Epilepsiome. Two weeks ago, we had a few teleconferences about what the purpose of our gene curation and gene review efforts should be. Should we be a blog, should we be OMIM, or should we be a new neuromuscular homepage? None of these resources feel quite right for the questions that we have in epilepsy genetics. While thinking about this question, it became clear to me that we need to address five different aspects of each epilepsy gene, a framework that I’ll refer to as the 5-P concept. What is this concept about? Follow me onto a journey into the essential dimensions of interpreting the relevance of a variant in a clinical context.
Framework. When we discussed the purpose of what we call the Epilepsiome project, we felt that our project should help you answer the following question: “My patient has a variant in Gene X – what does this mean?” Trying to zero in on the particular approach that we want to take, we felt that this would be the real question that clinicians and researchers approach us with. OMIM provides us with information about genes, ClinVar curates all the variants. And we will be trying to provide you with information about what it may mean in a clinical context when a variant in a particular gene is found in your patient. Discussing this idea with people in field, it became immediately clear that there are a few issues that we would need to address.
5 Ps. Our Epilepsiome effort likes to answer the question what it means if you find a variant in a given gene, trying to synthesize the available information that is out there. This is information that you will not get from gene-based reviews like OMIM or variant-based databases like HGMD or ClinVar. Here are the 5-Ps that will help you understand what we are currently thinking about. This acronym stands for (1) Pathogenicity of the variant, (2) Parental information, (3) Previously seen variants, (4) Phenotypic information, and (5) Precision medicine approaches. In order to interpret the relevance of variants appropriately in a clinical context, these are the issues that you need to be aware of.
1 – Pathogenicity. You have a variant in an epilepsy gene. So where do you go from there. The first step in our “5P strategy” is to look at the prediction of this variant that either your lab provides or that you can get yourself by looking at resources like annovar or other annotation programs. Let’s get one thing straight from the beginning: when labs tell you about a “pathogenic variant”, it is an assessment within the overall framework of how a particular variant is evaluated. It is worth noting that labs use established criteria for assessing the pathogenicity of variants. While there may be some discrepancies between labs, this is a standardized process and not a gut call most of the time. Most labs have variant assessment criteria based on the ACMG guidelines. The determined pathogenicity of a variant doesn’t necessarily mean anything for your patient – you can carry a pathogenic variant and be unaffected and a seemingly non-pathogenic variant may be the cause of your patient’s disease. This is why you need the other components of our variant assessment strategy.
2 – Parental information. Is the variant inherited or de novo? In some cases, this information may not be required, but for most variants, this question should figure into our assessment. A variant that is inherited from an unaffected parent is usually not considered the single cause of your patient’s disease. Maybe the only exception for this rule is a situation when you are dealing with a variant within a larger family that is known to be the cause of the familiar disease and may have unaffected carriers, as in the case of some SCN1A variants. Also, for some variants like microdeletions, the evidence from association studies is so compelling that the variant is often classified as explanatory despite being inherited from unaffected parents. In most cases, however, you need parental confirmation and you would be skeptical of transmitted variants. For example, an STXBP1 variant inherited from an unaffected parent in a child with severe epilepsy would probably considered an innocent bystander and we would continue our search for the cause of the patient’s disease.
3 – Previously seen variants. For some variants, there is strong evidence from other patients that the particular variant is disease-causing. Many of the recently discovered epilepsy genes carry recurrent mutations and we would not necessarily need parental information to call a known recurrent variant in KCNT1, KCNA2, SCN2A, or SCN8A explanatory for the patient’s disease if the same variant has been seen before in patient as a de novo variant, but has never been seen in controls.
4 – Phenotypic information. While the steps 1-3 have dealt with assessment of the actual genetic variant, we should not forget that phenotype is key in many genetic epilepsies. We would be skeptical of a missense variant in CDKL5 in a patient with absence epilepsy without intellectual disability and would probably call this variant an innocent bystander and a non-pathogenic variant that was found by chance. Likewise, we would probably pay little attention to an SCN2A variant in a child with late-onset focal epilepsy or an SCN1A variant in a patient with severe neonatal epilepsy. These variants may in fact be contributory or even causative, but you may want to look closer to see what the phenotype is. Genetic epilepsies are very rich and complex with respect to their clinical features that tell you a lot about how to interpret a variant. Sometimes, this important fact gets lost when a lab result is coming back and may seem at first glance based on interpretation of the pathogenicity of the variant, which causes confusion. However, in genetic epilepsies, correlating the genetic finding with the phenotype is one of the most important steps. Very rarely do genetic findings speak for themselves and you often need the context of the patient’s phenotype.
5 – Precision medicine approaches. In some cases, the considerations 1-4 may not be all that relevant if you genetic finding has a potential treatment implication. For example, independent of the phenotype, you may want to give vitamin B6 if you find a homozygous or compound heterozygous variant in ALDH7A1. Some clinicians may be reluctant to use lamotrigine as a first choice in patient with an SCN1A variant, and we often think twice before initiating valproic acid in a patient who carries a heterozygous POLG mutation. These variants may not be causative for the disease, but they may prompt us to alter our treatment or patient management. It should be noted, however, that most of these strategies are based on assumptions that have not been tested systematically yet. However, these studies may be feasible in the near future given the widespread use of genetic testing in patients with epilepsy.
Thank you for this very informative post. In your 3rd P (the phenotype), I would disagree that a variant of an unknown significance in a gene that has not been previously reported to cause a specific phenotype is less likely to be causative. We have seen several examples of expanding phenotype in epilepsy genes. There are also several examples about modifier effects or interaction between epilepsy genes. I would interpret the result in this case with more caution.