Genetic Determinants of Sudden Unexpected Death in Pediatrics

Abstract

Purpose:

To evaluate genetic contributions to sudden unexpected death in pediatrics (SUDP).

Methods:

We phenotyped and performed exome sequencing on 352 SUDP cases. We analyzed for variants in 294 “SUDP genes” with mechanisms plausibly related to sudden death. In a subset of 73 cases with parental data (trios), we performed exome-wide analyses. We conducted cohort-wide burden analyses.

Results:

In total, we identified likely contributory variants in 37/352 probands (11%). Analysis of SUDP genes identified pathogenic/likely pathogenic (P/LP) variants in 12/352 cases (SCN1A, DEPDC5[2], GABRG2, SCN5A[2], TTN[2], MYBPC3, PLN, TNNI3, and PDHA1) and variants of uncertain significance-favor pathogenic (VUS-FP) in 17/352. Exome-wide analyses of the 73 cases with family data additionally identified 4 de novo P/LP variants (SCN1A[2], ANKRD1, and BRPF1) and 4 de novo VUS-FP. Comparing cases vs. controls, we demonstrated an excess burden of rare damaging SUDP gene variants (OR 2.94; 95% CI, 2.37-4.21) and of de novo variants exome-wide in the subset of 73 with trio data (OR 3.13; 95% CI, 1.91-5.16).

Conclusion:

We provide strong evidence for a role of genetic factors in SUDP, involving both candidate genes and novel genes for SUDP, and expand phenotypes of disease genes not previously associated with sudden death.

INTRODUCTION

Over 10% of infant and child deaths in the US occur suddenly, unexpectedly, and without established cause, exceeding pediatric mortality from cancer and cardiac disease.¹ Typically affecting apparently healthy children during sleep, these deaths are certified as sudden infant death syndrome (SIDS), sudden unexpected infant death (SUID), or sudden unexplained death in childhood (SUDC), and can be conceptualized together as sudden unexpected death in pediatrics (SUDP).² SIDS/SUID (hereafter referred to as SIDS), the major component of SUDP, accounts for one in six infant deaths at a rate of 1/1000 live births.³ SIDS is the leading cause of post-neonatal mortality. Reductions in SIDS have been associated with changes in infant sleep practices, yet the reductions mirror declines in non-SIDS rates over the same time, indicating that other improvements in prenatal and infant health may have contributed.⁴ Globally, SIDS rates have remained virtually unchanged since the late 1990’s.⁵ SUDC, less familiar to medical and lay communities, and lacking an International Classification of Diseases designation, is estimated to affect 1.3/100,000 children.⁶ There is emerging consensus that SUDP represents a heterogeneous grouping of rare and undiagnosed diseases presenting with death, sometimes involving genetic mechanisms.^7,8

Genetic contributions to SUDP are supported by studies of families whose children died from SIDS during the “safe sleep” era showing increased risk of recurrence for subsequent siblings (odds ratio, [OR] 4.2) and within three generations (OR 9.3).⁹ The prevailing etiological model of SUDP postulates that modest extrinsic threats become fatal in infants and children who harbor intrinsic vulnerabilities.¹⁰ Early research described intrinsic vulnerabilities as biologically-mediated risk factors, such as prematurity, male sex, and prenatal alcohol and/or tobacco exposure. While these children are often diagnosed with “cardiac arrest” when they present in a hospital setting, and though estimates of deaths attributable to cardiac channelopathies and cardiomyopathy in SUDP vary widely,^11-14 more recent estimates in a European cohort are that there is a cardiac-related etiology in <5% SIDS.¹⁵ Although the detection of medium chain acyl-CoA deficiency in some cases led to optimism that metabolic diseases explained a significant component of SIDS, current estimates suggest that undetected metabolic disease accounts for only 1-2% of SIDS.¹⁶

While previous genetic research has focused on metabolic and cardiac etiologies,¹⁷ there is a substantial body of research demonstrating brainstem-mediated and epilepsy-like changes in SUDP. The ‘brainstem model’ for SIDS is based on serotonergic deficiencies in the neurotransmitter, its precursors, and its transporters in the ventral medulla in 40% of affected infants;^18,19 the implications of these findings are supported by animal work showing auto-resuscitative failure in mice with altered serotonin.^20,21 Neuropathological changes classically described in epilepsy, notably bilamination of the hippocampal dentate gyrus, are present in 41% of SIDS²² and 48% of SUDC.²³ The age overlap in this neuropathologic finding also challenges the traditional notion that SIDS and SUDC have different mechanisms and instead suggests that a substantial portion of SIDS and SUDC share mechanisms that present over the pediatric age continuum.²⁴

Past genetic studies of SUDP have also focused on genes related to cardiac or metabolic conditions.^15,17,25 However, recent findings of pathogenic variants in SCN1A, a gene associated with sudden unexpected death in epilepsy (SUDEP), suggest that epilepsy-related mechanisms for sudden death may play a role in SUDP.^26,27 These genetic findings provide further evidence for a relationship between SUDP and epilepsy, and for the role of neurological disease genes in SUDP pathogenicity that has been relatively unexplored.

Although exome sequencing studies have been reported for SIDS,^17,25 they have been limited to sequencing of probands; we are unaware of genetic reports to date from trio-based (proband and parents) cohorts. The proband-only design, often resulting from the way in which SUDP cases have been ascertained for research, limits the ability to interpret whether variants are de novo or inherited; trio analysis identifying de novo variants provides additional supportive evidence for pathogenicity, as reflected in American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines.^28,29 In addition, in-depth phenotyping, especially critical for interpreting genomic information in a heterogeneous condition such as SUDP, has not been consistently or systematically reported in genetic studies of SUDP. When evaluating a variant in a specific disease-associated gene that has not yet been associated with SUDP, phenotypic detail may support variant pathogenicity, thereby suggesting an expansion of the phenotype of the gene to include sudden death. These limitations become more substantial given the relatively undeveloped concepts of infants and children at risk for SIDS or SUDC, or the “SUDP phenotype”.

Robert’s Program on SUDP at Boston Children’s Hospital (BCH) is a translational research program that takes a novel, multidisciplinary “undiagnosed diseases” approach to discover and understand intrinsic vulnerabilities underlying SUDP, using detailed phenotypic analysis of pre-mortem and autopsy data in conjunction with genetic analysis to identify genetic contributions to SUDP. Here we present the results from our approach to the genetics of SUDP, focusing on neurological, and other systemic/syndromic conditions not previously interrogated, as well as cardiac and metabolic disease.

SUBJECTS AND METHODS

Case ascertainment

From 2012-2020 we ascertained SUDP probands from the San Diego County Medical Examiner’s Office and proband-parent trios referred to Robert’s Program on SUDP at BCH.² DNA samples were obtained from probands and available parents. We obtained post-mortem brain specimens, and additional tissues as indicated to investigate specific phenotypes associated with genetic findings. Informed consent was obtained from parents of participants for trio cases. Consent for remaining probands was obtained from parents or, in cases obtained from the San Diego Office of the Medical Examiners, in accordance with the California statute (SB 1067) for research in sudden infant death syndrome. Research was conducted with approval from the BCH Institutional Review Board.

SUDP proband phenotyping

Detailed phenotypic analysis of each case was conducted by a multidisciplinary team with expertise in pediatrics, genetics, metabolism, neurology, cardiac genetics, pathology, and neuropathology. We obtained data from parent interviews, autopsy reports, investigative reports, and medical records regarding circumstances of death; coincident illnesses; obstetric, birth, and medical history; three-generation family history; and physical findings. Histologic analysis with an emphasis on neuropathological review was conducted according to published methods to identify specific abnormalities associated with SUDP: bilamination of the dentate gyrus and/or other abnormalities of the hippocampal architecture (e.g., hyperconvolution).²⁴

Exome sequencing and variant identification and classification

Exome sequencing was performed using Agilent SureSelectXT Human All Exon V4 or Illumina Rapid Capture Exome enrichment on Illumina platforms. We conducted exome sequencing of all 352 probands and of their parents when available; in total, we sequenced 279 proband-only cases and 73 trios. We analyzed exome sequencing data for potentially pathogenic variants using the WuXi NextCode platform (currently Genuity Science, https://genuitysci.com) with standard filtering for rare damaging variants (details in Supplemental Materials and Methods).¹⁸

For all probands, we analyzed variants in 294 genes plausibly related to SUDP (‘SUDP genes’) (Table S1). The SUDP genes list was curated from Online Mendelian Inheritance in Man (OMIM) and Human Gene Mutation Database (HGMD) and grouped into three categories of conditions: neurological (epilepsy, neurodevelopmental, neuromuscular), cardiac (arrhythmia, cardiomyopathy), and systemic/syndromic (inborn errors of metabolism, multisystem syndromes). For the subset of probands for whom we had trio data, we additionally performed an exome-wide analysis for rare, damaging variants.

We prioritized variants using the following criteria: (a) minor allele frequency (MAF) <0.005% in the Genome Aggregation Database (gnomAD) for dominant, <0.1% for recessive inheritance; (b) absence of homozygous/hemizygous variants in gnomAD; (c) location (exonic or splicing regions); (d) genotype quality = 99 and mean allele read depth >10; (e) conservation (across ≥7 species for missense variants); (f) ‘deleteriousness’ of missense variants: Combined Annotation Dependent Depletion (CADD) score ≥20 and Variant Effect Predictor (VEP) max score >0.9; and (g) predicted splicing effects on ‘cryptic splicing variants’, SpliceAI delta score ≥0.2). We confirmed variants of interest using independent variant analysis in a second platform (Codified Genomics) and direct inspection using Integrative Genomics Viewer. We used the database Mutalyzer to review whether variants were compliant with Human Genome Variation Society (HGVS) guidelines. We defined the following as damaging variants: loss-of-function (stop-gain, frameshift, altered canonical splice site), deleterious missense, non-frameshift indels, and cryptic splicing variants.

We classified variants as pathogenic, likely pathogenic (P/LP), or variant of uncertain significance (VUS) according to American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines.²⁸ We classified variants as VUS-favor pathogenic (VUS-FP) if published functional data demonstrated altered function, another substitution affecting the same amino acid had been reported as pathogenic, or if a cryptic splice was affected. The VUS-FP designation is consistent with ACMG/AMP guidelines supporting the use of additional tiers in sequence variant classification and is in use in some molecular laboratory settings (personal communication, Heidi Rehm).²⁹

We reviewed variants with reference to case-specific phenotypic data in ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) and HGMD, including them only when gene-associated clinical phenotypes were consistent with those of the respective probands. Due to the high prevalence of loss-of-function TTN variants in the general population, and according to accepted practice,³⁰ we only considered variants in TTN exons constitutively expressed in the heart (proportion spliced in (PSI) >0.9). Principal component analysis (PCA) was performed to delineate the ancestry of the subjects.

Burden analysis

Next-generation sequencing (NGS) provides data and novel statistical approaches to disease gene discovery, including a burden analysis approach. In this approach, the proportion of individuals carrying variants in a given gene, group of genes, or exome-wide, is compared between case and control subjects.³¹ To determine whether there is an excess of rare damaging variants in the SUDP gene list in SUDP cases, we conducted a gene list burden analysis: we compared the proportion of SUDP probands vs. 1,433 healthy BCH controls³² with a rare variant in (a) any gene on the SUDP gene list, and in genes on the list related to (b) neurological, (c) cardiac, and (d) systemic/syndromic diseases. Further, we assessed the burden of rare de novo variants exome-wide in SUDP, comparing the proportion of SUDP trios vs. 2,317 control trios (from Simons Foundation Autism Research Initiative)³³ with a rare de novo variant exome-wide. Proportions of variants in cases vs. controls were compared using the two-tailed Pearson’s Chi-squared test. In families with more than one affected sibling who died from SUDP, only the oldest proband was included in the analysis.

RESULTS

Study cohort and phenotypic features

Our cohort included 320 SIDS and 32 SUDC probands (total 352). The majority of probands were 2 to 6 months old at death (average 6.0 ± 10.9 months, range 1 day to 11 years) and male (57%). Comparable numbers were found prone (42%) and supine (40%) at death. Death was associated with a sleep period in 346 of the children. The six deaths that were reported to occur during an awake period were in infants, four of them during or immediately following a feeding. A history of febrile seizures was reported in 14%. Three-generation family histories revealed SIDS or SUDC in 12% of families, febrile seizures in 41% of families, and two families with more than one child dying from SIDS. No consanguinity was reported (Table 1). Of 162 cases with adequate neuropathological tissue for examination, 93 had one or more abnormalities of hippocampal architecture (84 with bilamination of the dentate gyrus, 41 with other abnormalities).

Table 1. Demographics and Phenotypes of the SUDP cohort.

Details regarding age of death, demographic information, and circumstances of death for the 352 SUDP probands are provided. PCA, principal component analysis. ROSC, return of spontaneous circulation. SIDS, sudden infant death syndrome. SUDC, sudden unexplained death in childhood. SUDEP, sudden unexpected death in epilepsy. SCD, sudden cardiac death.

Total SUDP cohort (n = 352)
	number of probands	proportion
Age at Death	352
<2 months	79	22%
2-<6 months	181	51%
6-<12 months	60	17%
≥12 months	32	9%
Sex	352
Male	199	57%
Female	153	43%
Ancestry	347
European	263	76%
African	33	10%
East Asian	32	9%
Mixed race	19	5%
Gestational Age	337
≥37 weeks	283	83%
34-37 weeks	33	10%
<34 weeks	21	6%
Position Found	275
Prone	116	42%
Supine	110	40%
Side	37	13%
Upright/partially upright	12	4%
Sleep site	307
Crib	103	34%
Adult bed	137	45%
Couch	26	8%
Car seat	5	2%
Held	9	3%
Other	27	9%
Sleeping circumstances	344
Shared sleep surface	127	37%
Sleeping alone	217	63%
Additional phenotyping in trios (n = 73)
Personal History	73
Antecedent fever	11	15%
Antecedent minor illness	35	48%
Febrile seizures	10	14%
Other seizures	4	5%
Low birth weight (<2500 g)	3	4%
ROSC	11	15%
Family History	73
SIDS or SUDC	9	12%
SUDEP	1	1%
SCD < 50 years old	11	15%
Febrile seizures	30	41%
Seizures	19	26%
Syncope (excluding vasovagal)	8	11%

Proband analysis for contribution of rare damaging variants in genes on the SUDP gene list

We took a candidate gene approach using the 294 genes on the SUDP gene list on all of our proband cases as we had a substantial number (279) of proband-only cases for whom we could thus not determine whether variants were inherited vs. de novo. Analysis of proband-only data for variants on the SUDP gene list identified 109 rare damaging variants in 98/352 probands (28%) (Table S2). Of these 109, we classified 12 variants as P/LP in genes related to neurological disease (SCN1A, DEPDC5[2], and GABRG2), cardiac disease (SCN5A[2], TTN[2], MYBPC3, PLN, TNNI3), and systemic/syndromic disease (PDHA1[1 male]). We classified 17 variants as VUS-FP in genes associated with neurological disease (CACNA1A, DYRK1A, GABRB3, SCN1A, SCN4A, SCN8A), cardiac disease (SCN5A[2], TTN[3], CAV3, FLNC, KCNE1, MYBPC3, TNNI3), and systemic/syndromic disease (KCNJ2) (Table 2). The remaining 80 variants were classified as VUS. Burden analysis of rare damaging variants in the 294 genes on the SUDP gene list demonstrated an excess of variants in the 352 SUDP probands vs. 1,433 controls (OR 2.94; 95% CI, 2.20-3.94) (Figure 1). We further demonstrated an excess of variants for each disease group in cases vs. controls: neurological (OR 3.91; 95% CI, 2.54-6.02), cardiac (OR 2.16; 95% CI, 1.47-3.16), and systemic/syndromic (OR 2.54; 95% CI, 1.29-4.99) (Figure 1, and Table S3).

Table 2. Summary of Genes with Variants Identified in our SUDP Cohort and Associated Phenotypes.

We present each case with P/LP variants and VUS-FP implicated in our SUDP cohort by disease categories. ACMG, The American College of Medical Genetics and Genomics. AMP, The Association for Molecular Pathology. AF, allele frequency. P, Pathogenic. LP, Likely pathogenic. VUS, variant uncertain significance. M, male. F, female. URI, upper respiratory infection. GI, gastrointestinal symptoms. VUS-FP are indicated with notes regarding reasons for classifying them as such (column indicated with asterisk): functional evidence,^a validated in an exome-wide approach,^b deleterious splicing variant,^c different missense substitution at the same amino acid position that is established as pathogenic.^d

			Genotypes and ACMG/AMP classification			Phenotypes and clinical history						Neuropathology review
Disease Categories	Gene	Case	Variant	AF in normal population	ACMG/AMP classification and evidence*	Age in months	Sex	Personal and family history	Position discovered	Location discovere d	Shared sleep surface	Bilamination of the dentate gyrus	Abnormal hippocampal architecture
Neurological	ALG13	316	NM_001099922.3:c.2525A>G (p.Gln842Arg)	N	VUS-FPb	9	M	Febrile seizures; velopharyngeal dyscoordination; Maternal and sibling history of childhood seizures	Prone	Crib	N	N	N
	CACNA1A	297	NM_023035.3:c.3377G>A (p.Arg1126His)	0.000036	VUS-FPa	5	F	Paternal history of febrile seizures	Upright	Carseat	N	N/A	N/A
	DEPDC5	156	NM_001242896.3:c.2446C>T (p.Gln816*)	N	LP (PVS1, PM2)	3	F	Antecedent minor illness: URI	Supine	Crib	N	Y	Y
		182	NC_000022.11(NM_001242896.3):c.2105-1G>A	0.000004	LP (PVS1, PM2)	4	M	No pediatric primary care	Supine	Adult bed	Y	N/A	N/A
	DYRK1A	71	NM_001396.4:c.1528A>G (p.Ser510Gly)	N	VUS-FPc	3	M	None	Side	Crib	N	N/A	N/A
	GABRB3	242	NM_000814.6:c.650G>A (p.Arg217His)	0.000044	VUS-FPa	5	M	Concurrent fever at time of death	Side	Adult bed	N	Y	Y
	GABRG2	1	NC_000005.10(NM_198903.2):c.327+1G>A	N	LP (PVS1, PM2)	19	M	Febrile seizures; Family history of childhood epilepsy in mother and maternal grandmother, paternal cousin died of SIDS	N/A, Age	Toddler bed	N	Y	Y
	SCN1A	343	NM_001165963.3:c.182T>C (p.Leu61Pro)	N	LP (PS2, PM2)	20	F	Febrile seizures; concurrent fever at time of death. Male sibling clinically diagnosed with Dravet syndrome	N/A, Age	Adult bed	Y	N	N
		308	NM_001165963.3:c.4357T>G (p.Tyr1453Asp)	N	LP (PS2, PM2)	22	M	None	N/A, Age	Toddler bed	N	N	N
		155	NM_001165963.3:c.2045G>T (p.Gly682Val)	N	LP (PS3, PM2)	2	F	Antecedent minor illness: GI	Supine	Crib	N	N	N
		60	NM_001165963.3:c.3886T>A (p.Leu1296Met)	0.000004	VUS-FPa	2	F	None	Prone	Adult bed	N	N/A	N/A
	SCN4A	70	NM_000334.4:c.2045C>G (p.Ser682Trp)	N	VUS-FPa	3	M	None	Prone	Couch	Y	N	N
	SCN8A	144	NM_001330260.2:c.3284G>A (p.Arg1095Gln)	N	VUS-FPc	1	M	None	Supine	Crib	N	N/A	N/A
Cardiac	AKAP10	295	NM_007202.4:c.850A>G (p.Lys284Glu)	N	VUS-FPb	10	F	None	Supine	Crib	N	Y	Y
	CAV3	329	NM_033337.3:c.260T>C (p.Leu87Pro)	0.000004	VUS-FPa	4	F	None	Prone	Crib	N	N	N
	FLNC	215	NC_000007.14(NM_001458.4):c.3964+5G>A	N	VUS-FPc	6	M	None	Supine	Adult bed	Y	N/A	N/A
	KCNE1	181	NM_000219.6:c.173C>T (p.Thr58Ile)	0.000028	VUS-FPa	2	F	None	Side	Adult bed	Y	N/A	N/A
	MYBPC3	103	NC_000011.10(NM_000256.3):c.821+1G>A	0.000029	P (PS3, PVS1, PP1, PP5)	2	M	None	Supine	Couch	Y	N/A	N
		89	NM_000256.3:c.3791G>A (p.Cys1264Tyr)	N	VUS-FPc	3	F	None	Supine	Couch	Y	N/A	N/A
	PLN	277	NM_002667.5:c.40_42del (p.Arg14del)	N	P (PS3, PM4, PP1, PP5)	6	M	Family history of early cardiac death in 3 maternal family members	Supine	Crib	N	N/A	N/A
	SCN5A	164	NC_000003.12(NM_001099404.1):c.4299+1del	N	LP (PVS1, PM2)	1	F	None	Prone	Couch	Y	Y	Y
		269	NM_001099404.1:c.5287G>A (p.Val1763Met)	N	P (PS3, PS2, PM2)	1	M	Antecedent minor illness: URI	N/A, Awake	N/A	N/A	N/A	N/A
		179	NM_001099404.1:c.4895G>A (p.Arg1632His)	0.000008	VUS-FPa	1	F	Antecedent minor illness: URI	Supine	Couch	Y	N/A	N/A
		152	NC_000003.12(NM_001099404.1 ):c.3840+12C>T	0.000032	VUS-FPc	4	F	None	Supine	Crib	N	N	Y
	TNNI3	2	NM_000363.5:c.433C>T (p.Arg145Trp)	0.000011	P (PS3, PS1, PP1, PP5)	2	F	None	Supine	Playpen	N	N/A	N/A
		330	NM_000363.5:c.556C>T (p.Arg186Trp)	0.000014	VUS-FPd	8	M	Sibling with cardiac concerns lacking definitive diagnosis; Family history of SIDS (maternal cousin)	Supine	Crib	N	N	Y
	TTN	251	NM_001267550.2:c.98299_98300del (p.Arg32767Glyfs*2)	0.000007	P (PVS1, PM1, PP5)	7	M	Paternal history of heart transplant at 29 years old	Supine	Crib	N	Y	Y
		126	NC_000002.12(NM_001267550.2):c.97492+1G>C	N	P (PVS1, PM1, PM2, PP5)	2	M	None	Prone	Caregiver's lap	Y	N	N
		132	NM_001267550.2:c.91721A>T (p.Glu30574Val)	N	VUS-FPc	9	M	Concurrent fever at time of death	Supine	Adult bed	Y	N/A	N/A
		149	NM_001267550.2:c.43622C>T (p.Ser14541Leu)	N	VUS-FPc	5	F	None	Supine	Crib	N	Y	Y
		112	NM_001267550.2:c.64898G>A (p.Arg21633Gln)	N	VUS-FPc	2	M	None	Supine	Adult bed	Y	N/A	N/A
Systemic/Syndromic	ANKRD11	282	NM_001256182.1:c.7534C>T (p.Arg2512Trp)	N	P (PS2, PS1, PM2)	30	M	Tethered cord, vocal cord paralysis, strabismus, ventricular septal defect, aortic root dilation, boney defect at base of skull, hypotonia, developmental delay	N/A, Age	Toddler bed	N	N	Y
	BRPF1	319	NM_001003694.2:c.1182_1183del (p.Ala396Leufs*69)	N	P (PVS1, PS2, PM2)	38	M	Short stature, ptosis, blepharophimosis, broad thumbs, speech delay	N/A, Age	Toddler bed	N	N/A	N/A
	FLNA	339	NM_001110556.2:c.4772C>T (p.Pro1591Leu)	0.000011	VUS-FPb	17	M	Concurrent fever	N/A, Age	Crib	N	Y	N
	KCNJ2	67	NM_000891.3:c.119G>A (p.Arg40Gln)	0.000018	VUS-FPa	5	F	None	Prone	Couch	N	N/A	N/A
	PDHA1	53	NM_001173454.1:c.1246C>T (p.Arg416Cys)	N	P (PS3, PM2, PP1, PP2, PP5)	11	M	Nystagmus; Antecedent febrile illness with fatigue; Histology inconsistent with overwhelming viral illness	Supine	Couch	Y	N/A	N/A
	TCF4	300	NM_001243226.2:c.868T>A (p.Ser290Thr)	N	VUS-FPb	6	M	None	Supine	Adult bed	Y	Y	Y

Figure 1. Burden analysis reveals excess of rare damaging variants in SUDP.

A. Comparing all rare damaging variants in our cohort (n = 352) vs. controls (n = 1,433), we demonstrate an excess of rare damaging variants in the entire SUDP gene list (OR 2.94; 95% CI 2.20-3.91), as well as within each disease-related group: neurological, cardiac, and systemic/syndromic disease. B. Considering all genes exome-wide, we observed an excess of rare damaging de novo variants in SUDP cohort trios (n = 73) vs. control trios (n = 2,317) (OR 3.13; 95% CI, 1.91-5.16; Pearson’s Chi-squared two-tailed p = 2.56x10⁻⁶).

Trio analysis for contribution of de novo variants to SUDP

For the 73 cases with trio data, we conducted exome-wide analysis to assess for de novo or X-linked variants. The analysis of the 73 trios revealed: 50 de novo variants (34 probands with one de novo variant, five probands with two, and two with three), 13 X-linked maternally inherited variants, three homozygous variants (one case each), and three compound heterozygous rare damaging variants (Figure 2, and Table S4). Eight de novo variants were P/LP or VUS-FP. Six had been identified in proband-only analyses as VUS, and their de novo status resulted in reclassification, two to LP (SCN1A) and four to VUS-FP (ALG13, AKAP10, FLNA, TCF4). Two de novo variants in genes not on the SUDP list (BRPF1 and ANKRD11) were identified and classified as LP.

Figure 2. Rare damaging de novo and maternally inherited X-linked variants among 73 trios.

Analysis of 73 SUDP trios identified rare damaging de novo and maternally inherited X-linked variants in 16 genes with known associations with neurological (blue), cardiac (red), and systemic/syndromic (orange) disease (left) and in 46 additional genes without known disease relevance (right). Genes found in the SUDP gene list analysis are indicated by asterisks.

Burden analysis revealed a significantly greater proportion of SUDP trio cases (38/73) vs. controls (596/2,317) with rare damaging de novo variants exome-wide (OR 3.13, 95% CI 1.91-5.16) (Figure 1).

Summary of genetic analysis in proband-only and trio analyses

Overall, we identified rare damaging P/LP or VUS-FP variants in 37/352 SUDP probands (11%). We identified 16 P/LP (12 on analysis of the SUDP gene list; 2 VUS-FP reclassified to LP based on the de novo finding on exome-wide analysis; and 2 de novo variants in genes not on the SUDP list) and 21 VUS-FP (17 on analysis of the SUDP gene list; and 4 VUS reclassified to VUS-FP based on the de novo finding on exome-wide analysis) (see Table 2). Among the 37 variants in these cases, 13 were in genes related to neurological, 18 to cardiac, and 6 to systemic/syndromic disease. The implicated genes are displayed according to disease category and age of death in Figure 3.

Figure 3. Genes implicated in SUDP according to age of death.

Our SUDP cohort included 320 SIDS and 32 SUDC probands (total 352), among which we identified a pathogenic/likely pathogenic (P/LP) variant or variant of uncertain significance-favor pathogenic (VUS-FP) in 37 (11%). Each case with a likely contributory genetic variant is represented by a box with the associated gene name (bold for P/LP, non-bold for VUS-FP) displayed on a timeline indicating age of death. Each gene’s disease category is indicated by color: neurological (blue), cardiac (red), and systemic/syndromic (orange).

Genotype-phenotype correlation

Among the 37 cases with variants identified, relevant history or family history was observed in several cases (Table 2). Febrile seizures were reported in three probands with variants in genes associated with neurological disease (epilepsy) (ALG13, GABRG2, and SCN1A); among these, the child with the SCN1A variant had a sibling with epilepsy (reported previously²⁷). Two probands had a significant family history of cardiac disease: one with a TTN variant whose father had undergone heart transplant at 29 years of age, and one with a PLN variant and a family history of early cardiac death in three maternal family members.

Detailed phenotypic review revealed that the probands with de novo ANKRD11 and BRPF1³⁴ variants had features consistent with the related genetic syndromes, KBG syndrome and intellectual developmental disorder with dysmorphic facies and ptosis (IDDDFP), respectively; notably, neither were recognized to have these syndromes pre-mortem or at autopsy (Table 2). Ten of the 84 cases with dentate gyrus bilamination (12%) harbored a P/LP variant or VUS-FP: three cases had variants in genes associated with neurological disease (DEPDC5, GABRB3, GABRG2), five in genes associated with cardiac disease (AKAP10, SCN5A, TNNI3, TTN[2 cases]), and two in genes associated with systemic/syndromic disease (FLNA, TCF4). Ten of the 41 cases with abnormal hippocampal architecture harbored a P/LP variant or VUS-FP (24%): these included 8 of the above cases with variants in DEPDC5, GABRB3, GABRG2, AKAP10, SCN5A, TTN[2 cases], and TCF4, an additional case with a variant in SCN5A, and one with a variant in ANKRD11, associated with systemic/syndromic disease (Table 2).

DISCUSSION

Given our hypotheses and emerging evidence that heterogeneous genetic factors contribute to SUDP, we pursued an undiagnosed diseases approach to SUDP that included in-depth phenotyping and analysis of exome data. We undertook a candidate gene approach for all cases in a large SUDP cohort of 352 cases. Further, for a subset of 73 probands, we leveraged the availability of trio data and conducted exome-wide analyses. We identified genetic contributions to SUDP in 11% of our cohort, providing specific examples of intrinsic vulnerabilities to sudden death. Our exome-wide trio analysis of the subset with parental DNA identified de novo variants in genes not previously associated with SUDP. Incorporating parental data also allowed us to reclassify several VUS identified in the proband-only analysis to P/LP variants or VUS-FP.

We used a genetic burden testing approach that leverages data from unrelated probands, which can increase the power to identify novel genetic associations.³⁵ This approach aggregates variants across a gene or group of genes to improve discovery power by comparing the proportion of cases vs. controls with variants in the gene(s) of interest. Importantly, our cohort-wide analyses demonstrated increased genetic burden in SUDP, both with respect to rare damaging variants in targeted genes and de novo variants exome-wide. These results provide further evidence supporting a role of genetic factors in SUDP and, moreover, support the premise that children dying from SUDP may harbor intrinsic vulnerabilities reflecting differences from unaffected children, on a population level. Our findings support the hypothesis that diverse neurological, cardiac, and metabolic mechanisms play a role in SUDP. Categories in our SUDP gene list were based on the predominant clinical symptoms associated with the genes; we acknowledge that some genes are expressed in multiple tissues, including both the brain and heart. While past studies of SUDP have focused on genes related to cardiac or metabolic conditions, our candidate approach using the SUDP gene list additionally included genes related to neurological and other systemic/syndromic conditions not previously interrogated. Indeed, 19 of the 37 probands harboring P/LP variants or VUS-FP (51%) had variants in genes related to neurological disease and other systemic/syndromic conditions, supporting the validity of this approach. We hypothesize that epilepsy-related mechanisms may have contributed to death in cases with variants in genes associated with epilepsy (SCN1A, DEPDC5, ALG13, CACNA1A, GABRB3, GABRG2, SCN8A). Among these, SCN1A has previously been implicated in SIDS³⁶ and SUDC,²⁷ and DEPDC5 and SCN1A have been implicated in SUDEP.³⁷ In addition, while simple febrile seizures have not been shown to be associated with an increased risk of death, the history of febrile seizures in some SUDP cases with variants in epilepsy-related genes suggests the possibility that these previous episodes may have been seizures unmasked by fever in individuals with genetic risk for epilepsy.³⁸

Our finding of variants in genes related to cardiac disease, particularly arrhythmia and cardiomyopathy, is consistent with prior reports.^39,40 Eighteen (49%) of our 37 P/LP variants and VUS-FP were in cardiac disease genes: TTN(5), SCN5A (4), MYBPC3 (2), CAV3, FLNC, KCNE1, and TNNI3, all previously reported with sudden death. The presence of a VUS-FP in the additional cardiac-related gene AKAP10 in one proband suggests an expansion of its associated phenotype. Since the penetrance of some arrhythmia-related genes is incomplete, the identification of variants in these genes in SUDP may have implications for living family members unaware of their risk. Additionally, genetic variants known to disrupt cardiac electrical activity may also be expressed and affect function in the brain.

Our results demonstrate the importance of conducting trio exome-wide analysis when trio data are available. Trio analysis led to reclassification in six cases from our initial SUDP gene list proband-only analysis: when found to be de novo, two VUS were reclassified to LP (SCN1A) and four to VUS-FP (ALG13, AKAP10, FLNA, TCF4). In addition, standard candidate gene-based approaches may overlook the role of genes in SUDP as classification of variants relies in part upon disease associations, which are largely based on phenotypes described typically in people living with disease. The risk for SUDP may not be well reflected among the known phenotypes of many disease genes because children may die of SUDP before a genetic condition is recognized. A trio approach to SUDP, not restricted to genes with known or hypothesized associations with sudden death, allows for novel genotype-phenotype discoveries for an entity that is still largely not understood and requires a broad-based approach. Our trio analysis identified new associations between syndromic disease genes and sudden death, including BRPF1,³⁴ associated with IDDDFP, and ANKRD11, associated with KBG syndrome, both in cases not recognized as such pre-mortem.

We identified a male proband with a pathogenic variant in PDHA1, responsible for pyruvate dehydrogenase E1-alpha deficiency, a metabolic disorder with variable expressivity. The variant, assessed as pathogenic by our classification and in ClinVar, expands the phenotype of this condition to include sudden death in the absence of overt metabolic disease. We additionally observed a VUS-FP in ALG13, also related to epilepsy and metabolic disease (glycosylation disorder) in a proband whose pre-mortem history was positive only for seizures with fever. While we classified this gene as neurological because of the epilepsy association, it is also possible that the child had an occult glycosylation defect.

There is growing evidence that stillbirth, SIDS, and SUDC represent a continuum with shared etiologies presenting as unexplained death over a continuum from fetal life through childhood.⁴¹ Shared neuropathological changes and a gene (SCN1A) common to SIDS and SUDC provides further evidence that unexplained infant and child deaths may contribute to the continuum. A recent study reports causal variants in stillbirth in a similar proportion of cases to the present study, and shared genes in their cohort and ours (e.g., MYBPC3)⁴² provide a genetic connection between stillbirth and SUDP. Our observation that neurologic and syndromic cases span SIDS and SUDC, whereas cardiac genes clustered in the SIDS age range (Figure 3), may help further delineate the mechanisms related to sudden death along this continuum.

The majority of deaths in our cohort remained genetically unexplained, paralleling results in other studies of undiagnosed disease. Additional cases will be required to validate the association between SUDP and genes implicated here in sudden death. Although we categorized genes on the SUDP gene list based on the predominant clinical symptoms associated with them, some genes are expressed in multiple tissues, including both the brain and heart, and further work will be necessary to determine pathogenic mechanisms. Further, the presence of pathogenic variants, even those deemed pathogenic, does not in itself establish causality. Despite these limitations, collectively our findings demonstrate a genetic contribution to SUDP and highlight the need for future investigation into as yet unidentified genetic causes due to limited cohort size and numbers of trios thus far sequenced. Future genetic evaluation of SUDP cohorts should include trio analysis when possible, to identify additional de novo causes (hypothesized to be involved given the lethal nature of the condition) and inherited causes in genes with decreased penetrance. In addition, deep sequencing of candidate genes, genome sequencing, and copy number analyses, could lead to the identification of mosaic variants, non-coding variants, and structural variants, respectively.

Our finding of specific genetic contributions to SUDP in 11% of our cohort highlights the role of genetics in SUDP and indicates diverse mechanisms within the diagnosis. Additionally, we demonstrate a paradigm for the genetic evaluation of SUDP that benefits from engaging parents to obtain data about the deceased infant and the family history, thus maximizing the available phenotypic data to inform genetic analyses, as well as samples from the proband and parents so that trio analyses can be conducted. Ideally, such practice should be undertaken in collaboration with specialized teams who can deliver results in a clinical context, providing bereaved parents with an approach to search for answers to why their child died, medical surveillance for at-risk surviving family members, counseling about recurrence risks, and the opportunity to participate in a process that will ultimately lead to better understanding and prevention of SUDP.

In conclusion, we demonstrate evidence for diverse genetic contributions to SUDP through an undiagnosed disease approach. We advocate that, when resources permit, comprehensive evaluation for SUDP should include a comprehensive genetic evaluation.

Data availability

Variant level sequencing data has been submitted to ClinVar. Accession numbers and URLs are pending.