GRIN2A-related disorders: genotype and functional consequence predict phenotype

Abstract

Alterations of the N-methyl-d-aspartate receptor (NMDAR) subunit GluN2A, encoded by GRIN2A, have been associated with a spectrum of neurodevelopmental disorders with prominent speech-related features, and epilepsy. We performed a comprehensive assessment of phenotypes with a standardized questionnaire in 92 previously unreported individuals with GRIN2A-related disorders. Applying the criteria of the American College of Medical Genetics and Genomics to all published variants yielded 156 additional cases with pathogenic or likely pathogenic variants in GRIN2A, resulting in a total of 248 individuals. The phenotypic spectrum ranged from normal or near-normal development with mild epilepsy and speech delay/apraxia to severe developmental and epileptic encephalopathy, often within the epilepsy-aphasia spectrum. We found that pathogenic missense variants in transmembrane and linker domains (misTMD+Linker) were associated with severe developmental phenotypes, whereas missense variants within amino terminal or ligand-binding domains (misATD+LBD) and null variants led to less severe developmental phenotypes, which we confirmed in a discovery (P = 10-6) as well as validation cohort (P = 0.0003). Other phenotypes such as MRI abnormalities and epilepsy types were also significantly different between the two groups. Notably, this was paralleled by electrophysiology data, where misTMD+Linker predominantly led to NMDAR gain-of-function, while misATD+LBD exclusively caused NMDAR loss-of-function. With respect to null variants, we show that Grin2a+/- cortical rat neurons also had reduced NMDAR function and there was no evidence of previously postulated compensatory overexpression of GluN2B. We demonstrate that null variants and misATD+LBD of GRIN2A do not only share the same clinical spectrum (i.e. milder phenotypes), but also result in similar electrophysiological consequences (loss-of-function) opposing those of misTMD+Linker (severe phenotypes; predominantly gain-of-function). This new pathomechanistic model may ultimately help in predicting phenotype severity as well as eligibility for potential precision medicine approaches in GRIN2A-related disorders.

Figure 1

Distribution of variants. (A) Pathogenic or likely pathogenic null variants (red bars) are spread over nearly the entire gene. However, according to ACMG criteria, the last exon 14 is spared, which encodes nearly the complete C-terminal domain. Null variants in healthy gnomAD controls (black bars) occur primarily in the last exon 14 (probability loss-of-function intolerance 1.00 in ExAC). (B) Pathogenic or likely pathogenic missense variants (red bars) cluster in regions of GRIN2A encoding functionally important domains (S1 and S2 ligand binding domains as well as M1–M4 transmembrane domains and linker regions). The density of missense variants in healthy gnomAD controls (MAC = 2, black bars) is highest in the intracellular C-terminal domain.

Figure 2

Distribution of phenotypes. Individuals with GRIN2A-related disorders display a broad range of phenotype severity and expressivity with respect to (A) intellectual outcome, (B) epilepsy, (C) EEG patterns, and (D) speech or language impairments.

Figure 3

Severity of ID/DD. Comparison of severity of ID/DD in carriers of variants in different protein domains. (A) Missense (blue) and null (truncating) variants (yellow). (B) Missense variants in different protein domains in the order of the linear amino acid sequence and truncating variants (far right). Here, variants that were inherited are coloured black, de novo variants are red and unknown variants are grey. Violins are plotted to have the same maximum width. Bottom, middle and top of boxplots within violins show the 1st, 2nd and 3rd quartiles of the data; whiskers maximally extend to 1.5× interquartile range.

Figure 4

Phenotypes correlated with protein domains. Comparison of phenotypes associated with variants in different protein domains (Fisher’s exact test). Phenotype differences being significant after Bonferroni multiple testing correction for 17 × 2 tests are labelled red. For each phenotype, OR with 95% CI (grey/red bars) and number of patients with the phenotype and number of patients for whom the phenotype was assessed are shown. For clarity, OR and CIs are cut at ±1.7. (A) Comparison mis_ATD+LBD or null variants with mis_TMD+Linker. CSWS = continuous spike-and-wave during slow-wave sleep; CTS = centrotemporal spikes.

Figure 5

Variance of ID/DD phenotype in individuals with the same genetic variant. ID/DD phenotypes (y-axis) of all recurrent GRIN2A genetic variants are of similar degree suggesting that the same variant leads to similar ID/DD phenotypes despite considerable phenotype expressivity.

Figure 6

No compensation by GluN2B. GluN2B-mediated currents do not increase to compensate for GluN2A deficiency in cortical neurons from a Grin2a knock-out rat. (A) Western blot and quantification confirming the absence of GluN2A expression in Grin2a^−/− neurons, and an intermediate expression level in Grin2a^+/− neurons at 15 DIV. Tukey’s test reveals a significant difference between Grin2a^+/+ versus Grin2a^+/− (P = 0.0196) and versus Grin2a^−/− (P = 0.0014). Grin2a^+/+: n = 6; Grin2a^+/−: n = 5; Grin2a^−/−: n = 5. (B) Western blot and quantification confirming no changes in GluN2B expression in either Grin2a^+/− or Grin2^−/− neurons compared to Grin2a^+/+ neurons at 15 DIV. Grin2a^+/+, n = 6; Grin2a^+/−: n = 5; Grin2a^−/−: n = 5. (C) NMDA (150 µM) evoked currents were measured in cortical neurons of the indicated genotypes and periods of culture. Currents were calculated and normalized to cell capacitance to give a value for the current density within the neuron. Two-way ANOVA reports a significant developmental stage effect (P < 0.0001) and a significant genotype effect (P = 0.013) as well as a significant interaction between the two (P = 0.0059). Sidak’s post hoc test reveals a significant difference between Grin2a^+/+ versus Grin2a^+/− (P = 0.0007) and versus Grin2a^−/− (P = 0.0006). Grin2a^+/+: n = 38 (7–8 DIV), 38 (15–16 DIV) cells, eight animals; Grin2a^+/−: n = 40 (7–8 DIV), 35 (15–16 DIV) cells, nine animals; Grin2a^−/−: n = 48 (7–8 DIV), 31 (15–16 DIV) cells, 10 animals. (D) NMDA (150 µM) evoked currents were measured in cortical neurons of the indicated genotypes and periods of culture before and after the application of the GluN2B-selective antagonist ifenprodil (3 µM). The ifenprodil-sensitive current was calculated and normalized to cell capacitance. Two-way ANOVA reports a significant developmental stage effect (P < 0.0001) but no significant genotype effect (P = 0.880) nor a significant interaction between the two (P = 0.154). Grin2a^+/+: n = 13 (7–8 DIV), 13 (15–16 DIV) cells, four animals; Grin2a^+/−: n = 13 (7–8 DIV), 11 (15–16 DIV) cells, five animals; Grin2a^−/−: n = 17 (7–8 DIV), 16 (15–16 DIV) cells, five animals. (E) At 15–16 DIV the percentage inhibition of NMDA (150 µM) evoked currents by ifenprodil (3 µM) was significantly greater (Tukey’s test) in Grin2a^−/− neurons compared to Grin2a^+/+ neurons (P < 0.0001) and Grin2a^+/− neurons (P = 0.0038). DIV = day in vitro.