Abstract
Brünger T, Pérez-Palma E, Montanucci L, Nothnagel M, Møller RS, Schorge S, Zuberi S, Symonds J, Lemke JR, Brunklaus A, Traynelis SF, May P, Lal D. Brain. 2022:awac305. doi:10.1093/brain/awac305. Epub ahead of print. PMID: 36036558. Clinically identified genetic variants in ion channels can be benign or cause disease by increasing or decreasing the protein function. Consequently, therapeutic decision-making is challenging without molecular testing of each variant. Our biophysical knowledge of ion channel structures and function is just emerging, and it is currently not well understood which amino acid residues cause disease when mutated. We sought to systematically identify biological properties associated with variant pathogenicity across all major voltage and ligand-gated ion channel families. We collected and curated 3,049 pathogenic variants from hundreds of neurodevelopmental and other disorders and 12,546 population variants for 30 ion channel or channel subunits for which a high-quality protein structure was available. Using a wide range of bioinformatics approaches, we computed 163 structural features and tested them for pathogenic variant enrichment. We developed a novel 3D spatial distance scoring approach that enables comparisons of pathogenic and population variant distribution across protein structures. We discovered and independently replicated that several pore residue properties and proximity to the pore axis were most significantly enriched for pathogenic variants compared to population variants. Using our 3D scoring approach, we showed that the strongest pathogenic variant enrichment was observed for pore-lining residues and alpha-helix residues within 5 Å distance from the pore axis center and not involved in gating. Within the subset of residues located at the pore, the hydrophobicity of the pore was the feature most strongly associated with variant pathogenicity. We also found an association between the identified properties and both clinical phenotypes and functional in vitro assays for voltage-gated sodium channels (SCN1A, SCN2A, SCN8A) and N-methyl-D-aspartate (NMDA) receptor (GRIN1, GRIN2A, GRIN2B) encoding genes. In an independent expert-curated dataset of 1,422 neurodevelopmental disorder pathogenic patient variants and 679 electrophysiological experiments, we show that pore axis distance is associated with seizure age of onset and cognitive performance as well as differential gain vs. loss-of-channel function. In summary, we identified biological properties associated with ion-channel malfunction and show that these are correlated with in vitro functional read-outs and clinical phenotypes in patients with neurodevelopmental disorders. Our results suggest that clinical decision support algorithms that predict variant pathogenicity and function are feasible in the future.
Commentary
The majority of epilepsy cases have no obvious underlying etiology, such as stroke, infection, injury, or brain tumors. 1 It is hypothesized that genetics play a major role in these cases. Advances in DNA sequencing technology are revealing the genetic basis of common and rare epilepsies. In generalized epilepsies and familial epilepsies with complex inheritance patterns, it appears that the accumulation of common genetic variants each adding a small amount to overall genetic risk contributes to epilepsy development. 2,3 In contrast, rare variants in single genes have been established as causal for severe epilepsies like developmental and epileptic encephalopathies. 4 In the past decade, hundreds of genes have been identified as monogenic causes of epilepsy. 1
A major class of genes implicated in monogenic epilepsy are voltage-gated and ligand-gated ion channels, including sodium channels (e.g., SCN1A, SCN2A, SCN8A), potassium channels (e.g., KCNA2, KCNQ2, KCNQ3, KCNT1), N-methyl-D-aspartate receptors (e.g., GRIN1, GRIN2A, GRIN2B), and γ-aminobutyric acid receptors (e.g., GABRA1, GABRA2, GABRD, GABRG2). Arguably the most well studied ion channel-related epilepsy is Dravet syndrome. The majority of Dravet syndrome cases are caused by variants in SCN1A that result in haploinsufficiency, including many variants that predict an obvious loss-of-function such as deletion, splice-site, frameshift, and stop variants. 5 However, most genetic variation in ion channels associated with epilepsy are missense variants that make prediction of variant effects on protein function difficult.
Ion channels tend to have high evolutionary conservation. These transmembrane proteins have essential roles in cellular physiology, and large parts of the protein sequence remain under intense evolutionary selection. It can then be inferred that these intransient sequences are important for channel function and that variants occurring in these regions are likely pathogenic. Many pathogenicity prediction algorithms are weighted heavily on estimates of evolutionary conservation. However, sequence changes alone do not provide sufficient information about the functional effects of variants. Missense variants within the same gene may produce different effects, broadly categorized as gain-of-function (e.g., increased ion permeation, changes in activation or inactivation parameters, altered kinetics) or loss-of-function (e.g., decreased ion permeation, impaired channel trafficking).
Some prediction programs also incorporate structural attributes where available. Some functional domains of specific ion channels have been studied extensively. For example, the inactivation gate of sodium channel α-subunits has been mapped to the intracellular linker between D3 and D4 with molecular substitution studies indicating the critical amino acid motif Isoleucine-Phenylalanine-Methionine. 6 Ion channels have characteristic functional readouts that have been studied extensively using electrophysiological methods in vitro and in vivo. Combining identification of patient variants with functional studies has allowed investigators to attempt to correlate variant position with functional effects with some success. For example, variants in the inactivation site of sodium channels are associated with gain-of-function, whereas variants clustered in the pore loop region are associated with loss-of-function. 7 In addition to molecular studies, high-quality crystal structures for many ion channels have recently been published. Incorporation of structural data into variant prediction algorithms will likely increase prediction accuracy.
In this new paper by Brünger et al, the authors aim to improve both prediction of variant pathogenicity and interpretation of genotype–phenotype relationships using conserved biological properties across ion channels. 8 They utilized 3049 variants classified as pathogenic or likely pathogenic and 12 546 population variants in 30 ion channels to ascertain biological parameters that were associated with pathogenicity. For each amino acid residue in a given channel, they computed 13 features such as secondary protein structure, location within a known functional domain (e.g., ligand-binding site), or location within a structurally defined protein region (e.g., lining the pore). The researchers identified residues that share the same features in various combinations and grouped these into 163 residue sets. For example, one residue set included 104 residues that were located at the pore, formed helices, and formed protein–protein interactions.
They found that 52 of these residue sets were enriched in pathogenic versus population variants. They further sought to identify residue sets that were strongly associated with pathogenic variants in all 30 types of ion channels. The features that contributed to pathogenicity across all ion channels were “alpha-helix secondary structure,” “localization at the pore,” and “coiled secondary structure.” Pore-lining residues and residues located in the membrane demonstrated the highest enrichment of pathogenic variants. The highest enrichment was seen in variants closest to the pore axis. They calculated 7 biophysical pore properties and found that hydrophobicity of the pore was correlated with pathogenicity of the pore region. The researchers also investigated whether the location of missense variants was correlated with functional effect and clinical outcome using ∼700 functionally tested variants and more than 1400 pathogenic patient variants in 6 ion channel genes: SCN1A, SCN2A, SCN8A, GRIN1, GRIN2A, and GRIN2B. Again, they found that patient variants clustered within and close to the pore, and they found that functional effects were more pronounced for variants located closer to the pore or closer to the membrane center. Overall, the study identified new conserved structural features of residues across disease-associated ion channels and showed that these features are correlated with variant pathogenicity.
This study identified pore features and residues close to the pore as being most correlated with pathogenicity in ion channels. However, many pathogenic mutations in ion channels are not located at the pore. For example, many pathogenic mutations in sodium channels are located away from the pore in other transmembrane regions such as the voltage sensors, in linkers such as the inactivation gate, and in regions of protein–protein interaction such as the C-terminus. 9 It is possible that some of these residues and regions are less highly evolutionarily conserved than those that are located at the pore and interfere with pore function. This study utilized variant annotation in databases that would have relied on algorithms skewed toward evolutionary conservation to call variants pathogenic versus variant of unknown significance. Thus, variants outside of pore regions may be less likely to be marked as pathogenic or likely pathogenic.
Another factor limiting improvement of variant calling algorithms is lack of functional data. Although the authors were able to find a seemingly high number of functional studies for 6 of the ion channel genes, they still represent only a fraction of the variation identified in disease across all ion channels. Additionally, some disease-associated variants exhibit subtle or no biophysical effects in heterologous systems. 10 For genes that do not have relatively straight-forward functional readouts, the number of variants with functional studies is very small. Some channels have proven hard to crystallize and do not have very good 3D structural data (e.g., SCN8A). More functional and structural data are needed to feed into bioinformatic pipelines to increase our knowledge of how disease-associated variants impact protein function and to improve our ability to make genotype–phenotype correlations.
Improvement of pathogenicity prediction and genotype–phenotype correlations remains an important goal for epilepsy genetics. These improvements will increase diagnoses for genetic epilepsies and further our knowledge of epilepsy gene function, which will aid in the development of therapeutic strategies for severe monogenic epilepsies and complex familial epilepsies. Genotype–phenotype correlations can influence clinical care and choice of treatment. For example, sodium channel blockers are contraindicated for Dravet syndrome caused by loss-of-function SCN1A variants but are recommended for developmental and epileptic encephalopathies caused by gain-of-function SCN2A and SCN8A variants. 11 –13 The study by Brünger and colleagues indicates that the more functional and structural data that are generated and incorporated into bioinformatic analyses, the better we will get at identifying variant contributions to pathogenicity and functional effects that underlie genetic epilepsies.
