INTRODUCTION
Epilepsy is a chronic disorder characterized by a predisposition to recurrent unprovoked seizures. Seizures are caused by abnormal neuronal discharges in the brain. Despite appropriate pharmacotherapy, about one third of newly treated epilepsy patients have refractoriness to medical treatment (Elger & Schmidt 2008). The most prominent hypothesis for explaining refractoriness to antiepileptic drugs (AEDs) in epilepsy is the multidrug-transporter hypothesis, which was first explored in chemotherapy-resistant cancer. Excess efflux of AEDs across the blood-brain barrier (BBB) caused by overexpression of efflux transporters suggests that alterations in entry and extrusion of drugs into and out of cells may be an important mechanism of pharmacoresistance (Schmidt & Löscher 2005). P-glycoprotein 170 (P-gp) is the most studied protein among the ATP-binding cassette (ABC) efflux transporters (Schinkel et al. 1994), encoded by ABC subfamily B member 1 transporter or multidrug resistance (ABCB1/ MDR1) gene. Evidence has shown that ABCB1/MDR1 expression was elevated in the epileptic foci of drug resistant epilepsy up to ten times indicating its main role in pharmacoresistance (Tishler et al. 1995; Gottesman et al. 2002).
Different responses to pharmacotherapy among patients with similar clinical phenotypes and taking the same doses of medication may be explained, in part, by single-nucleotide polymorphisms (SNPs) as the most frequent type of genetic variation (Schmidt & Löscher 2005). Numerous SNPs have been identified in ABCB1/MDR1 gene among which the C3435T variant is the most common polymorphism (Fung & Gottesman 2009). The key question of this review is “is the C3435T variant of the ABCB1/MDR1 gene associated with outcomes of AEDstreatment in epilepsy?” The initial pharmacogenetic study demonstrated an association of C3435T polymorphism with response to AEDs while the replication of this study has been controversial with consistent and conflicting studies. Herein, the sources of bias in association studies for the C3435T variant and refractoriness to antiepileptic drug (AED) therapy are briefly assessed.
P-gp Structure
P-gp is the first discovered human ABC transporter more than 30 years ago in drug-resistant ovarian cells obtained from Chinese hamsters (Juliano & Ling 1976). This protein is the product of the ABCB1/MDR1 gene, located at 7q21 and containing 28 exons with two transcription start sites. It is a phosphorylated and glycosylated trans-membrane protein consisting of approximately 1280 amino acids and with a molecular mass of around 170 kDa. P-gp contains two homologous halves in which a short linker region joining the symmetric sequences together (Figure 1). Each half is composed of one trans-membrane domain (TMD) containing six α-helix segments and one ATP-binding motif or ABC unit located at the cytoplasmic site. The two ATP-binding domains generate energy for interaction of drugs with drug-binding sites and for efflux transport function. P-gp mutations in TMDs 5, 6, 11 and 12 affect substrate specificity and indicate the presence of a drug-binding site in this region (Fung & Gottesman 2009).
P-gp function
P-gp functions as a transmembrane efflux pump, by moving drugs from the intracellular to the extracellular domain. It may also interact with drug molecules trapped within the cell membrane lipid bilayer. This protein is expressed in many normal tissues such as the liver, gastrointestinal tract, kidney, adrenal glands and the blood–brain barrier (BBB) (Löscher & Potschka 2005a). Anatomic localization and ability of P-gp in transporting a broad range of compounds suggest that this transporter acts as an effective cellular protector against toxic xenobiotics, drugs and metabolites that are its substrates. It does this by secreting these compounds into the bile, urine and intestinal lumen, thereby preventing their accumulation in critical organs such as the brain. Altered P-gp expression and activity may influence an individual's susceptibility to drugs (Löscher & Potschka 2005b). P-gp transports a broad spectrum of substrates in the plasma membrane ranging in size from 250 to almost 4000 Da including many classes of drugs such as calcium channel blockers, lipid-lowering statins, opioids, chemotherapeutic agents, HIV protease inhibitors, antibiotics, immunosuppressive agents and b-adrenergic antagonists. Most AEDs are planar lipophilic agents and thus substrates for the P-gp transporter (Kwan & Brodie 2005).
ABCB1/MDR1 polymorphisms
Any two unrelated humans have millions of genetic differences caused by insertion/deletions or SNPs. SNPs are the most abundant form of genetic variation. About 0.1% of an individual’s DNA sequences are different among which approximately 80% is represented by SNPs. SNPs may be the causal factor of about 40% of the variation of the most common diseases and 70% of schizophrenia (Goldstein & Cavalleri 2005). Do the ABCB1/MDR1 variants account for the inter-individual variability in the pharmacokinetics and pharmacodynamics of drugs? The level of P-gp expression is highly variable between different individuals. Inter-individual variability of P-gp activity might affect blood levels and drug distribution to the specific target compartment. Insufficient or excessive drug penetration could cause adverse side effects or inactivity of drugs, respectively (Löscher & Potschka 2005a). The ABCB1/MDR1 gene is highly polymorphic. Numerous SNPs have been found in the ABCB1/MDR1, the majority of which involve noncoding regions thus not affecting the P-gp amino acid sequence. More than 50 variants reside in the coding region which can possibly cause altered function (Fung & Gottesman 2009).
Allele frequency for most of the coding region SNPs is less than 8% in different ethnic populations with the exception of 3 SNPs in exons 12 (C1236CT), 21 (G 2677T/A), and 26 (C3435T) in ABCB1/MDR1 gene. Both synonymous C1236T and C3435T polymorphisms change a GGC codon to GGT (Gly412 Gly) and ATC to ATT (Ile1145Ile), respectively (Table 1). So, what is the role of C3435T polymorphism in pharmacoresistance in epilepsy? An original report suggested a strong and significant association (P=0.006) between the CC genotype at 3435 in ABCB1/MDR1 gene and drug-resistant epilepsy (Siddiqui et al. 2003). Zimprich et al. (2004) confirmed the results of the first report (p = 0.035), but Tan et al. (2004a), with the exact replication of the first study and almost twice the sample size, failed to confirm the original findings (Zimprich et al. 2004; Tan et al. 2004a). Furthermore, Leschziner et al. (2008) in a comprehensive genome wide approach failed to confirm the association of C3435T polymorphism and response to AEDs (Leschziner et al. 2008). In total, 26 studies attempted to examine this hypothesis in epilepsy, but only 8 found a positive association (Table 2). The studies are divided into two groups: positive and negative associations, some of which are replications of the initial positive study (Table 2). It is unclear why these reports have found such contradictory results. How can such contradictory results be interpreted? Is there any obvious effect of C3435T polymorphism on response to AEDs even if the phenotypes are almost similar? There may be several reasons for such discordant findings which are discussed below.
Publication bias
Authors and journal editors prefer to publish papers with positive findings which can lead to publication bias. Methodologic weaknesses and consequent bias or even robust study designs and analyses with significant positive results at the 5% level in association studies may lead to false positive data probably arising by chance (Cardon & Palmer 2003; Tan et al. 2004b). The initial positive association study between C3435T polymorphism and multidrug resistance in epilepsy (Siddiqui et al. 2003) is a publication bias and may be a false positive result. Hence 17 subsequent studies could not confirm the first finding by improving the statistical power through increasing sample size (Colhoun et al. 2003). To avoid such a false positive association report, it is recommended to publish the original study only with a= 0.005 or lower. Even at this significance level and 50% power, the chance of true positive findings is about 10%, and so the findings should be interpreted with caution. In association studies between C3435T polymorphism and multidrug resistance in epilepsy, most authors have pointed to low power due to small sample size as a limitation. The total sample size in the 26 association studies is 7352 (282.77 ± 31.29), ranging from 45 to 609 epilepsy cases (Table 2). In order to have a power of 80%, these studies require a large sample size or collaboration between multi-centres and countries (Otto 2004).
Definition of drug-responsiveness and drug-resistant epilepsy
There have been a number of definitions for drug-resistant epilepsy. Selection of the most appropriate definition for a particular study depends on the type of seizure and epilepsy syndrome and the purpose of using the definition. According to the National Association of Epilepsy Centres (NAEC), pharmacoresistance in epilepsy is defined when seizures do not come under control after 9 months of treatment under the care of a neurologist or when a patient does not become seizure-free for 12 months during long-term state-of-the-art treatment with several suitable AEDs at maximal tolerated doses. Surgery, as an alternative to pharmacotherapy for about 5–10% of newly diagnosed patients can be used to define drug resistance because it would be assumed that the seizures remained uncontrolled after treatment with various drugs (Elger 2003; Schmidt & Löscher 2005). Most association studies for C3435T and multidrug resistance epilepsy have divided patients into “drug-resistant” and “drug-responsive” groups. The definition of drug-responsiveness and drug-resistance varies amongst these studies. In summary, drug-responsiveness is defined as complete seizure freedom over three months to one year and/or more or 50% reduction or more in seizure frequency within one year, while drug-resistance is defined as having one or more seizures per month or four to 10 or more seizures in six months or one year, or less than 50% reduction of seizures per year taking at least two AEDs, and also surgery for seizure control (Table 2). Additionally, some studies have used healthy individuals instead of drug-responsive patients as controls (Kasperaviciute & Sisodiya 2009). Since the definition of drug-responsiveness and drug-resistance is different between studies, patients who are classified as drug resistant in some studies may be drug-responsive in others. Therefore such variation in patients’ classification may lead to wrong conclusions. To have reliable findings, there need to be an agreement on the definition of drug responsiveness and drug resistance in epilepsy.
AED-therapy
a. AEDs–P-gp interactions
There is discrepancy in the transporter hypothesis as a mechanism for AED-resistant epilepsy. Some reports suggest that P-gp plays a significant role in mediating resistance to AEDs. This includes drugs such as phenytoin (PHT), carbamazepine (CBZ), phenobarbitone (PHB), and lamotrigine (LTG) which, as P-gp substrates induce upregulation of P-gp expression in the brain. However, valproic acid (VPA) and vigabatrin do not produce the same effects (Sills et al. 2005). Therefore, based on this hypothesis, in some association studies for ABCB1/MDR1 gene polymorphism and response to CBZ, VPA or other drugs that may not be substrates for P-gp, this polymorphism probably may not play any role in determining drug resistance (Kimura et al. 2007) and the patients, classified as drug resistant, may not really be pharmacoresistant. However, in-vitro observations do not support this hypothesis as CBZ, PHT, and VPA (Baltes et al. 2007a, Baltes et al. 2007b, Rivers et al. 2008), possibly belong to a class of “weak substrates” of human P-gp and may compete for drug-binding sites on P-gp and inhibit the transport of other substrates (Luna-Tortós et al. 2008). Determination of which AEDs are real substrates of P-gp requires in vivo experiments.
b. AED–AED and AED-nonAED interactions
In patients who are on multiple drug therapy, drug interactions may occur in which one drug may modify the activity of another by enhancing or reducing its pharmacologic effects with beneficial or harmful outcomes. AED–AED or AEDs and non-AED interactions in polytherapy regimen can significantly influence the clinical management in patients with epilepsy, pharmacokinetically and pharmacodynamically. For example, phenobarbitone (PHB), PHT, and CBZ as enzyme inducers, can increase the level of activities of various cytochrome P450 and UGT isoenzymes. Therefore, a combination of PHB, PHT, and CBZ with VPA, LTG, topiramate, and tiagabine in polytherapy increases and decreases their metabolism and half-lives, respectively (Patsalos et al. 2002). In association studies for C3435T polymorphism and refractory epilepsy, some authors have not differentiated between monotherapy and polytherapy treatments which could have influenced the results of their study. To prevent such a problem, the possible effect of ABCB1/MDR1 polymorphism on refractoriness to each AED monotherapy must be established.
Haplotype
SNPs are often associated with particular variants existing on the same chromosome known as a ‘haplotype’ which may play a role in response to drugs and disease susceptibility (Goldstein & Cavalleri 2005). Such a non-random association of SNPs is called linkage disequilibrium (LD) (Goldstein & Cavalleri 2005). Some reports have demonstrated significant LD between synonymous 3435C>T and 1236C>T and nonsynonymous 2677G>T/A (Ala893Ser or Ala893Thr), suggesting that C3435T and 2677G>T/A may influence P-gp function in epilepsy patients in response to AEDs (Kim et al. 2001). In association studies for C3435T polymorphism and refractory epilepsy with negative results, some authors have suggested that since the C3435T variant is silent, its effect is possibly caused by specific haplotypes (CGC, TGC and TTT) of C1236T and G2677T/A (Hung et al. 2005). It was hypothesized that the presence of the C3435T variant within a haplotype block can affect the timing of co-translational folding and P-gp structure which could influence its insertion into the membrane and subsequently change the structure of substrate and inhibitor interaction sites (Kimchi-Sarfaty et al. 2007). Haplotypes may potentially improve the statistical power under certain circumstances. However, the debate still exists whether SNPs or haplotypes are more appropriate for genetic association studies (Tan et al. 2004b). Finally, many genes including drug-transporters,drug-metabolizing enzymes, and drug targets are responsible for drug responsiveness in epilepsy (Löscher et al. 2008) Therefore, one need to investigate the entire genetic effect of these genes and their SNPs together, rather than study a particular variant or haplotype in a single gene.
Population heterogeneity
a. Epilepsy types
Epilepsy is heterogenous in nature, influenced by different genes and environment. Drug resistance differ amongst patients with different types of seizures and epilepsy syndromes, and the population (Duncan et al. 2006). A study on Finnish epilepsy cases performed by Sillanpää et al. (1999) showed that pharmacoresistance to AEDs was seen inabout 78% of patients with symptomatic partial or symptomatic generalised epilepsies and 13% of patients with idiopathic generalised epilepsies but not in patients with idiopathic partial epilepsies (Sillanpää et al. 1999). Such a high frequency of drug resistance in symptomatic epilepsy (for example structural brain lesions) could be caused by overexpression of P-gp due to the effect of frequent seizures or AEDs therapy (Löscher 2005c). This will release excitotoxic metabolites from the altered tissue and inhibit the efficacy of AEDs for treatment, irrespective of the C3434T gene polymorphism involvement (Schmidt & Löscher 2005). Most association studies for C3434T gene polymorphism and refractoriness have been conducted on either a specific subset of epilepsy types/syndromes or all the various types of epilepsies which could therefore confound the results. To avoid such a bias, the relationship of pharmacoresistance with C3434T gene polymorphism should be separately calculated for different epilepsy types including idiopathic, symptomatic and cryptogenic.
b. Ethnicity effect
The distribution of allele and haplotype frequency is different among populations, reflected by differing ancestral histories. This difference may be caused by responses to natural selection, migration patterns, stochastic effects and founder events. Therefore, different ethnic groups may have various allele frequencies in both disease and non-disease genes (Cardon & Palmer 2003). Genotyping of the C3435T polymorphism in different normal populations have shown that the C-variant is more frequent than T-variant and is the dominant ancestral allele, carried by more than 74% Africans (Figure 2) (Ameyaw et al. 2001). The wild haplotype (CGC) frequency in South American and African populations is more than the mutated type (TTT) while the opposite is true for the Asians (Fung & Gottesman 2009). The prevalence of the C-variant in the initial report was more in the drug-resistant than the drug-responsive Caucasian epilepsy patients (Siddiqui et al. 2003). However, in most of the studies with Asian cases (Seo et al. 2006; Kwan et al. 2007; Kim et al. 2008a, Kim et al. 2008b, Chen et al. 2007; Lakhan et al. 2009; Vahab et al. 2009; Szoeke et al. 2009) the opposite was seen, which was similar to the normal population (Figure 2), probably reflecting the complex role of C3435T polymorphism in the response of AEDs in different ethnic populations. In summary, based on the C allele and LD differences between various epilepsy and normal populations, inclusion of patients with various ethnicities in the association studies for C3435T variant and refractoriness in epilepsy can potentially introduce heterogeneity which would confound the findings (Tan et al. 2004a, Shahwan et al. 2007).
CONCLUSION
Despite more than 26 association studies in epilepsy involving the common ABCB1/MDR1 C3435T variant and refractoriness, 16 of the replicated studies failed to confirm the original report. This review has provided an opportunity for better understanding of the methodological problems in these association studies by highlighting five sources of bias which may confound the results. First, publication bias which can be caused by false positive results, found by chance in the initial report as a result of insufficient sample size, and low statistical power. Second, varying definitions of drug-responsive and drug-resistant epilepsy in the different studies could lead to different results. Third, for some AEDs, it is not clear whether they are substrates of P-gp; hence justification of the results is difficult. Moreover, the interaction between AED–AED and AED-nonAED in polytherapy regimen can influence response to AEDs. Fourth, certain haplotypes, formed by C3435T with specific variants in ABCB1/MDR1 or other genes and SNPs may affect the response to AEDs. Finally, the studies may involve a heterogeneous population made up of combinations of various cases with different epilepsy syndromes or ethnicities. In order to obtain reliable results in the genetic association studies for C3435T polymorphism and refractoriness, these factors need to be controlled.