High-resolution molecular genomic autopsy reveals complex sudden unexpected death in epilepsy risk profile



Advanced variant detection in genes underlying risk of sudden unexpected death in epilepsy (SUDEP) can uncover extensive epistatic complexity and improve diagnostic accuracy of epilepsy-related mortality. However, the sensitivity and clinical utility of diagnostic panels based solely on established cardiac arrhythmia genes in the molecular autopsy of SUDEP is unknown. We applied the established clinical diagnostic panels, followed by sequencing and a high density copy number variant (CNV) detection array of an additional 253 related ion channel subunit genes to analyze the overall genomic variation in a SUDEP of the 3-year-old proband with severe myoclonic epilepsy of infancy (SMEI). We uncovered complex combinations of single nucleotide polymorphisms and CNVs in genes expressed in both neurocardiac and respiratory control pathways, including SCN1A, KCNA1, RYR3, and HTR2C. Our findings demonstrate the importance of comprehensive high-resolution variant analysis in the assessment of personally relevant SUDEP risk. In this case, the combination of de novo single nucleotide polymorphisms (SNPs) and CNVs in the SCN1A and KCNA1 genes, respectively, is suspected to be the principal risk factor for both epilepsy and premature death. However, consideration of the overall biologically relevant variant complexity with its extensive functional epistatic interactions reveals potential personal risk more accurately.

Children with epileptic encephalopathy and uncontrolled seizures are at increased risk for sudden unexpected death in epilepsy (SUDEP).[1] Yet, the clinical risk factors do not provide a pathogenic mechanism, and they are not strongly predictive of the individual mortality hazard.

Ion channel genes that modulate cardiac, autonomic, and respiratory functions are prime molecular risk factors for SUDEP. The causative mechanistic link between epilepsy, arrhythmias, sudden death, and the most common LQT gene, the potassium channel KCNQ1, was originally demonstrated in transgenic mice[2] and subsequently clinically validated.[3]

Because many ion channel genes that are critical for the regulation of neurocardiac and neurorespiratory pacemaking are also expressed within brain networks underlying epilepsy, a potential number of novel SUDEP candidate genes extends beyond the cardiac LQT genes.[4] For example, the voltage-gated potassium channel KCNA is coexpressed in brain and vagus nerve, and Kcna1-null mice have seizures, cardiac arrhythmias, vagal hyperexcitability, and die prematurely.[5, 6] Similarly, the voltage-gated sodium channel SCN1A is dually expressed in the brain and the cardiac sinoatrial node and ventricular myocytes.[7] Scn1a-deficient mice also show autonomic instability and seizure-driven vagal activation preceding sudden death,[7] paralleling the clinical observations in children with SCN1A mutations and severe myoclonic epilepsy of infancy (SMEI).[8] However, most SMEI patients do not die suddenly, suggesting the modulating influence of other candidates in the genetic background, beginning with ion channels themselves.

We identified a SUDEP patient who displayed multiple established clinical-pathologic risk factors for SUDEP, including pharmacoresistant epileptic encephalopathy of the SMEI spectrum,[9] recurrent periictal respiratory compromise, and a suspected cardioautonomic clinical phenotype. In order to comprehensively assess the SUDEP risk embedded within this SMEI phenotype, we designed and performed an extensive postmortem search for deleterious variants in candidate ion channel subunit genes that regulate excitability within neural cardiorespiratory regulatory pathways.


The 11-month-old patient and his parents were recruited into the institutional review board–approved Ion Channels in Epilepsy Project at Baylor College of Medicine.[10] Genomic DNA prepared from blood lymphocytes was submitted for commercial diagnostic exome sequencing in five LQT genes—KCNQ1, SCN5A, KCNH2, KCNE2, and ANK2 (Transgenomics); whole genome copy number variants (CNV) analysis at the Medical Genetics Laboratory at Baylor College of Medicine; exome sequencing of 237 ion channel genes[10]; and screening on a custom-designed Ion Channel Comparative Hybridization (ICCH) 4 × 44K microarray (Agilent Technologies, Santa Clara, CA, U.S.A.).[11] (See Data S1.)


Index case clinical report

The proband was a healthy, full-term Latin American male born to a G1P1 mother. At 4 months of age the child developed a prolonged, afebrile hemiclonic seizure that subsided spontaneously but was followed by cessation of respiration. Cardiopulmonary resuscitation (CPR) was administered by a family member and the child promptly and fully recovered. General physical and neurologic examinations, a head computerized tomography (CT) (Acquilion Toshiba 320, Toshiba America Medical Systems, Inc., Tustin, CA, U.S.A.); and electroencephalography (EEG) (24-channel Nicolet/VIASYS, VIASYS Healthcare, Madison, WI, U.S.A.) were normal, and treatment was deferred. Within a month he started experiencing weekly, treatment-resistant hemiclonic seizures involving either side. Serial electroencephalography and brain magnetic resonance imaging (MRI) (Philips Achieva 1.5T, Philips, Andover, MA, U.S.A.) studies remained unremarkable. Karyotyping confirmed a normal male chromosomal pattern. Routine serum and cerebrospinal fluid (CSF) studies were repeatedly normal and a comprehensive diagnostic workup for inborn metabolic errors was noncontributory. His development remained normal. Detailed family history was positive for migraine headaches in the mother. An episode of elevated temperature of 100.8°F triggered the first generalized tonic–clonic seizure at 9 months. Treatment-resistant daily myoclonic jerks associated with loss of tone began at 11 months of age, and the EEG showed epileptiform bursts of frontocentrally dominant generalized 2–3 Hz abortive spike and slow wave activity. The monthly, prolonged partial seizures were associated with cyanosis and frequent secondary generalization. By 18 months, global developmental delay became evident and the clinical evolution led to the diagnosis of SMEI.[12] A cardiac murmur was noted during a follow-up visit. Routine electrocardiography (ECG) was unremarkable, but he was referred to a cardiologist for further evaluation. The proband was 3 years and 3 months old and in his usual state of health when he was found cyanotic and unresponsive in bed. Full autopsy showed only pulmonary congestion, a frequent finding in sudden death.[13] SUDEP was confirmed as the official cause of death.

Integrated genomic analysis

Our initial search centered on the five principal LQTS genes. It showed inherited nonsynonymous single nucleotide polymorphisms (nsSNPs) of unknown clinical significance in KCNH2 (LQT2), SCN5A (LQT3), and KCNE1 (LQT5) (Table 1) and in the context of the LQT genotype (Table S1), it failed to reveal a plausible molecular diagnosis. We next evaluated the channel variant profile (channotype) of the proband through parallel Sanger sequencing of 237 ion channel genes.[10] This step confirmed the previously detected LQTS polymorphisms and additionally uncovered a maternally transmitted heterozygous ryanodine receptor 2 (RYR2) nsSNP Q2958R (rs34967813) that has been reported previously in association with catecholaminergic polymorphic ventricular tachycardia (CPVT)[14] (Table 1; Fig. S1A).

Table 1. Nonsynonymous single nucleotide polymorphisms in candidate genes for SUDEP in the proband (IE124) compared to parental profiles
SyndromeGENEPROTEINPolyphenSIFTIE124p (dbSNP)IE125mIE126f
Cardiac LQT Gene SNP Sequencing (gene dosage: heterozygous = 1; homozygous = 2)       
LQT2 KCNH2 hERG/Kv11.1ToleratedBenignK897T (2) (rs1805123)K897T (2)K897T (1)
LQT3 SCN5A Nav1.5ToleratedBenignH558R (1) (rs1805124)H558R (1)
LQT5 KCNE1 MinKToleratedBenignS38G (1) (rs1805127)S38G (2)
CPVT RYR2 RyR2Probably damagingQ2958R (1) (rs34967813)Q2958R (1)
Human Epilepsy Gene Sequencing (gene dosage: heterozygous = 1; homozygous = 2)       
ADNFLE CHRNA2 nAChR2ToleratedBenignT125A (1) (rs891398)T125A (1)T125A (2)
IGE CLCN2 CLC-2ToleratedBenignT668S (1) (rs9820367)T668S (1)T668S (2)
DEND KCNJ11 Kir6.2ToleratedBenign

K23E (1); (rs5219)

V337I (1) (rs5215)

–; –K23E (2); V337I (2)
Dravet/SMEI/GEFS+ SCN1A Nav1.1Tolerated; DeleteriousBenign; Probably damaging (benign)A1067T (1) (rs2298771); A1783V (1) (rs121917980)A1067T (1); –A1067T (1); –
Respiratory Serotonin Receptor Gene Sequencing (gene dosage: (heterozygous = 1; homozygous = 2)       
N/A HTR3C 5-HT3CToleratedBenignG405A (1) (rs6807362)G405A (2)
N/A HTR3D 5-HT3DTolerated; ToleratedBenign; Possibly damaging (benign)G36A (2) (rs6443930); R260H (2) (rs6789754)G36A (1); R260H (2)G36A (1); R260H (1)

The known epilepsy genes—SCN1A, KCNA1, and SCN8A—have also been implicated in SUDEP.[5, 7, 15-17] Proband channotype analysis[10] revealed an inherited common polymorphism A1067T and a de novo nsSNP, A1783V in SCN1A (Table 1; Fig. S1B) previously found in SMEI (http://www.molgen.vib-ua.be/SCN1AMutations/Home). This predicted deleterious de novo mutation in our case suggests a contribution to the epileptic encephalopathy phenotype, yet its influence on the lethality is uncertain. We also uncovered a paternally inherited, novel, nsSNP, C1288Y, in the RYR3 gene that is preferentially expressed in hippocampus and smooth muscle cells of the pulmonary artery,[18, 19] and the animal models support its role in learning, cognition, and in hypoxia-induced pulmonary vasoconstriction. Therefore, a dysfunctional RYR3 channel could contribute to the cognitive impairment and respiratory compromise of our patient, and targeted RYR3 analysis in SMEI cohorts will be essential to validate this assumption.

Given the clinical history of recurrent, seizure-related apnea, we also analyzed genetic variation in all 18 of the known 5-HT ligand gated ion channels (HTR1A-F, HTR2A-C, HTR3A-E, HTR4, HTR5A, HTR6, and HTR7) and found three inherited nsSNPs (Table 1), of which only the R260H variant is predicted to be possibly damaging by Sorting Intolerant From Tolerant ([SIFT] sift.jcvi.org) SIFT.

Given the modulating role of genetic background on clinical phenotype, we also examined the whole genome for structural gene rearrangements. The clinical aCGH screen identified eight inherited autosomal copy number changes in the proband, such as the paternally inherited duplication in SLC6A10P, a gene recently implicated in autistic spectrum disorder,[20] and the recurrent deletion at 15q11.2, which was found previously in excess in children with congenital heart defects[21] (Table 2). Because all eight CNVs were inherited and their pathogenic relevance to epilepsy or SUDEP was uncertain, we applied our custom high resolution custom designed Ion Channel Comparative Hybridization (ICCH) Array, which has minimal detection threshold of 50 bp and an ultra-dense coverage across the exome of 253 ion channel genes, their structurally related family members, and known accessory subunits. Eleven novel duplications in nine known SUDEP genes were confirmed by quantitative polymerase chain reaction (qPCR; Table 2). Duplication size ranged from to 60 to 3,059 bp. Four CNVs were de novo. Two rearrangements were independent gains in RYR2, and one was a duplication in GABRG3. They were restricted to introns. The single coding de novo CNV, confirmed by qPCR, was at the 3′ end of exon 2 in KCNA1 (Fig. 1A–C), a gene encoding the Kv1.1 pore forming alpha subunit, the loss of function of which causes severe epilepsy and SUDEP in animal models[5] (Fig. 1C). Normalization with two reference genes revealed that the proband harbored five extra copies of this exonic region as compared to the diploid genomes of both parents (Fig. 1D). This gain has a direct impact on the protein coding sequence of the KCNA1 gene. It extends from the highly conserved proline hinge motif (Pro-X-Pro) to the end of the S6 transmembrane helix of the Kv1.1 subunit (Fig. 1E). The PVP motif in this membrane-spanning helix forms a flexible hinge in the transmembrane domain and is directly involved in channel function. Mutations in this region have previously been shown to cause epilepsy (V408T),[22] and premature C-terminal truncation or deletion of the Kv1.1 gene leads to aberrant protein expression resulting in epilepsy, ataxia, megalencephaly, and SUDEP in mice.[23] The repeated gain of this transmembrane helix in the Kv1.1 subunit is likely to impact protein packing and lipid membrane insertion, and thus is an attractive candidate mechanism for Kv1.1 dysfunction contributing to both the seizure and SUDEP phenotype of the proband.

Table 2. Copy number variation in SUDEP proband (IE124) compared to parental profiles using a clinical diagnostic microarray and a custom high-density ion channel comparative hybridization array
CNV no.ChromosomeCytoBandStart position (Hg19)End position (Hg19)Gain/LossNumber of probesLength of CNVKnown CNVaNumber of genesRefSeq (HUGO) gene namesIn proband (IE124)In mother (IE125)In father (IE126)
  1. a

    Known aberrations Toronto Database of Genomic Variants (dgv.tcag.ca) (1 = region has reported gain; 2 = region has reported loss; 3 = region has reported indels; 4 = region has reported inversions).

Clinical aCGH Diagnostic Array (BCM Molecular Diagnostics Core – Director Dr. Ankita Patel)             
1chr1p21.1104012051104012498Loss8447N (1,2,4)  yny
2chr4q125774641557988228Gain9241813N (1,2,3)  yny
3chr8p11.233936994239499498Gain8129556N (1,2,3)2ADAM5P, ADAM3Ayyy
4chr10q11.224638497946506801Gain4121822N (1,2,3,4)3SYT15,GPRIN2,PPYR1yyy
5chr14q11.22160964422028409Loss14418765N (1,2,3)  yyy
6chr14q11.21886456119459230Gain3594669N (1,2,3)6P704P, OR4Q3, OR4M1, OR4N2, OR4K2, OR4K5yyy
7chr15q11.21910876319464920Loss113356157N (1,2,3,4)3LOC646214, CXADRP2, POTEByyy
8chr16p11.23248130833528443Gain1061047135N (1,2,3,4)5ZNF267, HERC2P4, LOC729355, TP53TG3, SLC6A10Pyny
Ion Channel Comparative Hybridization (ICCH) Custom High Density aCGH Array (Translational Neurogenetics in Epilepsy Laboratory - Director Dr. Alica Goldman)             
1chr1q43235879084235879144Gain160N1RYR2 (intron 49/50) NM_001035.2ynn
2chr1q43236028304236028634Gain3330N (1,2)1RYR2 (intron 97/98) NM_001035.2ynn
3chr1q431986462119864666Gain245N1HTR6 (exon 1 5′ UTR) NM_000871.1yyn
4chr7q36.1150285908150285968Gain160N (1,2)1KCNH2 (intron 4/5) iso1 NM_000238.3yny
5chr12p13.3321171632123823Gain36660N (1,2)1CACNA1C (intron 2/3) iso1 NM_199460.2yny
6chr12P13.3248921744892251Gain277N1KCNA1 (exon 2): NM_000217.2ynn
7chr15q122456864224569014Gain3372N (1)1GABRB3 (exon 2-3) iso1 NM_000814.5yyy
8chr15q122505265825055717Gain23059N (2)1GABRG3 (intron 3/4) iso1 NM_033223.4ynn
9chr19P13.131317878813179169Gain3381N1CACNA1A (exon 47 3′ UTR) iso1 NM_000068.3yyy
10chr19p13.131347801413478059Gain145N (2)1CACNA1A (exon 1) NM_000068.3yYy
11chr19p13.3565234565928Gain5694N (1,2)1HCN2 (intron 3/4) NM_001194.3yyy
Figure 1.

A de novo gain in the human epilepsy and SUDEP gene KCNA1 was identified in the proband. (A) Chromosomal location of the human KCNA1 gene using Hg18 as the reference genome. The region of the detected genomic gain is in the gray box with the probe positions located beneath the coding exon. (B) Higher magnification view of the 3′ end of the KCNA1 gene showing the region of the gain relative to the Ion Channel Comparative Hybridization (ICCH) microarray probes. The qPCR primers TLK1822 and TLK1823 used to validate the CNV are shown where primer TLK1823 overlaps with the CGH probe. (C) Sybr green standard curve shown against known concentrations of human gDNA to establish qPCR assay efficiency (91.3%). All qPCR assays underwent optimization and efficiency analysis prior to validation experiments in proband genomic DNA. (D) Quantification of the gain in the KCNA1 gene showed five additional copies of this region in the proband and not in either parent. Normalization of genomic copy number was performed using two reference genes that are known to be free of copy number variants and compared to a normal diploid control. (E) Homology model of the human Kv1.1 ion channel subunit showing two opposing subunits in the tetrameric channel. The ribbon is colored the same as the 6TM schematic diagram top right for orientation. The S4 voltage sensor is gray with the positively charged arginine and lysine residues shown in red. The S6 domain is shown in red and the region of gain is in black. The S6 PVP hinge sequence is highly conserved from jellyfish to man. The residues V404 and V408 are shown in black where amino acid substitutions at these positions cause epilepsy, ataxia, and myokymia.


As the list of validated risk genes for SUDEP expands beyond those currently linked to cardiac-related mortality, robust diagnostic platforms must be developed for optimal assessment of integrated genetic risk.

Herein we show that constructing the genetic variation risk profile for SUDEP benefits from complementary, comprehensive, candidate ion channel gene–focused detection platforms. Both single base pair substitutions and architectural defects contribute to the risk of epilepsy and SUDEP, as evidenced by the discovery of two biologically plausible pathogenic de novo variants in known SUDEP candidates: SCN1A and KCNA1. Mutations in both genes play a critical role in autonomic destabilization described in clinical reports[16, 24] and experimental models of SUDEP,[5, 7] and likely contributed to lethality in our patient. Yet, the co-occurrence of epileptic encephalopathy, ictal apnea, suspected cardiac compromise, and SUDEP in this patient may not be explained solely by the molecular mechanisms elucidated through the SCN1A and KCNA1 models,[5, 7] but may also reflect the compound effect of these mutations together with the transmitted nsSNPs and CNVs of the cardiac arrhythmia and serotonin receptor genes, RYR3 gene variant, and the 15q11.2 region variant associated with structural heart defects. Because clinical phenotypes reflect the pattern of both the individually unique (de novo) and inherited ion channel variants[10] (Fig. S2), resolving the full genetic context is essential for accurate assessment of risk. The integration of ion channel exome sequencing, high resolution ion channel specific CNV survey, and subsequent analysis of 54 candidate SUDEP genes in the neuro-cardiac-respiratory network in this case shows the need for multiscale channel-based risk prediction for SUDEP.

We present the first comprehensive genomic interrogation of ion channel candidate gene pathways to dissect and personalize SUDEP risk prediction in pediatric patients with epilepsy. This case harbored combination of de novo SNPs and CNVs in the SCN1A and KCNA1 genes potentially acting as the principal risk factors for premature death. The larger complexity of the risk load was revealed by additional inherited structural rearrangements and missense polymorphisms within the clinically evident neurocardiac and respiratory pathways. As we continue to refine our understanding of the specific biologic pathways and genetic risk factors leading to SUDEP, comprehensive assessment of genomic variation in cardiac and respiratory networks using detailed gene profiling can enhance predictive value of gene testing in the routine neurologic care of individuals with epilepsy.


Funding: Supported by National Institute of Neurological Disorders and Stroke grants NS067013 and NS067013S (A.M.G.), NS049130 and NS076916 (J.L.N.), and T32NS043124-09 (VCB); CURE; Fiorito Foundation and the Emma Bursick Memorial Fund Q2 (A.M.G.); and the Blue Bird Circle Foundation (J.L.N.). The authors wish to extend special acknowledgment to Dr. Benjamin Salisbury for his expert advice and assistance with the genetic screening of the samples on the Familion long QT gene panel.


J.R.L. is a paid consultant for Athena Diagnostics, has stock ownership in 23 and Me and Ion Torrent Systems, and is a co-inventor on multiple United States and European patents related to molecular diagnostics for inherited neuropathies, eye diseases, and bacterial genomic fingerprinting. The Department of Molecular and Human Genetics at Baylor College of Medicine derives revenue from the chromosomal microarray analysis (CMA) and clinical exome sequencing offered in the Medical Genetics Laboratory (MGL; http://www.bcm.edu/geneticlabs/). The remaining authors have no potential conflicts of interest. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.


  • Image of creator

    Dr. Tara L. Klassen, Ph.D. is an Assistant Professor at the Faculty of Pharmaceutical Sciences at the University of British Columbia, Canada.