|Home | About | Journals | Submit | Contact Us | Français|
Disruptions of LGI1 in glioblastoma (GBM) cell lines and LGI1 mutations in families with autosomal dominant epilepsy imply a role for LGI1 in glial cells as well as in neurons. Although we and others could not find LGI1 mutations in malignant gliomas, our initial studies appeared to support the idea that LGI1 is poorly expressed or absent in these tumors. Microarray data suggested that LGI1 could be involved in the control of matrix metalloproteinases, and we found that tumors derived from U87 glioblastoma cells overexpressing LGI1 were less aggressive than U87 control tumors. To our surprise, we observed that LGI1 expression after differentiation of murine neural stem cells was robust in neurons but negligible in glial cells, in agreement with immunohistochemistry studies on rodent brain. This observation could suggest that the variable levels of LGI1 expression in gliomas reflect the presence of neurons entrapped within the tumor. To test this hypothesis, we investigated LGI1 expression in parallel with expression of the neuronal marker NEF3 by real-time PCR on 30 malignant gliomas. Results showed a strong, positive correlation between the expression levels of these two genes (P < 0.0001). Thus, our data confirm that LGI1 is involved in cell-matrix interactions but suggest that its expression is not relevant in glial cells, implying that its role as a tumor suppressor in gliomas should be reconsidered.
LGI1,4 the leucine-rich, glioma-inactivated 1 gene, was first identified by studying rearrangements caused by a t(10;19)(q24;q13) balanced translocation in the T98G glioblastoma multiforme (GBM) cell line; rearrangements of the LGI1 gene were also detected in the A172 GBM cell line and in several primary GBMs (Chernova et al., 1998). The LGI1 gene was subsequently found altered, usually by point mutations, in ADPEAF (autosomal dominant partial epilepsy with auditory features [OMIM entry 600512]), a rare form of idiopathic lateral temporal lobe epilepsy characterized by partial seizures with auditory disturbances (Gu et al., 2002; Kalachikov et al., 2002; Morante-Redolat et al., 2002). These observations imply that LGI1 is expressed in glial cells and in neurons and that its altered function can be at least partly responsible for two such different diseases as glioblastoma multiforme and epilepsy. Presently, available data provide only partial hints at an understanding of LGI1 function. LGI1 is part of a gene family, including other LGI1-like genes. Senechal et al. (2005) have shown that all LGI genes encode secreted protein products. LGI1, in particular, encodes a protein with a calculated molecular mass of 60 kDa that contains 3.5 leucine-rich repeats (LRRs) with conserved flanking sequences. In the LRR domain, LGI1 protein has the highest homology with a number of transmembrane and extracellular proteins that function as receptors and adhesion proteins (Chernova et al., 1998). Sequence analysis of proteins encoded by LGI1 and MASS1 (monogenic, audiogenic seizure susceptibility 1), a gene mutated in the Frings mouse model of audiogenic epilepsy and in one family with febrile and afebrile seizures (Nakayama et al., 2002; Skradski et al., 2001), identified another homology domain called EAR (epilepsy-associated repeat) consisting of a sevenfold repeated 44-residue motif (Scheel et al., 2002). In the MASS1 gene product, a fragment of the very large G-protein-coupled receptor 1 (VLGR1), the EAR domain is part of the ligand-binding ectodomain.
Previous data have suggested that there is a loss or reduction of LGI1 expression at the transition from low-grade to high-grade gliomas, and LGI1 was therefore proposed to be a tumor suppressor gene (Kunapuli et al., 2003). Expression of the LGI1 gene product has also been linked to downregulation of a number of matrix metalloproteinases, further suggesting a role for LGI1 in suppressing the invasion phenotype of glioma cells (Kunapuli et al., 2004). However, mutations in the LGI1 coding region were never found in malignant gliomas, and data following gene transfer in cell culture were conflicting (Krex et al., 2002). To gain further insight on the biologic function of LGI1, we have studied the expression pattern of the LGI1 gene in differentiating embryonic stem (ES) cells, adult brain, and tumor tissues. Further, we have used the Affymetrix (Santa Clara, Calif.) GeneChip system U95Av2 on U87 cells engineered to express the LGI1 gene product. In vivo experiments with CD1 nude mice showed a slower tumor proliferation in those animals inoculated subcutaneously and intracranially with U87 cells forced to express LGI1 than in those injected with U87 transfected with vector only.
Our data point to a role for LGI1 in the organization of cell-matrix interactions and particularly in the relationships of the tissue inhibitor of matrix metalloproteinase 3 (TIMP-3) with other related factors. Thus, the LGI1 gene product, which is expressed mainly in neural tissues, may contribute to epileptogenesis by affecting neural migration. Its low expression in malignant gliomas appears as a consequence of LGI1 downregulation in glial cells rather than of inactivation by tumor development.
Total RNA was isolated from cell lines and tumor samples with Eurozol (EuroClone, Milan, Italy). For real-time (RT)-PCR analysis, 1 μg of total RNA was first treated with DNase I (Roche, Basel, Switzerland) at 37°C for 40 min and then reverse transcribed with Expand Reverse Transcriptase (Roche) according to manufacturer’s instructions. PCR was performed with AmpliTaq Gold (Applied Biosystems, Foster City, Calif.) for 30 cycles by using specific primers on exon 5 and exon 6 (GT3: 5′-cgaactcctttgatgtgatc-3′ and GT4: 5′-cttcgcagtagatgtcttca3′, respectively). The results were normalized on beta-actin.
For most RT-PCR analysis, we used the TaqMan system (GeneAmp 5700 Detection Systems from Applied Biosystems). Primers and probes for the LGI1 and neurofilament 3 (NEF3) genes were obtained from the Assays-on-Demand choice of Applied Biosystems. We used the TaqMan Universal PCR master mix with AmpErase UNG (uracil-N-glycosylase) in a 20-μl reaction volume following cycling conditions recommended by the manufacturer (Applied Biosystems). Each reaction was carried out in triplicate, together with a negative control (water as template). The 18S gene was used as an endogenous control (Applied Biosystems). The amplification values obtained from NEF3 and LGI1 genes were divided by the corresponding amplification values for 18S to produce an expression index. Statistical comparisons of the two groups were performed by using the Fisher’s exact test within the StatView 4.5 software (Abacus Concepts, Inc., Berkeley, Calif.). We analyzed the coding sequence of LGI1 by direct sequencing of PCR products from tumor genomic DNA. Primer sequences, PCR, and sequencing conditions are available on request.
U87 cells were originally obtained from ATCC (American Type Culture Collection, Manassas, Va.) and cultured as suggested by the supplier. LGI1 was first cloned in the retroviral vector pLXSN (Clontech, Mountain View, Calif.). Cells were transfected with pLXSN (mock) and pLX-LGI1 by FuGene 6 Transfection Reagent (Roche). Transfected cells were selected by G418. Proliferation studies were done by cell counts (trypan blue) and by colorimetric WST assay (Roche).
Tumor spheroids were initiated as described elsewhere (Goldbrunner et al., 1997; Pilkington et al., 1997). Briefly, both labeled U87-mock and U87-LGI1 suspensions were seeded into culture flasks base coated with 0.75% agar noble in RPMI medium. After five to seven days, 200-μm spheroids were chosen for coculture assay. Brain cell aggregates were made from brains of fetal BD-IX rats as previously described (Bjerkvig et al., 1986). Briefly, mature brain cell aggregates (300 μm in diameter) were transferred individually into agar-coated 96-well culture dishes (Nunc, Rochester, N.Y.). Thereafter, similarly sized tumor spheroids were transferred to the wells containing aggregates, and the tumor spheroids were brought into proximity with the aggregates by using a sterile syringe. The cocultures were visualized, after 48 h, by fluorescence microscopy or confocal microscopy (Leica TCS-NT, Heidelberg, Germany).
U87 cell migration and invasion were evaluated by means of chemotaxis and chemoinvasion experiments in a 48-well, modified Boyden chamber (NeuroPore, Gaithersburg, Md.). The chemotaxis experiments were performed by using 8-μm Nucleopore (Nucleopore, Pleasanton, Calif.) polyvinylpyrrolidone-free pyro-carbonate filters coated with 10 μg/ml of type IV collagen and placed over a bottom chamber containing 20 ng/ml of platelet-derived growth factor (PDGF) as the chemoattractant factor. U87-LGI1 and U87 mock cells were added to the upper chamber at a density of 3 × 10−4 cells per well. After 6 h of incubation at 37°C, the nonmigrated cells were scraped, and cells that had migrated to the lower side of the filter were stained by Diff-Quick stain (VWR Scientific, Bridgeport, N.J.): Five to eight unit fields per filter were counted at 160× magnification by using a Zeiss microscope (Zeiss, Göttingen, Germany). The assays were run in triplicate. For the chemoinvasion assays, the filters were coated with a layer of Matrigel (BD Biosciences, San Jose, Calif.) (100 μg/filter), a solubilized basement membrane preparation extracted from the Engelbreth-Holm-Swarm mouse sarcoma, which reproduces in vitro the peritumoral matrix that the cells must degrade in vivo to invade the neighboring tissue. The Boyden chamber was assembled with the coated side of the filter facing the cells in the upper compartment. The assays were then performed as described for chemotaxis.
Mice were housed in a ventilated cabinet with constant and controlled temperature under a 12-h light schedule. Subcutaneous injections were done on nu/nu female mice (20 g; Charles River Italia, Calco, Italy) with 3 × 10−5 transfected U87 cells (21 U87-mock, 21 U87-LGI1). Tumor growth was measured with a caliper. After 30 days, the animals were sacrificed, and tumors were removed and weighed. Tumors were partially fixed with Carnoy’s fixative and were paraffin embedded or partially frozen to isolate RNA.
Intracerebral inoculi, containing 1 × 10−5 transfected U87 cells in 3 μl of phosphate-buffered saline (PBS), were introduced into nu/nu mice (23 U87-mock and 24 U87-LGI1 in three experiments) placed in a stereotaxic frame. Survival was analyzed by the log-rank (Mantel-Cox) test in Kaplan-Meier nonparametric analysis, using the StatView 4.5 software. Whole brains were fixed in Carnoy’s fixative or frozen. Other nude mice were inoculated intracerebrally (i.c.) with 1 × 10−5 cells and sacrificed on day 15 (one U87-mock and two U87-LGI1) and day 21 (two U87-mock and two U87-LGI1) after injection. Adult C57Bl6J mice were sacrificed and processed for in vivo analysis of LGI1 expression. All the mice were terminally anesthetized with CO2 and trans-cardially perfused with 50 ml of 0.1% heparin-PBS followed by 4% paraformaldehyde-PBS.
We compared the results obtained from the chips hybridized with U87-mock and U87-LGI1 cRNAs. To obtain cRNAs, total RNA was first isolated from U87-mock and U87-LGI1 cells with the RNeasy kit (Qiagen, Venlo, The Netherlands). The RNA status was controlled on a 1% agarose gel stained with ethidium bromide. First, 15 μg of total RNA was reverse transcribed to cDNA with Superscript Choice System (Gibco BRL Life Technologies, Gaithersburg, Md.) by using T7 -(dT)24 primer. Double-stranded cDNA was synthesized and then purified by separation with phenol-chlorophorm-isoamyl alcohol. This cDNA was used for in vitro transcription to cRNA by using the Enzo BioArray kit (Affymetrix). During this step, the cRNA product was labeled with biotin. Following a cleanup with Qiagen RNeasy spin columns, the labeled cRNA was quantified, and 20 μg was fragmented at 94°C for 35 min. The product was controlled on a 1% agarose gel to verify the distribution of RNA fragments from approximately 35 to 200 bases. The fragmented, biotinylated cRNA was used to hybridize the Human Genome U95A chip (oligonucleotide microarray technology, Affymetrix) following the manufacturer’s instructions.
Within each experiment, a dilution series of the transcript and of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) PCR products was processed to demonstrate that amplification efficiencies of target and reference DNA were approximately equal.
Raw data were analyzed by using the MAS5 program (Affymetrix). Results were validated by RT-PCR by using the SYBR Green kit with GeneAmp 5700 Detection Systems (Applied Biosystems) and primer designed by PrimerExpress (Applied Biosystems). Results were normalized on GAPDH values.
Anti-LGI1 antibodies were raised in rabbits injected with two human LGI1 peptides conjugated with ovalbumin. Peptide sequences were the following: nh2-FIKIQDIE-ILKIRK-cooh (peptide 716, corresponding to aa 318–331) and nh2-NTQIYKHVIVDLSA-cooh (peptide 717, corresponding to the last amino acid residues of the C-terminal region). Both peptides were conjugated with ovalbumin and mixed in a 1:1 ratio before injection.
For immunocytochemical analysis, cells were fixed for 20 min with 4% paraformaldehyde in PBS pH 7.4, washed in PBS, permeabilized for 5 min in 0.2% NP-40 in PBS, and then blocked for 45 min with 2% normal human serum and 0.5% bovine serum albumin (BSA) in 0.1% PBS-Triton X-100. Cells were then incubated at 4°C overnight with the primary antibodies (rabbit anti-LGI1 polyclonal, 1:300; and mouse antihuman TIMP-3 monoclonal, 1:5000 [Chemicon, Temecula, Calif.]) in 0.5% BSA-PBS. Cells were washed in PBS and incubated at room temperature for 60 min with indocarbocyanine 3 (Cy3)–conjugated goat antimouse (1:200), Cy3-conjugated goat antirabbit (1:400), fluorescein isothiocyanate (FITC)-conjugated goat antimouse (1:100), or FITC-conjugated goat antirabbit (1:200) secondary antibodies (all from Jackson Immunoresearch Laboratories, West Grove, Pa.) prepared in 0.5% BSA-PBS. Samples were washed three times in PBS and coverslipped with a mounting medium containing 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, Calif.) to counter-stain the nuclei before examination by fluorescence microscopy. Digital images were acquired with a Zeiss Axioskop fluorescence microscope. For bright-field staining after the primary antibody incubation, cells were incubated at room temperature for 60 min with biotin-labeled secondary goat antirabbit (1:200 [Dako, Glostrup, Denmark]) or goat antimouse (1:100 [Dako]) antibodies, and the signal was detected with streptavidin-conjugated horseradish peroxidase (1:300, Dako) and diaminobenzidine.
Experiments on rat embryos (Sprague-Dawley, Charles River, Calco, Italy) were performed on embryonic day (E) 15 and E19. The day of detection of a vaginal plug was designated as E1. All the experiments were undertaken in accordance with the guidelines established in the European Union principles of laboratory animal care (Council of European Communities, directive 86/609/EEC). The embryos were obtained from pregnant rats anesthetized with chloral hydrate (4%; 1 ml/100 g of body weight; i.p.) perfused with mixed aldehydes (for details, see Ortino et al. ). The forebrain was dissected out and coronally cut with a Vibratome (St. Louis, Mo.) in 50- to 60-μm-thick serial sections or paraffin embedded and cut into 7-μm-thick sections on gelatin-coated slides. For single immunolabeling, paraffin-embedded sections were incubated overnight with polyclonal antibody LGI1 (1:300). The avidin-biotin-peroxidase complex protocol (ABC method, Vector) was used for the immunohistochemical procedure with 3,3′-diaminobenzidine as chromogen. Immunoreacted sections were then counterstained with thionine for identification of the developing region of the telen-cephalon. For double-immunofluorescence experiments, free-floating sections were incubated in a cocktail (mixture) of the LGI1/TuJ11 (BAbCO, Richmond, Calif.) (1:300/1:8000) or LGI1/VIM (Dako) (1:300/1:200) and subsequently in a mixture of the corresponding secondary antibodies (Jackson Immunoresearch Laboratories), Cy2-conjugated horse antimouse (1:200) and Cy3-conjugated goat antirabbit (1:600). After repeated rinsing, sections were mounted in Fluorsave (Calbiochem, San Diego, Calif.) and examined in a confocal laser scanning microscope (Bio-Rad, Hemel Hempstead, UK) equipped with an argon-krypton mixed-gas laser with excitation peaks at 510 nm (for Cy2) and 550 nm (for Cy3). Confocal image series were recorded through separate channels to avoid cross talk and merged with Bio-Rad Lasersharp 2000 software.
For experiments on mouse brain, paraffin-embedded brains were cut at 4-μm thickness on a microtome and stained for histological examination. Tumor size and distribution were evaluated on routine hematoxylin and eosin–stained slides. The expression of LGI1 throughout the brain was assayed by immunohistochemistry by using the same antibodies and dilutions described previously.
Mouse J1 ES cells were differentiated in vitro into neuronal and glial committed precursor cells according to previously described protocols (Brustle et al., 1999; Okabe et al., 1996). Briefly, proliferating mouse ES cells (stage 1) were expanded on irradiated mouse embryonic fibroblasts in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen, Karlsruhe, Germany) supplemented with 20% fetal bovine serum (AG Seromed, Biochrom, Berlin, Germany), 1× MEM-nonessential amino acids (Invitrogen), 8 mg/liter adenosine, 8.5 mg/liter guanosine, 7.3 mg/liter cytidine, 7.3 mg/liter uridine, 2.4 mg/liter thymidine, 0.1 mM 2-mercaptoethanol, 26 mM HEPES (all from Sigma, Taufkirchen, Germany), and 103 U/ml leukemia inhibitory factor (Chemicon, Hofheim, Germany). After passaging onto gelatin-coated dishes (0.1% gelatin; Sigma), ES cells were trypsinized and transferred to bacterial dishes to allow embryoid body formation. Embryoid bodies were propagated for four days in the absence of leukemia inhibitory factor and subsequently plated onto tissue culture dishes. One day after plating, the medium was replaced by ITSFn (insulin, transferrin, selenium, and fibronectin) (stage 2) (i.e., DMEM/F12, Invitrogen) supplemented with 5 g/ml insulin, 50 g/ml human apotransferrin (both from Intergene, Purchase, N.Y.), 30 nM sodium selenite (Sigma, St. Louis, Mo.), 2.5 g/ml fibronectin (Life Technologies, Karlsruhe, Germany), and penicillin-streptomycin (Invitrogen). This medium has been shown to favor the survival of neuroepithelial cells (Okabe et al., 1996). After five days, cells were triturated to a single-cell suspension and further propagated on polyornithine-coated dishes (15 μg/ml, Sigma) in the presence of 10 ng/ml fibroblast growth factor 2 (FGF-2). This permits the proliferation of pan-neural precursors, which can be differentiated by growth factor withdrawal into neuronal as well as glial cells. The proliferating pan-neural precursor cells were either growth factor withdrawn for four days (stage 3) and eight days (stage 4) or further proliferated in different media for glial cell enrichment. To obtain proliferating glial precursors, ES cells were again dissociated and proliferated in the presence of 10 ng/ml FGF-2 and 20 ng/ml epidermal growth factor (EGF) (stage 5). After four days of growth factor withdrawal, these cells were driven into a mixed population of astroglial and oligodendroglial cells (stage 6). Proliferating cells from stage 5 were further grown in a medium containing 10 ng/ml FGF-2 and 10 ng/ml PDGF (stage 7), which has been shown to enhance oligodendroglial differentiation upon growth factor withdrawal (Glaser et al., 2005). Cells from stage 7 were differentiated by four days of growth factor withdrawal (stage 8). All growth factors (FGF-2, EGF, and PDGF) were purchased from R&D Systems (Wiesbaden, Germany).
Embryonic stem cell–derived neural precursors (ESNPs) from proliferating and differentiated pan-neural (stages 3 and 4) and glial (stages 5–8) precursor stages were used for Trizol (Invitrogen) extraction of RNA. Total RNA from cultured cells was phenol extracted with TriFast (Peqlab, Erlangen, Germany) and treated with RNase-free DNase (Promega, Madison, Wis.). From each sample, 1 μg of total RNA was transcribed into cDNA by using an iScript cDNA synthesis kit (Bio-Rad, Munich, Germany) according to the manufacturer’s instructions. All qRT-PCR experiments were performed in triplicate on an i-cycler (Bio-Rad) with GAPDH as the reference gene for normalization and SYBR I Green for fluorescence detection. To test whether the expression of GAPDH remains constant in the analyzed samples, we quantified the expression levels of 18S RNA. The differences in expression of these two genes remained similar in all of the cell populations examined. Thresh-old cycle (Ct) was measured as the cycle number at which the SYBR Green emission increases above the threshold line. Specific mRNA transcript levels were calculated by using Q-Gene Core Module 1.2 software (Muller et al., 2002). PCR product specificity was controlled by melt curve analysis and visualized by agarose gel electrophoresis. All used primers, with the exception of GAPDH, all were intron spanning, and sequences were as follows: GAPDH forward 5′-acgaccccttcattgacctcaact-3′ and reverse 5′-atatttctcgtggttcacacccat-3′; LGI1 forward 5′-attttcctggaatgggaccacgtag-3′ and reverse 5′-tgaggtgttct-gagggccaatggt-3′; glial fibrillary acidic protein (GFAP) forward 5′-gagtaccacgatctactcaac-3′ and reverse 5′-cca-cagtctttaccacgatgt-3′; and Tubb3 forward 5′-gatgatgac-gaggaatcggaa-3′ and reverse 5′cagatgctgcttgtcttggc-3′.
Glial fibrillary acidic protein–enhanced green fluorescent protein (GFAP-eGFP) transgenic mice (Nolte et al., 2001) at the age of 8 and 14 days were sacrificed. The brain regions were cut into 1-mm3 pieces and papain digested for 40 min before single-cell trituration. Media for cell dissociation were used as described for cell dissociation of embryonic cortical neurons (Polleux and Ghosh, 2002). The cell suspensions were kept on ice and were purified by fluorescence-activated cell sorting (FACS) of GFAP-eGFP-positive cells by using a FACSDiVa cell sorter (BD Biosciences). After sorting, the GFP-positive and GFAP-negative cells were FACS analyzed for sorting efficiency and immediately frozen for RNA purification.
We analyzed LGI1 expression by semiquantitative PCR in a panel of eight different normal tissues, in 70 gliomas (5 grade I, 19 low-grade glioma, 11 grade III, and 35 GBMs) and in six cell lines (human malignant gliomas U373, U138, SW 1088, U87, and U118 and murine malignant glioma GL261). In 16 (45.7%) of 35 GBMs, LGI1 expression was very low or absent and was completely absent in all cell lines. In only two of 35 low-grade gliomas, LGI1 expression was decreased or absent, while in other cases, expression levels were comparable to normal brain (two samples examined), confirming previous data (Besleaga et al., 2003; Chernova et al., 1998). (See supplementary data, Fig. 1S.5)
To look for mutations of LGI1, the complete coding sequence was analyzed in 10 GBMs after PCR amplification with intronic primers. No mutations were found, which was in agreement with previous observations (Krex et al., 2002; Somerville et al., 2000).
The effect of LGI1 on cell proliferation was studied on the human GBM cell line U87, which does not express LGI1, as shown by RT-PCR (Fig. 1A). LGI1 was first inserted into the retroviral plasmid pLXSN, creating pLX-LGI1, and then U87 cells were transfected with pLX-LGI1 or pLXSN as a control and selected by G418. When checked around passage 5, U87-LGI1 cells showed strong LGI1 expression (Fig. 1A).
The clonal analysis performed two days after transfection showed that U87-LGI1 clones formed after 15 days are 16% to 34% of the number formed by U87-mock cells (Fig. 1B). Analysis by the WST1 proliferation assay on cells that had been passaged more than five times (i.e., at least two months after transfection) did not show any significant difference in the proliferation rate (data not shown).
To look for in vivo tumorigenicity of U87-LGI1 and U87-mock cells, we inoculated female CD1 nude mice subcutaneously with 3 × 105 cells of the two types (n = 21 for each group). After 30 days, the weight of U87-LGI1 tumors was 45% that of U87-mock tumors (28.8 ± 39.8 vs. 63.8 ± 68.6 mg; P = 0.0499). LGI1 expression measured by RT-PCR in four U87-LGI1 tumors was 24.2% ± 17.2% that in U87-LGI1 cells, which suggests that in vivo LGI1 expression was selected against (data not shown). We then performed three experiments inoculating 3 × 10−5 U87-LGI1 or U87-mock cells into the brain of nude mice (n = 47 overall; see Materials and Methods): Animals inoculated with U87-LGI1 cells survived significantly longer than controls inoculated with U87-mock cells (P < 0.0001; Fig. 1C).
To gain insight on LGI1 function, we performed two independent experiments comparing expression profiles of U87-LGI1 and U87-mock cells using the human gene chip U95 from Affymetrix. With this approach, we identified 18 probe sets downregulated and 23 probe sets upregulated in LGI1-transfected U87 cells as compared to those in mock-transfected cells (see supplementary data, Table 1S5). We then focused on transcripts that could be linked to the role played by LGI1 in GBM containment: One of these 41 genes, TIMP-3, appeared as a good candidate, since it is an important negative regulator of tumor invasion, promotes tumor apoptosis (Baker et al., 1999), and is often silenced in different malignancies, including gliomas (Bachman et al., 1999).
Using RT-PCR on cDNAs from U87-mock and U87- LGI1 cells used in microarray experiments and on cDNAs obtained independently from the same cells, we confirmed that TIMP-3 is upregulated when LGI1 is expressed (Fig. 2). We then looked for other genes whose expression could potentially be influenced by TIMP-3. Matrix metalloproteinase 3 (MMP-3, or stromelysin 1) was a first, obvious candidate because this matrix metalloproteinase is controlled by TIMP-3 (Apte et al., 1995) and is active in gliomas (Arato-Ohshima and Sawa, 1999; Nakano et al., 1993). In both experiments, the expression of MMP-3 (accession no. X05232) was lower in U87-LGI1 cells than in U87-mock cells. We then investigated by RT-PCR the MMP-3 expression in U87-LGI1 and U87-mock cells and found that indeed LGI1 expression was associated with a fivefold decrease in MMP-3 expression (Fig. 2). The ELISA analysis performed on the supernatant of U87-mock and U87-LGI1 cells showed higher levels of MMP-3 in the supernatant of U87-mock cells (68 ng/ml) than in that of U87-LGI1 cells (34 ng/ml; see supplementary material, Fig. 2S5).
We then studied by immunocytochemistry LGI1 and TIMP3 expression in U87-mock and U87-LGI1 cells. The LGI1 antiserum that we obtained by immunization against LGI1 peptides could effectively stain early- passage U87-LGI1 cells but not U87-mock cells. The same pattern was obtained by staining with commercially available anti-TIMP3 antibodies (see supplementary material, Fig. 3S5). Immunohistochemistry was also performed 15 days after injection of U87-LGI1 and U87-mock cells into the brain of CD1 nude mice. U87-LGI1 tumors, stained by anti-LGI1 antibodies, were smaller than U87-mock tumors and showed higher TIMP-3 expression (data not shown).
To validate the role of LGI1 in the regulation of invasion processes, we set up an in vitro experiment to test interactions of U87-mock and U87-LGI1 spheroids with fetal brain aggregates. After 48 h, U87-mock cells showed an extensive invasion into the brain cell aggregates (Fig. 3A). In contrast, the U87-LGI1 cells showed a reduced invasion with a clear-cut demarcation with respect to normal brain tissue (Fig. 3B), confirming the role of LGI1 in controlling invasion-migration processes.
To quantify cellular invasive activity, we performed Matrigel chemoinvasion experiments in Boyden chambers (Albini et al., 2004). In this assay, we compared the invasive properties of U87-mock and U87-LGI1 cells both in the absence and in the presence of PDGF (20 ng/ml), a growth factor able to induce motility in various cell types, including gliomas (Heldin et al., 1998). As shown in Fig. 3C, the two cell lines retained their ability to invade Matrigel, both under basal conditions and after stimulation with PDGF. The expression of LGI1 significantly reduced this response, with an inhibition of 50% under basal conditions and 36% in PDGF-stimulated cells (P < 0.01).
Phosphoinositide 3-kinases (PI3-Ks) play a crucial role in many different physiological events, including cell migration (Cantley, 2002). AKT is a protein kinase recruited to the membrane and activated by phosphorylation as a consequence of PI3-K activity. Phosphorylated AKT is therefore taken as an index of the activity of class I PI3-Ks. We found that PDGF-induced AKT phosphorylation in U87-LGI1 cells was strongly reduced in comparison with the response detected in U87-mock cells (see supplementary material, Fig. 4S-A5). On the contrary, ERK (extracellular signal-related kinase) activation by PDGF was similar in the two cell lines, indicating that the impairment of the PI3-K pathway in transfected cells is not due to a general stress of the cells, but it is selective for this pathway. That the PI3-K signaling pathway plays a pivotal role in controlling glioma cell movement is strongly indicated by the experiments showing that the PI3-K inhibitor LY294002 reduces by 80% the PDGF-stimulated migration of U87-mock cells and blocks it in U87-MG/LGI1 cells (see supplementary material, Fig. 4S-A and B5).
Following published protocols (Brustle et al., 1999; Okabe et al., 1996), mouse ES cells were differentiated in vitro into pan-neural (i.e., with potential to differentiate into neuronal as well as glial cells) and subsequently glial cells (for details, see Materials and Methods). FGF-2 withdrawal from pan-neural precursors for four days (stage 3) and eight days (stage 4) induces neuronal differentiation, achieving approximately 60% neurons (Okabe et al., 1996). In the subsequent stages (stages 5–8), the neural precursors drift predominantly into a glial cell fate. Proliferation under EGF/FGF-2 influence (stage 5) induces an astroglial fate upon growth factor withdrawal (stage 6), whereas proliferation under PDGF/FGF-2 administration (stage 7) induces an oligodendroglial fate upon growth factor withdrawal (stage 8) (Glaser et al., 2005).
To confirm the expected differentiation of the mouse ES cells into neuronal and subsequently glial cells, the neuronal marker Tubb3 and the astroglial marker GFAP were quantified by RT-PCR. Tubb3 was found at least 4-fold to 8-fold higher in stages 3 and 4 than in proliferating or differentiated glial stages (Fig. 4). GFAP was found 14-fold to 120-fold higher in proliferating/ differentiated glial stages than in stages 3 and 4 (Fig. 4). LGI1 was at least 18-fold higher expressed in stages 3 and 4 as compared to levels in the subsequent glial stages (Fig. 4), which supports the idea that LGI1 is predominantly expressed in neuronal cells.
GFAP-GFP transgenic mice expressing GFP under the human GFAP promoter (Nolte et al., 2001) were used for acute isolation and flow cytometric separation of GFP-positive and GFP-negative cells from P8 and P14 brain tissues. The sorting efficiencies are shown in Fig. 5A. We compared sort-purified GFP-positive and GFP-negative cells of neocortical, striatothalamic, and hindbrain regions. Fivefold increased GFAP levels in the GFP-positive cells confirmed sort purification. The GFP-negative-sorted cell fraction revealed at least twofold increased levels of Tubb3, indicating an enrichment of neuronal cells (Fig. 5C). LGI1 was found at the highest levels in the GFP-negative cell fractions of cortex and striatum (Fig. 5B), again indicating a preferentially neuronal expression of LGI1. Regarding the analyzed ages of 8 and 14 days, we found that LGI1 expression was significantly higher in older mice. In summary, we conclude that LGI1 expression is highest in neuronal cells and seems to be upregulated during development.
Double-labeling experiments in E15 embryos, using antibodies against LGI1 and TuJ1 (Tubb3), a marker of migrating and differentiating postmitotic neurons (Lee et al., 1990), revealed that LGI1 was intensely expressed by the migratory cells of the cortical plate (CP) (Fig. 6A). Vimentin-positive, radially oriented glial fibers, which at this developmental stage cross the whole cortex, were interposed within LGI1-expressing cells in the CP (Fig. 6B), which suggests that some of these young neurons reach their final position in the CP by using radial glial fibers. Immune labeling for LGI1 was found at the end of the embryonic development, when the CP is expanding and layer VI is clearly identifiable. Intense immuno-histochemical staining was present at E19 in the CP as well as in neurons of layer VI (Fig. 6C and D). In addition, some neuroepithelial precursor cells localized in the ventricular zone were LGI1 positive (Fig. 6E).
We also looked for the expression patterns and the anatomical distribution of LGI1 in adult mouse CNS. LGI1 was highly expressed in neurons of different CNS regions and particularly in cortical, cerebella, hippocampal, and spinal cord neurons (Fig. 6F–I).
Since neurons express LGI1 at a high level, we decided to test whether LGI1 expression in gliomas was consequent to the presence of neurons entrapped into the tumor. To study this, we explored by RT-PCR the expression pattern of the neuronal cell marker NEF3 and of LGI1 in 30 malignant gliomas (28 GBMs and two anaplastic astrocytomas). Reactions were normalized according to 18S expression, and expression values were compared to those of commercially available normal brain (Ambion, Austin, Tex.). Simple regression analysis indicated that a significant correlation was present in the expression of these two genes (r = 0.68, P < 0.0001; Fig. 7).
We also evaluated results considering as low or high the values that were, respectively, below or above the median of the delta cycle threshold (see supplementary data, Table 2S5). Again the data indicated a strong correlation between LGI1 expression and NEF3 expression in GBM by Fisher’s exact test (P = 0.0001).
Two reports first established that mutations of LGI1 are associated with a dominantly inherited form of epilepsy (Kalachikov et al., 2002; Morante-Redolat et al., 2002). After these findings were published, a number of studies confirmed that LGI1 mutations do play a role in inherited epilepsy (Brodtkorb et al., 2005; Fertig et al., 2003; Gu et al., 2002; Hedera et al., 2004; Michelucci et al., 2003; Pizzuti et al., 2003). On the other hand, the initial suggestion that LGI1 alterations are implied in glioma tumorigenesis did not receive any direct confirmation. On the contrary, a few reports have cast doubts on such a role (Brodtkorb et al., 2003; Gu et al., 2005; Krex et al., 2001, 2002). In recent years, we have compiled the data presented in this article that, while confirmatory of a role of LGI1 in cell-matrix interactions and migratory processes in the CNS, seem to restrict this role to neuronal and not to glial cells.
Our data do confirm that in malignant gliomas LGI1 expression is often low and, interestingly enough, that it is completely absent in all the glioma cell lines that we have tested. These include U87 GBM cells, one of the cell lines most frequently used in experimental neuro-oncology. Using RT-PCR, we confirmed the absence of LGI1 expression in two U87 lines used in two independent laboratories collaborating in this study (Istituto Nazionale Neurologico C. Besta [G.F.] and University of Hong Kong [S.Y. Leung]). Because of these results and considering the in vivo tumorigenicity of U87 cells, we decided to base a relevant part of our work on the study of LGI1 overexpression in this cell line. Kunapuli et al. (2003), however, found that U87 cells express low levels of LGI1 and suggested that this endogenous expression was an obstacle to displaying the results of LGI1 transfection. Interestingly enough, Carson and Pirruccello (1998), using dilutional cloning, obtained five distinct sublines from the parent U87 population; thus, cell culture heterogeneity may explain differences in LGI1 expression found by us and by Kunapuli et al. We have found that in early-passage U87-LGI1 cells the forced expression of LGI1 reduces the clonogenic potential of U87 cells and their migration potential in Matrigel (see Figs. 1B and and3C,3C, respectively). This is in agreement with the data obtained by Kunapuli et al. on A172 and T98 glioblastoma cells overexpressing LGI1. When we repeated experiments in U87-LGI1 cells passaged more than five times, the effects on both proliferation and migration progressively disappeared. This was in close association with decreased LGI1 expression taking place during culturing in spite of G418 selection, as evaluated by RT-PCR. U87 cells can escape the action of exogenous genes negatively regulating the cell cycle (Tenan et al., 1995). Thus, it is possible that Kunapuli et al. (2003) could not confirm in U87 cells the observations made by LGI1 overexpression on A172 and T98 cells because U87 cells can silence the exogenous expression of LGI1 more effectively than can the other cell lines. This could also explain the difficulties experienced by Krex et al. (2002) in suppressing S-phase transition by LGI1 overexpression.
Our in vivo observations obtained by injecting U87-LGI1 and U87-mock cells in nude mice s.c. or into the brain agree with the in vitro data, showing for the first time that LGI1 expression can decrease aggressiveness of U87 tumors (Fig. 1C). Microarray data suggested that this function is mediated by inhibition of MMP-3 and by increased expression of TIMP-3. The data were also validated by PCR, ELISA, and immunocytochemistry on U87-LGI1 and U87-mock cells and by in vitro experiments on brain spheroids and agree with similar observations performed by Kunapuli et al. (2004).
At the time of these investigations, then, we obtained a set of data coherent with the attribution of a tumor suppressor role to LGI1. Discrepancies, however, were present. We and others could not find point mutations or microdeletions supporting the relevance of LGI1 as a suppressor (this article; Krex et al., 2002; Somerville et al., 2000). According to the Knudson model, two hits are inactivating tumor suppressor genes: The first one is usually a chromosomal deletion including one gene copy, detectable as a loss of heterozygosity (a frequent finding in the 10q24 region, where LGI1 is located); the second can be a mutation or a chromosomal translocation, as initially reported for LGI1 (Chernova et al., 1998). A further report, however, could not confirm translocations as a general mechanism of LGI1 inactivation (Krex et al., 2002). As exemplified by studies on CDKN2A, promoter methylation may be an important mechanism for gene inactivation (Merlo et al., 1995). Analysis on the LGI1 promoter, however, did not show significant differences in the methylation status of normal brain and that of glioblastomas (Somerville et al., 2000).
Another important question on the LGI1 role in gliomas was raised by histological studies on adult mouse brain, showing that LGI1 expression is confined to neurons (Kalachikov et al., 2002; Morante-Redolat et al., 2002), as also confirmed by our observations (Fig. 6F–I). In acutely isolated cells from the juvenile mouse brain, LGI1 expression was highest in neuronal cell fractions of neocortex and striatum. During neural differentiation of murine embryonic stem cells, the LGI1 expression profile strongly corresponded with the profile of the neuronal marker Tubb3, whereas LGI1 levels were clearly downregulated in predominantly glial cell stages.
Thus, three different lines of evidence support the idea that LGI1 is not expressed primarily in glial cells, implying that its lack in malignant gliomas is not the consequence of genetic alterations but rather of expression patterns normally occurring during brain development. The results obtained by RT-PCR of LGI1 and NEF3 expression on 30 malignant gliomas explain the apparent discrepancy in the different expression levels of different gliomas in vivo, indicating that such difference is consequent to neuronal cell “contamination” in the tumor specimen.
If this is the case, should we consider as contradictory the data attributing to LGI1 a function in controlling cell-matrix interactions and, therefore, migration/invasion processes? Our results, as well as those of Kunapuli et al. (2003, 2004), may very well allude to the actual function of LGI1, but to such function as it physiologically occurs in neurons and not in glial cells. In this regard, it is quite intriguing that LGI1 bears considerable sequence similarity with SLIT1, an important repulsive guidance cue for neurons (Cloutier et al., 2004; Yuan et al., 1999) in close collaboration with the slit receptor Rig-1/Robo3 (Marillat et al., 2004).
Although direct evidence linking mutations of guidance genes to epilepsy is not available, recent data provide important suggestions that this may well be the case (Ponnio and Conneely, 2004; Yang et al., 2005). A knockout mouse for LGI1 could provide a relevant tool for these investigations. Unfortunately, our attempts in this direction, performed by deletions of the first exon of LGI1, have been repeatedly unsuccessful, suggesting that the LGI1 locus maybe resistant to homologous recombination.
We thank Suet Y. Leung (Departments of Pathology and Surgery, University of Hong Kong, Hong Kong, China) for contributing data on LGI1 expression in gliomas; Christiane Nolte and Helmut Kettenmann for providing the hGFAP-eGFP transgenic mouse strain; Cristina Longinotti, Patrizia Tunici, Ettore Salsano, Laura Cajola, Stefania Mazzoleni (all at the Istituto Neurologico Besta, Milan, Italy), and Monika Rade (Institute of Reconstructive Neurobiology, University of Bonn, Bonn, Germany) for help with in vivo and in vitro experiments; and Elmar Endl for management of flow cytometry (University of Bonn, Bonn, Germany; http://www.ukb.uni-bonn.de/IMMEI/flowcore/index.html).
1This work has been supported by a grant to G.F. from Associazione Italiana per la Ricerca sul Cancro (Italian Association for Cancer Research).
Oliver Brüstle holds patent EP 98 966 877.3 relating to embryonic stem cell differentiation into neural precursors.
4Abbreviations used are as follows: BSA, bovine serum albumin; CP, cortical plate; Cy, indocarbocyanine; DMEM, Dulbecco’s modified Eagle’s medium; E, embryonic day; EAR, epilepsy-associated repeat; EGF, epidermal growth factor; eGFP, enhanced green fluorescent protein; ES, embryonic stem; ESNP, embryonic stem cell–derived neural precursor; FACS, fluorescence-activated cell sorting; FGF-2, fibroblast growth factor 2; FITC, fluorescein isothiocyanate; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; GBM, glioblastoma multiforme; GFAP, glial fibrillary acidic protein; LGI1, leucine-rich, glioma-inactivated 1 gene; LRR, leucine-rich repeat; MASS1, monogenic, audiogenic seizure susceptibility 1 gene; MMP-3, matrix metalloproteinase 3; NEF3, neurofilament 3; OMIM, Online Mendelian Inheritance in Man; PBS, phosphate-buffered saline; PDGF, platelet-derived growth factor; PI3-K, phosphoinositide 3-kinase; RT, real time; TIMP-3, tissue inhibitor of matrix-metalloproteinase 3.
5Supplementary figures and tables for this article may be viewed online at http://www.brainlife.org/supplemental/piepoli_t2006.htm.