The clinical impact of pharmacogenetics on the treatment of epilepsy

Transcription

1 CRITICAL REVIEW AND INVITED COMMENTARY The clinical impact of pharmacogenetics on the treatment of epilepsy Wolfgang Löscher, yulrich Klotz, zfritz Zimprich, and xdieter Schmidt Department of Pharmacology, Toxicology and Pharmacy, University of Veterinary Medicine, Hannover, Germany; ydr. Margarete Fischer-Bosch-Institute of Clinical Pharmacology and University of Tübingen, Stuttgart, Germany; zdepartment of Clinical Neurology, Medical University of Vienna, Austria; and xepilepsy Research Group, Berlin, Germany SUMMARY Drug treatment of epilepsy is characterized by unpredictability of efficacy, adverse drug reactions, and optimal doses in individual patients, which, at least in part, is a consequence of genetic variation. Since genetic variability in drug metabolism was reported to affect the treatment with phenytoin more than 25 years ago, the ultimate goal of pharmacogenetics is to use the genetic makeup of an individual to predict drug response and efficacy, as well as potential adverse drug events. However, determining the practical relevance of pharmacogenetic variants remains difficult, in part because of problems with study design and replication. This article reviews the published work with particular emphasis on pharmacogenetic alterations that may affect efficacy, tolerability, and safety of antiepileptic drugs (AEDs), including variation in genes encoding drug target (SCN1A), drug transport (ABCB1), drug metabolizing (CYP2C9, CYP2C19), and human leucocyte antigen (HLA) proteins. Although the current studies associating particular genes and their variants with seizure control or adverse events have inherent weaknesses and have not provided unifying conclusions, several results, for example that Asian patients with a particular HLA allele, HLA-B 1502, are at a higher risk for Stevens- Johnson syndrome when using carbamazepine, are helpful to increase our knowledge how genetic variation affects the treatment of epilepsy. Although genetic testing raises ethical and social issues, a better understanding of the genetic influences on epilepsy outcome is key to developing the much needed new therapeutic strategies for individuals with epilepsy. KEY WORDS: Antiepileptic drugs, Gene variation, Single nucleotide polymorphism, Drug resistance, Tolerability. Epilepsy is the most prevalent chronic neurological disorder, affecting at least 50 million people worldwide (Duncan et al., 2006). Studies in new-onset epilepsy indicate that seizures can be controlled by antiepileptic drugs (AEDs) in up to 70% of patients (Sillanpää & Schmidt, 2006). Yet, for someone developing the disease, key questions are whether or not their seizures will cease, what the optimal dose of an AED is, and whether serious adverse effects will occur. Although the incidence of rash differs among AEDs, predicting serious hypersensitivity reaction Accepted May 6, 2008; Early View publication July 9, Address correspondence to Wolfgang Lçscher, Department of Pharmacology, Toxicology and Pharmacy, University of Veterinary Medicine Hannover, Bünteweg 17, Hannover, Germany. wolfgang.loescher@tiho-hannover.de Wiley Periodicals, Inc. ª 2008 International League Against Epilepsy in an individual patient exposed to a drug that may cause a rash is currently not possible. Furthermore, the optimal doses of AEDs may differ four-fold among individuals (Kwan & Brodie, 2001). In addition, prognosis varies considerably among the different types of epilepsy (Semah et al., 1998). Moreover, the treatment outcome may vary even between patients with seemingly the same epilepsy syndrome, and its determinants are largely unknown (Schmidt & Lçscher, 2005). Although many factors may contribute to variability of clinical outcome in individual patients, unpredictability may, at least in part, result from genetic variation. The influence of genes on outcome of drug treatment is a rapidly evolving field termed pharmacogenetics over 40 years ago by the German geneticist Friedrich Vogel (Vogel, 1959). The ultimate goal of pharmacogenetics is to use the genetic makeup of an individual to predict drug response and efficacy, as well as potential 1

2 2 W. Löscher et al. Figure 1. The course of an AED tablet from drug absorption in the gastrointestinal tract to drug distribution to the brain, drug actions at brain targets, and finally drug metabolism and elimination, and how genetic variation may affect these processes. The ultimate goal of pharmacogenetics is to optimize the treatment of the individual patient and to predict the clinical outcome including drug response, adverse effects, comorbidity, and health status. Epilepsia ILAE adverse drug events. The topic of pharmacogenetics in epilepsy has recently been covered in several excellent review articles (Ferraro & Buono, 2005; Sisodiya, 2005; Depondt, 2006; Ferraro et al., 2006; Szoeke et al., 2006; Mann & Pons, 2007; Tate & Sisodiya, 2007), and several aspects of this topic are described in much greater detail in some of these recent reviews. The primary aim of the present review is to critically evaluate if and to what degree genetic variation is affecting outcome of medical treatment of epilepsies in the individual patient. For this purpose, we subdivided the review in two main sections: (1) genetic variation that has a potential effect on clinical efficacy of AEDs and (2) gene variants that may affect tolerability and safety of AEDs, with particular emphasis on more recent studies that have not yet been covered by previous reviews. The Promise of Pharmacogenetics in Clinical Practice The concept of personalized medicine is receiving much attention, and expectations have been raised that pharmacogenetics may be an important tool to optimize the treatment of epilepsy for the individual patient. As outlined above, treatment of epilepsy with AEDs is complicated by unpredictability of efficacy, adverse drug reactions (ADRs), and optimal doses in individual patients. At least in part, this unpredictability may result from individual genes whose variations exert a measurable influence on the effect of a given drug. It is becoming increasingly clear that genetic polymorphisms play an integral role in variability of both AED pharmacokinetics and pharmacodynamics. Single nucleotide polymorphisms (SNPs), variations at a single site in the DNA, are the most frequent form of sequence variations in the human genome and may affect the efficacy, tolerability, safety, and duration of action of AEDs. With a better molecular understanding of the variability of drug action, there is some hope for the future to cope better with this pertinent clinical problem by incorporating pharmacogenetic principles (Pirazzoli & Recchia, 2004; Kirchheiner et al., 2005; Lewis, 2005; Phillips & van Bebber, 2005; Kalow, 2006; Goldstein et al., 2007; Lesko, 2007). The present article will critically evaluate the clinical evidence how the treatment of epilepsy could in fact benefit from considering and applying pharmacogenetic principles. For this review, we will follow the course of an AED tablet from drug absorption in the gastrointestinal tract to drug distribution to the brain, drug actions at brain targets, and finally hepatic metabolism and renal excretion (both processes contributing to drug elimination) as illustrated in Fig. 1. Following this course, we will consider how genetic variations of these individual processes may affect the efficacy of AED treatment. In addition, we will discuss the impact of gene variation on side effects of AEDs, which may involve several of the above processes. At the end of each subsection, we will briefly assess the potential clinical impact. Finally, we will briefly consider medicoeconomic issues, availability of genetic testing, legal aspects, and the ethical issues at stake. Genetic Variation That May Affect Clinical Efficacy of Antiepileptic Drugs In general, as with other drugs, the absorption and distribution and, hence clinical efficacy, of AEDs depend on their physicochemical characteristics such as lipophilicity,

3 Pharmacogenetics in Epilepsy 3 solubility, molecular weight, and ionic state. Most AEDs are sufficiently lipophilic to penetrate biomembranes by passive diffusion (Lçscher & Potschka, 2005a). However, drug efflux transporters at the gastrointestinal tract and blood-brain barrier (BBB) may limit absorption and brain uptake of AEDs, so that genetic variation in the expression and functionality of such transporters may determine clinical outcome (Fig. 1). Furthermore, polymorphisms in brain targets of AEDs may affect their effectiveness. Finally, genetic variation in drug metabolism and elimination may contribute to interindividual variability in drug response (Fig. 1). Drug absorption Traditionally, drug absorption has been considered as a passive process governed mainly by the physicochemical properties of the drug. However, in addition to passive transcellular and paracellular transport mechanisms, carrier-mediated transport across membranes plays an important role in drug and nutrient absorption (Anderle et al., 2004). The role of such transporters for absorption of AEDs from the intestine in the blood stream is largely unknown, with two prominent exceptions, namely, gabapentin (GBP) and pregabalin (PGB). GBP is absorbed via the large neutral amino acid carrier, system L, in the proximal small intestine, resulting in dose-dependent absorption (due to saturation of this transport system) and inhibition of intestinal uptake of GBP by several large neutral amino acids (Piyapolrungroj et al., 2001; Anderle et al., 2004). Despite the structural similarity between GBP and PGB, PGB absorption exhibits nearly linear pharmacokinetics, which is explained by lower affinity of PGB for the L-type system and involvement of other amino acid transport systems in the intestinal uptake of PGB (Piyapolrungroj et al., 2001; Su et al., 2005). Whether polymorphisms in these carriers affect the intestinal uptake of GBP or PGB is not known. Transporters can also play a crucial role in limiting drug absorption through drug secretion into the intestinal lumen. A number of drug efflux transporters are expressed by enterocytes, including P-glycoprotein (Pgp), members of the multidrug resistance-associated protein (MRP) family, and breast cancer-related protein (BCRP) (Anderle et al., 2004; Fromm, 2004; Ito et al., 2005). Furthermore, enterocytes express the major drug-metabolizing enzymes CYP3A4 and CYP2C9/19, which, in concert with efflux transporters, may restrict the oral bioavailability of drugs that are substrates for CYP enzymes and/or drug efflux transporters such as Pgp (Fromm, 2004). Drugs, including AEDs such as carbamazepine (CBZ), may induce the expression of drug-metabolizing enzymes and drug efflux transporters in the intestinal tract, thereby reducing their own absorption (Giessmann et al., 2004; Christians et al., 2005). Furthermore, genetic and environmental factors probably have a major role in the regulation of the basal expression and function of intestinal CYP3A4 and Pgp (Fromm, 2004). Many drugs, including several AEDs, are Pgp substrates, so that genetic variation in the ABCB1 [multidrug resistance 1 (MDR1)] gene that encodes Pgp may have dramatic consequences for the pharmacological behavior of substrate drugs (Marzolini et al., 2004; Lçscher & Potschka, 2005a, 2005b). The human ABCB1 gene is composed of 29 exons (for details see latest data releases at and A synonymous SNP in exon 27 (C3435T) was the first variant to be associated with altered protein expression in the human intestinal tract, although the SNP does not change the encoded amino acid (Hoffmeyer et al., 2000). Pgp expression in the duodenum of individuals with the CC genotype was noted to be two-fold higher when compared with that in individuals with the TT genotype, which was associated with significantly decreased plasma concentrations of the Pgp substrate digoxin after oral administration, suggesting lower drug absorption in individuals with high intestinal Pgp levels (Hoffmeyer et al., 2000). The observation that the 3435C allele in exon 27 is associated with lower digoxin levels was confirmed by some, but not all subsequent studies (Marzolini et al., 2004). The synonymous 3435C>T polymorphism is in linkage disequilibrium with a synonymous SNP in exon 13 (1236C>T) and a nonsynonymous SNP in exon 22 (2677G>TA), suggesting that the observed functional differences in Pgp, initially attributed to the exon 27 synonymous SNP, may be the result of the associated nonsynonymous polymorphism in exon 22, which results in amino acid exchanges (Ala893Ser or Ala893Thr) (Marzolini et al., 2004). However, a recent study by Gottesman s group showed that the synonymous C3435T SNP in exon 27, although not resulting in amino acid changes itself, is not silent, but results in Pgp with altered drug and inhibitor interactions (Kimchi-Sarfaty et al., 2007). Similar messenger RNA (mrna) and protein levels, but altered conformations, were found for wild-type and polymorphic Pgp. Kimchi-Safarty et al. (2007) hypothesized that the presence of a rare codon, marked by the synonymous polymorphism, affects the timing of cotranslational folding and insertion of Pgp into the membrane, thereby altering the structure of substrate and inhibitor interaction sites. This study is of immense importance, as it demonstrates for the first time that naturally occurring silent SNPs can lead to the synthesis of protein product with the same amino acid sequence but different structural and functional properties. Thus, silent SNPs should no longer be neglected in determining the likelihood of development of various diseases and should be taken into account in personalized drug treatment and development programs (Komar, 2007). Polymorphisms in ABCB1 change not only the oral bioavailability of digoxin, but also that of several other Pgp substrates, although data are conflicting (Marzolini et al.,

4 4 W. Löscher et al. 2004). Kerb et al. (2001) studied whether levels of phenytoin (PHT), which is a substrate of Pgp, correlate with the C3435T polymorphism in the ABCB1 gene. Genotyping and analyses of plasma levels of PHT after oral administration in 96 healthy Turkish volunteers showed that the 3435C>T polymorphism affects PHT plasma levels. The CC genotype was significantly more common in volunteers with low PHT levels. The effect of the C3435T polymorphism on oral bioavailability of PHT was also determined in 35 PHT-treated patients with epilepsy (Kerb et al., 2001). In a more recent study by Simon et al. (2007) in patients with epilepsy, intestinal Pgp expression and PHT and CBZ dose requirements were influenced by the genotype in position 3435 and 2677 of the ABCB1 gene, thus confirming the data of Kerb et al. (2001). Furthermore, Ebid et al. (2007) reported that the 3435 genotype affected plasma levels of PHT in epilepsy patients in that subjects with the CC genotype were more likely to have low PHT levels (<10 lg/ml) than patients with the TT phenotype. In this respect, it is also interesting to note that Lazarowski and colleagues reported that some patients with refractory epilepsy and high expression of brain Pgp have persistently subtherapeutic plasma levels of CBZ, valproate (VPA) or PHT despite administration of high doses of these AEDs, suggesting that absorption and/or elimination of AEDs in such patients may be affected by increased expression of Pgp in the periphery (Lazarowski et al., 1999, 2004, 2007). However, data on whether ABCB1 3435CC genotype is indeed associated with increased Pgp level in duodenal enterocytes are conflicting (Goto et al., 2002; Nakamura et al., 2002; Sakaeda et al., 2002). As a result, it remains circumferential that any reduction in bioavailability of AEDs observed can be attributed to ABCB genotype. As mentioned above, presystemic drug elimination can occur already during the intestinal absorption process of drugs, because drug metabolizing enzymes are also expressed along the human gastrointestinal tract (Ding & Kaminsky, 2003; Glaeser et al., 2005; Thçrn et al., 2005). However, this intestinal metabolic capacity is regarded as much less effective as that in the liver (Lin et al., 1999), and there are no data demonstrating that the oral bioavailability of any AED is affected to a significant extent by an intestinal first-pass effect. Clinical impact Overall, one may conclude that the clinical contribution of genetic mutations of drug metabolizing enzymes to the interindividual variability in drug absorption can be neglected for most patients. However, it is possible that mutations in genes encoding proteins involved in drug absorption are more relevant in patients with compromised liver function taking a CYP-metabolized AED, or for patients with reduced kidney function who may be taking a drug that undergoes primarily renal excretion. It remains unclear if the increased expression of intestinal Pgp, which has been suggested to lead to poor biovailability of AEDs, contributes to poor seizure control in some patients. Drug distribution For drugs such as AEDs that act on targets in the brain, sufficient penetration through the BBB is a prerequisite for therapeutic efficacy (Fig. 1). The BBB is a physical and metabolic barrier between the brain and the systemic circulation, which serves to protect and regulate the microenvironment of the brain (Huber et al., 2001). Most AEDs are quite lipophilic (c.f., DrugBank, redpoll.pharmacy.ualberta.ca/drugbank/index.html), so that they can easily penetrate through the brain capillary endothelial cells that form the BBB (Lçscher & Potschka, 2005a). However, efflux transporters such as Pgp, which are located at the apical (luminal) membrane of brain capillary endothelial cells and protect the brain from intoxication by lipophilic xenobiotics, may restrict the brain uptake of AEDs and mediate extrusion of AEDs from the brain (Kwan & Brodie, 2005; Lçscher & Potschka, 2005a). Because most AEDs are only weak substrates for Pgp, the basal (constitutive) expression of Pgp at the BBB is unlikely to restrict brain penetration of AEDs to any clinically important extent (Lçscher & Potschka, 2005a). However, intrinsic or acquired overexpression of Pgp in the BBB may critically limit drug penetration into the brain, leading to resistance against all AEDs that are substrates of Pgp (Kwan & Brodie, 2005; Lçscher & Potschka, 2005a, 2005b). Such Pgp overexpression can result from the effects of disease or drug treatment on Pgp expression or from ABCB1 polymorphisms and might explain the clinical observation that patients with refractory epilepsy are usually resistant to a broad range of AEDs with different mechanisms of action (Kwan & Brodie, 2005). Increased expression of Pgp and other drug efflux transporters has been determined in epileptogenic brain tissue of patients with refractory epilepsy (Fig. 2A) and in rodent models of AED-resistant epilepsy (Kwan & Brodie, 2005; Lçscher & Potschka, 2005a). In rodent models of temporal lobe epilepsy (TLE), the increased Pgp expression in the hippocampus and parahippocampal regions was associated with significantly decreased concentrations of AEDs in these regions (Rizzi et al., 2002; Van Vliet et al., 2007). In patients with oxcarbazepine (OXC)-resistant epilepsy, the brain tissue expression of ABCB1 mrna was found to be inversely correlated with brain levels of 10,11-dihydro-10-hydroxy-5H-dibenzo(b,f)azepine-5-carboxamide (10-OHCBZ), the active metabolite of OXC, indicating that Pgp may play a role in the pharmacoresistance to OXC by causing insufficient concentrations of its active metabolite at neuronal targets (Marchi et al., 2005). Using an in vitro BBB model with human capillary endothelial cells from either normal brain

5 Pharmacogenetics in Epilepsy 5 or drug-resistant epileptic brain, Cucullo et al. (2007) recently reported a dramatically reduced permeability of PHT across the in vitro BBB formed from endothelial cells of patients with refractory epilepsy, which could be partially counteracted by the selective Pgp inhibitor tariquidar (Fig. 2B). In line with this finding, the decrease in brain concentrations and resistance to AEDs, such as PHT or phenobarbital (PB), associated with Pgp overexpression in rodent models could be counteracted by tariquidar in vivo, suggesting a causal association between Pgp overexpression and AED resistance (Brandt et al., 2006; van Vliet et al., 2006; van Vliet et al., 2007). In 2003, Siddiqui et al. reported the C3435T polymorphism in the ABCB1 gene as being associated with resistance to multiple AEDs, leading to the suggestion that drug resistance in epilepsy might be genetically determined, which could open new therapeutic avenues. In the genetic association study of Siddiqui et al. (2003), which was performed as a retrospective case-control study by comparing the frequencies of the ABCB1 C3435T variant in 115 AED responders with 200 AED-resistant patients and 200 nonepileptic controls, it was shown that patients with multidrug-resistant epilepsy were significantly more likely to be homozygous for the C allele than the T allele. Because the CC genotype has been associated with increased expression of intestinal Pgp (Hoffmeyer et al., 2000), the data of Siddiqui et al. (2003) suggested that the CC genotype may be associated with increased expression and functionality of Pgp also at the BBB, leading to reduced AED levels at their brain targets. In a follow-up study by the same group (Soranzo et al., 2004), the association of AED resistance with the 3435C>T polymorphism was confirmed in a larger group of patients, and intronic sites that are strongly associated with the 3435C>T polymorphism were identified. The increased prevalence of the CC genotype of ABCB in patients with drug-resistant epilepsy reported by Siddiqui et al. (2003) initiated several subsequent genetic association studies, using a candidate gene approach with either one SNP or a haplotype (Table 1). Six of these studies, genotyping either ABCB or the common haplotype combination, ABCB1 3435C>T-2677G>T-1236C>T, confirmed the association between the 3435 SNP or the three-snp haplotype (containing the 3435 SNP) and AEDresistant epilepsy. However, in two studies in non-caucasian subjects, the association was in the reverse direction compared to studies in Caucasian subjects in that patients with drug-resistant epilepsy were more likely to have the TT genotype compared with those with drug-responsive epilepsy (Seo et al., 2006; Kwan et al., 2007), highlighting the complexity of the possible role of ABCB1 polymorphisms in AED response in different ethnic populations. In contrast to studies showing an association of ABCB1 polymorphisms and AED resistance, six other retrospective association studies and one prospective cohort study in either Caucasian or non-caucasian subjects did not identify any significant association between ABCB1 polymorphisms and response to AEDs (Table 1). However, only one of these negative studies (Tan et al., 2004) was an exact replication of the first report by Siddiqui et al. (2003), so that the authors of the first report argued that phenotypic differences between studies may explain the failure to robustly identify a role for Pgp from such genetic association studies (Sisodiya et al., 2005; Sisodiya & Goldstein, 2007; Tate & Sisodiya, 2007). In addition to inconsistent phenotype definition (i.e., definition of resistance versus response to AEDs) among studies, there are various other potential explanations for the discordant results, including inadequate power, potential confounding by comorbidity and comedication, population substructure, genotyping error, overlap in substrate specificity between Pgp and other drug efflux transporters, and inclusion of AEDs that might not be Pgp substrates (Leschziner et al., 2007). Twelve of the 15 genetic association studies summarized in Table 1 included patients on treatment with various AEDs, for several of which it is either not yet known whether they are transported by Pgp or which do not seem to be transported by Pgp (e.g., VPA) (Baltes et al., 2007). Only three studies (Seo et al., 2006, Ebid et al., 2007; Ozgon et al., 2007) included patients on a single AED, either PHT or CBZ. For CBZ, data on transport by Pgp are at best equivocal (Owen et al., 2001; Potschka et al., 2001; Rizzi et al., 2002), whereas there is ample evidence that PHT is transported by Pgp (Lçscher & Potschka, 2005a), which could explain the recent finding of Ebid et al. (2007) that the CC genotype of ABCB is significantly associated with PHT resistance in patients with epilepsy (Fig. 2C). It should be noted that the results of the Ebid et al. (2007) study differed substantially from other similar studies, in that a highly significant association between 3435CC and resistance to PHT was found despite a smaller sample size (100 patients only compared with >300 in most other studies). A number of concerns exist about the clarity of the report and the study design. The original report did not make it clear whether PHT was used as monotherapy, and response was evaluated over a 3-month period only. Patients were already taking PHT at enrollment, but the baseline seizure frequency was unknown. The underlying epilepsy syndromes, which strongly influence outcome, were unknown and not accounted for in the analysis. All except one of the studies on ABCB1 polymorphisms and AED resistance summarized in Table 1 were retrospective association studies with the inherent potential biases of a retrospective case-control design. Only in one study the association of the ABCB1 gene with drug withdrawal due to poor seizure control or adverse effects was determined prospectively in 503 patients with mostly new-onset epilepsy (Leschziner et al., 2006). Although all patients were followed prospectively to determine the

6 6 W. Löscher et al. response to medication, ABCB1 3435C>T polymorphism and three-snp haplotype, plus a comprehensive set of tag SNPs across ABCB1 and adjacent ABCB4, were genotyped retrospectively (i.e., after outcomes were known). Randomly selected genome-wide HapMap SNPs (n ¼ 129) were genotyped in all patients for genomic control. There was no association of the ABCB1 Figure 2. Continued. 3435C>T polymorphism, the three-snp haplotype, or any gene-wide tag SNP with time to first seizure after starting drug therapy, time to 12-month remission, or time to drug withdrawal due to unacceptable side-effects or to lack of seizure control (Leschziner et al., 2006). The limitations of this important first prospective study include a wide heterogeneity of the study population and the drug treatment. The study population was taking a number of different AEDs, none accounting for more than 25%. Topiramate (TPM) and lamotrigine (LTG) were the most commonly prescribed AEDs, followed by CBZ, GBP, OXC, and VPA. Thus, one explanation why this study did not find the reported association between ABCB1 genotype and drug resistence is that Pgp may only be of relevance in the context of some AEDs and not others. The study population was also heterogeneous in terms of their treatment, 80% of patients were previously untreated, 17% had monotherapy with suboptimal seizure control, and 3% had recent seizures after remission. Thus it is difficult to compare the conflicting results of studies in chronic drug-resistant epilepsy with the results in mostly newly treated epilepsy. ƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒƒ Figure 2. Alterations in drug efflux transporter expression affect the distribution and efficacy of phenytoin (PHT) in patients with intractable epilepsy. (A) Overexpression of drug efflux transporter genes in brain capillary endothelial cells (that form the BBB) from patients with drug-resistant epilepsy undergoing temporal lobectomies to relieve medically intractable seizures. Patients with drug-resistant epilepsy had significantly higher expression of the genes encoding Pgp (MDR1; ABCB1), MRP2, and MRP5 than controls (age-matched nonepileptic patients with aneurysms). Asterisks indicate significant difference to control (p < 0.05). Based on data from Dombrowski et al. (2001). (B) The overexpression of drug efflux transporter genes in brain capillaries of patients with refractory epilepsy reduces the permeability to PHT. Transport of PHT was studied in an in vitro BBB model. The reduced penetration of PHT could be partially counteracted by the selective Pgp inhibitor tariquidar, indicating that Pgp was involved in the reduced transport of PHT. Asterisk indicates significant difference to control, circle significant difference to PHT alone (p < 0.05). Data are from Cucullo et al. (2007). (C) One hundred patients with partial or generalized tonic clonic seizures receiving PHT monotherapy were genotyped for ABCB1 C3435T polymorphisms. Patients with PHT-resistant epilepsy were significantly more likely to have the CC genotype than the TT genotype compared with responsive patients (p < ). Data are from Ebid et al. (2007). Epilepsia ILAE

7 Pharmacogenetics in Epilepsy 7 Table 1. Association between polymorphisms in the human multidrug resistance (MDR1, ABCB1) gene and resistance to antiepileptic drug treatment in patients with epilepsy Polymorphism Number of epilepsy patients Type of Association of Authors Origin in MDR1 Responders Nonresponders epilepsy/aeds polymorphism with resistance Positive studies Siddiqui et al., 2003 UK 3435C>T Various/various Yes Soranzo et al., 2004 UK 3435C>T Various/various Yes IVS T>C Yes Zimprich et al., 2004 Austria 3-SNP haplotype a 210 TLE/various Yes (within the resistant group) Hung et al., 2005 Taiwan 3-SNP haplotype a Various/various Yes Seo et al., 2006 Japan 3-SNP haplotype a Various/CBZ Yes (but in reverse direction) Kwan et al., 2007 China (Hong 3435C>T Various/various Yes (but in reverse direction) Kong) T Ebid et al., 2007 Egypt 3435C>T Various/PHT Yes Hung et al., 2007 Taiwan 3435C>T and 2677G>T Various/various Yes Negative studies Tan et al., 2004 Australia 3435C>T Various/various No Sills et al., 2005 Scotland 3435C>T Various/various No Kim et al., 2006a Korea 3435C>T Various/various No Kim et al., 2006b Korea 3-SNP haplotype a Various/various No Leschziner et al., 2006 UK 3435C>T 503 Various/various No 3-SNP haplotype a (prospective response evaluation) No Ozgon et al., 2007 Turkey 3435C>T Various/CBZ No Shahwan et al., 2007 Ireland 3435C>T and other SNPs Various/various No and SNP haplotypes a Three-SNP haplotype ¼ 3435C>T, 2677G>T, and 1236C>T. For directly addressing whether ABCB1 polymorphisms affect drug distribution into the human brain, positron emission tomography (PET) with the Pgp substrate 11 C-verapamil was used. Healthy volunteers differing in ABCB1 haplotypes did not differ in brain distribution of 11 C-verapamil (Brunner et al., 2005; Takano et al., 2006). However, verapamil is not an ideal PET ligand to study functional consequences of overexpression of Pgp at the BBB, because it hardly crosses the BBB in humans at constitutive BBB expression of Pgp (Sasongko et al., 2005). Using single photon emission computed tomography (SPECT) with the Pgp substrate [ 99m Tc]-sestamibi, Jensen et al. (2006) recently reported in an abstract reduced brain uptake of [ 99m Tc]-sestamibi in epilepsy patients with the C3435C and G2677G ABCB1 genotypes, which was correlated with AED resistance. Although the latter study may indicate that the 3535C>T SNP is associated with higher expression of Pgp in the brain, direct evidence for such an association is missing as yet. To our knowledge, there are only two studies that determined expression of Pgp in brain tissue of patients genotyped for the C3435T polymorphism (Vogelgesang et al., 2002, 2004). In one study, brain tissue samples were obtained at autopsy from 243 nondemented subjects for determination of vascular Pgp expression in the medial temporal lobe. The highest Pgp expression was determined in the 3435CC genotype, but the difference to the other 3435 genotypes was not statistically significant (Vogelgesang et al., 2002). In the second study, brain specimens were obtained from 14 patients with dysembryoplastic neuroepithelial tumors (DNT) undergoing temporal lobectomy because of intractable epilepsy. In all patients, the expression of Pgp was significantly higher in DNT compared with peritumoral tissue. The by far highest Pgp expression was found in one patient with the 3435CC genotype, who exhibited three times higher Pgp expression than the 10 CT and 3 TT carriers (Vogelgesang et al., 2004). However, the small sample size did not allow concluding whether there was a significant correlation between Pgp expression and C3435T genotype in patients with epilepsy. The most convincing evidence for an association between ABCB1 genotype and Pgp expression, function and therapeutic drug response was recently reported by Basic et al. (2008), who studied in a prospective fashion whether the C3435T polymorphism affects the brain uptake of PB in patients with generalized epilepsy. Genotyping was performed in 60 patients with idiopathic generalized epilepsy with tonic clonic seizures on PB monotherapy. PB analysis in plasma and cerebrospinal fluid (CSF) demonstrated that, while the C3435T polymorphism did not affect plasma levels of PB (Fig. 3A), the CC genotype of 3435 was associated with significantly lower PB levels in CSF (Fig. 3B) and a significantly lower CSF/plasma ratio (Fig. 3C) than the CT or TT genotypes (Basic et al., 2008). Furthermore, patients with the CC genotype had a significantly higher seizure frequency than patients with the CT or TT genotype (Fig. 3D). The daily doses of PB did not differ significantly between genotypes.

8 8 W. Löscher et al. Figure 3. The influence of the ABCB1 3435C>T polymorphism on CNS levels and efficacy of phenobarbital (PB) in patients with idiopathic primary generalized epilepsy with tonic clonic seizures. Genotyping was performed in 60 patients, who were on monotherapy with PB. The CC, CT, and TT genotypes of 3435 were analyzed in 16, 31, and 13 patients, respectively. Data in panels A, B, and C are shown as means ± SEM; data in panel D is shown as proportion of patients. Plasma levels of PB were not affected by 3435 genotype (A). However, analysis of variance (ANOVA) indicated significant differences between genotypes for PB levels in CSF (p ¼ 0.006) (B) and the CSF/plasma ratio of PB (p < 0.001) (C). Post hoc analysis indicated significantly lower CSF levels and CSF/plasma ratios in patients with the CC genotype compared to the other genotypes. (D) Six-month seizure rates in patients under treatment with PB. Patients with the CC genotype showed a significantly higher seizure frequency, compared to patients with the CT or TT genotype (p < 0.001). Data are from Basic et al. (2008). Epilepsia ILAE Furthermore, in the same group of patients, the G2677T/A polymorphism in the ABCB1 gene had no effect of CSF levels of PB. Although valuable, the study by Basic et al. (2008) has its limitations. Important clinical data are missing, for example if the patients had other seizure types such as absence or myoclonic seizures in addition to the reported generalized tonic seizures, which would be common for idiopathic generalized epilepsies. Another concern is that the patients were treated with PB, which is not recommended for first-line treatment of idiopathic generalized epilepsies mainly because of its sedative side effects and poor efficacy against absence seizures (Benbadis, 2005). The study of Basic et al. (2008) seems to confirm the association between the CC genotype at ABCB and AED resistance described by Siddiqui et al. (2003) and several other groups, although the effect size may be small. However, causality has not been proven in any of these studies, but all reported findings remain associations. Furthermore, several of the many association studies on ABCB1 in epilepsy not only analyzed drug responsive and nonresponsive epilepsy patients, they also analyzed ABCB1 genotypes in control subjects without epilepsy and found that in fact ABCB1 genotypes were associated with the disease per se rather than specifically with drug responsiveness. This suggests that the reported association studies on this gene may be looking simply at the random segregation of epilepsy patients into the two drug response groups such that in half of the studies an association was found and in the other half of the studies the results were negative. Almost all previous studies on gene variation that might affect AED distribution have dealt with polymorphisms in the ABCB1 gene that encodes Pgp. However, there are polymorphisms in other drug efflux transporters, such as

9 Pharmacogenetics in Epilepsy 9 members of the MRP family that may affect the distribution of AEDs (Lçscher & Potschka, 2005a) and need further investigation. One of these transporters, RLIP76, has been suggested to be involved in AED resistance by transporting both CBZ and PHT at the BBB (Awasthi et al., 2005), but a recent genetic analysis of RLIP76 genotypic and haplotypic frequencies in 783 patients with epilepsy and 359 healthy controls showed no significant differences for genotypic frequencies between drug-resistant and drug-responsive patients (Soranzo et al., 2007). In conclusion, the role of genetic variation in drug efflux transporter genes for AED distribution and efficacy remains uncertain at present. The recent prospective study of Basic et al. (2008) showing a significant and clinically relevant effect of the ABCB1 C3435T polymorphism on CSF levels of PB in patients with epilepsy (Fig. 3) indicates that the study design is of crucial importance for any definite conclusions. However, even though certain polymorphisms, such as the 3435C>T in ABCB1, may play a role in the phenomenon of refractory epilepsy, this cannot explain the fact that by definition such patients are refractory to multiple AEDs, which are generally chemically and pharmacologically diverse. Not all AEDs have a common element of pharmacology (e.g., not all AEDs are substrates for Pgp, and not all AEDs have the same therapeutic target), so that a single gene variant cannot affect responsiveness to multiple and diverse drugs. Rather, as a complex trait, drug responsiveness is expected to be determined in part by the effects of multiple genes acting both independently and in interaction with each other, so that in the end, individual responsiveness will be determined by the effects of many variations, including genes involved in drug distribution and genes encoding drug targets. Clinical impact In summary, with so many methodological problems of previous studies and inconsistent results, the evidence of association between ABCB1 polymorphisms and AED response is at best conflicting and does not support a major role for the C3435T polymorphism. However, more recent genetic association studies on AEDs known to be transported by Pgp are encouraging (Ebid et al., 2007; Basic et al., 2008), but these studies need to be replicated. Even if these data can be replicated, it should be noted that, as discussed above, association is not causation, so that studies that directly address whether ABCB1 polymorphisms significantly alter brain uptake of AEDs in patients with refractory epilepsy by PET imaging or analysis of brain tissue are needed. Drug targets Once an AED has successfully passed the BBB, its next step is to reach its molecular target in the brain. Although presently available AEDs appear to be directed against a relatively small number of targets mainly ion channels or other components of the synaptic machinery matters are complicated by the fact that many AEDs seem to work through multiple mechanisms, some of which are still unresolved (Rogawski & Lçscher, 2004). Recent efforts have revealed interesting genetic polymorphisms in some of these AED targets. However, it has to be stressed that, at present, we still know very little about the functionality of such polymorphisms and to what extent such variations have a clinically relevant impact on AED treatment. So far the most interesting data have been accrued for voltage-dependent Na + channels, which are the primary targets for a number of important AEDs such as CBZ, PHT, or LTG (Rogawski & Lçscher, 2004). Early observations had indicated that mutations in sodium channels may affect the clinical response to AEDs. The Dravet syndrome, which can be caused by de novo truncation mutations in the SCN1A sodium channel gene, is characterized by a marked aggravation of seizures upon treatment with LTG (Guerrini et al., 1998). This was explained by the preferential expression of SCN1A on inhibitory interneurons and the further reduction of this inhibitory component by Na + channel blocking agents such as LTG (Yu et al., 2006). Recently, genetic variations in sodium channels have received further attention because of suggestions that Na + channels might exhibit different electrophysiological properties in patients with pharmacoresistant epilepsy as compared to responsive patients. This idea originated from a study in which sodium currents were recorded in hippocampal neurons collected from 10 surgical specimens of patients with AED-resistant TLE and Ammon s horn sclerosis (AHS) (Remy et al., 2003). Recordings from three AED-responsive patients with partial epilepsy but without AHS were taken as controls. The authors then investigated the use-dependent block of CBZ on Na + channels, which is considered to be the main mechanism responsible for the antiepileptic action of this drug. In neurons from AHS-patients who were preoperatively shown to be resistant to CBZ, the use-dependent block of CBZ on voltage dependent sodium channels was lost and seizure activity elicited in vitro was resistant to CBZ. In the three non-ahs control cases, who were considered CBZ-responsive, the use-dependent block of CBZ was either still detectable (one case) or the in vitro seizure activity was responsive to CBZ (two cases). The main conclusion of this study that these functional differences were the cause of the pharmacoresistance was, however, questioned by many other researchers given the (inevitable) limitations of the few control samples. (The controls exhibited a different pathology, and the clinical CBZ responsiveness was not adequately established as one patient received no CBZ at all and two other patients only for 2 to 3 months). Still, even if the causal connection between these functional differences in Na + channels and the response to AED treatment remains as yet unsolved, the mechanism underlying the loss (or lack) of the use-dependent block

10 10 W. Löscher et al. appears very interesting. In particular, the question arose as to whether genetic factors might influence this phenomenon, by, for instance, determining the subunit composition or the structure of voltage-dependent Na + channels (Remy & Beck, 2006). The fact that functional properties of ion channels can be dynamically regulated via genetic mechanisms is actually well established. The best known example is probably the evolutionary well-conserved alternative splicing of exons coding for voltage sensors of ion channels (Plummer & Meisler, 1999). The voltage sensor regions determine the gating properties of the channels and, in the case of sodium channels, are thought to influence the interaction with classical Na + channel AEDs such as CBZ and PHT. In insects it has been shown that such alternatively spliced Na + channels can have strikingly different pharmacological properties by exhibiting markedly different sensitivities to pyrethroid insecticides (Tan et al., 2002). Given these data, recent studies investigated alternative splicing processes in humans, in the SCN1A gene, which is one of the Na + channel genes of the brain. In particular, the question was addressed whether the alternative splicing affecting the voltage sensor regions could be influenced by genetic polymorphisms and if there is a relationship to pharmacoresistance in epilepsy. In the SCN1A gene, exon 5 codes for one of the four voltage sensors of the channel, the domain I-S4 voltage sensor. Two alternatively spliced versions of this exon are present in the genomic DNA, a neonatal and an adult copy, which differ by three amino acids in the final product (Tate et al., 2005). Normally both exons are coexpressed in the adult brain. The neonatal exon can be drastically upregulated under several circumstances which also include seizures according to some but not all studies (Gastaldi et al., 1997; Aronica et al., 2001; Tate et al., 2005; Heinzen et al., 2007). The focus of two recent studies was a functional SNP (rs ) in the intron adjacent to exon 5 (Fig, 4). This SNP lies within a 5 slice donor site, a consensus sequence important for the splicing process (Tate et al., 2005; Heinzen et al., 2007). Apparently, this SNP determines whether the neonatal or the adult version of exon 5 is incorporated into the final channel. The ancestral G allele allows both exons to be expressed, whereas the mutant A allele almost abolishes the expression of the neonatal exon by disrupting the consensus sequence. The fact that the expression of these two alternatively spliced products is tightly controlled by the regulatory protein Nova-2 points to the existence of some functionality (Heinzen et al., 2007). However, at present we do not have any further information on how these alternatively spliced channels might differ from each other, as electrophysiological or pharmacological data are still lacking. Given the probable functionality, this SCN1A-SNP was further investigated in a large pharmacogenetic study involving 425 patients from the UK (Tate et al., 2005). The maximal doses of two AEDs (CBZ and PHT), prescribed during a routine clinical setting, were retrospectively associated with the genotypes. Although the retrospective design prevented the authors from establishing that the maximal doses reported were actually needed for seizure control, the presence of the splice site interrupting ( mutant ) A alleles was linearly correlated with a higher dose of either AED. AA homozygote patients were on average prescribed higher doses of CBZ or PHT than heterozygotes, and heterozygotes showed higher doses than wild-type GG homozygotes. A further drawback of this study was that no information was given how many patients were seizure-free or had side effects at the maximal doses reported. A second, smaller study by the same researchers in 71 Chinese patients failed to provide a similar association with prescribed PHT maintenance dosages (Tate et al., 2006). However in the same patients there was a marginally significant association with PHT serum levels. These results were therefore considered by the authors as a tentative confirmation of their initial findings, although in a strict statistical sense, the study did not provide a positive replication. As a further note of caution, it has to be added that a third association study has failed to detect similar results (Zimprich et al., 2008). In this study, the maintenance dosages of CBZ were analyzed in 369 patients with mainly pharmacoresistant focal epilepsies and did not correlate with the genotype. Further replication studies are still absent. And finally it has to be noted, that the magnitude of the observed effect in the initial study was on average modest 230 mg difference between the two homozygote genotypes for CBZ and 47 mg for PHT with a large overlapping range. The clinical relevance of this polymorphism in terms of its ability to predict an individuals dosage requirement is therefore still doubtful. However, it has to be kept in mind that even truly functional polymorphisms in single AED targets might be very hard to catch in clinical studies if the AED under investigation really works through multiple mechanisms. In the case of CBZ, several additional targets have been demonstrated, among them acetylcholine receptors, which also underlie genetic variations. Missense mutations in a4 subunits, for example, have been linked to a rare form of epilepsy, dominant nocturnal frontal lobe epilepsy (ADNFLE). Such mutated a4b2 acetylcholine receptor channels when reconstituted in Xenopus oocytes display a three-fold higher sensitivity to CBZ than wild-type channels (Picard et al., 1999). This was taken as a possible explanation why patients with ADNFLE respond better to CBZ than other AEDs (Picard et al., 1999; Bertrand et al., 2002). Such complementary modes of AED-actions might differ between individuals depending on their genetic background. Obviously, such a constellation would make it much more difficult to detect variations in individual

11 Pharmacogenetics in Epilepsy 11 Figure 4. Many pharmacogenetically relevant polymorphisms are expected in nonexonic regulatory regions of the DNA. The figure shows the example of an intronic SNP in the SCN1A gene (IVS5N+5 G>A; rs ) that, by affecting the splicing of the gene, is likely to alter the functional properties of the resulting Na + channel. Subtle differences in Na + channel properties might in turn result in different AED dosage requirements depending on the genotype (Tate et al., 2005; Heinzen et al., 2007). Epilepsia ILAE AED targets and would necessitate larger pharmacogenetic studies investigating the effects of several AED-target polymorphisms at the same time. Other AED targets, where functional polymorphisms could possibly influence the clinical treatment response, are GABA A receptors. On the one hand, GABA A receptors might themselves display pathogenic mutations or variations that determine the type of AED to be used. Animal models have taught us that the disruption of GABA A b3 subunits may lead to abnormally synchronized thalamocortical oscillations, which are considered the basis for absence seizures. Ethosuximide, a blocker of T-type calcium channels, has been shown to effectively dampen these pathological oscillations and is therefore a drug of choice for absence seizures (DeLorey & Olsen, 1999; Handforth et al., 2005). However, at present, the decision when to treat with ethosuximide is solely based on clinical experience and not on any genetic data. On the other hand, GABA A receptors also constitute the targets for benzodiazepines and other AEDs (Rogawski & Lçscher, 2004). In several studies it could be shown that the pharmacological properties of the final pentameric GABA A receptors depend critically on the combination of subunits of which there are more than 20 to choose from (Brooks-Kayal et al., 1998). In dentate gyrus granule cells of chronically epileptic rats a downregulation of a1 subunits and a concomitant upregulation of a4 subunits has been observed which is paralleled by a decreased sensitivity to the GABA A receptor agonist zolpidem (Brooks-Kayal et al., 1998). In human samples the subunit composition of GABA receptors has also been shown to differ markedly between TLE and control tissue (Loup et al., 2000). The differential expression of GABA A receptor subunits is very complex being dynamically regulated by the interaction of a multitude of transcription factors with an even wider range of cis-regulatory DNA elements (Steiger & Russek, 2004). Functional polymorphisms at expressionregulating DNA sites are known to be abundant and are thought to be main factors determining interindividual diversity in general (Wray, 2007). To what extent cisregulatory polymorphisms in GABAergic genes, like the recently discovered one for the GABA A receptor b3 subunit, are pharmacogenetically relevant remains to be determined (Urak et al., 2006). Clinical impact At present, we still know very little about the functionality of polymorphisms of drug targets and to what extent such variations have a clinically relevant impact on AED treatment. Drug metabolism and elimination Elimination of drugs including AEDs is accomplished by the hepatic (metabolism) and/or renal (excretion) route (Fig. 1). In general, drug metabolism represents the prominent pathway, both in qualitative and quantitative terms, which comprises so-called phase I (e.g., oxidative reactions catalyzed by various cytochrome P-450 enzymes) and phase II (e.g., conjugations like glucuronidation) reactions. Within the most important cytochrome P(CYP)-450