Epilepsy-Related Ligand/Receptor Complex LGI1 and ADAM22 Regulate Synaptic Transmission

See allHide authors and affiliations

Science  22 Sep 2006:
Vol. 313, Issue 5794, pp. 1792-1795
DOI: 10.1126/science.1129947


Abnormally synchronized synaptic transmission in the brain causes epilepsy. Most inherited forms of epilepsy result from mutations in ion channels. However, one form of epilepsy, autosomal dominant partial epilepsy with auditory features (ADPEAF), is characterized by mutations in a secreted neuronal protein, LGI1. We show that ADAM22, a transmembrane protein that when mutated itself causes seizure, serves as a receptor for LGI1. LGI1 enhances AMPA receptor-mediated synaptic transmission in hippocampal slices. The mutated form of LGI1 fails to bind to ADAM22. ADAM22 is anchored to the postsynaptic density by cytoskeletal scaffolds containing stargazin. These studies in rat brain indicate possible avenues for understanding human epilepsy.

Physiological functioning of the mammalian brain involves a finely tuned balance between excitation and inhibition in neural circuits. Upsetting this delicate balance can cause epilepsy, which is a devastating and poorly treated disease. Because many genes that cause epilepsies encode synaptic ion channels, characterization of synaptic protein complexes in rat brain can provide essential insights into molecular mechanisms underlying epilepsy.

The postsynaptic density-95 (PSD-95) is a scaffolding protein at excitatory synapses and plays critical roles in synaptogenesis and synaptic plasticity (15). PSD-95 contains an array of protein/protein interaction domains and forms protein complexes with various synaptic proteins, which help organize AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid) and NMDA (N-methyl-d-aspartate)–type glutamate receptors and cell adhesion molecules at synapses. Immunoprecipitation of PSD-95 from rat brain extracts resulted in selective purification of proteins with molecular masses of 95 kD (p95) and 60 kD (p60) (Fig. 1A). Mass spectrometry indicated that p95 contained PSD-95 and ADAM22 (6), and that p60 was LGI1 (710) (table S1). Western blotting showed that stargazin (1113), a transmembrane AMPA receptor (AMPAR) regulatory protein, also coprecipitated (Fig. 1B). The recovery of ADAM22, LGI1, and stargazin showed similar efficiency (Fig. 1B, ∼10% of input). In contrast, other reported PSD-95–binding proteins, neuroligin and NR1, were hardly detected under our conditions. The interaction of ADAM22 and LGI1 with PSD-95 is specific, as PSD-95 and LGI1 quantitatively coimmunoprecipitated with ADAM22 (fig. S1 and table S1). Synapse-associated protein 102 (SAP102), another postsynaptic scaffolding protein, did not interact with ADAM22 and LGI1.

Fig. 1.

Identification of a PSD-95–associated protein complex containing ADAM22 and LGI1. (A) Immunoprecipitation of PSD-95 from adult rat brain extracts showed a series of bands. Specific bands shared by two independent PSD-95 antibodies were identified by mass spectrometry. p95 (by arrows) contained PSD-95 and ADAM22, and p60 (by arrowheads) was LGI1. A PSD-95 degradation product (p75) is shown with asterisks. IP, immunoprecipitation. (B) Western blotting showed that ADAM22, LGI1, and stargazin specifically coprecipitated with PSD-95.

All three proteins that associate with PSD-95 in our immunoprecipitation are genetically linked to epilepsy. Stargazin is mutated in stargazer mice with absence epilepsy and ataxia (14), and stargazin regulates AMPAR trafficking and gating as an auxiliary subunit (1113). LGI1 is a secreted neuronal protein (7), and its mutations have been found in patients with autosomal dominant partial epilepsy with auditory features (ADPEAF) (8, 9, 15). ADPEAF is a rare form of familial idiopathic lateral temporal lobe epilepsy characterized by partial seizures with auditory disturbances (OMIM 600512). ADAM22 shares homology to a large family of transmembrane ADAM metalloproteases but is catalytically inactive (6, 16) and is considered either a cell adhesion molecule or an orphan receptor (16). ADAM22-deficient mice show cerebellar ataxia and die around 2 to 3 weeks after birth because of multiple seizures (6).

To understand the function of this PSD-95–associated protein complex, we defined the modes for interaction. The C-terminal tail of stargazin binds to the first two PDZ domains of PSD-95 (17) (figs. S3A and S3B), and we found that one of the ADAM22 splicing variants has a C-terminal PDZ binding motif (-Glu-Thr-Ser-Ile, -ETSI) that interacts selectively with the C-terminal half containing the third PDZ domain of PSD-95 (Fig. 2A and fig. S3B). LGI1 has an N-terminal signal sequence (Fig. 2A) and is secreted from transfected hippocampal neurons (fig. S6C) and from transfected HEK293 cells (7, 18) (fig. S4, A and B) as an oligomer (fig. S4C). As PSD-95 occurs on the inner surfaces of postsynaptic membranes, extracellular LGI1 must interact with a transmembrane protein in the PSD-95 immunoprecipitates. Using cDNA transfection, we found that ADAM22, but not stargazin, specifically interacted with LGI1 (fig. S4D). Furthermore, we found that transfected LGI1, ADAM22, and PSD-95 form a tripartite complex (Fig. 2B). As further evidence for this interaction, we stained cells without permeabilization and found that LGI1 interacts specifically with ADAM22 on the cell surface, indicating that secreted LGI1 binds to the ectodomain of ADAM22. As expected, our extracellular domain binding assay readily detects the interaction of Slit2 with its receptor Robo2 (19) (Fig. 2C).

Fig. 2.

Tripartite complex formation of PSD-95, ADAM22, and LGI1. (A) Domain structures of ADAM22, PSD-95, and LGI1. SS, signal sequence; Pro, prodomain; MP, inactive metalloprotease domain; DI, disintegrin domain; CR, cysteine-rich domain; EGF, EGF-like domain; TM, transmembrane domain. ETSI represents the type I PDZ binding motif of ADAM22. WT, wild type; ΔC4, missing ETSI; ED, extracellular domain; GuK, guanylate kinase domain. LRR, leucine-rich repeat; EPTP, Epitempin repeat; E383A, a point mutant changing Glu (amino acid 383 in the fourth EPTP repeat) to Ala. (B) Tripartite complex of PSD-95/ADAM22/LGI1. PSD-95-GFP and ADAM22-HA were cotransfected with or without LGI1-Flag, and PSD-95-GFP was immunoprecipitated (lowest panel, stained by Coomassie brilliant blue). LGI1 indirectly binds to PSD-95 through ADAM22. An arrow and an arrowhead indicate the position of immature and mature ADAM22, respectively. (C and D) Interaction between secreted LGI1 and ADAM22 on the cell surface. Indicated cDNAs were cotransfected into COS7 cells. At 24 hours after transfection, surface-bound Flag-tagged proteins (red) were labeled before cell permeabilization, and then HA-tagged proteins were stained (green). The EPTP domain of LGI1 mediates ADAM22 binding. LGI1 E383A, an ADPEAF mutant, failed to bind to ADAM22. Scale bars, 10 μm.

LGI1 has two structural domains, LRR (leucine-rich repeat) and EPTP (Epitempin) repeat (20, 21) (Fig. 2A). The LRR domains show high homology to Slit, a repulsive ligand for the Robo receptor (19); the EPTP repeat domain is shared with Mass1/VLGR/USH2C, genes that cause audiogenic epilepsy in Fringe mice and Usher syndrome in humans (15, 2023) (OMIM 602851). We found that the EPTP domain (amino acids 224 to 557) mediates LGI1 binding to ADAM22. The point mutation (E383A) observed in ADPEAF (8) prevented its secretion (7) (figs. S4, B and C) and binding to ADAM22 (Fig. 2D and fig. S4E). ADAM22 did not interact with another EPTP domain (amino acids 3194 to 3530) of Mass1/VLGR/USH2C. We also found that LGI1 bound to ADAM23, the closest homolog of ADAM22, but not to the more distantly related ADAM9 (fig. S5). The disintegrin domain of ADAM22 is essential for LGI1 binding, as ADAM22(D509N) harboring a mutation in its disintegrin domain did not bind to LGI1 (fig. S5A).

To demonstrate directly the receptor/ligand relationship of ADAM22/LGI1, we constructed a secreted alkaline phosphatase (AP) fusion protein of LGI1 (LGI1-AP). LGI1-AP bound to the surface of cells only when transfected with ADAM22 (Fig. 3A). Slit2-AP did not bind to ADAM22-transfected cells. Under the conditions, Slit2-AP specifically bound to Robo2-transfected cells (19). To test whether the interaction of LGI1 with ADAM22 is stoichiometric, the ADAM22 immunoprecipitate from brain was evaluated by Coomassie blue staining and quantitative Western blotting. The stoichiometry of LGI1 binding to ADAM22 was at least 1.0 (fig. S6, A and B). Furthermore, the secreted LGI1 accumulated with ADAM22 at synaptic puncta in hippocampal neurons, where PSD-95 was localized (fig. S6, C and D). Taken together, these results imply that secreted LGI1 serves as a specific extracellular ligand for ADAM22 and that the LGI1/ADAM22 complex is scaffolded by PSD-95.

Fig. 3.

ADAM22 is a neuronal receptor for secreted LGI1. (A) Binding of LGI1-AP to ADAM22-expressing COS7 cells. LGI1-AP or Slit2-AP bound to the cell surface was detected by AP reaction. Scale bar, 20 μm. (B) Immunohistochemical staining of ADAM22 in the dentate gyrus (DG) in hippocampus and cerebellar cortex. Mo, molecular layer; Gr, granule cell layer; Pol, polymorphic cell layer; PC, Purkinje cell layer. (C) Receptor activity of LGI1 in the hippocampus and cerebellar cortex. Mouse brain sections were treated with conditioned media containing LGI1-AP. LGI1 receptor activity corresponded to the regions where ADAM22 is expressed. (D) Preincubation of LGI1-AP with the soluble form of ADAM22 (ADAM22-ED) significantly blocked the LGI1-AP binding in cerebellar cortex. Scale bars in (B), (C), and (D), 50 μm.

LGI1 mRNA is coexpressed with ADAM22 and PSD-95 mRNAs in hippocampus, cerebellum, and cerebral cortex (fig. S7, A to C). Immunohistochemical analysis with a specific antibody (fig. S7D) showed that ADAM22 protein occurs in hippocampus and cerebellum (Fig. 3B). We used the LGI1-AP fusion to detect LGI1 receptor activity directly in brain. LGI1-AP detected high receptor activity in the hippocampus and cerebellar cortex (Fig. 3C and fig. S7E). The molecular layers of dentate gyrus (DG) and CA1 regions in the hippocampus were labeled. In the cerebellar cortex, labeling occurred in neuropil of the molecular layer and synaptic glomeruli of the granular layer. These regions corresponded to the regions where ADAM22 is expressed. Preincubation of LGI1-AP with the soluble extracellular domain of ADAM22 (ADAM22-ED, depicted in Fig. 2A) inhibited LGI1-AP binding (Fig. 3D and fig. S7E), consistent with ADAM22 being a receptor for LGI1.

Because PSD-95 controls synaptic AMPA receptor number (4), we next asked whether application of LGI1 to hippocampal slices would influence glutamatergic transmission. Incubation of hippocampal slices in LGI1-AP significantly increased the synaptic AMPA/NMDA ratio (Fig. 4A) (control = 0.60 ± 0.046; LGI1 = 0.90 ± 0.10; n = 23 for each group; P < 0.05); nontagged LGI1 showed a similar effect, so we pooled the data. The effects of LGI1 on synaptic currents could be prevented by preincubation of LGI1 with ADAM22-ED (Fig. 4A) (ADAM22-ED = 0.59 ± 0.06; LGI1+ADAM22-ED = 0.65 ± 0.07), suggesting that an interaction between LGI1 and ADAM22 was required for the increased AMPA/NMDA ratio. To determine whether LGI1 directly affects the number of synaptic AMPARs, we measured AMPAR-mediated spontaneous miniature excitatory postsynaptic currents (mEPSCs). LGI1 incubation increased the average amplitude of these events (n = 17, both groups, P < 0.005), and preincubation of LGI1 with ADAM22-ED blocked this increase (Fig. 4, B and D). The frequency of spontaneous events was also increased, likely due to the increased detection of enlarged events that now reached threshold (Fig. 4C) (n = 17, each group, P < 0.005). Supporting the electrophysiological recording, LGI1 expression significantly increased AMPA receptor surface expression in cultured hippocampal neurons (fig. S8).

Fig. 4.

LGI1 selectively enhances AMPAR-mediated synaptic currents. (A) Incubation of slices in buffer containing LGI1 media significantly increased the synaptic AMPA/NMDA ratio (P < 0.05), and the effect was blocked by preincubation of LGI1 with the soluble form of ADAM22 (AD22-ED). (B to D) Incubation of hippocampal slices with LGI1-containing media increases synaptic AMPA receptor numbers. (B) Cumulative distribution plot of mEPSCs from cells in slices incubated in LGI1 as compared with control (P < 0.005). (C) Cumulative distribution plot of the interevent interval of mEPSCs in the same cells as in (B) (P < 0.005). (D) The increase in mEPSC amplitude by LGI1 was reduced by preincubation with the extracellular domain of ADAM22. (E) LGI1 incubation does not alter the magnitude of pairing-induced whole-cell LTP. (F) No change in paired-pulse ratio is seen in slices incubated in LGI1, ADAM22-ED, or both (P = 0.68).

Might the potentiation of synaptic AMPA currents by LGI1 share a mechanism with long-term potentiation (LTP), an activity-dependent process that involves synaptic insertion of AMPARs? To address this, we determined whether LGI1 incubation occludes LTP. No significant change in LTP induction was found between control and LGI1-treated slices (Fig. 4E) (control = 247 ± 30%, n = 7; LGI1 = 242 ± 43%, n = 8at 30 min after LTP induction, P = 0.93), suggesting that LGI1 strengthens excitatory synapses by a mechanism distinct from LTP. Finally, we tested whether LGI1 incubation affects presynaptic properties by measuring paired-pulse facilitation. No difference was found between control and LGI1-treated groups (Fig. 4F) (control = 1.50 ± 0.46, LGI1 = 1.58 ± 0.05, ADAM22-ED = 1.45 ± 0.07, LGI1+ADAM22-ED = 1.56 ± 0.08, P = 0.68). Taken together, our data indicate that the effects of LGI1 on synaptic transmission are exclusively postsynaptic.

This study establishes a neuronal ligand-receptor interaction between LGI1 and ADAM22, both of which are genetically related to epilepsy. This study also identifies LGI1 as an extracellular factor that controls synaptic strength at excitatory synapses. Stargazin controls the trafficking and gating of AMPARs, and PSD-95 anchors the AMPAR/stargazin complex at postsynaptic sites (11). Because the ADAM22 and stargazin binding sites on PSD-95 do not overlap, the LGI1/ADAM22 complex may stabilize the AMPAR/stargazin complex on the PSD-95–scaffolding platform (fig. S9). Supporting the idea, ADAM22 interacted with stargazin through PSD-95 (fig. S3C). Very recently, LGI1 was reported to be a subunit of Kv1.1-containing voltage-gated potassium channels and to inhibit channel inactivation by a cytoplasmic regulatory protein, Kvβ1 (24). As LGI1 is secreted, it remains unclear how it might modulate a cytosolic potassium channel mechanism.

This study defines a potentially general mode for protein-protein interaction between EPTP domains and the ectodomain of some ADAM family proteins. LGI4, another member of the LGI family, is mutated in claw paw (clp) mice, which show hypomyelination throughout the peripheral nervous system (18). Mice lacking ADAM22 display similar hypomyelination of the peripheral nervous system (6). Knockouts of Mass1/VLGR/USH2C, ADAM23 (25), or ADAM22 (6) alldisplay a seizure phenotype. These phenotypes are consistent with EPTP domain interactions with ectodomains of certain ADAM family proteins. Future binding and structural analysis will be needed to clarify the nature of the EPTP domain/ADAM family interaction. This epileptic ligand/receptor complex LGI1/ADAM22 could become a therapeutic target for synaptic disorders.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S9

Table S1


References and Notes

Stay Connected to Science

Navigate This Article