-specific cytotoxic T cells. J Allergy Clin Immunol. 2004;114(5):1209–1215.

54. Payne JA, Rivera C, Voipio J, et al. Cation-chloride co-transporters in neuronal communication, development and trauma. Trends Neurosci. 2003;26(4):199–206.

55. Pichler WJ. Modes of presentation of chemical neoantigens to the immune system. Toxicology. 2002;181–182:49–54.

56. Potschka H, Fedrowitz M, Loscher W. P-Glycoprotein-mediated efflux of phenobarbital, lamotrigine, and felbamate at the blood-brain barrier: Evidence from microdialysis experiments in rats. Neurosci Lett. 2002;327(3):173–176.

57. Potschka H, Loscher W. In vivo evidence for P-glycoprotein-mediated transport of phenytoin at the blood-brain barrier of rats. Epilepsia. 2001;42(10):1231–1240.

58. Prince BS, Goetz CM, Rihn TL, et al. Drug-related emergency department visits and hospital admissions. Am J Hosp Pharm. 1992;49(7):1696–1700.

59. Rettie AE, Haining RL, Bajpai M, et al. A common genetic basis for idiosyncratic toxicity of warfarin and phenytoin. Epilepsy Res. 1999;35(3):253–255.

60. Rhodes TH, Lossin C, Vanoye CG, et al. Noninactivating voltage-gated sodium channels in severe myoclonic epilepsy of infancy. Proc Natl Acad Sci U S A. 2004;101(30):11147–11152.

61. Rivera C, Voipio J, Payne JA, et al. The K⁺/Cl- co-transporter KCC2 renders GABA hyperpolarizing during neuronal maturation. Nature. 1999;397(6716):251–255.

62. Roujeau JC, Stern RS. Severe adverse cutaneous reactions to drugs. N Engl J Med. 1994;331(19):1272–1285.

63. Sander JW. Some aspects of prognosis in the epilepsies: A review. Epilepsia. 1993;34(6):1007–1016.

64. Schinkel AH, Wagenaar E, Mol CA, et al. P-glycoprotein in the blood-brain barrier of mice influences the brain penetration and pharmacological activity of many drugs. J Clin Invest. 1996;97(11):2517–2524.

65. Schnyder B, Mauri-Hellweg D, Zanni M, et al. Direct, MHC-dependent presentation of the drug sulfamethoxazole to human alphabeta T cell clones. J Clin Invest. 1997;100(1):136–141.

66. Schumacher TB, Beck H, Steinhauser C, et al. Effects of phenytoin, carbamazepine, and gabapentin on calcium channels in hippocampal granule cells from patients with temporal lobe epilepsy. Epilepsia. 1998;39(4):355–363.

67. Shimizu-Okabe C, Tanaka M, Matsuda K, et al. Changes in the expression of mRNAs for glycine and GABA-A receptor subunits in the rat freeze-lesion model of cortical malformation. Soc Neurosci Abstr. 2004;272(22).

68. Shorvon SD, van Rijckevorsel K. A new antiepileptic drug. J Neurol Neurosurg Psychiatry. 2002;72(4):426–429.

69. Siddiqui A, Kerb R, Weale ME, et al. Association of multidrug resistance in epilepsy with a polymorphism in the drug-transporter gene ABCB1. N Engl J Med. 2003;348(15):1442–1448.

70. Siepel A, Haussler D. Combining phylogenetic and hidden Markov models in biosequence analysis. J Comput Biol. 2004;11(2–3):413–428.

71. Sillanpaa M. Carbamazepine, 2nd ed. Oxford, UK: Blackwell Publishing; 2004.

72. Sisodiya SM, Lin WR, Harding BN, et al. Drug resistance in epilepsy: Expression of drug resistance proteins in common causes of refractory epilepsy. Brain. 2002;125(Pt 1):22–31.

73. Soranzo N, Cavalleri GL, Weale ME, et al. Identifying candidate causal variants responsible for altered activity of the ABCB1 multidrug resistance gene. Genome Res. 2004;14(7):1333–1344.

74. Takanashi K, Tainaka H, Kobayashi K, et al. CYP2C9 Ile359 and Leu359 variants: Enzyme kinetic study with seven substrates. Pharmacogenetics. 2000;10(2):95–104.

75. Tan NC, Heron SE, Scheffer IE, et al. Failure to confirm association of a polymorphism in ABCB1 with multidrug-resistant epilepsy. Neurology. 2004;63(6):1090–1092.

76. Tate SK, Depondt C, Sisodiya SM, et al. Genetic predictors of the maximum doses patients receive during clinical use of the anti-epileptic drugs carbamazepine and phenytoin. Proc Natl Acad Sci U S A. 2005;102(15):5507–5512.

77. Tatum WO, French JA, Faught E, et al. Postmarketing experience with to-piramate and cognition. Epilepsia. 2001;42(9):1134–1140.

78. Tishler DM, Weinberg KI, Hinton DR, et al. MDR1 gene expression in brain of patients with medically intractable epilepsy. Epilepsia. 1995;36(1):1–6.

79. Trinklein ND, Aldred SJ, Saldanha AJ, et al. Identification and functional analysis of human transcriptional promoters. Genome Res. 2003;13(2):308–312.

80. Twombly DA, Yoshii M, Narahashi T. Mechanisms of calcium channel block by phenytoin. J Pharmacol Exp Ther. 1988;246(1):189–195.

81. van der Weide J, Steijns LS, van Weelden MJ, et al. The effect of genetic polymorphism of cytochrome P450 CYP2C9 on phenytoin dose requirement. Pharmacogenetics. 2001;11(4):287–291.

82. Vinod KY, Subhash MN. Lamotrigine induced selective changes in 5-HT(1A) receptor mediated response in rat brain. Neurochem Int. 2002;40(4):315–319.

83. Wallace RH, Scheffer IE, Barnett S, et al. Neuronal sodium-channel alpha1-subunit mutations in generalized epilepsy with febrile seizures plus. Am J Hum Genet. 2001;68(4):859–865.

84. Wang D, Johnson AD, Papp AC, et al. Multidrug resistance polypeptide 1 (MDR1, ABCB1) variant 3435C>T affects mRNA stability. Pharmacogenet Genomics. 2005;15(10):693–704.

85. Weiss J, Kerpen CJ, Lindenmaier H, et al. Interaction of antiepileptic drugs with human P-glycoprotein in vitro. J Pharmacol Exp Ther. 2003;307(1):262–267.

86. White JR, Walczak TS, Leppik IE, et al. Discontinuation of levetira-cetam because of behavioral side effects: A case-control study. Neurology. 2003;61(9):1218–1221.

87. Wu WK, Pantaleo N. Evaluation of outpatient adverse drug reactions leading to hospitalization. Am J Health Syst Pharm. 2003;60(3):253–259.

88. Zeggini E, Rayner W, Morris AP, et al. An evaluation of HapMap sample size and tagging SNP performance in large-scale empirical and simulated data sets. Nat Genet. 2005;37(12):1320–1322.

89. Zimprich F, Sunder-Plassmann R, Stogmann E, et al. Association of an ABCB1 gene haplotype with pharmacoresistance in temporal lobe epilepsy. Neurology. 2004;63(6):1087–1089.

	Epilepsy: A Comprehensive Textbook 2nd Edition
	© 2008 Lippincott Williams & Wilkins

Epilepsy: A Comprehensive Textbook2nd Edition

Epilepsy: A Comprehensive Textbook
2nd Edition