Abstract
Identifying genetic risk factors for highly heterogeneous disorders such as epilepsy remains challenging. Here we present, to our knowledge, the largest whole-exome sequencing study of epilepsy to date, with more than 54 000 human exomes, comprising 20 979 deeply phenotyped patients from multiple genetic ancestry groups with diverse epilepsy subtypes and 33 444 controls, to investigate rare variants that confer disease risk. These analyses implicate seven individual genes, three gene sets, and four copy number variants at exome-wide significance. Genes encoding ion channels show strong association with multiple epilepsy subtypes, including epileptic encephalopathies and generalized and focal epilepsies, whereas most other gene discoveries are subtype specific, highlighting distinct genetic contributions to different epilepsies. Combining results from rare single-nucleotide/short insertion and deletion variants, copy number variants, and common variants, we offer an expanded view of the genetic architecture of epilepsy, with growing evidence of convergence among different genetic risk loci on the same genes. Top candidate genes are enriched for roles in synaptic transmission and neuronal excitability, particularly postnatally and in the neocortex. We also identify shared rare variant risk between epilepsy and other neurodevelopmental disorders. Our data can be accessed via an interactive browser, hopefully facilitating diagnostic efforts and accelerating the development of follow-up studies.
Commentary
A decade has passed since the first reports of large-scale whole-exome sequencing for genetic variants in developmental and epileptic encephalopathy (DEE), and in subsequent years, the number of genes and confidence in them has steadily risen. With around 80-100 genes now firmly recognized as harboring variants that explain up to 40% of all DEE cases (plus or minus many genes, depending on the criteria applied), realistic hope emerges in the form of effective precision medicines and genetic therapies for this devastating and mostly intractable disease.
Still, this big leap forward for a Mendelian epilepsy remains in stark contrast to solving the genetic contribution to more common epilepsies, a noted frustration given that they have long been suspected to have significant heritability. 1 The explanation presumably lies with genetic complexity, a term that encompasses various factors including oligogenic or polygenic inheritance and likely interactions with environmental factors. It may also be true that current sequence analysis falls short, in the respect that underlying variants for common disease may be subtle regulatory alterations that are not detected in exome arrays, nor understood well enough to reliably predict the impact on gene expression even when captured by whole-genome sequencing.
Therefore, short of sequencing a half million or more epilepsy cases to increase power of detection, plus a magical advance in our ability to parse deep genomic space, one can understand why the way ahead thus far has been to increase sampling as much as funding agencies will allow and to compare & contrast DEE with other epilepsy types that are more reflective of common disease.
Enter the Epi25 Collaborative's 2024 Nature Neuroscience tome, which utilized almost 21 000 exomes from individuals with DEE, genetic generalized epilepsy, non-acquired focal epilepsy), and other epilepsy types and exhaustively explored the data space with sophisticated and robust data analysis. 2 In addition to assessing individual variants, the team also ran sets of genes (e.g., GABA receptors, MTOR pathway, ion channels) which, at the expense of granularity, provided en masse detection power including more opportunity to nominate convergent mechanisms based on the function of a gene set. While for rigor's sake Epi25's taut machines were tuned to examining the rarest variants, by combining an investigation of epilepsy type and demographics together with CNVs and common variants that were nominated in recent meta-analysis of GWAS studies, 3 a working model emerges whereby non-Mendelian genetic epilepsies arise from rare mutations that co-exist with common inherited variants that share function or expression space.
Despite the prodigious and sophisticated effort of Epi25 2025, evidenced quite elegantly in the paper, the results on the surface were almost all confirmatory—disappointing to those who expected new genes and mechanisms to arise from such a program. Only one new gene, ANKRD11, reached exome-wide significance. Similarly, there were no new gene sets that beat a high significance threshold. Another confirmation, which has been recognized for a while now, was the overlap with autism spectrum and neurodevelopmental disorder genes. Nevertheless, a real strength of the study was the transparent consideration of residual gene and gene set matches that lie just under rigorous significance threshold, for example, NSL and phosphodiesterase gene sets, implying that they may indeed reach certainty with additional sampling. Another real plus were the extremely thoughtful and clear data representations and accompanying tools and resources, including an excellent online front-end to a comprehensive database (https://epi25.broadinstitute.org).
The Collaborative launched several deeper dives, including in the “ion channels” gene set to analyze the relationship between the apparent genetic mechanism (e.g., loss-of-function vs missense) and the respective type of epilepsy. While some useful distinctions were made, such an approach cannot come close to resolving biochemical or physiologic subtleties that really do exist and are likely key factors in pathogenesis. For example, it is well known that significant partial or mixed biophysical features accompany missense mutations (e.g., see Thompson et al. 4 for SCN2A). Even putative loss-of-function alleles have degrees of loss, the level of which contribute to different numbers of subunit molecules that participate in a heteromultimeric channel, which in turn may prompt different threshold effects. There may be no substitute for examination of channel variant function at the goriest level and in physiologically relevant contexts to even have a clue of what is actually happening, at least until artificial intelligence begins to compete with traditional wet bench studies in this area. 5
All criticism aside of this otherwise tour de force effort, it is important to get to the bottom of gene discovery in heritable non-Mendelian epilepsy as this represents the vast majority of individuals with epilepsy. Epi25 2024 suggests that rare variant risk in epilepsy is far from saturation. The reader will have to decide whether or not this is the genetic understatement of the year, but at least for unexplained DEE cases, one can easily envision that much of the remaining signal comes from regulatory variants that are beyond exome detection, probably in the some of the same genes that harbor more obvious variants.
But the real issue is this. If most heritable epilepsy arises from complex interactions between rare variants and common alleles, how on earth do we get there? That is, what sample sizes are required to detect both individual effects and genetic interactions (which may be two-way, three-way, multi-way)? An alternate model is also plausible—that heritable non-Mendelian epilepsy is largely polygenic as might be predicted from initial assessments of polygenic risk for some forms of this disease, 6 whether or not the same genes have more severe variants that cause Mendelian epilepsy. Combinatorial genetic complexity has been observed—albeit molecularly unsolved—in epilepsy mouse models: e.g., strains that individually do not experience ready seizures, change their tune when their existing variants are mixed-up in genetic crosses.7,8 Why shouldn’t this be the case for human disease? And if it is mostly down to polygenic risk, then how many combinations and permutations are we talking about, and what are the sample sizes required for confident detection. Most importantly, will it be possible to identify discrete, rate-limiting functions, pathways, or genes that can be made actionable in the clinical setting? To move ahead the field seems to require another big leap forward, with brute force, super-sized sampling only being part of it.
Footnotes
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.