Comparison and optimization of in silico algorithms for predicting the pathogenicity of sodium channel variants in epilepsy

Structural mapping of patient-associated KCNMA1 gene variants

KCNMA1-linked channelopathy is a neurological disorder characterized by seizures, motor abnormalities, and neurodevelopmental disabilities. The disease mechanisms are predicted to result from alterations in KCNMA1-encoded BK K+ channel activity; however, only a subset of the patient-associated variants have been functionally studied. The localization of these variants within the tertiary structure or evaluation by pathogenicity algorithms has not been systematically assessed. In this study, 82 nonsynonymous patient-associated KCNMA1 variants were mapped within the BK channel protein. Fifty-three variants localized within cryoelectron microscopy-resolved structures, including 21 classified as either gain of function (GOF) or loss of function (LOF) in BK channel activity. Clusters of LOF variants were identified in the pore, the AC region (RCK1), and near the Ca2+ bowl (RCK2), overlapping with sites of pharmacological or endogenous modulation. However, no clustering was found for GOF variants. To further understand variants of uncertain significance (VUSs), assessments by multiple standard pathogenicity algorithms were compared, and new thresholds for sensitivity and specificity were established from confirmed GOF and LOF variants. An ensemble algorithm was constructed (KCNMA1 meta score (KMS)), consisting of a weighted summation of this trained dataset combined with a structural component derived from the Ca2+-bound and unbound BK channels. KMS assessment differed from the highest-performing individual algorithm (REVEL) at 10 VUS residues, and a subset were studied further by electrophysiology in HEK293 cells. M578T, E656A, and D965V (KMS+;REVEL-) were confirmed to alter BK channel properties in voltage-clamp recordings, and D800Y (KMS-;REVEL+) was assessed as benign under the test conditions. However, KMS failed to accurately assess K457E. These combined results reveal the distribution of potentially disease-causing KCNMA1 variants within BK channel functional domains and pathogenicity evaluation for VUSs, suggesting strategies for improving channel-level predictions in future studies by building on ensemble algorithms such as KMS.

Structural mapping of patient-associated KCNMA1 gene variants

KCNMA1-linked channelopathy is a neurological disorder characterized by seizures, motor abnormalities, and neurodevelopmental disabilities. The disease mechanisms are predicted to result from alterations in KCNMA1-encoded BK K+ channel activity; however, only a subset of the patient-associated variants have been functionally studied. The localization of these variants within the tertiary structure or evaluation by pathogenicity algorithms has not been systematically assessed. In this study, 82 nonsynonymous patient-associated KCNMA1 variants were mapped within the BK channel protein. Fifty-three variants localized within cryo-EM resolved structures, including 21 classified as either gain-of-function (GOF) or loss-of-function (LOF) in BK channel activity. Clusters of LOF variants were identified in the pore, the AC region (RCK1), and near the Ca 2+ bowl (RCK2), overlapping with sites of pharmacological or endogenous modulation. However, no clustering was found for GOF variants. To further understand variants of uncertain significance (VUS), assessments by multiple standard pathogenicity algorithms were compared, and new thresholds for sensitivity and specificity were established from confirmed GOF and LOF variants. An ensemble algorithm was constructed (KCNMA1 Meta Score), consisting of a weighted summation of this trained dataset combined with a structural component derived from the Ca 2+ bound and unbound BK channels. KMS assessment differed from the highest performing individual algorithm (REVEL) at 10 VUS residues, and a subset were studied further by electrophysiology in HEK293 cells. M578T, E656A, and D965V (KMS+;REVEL-) were confirmed to alter BK channel properties in voltage-clamp recordings, and D800Y (KMS-;REVEL+) was assessed as benign under the test conditions. However, KMS failed to accurately assess K457E. These combined results reveal the distribution of potentially disease-causing KCNMA1 variants within BK channel functional domains and pathogenicity evaluation for VUS, suggesting strategies for improving channel-level predictions in future studies by building on ensemble algorithms such as KMS.

Severe epilepsy phenotype with SCN1A missense variants located outside the sodium channel core region: Relationship between functional results and clinical phenotype

PURPOSE

Most SCN1A missense variants located outside the sodium channel core region show a mild phenotype. However, there are exceptions, because of which it is challenging to determine the correlation between genotype and phenotype. In this study, we aimed to determine whether functional study could be used to determine disease severity in cases with such variants, and elucidate possible genotype-phenotype relationships.

METHODS

Forty-seven patients with SCN1A missense variants were recruited, and one with a Dravet syndrome phenotype with an SCN1A missense variant (c.3811T>C/ p.W1271R) located outside the core region was screened with electrophysiological tests. We also reviewed functional SCN1A studies on patients with inconsistent phenotypes and genotypes, and studied the relationship between electrophysiological measurements and clinical phenotype.

RESULTS

Patch clamp experiments showed that the W1271R variant caused significantly reduced sodium current, decreased channel voltage sensitivity, loss of channel availability, and prolonged recovery time from inactivation compared with wild type (WT), which ultimately caused a change in loss of function (LOF). Twelve cases of severe SCN1A-related epilepsy with missense variants located outside the channel core region were also included from the functional studies. Nine patients with missense SCN1A variants showed complete (3/9) or partial (6/9) physiological LOF. Two missense SCN1A variants caused physiological gain-and-loss of function (G-LOF), and one caused decreased excitability (DE).

CONCLUSIONS

Not all missense variants located outside the core region cause a mild phenotype. Although current functional studies in heterologous expression systems do not accurately reflect disease severity caused by SCN1A missense variants, they could be an effective model for generation of data to study the initial effects of SCN1A missense variants.

Computational Studies of the Structural Basis of Human RPS19 Mutations Associated With Diamond-Blackfan Anemia

Diamond-Blackfan Anemia (DBA) is an inherited rare disease characterized with severe pure red cell aplasia, and it is caused by the defective ribosome biogenesis stemming from the impairment of ribosomal proteins. Among all DBA-associated ribosomal proteins, RPS19 affects most patients and carries most DBA mutations. Revealing how these mutations lead to the impairment of RPS19 is highly demanded for understanding the pathogenesis of DBA, but a systematic study is currently lacking. In this work, based on the complex structure of human ribosome, we comprehensively studied the structural basis of DBA mutations of RPS19 by using computational methods. Main structure elements and five conserved surface patches involved in RPS19-18S rRNA interaction were identified. We further revealed that DBA mutations would destabilize RPS19 through disrupting the hydrophobic core or breaking the helix, or perturb the RPS19-18S rRNA interaction through destroying hydrogen bonds, introducing steric hindrance effect, or altering surface electrostatic property at the interface. Moreover, we trained a machine-learning model to predict the pathogenicity of all possible RPS19 mutations. Our work has laid a foundation for revealing the pathogenesis of DBA from the structural perspective.

In Silico Predictions of KCNQ Variant Pathogenicity in Epilepsy

BACKGROUND

Variants in KCNQ2 and KCNQ3 may cause benign neonatal familial seizures and early infantile epileptic encephalopathy. Previous reports suggest that in silico models cannot predict pathogenicity accurately enough for clinical use. Here we sought to establish a model to accurately predict the pathogenicity of KCNQ2 and KCNQ3 missense variants based on available in silico prediction models.

METHODS

ClinVar and gnomAD databases of reported KCNQ2 and KCNQ3 missense variants in patients with neonatal epilepsy were accessed and classified as benign, pathogenic, or of uncertain significance. Sensitivity, specificity, and classification accuracy for prediction of pathogenicity were determined and compared for 10 widely used prediction algorithms program. A mathematical model of the variants (KCNQ Index) was created using their amino acid location and prediction algorithm scores to improve prediction accuracy.

RESULTS

Using clinically characterized variants, the free online tool PROVEAN accurately predicted pathogenicity 92% of the time and the KCNQ Index had an accuracy of 96%. However, when including the gnomAD database as benign variants, only the KCNQ Index was able to predict pathogenicity with an accuracy greater than 90% (sensitivity = 93% and specificity = 98%). No model could accurately predict the phenotype of variants.

CONCLUSION

We show that KCNQ channel variant pathogenicity can be predicted by a novel KCNQ Index in neonatal epilepsy. However, more work is needed to accurately predict the patient's epilepsy phenotype from in silico algorithms.

SCN1A/NaV 1.1 channelopathies: Mechanisms in expression systems, animal models, and human iPSC models

Pathogenic SCN1A/Na_V 1.1 mutations cause well-defined epilepsies, including genetic epilepsy with febrile seizures plus (GEFS+) and the severe epileptic encephalopathy Dravet syndrome. In addition, they cause a severe form of migraine with aura, familial hemiplegic migraine. Moreover, SCN1A/Na_V 1.1 variants have been inferred as risk factors in other types of epilepsy. We review here the advancements obtained studying pathologic mechanisms of SCN1A/Na_V 1.1 mutations with experimental systems. We present results gained with in vitro expression systems, gene-targeted animal models, and the induced pluripotent stem cell (iPSC) technology, highlighting advantages, limits, and pitfalls for each of these systems. Overall, the results obtained in the last two decades confirm that the initial pathologic mechanism of epileptogenic SCN1A/Na_V 1.1 mutations is loss-of-function of Na_V 1.1 leading to hypoexcitability of at least some types of γ-aminobutyric acid (GABA)ergic neurons (including cortical and hippocampal parvalbumin-positive and somatostatin-positive ones). Conversely, more limited results point to Na_V 1.1 gain-of-function for familial hemiplegic migraine (FHM) mutations. Behind these relatively simple pathologic mechanisms, an unexpected complexity has been observed, in part generated by technical issues in experimental studies and in part related to intrinsically complex pathophysiologic responses and remodeling, which yet remain to be fully disentangled.

Optimization of in silico tools for predicting genetic variants: individualizing for genes with molecular sub-regional stratification

Genes are unique in functional role and differ in their sensitivities to genetic defects, but with difficulties in pathogenicity prediction. This study attempted to improve the performance of existing in silico algorithms and find a common solution based on individualization strategy. We initiated the individualization with the epilepsy-related SCN1A variants by sub-regional stratification. SCN1A missense variants related to epilepsy were retrieved from mutation databases, and benign missense variants were collected from ExAC database. Predictions were performed by using 10 traditional tools with stepwise optimizations. Model predictive ability was evaluated using the five-fold cross-validations on variants of SCN1A, SCN2A, and KCNQ2. Additional validation was performed in SCN1A variants of damage-confirmed/familial epilepsy. The performance of commonly used predictors was less satisfactory for SCN1A with accuracy less than 80% and varied dramatically by functional domains of Nav1.1. Multistep individualized optimizations, including cutoff resetting, domain-based stratification, and combination of predicting algorithms, significantly increased predictive performance. Similar improvements were obtained for variants in SCN2A and KCNQ2. The predictive performance of the recently developed ensemble tools, such as Mendelian clinically applicable pathogenicity, combined annotation-dependent depletion and Eigen, was also improved dramatically by application of the strategy with molecular sub-regional stratification. The prediction scores of SCN1A variants showed linear correlations with the degree of functional defects and the severity of clinical phenotypes. This study highlights the need of individualized optimization with molecular sub-regional stratification for each gene in practice.

RET S409Y Germline Mutation and Associated Medullary Thyroid Carcinoma

Background: Inherited medullary thyroid carcinoma (MTC) is primarily caused by RET mutations that are commonly localized in exons 5, 8, 10, 11, and 13-16. In this study, we report pedigrees for individuals with MTC that harbor a germline S409Y variant within exon 6 of the RET proto-oncogene. Methods: Targeted sequencing was used to diagnose four apparently sporadic MTC index cases carrying the germline RET S409Y (c.1226 C>A) variant. Subsequently, 27 relatives of these individuals underwent clinical and genetic assessments and/or thyroid surgery. Furthermore, in silico analyses and in vitro assays were performed to predict or verify the potential oncogenic activity of the S409Y variant. Results: Overall, 15 of 31 participants were found to carry the RET S409Y variant. Of these, 6 presented with isolated MTC (mean age 50.2 years; range 41-75 years), of which 3 presented with neck lymph node metastases and 2 presented with distant liver or lung metastases. Among the remaining 9 carriers, 3 (mean age 56 years; range 41-76 years) had elevated serum calcium-stimulated calcitonin (sCtn) or concurrent marginally elevated serum calcitonin (Ctn) levels, whereas the other 6 (mean age 37.5 years; range 14-52 years) exhibited typical Ctn/sCtn levels (p < 0.05). None of the 15 carriers in these 4 families presented clinical evidence of pheochromocytoma, hyperparathyroidism, or Hirschsprung's disease. In silico analyses revealed that S409Y was a "possibly damaging" mutation that could affect the RET protein inter-domain interface. An in vitro assay revealed that the phosphorylation level of RET tyrosine 905 was relatively higher in the RET S409Y mutant than in wild-type (WT) RET. Moreover, transfection of HEK 293 cells with S409Y enhanced the phosphorylation activity of AKT, ERK pathways, and it increased cell proliferation compared with WT RET, but to a lesser degree than that for the RET C618Y and C634Y mutations. Conclusions: This study demonstrates that the novel germline RET S409Y variant is likely pathogenic and is associated with lower penetrance of MTC than that for the C618Y and C634Y mutations. Individuals with S409Y should be managed using a personalized approach, and additionally, "at-risk" family members should be evaluated. Additional studies are needed to elucidate the correlation between the S409Y mutation and multiple endocrine neoplasia type 2-specific tumors.

Structural Biology Helps Interpret Variants of Uncertain Significance in Genes Causing Endocrine and Metabolic Disorders

CONTEXT

Variants of uncertain significance (VUSs) lack sufficient evidence, in terms of statistical power or experimental studies, to allow unequivocal determination of their damaging effect. VUSs are a major burden in performing genetic analysis. Although in silico prediction tools are widely used, their specificity is low, thus urgently calling for methods for prioritizing and characterizing variants.

OBJECTIVE

To assess the frequency of VUSs in genes causing endocrine and metabolic disorders, the concordance rate of predictions from different in silico methods, and the added value of three-dimensional protein structure analysis in discerning and prioritizing damaging variants.

RESULTS

A total of 12,266 missense variants reported in 641 genes causing endocrine and metabolic disorders were analyzed. Among these, 4123 (33.7%) were VUSs, of which 2010 (48.8%) were predicted to be damaging and 1452 (35.2%) were predicted to be tolerated according to in silico tools. A total of 5383 (87.7%) of 6133 disease-causing variants and 823 (55.8%) of 1474 benign variants were correctly predicted. In silico predictions were noninformative in 5.7%, 14.4%, and 16% of damaging, benign, and VUSs, respectively. A damaging effect on 3D protein structure was present in 240 (30.9%) of predicted damaging and 40 (9.7%) of predicted tolerated VUSs (P < 0.001). An in-depth analysis of nine VUSs occurring in TSHR, LDLR, CASR, and APOE showed that they greatly affect protein stability and are therefore strong candidates for disease.

CONCLUSIONS

In our dataset, we confirmed the high sensitivity but low specificity of in silico predictions tools. 3D protein structural analysis is a compelling tool for characterizing and prioritizing VUSs and should be a part of genetic variant analysis.