Meta-Analysis Revives Genome-Wide Association Studies in Epilepsy

Genetic Determinants of Common Epilepsies: A Meta-Analysis of Genome-Wide Association Studies.

International League Against Epilepsy Consortium on Complex Epilepsies. Lancet Neurol 2014;13:893–903.

BACKGROUND: The epilepsies are a clinically heterogeneous group of neurological disorders. Despite strong evidence for heritability, genome-wide association studies have had little success in identification of risk loci associated with epilepsy, probably because of relatively small sample sizes and insufficient power. We aimed to identify risk loci through meta-analyses of genome-wide association studies for all epilepsy and the two largest clinical subtypes (genetic generalised epilepsy and focal epilepsy). METHODS: We combined genome-wide association data from 12 cohorts of individuals with epilepsy and controls from population-based datasets. Controls were ethnically matched with cases. We phenotyped individuals with epilepsy into categories of genetic generalised epilepsy, focal epilepsy, or unclassified epilepsy. After standardised filtering for quality control and imputation to account for different genotyping platforms across sites, investigators at each site conducted a linear mixed-model association analysis for each dataset. Combining summary statistics, we conducted fixed-effects meta-analyses of all epilepsy, focal epilepsy, and genetic generalised epilepsy. We set the genome-wide significance threshold at p<1·66 × 10(−8). FINDINGS: We included 8696 cases and 26 157 controls in our analysis. Meta-analysis of the all-epilepsy cohort identified loci at 2q24.3 (p=8·71 × 10(−10)), implicating SCN1A, and at 4p15.1 (p=5·44 × 10(−9)), harbouring PCDH7, which encodes a protocadherin molecule not previously implicated in epilepsy. For the cohort of genetic generalised epilepsy, we noted a single signal at 2p16.1 (p=9·99 × 10(−9)), implicating VRK2 or FANCL. No single nucleotide polymorphism achieved genome-wide significance for focal epilepsy. INTERPRETATION: This meta-analysis describes a new locus not previously implicated in epilepsy and provides further evidence about the genetic architecture of these disorders, with the ultimate aim of assisting in disease classification and prognosis. The data suggest that specific loci can act pleiotropically raising risk for epilepsy broadly, or can have effects limited to a specific epilepsy subtype. Future genetic analyses might benefit from both lumping (ie, grouping of epilepsy types together) or splitting (ie, analysis of specific clinical subtypes).

Commentary

We have known for centuries that epilepsy can be caused by both acquired and genetic factors. Long-held hypotheses about the genetic basis of epilepsy were substantiated by the last century's twin studies, familial aggregation studies, recognition of large pedigrees with apparently monogenic epilepsy, and positional cloning efforts.1 With advances in genetic and genomic techniques at the beginning of this century came the hope that genome-wide association studies (GWAS) would unlock the genetic basis of many diseases, including epilepsy. This was based on the notion that common diseases, such as epilepsy, were not mediated only or even usually by inherited genetic factors that would segregate in Mendelian fashion but that genetic variability (measured as single nucleotide polymorphisms, representing naturally occurring common variants) at many loci could influence the risk of developing epilepsy. Unlike traditional Mendelian genetic disease studies that seek typically rare mutations with large effect size (dominant or recessively inherited disease), the goal of GWAS studies of complex epilepsy is to identify commonly occurring susceptibility loci that confer an increased risk of developing disease, often with effect sizes (reflected in odds ratios) that can be quite modest. Often, the associated regions can include more than one gene. The gene content of these loci may then be interrogated to reveal genes that are likely to play a role in mediating complex epilepsy risk.

Most attempts at GWAS in epilepsy were either not successful, or initially suggested that findings could not be reproduced.2 Prevailing explanations for why the field had not explained epilepsy by invoking common variants included the diversity of phenotypes inherent in the epilepsies (phenotypic heterogeneity) and lack of sufficient power to detect signals of small effect. But even notably well-powered, rigorous attempts to determine the genetic determinants of genetic generalized epilepsy, for example, did not yield evidence for the influence of common variants on this form of epilepsy.3 A single exception came just three years ago with the EPIGEN group's study, which identified four epilepsy-associated susceptibility loci.4

In an example of consortium science reaching a new level, a consortium of consortia came together through the International League Against Epilepsy (ILAE) Consortium on Complex Epilepsies to address the issue of common variants playing a role in epilepsy. Here, we feature their report, published in Lancet Neurology last year. The effort included the analysis of 8,696 individuals with epilepsy and more than 26,000 control individuals and took advantage of the phenotypic and genotypic data that had been collected over many years for 12 cohorts. To meaningfully combine the data from these cohorts, with genotype data at five sites, the investigators took care to make sure that they were indeed comparing “apples to apples,” ensuring that the same genotypes were included or inferred from all sites’ data before engaging all of the data in a meta-analysis.

The robustness of the statistical methods used in this study have already been commended in detail.5 The group harnessed phenotypic expertise in the ILAE Consortium to divide the nearly 9,000 individuals with non-acquired epilepsy into three groups—genetic generalized epilepsy (formerly called “idiopathic generalized epilepsy”), focal epilepsy, and unclassified epilepsy (for those without EEG evidence for generalized epilepsy or evidence for focal onset of seizures OR for those with features of both generalized and focal epilepsy). That alone is a feat not to be overlooked. Because GWAS analysis relies on a comparison of genotypes in the disease group versus a control group, the extent to which a control population is truly unaffected is also important—in this study, population-based datasets were used, including some populations who had been screened specifically for neurological diseases.

The results of the meta-analysis study included a locus containing an expected gene, giving the sense that this approach works, and some surprises, providing some potential new avenues for exploration. When all epilepsy was considered, there was genome-wide significance for an association between a locus on chromosome 2 (2q24.3) containing SCN1A, the most likely candidate gene in that region, as well as a locus on chromosome 4 (4p15.1) containing PCDH7, a protocadherin gene not previously known to have an association with epilepsy. The genetic generalized epilepsy subset consisted of 2,606 cases from 8 of the original 12 cohorts, analyzed with 18,990 controls. Analysis of this cohort revealed a significant associated between genetic generalized epilepsy and a locus at 2p16.1 that contains the genes FANCL, Fanconi anemia complementation group L, and VRK2, a serine-threonine kinase previously associated with epilepsy and schizophrenia. Interestingly, the authors note the allele that appears to confer risk for schizophrenia is actually the protective allele for epilepsy, adding yet another layer of complexity in the overlapping genetic architecture of neurodevelopmental disorders.

While there is a growing list of genes that explain monogenic forms of familial and sporadic focal epilepsy, this large meta-analysis did not detect any loci significantly associated with focal epilepsy. The reasons for this lack of association can only be speculated. It is possible, despite the impressive size of this study and the major phenotypic efforts that were undertaken, that some of the individuals included in the original cohorts had focal epilepsy that was in fact acquired from early life exposures (e.g., infectious or traumatic), thus reducing the power to detect genetic risk factors from among the focal epilepsy group. Another possibility, since the focal epilepsy group included patients with focal structural lesions, is that some portion of those individuals have somatic mutations that arose postzygotically, such that their risk for epilepsy is not determined by a haplotype that can be detected from DNA collected from leukocytes, the typical source for studies such as this one.

The ILAE study featured is likely to be as definitive as we will see for some time, until the next several thousand patients with epilepsy are sequenced and compared to the ever-growing available repositories of non-epilepsy controls available for analysis. While findings in GWAS do not translate immediately to the bedside in that predicting risk for an individual carrying an at-risk allele is not as simple as it is for dominant and recessive genotypes, scientifically, the ILAE GWAS meta-analysis represents a move toward a deeper understanding of the highly prevalent yet genetically elusive complex epilepsies.