Reelin, synthesized and secreted by Cajal–Retzius (CR) cells in the marginal zone of the cortex, is an extracellular matrix protein important for the development of cortical layers. In reeler mutant mice lacking Reelin, there are severe malformations of neocortical and hippocampal lamination. It has been assumed that Reelin acts as a stop signal for migrating neurons. Here we show, by using the dentate gyrus as a model in in vivo studies and in vitro assays, that Reelin exerts its effects, at least in part, by acting on the radial glial scaffold required for neuronal migration. Migration defects of dentate granule cells, reminiscent of those seen in reeler mutants, are observed in tissue from patients with temporal lobe epilepsy (TLE). The extent of granule cell dispersion in TLE was found to be inversely correlated with the number of reelin mRNA synthesizing CR cells and reelin mRNA expression as revealed in quantitative RT–PCR studies. These findings show that the Reelin signaling pathway is essential for the correct positioning of human hippocampal neurons and that a Reelin deficiency is involved in the pathological changes associated with epilepsy.
In the present study we have dealt with the role of Reelin in the formation of the granule cell layer in the rodent and human dentate gyrus. There are several reasons to focus on the dentate gyrus when trying to understand the principles of cortical development. First, the dentate gyrus offers a relatively simple organization with the principal cells, the granule cells, arranged in a densely packed layer. Second, the majority of the granule cells are generated late, largely post-natally (Altman and Das, 1965, 1966; Altman, 1966; Schlessinger et al., 1975; Bayer, 1980), offering the possibility of studying neurogenesis, neuronal migration and the formation of the densely packed granule cell layer in post-natal animals and in adult human beings. Third, and most relevant to the present study on the role of Reelin in the formation of the granule cell layer, there are characteristic malformations of the dentate gyrus in reeler mice lacking Reelin. In these mutants, the granule cells are not arranged in a densely packed layer, but are distributed over the entire hilar region (Stanfield and Cowan, 1979; Drakew et al., 2002).
Reelin is synthesized and secreted by Cajal–Retzius (CR) cells (d’Arcangelo et al., 1995, 1997), early generated horizontal neurons located in the marginal zone of the cortex (Retzius, 1893, 1894; Ramón y Cajal, 1911). As part of the cortex, the dentate gyrus contains CR cells in its marginal zone, the outer molecular layer (del Rio et al., 1997). In the present study we have focused on Reelin effects on the radial glial scaffold in the dentate gyrus required for the migration of granule cells. In addition, we show that granule cell migration defects in the dentate gyrus of human patients suffering from temporal lobe epilepsy (TLE) are associated with a reduced expression of reelin mRNA.
Materials and Methods
Preparation of Hippocampal Slice Cultures
Hippocampal slices were prepared according to a standard procedure (Förster et al., 1998). Briefly, brains were removed from young post-natal mice. Hippocampi were dissected using a fine spatula and sliced perpendicular to their longitudinal axis with a McIllwain tissue chopper. Section thickness was 400 μm. Hippocampi from mutant mice were identified by their characteristic morphological alterations. In addition, the genotypes of wild-type, heterozygous and reeler mice were characterized by PCR amplification of genomic DNA fragments, as described (Deller et al., 1999). Slices were transferred onto stripe matrices (see below) and then incubated on Millipore membranes for up to 14 days according to the method of Stoppini et al. (Stoppini et al., 1991).
Brains of young post-natal mice were immersion fixed in 4% paraformaldehyde (PFA) at 4°C overnight. Coronal sections (50 μm) were cut on a vibratome and then subjected to immunostaining. For analysis of cell migration and process outgrowth, slice cultures of hippocampus were fixed with 4% PFA for 1 h at room temperature. Immunostaining of sections and cultures was performed with antibodies against the radial glial markers glial fibrillary acidic protein (GFAP) (DAKO), RC2 and nestin (Developmental Studies Hybridoma Bank). The antibody TUJ1 (Babco), which recognizes neuron-specific βIII-tubulin was used to stain outgrowing neurites. For visualization of immunostaining, Cy2- or Cy3-labeled fluorescent secondary antibodies (Dianova) were used, according to the manufacturer’s instructions. In addition, cell nuclei were stained with the fluorescent dye DAPI (Roche). Nucleopore membranes (Corning-Costar) with the stained cultures were transferred to a microscope slide, coverslipped with Moviol (Hoechst) and analyzed under a fluorescence microscope.
In Situ Hybridization Histochemistry for reelin and Dab1 mRNA
In situ hybridization for reelin and disabled 1 (Dab1) mRNA was performed according to a previously described protocol (Haas et al., 2000), using digoxigenin (DIG)-labeled riboprobes. Some of the sections stained for Dab1 mRNA were subjected to GFAP immunolabeling.
Transfection of 293-cells with reelin cDNA
Preparation of Reelin-containing and Control Supernatants
To obtain Reelin-enriched supernatants and control supernatants not containing Reelin, the incubation medium [DMEM (Gibco-BRL), 10% fetal calf serum (FCS), 0.9 g/l G418] from reelin-transfected 293-cells or green fluorescent protein (GFP)-transfected control 293-cells was replaced by serum-free medium (QBSF51; Sigma) and cells were incubated for 2 days at 37°C, 5% CO2. The conditioned medium was collected and the Reelin content (absence of Reelin in control cell supernatants) was confirmed by western blotting using mAb G10 (kindly provided by A. Goffinet).
Preparation of Reelin-containing Stripe Carpets
Nucleopore membranes were coated with alternating stripes of either serum-free medium from Reelin-secreting 293-cells or, as a control, from GFP-transfected cells. Coating of Nucleopore membranes was performed according to Walter et al. (Walter et al., 1987). Presence of Reelin in the medium and its absence in control medium was confirmed by western blot analysis of the conditioned media. Correct coating of Reelin stripes was controlled by staining with mAb G10 and a fluorescent secondary antibody.
Microglial Staining with Griffonia simplicifolia Agglutinin (GSA I-B4) Lectin
Microglial staining was performed with GSA I-B4 conjugated to FITC (Sigma). PFA-fixed cultures were incubated with FITC-labeled GSA I-B4 as previously described (Hollerbach et al., 1998) and GSA-stained cells were visualized under a fluorescence microscope.
A total of 22 patients (mean age 36.9 ± 9.6 years) undergoing amygdalohippocampectomy or temporal lobectomy with medically intractable TLE were included in this study. All patients experienced pharmaco-resistant complex partial seizures. For comparison, the hippocampi from seven subjects (mean age 28.3 ± 9.2 years) with no history of neurological disorder were collected at autopsy within 48 h of death.
Preparation of Human Tissue Samples and Cell Counts
Hippocampi were collected in isotonic saline and 2 mm coronal sections at the mid level of the hippocampus were cut. Slices for PCR were immediately frozen and kept at −80°C. Tissue for morphological analysis was immersion fixed in buffered 4% PFA followed by cryoprotection. Serial sections (coronal plane, 40 μm) were cut on a cryostat and alternately processed for cresyl violet staining or for in situ hybridization.
Reelin mRNA-positive cells were counted along the hippocampal fissure in five consecutive sections of epilepsy (n = 15) and autopsy (n = 7) hippocampal specimens at a magnification of 200× using a counting grid, defining an area of interest to a width of 500 μm along the hippocampal fissure.
Measurement of Granule Cell Dispersion
The average width of the granule cell layer of the dentate gyrus was determined in cresyl violet-stained sections of epileptic (n = 22) and autopsy hippocampi (n = 7) as described by Houser (Houser, 1990) using an image analysis system. Mean and standard deviation of 20 measurements in five sections were calculated for each case.
Microdissection of the Human Dentate Gyrus for PCR Analysis
Cryosections (50 μm) of the hippocampus were collected on RNase-free slides and fixed in −20°C ethanol. Three sections were used for measurement of granule cell dispersion and three sections were stained with toluidine blue for microdissection. For this the dentate gyrus was excised using a scalpel blade and collected in Trizol (Invitrogen).
RNA Extraction, Reverse Transcription and Real-time Quantitative RT–PCR
Total RNA was isolated along with 0.5 ng Drosophila poly(A)+ RNA (Clontech), added as an external standard, and was reverse transcribed with oligo(dT) primers. Abundance of transcripts was determined by real-time quantitative PCR on a GeneAmp 5700 System with SYBR Green (Applied Biosystems). Specific primers for human reelin, very low density lipoprotein receptor (VLDLR), apolipoprotein E receptor 2 (ApoER2), Dab1 and Drosophila glucose 6-phosphate dehydrogenase mRNAs were used for amplification as described (Haas et al., 2002). Monitoring the fluorescence signal, which is proportional to the amount of double-stranded product, yielded complete amplification profiles. Melting curves of the amplified products were used to control for specificity of the amplification reaction. Due to heterogeneity of the tissue samples with respect to disease-related cell loss, the external standard and the dissected tissue volume were used to normalize gene expression levels from different patients (Haas et al., 2002).
Results and Discussion
Role of Reelin in the Migration of Hippocampal Cells In Vitro
In an attempt to study the effects of Reelin on the development of the dentate gyrus, we first established a stable Reelin-synthesizing cell line (Förster et al., 2002). For this, 293-cells were transfected with the full-length reelin cDNA. Reelin synthesis and secretion were tested by immunocytochemistry and western blot analysis of the supernatant by using a monoclonal Reelin antibody (G10), kindly provided by Dr A. Goffinet (Bruxelles, Belgium) (Fig. 1). Next, we took advantage of the stripe choice assay, an assay originally developed to study attraction and repulsion of outgrowing axons (Walter et al., 1987). In order to study Reelin effects on the migration of hippocampal cells, slices of hippocampus were placed on Nucleopore membranes containing Reelin stripes next to control stripes made from the supernatant of GFP-transfected cells. Cells migrating out of the slices were then identified as neurons, radial glial cells and microglial cells by applying appropriate, cell-specific markers (TUJ1 antibody, recognizing βIII-tubulin in neurons, antibodies against the radial glial markers GFAP, nestin and RC2, and GSA I-B4 for microglial cells). The identified cells were then studied with respect to their preference for the Reelin stripes or control stripes.
Neuronal cells did not show a response to the Reelin stripes. Only occasionally did we detect neurons that had migrated out of the slice cultures, and their cell bodies were located on either the Reelin stripes or the control stripes. Their processes crossed randomly over both types of stripes (Förster et al., 2002). Similarly, microglial cells, which were the most abundant cells migrating out of the slices, did not show a preference for the Reelin or control stripes. In contrast, GFAP-positive cells clearly accumulated on the Reelin stripes (Fig. 2A). A large number of these GFAP-positive cells on the Reelin stripes showed morphological characteristics of astrocytes. However, many of them also gave rise to long processes reminiscent of radial glial fibers. Moreover, double labeling for neuronal βIII-tubulin and GFAP revealed neurons attached to GFAP-positive fibers, suggesting neuronal migration along radial glial fibers (Fig. 2B). Another most important observation was that GFAP-positive fibers preferentially branched on the Reelin stripes. There were significantly more branching points of GFAP-positive glial fibers on the Reelin stripes than on the control stripes (Förster et al., 2002). This finding in particular reminds one of radial glial branching in the marginal zone, as already observed by Gustaf Retzius >100 years ago (Retzius, 1893). Comparable with a Reelin stripe in our choice assay, Reelin is enriched in the marginal zone containing CR cells synthesizing and secreting this glycoprotein. In sum, we found specific effects of Reelin not on hippocampal neurons, but on GFAP-positive cells, presumably radial glial cells required for the migration of hippocampal neurons (Förster et al., 2002). One is tempted to speculate that the migration defects observed in reeler mutants are due, at least in part, to a dysfunction of radial glial cells that were found to be responsive to Reelin in the stripe choice assay.
The Radial Glial Scaffold is Altered in Mice with Defects in the Reelin Signaling Pathway
These findings in the stripe choice assay were supported by subsequent studies of the radial glial scaffold in reeler mice, which was found to be severely altered when compared with wild-type animals (Fig. 3A,B). Along this line, an altered radial glial scaffold was also noticed in scrambler mice, a mutant lacking Dab1, an intracellular adapter molecule in the Reelin signaling cascade (Howell et al., 1997; Sheldon et al., 1997; Ware et al., 1997). Together these findings indicate that the Reelin signaling cascade is required for the normal formation of the glial scaffold in the hippocampus, and the radial glia malformation seen in these mutants is likely to underlie, at least in part, the severe neuronal migration defects in the dentate gyrus of reeler mice (Stanfield and Cowan, 1979; Drakew et al., 2002). Dab1 mRNA expression in radial glial cells was confirmed by co-localizing Dab1 mRNA and GFAP protein in dentate radial glial cells (not shown) (Förster et al., 2002), and radial glial cells from Dab1-deficient scrambler mice did not show a preference for Reelin in the stripe choice assay, as was observed in wild-type animals. These findings suggest that the Reelin signaling cascade is active in dentate radial glial cells (Förster et al., 2002). Some of the changes observed in reeler mice and Dab1-deficient animals, i.e. minor, localized migration defects of the granule cells, were observed in mice lacking β1-integrins (Graus-Porta et al., 2001; Förster et al., 2002), which are putative Reelin receptors (Dulabon et al., 2000). We are currently studying mice deficient of the Reelin receptors VLDLR and ApoER2. First data indicate that there are gradual defects in the formation of the radial glial scaffold in the dentate gyrus of these mutants. Double knockouts lacking both lipoprotein receptors phenocopy the radial glia malformation observed in reeler and scrambler mice.
A Role for Reelin in Granule Cell Dispersion in Epilepsy
Migration defects of the granule cells, reminiscent of the ones observed in mouse mutants with genetic defects in the Reelin signaling pathway, were often observed in tissue removed from epileptic patients for therapeutic reasons (Fig. 4A–D). This granule cell dispersion, first described by Houser (Houser, 1990), could indicate a localized dysfunction of the Reelin signaling cascade, either during development or in adulthood, in the latter case probably affecting persisting radial glial cells required for the migration of late generated granule cells.
In order to study a possible involvement of the Reelin pathway in granule cell dispersion, two different approaches were chosen (Haas et al., 2002). First, reelin mRNA-synthesizing cells in the outer molecular layer of the dentate gyrus were counted in tissue samples from epileptic patients and in control tissue from tumor patients or in autoptic human tissue. The number of reelin mRNA-synthesizing cells was correlated with the width of the granule cell layer, taken as an indicator of the extent of granule cell dispersion (Haas et al., 2002). In a second approach, the dentate gyrus was microdissected and the relative amounts of reelin, ApoER2, VLDLR and Dab1 mRNA were estimated by quantitative real-time RT–PCR. Both approaches provided evidence of reduced reelin expression in the dentate gyrus of patients with granule cell dispersion. In the cell counts, there was an inverse correlation of the number of reelin mRNA-synthesizing CR cells and the width of the granule cell layer (Fig. 5). Similar results were obtained with the quantitative RT–PCR studies of reelin mRNA. Thus, large relative amounts of reelin mRNA were found in hippocampi exhibiting a densely packed granule cell layer, whereas relative reelin mRNA amounts were low in cases with pronounced granule cell dispersion (not shown) (Haas et al., 2002). No correlation between the width of the granule cell layer and mRNA expression was observed with other molecules of the Reelin cascade analyzed. These data indicate that: (i) Reelin-synthesizing CR cells are present in the adult human dentate gyrus; (ii) a gradual Reelin deficiency correlates with the extent of granule cell dispersion suggesting that Reelin is required for the dense packing of granule cells in the human dentate gyrus, as is known for the dentate gyrus of rodents.
In this study we have shown that Reelin is required for the proper migration of dentate granule cells in the rodent and human hippocampus. Reelin is an ancient, large molecule of the extracellular matrix. It has been found in lower vertebrates such as turtles and lizards (Bar et al., 2000). Using the stripe choice assay we have shown that Reelin acts on radial glial cells and is required for the terminal branching of their long processes in the marginal zone. It is probably also important for the anchorage of radial glia end-feet to the pial surface (Graus-Porta et al., 2001). Our findings in the stripe choice assay suggest that the effects of Reelin on neuronal migration are indirect, likely mediated by Reelin acting on the radial glial scaffold. However, our results do not exclude direct effects of Reelin on neurons. The stripe choice assay is an artificial in vitro system that cannot reflect all complex cellular and molecular interactions in vivo. In the developing cortex, Reelin seems to stop the migration of neurons as soon as they reach the marginal zone (Frotscher, 1997, 1998). In the absence of Reelin, the marginal zone is densely populated with neurons, whereas neurons are rare in this layer of the cortex in wild-type animals (Caviness and Rakic, 1978; Caviness et al., 1988; Rakic and Caviness, 1995; Frotscher, 1998).
A role of Reelin as a stop signal for migrating neurons cannot account for the migration defects seen in the dentate gyrus of reeler mice. In this mutant, the granule cells are distributed all over the hilar region and do not complete their migration to the granule cell layer. Our results indicate that a malformation of the radial glial scaffold underlies this migration defect. A similar loose distribution of the granule cells, granule cell dispersion, was found in hippocampal tissue from TLE patients and was accompanied by reduced reelin expression, strongly suggesting that Reelin signaling is required for normal neuronal migration in the human hippocampus (Haas et al., 2002). Studies are in progress to find out whether or not Reelin deficiency and granule cell dispersion in epileptic patients are associated with a disorganized radial glial scaffold similar to that we have observed in rodents with genetic defects in the Reelin signaling cascade.
These studies were supported by grants from the Deutsche Forschungsgemeinschaft (SFB 505, SFB TR3 and FO 223/4 to E.F.) and a Max Planck Research Award to M.F.