Biallelic LGI1 and ADAM23 variants cause hippocampal epileptic encephalopathy via the LGI1-ADAM22/23 pathway

Abstract

Monoallelic pathogenic variants in LGI1 cause autosomal dominant epilepsy with auditory features with onset in childhood/adolescence. LGI1 is a secreted neuronal protein, functions as a ligand for ADAM22/23, and regulates excitatory synaptic transmission and neuronal excitability in the brain. While biallelic ADAM22 variants cause developmental and epileptic encephalopathy (DEE), the whole picture of LGI1-ADAM22/23 pathway-related diseases remains incompletely understood. Through international genetic data sharing, we identified the first ultra-rare biallelic LGI1 variants in six individuals from four consanguineous families. Affected individuals presented DEE with neonatal/infantile-onset epilepsy (n = 6/6), global developmental delay/intellectual disability (n = 6/6) and infant/premature death (n = 5/6). Brain MRI showed mild cerebral atrophy in a subset of patients (n = 3/6). Functional analyses revealed that all LGI1 variants result in reduced secretion and ADAM22-binding. Residual LGI1 function levels correlated with clinical severity, ranging from infantile lethality to intermediate phenotypes. Further, we observed epileptic discharges from the isolated whole hippocampus of Lgi1-/- knockout mice, experimentally modelling the hippocampal origin of LGI1-related epilepsy. Automated behavioural analysis of a mouse model for ADAM22-related DEE revealed its impaired cognitive function. Furthermore, we report the first ADAM23 variant associated with lethal neonatal-onset epilepsy and myopathy. Collectively, this study defines the LGI1-ADAM22/23 pathway-related disease spectrum.

Keywords: ADAM22; ADAM23; LGI1; MAGUK; developmental and epileptic encephalopathy; drug-resistant seizures.

Figure 1

Pedigrees and functional analyses for biallelic LGI1 and ADAM23 variants. (A) Pedigrees of seven individuals from five families with segregating biallelic LGI1 (Families 1–4) or ADAM23 (Family 5) variants are shown. (B) Schematic diagram of LGI1 gene and protein. LGI1 consists of the N-terminal leucine-rich repeat (LRR, light blue) and the C-terminal epitempin-repeat (EPTP, dark blue) domains. The N-terminal secretion signal peptide (SP, enclosed by dotted lines) is removed in the secreted LGI1. The RefSeq ID NM_005097.4 is used to indicate all variants. Grey = reported monoallelic autosomal dominant epilepsy with auditory features (ADEAF) variants; red = new biallelic variants (B, C, E and G). Ter558Gext* means Ter558GlyextTer23 (B, C, E and F). (C) Secretion test for LGI1 variants. Shown are western blots, with anti-LGI1 antibody, of conditioned medium from transfected HEK293T cells (top) and cell lysates (bottom). A closed arrowhead indicates the position of the secreted LGI1 protein. An asterisk indicates the non-specific signal. The lower band of R311 variant in the bottom panel seems to be its degradation product. The data shown are representative of four experiments. Glu383Ala is a reported monoallelic ADEAF variant. WT = wild-type. (D) Principle of the HiBiT system. When a HiBiT-tagged LGI1 secreted in the culture medium interacts with a cell-impermeable LgBiT, they form an active nanoluciferase to produce luminescent signals in the presence of its substrate, furimazine. (E) The HiBiT system shows that secretion levels of LGI1 variants are variably reduced compared with WT. P-values were determined by Kruskal-Wallis test: ^##P < 0.01 (versus WT, Steel test); *P < 0.05, **P < 0.01 (between variants, Steel-Dwass test); n = 8. Monoallelic variants in ADEAF: Leu232Pro, with partial secretion and incomplete penetrance; Phe318Cys, no secretion and complete penetrance. (F) The ADAM22-binding of LGI1 variants. Indicated cDNAs of LGI1 variants and ADAM22 were co-expressed into COS-7 cells, and cell-surface bound LGI1 through ADAM22 was live-labelled by anti-LGI1 antibody (magenta). After fixation and permeabilization of cells, protein expression of ADAM22 (green, total) and LGI1 [blue (pseudocolour) in insets, total] was validated. Nuclear DNA was stained by Hoechst 33342 (blue) to distinguish transfected from untransfected cells (top). HiBiT-tagged LGI1 WT (Fig. 1D) showed the intact ADAM22 binding. Scale bar = 20 μm. (G–I) Mapping of three biallelic LGI1 variants on the LGI1 structure (extracted from PDB #5Y31). The EPTP β-propeller domain (G) is responsible for the ADAM22 binding. The corresponding amino acid residues are shown as magenta or light grey spheres. Close-up views of Cys48 (H) and Ser524 (I) are shown.

Figure 2

In toto whole-cell patch-clamp and LFP recordings from hippocampal CA3 pyramidal neurons and CA3 stratum pyramidale of Lgi1 ^−/−  mice, respectively. (A) Representative traces of membrane potentials (Vm) and local field potentials LFPs [A(i)], spontaneous excitatory postsynaptic current (sEPSC) and LFPs [A(ii)], spontaneous inhibitory postsynaptic current (sIPSC) and LFPs [A(iii)] recorded in the same neuron. Scale bars = 20 mV, 0.1 mV, 1 s [A(i)], 50 pA, 0.1 mV, 1 s [A(ii)], 100 pA, 0.1 mV, 1 s [A(iii)]. WT = wild-type; KO = Lgi1^−/− mouse. (B) Cumulative frequency distribution curve of sEPSC amplitude and inter-event interval (IEI) [B(i)], sIPSC amplitude and IEI [B(ii)] (WT, n = 6; KO, n = 9). There are no significant differences between them (Kolmogorov-Smirnov test, P > 0.1). (C) Quantification of sEPSC charge transfer per second (WT, n = 6; KO, n = 9). (D) Power spectrum of LFPs. While delta and theta oscillations were recorded from hippocampal CA3 in WT mice, gamma and fast ripple oscillations were additionally recorded from hippocampal CA3 in Lgi1 KO mice. Postnatal 16- to 18-day-old mice were used.

Figure 3

Behavioural analysis of Adam22^ΔC5/ΔC5 knock-in mice. (A) Experimental schedule of behavioural analysis using IntelliCage. SP-FLEX test = self-paced behavioural sequencing learning and behavioural flexibility test. (B) Exploratory behaviour of Adam22^ΔC5/ΔC5 (ΔC5/ΔC5) mice and wild-type (+/+) (WT) control mice. The total number of corner visits during the 24-h period of the exploratory behaviour task is shown in graph (B, right). Two-tailed Welch's t-test. n = 31 (WT mice) and n = 37 (Adam22^ΔC5/ΔC5 mice). Mean ± standard error of the mean (SEM). (C and D) Adam22^ΔC5/ΔC5 mice show high basal activity levels. (C) The number of corner visits during the middle phase of the dark period (22:00–04:00) was significantly higher in Adam22^ΔC5/ΔC5 mice than in WT mice (C, left). The mean number of corner visits for the 24-h, 12-h light period, and 12-h dark period is shown in graph (C, right). Two-tailed Welch's t-test. Outliers are represented as closed (WT) and open (Adam22^ΔC5/ΔC5) dots. (D) Higher basal activity levels during the dark phase were stably observed over 10 days in Adam22^ΔC5/ΔC5 mice. Two-way repeated-measures ANOVA with post hoc Tukey-Kramer test. Group effect, F(1,65) = 3.293, P = 0.0742. n = 32 (WT mice) and n = 35 (Adam22^ΔC5/ΔC5 mice). Mean ± SEM. (E) Behavioural sequencing task scheme for SP-FLEX test, showing opposite rewarded corners where the mouse has to go back and forth. In the complete shift (CS) task, the diagonally opposite rewarded corners are alternated sequentially with the other, upon achieving the successful visit-rate criterion at one diagonal. Black-filled circle = rewarded corner (active); open circle = rewarded corner (inactive); crossed-out circle = non-rewarded corner; grey-filled circles connected with double arrow = rewarding sequence. (F) Adam22^ΔC5/ΔC5 mice have defects in learning and behavioural flexibility. The number of trials to reach the criteria was significantly higher in Adam22^ΔC5/ΔC5 mice than in WT mice. Two-way repeated-measures ANOVA with post hoc Tukey-Kramer test. CS (CS1–7) group effect, F(1,61) = 7.475, P = 0.0082. n = 31 (WT mice) and n = 32 (Adam22^ΔC5/ΔC5 mice). Mean ± SEM.