Whole exome sequencing in patients with white matter abnormalities

Abstract

Here we report whole exome sequencing (WES) on a cohort of 71 patients with persistently unresolved white matter abnormalities with a suspected diagnosis of leukodystrophy or genetic leukoencephalopathy. WES analyses were performed on trio, or greater, family groups. Diagnostic pathogenic variants were identified in 35% (25/71) of patients. Potentially pathogenic variants were identified in clinically relevant genes in a further 7% (5/71) of cases, giving a total yield of clinical diagnoses in 42% of individuals. These findings provide evidence that WES can substantially decrease the number of unresolved white matter cases.

Introduction

Patients with white matter abnormalities in the central nervous system (CNS) may have one of over a hundred genetic disorders, including the leukodystrophies (Supplemental Table S1).¹ Over the past two decades MRI pattern recognition has transformed the diagnosis of leukodystrophies.^2,3 Despite these advances, the breadth of conditions that present as a possible leukodystrophy continues to challenge even the most astute clinician.⁴ Nearly half of these patients will remain unresolved, resulting in a prolonged diagnostic odyssey for affected families.^5–7

A number of recent reports have provided evidence that whole exome sequencing (WES) can resolve previously intractable genetic disorders, with diagnostic yields ranging from 16% to 53%.^8–19 Given the unmet diagnostic need amongst patients with leukodystrophy, and the potential for agnostic next generation sequencing (NGS) to clarify these cases, we performed whole exome sequencing (WES) on a cohort of 71 patients referred to the Myelin Disorders Bioregistry Project (MDBP) for unresolved leukoencephalopathy of presumed genetic etiology. These patients were collected prospectively from 8.1.2009 to 7.31.2013 in the MDBP or the Amsterdam Database of Unclassified Leukoencephalopathies with approval from the institutional review boards at all collaborating institutions including Children’s National Medical Center, the Baylor Neurogenetic Institute, and VU University Medical Center.

Methods

Cohort description

A total of 191 persistently unresolved cases were identified during the course of the study. 101 patients were diagnosed using MRI pattern recognition followed by biochemical or other molecular approaches. These testing approaches included lysosomal enzymes, very long chain fatty acids, specific electron transport chain or mitochondrial enzyme assays, urine organic acids, microarray testing of copy number variations, gross chromosomal abnormality testing by karyotype or microarry, plasma amino acids, cerebrospinal neurotransmitters and alpha interferon, urine mucopolysaccaride or sialic acid testing, and targeted molecular testing, for example for EIF2B1–5, PLP1 or GFAP sequencing based on MRI pattern recognition. Of the 90 persistently unsolved cases, 19 were excluded from WES testing: nine families obtained access to WES at other facilities, and an additional 10 families were excluded because DNA quality for all members of the trio did not meet stringency criteria and attempts to collect additional samples during the course of the study were unsuccessful. Seventy-one families remained for which high quality samples were available for complete trios. These 30 female and 47 male individuals all had abnormal white matter signal on neuroimaging. Individuals ranged in age from 3 years to 26 years at the time of sequencing, but symptom onset ranged from birth to 19 years (see complete Supplemental Case Reports available at http://imb.uq.edu.au/download/Vanderver_AON_2016.case_reports.pdf). Ethnicities were varied and included individuals of mixed and northern European descent, as well as African American, Arab, African, Asian, and Latin American origin. Radiological images and clinical summaries are provided in Supplemental Case Reports.

WES sequencing was performed at the Queensland Centre for Medical Genomics. Exomes were captured using either Illumina Nextera Rapid Capture kit or the SeqCap EZ Human Exome Library v.3.0. Captured libraries were sequenced on an Illumina HiSeq 2000 (2 × 100nt) or on an Illumina NextSeq 500 (2 × 150nt). WES sequencing was performed such that a minimum of 80% of targeted bases were sequenced to a read depth of 20× or greater (average: 88%). Reads were aligned to the reference human genome (GRCh37) and pedigree informed variant calling was performed using the Real Time Genomics (RTG) integrated analysis tool rtgFamily v3.2.²⁰ All variants were annotated using SnpEff v3.4²¹ and filtered using data from the SnpEff GRCh37.72 database, dbSNP138, and dbNSFP v2.4.

We utilized a custom-built variant annotation and interpretation interface to identify possible causal mutations in each case, incorporating evidence including minor allele frequency, conservation, predicted pathogenicity, disease-association (in public databases or the published literature), established or predicted biological function, confirmation with Sanger sequencing and familial segregation (Supplemental Tables S2–S4).²² Cases with variants in known disease genes meeting the ACMG criteria for pathogenic or likely pathogenic (see Supplemental Case Reports), and whose clinical features were concordant with the established gene:disease relationship were classified as diagnostically resolved.

Results

In this cohort we were able to unambiguously resolve 25 cases (Table 1, Supplemental Table S2, and Supplemental Case Reports). In three cases we were able to confirm pathogenicity with downstream biochemical testing. For example, we identified a compound heterozygous mutation in TERT (MIM:187270) in a patient that presented with white matter changes, frequent infections, mild developmental delay and hypogammaglobulinemia which was validated by Flow-FISH telomere length analysis and confirmed a diagnosis of atypical Dyskeratosis Congenita with hypomyelination (MIM:613989) (Figure 1, Supplemental Case Reports, LD_0607.0).²³ Likewise, an individual with a compound heterozygous mutation in GLB1 (MIM:611458) had confirmatory lysosomal enzyme testing (Supplemental Case Reports, LD_0846.0), and an individual with mutations in ATM (MIM:607585) had confirmatory elevated alpha-fetoprotein levels (Supplemental Case Reports, LD_0678.0).

Table 1.

Cases with pathogenic variants identified by exome sequencing in classical leukodystrophy genes.

Family	Gene	Zygosity	Protein	Classification
LD_0139	TUBB4A	Het, de novo	p.Arg391His	Likely pathogenic
LD_0181	DARS	Het	p.Arg494Gly	Likely pathogenic
		Het	p.Arg460His	Likely pathogenic
LD_0604	POLR3B	Het	p.Glu271fs	Pathogenic
		Het	p.Val523Glu	Pathogenic
LD_0672	TUBB4A	Het, de novo	p.Val180Gly	Likely pathogenic
LD_0764	EIF2B5	Het	p.Gln562*	Pathogenic
		Het	p.Arg339Trp	Pathogenic
LD_0774	POLR1C	Het	p.Lys295del	Likely pathogenic
		Het	p.Cys146Arg	Likely pathogenic
LD_0869	EIF2B2	Het	p.Gly200Val	Pathogenic
		Het	p.Glu213Gly	Pathogenic

Key: Het –Heterozygous; Hom – Homozygous; rs – Reference Single Nucleotide Polymorphisms

Figure 1. MRI and pathogenic variants for three cases.

A) MRI of individual LD_0607.0 – a male of mixed European descent with a multisystem disorder characterized by elevated creatine kinase, recurrent infection with hypogammaglobulinemia, dyskeratosis congenita, and mild transaminase abnormalities. MRI revealed moderate cerebellar atrophy and diffuse multifocal white matter changes. Follow-up MRI one year later showed unchanged T2 hyperintensities. The variants found in TERT in this patient were classified as pathogenic per the ACMG guidelines and the diagnosis was supported by telomere length analysis (data not shown). B) Schematic of the TERT protein showing heterozygous variants identified in LD_0607.0. Predicted domains: GQ, GQ motif; QFP, QFP motif; RT, Reverse transcriptase domain. C) Clustalo alignment of vertebrate homologs of TERT showing conservation of mutated residues. D) MRI of individual LD_0756.0 – male of Turkish descent with motor delays were noted at birth who abruptly decompensated at 7 months of age, and a history of ataxia, hypotonia, and spasticity. MRI at 3 years and 6 months of age was significant for signal abnormality of the supratentorial white matter with sparing of the U fibers, a swollen appearance to the corpus callosum, involvement of the cerebellar white matter, and the brain stem. This pattern has been seen in previously published cases and supports the SDHB variant categorization as potentially pathogenic. E) Schematic of the SDHB protein showing a homozygous variant identified in LD_0756.0. Predicted domains: MTS, mitochondrial targeting signal; SDH, Succinate dehydrogenase domain. F) Clustalo alignment of vertebrate homologs of SDHB showing conservation of mutated residues. G) MRI of individual LD_0286.0B – male of mixed-European descent with leukoencephalopathy and a history of sensorineural hearing loss, developmental delay and febrile seizures. MRI is significant for bilateral temporal lobe cysts, small corpus callosum, and periatrial white matter abnormalities. Hearing loss and other clinical manifestations were consistent with the phenotype reported for mutations in RMND1, and the variant was classified as likely pathogenic based on supporting evidence. H) Schematic of the RMND1 protein showing heterozygous variants identified in LD_0286. Predicted domains: MLS, mitochondrial localization signal; DUF155, domain of unknown function 155; CC, coiled-coil domain; and TM, trans-membrane domain. I) Clustalo alignment of vertebrate homologs of RMND1 showing conservation of mutated residues.

Nine of the twenty-five cases had mutations in genes associated with disorders classically defined as leukodystrophies³ (Table 1 & Supplemental Table S1): two patients were identified with TUBB4A (MIM:602662) related hypomyelination (Hypomyelination with atrophy of the basal ganglia and cerebellum [MIM:612438])²⁴; two patients were identified with early onset Vanishing White Matter Disease (MIM:603896) (genotype EIF2B2 [MIM:606454] and EIF2B5 [MIM:603945])^{25, 26}; three families were identified with t-RNA synthetase disorders (AARS [MIM:601065] and DARS [MIM:603084])^27–29; and two families identified with a POLR3-related disorder (POLR3B [MIM:614366] and POLR1C [MIM:610060]) (Table 1, Supplemental Table S2 and Supplemental Case Reports).³⁰ The remaining individuals had mutations in genes associated with genetic leukoencephalopathies. In these cases expert review confirmed that the clinical presentation and MR imaging was consistent with published phenotypes. These findings are consistent with the estimation that a majority of disorders associated with abnormal white matter on neuroimaging are not classic leukodystrophies.¹ This suggests that testing of leukodystrophy-associated genes on NGS panels may have limited diagnostic efficacy (predicted to be only 13% in this cohort), which may outweigh the perceived cost benefit and limiting exposure to incidental findings.

In a further four cases we identified one or more potentially damaging variants of uncertain significance (VUS) that did not reach the strict burden of proof required to be classified as pathogenic or likely pathogenic. In each of these cases the neuroradiological findings, clinical features and familial segregation of the variants in these individuals were consistent with the published phenotype (Table 2, Supplemental Table S3, and Supplemental Case Reports). We therefore classified these variants as potentially pathogenic and considered the cases clinically resolved by expert assessment. This included cases with variants in AMPD2, FLNA, and NDUFA2 (Supplemental Table S4). This also included one case (LD_0675) where the individual was reported as part of cohort describing a novel disease due to mutations in AARS2.²⁷

Table 2.

Cases with pathogenic variants identified by WES in genes not associated with leukodystrophy.

Family	Gene	Zygosity	Protein	Classification
LD_0106	GRIN1	Het, de novo	p.Arg865Cys	Likely pathogenic
LD_0115	AARS	Het	p.Arg751Gly	Pathogenic
		Het	p.Lys81Thr	Pathogenic
LD_0119	KCNT1	Het, de novo	p.Phe932Ile	Pathogenic
LD_0157	SZT2	Het, de novo	p.Pro1833fs	Pathogenic
		Het	p.Gly2306Arg	Likely pathogenic
LD_0158	CNTNAP1	Hom	p.Arg388Pro	Likely pathogenic
LD_0232	MRPS22	Het	p.Lys248fs	Pathogenic
		Het	p.Arg191Gln	Likely pathogenic
LD_0286	RMND1	Hom	p.Asn238Ser	Likely pathogenic
LD_0333	CNTNAP1	Het	p.Arg107*	Pathogenic
		Het	p.Cys323Arg	Likely Pathogenic
LD_0358	STXBP1	Het, de novo	p.Arg367*	Pathogenic
LD_0366	GATAD2B	Het, de novo	p.Gln274*	Pathogenic
LD_0463	ALS2	Het	p.Arg1139*	Pathogenic
		Het	p.Gly1083Glu	Pathogenic
LD_0607	TERT	Het, de novo	p.Arg819Cys	Pathogenic
		Het	p.Val664Met	Pathogenic
LD_064619	NDUFS7	Hom	p.Arg135Cys	Likely pathogenic
LD_0678	ATM	Het	p.Leu275fs	Pathogenic
		Het	p.Lys2756*	Pathogenic
LD_0755	SDHAF1	Hom	p.Arg55Pro	Pathogenic
LD_0756	SDHB	Hom	p.Asp48Val	Pathogenic
LD_0846	GLB1	Het	p.Arg196Ser	Pathogenic
		Het	p.Tyr240His	Pathogenic
LD_0857	AARS	Hom	p.Arg751Gly	Pathogenic

Key: Het –Heterozygous; Hom – Homozygous;

A final case had a de novo variant in FUS (MIM:205100) classified as pathogenic by ACMG criteria, but because this gene has previously only been associated with only juvenile or adult onset Amyotrophic Lateral Sclerosis (ALS), this was not considered an unambiguous resolution (Table 2, Supplemental Table S3 & Supplemental Case Reports). However, because mutations in other ALS associated genes are associated with early hypomyelination (including ALS2 [MIM:205100] in this cohort), and because the de novo finding segregated in this family, it was classified as a potentially pathogenic variant and a clinically resolved case.

To investigate the burden of actionable incidental findings that may be identified during trio-based WES investigation of rare genetic disorders, the Illumina Clinical Services Laboratory screened all 56 adult and 49 pediatric ACMG recommended genes for potential incidental variants.³¹ This analysis revealed pathogenic or likely pathogenic variants in 3 of the 71 families screened (Table 3). The identified variants were restricted to KCNQ1 (MIM:607542) associated with long QT syndrome and SDHB (MIM:115310) associated with hereditary paragangliomas (Table 2, Supplemental Table S3 and Supplemental Case Reports). Interestingly, mutations in SDHB are also now associated with autosomal recessive succinate dehydrogenase deficiency associated leukoencephalopathy, although the lack of a second mutation in LD_0315 precluded definitive association of this genotype with the patient’s phenotype.³² We found less than one known pathogenic or likely pathogenic variant per 46 adult exomes analyzed, suggesting that the impact of incidental findings is likely to be minimal, especially when weighed against the potential benefits of a successful genetic diagnosis in families with severe, life-threatening neurologic illnesses (Table 4).

Table 3.

Cases with potentially pathogenic variants identified with whole exome sequencing.

Family	Gene	Zygosity	Protein	Classification
LD_0493	FLNA	Hem, de novo	p.Leu2271Arg	VUS
LD_0664	FUS	Het	p.Gly500fs	Pathogenic
LD_0673	AMPD2	Hom	p.Arg843His	VUS
LD_0675	AARS2	Het	p.Gly965Arg	Likely pathogenic
		Het	p.Glu405Lys	VUS
LD_0821	NDUFA2	Het	p.Lys45Thr	VUS
		Het	p.Thr75fs	VUS

^*Key: Het –Heterozygous; Hom – Homozygous;

Table 4.

Incidental Findings Identified in Cohort

	Number of individuals screened	Number of individuals with incidental findings	Reported Incidental Finding
Unaffected Adults	142	3 (2.1%)	KCNQ1 (NM_000218.2) c.514G>A KCNQ1 (NM_000218.2) c.1189C>T SDHB (NM_003000.2) c.541-2A>G
Affected Children	79	3 (3.7%)	As above
Unaffected Siblings	10	0	NA

Discussion

Using an intention to treat analysis,³³ and if the combined initial cohort of 191 families is considered in which 101 families achieved a diagnosis through standard approaches, the use of WES approach would result in an ~20% diagnostic increase. This yields an overall rate of diagnosis of ~72% for the combination of standard and WES approaches. Clinical integration of WES (or whole genome sequencing) therefore, may decrease the number of patients with unsolved genetic white matter disorders from 50% to less than 30%. Taking into consideration the clinical and psychosocial costs of prolonged diagnostic odysseys in these families, this is substantial.

Additionally, while the clinical utility of WES as measured by changes to patient care was not addressed in this study, it should be noted that in several cases the results of WES directly influenced clinical care. For example, patients with ATM and TERT mutations were both referred to specialist clinics where they now undergo oncologic monitoring, and the patient with a de novo KCNT1 mutation was treated with a potassium channel anticonvulsant to control refractory epilepsy.³⁴

The use of WES in large cohorts enables the identification of sequence variants of varying degrees of clinical certainty. ACMG criteria classify the spectrum of identified variants into four tiers; pathogenic, likely pathogenic, variant of unknown significance, or unresolved.³⁵ However, our study, in which 4 of 31 cases were resolved with MRI pattern recognition^{2, 3, 36, 37} and clinical review of the identified variant, suggests that a variant classification system that takes into account clinical context and downstream clinical evaluation and testing (e.g. MRI interpretation) should be considered.

In summary, WES has the potential to decrease the number of unsolved cases of leukodystrophy and genetic leukoencephalopathies. Additional research is needed establish the potential value of NGS as a first-line diagnostic tool, and to assess the comparative effectiveness of WES, WGS and targeted panels in this disease population.