Clinical exome sequencing uncovers a high frequency of Mendelian disorders in infants with stroke: a retrospective analysis

Abstract

Background:

Stroke causes significant disability and is a common cause of death worldwide. Previous studies have estimated that 1-5% of stroke is attributable to monogenic etiologies. We set out to assess the utility of clinical exome sequencing (ES) in the evaluation of stroke.

Methods:

We retrospectively analyzed 124 individuals who received ES at the Baylor Genetics reference lab between 2012 and 2021 who had stroke as a major part of their reported phenotype.

Results:

Ages ranged from 10 days to 69 years. 8.9% of the cohort received a diagnosis, including 25% of infants less than 1 year old; an additional 10.5% of the cohort received a probable diagnosis. We identified several syndromes that predispose to stroke such as COL4A1-related brain small vessel disease, CBS-related homocystinuria, POLG-related disorders, TTC19-related mitochondrial disease, and RNASEH2A associated Aicardi-Goutieres syndrome. We also observed pathogenic variants in NSD1, PKHD1, HRAS and ATP13A2, which are genes rarely associated with stroke.

Conclusions:

Although stroke is a complex phenotype with varying pathologies and risk factors, these results show that use of exome sequencing can be highly relevant in stroke, especially for those presenting <1 year of age.

Background

Stroke is a leading cause of death and long-term disability worldwide. The risk of stroke for children and young adults is lower than for older adults, but is still 1/4,000 [Ferriero et al., 2019]. Especially for children, “stroke” may include ischemic, hemorrhagic, or metabolic pathology. Not surprisingly, the molecular basis of stroke remains partially understood. Pediatric stroke sometimes occurs as part of an inherited genetic syndrome [Barrett and Meschia, 2014], and can follow Mendelian inheritance patterns [Ilinca et al., 2020]. Monogenic causes of stroke include CADASIL, homocystinuria, Fabry disease, TREX1, COL4A1/COL4A2-related syndromes, and others [Chojdak-Łukasiewicz et al., 2021; Bersano et al., 2021]. One study of 38 patients with pediatric and perinatal stroke diagnosed 10% of cases using a panel of 15 genes focused on Mendelian causes of stroke [Grossi et al., 2020]. Others have similarly shown the importance of gene-panel testing in evaluation of stroke [Fang et al., 2020].

To our knowledge, exome sequencing (ES) has not been systematically utilized for evaluation for monogenic causes of stroke [Gudenkauf et al., 2020]. ES has been used successfully to diagnose patients with extensive prior testing [Kumar et al., 2022], and we hypothesized that ES would have high diagnostic rates for infants, children, and young adults with stroke as a presenting complaint. Here we describe the results of ES in 124 individuals with personal histories of stroke referred to our clinical testing facility.

Materials and Methods

Editorial Policies and Ethical Considerations

This study adheres to the principles of the Declaration of Helsinki. The study was ethically approved by the Baylor College of Medicine Institutional Review Board (protocol H-41191).

Clinical Samples

We retrospectively identified 124 patients from over 8,000 clinical samples exome sequenced by Baylor Genetics (Houston, TX) between June 2012 and May 2021. Inclusion criteria included personal histories of stroke or cerebral infarct based on the provided clinical phenotype. Patients who only had family histories of stroke were excluded, as were patients diagnosed by methods besides exome sequencing (e.g. concurrent microarray or mitochondrial DNA sequencing). There was a single prenatal exome which was retained (Patient ID 87642) with maternal demographic data. Patients provided informed consent for genetic testing, as well as receiving secondary findings or carrier status as recommended by the American College of Medical Genetics and Genomics (ACMG) [Kalia et al., 2017]. Patient identification numbers were randomized.

Exome Sequencing and Analysis

The 124 patients underwent one of proband-only ES (available since December 2011), trio ES (available since October 2014), or critical-trio ES (an expedited test available since April 2015), all of which were performed by Baylor Genetics as clinical assays. Trio ES involves sequencing the proband and both parents, allowing variants to be described as inherited or de novo, and in cis or trans. Critical-trio ES is analytically identical to trio ES, but follows an expedited schedule and is most often used for critically ill patients. A few patients included in this study had Total Blueprint panels, which encompass roughly 5,000 genes associated with Mendelian disease. In some cases parental samples were provided and used to phase and determine inheritance of significant variants identified with proband-only ES. All samples were concurrently analyzed by Illumina HumanOmni1-Quad, HumanExome-12v1 or HumanCoreExome-24v1 chromosomal microarrays to exclude large copy number variants. Exome data were interpreted according to ACMG guidelines and variant interpretation guidelines of Baylor Genetics as previously described [Lalani et al., 2016].

Clinical Exome Reporting

A case was classified as molecularly diagnosed when (1) a pathogenic or likely pathogenic (P/LP) variant was detected, (2) in a gene associated with a syndrome that substantially matched the information submitted to the laboratory, and (3) the zygosity of the variant matched the gene’s established inheritance pattern. Some cases were designated as “probably diagnosed” if there was a variant of unknown significance (VUS) with appropriate zygosity in a gene that substantially matched the reported phenotype, or if there was P/LP variant with appropriate zygosity in a gene that partially matched the reported phenotype. Patients with P/LP variants in genes that cause a known syndrome which does not match the reported phenotype were considered “undiagnosed” for the purposes of this study.

Statistics

R x64 4.0.3 with base package was used for statistical analyses. Fisher exact tests were used unless otherwise specified.

Results

Demographics, Indication, Test Type

Of 124 patients in the study, 67 were male. Ages ranged from 10 days to 69 years. The median age was 9.0 years, with 24 patients younger than 1 year and 74 patients below 12 years. Sixteen patients had metabolic stroke, 9 had prenatal stroke, and 14 had hemorrhagic stroke or stroke associated with vessel dissection. Ninety-three patients had non-neurologic symptoms or findings (Table S1). Eighty-nine patients had proband-only ES, 19 trio ES and 15 critical-trio ES (Table 1). One patient had a total blueprint assay.

Table 1.

Diagnosis status by test type and patient age. Diagnosed cases had a pathogenic or likely pathogenic (P/LP) variant identified in a gene that causes a syndrome that matched the reported patient phenotype. Probably diagnosed cases had P/LP variants in a gene causing a syndrome that partially matches the observed phenotype, or a variant of unknown significance in a gene causing a syndrome that fully match the observed phenotype.

Age:	Diagnosis:	Critical-trio ES	Trio ES	Proband ES	TotalBluePrint	Totals
<3m	Undiagnosed	3	2	5	0	10
	Probably diagnosed	0	0	0	0	0
	Diagnosed	0	1	0	0	1
3m-1y	Undiagnosed	3	3	1	0	7
	Probably diagnosed	0	0	1	0	1
	Diagnosed	2	0	3	0	5
1-5y	Undiagnosed	0	3	14	0	17
	Probably diagnosed	0	1	1	0	2
	Diagnosed	0	0	0	0	0
5-12y	Undiagnosed	0	5	18	1	24
	Probably diagnosed	1	0	4	0	5
	Diagnosed	2	0	0	0	2
12-18y	Undiagnosed	1	2	11	0	14
	Probably diagnosed	0	0	1	0	1
	Diagnosed	1	0	0	0	1
18-40y	Undiagnosed	2	2	10	0	14
	Probably diagnosed	0	0	2	0	2
	Diagnosed	0	0	2	0	2
>40y	Undiagnosed	0	0	14	0	14
	Probably diagnosed	0	0	2	0	2
	Diagnosed	0	0	0	0	0
Totals		15	19	89	1	124

Diagnostic Rates

Eleven (8.9%) patients were fully diagnosed; though an additional 13 cases were probably diagnosed, resulting in a 19.4% diagnostic rate when included (Table 1, Table S2). Diagnostic rates varied by patient age. For example, the average age of fully diagnosed cases was 6.7 years, versus an average age of 15.4 for the undiagnosed cases (t-test p-value=0.01, n=11, 100). The diagnostic rate was highest in infants: 6/24 for patients under 1 year-old versus 5/100 for all other patients (p=0.007, OR=6.2 [95% CI 1.4-28.7]).

The diagnostic rate for critical-trio ES was also higher than proband-only ES at 5/15 versus 5/89 cases fully diagnosed, respectively (p=0.005, OR=8.1 [95% CI 1.6-42.6]). However, critical-trio ES also trended towards outperforming standard trio ES (5/15 versus 1/19, p=0.066, OR=8.4 [95% CI 0.8-447.4]). Since these two tests are analytically identical, we assessed for possible confounding factors that could explain the difference in performance, such as critical-trio ES being used in younger or more critically ill patients with higher rates of genetic disease.

The average age for patients receiving trio studies was younger than patients receiving proband-only studies (7.1 versus 17.9 years, t-test p<0.0001, n=34, 89). Although trio studies (critical and non-critical) make up 27% of the dataset (34/124), they make up 58% of the tests sent in infants less than 1 year-old (14/24, p=0.0007, OR=5.2, [95% CI 1.8-15.1]). The selection of proband-only versus trio testing is sufficiently correlated with patient age that we could not quantify separate diagnostic rates for these tests while also correcting for age.

Diagnoses

A P/LP variant that substantially matched the reported phenotype was identified in 11/124 cases, (Table 2). Six solved cases were under 1-year-old, with diagnoses that included Aicardi-Goutieres and POLG-related syndromes. Additionally, in six solved cases the gene was previously associated with stroke, based on reviews of OMIM, GeneReviews and literature searches. Among that group are patients 31162 with CBS-related homocystinuria (MIM# 236200), patient 84711 with TTC19-related mitochondrial complex III deficiency (MIM #615157), and patient 87229 with POLG-related mitochondrial disease.

Table 2.

The 11 fully diagnosed cases. The pathogenic or likely pathogenic variant and the affected gene are shown, as well as additional data that were clinically reported. When available, parental samples were used to determine phase and inheritance of P/LP variants for proband-only ES.

Patient ID	Sex	Age Range	Test	Gene	Variant	Zygosity	Inheritance	Disease	MIM#	Stroke-associated
62614	F	<3m	Trio ES	ALDH7A1	c.834G>A; c.1279G>C (p.E427Q)	Compound Het	Inherited; c.834G>A father, p.E427Q mother	Epilepsy, pyridoxine-dependent	266100	No
90361	F	3m-1y	Proband ES	HRAS	c.64C>A (p.Q22K)	Heterozygous	De novo	Costello syndrome	218040	No
76670	M	3m-1y	Critical-trio ES	PKHD1	c.983G>A (p.R328Q); c.1964+1G>T	Compound Het	Inherited; p.R328Q mother, c.1964+1G>T father	Polycystic kidney disease 4, with or without hepatic disease	263200	No
34275	M	3m-1y	Proband ES	GNAS	c.1096G>A (p.A366T)	Heterozygous	De novo	Pseudohypoparathyroidism Ia	103580	No
17464	F	3m-1y	Proband ES	RNASEH2A	c.557G>A (p.R186Q)	Homozygous		Aicardi-Goutieres syndrome 4	610333	Yes
87229	F	3m-1y	Critical-trio ES	POLG	c.1399G>A (p.A467T); c.3483-4_3497del	Compound Het	Inherited; c.3483-4_3497del mother, p.A467T father	POLG-related disorders	Multiple	Yes
84711	M	5-12y	Critical-trio ES	TTC19	c.150_153delCGGC (p.R52fs)	Homozygous	Inherited	Mitochondrial complex III deficiency, nuclear type 2	615157	Yes
30931	F	5-12y	Critical-trio ES	SERAC1	c.202C>T (p.R68X); c.1598C>G (p.P533R)	Compound Het	Inherited; p.P533R mother; p.R68X father	3-methylglutaconic aciduria with deafness, encephalopathy, and Leigh-like syndrome	614739	No
31162	F	12-18y	Critical-trio ES	CBS	c.1007G>A (p.R336H)	Homozygous	Inherited	Homocystinuria, B6-responsive and nonresponsive types	236200	Yes
88384	M	18-40y	Proband ES	NSD1	c.7966C>T (p.Q2656X)	Heterozygous	De novo	Sotos syndrome 1	117550	Yes
96169	M	18-40y	Proband ES	COL4A1	c.1973G>A (p.G658D)	Heterozygous	De novo	Brain small vessel disease with or without ocular anomalies	175780	Yes

Recurrent genes observed in the cohort:

No genes recurred among the 11 diagnosed cases. However, an additional 13 cases were designated as “probably diagnosed” (Table S2). Including these additional cases, the only recurrent genes were COL4A1 and ATP13A2.

COL4A1 causes several autosomal dominant syndromes that have been associated with stroke [Grossi et al., 2020]. Patient 96169 was an adult male with developmental delay, hypotonia, seizures, dysmorphic facies, cataracts and brainstem strokes. Brain MRI was concerning for agenesis of the corpus callosum, porencephalic cysts, and cerebellar hypoplasia. Through proband ES, he was diagnosed with brain small vessel disease with or without ocular anomalies (BSVD1, MIM# 175780, Table 2). Two additional patients were considered probably diagnosed for this condition (Table S2). Patient 87642 was a male fetus for whom proband ES was requested on a sample collected by amniocentesis. The clinical description included intrauterine growth restriction, microcephaly, congenital heart disease, and brain MRI consistent with cerebral hemorrhages and infarcts. Testing showed a c.2879G>T (p.G960V) variant in COL4A1 which was called likely pathogenic for BSVD1.Patient 71712 was a 5-year-old girl with developmental delay, intellectual disability, seizures, dysmorphic features, microcephaly, congenital cataracts and stroke-like episodes. A novel de novo heterozygous c.3555A>G (p.K1185K) VUS was identified in COL4A1 and predicted to affect splicing.

Two patients were considered probably diagnosed with variants in ATP13A2. This gene is associated with Kufor-Rakeb syndrome and spastic paraplegia 78 (MIM# 606693, 617225), autosomal recessive disorders associated with juvenile and adult-onset Parkinsonism, respectively, although neither is associated with stroke. Patient 51411 was an adult male with seizures, apraxia, memory problems, confusion and progressive cognitive decline and history of stroke; proband-only testing revealed a heterozygous c.348-9_351del pathogenic splice-site variant and c.943G>A (p.G315R) VUS of ATP13A2. Parental samples were not available to phase these variants. Patient 20796 was a male infant with seizures, microcephaly, failure-to-thrive, and cataracts. Brain MRI showed evidence of cerebral atrophy, periventricular leukomalacia, and stroke. Proband ES showed a homozygous c.3056delG (p.G1019fs) variant in ATP13A2.

Syndromes not previously associated with Stroke:

In our study, 5/11 fully diagnosed cases involve disorders that, to our knowledge, have not been previously associated with stroke (Table 2). We will highlight two of these cases.

Patient 90361 was a female infant with intrauterine stroke, delayed motor milestones, dyskinetic movements, dilated aorta and acquired obstructive hypertrophic cardiomyopathy. Proband ES revealed a de novo heterozygous c.64C>A (p.Q22K) pathogenic variant in HRAS which is associated with autosomal dominant Costello syndrome (MIM# 218040). Costello syndrome can cause growth failure, intellectual disability, and dysmorphic facial features in addition to other skin and musculoskeletal anomalies. There is no known association between Costello syndrome and stroke, though related disorders Noonan and Neurofibromatosis 1, which also affect the Ras signaling pathway, have been reported with stroke [Robertson et al., 1997; Rukavina et al., 2019].

Patient 76670 was a male infant with premature birth, oligohydramnios, anisocoria, seizures, polycystic renal dysplasia, and recurrent vascular ischemic strokes. Critical-trio ES revealed heterozygous pathogenic c.983G>A (p.R328Q) and c.1964+1G>T variants in PKHD1 which were inherited in trans. Defects of PKHD1 are a cause of autosomal recessive polycystic kidney disease (ARPKD, MIM# 263200). ARPKD typically presents in the neonatal period with oligohydramnios, massively enlarged kidneys, with perinatal death in ~30% of affected newborns. Although autosomal dominant polycystic kidney disease can present with hemorrhagic stroke in young adulthood, the recessive forms have been associated with nonbleeding aneurysms [Gately et al., 2021], or rarely with optic nerve ischemia [Thomas et al., 2003]. Stroke due to sequelae of hypertension cannot be excluded.

Discussion

This study represents one reference laboratory’s experience using clinical ES for patients with stroke. Our study indicates that patients with stroke of all ages, and especially patients with perinatal or pediatric stroke might benefit from ES. We diagnosed nearly 9% of cases, and 25% of infants under the age of 1; an additional 10% of cases received probable diagnoses. In our cohort, P/LP variants associated with COL4A1-related brain small vessel disease was the most frequent finding. Mitochondrial diseases due to biallelic P/LP variants in POLG and TTC19 were found in two children in our study, presenting with ischemic stroke and multiple stroke-like episodes respectively. Variants in TTC19 cause complex III deficiency and have recently been shown to cause recurrent stroke-like episodes with basal ganglia changes [Conboy et al., 2018]. Homocystinuria due to CBS P/LP variants and Aicardi-Goutières syndrome due to RNASEH2A P/LP variants were also identified. Interestingly, we identified a subject with Sotos syndrome due to a pathogenic NSD1 variant, who had a stroke at age 18 years. Sotos syndrome is an overgrowth syndrome which is not classically associated with stroke, however it has recently been associated with arterial thrombosis, hemorrhagic stroke, and subdural hematoma in rare patients [Foster et al., 2019]. Our data indicate that NSD1 evaluation should be included in the genetic work up for stroke.

Importantly, ES also identified P/LP variants in genes not currently known to increase the risk of stroke. Beyond PKHD1 and HRAS discussed earlier, our study found a subject with a neurodegenerative disorder affecting mitochondrial function, SERAC1-related MEGDHEL syndrome (3-methylglutaconic aciduria with deafness-dystonia, hepatopathy, encephalopathy, and Leigh-like syndrome), which is not recognized to be associated with stroke at present. Further studies could elucidate the relationship between SERAC1 and stroke, if any.

There are some important limitations to our study. Stroke is a challenging phenotype for genetic analysis. Foremost, “stroke” may represent a number of pathologies, such as hemorrhage or ischemia, and our dataset had limited clinical details to separate these entities. Especially in the pediatric population, the presence of infection, vasculopathy, congenital heart disease, coagulopathy and other risk factors may influence stroke risk irrespective of the genetic determinants. However, clinicians often must order testing without access to these data. The perinatal period is an excellent example, where the original insult can occur prior to delivery with limited characterization. Our analysis indicates that ES is particularly well suited to assessing stroke in the first year of life. Another limitation is that ES primarily assesses for small insertions/deletions and single nucleotide variants within or near exons. Alternative testing is needed if deep intronic variants, repeat expansions, large copy number variants or other classes of variation are of interest [Kumar et al., 2021].

Given recent guidance encouraging the use of exome and genome sequencing, it is highly likely that the number of patients undergoing these genetic assays will increase rapidly [Manickam et al., 2021]. Ongoing analysis of the generated data will be essential to assessing stroke risk for patients with previously described genetic disorders, and eventually to help identify new heritable stroke syndromes.

Conclusions

In this study we retrospectively analyzed 124 patients with personal history of stroke referred for exome sequencing. Nearly one fifth of patients received a complete or probable diagnosis, and one quarter of patients less than one year old were fully diagnosed. Several individuals have Mendelian disorders that are not typically associated with stroke, including Sotos syndrome, autosomal recessive polycystic kidney disease, and Costello syndrome. This analysis indicates that patients with stroke, especially at younger ages, may to benefit from broad genetic testing.

Data Availability

All data described in this study are provided within the article and supplementary materials. Additional de-identified clinical data is available upon request to the corresponding author.