Epilepsy is one of the most common neurological diseases worldwide and is considered a major public health problem (1,2). The International League Against Epilepsy (ILAE) has established that the term ‘epilepsy’ refers to a disease of the brain that meets any of the following conditions: i) At least two non-induced seizures, or reflexes, that occur ≥24 h apart; ii) one non-induced seizure, or reflex, and a risk of further seizures similar to the general recurrence risk (≥60%) following two non-induced seizures that occur over the next 10 years; or iii) the diagnosis of an epilepsy syndrome (3,4). Epilepsy is considered to be resolved when an individual with epilepsy has remained seizure-free for 10 years and without antiepileptic drug treatment for ≥5 years (4).

For the most accurate study of epilepsy, the ILAE (5,6) has organized and classified seizures and several epilepsy types as focal, generalized and of unknown onset, based on certain characteristics, including seizure type, electroencephalography (EEG) features, imaging studies, age-related features and triggering factors, such as comorbidities and prognosis (6,7); this classification involved the work of epileptologists, neurophysiologists and epilepsy researchers (6).

A previous study demonstrated that epileptic seizures are associated with several mechanisms that involve the glutamate excitotoxicity process, microglial activation, mitochondrial dysfunction, degenerative processes, and the presence of reactive oxygen species and oxidative stress (8). In addition, it has previously been reported that certain regulatory processes are involved at the transcriptional level; for example, the nuclear transcription factor erythroid-derived 2-like 2 was revealed to serve a role in epileptic seizures (9,10). A recent study conducted by our group used microarray analysis in children with epilepsy to demonstrate that those with epilepsy overexpressed genes that were related to the transcriptional factor cAMP-response element binding protein (CREB) compared with normal children, in addition to significantly altered expression levels of genes involved in energy metabolism, redox balance and the immune response (11). The differential gene expression, particularly genes related to CREB, observed in children before and after the administration of valproic acid indicated that the activity of antiepileptic drugs (AED) is dependent on target genes. These data suggested a role for genetics in epilepsy development and highlighted the importance of studying the genetic mechanisms associated with drug resistance. This would provide an improved understanding of the impact of pharmacological treatment on epilepsy and in the patient's daily activities, of which both are influenced by: i) The patient's response to treatment; ii) the relationship between the number and type of seizures, and the modified transportation of proteins or their receptors due to the drug's activity and the presence of genetic variations; iii) the possible influence of the phenotypic characteristics of the patient in response to the treatment; iv) the impact of the presence of the genetic variants in the functionality of the transporting proteins and AED target proteins; v) the interference of the potential alterations in the target protein on the mechanism of action of the AED due to the presence of the variant; vi) the influence of the inflammatory and immunological response; vii) the predisposition to some of the different aspects of epilepsy, including refractoriness or decreased sensitivity to the AED effect; and viii) the result following the combination of several of these or other mechanisms (12–15). In accordance with the last point, the objective of the present review was to focus on describing the findings of genetic alterations involved in pharmacoresistant epilepsy.

The term pharmacoresistant epilepsy refers to a type of epilepsy that does not respond to at least two AEDs, which were chosen and used in monotherapy or combination therapy (bi- or polytherapy) and fail to fully control seizures for an adequate period (16). In 2011, the ILAE proposed to standardize the definition of pharmacoresistant epilepsy as the presence of seizures in a period of 6 months, even under proper therapeutic regimens (either monotherapy or in combination) (16). Two studies performed in 2000 and 2012 reported that people with epilepsy (PWE) responded differently to AED treatment, since 47–49.5% of the patients required one AED to control the seizures, 13–13.3% required a second AED, and 3.7–4% needed a third AED, which was administered either alone or in combination (17,18). To summarize, Kwan and Brodie (17) observed that in a prospective study of 525 PWE (age, 9–93 years), 333 of them (63%) remained seizure-free during AED administration and seizures that did occur were more persistent in patients with symptomatic and cryptogenic epilepsy (40%) compared with those with idiopathic epilepsy (26%). Moreover, among 470 previously untreated patients, 222 of them (47%) became seizure-free during treatment with their first AED, 67 patients (14%) became seizure-free during treatment with a second or third AED and 12 patients (3%) were controlled with two AEDs administered together (17). Brodie et al (18) subsequently discovered that patients have differential responses to AEDs; out of 1,098 PWE (ages, 9–93 years) studied, 749 of them (68%) were seizure-free with AED monotherapy, but in 272 patients (25%), freedom from seizures was never attained. Moreover, <2% of patients became seizure-free with the use of up to six or seven AEDs (18). In addition to the above findings, epidemiological studies conducted among children and adults have discovered that 20–40% of PWE present with pharmacoresistant epilepsy (19–22), which negatively impacts their quality of life because it is also associated with increases in psychiatric comorbidities and the risk of premature death and social discrimination (23).

It is important to highlight that pharmacoresistant epilepsy may also cause serious socioeconomic problems. For instance, Argumosa and Herranz (24) evaluated the economic cost of controlled and uncontrolled epilepsy in Spain (participants were <14 years old) and reported that the mean annual cost of controlled epilepsy was $2,002.36 USD, whereas the cost of uncontrolled epilepsy was $5,348.50 USD; thus, uncontrolled epilepsy was 2.7 times more expensive compared with controlled epilepsy. Given the elevated costs of treatment, alternative therapeutic approaches, such as the ketogenic diet (25), high doses of steroids (26) and brain surgery (27) have all been implemented The ketogenic diet has proved beneficial in PWE in which pharmacological and/or surgical treatment is not effective; this diet is centered around a very high-fat and low-carbohydrate intake, reducing carbohydrates to <10% of used energy (90% of the total caloric intake comes from fat, 6% from protein and 4% from carbohydrates). These proportions trigger a systemic shift from glucose metabolism towards the metabolism of fatty acids, which yields ketone bodies that serve as the energy source to replace glucose in the brain (28). In patients with pharmacoresistant epilepsy, a ketogenic diet has been observed to improve the quality of life and decreases seizure frequency in ~30% of patients (25). In a significant number of patients with pharmacoresistant epilepsy, curative epilepsy surgery cannot be offered since there are multiple epileptogenic zones; for these patients, neurostimulation techniques, such as vagus nerve stimulation, deep brain stimulation and responsive neurostimulation, are viable treatment options that should be considered in every patient with this type of epilepsy that is unsuitable for surgery (27,29). These techniques provide palliative care, resulting in a 10–80% reduction in seizure occurrence (29). Furthermore, if all the aforementioned treatment approaches fail to control the seizures, cannabidiol (Epidiolex^®) can be prescribed; this is a pharmaceutical product approved by the U.S. Food and Drug Administration that consists of 99% cannabidiol derived from cannabis (30,31). A previous study reported that out of 43 Mexican children with pharmacoresistant epilepsy, a decrease in the number of epileptic seizures were observed in 81.3% of patients and 20.9% patients displayed a reduction in the number of AEDs with the use of cannabidiol. Significant adverse effects related to appetite or sleep were only observed in 42% of patients following the use of cannabidiol (31).

Previous studies have suggested that pharmacokinetic and pharmacodynamic mechanisms form the physiopathological basis of pharmacoresistant epilepsy (Table I) (22,32,33). Advances in genomic technologies have facilitated the genome-wide discovery of common and rare variants and have increased our understanding of genetics in epilepsy (34); however, the mechanisms underlying pharmacological resistance have not been fully elucidated (35). Some of the most important genes associated with the physiopathology of epilepsy were associated with the neuronal acetylcholine receptor, neuronal potassium channels (KCNs), voltage-dependent sodium channels (SCNs), calcium channels and γ-aminobutyric acid (GABA) receptors (32,36,37). Table II summarizes some of the genetic variants that occur in the main genes linked to epilepsy from recent studies.

In the particular case of pharmacoresistant epilepsy, the most frequently studied polymorphisms are those associated with multidrug resistance genes (MDR): ATP-binding cassette subfamily B member 1 (ABCB1 or MDR1) and ATP-binding cassette subfamily C member 2 (ABCC2 or MRP2); SCN α subunits 1, 2 and 3 (SCN1, SCN2 and SCN3); and metabolizers of endogenous and xenobiotic substances, cytochromes P450 families 2 and 3 (CYP2 and CYP3). Additional details related to the findings of these studies are found in Table III.

Currently, a patient's medical history and EEG results are used to diagnose the type of seizure, but they must be interpreted with caution so that diagnostic errors are not made; this suggests the use of complementary studies (38). Emerging genomic technologies, high-throughput screening and chip technologies have accelerated our understanding of the genetic makeup of epilepsy (34); for instance, the identification of mutations or polymorphisms in specific genes, such as those that encode ion channels, aforementioned, that are mainly expressed in brain neurons, specific neurotransmitter receptors and molecules that have functions in intercellular communication (39). For example, a study by Wang et al (40) investigated the genetic etiology of epilepsy in a cohort of 120 children with unexplained epilepsy using whole-exome sequencing (WES); it was found that the pathogenic variant c.1174G>A in the KCN subfamily D member 3 (KCND3) gene may be responsible for a broader phenotypic spectrum than was previously thought, including infantile epileptic encephalopathy. In addition, this study discovered that the glutamate receptor, ionotropic glutamate ionotropic receptor NMDA-type subunit 1 (GRIN1) and hyperpolarization-activated cyclic nucleotide-gated potassium channel 1 (HCN1), were candidate gene variants (c.2530C>T and c.1138A>T for GRIN1 and HCN1, respectively) for the Dravet and Dravet-like phenotypes (40). In intractable epilepsy and other mental disabilities, WES identified de novo variants in the Bernardinelli-Seip congenital lipodystrophy 2 (BSCL2) gene in two patients (41), of which one of the variants (c.985C>T) has been observed in other populations of epilepsy and developmental regression, regressive autism spectrum disorder, motor stereotypies, lower limb hypertonia and frontal lobe syndrome (41). As it was discovered that BSCL2 serves a role in neuronal function, it was suggested to be a potential candidate gene for epileptogenesis (41).

In pyridoxine-dependent epilepsy (PDE), despite seizure control, ≥75% of patients experience intellectual disability and developmental delay, which emphasizes the importance of early diagnosis. Genetic tests are increasingly being used as first-level tests for epileptic encephalopathies, which aim to provide a general description of the mutations in the aldehyde dehydrogenase 7 family member A1 that causes PDE (42). Epilepsy with myoclonic-atonic seizures (EMAS) accounts for 1–2% of all childhood onset epilepsies (43). EMAS has been demonstrated to have an underlying genetic component and several genes have been associated with this disease, such as SCN1A, SCN2A, CHD2, STX1B, SLC2A1, SLC6A1, POLG1, NRXN1, PIGN, CSNK2A1, GABRG2 and GABRB3; however, the genetic basis for this disorder remains unknown and the diagnostic potential of genetic tests remains low. This could be explained by the lack of several of the genes that may be associated with EMAS in the most commonly ordered epilepsy panels, although some have recently been added (43–47). Furthermore, a study conducted by Ortega-Moreno et al (48) analyzed a multigenic panel of 87 PWE and developmental delay, including classified and unclassified epileptic encephalopathies, epileptic spasms, severe myoclonic epilepsy of infancy, Lennox-Gastaut and Landau-Kleffner syndromes; they found mutations in various genes, such as potassium voltage-gated channel subfamily Q member 2 (KCNQ2), syntaxin binding protein 1, UDP-N-acetylglucosaminyltransferase subunit, cyclin dependent kinase-like 5 (CDKL5), protocadherin 19, SCN1A, CHD2, SLC2A1, synaptic Ras GTPase activating protein 1, aristaless related homeobox, DNA polymerase gamma, catalytic subunit and GRIN1. Although a high proportion of these patients had unclassified epilepsies, the results supported the use of the multigene epilepsy panel because it offered rapid testing with a good diagnostic efficiency (48).

It is hypothesized that genetic variants may also contribute to the efficacy of drug treatments for epilepsy; for example, adverse or toxic reactions, teratogenic risk in pregnancy, as well as long-term outcomes have been observed among PWE (49–63). Consistent with the findings in numerous other disorders with complex genetic backgrounds, the associated genetic variants that have been successfully verified are limited. Nevertheless, it is likely that new techniques and improved research approaches will increase this number in the near future (49). In recent studies, the association between genetic polymorphisms, treatment responses in epilepsy and AEDs reactions (toxic, adverse or those related with its efficacy) have been investigated; it was reported that polymorphisms in the human leukocyte antigen (HLA) gene were associated with severe cutaneous adverse AED reactions (50), and polymorphisms in a number of other genes, including ABCB1, ABCC2, GABRA6, GABRG2, CYP2C9, CYP3A4, UDP-glucuronosyltransferase (UGT)1A1, UGT1A4, UGT1A6, UGT2B7, SCN2A and SCN1A, have been associated with the concentration, response and efficacy of some of the most commonly used AEDs in clinical practice, including carbamazepine, oxcarbazepine, phenytoin, lamotrigine and valproic acid (51–63). Esmaeilzadeh et al (50) reported an association between HLA polymorphisms and severe cutaneous adverse reactions (SCARs) induced by drugs; in this study, 61 patients with diverse SCARs were recruited, and it was found that the hypersensitivity to different AEDs, including phenytoin, carbamazepine, valproic acid, topiramate and lamotrigine, was associated with HLA-A gene polymorphisms. Berghuis et al (64) recruited 1,328 adult PWE who had received oxcarbazepine (n=1,031) and carbamazepine (n=297) and performed a genome-wide association study to demonstrate the association between genetic factors and sodium levels and AED metabolism. The authors did not observe significant associations between sodium levels and other clinical variables, but in relation to carbamazepine metabolism, they observed a significant association with the intronic rs2234922 polymorphism in the epoxide hydrolase 1 gene. Furthermore, the same authors reported in 2017 that carbamazepine and oxcarbazepine induced hyponatremia in those with epilepsy (65); in this study, 1,782 adult patients who carbamazepine (n=1,424) and oxcarbazepine (n=358) were recruited; using an electronic database designed for pharmacogenomics studies, it was found that sodium levels were significantly associated with serum levels of carbamazepine (P<0.001) and oxcarbazepine (P=0.001), whilst age, sex and the number of concomitantly used AEDs did not influence this association. Serum levels of carbamazepine [Odds Ratio (OR)=1.2; 95% CI=1.12–1.28; P<0.001] and oxcarbazepine (OR=1.06; 95% CI=1.02–1.1; P=0.001) were significantly associated with hyponatremia. The co-treatment and the sequential use of carbamazepine and oxcarbazepine were also related to severe hyponatremia (65). McCormack et al (66) recruited patients with maculopapular exanthema (MPE) associated with AED use (all aromatic AEDs: n=259 European and n=116 Han Chinese patients; carbamazepine: n=95 European and n=85 Han Chinese patients; lamotrigine: n=118 European and n=16 Han Chinese patients; phenytoin: n=52 European and n=22 Han Chinese patients) and 1,321 controls and performed a genome-wide association to analyze the association between AED use and MPE. It was found that within the European population, variations in HLA-A*31:01 were significantly associated with carbamazepine-induced MPE (OR=5.5; 95% CI=3–10; P=1.47×10⁻¹⁰). Regarding phenytoin use, a significant association was identified between the rs78239784 polymorphism and an intronic variant of the complement factor H-related 4 gene in the European population (OR=8.8; 95% CI=4–19; P=2.94×10⁻¹⁰) (66). Bai et al (67) observed that VPA induced obesity in PWE, as following the recruitment of 225 Chinese Han patients with epilepsy receiving VPA treatment, 19.6% were found to be obese. The authors also found genotypic associations of rs1194197 in the CD36 gene and rs10865710 in the peroxisome proliferator-activated receptor γ gene following weight gain. In another study conducted by Li et al (68), associations between the rs1137101 polymorphism in the leptin receptor (P<0.001), the rs1800497 polymorphism in the ankyrin repeat kinase domain-containing 1 (P=0.017) and the rs10789038 polymorphism in AMP protein kinase (P=0.02) with valproic acid-induced weight gain were observed in 212 PWE (68). Wang et al (69) discovered that some polymorphisms were associated with the adverse effects of valproic acid in Chinese PWE by direct sequencing; following the recruitment of 254 Chinese PWE that received valproic acid monotherapy, a correlation was identified between CYP2C9 and acyl-coenzyme A synthetase 2A (ACSM2A) gene polymorphisms with serum alanine aminotransferase and aspartate aminotransferase levels (P<0.03) indicating that these gene polymorphisms can be used to identify liver dysfunction (69).

The therapeutic effect of valproic acid among children with focal epilepsy (89 children) was also studied, and the results identified 66 single nucleotide polymorphisms (SNPs) that were involved in the metabolism and transport of valproic acid target receptors (54); however, among the children with focal seizures, the selected genetic polymorphisms were not significantly associated with the response to valproic acid. Nonetheless, three variants of GABRA6 (rs9313892, rs4921195 and rs424740) demonstrated the potential to be associated with the response to valproic acid (54). In addition, although polymorphisms in the SCN1A gene are thought to influence the efficacy of carbamazepine and phenytoin, Manna et al (70) found that the rs3812718 variant in SCN1A was not associated with the response to carbamazepine in patients with focal epilepsy. It has been reported that disorders related to changes in the KCNQ2 gene included both benign seizure disorders and early onset epileptic encephalopathies (EOEE), especially the latter, which includes patients who present refractory seizures following standard AED treatment and development delay (71–73). Kuersten et al (74) conducted a systematic review (52 studies including data from 217 patients), in which they analyzed AEDs in KCNQ2-related epilepsies; it was discovered that seizures associated with KCNs could be controlled upon treatment with carbamazepine, lamotrigine, oxcarbazepine, phenytoin, valproic acid, levetiracetam, topiramate, phenobarbital, piracetam, vigabatrin, clonazepam, diazepam and midazolam in patients with benign infantile or neonatal seizures (n=133 patients, including 74 who were seizure-free with monotherapy, 4 who were unsuccessfully treated with monotherapy, 11 who were seizure-free with polytherapy, 4 patients with no response to polytherapy or any treatmet and 40 showed seizure cessation spontaneously without AEDs). Moreover, the results also demonstrated that moderate control of seizures was achieved with the use of carbamazepine, oxcarbazepine, lamotrigine, lacosamide, phenytoin, phenobarbital, valproic acid, topiramate, levetiracetam, retigabine, zonisamide, sultiame, ethosuximide, acetazolamide, clonazepam, diazepam, clobazam, nitrazepam and midazolam in patients with EOEE (n=84, including 48 who were seizure-free with monotherapy, 12 that did not respond to monotherapy, 20 who were seizure-free following polytherapy and 4 patients exhibited seizure reduction without AED). Phenobarbital was the most common prescribed monotherapy in the majority of patients with benign seizures and EOEE (n=65 in benign seizures and n=35 in patients with EOEE); however, the use of sodium channel blockers, such as carbamazepine, lamotrigine, oxcarbazepine and phenytoin, led to seizure cessation in the majority of patients with benign seizures and EOEE (n=21 benign seizures and n=45 with EOEE). With regards to the genetics, 25.6 and 67.9% of patients with benign seizures and EOEE, respectively, were reported to have de novo mutations, including missense, frameshift, splice site, deletion and truncation mutations. However, sparse systemic data are available on the response of treatment in KCNQ2-related epilepsy in larger cohorts (74), which limit our ability to comment on the efficacy of personalized medicine approaches to treat the large number of newly discovered genetic channelopathies.

The effects of SNPs in three KCN genes, including KCNA1 (rs112561866, rs2227910 and rs7974459), KCNA2 (rs3887820) and KCNV2 (rs7029012, rs10967705 and rs10967728), and their association with the susceptibility to epilepsy and their ability to respond to AEDs (carbamazepine for partial epilepsy and valproic acid for generalized epilepsy) was analyzed in a pharmacogenetic cohort of 595 patients (75). The results suggested that KCNA1, KCNA2 and KCNV2 did not influence the susceptibility of the disease or the capacity to respond to drugs (75). Mutations in the SCN2A gene have also been associated with neonatal seizures and a wide number of epileptic syndromes (76,77). Recently, an association between rs17183814 in SCN2A and the function of oxcarbazepine was demonstrated in a cohort of 218 patients; the results indicated that the presence of the SNP was associated with higher oxcarbazepine maintenance doses in patients with lower body weights and lower oxcarbazepine maintenance doses in patients who were overweight (51). In a cohort of 201 patients treated with valproic acid, an association was reported between the presence of the SNP rs230416 in SCN2A and the response to the drug (78). Similarly, another study reported an association between the SNPs in SCN2A (rs2304016) and SCN3B (rs3851100) and AED responsiveness in a cohort of 595 patients who were treated with valproic acid for generalized epilepsy and carbamazepine for partial epilepsy; the results demonstrated that none of these SNPs were significantly associated with a response to the AED (79).

Previous studies have indicated that to improve the efficacy and safety of epilepsy treatments, it is necessary to conduct studies to identify the following: i) The possible influence of some of the phenotypic characteristics of patients in response to the treatments; ii) the relationship between the number and type of seizures; iii) the differences in drug transporting protein activity or receptor activity caused by the genetic variations; iii) the impact of the presence of a genetic variant on the functionality of transporting proteins and AED target proteins; iv) the effects of potential alterations in target proteins, owing to a genetic variant, in the mechanisms of action of an AED; and v) factors that may contribute to susceptibility and treatment outcomes.

Personalized medicine is treatment of patients with therapy aimed at targeting their specific pathophysiology; however, currently, this has limited applications in clinical practice. Advancements in genetic epilepsy models and deep phenotyping techniques have the potential to revolutionize translational research, and will bring precision medicine to the forefront of clinical practice (80). In a needs assessment aimed at identifying the clinical challenges faced by physicians in diagnosing and treating children with epilepsy in Germany, Spain and the United States, it was reported that the main challenges were the application of guidelines in clinical practice, the identification of epilepsy and epileptic events, the integration of genetic tests into clinical practice, the integration of non-pharmacological treatments, the transition from pediatric to adult care and the participation and commitment of caregivers (80). These findings may support neuropediatricians who wish to specialize in epileptology to address these identified challenges based through precision medicine treatments (80). As we continue to gain an improved understanding of the true complexity underlying the physiopathology of genetic epilepsy and the identification of factors that are involved in phenotypic variations, it will be easier to address and understand genotype-phenotype correlations (81).

Previous studies on genetic epilepsy syndromes have provided insight into the mechanisms of epileptogenesis, and have suggested roles for a number of genes with different functions, including ion channel proteins and those associated with the synaptic vesicle cycle and energy metabolism (82). In addition, advanced genomic technologies, high-throughput sequencing and molecular diagnostics are increasingly improving our understanding of the genetic architecture in epilepsy, and molecular confirmation may influence the treatment prescribed for some monogenic epilepsies. Moreover, it is of crucial importance that genetic methods that are able to analyze all known genes at a reasonable cost be developed to discover novel therapeutic options and to implement individualized precision medical treatment regimens (82).

Genetic variations in the most common genes that encode channels, transporters, drug-metabolizing enzymes and receptors have been discussed in this Review with regards to their association with drug-resistant epilepsy, including: Sodium voltage-gated channels SCN1A, SCN2A, SCN3A, SCN8A and SCN1B); potassium voltage-gated channels (KCNA1, KCNA2, KCNB1, KCND7, KCNH5, KCNJ10, KCNQ2 and KCNT1); calcium voltage-gated channel subunit α1 H; ATP binding-cassette transporters (ABCB1, ABCC2, ABCC5 and ABCG2); mitochondrial transporter family members (SLC2A1, SLC6A1, SLC6A4, SLC6A11, SLC9A6, SLC25A22 and SLC35A2); drug-metabolizing enzymes (CYP2C1, CYP2C9, CYP2C19, CYP2D6, CYPP3A4 and CYP3A5); CDKL5; and GABA receptors (GABRA1 and GABBR1). These data may prove useful for future studies of drug resistance in epilepsy and may contribute to the generation of new diagnostic methods. In addition, these methods could subsequently support the development of improved treatment regimens, including novel pharmacological targets and pharmacological therapeutics. Overall, these findings may also improve the application of more personalized therapies, which would lead to the reduction in treatment and medical care costs, and increase the quality of life for patients and caregivers.