GRIN2A mutations cause epilepsy-aphasia spectrum disorders

Letter to the editor

The epilepsy-aphasia syndromes (EAS) are a group of rare, severe epileptic encephalopathies of unknown etiology with a characteristic EEG pattern and developmental regression, particularly affecting language. Rare pathogenic deletions that include GRIN2A have been implicated in neurodevelopmental disorders. We sought to delineate the pathogenic role of GRIN2A in 519 epileptic encephalopathy probands with diverse epilepsy syndromes. We identified four probands with GRIN2A variants that segregated with the disorder in their families. Strikingly, all four families presented with EAS, accounting for 9% of epilepsy-aphasia cases. We did not detect pathogenic variants in other epileptic encephalopathies (n=475), nor in 81 probands with benign childhood epilepsy with centro-temporal spikes. We report the first monogenic cause for EAS. GRIN2A mutations are restricted to this group of patients, with important ramifications for diagnostic testing and treatment, and novel insights into the pathogenesis of this debilitating group of conditions.

The epileptic encephalopathies are a severe group of disorders characterized by seizures and abundant epileptiform activity that contribute to cognitive and behavioral impairment¹. The epileptic encephalopathies comprise a range of electroclinical syndromes with characteristic ages of onset, clinical and EEG manifestations. Two syndromes with overlapping manifestations have the remarkable EEG signature of continuous spike-wave during slow wave sleep (CSWS) in which the non-REM sleep EEG shows virtually continuous (≥85%) high voltage bilateral slow spike wave activity that largely remits on awakening. In Landau-Kleffner syndrome (LKS), children who were previously normal or isolated language delay present with an acquired epileptic aphasia; focal motor seizures occur in 70% of cases and are usually easily controlled. In contrast, in the syndrome of epileptic encephalopathy with continuous spike-wave during slow wave sleep (ECSWS), prior development is delayed in half the children and refractory epilepsy with multiple seizure types is usual. Regression is more global with language, behavior and motor impairment². MRI brain studies are often normal or may show a malformation of cortical development such as perisylvian polymicrogyria.

In clinical practice, there are patients who do not meet the criteria on EEG or clinical grounds for LKS and ECSWS, usually because their EEG abnormalities do not occupy 85% of slow sleep, yet they have significant language or learning difficulties which may fluctuate in severity. There is debate whether <85% of bilateral epileptiform activity in non-REM sleep is diagnosable as CSWS or whether it should be regarded as an intermediate epilepsy-aphasia disorder (IEAD)³. These disorders can be conceptualized as falling along a spectrum with LKS, ECSWS at the severe end, IEAD in the middle and benign childhood epilepsy with centro-temporal spikes (BECTS) at the mild end of the spectrum³. BECTS is the most common focal epilepsy syndrome in childhood and occurs in normal children who present with focal motor rolandic seizures. The EEG shows unilateral or bilateral centro-temporal spikes that are activated by sleep but do not show the almost continuous bilaterally synchronous pattern of CSWS, and the children do not show cognitive decline. The presence of subtle oral dyspraxia has been noted in some patients with BECTS⁴.

Until recently there has been scant evidence for a genetic etiology for the disorders of the epilepsy-aphasia spectrum. To date, only four families have been reported with monogenic inheritance of rolandic epilepsy and speech or language difficulties. We reported an autosomal dominant family in 1995 with the syndrome of autosomal dominant rolandic epilepsy with speech dyspraxia (ADRESD)⁵. An additional three-generational family with a strikingly similar phenotype was reported more recently⁶. Finally a family with dysphasia and epilepsy with generalized and focal manifestations was reported⁷. A causal gene has not been implicated in these families. Conversely, a fourth family, presenting with X-linked rolandic epilepsy, oral and speech dyspraxia and intellectual disability (ID) was identified with a gain-of-glycosylation SRPX2 mutation⁸. Besides a SRPX2 mutation in an unrelated proband with perisylvian polymicrogyria and rolandic seizures and female relatives with mild ID, no additional SRPX2 variants in epilepsy-aphasia phenotypes have been described.

Clinical genetic studies of probands with BECTS or EAS provide little support for genes of major effect. Investigation of relatives up to three degrees of relatedness to probands with BECTS or the epilepsy-aphasia spectrum suggest that complex inheritance is most likely with febrile seizures being the most common phenotype in relatives of probands^{3, 9}.

While there has been strong contention that the epilepsy-aphasia syndromes have an immune basis, partly due to their resolution with high dose steroids, a genetic etiology is supported by the rare familial forms described. Furthermore, recent evidence for a genetic etiology has come from copy number variant (CNV) studies. An excess of rare CNVs was noted in a cohort of LKS and CSWS probands¹⁰, including a single LKS proband with a 16p13 deletion containing one gene, GRIN2A (NM_000833.3)¹⁰. Furthermore, three children with complex dysmorphic phenotypes were reported with 16p13 deletions that included GRIN2A¹¹. GRIN2A encodes the NR2A subunit of the N-methyl-D-aspartate (NMDA) receptor, a neurotransmitter-gated ion channel that mediates excitatory transmission in the mammalian brain, making it an attractive candidate for epileptogenesis. GRIN2A mutation screening in 127 probands with epilepsy or an abnormal EEG and/or ID detected two pathogenic mutations: a nonsense mutation segregating with epilepsy or an abnormal EEG in three family members and a de novo missense mutation in a patient with severe early-onset epileptic encephalopathies ¹². Furthermore, a missense mutation was recently reported in a single proband in a large exome sequencing cohort with ID¹³. While these observations strongly support a role for GRIN2A in epilepsy and ID, no clear genotype-phenotype correlations have emerged. Therefore, we sought to delineate the phenotypic spectrum of GRIN2A mutations by screening a large cohort of patients with epileptic encephalopathy.

We performed high-throughput sequence analysis of GRIN2A in 519 probands with a range of epileptic encephalopathies (Table 1). As part of a larger study ¹⁴ we performed targeted gene capture of 18 genes associated with epilepsy, including GRIN2A. Briefly, we re-sequenced all exons and flanking 5 base pairs using molecular inversion probes (MIPs), highly multiplex PCR and next generation sequencing as described previously with minor exceptions (online methods)¹⁵. Using this approach we achieved, on average, 98% coverage (>25X) across GRIN2A for all probands.

We identified four probands with GRIN2A mutations, each of which was confirmed by Sanger sequencing. Segregation analysis in additional family members showed that each variant segregated in an autosomal dominant manner (Table 2, Figure 1). These GRIN2A variants are not present in 6500 control exomes (see Resources). Two families (A, C) carried the same c.1005-1C>T variant, affecting the highly conserved donor splice site. Genotyping of microsatellite makers and a rare SNV flanking this GRIN2A mutation revealed an identical haplotype in these families, suggesting a common founder mutation (Supplementary Figure 1). The c1005-1C>T variant was predicted in silico to cause skipping of exon four during pre-mRNA splicing, resulting in the removal of 593 exonic nucleotides from the mature transcript and thus a frameshift mutation, Phe139Ilefs*15 (predicted) (see Supplementary Table1). We tested for the presence of a rare exonic SNV (rs61753382), encompassed by the common haplotype in affected individuals, in the RNA transcripts of three affected individuals from both families. We detected monoallelic expression of the wild-type variant, suggesting nonsense mediated decay of the mutant transcript (Supplementary Figure 2).

We detected a p. Met1Thr variant in family B. The alteration of this translation start codon is likely to have detrimental effects on GRIN2A protein synthesis, resulting in either complete absence of product due to failure of translation initiation at the start codon, or a truncated protein stemming from translation initiation at an alternate start codon. We were unable to test this as proband RNA was unavailable.

Finally we describe a p.Thr531Met variant that affects a highly conserved residue (as predicted by high GERP and Grantham scores), that is predicted to be probably damaging by Polyphen2 and SIFT (Table 2). This variant is located in the extracellular ligand-binding domain of NR2A. Specific sites within this domain are known to influence gating and kinetic properties of NMDA receptors^{16, 17}. We assessed the effect of the p.Thr531Met mutation on NR2A function by co-expression with NR1 in COS-7 cells to form a mutant heteromeric NMDA receptor. A resultant shift in NMDA receptor kinetics was observed by single channel recordings with a four-fold increase in mean open time of the mutant channels (36.7±2.5 ms ; n=2299 channel events), as compared to the wild-type channels (9.1±0.2 ms; n=6715 channel events) (P<0.0001, Mann Whitney test, two-tailed) (Figure 2). This novel variant displayed similar clinical and functional consequences to missense mutations in the same domain in a parallel study by Lesca and colleagues (this issue).

The c1005-1C>T and Met1Thr variants likely cause disease as a result of haploinsufficiency of the NR2 subunit of the NMDA receptor, possibly by aberrant NMDA receptor composition or distribution in the brain. Furthermore, we show that the p. Thr531Met variant has a profound effect on NMDA receptor kinetics. Given the pathogenic effect of these mutations and their segregation with the disorder, we conclude that the GRIN2A mutations ascribe causality in these families.

Remarkably, all four GRIN2A positive families presented with EAS, yielding a 9% (4/44) mutation rate in patients with this group of EE. No additional pathogenic variants were detected in the remaining epileptic encephalopathy phenotypes (Table 1). In the 40 remaining EAS patients, we performed array-CGH using a custom microarray with probes spanning GRIN2A at an average density of one probe every ~350bp. No copy number alterations were detected.

Given that BECTS lies at the mild end of the EAS, we next screened 81 probands with BECTS for GRIN2A variants using Sanger sequencing. No additional pathogenic variants were identified.

There were 16 subjects with GRIN2A mutations. Segregation was perfect in the 7 affected members of the original family with autosomal dominant rolandic epilepsy with speech dyspraxia (Family A, Fig. 1)⁵. The same mutation was found in a father-son pair with ECSWS (Family C). Interestingly the GRIN2A mutations were associated with a range of EAS phenotypes including LKS, ECSWS and IEAD (Table 3). All individuals with LKS and ECSWS showed CSWS on EEG studies. Individuals with IEAD had not had a sleep EEG performed to detect CSWS. Affected family members had a complex phenotype including epilepsy (14/16), speech and language difficulties (16/16). While intellectual disability occurred in 6/16 mutation carriers, a further two were of borderline intellect (Supplementary Table 2).

Previous cases implicating GRIN2A have not identified a consistent epilepsy phenotype but have shared features with our cases. Four cases with 16p13 microdeletions including GRIN2A have been reported; one had LKS¹⁰. The remaining three were more complex with dysmorphic features and moderate to severe ID; two were non-verbal and only one walked independently ³. All had seizures; one had atypical benign partial epilepsy, which is part of the EAS. One had rolandic seizures without regression and EEG studies were not available. In two patients, eyelid myoclonias were noted which is somewhat atypical for EAS. Two had an EEG pattern suggestive of CSWS. In another study, a three generation family with a translocation disrupting GRIN2A was associated with childhood and adolescent onset convulsions in the setting of learning difficulties or ID. There was no suggestion of CSWS on their EEG studies and no epilepsy syndrome was determined.

We conclude that GRIN2A mutations are highly predictive of EAS that include LKS, ECSWS and IEAD. Furthermore, in a separate study Lesca and colleagues report GRIN2A mutations in 20% of LKS, ECSWS and atypical rolandic epilepsy with speech impairment confirming the importance of GRIN2A to the EAS (this issue). Of note, we did not detect any GRIN2A variants in 475 probands with other epileptic encephalopathy phenotypes, or in 81 probands with BECTS. Furthermore, in a large series (n= 1703) of autism probands no GRIN2A mutations were identified¹⁸. These results demonstrate that the genetic etiology of EAS may well be distinct, an observation that balks the current trend towards an overlapping etiology for neurodevelopmental disorders. We hypothesize that altered NMDA receptor activity due to GRIN2A haploinsufficiency or missense mutations results in aberrant ion flux and disruption of the downstream signaling cascade. The role of NMDA receptor aberration and its potential role in the corticothalamic network disrupted in slow sleep will be an important area of future research. This study is the first to detect a monogenic cause for epilepsy-aphasia syndromes with a mutational rate of 9%. These results strongly suggest that GRIN2A diagnostic testing is warranted in patients with epilepsy-aphasia and will enhance prognostic and genetic counseling for families.

Online methods

Supplementary Material