|Home | About | Journals | Submit | Contact Us | Français|
The epilepsies are a common, clinically heterogeneous group of disorders defined by recurrent unprovoked seizures1. Here we describe identification of the causative gene in autosomal-dominant partial epilepsy with auditory features (ADPEAF, MIM 600512), a rare form of idiopathic lateral temporal lobe epilepsy characterized by partial seizures with auditory disturbances2,3. We constructed a complete, 4.2-Mb physical map across the genetically implicated disease-gene region, identified 28 putative genes (Fig. 1) and resequenced all or part of 21 genes before identifying presumptive mutations in one copy of the leucine-rich, glioma-inactivated 1 gene (LGI1) in each of five families with ADPEAF. Previous studies have indicated that loss of both copies of LGI1 promotes glial tumor progression. We show that the expression pattern of mouse Lgi1 is predominantly neuronal and is consistent with the anatomic regions involved in temporal lobe epilepsy. Discovery of LGI1 as a cause of ADPEAF suggests new avenues for research on pathogenic mechanisms of idiopathic epilepsies.
In 1995 we mapped the ADPEAF locus to a 10-cM region on chromosome 10q24 in a single extended pedigree2. Linkage was subsequently reported to an overlapping interval in another large family, narrowing the minimal genetic region to approximately 3 cM, assuming the causative gene was the same4. Analysis of additional pedigrees confirmed the linkage but failed to narrow the region further5–7. To screen for disease-related mutations, we resequenced all coding-exon and bordering-intron sequences from positional candidate genes in the overlap interval in one affected individual from each of three ADPEAF pedigrees showing linkage to chromosome 10q24 (families 6610, A and B; Fig. 2)2,7. We then genotyped putative disease-related mutations in all available family members from the three linked pedigrees, all family members from two smaller families with ADPEAF (families C and D; Fig. 2) and 123 unrelated control individuals.
Resequencing of LGI1 identified presumptive mutations in each of the five families with ADPEAF (Table 1 and Fig. 2). All tested affected individuals from the five families harbored a single copy of a putative disease mutation, as did all obligate carriers and individuals classified as ‘unknown’ who were found to carry the disease-linked haplotype (Fig. 2). Several unaffected individuals also carried the disease haplotype and presumptive mutation. Whether these individuals manifest subclinical signs of disease or have undergone recent changes in affection status is not yet known, but the results are consistent with our previous estimate of 71% disease-gene penetrance in family 6610 (ref. 2).
To distinguish disease-related mutations from polymorphic variants, we analyzed a panel of unrelated control individuals by resequencing the entire coding region of exon 8, all of exon 6 and all of exons 3 and 4 plus intronic sequences, each in a total of 123 control individuals. Resequencing revealed several polymorphisms (none encoding amino-acid changes), but none of the five putative mutations was detected.
Three mutations predictably altered the codon frame, leading to missense mutations and premature truncation of the Lgi1 protein, whereas one point mutation in exon 8 (family D) changed a glutamic acid residue to an alanine (Fig. 3). This amino-acid residue, as well as those extending ten residues upstream and three residues downstream, was conserved in the only identifiable, full-length LGI1 homolog, the highly conserved mouse Lgi1 (data not shown). One putative mutation occurred in intronic DNA, changing a single nucleotide at the third base from the acceptor intron–exon junction of exon 4 (Table 1; family B). We speculated that the alteration would either lead to aberrant splicing or, if left unspliced, encode a stop codon at amino-acid residue 118, leading to truncation of the protein (Fig. 3c). We carried out RT–PCR on mRNA isolated from lymphoblasts prepared from a control individual and four affected individuals from family B. PCR amplification spanning the region between exons 3 and 6 identified one normal and one aberrant band in affected individuals but not in the control sample (Fig. 4a). DNA sequencing of this aberrant band showed that the IVS3(–3)C→A alteration led to retention of the entire intron 3 in a portion of the LGI1 transcript produced in individuals who carried this putative mutation (Fig. 4b,c).
Chernova et al.8 were the first to describe LGI1, upon observing that the gene was disrupted by translocation in the T98G glioblastoma multiforme (GBM) cell line and in over one-quarter of primary tumors. LGI1 expression is absent or significantly downregulated in many high-grade but not low-grade gliomas, suggesting a role for LGI1 in glial tumor progression8,9. The functional inactivation of LGI1 in high-grade GBM tumors provides the only evidence of the protein’s function. Assuming that LGI1 functions primarily as a tumor progressor as opposed to a tumor suppressor, mutations in LGI1 would predictably increase the severity or progression of glial tumors rather than increase glial tumor frequency. Although we found no clear cases of glioblastomas in these families with ADPEAF, one affected male died of a brain tumor 18 months after his diagnosis. GBM is the most common malignant tumor of the adult central nervous system and has a median post-treatment survival of less than two years. Unfortunately, the affected individual died over 30 years ago and all records and samples have been discarded.
The genomic structure of LGI1 and the highly conserved mouse homolog (97% amino-acid homology) were recently reported9; a modified representation of the human gene is depicted in Fig. 3b. The predicted protein structure consists of a 5′ signal peptide and three functional leucine-rich repeats (LRRs), flanked by cysteine-rich repeat sequence clusters. The carboxy terminus shows no significant homology to any known protein other than the mouse homolog and is composed of two direct tandem repeats and a possible transmembrane domain. All LRR-containing proteins seem to be involved in ligand binding or protein–protein interactions, and at least half are involved in signal transduction pathways10,11. The presence of four cysteine residues in each of the LRR flanking repeat sequences may indicate that LGI1 belongs to the largest group in the LRR superfamily, the adhesive proteins and receptors10. It seems that LGI1 encodes a protein consisting of an extracellular domain with LRR repeat motifs, a transmembrane segment and an intracellular segment of unknown function. The extracellular portion of Lgi1 aligns most closely with a small group of proteins including the Drosophila proteins slit, toll and tartan, and the mammalian Trk gene family, in which the LRRs are known to bind nerve growth factor and other neurotrophins10–12.
The genes tartan and slit seem to be essential in the development of the central nervous system. The tartan protein presumably functions as a membrane-bound cell-adhesion protein, and slit as a diffusible inhibitory cue for both neuronal growth-cone guidance and neuronal migration. The Drosophila slit protein is produced in midline glial cells where the LRR motif is required for the directional guidance of neuronal migration13,14. The Drosophila and vertebrate slit genes are involved both in mechanisms of growth-cone guidance and migration of entire neurons15. Although there is no direct evidence of functional conservation among LGI1 and these LRR-containing genes, the role of slit, and perhaps tartan, in neuronal cell migration and neuronal growth-cone guidance is consistent with a presumptive role for LGI1 in both epilepsy and tumor metastasis.
In humans, LGI1 is expressed primarily in brain (cerebellum, cortex, medulla, occipital pole, frontal lobe, temporal lobe and putamen) and muscle and at lower levels in spinal cord (data not shown)8. Mouse Lgi1 was also predominantly expressed in fetal brain but not in fetal lung, liver or kidney, as detected by northern-blot analysis of whole mouse embryos on embryonic days 7, 11, 15 and 17 (data not shown).
To determine whether mouse Lgi1 mRNA was predominantly expressed in neurons or glial cells, we carried out high-resolution, chromogenic RNA in situ analysis (Fig. 5). The expression pattern of Lgi1 was predominantly neuronal and was consistent with our current understanding of the anatomic regions involved in temporal lobe epilepsy. Mouse Lgi1 was expressed in a highly specific pattern in the neocortex and limbic regions. In the hippocampus, the amount of expression was highest in the granule cells of the dentate gyrus, the large-bodied cells within the hilus of the dentate gyrus and the pyramidal cells of the CA3 region, and much lower in the CA1 region. Lgi1 was also expressed in the amygdala, the piriform cortex and in distinct laminae of the dorsal lateral cortex, including the auditory cortex of mice. The degree of expression of Lgi1 varied among different brain nuclei and among individual cells within distinct brain nuclei (Fig. 5i), indicating a finely tuned transcriptional regulation and raising the possibility that differences in the amount of Lgi1 produced by individual cells are of functional importance.
All of the genes previously identified as causing idiopathic epilepsy syndromes in humans have been voltage-gated or ligand-gated ion channels16, and only one gene with a different (but unknown) mechanism has been reported in a mouse model17. LGI1 is not homologous to any known ion channel gene. Although there are no direct experimental data addressing the function of LGI1, its homology to other genes encoding LRR-containing proteins and its presumptive role in glioma tumor progression suggest a role in cell–cell communication. It is intriguing to speculate whether LGI1 functions like slit and tartan, or other LRR-containing adhesion proteins, in neuronal-cell migration and axon growth-cone guidance in the developing central nervous system, such that loss of a single gene copy would interfere with normal neuronal development and lead to focal seizures and other symptoms of ADPEAF. Loss of both LGI1 gene copies might lead to glial cell metastasis by upsetting neuron–glia homeostasis, perhaps through loss of neuronal inhibition of glial cell proliferation or differentiation.
We have reported five putative disease mutations in five families with ADPEAF2,3,7. Based on segregation patterns in affected families, predicted effects of mutations on protein sequence and lack of detection in control samples, it is likely that several, if not all, of the candidate mutations lead to temporal lobe epilepsy through haploinsufficiency. Although preliminary data indicate that loss of both copies of human LGI1 promotes glial tumor progression, it is not clear from the current study how the homozygous loss of a predominantly neuronal gene produces this effect. Such an effect is clearly possible, however, because neurons are known to inhibit glial mitosis, and interactions between neurons and glia apparently establish precisely regulated homeostasis in both tissues.
This study was approved by the Columbia Presbyterian Medical Center Institutional Review Board.
Since our original linkage report, we have actively sought families with ADPEAF through a variety of methods, including solicitation of referrals through a letter to the members of the American Epilepsy Society, presentations at Epilepsy Foundation meetings and a study web site. Given the distinct auditory symptoms in the original family (and the relative rarity with which they are reported), we specifically selected families for inclusion if they contained more than one individual with auditory symptoms.
We screened each subject for occurrence of seizure disorders through a telephone interview administered either directly or to a close relative. To ensure complete ascertainment of childhood seizures, we also, whenever possible, administered screening interviews to a parent. A neurologist or physician with special training in epilepsy then administered a semistructured diagnostic interview to each subject reported to have had afebrile seizures. The diagnostic interview obtained information on seizure semiology through both verbatim descriptions and structured questions about signs and symptoms, and on seizure etiology through questions about the history and timing of specific risk factors for epilepsy. We requested medical records when information from the diagnostic interview was ambiguous or further clarification was needed. We used information from the medical records on seizure histories and the results of EEGs, imaging studies and neurological examinations to supplement the information from the diagnostic interview when available.
Final diagnoses were assigned by expert neurologists (W.A.H. and T.A.P.), based on a review of all of the data collected on each subject. Epilepsy was defined as a lifetime history of two or more unprovoked seizures. We classified subjects with epilepsy who had a history of an insult to the central nervous system prior to the first unprovoked seizure as ‘remote symptomatic’ and those with no identified cause as ‘idiopathic’. Seizures precipitated by acute alterations in homeostasis or insults to the central nervous system (including febrile seizures) were excluded from the definition of epilepsy and classified as ‘acute symptomatic’. Epilepsies were classified according to the 1989 criteria of the International League Against Epilepsy for classification of epileptic syndromes18. To ensure that diagnoses were made blindly with respect to those of other family members, we removed identifying information prior to consensus review and reviewed subjects from different families (including many with epilepsies other than ADPEAF) in random order.
A detailed clinical description of the families and the results of linkage analysis has been presented elsewhere2,3,7. Briefly, epilepsy was related to localization in all of those with idiopathic epilepsy who could be classified. Among those with idiopathic epilepsy, auditory symptoms were reported by 6 of 11 (55%) in family 6610, 4 of 4 (100%) in family A, 6 of 9 (67%) in family B, 2 of 3 (67%) in family C and 2 of 3 (67%) in family D. Other (primarily sensory) symptoms were also reported in all five families.
We extracted DNA from blood and lymphoblastoid cell lines by established methods and generated and scored genotypes using a semi-automated high-throughput approach with fluorescently labeled microsatellite markers and ABI 377 sequencers19.
For linkage analysis in families A, B and C, we used seven microsatellite markers: D10S185, D10S200, D10S198, D10S603, D10S192, D10S222 and D10S566 (ref. 7). We initially conducted a two-point parametric analysis using the ANALYZE package20. As in our previous analysis, we assumed an autosomal dominant model with 71% penetrance and a disease allele frequency of 0.001. We computed allele frequencies using all family members. We subsequently conducted multipoint parametric and nonparametric linkage analyses as implemented in GENEHUNTER2 (ref. 21), using the scoring function based on allele-sharing among all affected relatives simultaneously. For locus order and intermarker distance, we used the maps from the Marshfield Medical Research Foundation.
In the two-point parametric analysis the maximum lod score was 1.86 for D10S603, providing significant evidence for confirmation of our earlier finding. Both the multipoint parametric analysis and nonparametric analysis strengthened the findings from the two-point analysis. The strongest support for linkage was observed for D10S603 (lod=2.93; NPL=8.06, P=0.001354). The support for linkage for adjacent markers decreased, although not substantially.
We isolated DNA samples for sequence analysis from subjects’ blood or Epstein–Barr virus (EBV)–transformed lymphoblastoid cell lines. We prepared control DNAs from a panel of 123 unrelated individuals composed of one healthy married-in individual from each of 123 families collected predominantly from 13 states in the US. The majority of families were of European descent, and the families contained no known cases of epilepsy. We designed oligonucleotide PCR primers to amplify intron–exon junctions and complete exonic coding sequences. We purified PCR amplification products over 96-well glass-fiber plates (Whatman) and sequenced in both directions using dye-terminator chemistry and ABI 373 automated sequencers. We determined sequence variants using the SEQUENCHER3 program (Gene Codes Corporation) and verified by manual inspection. Because detection of heterozygous alleles is complicated by preferential PCR amplification of one allele and by context-dependent bias in dye-terminator incorporation during cycle sequencing, we limited PCR cycles to 30 and inspected all sequence traces manually. We carried out sequence analysis blind to disease status or haplotype, analyzing control samples together with known mutation carriers. In all cases, ‘blind’ analysis detected mutations in the known mutation carriers.
We used the GeneBridge 4 radiation hybrid mapping panel (Research Genetics) and the Whitehead radiation hybrid map server to physically map the region between flanking genetic markers D10S200 and D10S577. We used sequence-tagged sites and expressed sequence–tagged (EST) sequences from the radiation hybrid map as BLASTN queries against the HTGS division of the National Centre for Biotechnology Information (NCBI) GenBank database. To identify regions of overlap, we retrieved the corresponding BAC clones from this search and systematically aligned and compared them by BLAST analysis. We carried out a similar search on genomic DNA downloaded from the Celera Discovery database, which generated a series of 0.5-Mb genomic fragments spanning the region. Comparison and organization of chromosome 10q24 DNA from HTGS and Celera allowed us to compile a single uninterrupted BAC contig across the minimal genetic region.
We used BAC sequences and Celera genomic sequences to search EST databases for homologs. Several published maps of 10q24 greatly facilitated our efforts22–24. We obtained 108 Unigene EST clusters from the human GeneMap 99 web site at NCBI. In addition, we retrieved 22,500 EST sequences directly from dbEST after BLASTN searches with BAC and genomic sequences. We used in-house software to cluster these sequences into longer cDNA sequences and then aligned them with the genomic scaffold using pairwise BLASTN alignments25. In a few cases, we carried out RACE to extend cDNA sequences.
We used RNeasy Midiprep columns (Qiagen) according to the manufacturer’s suggestions to isolate total RNA from EBV-transformed lymphoblastoid cell lines. We removed DNA contamination from the total RNA samples by DNAse I treatment. We eluted total RNA from the RNeasy columns and further purified remaining poly(A)+ RNA using the Oligotex kit (Qiagen). We converted poly(A)+ RNA to cDNA by the process of oligo(dT)-primed reverse transcription using SuperScript II (Invitrogen/LTI). We amplified the LGI1 cDNA segment bounded by exons 3 and 6 by an initial 35-cycle primary PCR amplification, followed by a 35-cycle nested PCR amplification. We separated the resulting fragments on 3% agarose gels and sequenced them using Big Dye Terminator chemistry (ABI).
We generated a mouse Lgi1 probe by PCR amplification of mouse brain cDNA. The probe consisted of an 860 bp fragment generated from the 3′ end of the mouse RNA transcript and included both coding DNA and flanking 3′ noncoding sequences. We generated oligonucleotide primers with flanking T3 and T7 RNA promoter sequences (primer sequences are available upon request).
We purified the PCR-amplified product by agarose gel electrophoresis and then used it for the synthesis of digoxigenin-labeled sense and anti-sense in situ hybridization probes. Coronal cryosections, 16 μm thick, of fresh frozen brains derived from 10-wk mice were hybridized with these riboprobes26.
Transmembrane region prediction methods: http://www.hgmp.mrc.ac.uk/GenomeWeb/prot-transmembrane.html; Whitehead RH map server: http://www.genome.wi.mit.edu/cgi-bin/contig/rhmapper.
This work was supported by grants from the National Institutes of Health, National Institute of Neurological Disorders and Stroke and by funds from the Columbia Genome Center. We thank A. Efstratiadis, I. Dragatsis, I. Lipkin, M. Hornig and H. Scharfman for their critical discussions and helpful suggestions; J. Ju, A.K. Tong and C. Wang for timely technical assistance; P. McCabe, C.D. McNew and S.R. Resor for family referrals and W. Jimenez for assistance with database management. This research would not have been possible without the generous participation of the families described.