Neuronal ectopia, such as granule cell dispersion (GCD) in temporal lobe epilepsy (TLE), has been assumed to result from a migration defect during development. Indeed, recent studies reported that aberrant migration of neonatal-generated dentate granule cells (GCs) increased the risk to develop epilepsy later in life. On the contrary, in the present study, we show that fully differentiated GCs become motile following the induction of epileptiform activity, resulting in GCD. Hippocampal slice cultures from transgenic mice expressing green fluorescent protein in differentiated, but not in newly generated GCs, were incubated with the glutamate receptor agonist kainate (KA), which induced GC burst activity and GCD. Using real-time microscopy, we observed that KA-exposed, differentiated GCs translocated their cell bodies and changed their dendritic organization. As found in human TLE, KA application was associated with decreased expression of the extracellular matrix protein Reelin, particularly in hilar interneurons. Together these findings suggest that KA-induced motility of differentiated GCs contributes to the development of GCD and establish slice cultures as a model to study neuronal changes induced by epileptiform activity.
Human temporal lobe epilepsy (TLE) is associated with ectopic dentate granule cells (GCs) that give rise to a broadening of the normally compact GC layer (GCL), called granule cell dispersion (GCD; Houser 1990). In addition, many GCs in the human epileptic hippocampus show recurrent basal dendrites originating from the basal pole of the cell body but then returning to the molecular layer (Freiman et al. 2011). The pathomechanisms leading to GCD and recurrent basal dendrites of the GCs in epilepsy are poorly understood.
Recent studies reported that febrile seizures during the development induced aberrant migration of neonatal-generated GCs, resulting in GC ectopia persisting into adulthood (Koyama et al. 2012). Moreover, this study provided a statistical correlation between ectopic GCs and the development of epilepsy in adult animals. The data indicate that a developmental migration defect of the GCs induced by febrile seizures results in GCD, thereby altering the circuitry of the dentate gyrus and increasing the animals' susceptibility to develop epilepsy later in life (Koyama et al. 2012). The findings are in line with previous studies showing an increased number of excitatory synapses and epileptiform activity in hilar ectopic GCs (Dashtipour et al. 2001; Scharfman et al. 2007).
On the contrary, GCD and aberrant basal dendrites can be induced experimentally in adult animals (Spigelman et al. 1998; Bouilleret et al. 1999; Ribak et al. 2000; Heinrich et al. 2006; Duveau et al. 2011; Häussler et al. 2012). Unilateral injection of kainate (KA), an agonist of the excitatory neurotransmitter glutamate, into the hippocampus of adult mice not only induced epileptic GC activity and epileptic seizures, but also GCD, on the side of KA injection but not on the contralateral side (Bouilleret et al. 1999; Heinrich et al. 2006). It has been shown previously that postnatal neurogenesis of dentate GCs is abolished near the site of KA injection where GCD is maximal (Heinrich et al. 2006; Nitta et al. 2008; Murphy et al. 2012; Häussler et al. 2012; Sibbe et al., 2012), making it unlikely that GCD develops by the aberrant migration of adult-generated GCs. Does GCD in epilepsy result from migratory activity of fully differentiated GCs despite that they are fixed in place by thousands of synapses?
In the present study, we hypothesized that differentiated GCs might become motile and change their positions following the induction of epileptiform activity, eventually resulting in GCD. To monitor the motility of differentiated GCs following KA application by real-time microscopy, we used slice cultures of hippocampus from mice that express enhanced green fluorescent protein (eGFP) under the control of the Thy1 promoter in differentiated telencephalic neurons, including dentate GCs (Feng et al. 2000). In these mice, eGFP is expressed in a mosaic manner, allowing one to follow dendritic and axonal processes of individual neurons in a way similar to Golgi impregnation.
Previous studies have revealed that GCD formation in human TLE and KA-induced epilepsy in mice is associated with decreased expression of the extracellular matrix protein Reelin (Haas et al. 2002; Heinrich et al. 2006; Gong et al. 2007; Duveau et al. 2011; Tinnes et al. 2011). Moreover, application of Reelin-neutralizing antibodies to the hippocampus of naïve, mature animals was sufficient to induce GCD (Heinrich et al. 2006). These findings pointed to a causal role of Reelin in the stabilization of a compact GCL, likely by phosphorylation of the actin-depolymerizing protein cofilin (Chai et al. 2009). Reelin deficiency, in turn, was found to result in decreased cofilin phosphorylation associated with destabilization of the actin cytoskeleton and increased neuronal motility (Chai et al. 2009; Frotscher 2010). Therefore, we also studied Reelin expression in control slice cultures and in slice cultures exposed to KA.
Materials and Methods
Experiments were carried out with transgenic mice in which eGFP is expressed under the control of the Thy1 promoter (M-line, C57BL/6 background). All animal procedures were carried out in accordance with the guidelines of the European Community's Council Directive of 22 September 2010 (2010/63/EU) and approved by the regional council and local animal welfare officer according to the German animal protection act. Genotypes were confirmed by PCR analysis of genomic DNA (Vuksic et al. 2008).
Intrahippocampal KA Injection
Adult male (9–12 weeks of age) Thy1-eGFP mice were used for unilateral, intrahippocampal KA injections as described previously (Heinrich et al. 2006; Häussler et al. 2012). In brief, anesthetized mice (ketamine hydrochloride 100 mg/kg, xylazine 5 mg/kg, atropine 0.1 mg/kg body weight, i.p.) were stereotaxically injected with 50 nL (1 nmol) of a 20 mM KA solution (Tocris, Bristol, UK) in 0.9% sterile saline into the right dorsal hippocampus (coordinates relative to bregma: anterioposterior (AP) = −2.0 mm, mediolateral (ML) = −1.4 mm, dorsoventral (DV) = −1.8 mm). After recovery from anesthesia, mice were kept under observation for several hours. Behavioral status epilepticus was verified, characterized by mild convulsive movements, chewing, rotations, or immobility, as previously described (Riban et al. 2002; Heinrich et al. 2006). Only mice that had experienced status epilepticus after KA injection were allowed to survive for different periods of time (7 days, n = 5; 14 days, n = 4; 21 days, n = 5).
Perfusion and Tissue Preparation
Mice were deeply anesthetized (see above) and transcardially perfused with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB, pH 7.4) for 10 min followed by postfixation for 4 h at 4 °C. Brains were rinsed in PB (overnight, 4 °C) and were cut (coronal plane, 50 µm) on a vibratome (VT1000S, Leica, Bensheim, Germany).
Tissue sections (frontal brain sections, 50 µm, or resliced cultures, 20 µm) were immunolabeled using a free-floating protocol (Heinrich et al. 2006). After preincubation (0.25% Triton X-100, 10% normal serum in PB, 30 min), sections were incubated for 4 h at room temperature (RT) and subsequently overnight (4 °C) with the following primary antibodies: rabbit polyclonal anti-Prox1 (1:1000; Chemicon), mouse monoclonal anti-glutamate decarboxylase-67 (GAD67; 1:1000; Sigma) and mouse monoclonal anti-Reelin (G-10, 1:1000; Chemicon). Antibody binding was visualized by incubation with appropriate secondary antibodies conjugated with Cy3 (1:400, Jackson ImmunoResearch Laboratories) or Alexa Fluor 488, 568, and 647 (1:300; Invitrogen), in the dark (4 h, RT). Sections were coverslipped with antifading mounting medium (IMMU-Mount, Thermo Fisher Scientific) and photographed using a Zeiss Axioplan 2 fluorescence microscope equipped with an ApoTome setting. Alternatively, confocal images were captured with an inverted laser-scanning microscope (TCS SP2, Leica).
Preparation of Organotypic Hippocampal Slice Cultures and KA Application In Vitro
Slice cultures of hippocampus have been shown to maintain organotypic characteristics such as the arrangement of pyramidal neurons and GCs in compact layers and cell-type-specific features, including characteristic input- and output synapses of the GCs (Frotscher and Gähwiler 1988; Stoppini et al. 1991; Gähwiler et al. 1997). For the preparation of hippocampal slice cultures, 5- to 8-day-old mouse pups (n = 80) were used. Brains were removed following decapitation under hypothermic anesthesia. The hippocampi were dissected, in most cases together with the entorhinal cortex to preserve a major afferent fiber projection to the hippocampus and dentate gyrus. The tissue was sliced (300 μm) perpendicular to the longitudinal axis of the hippocampus using a McIlwain tissue chopper. The slices were placed onto culture inserts (Millipore) and transferred to 6-well plates with 1 mL/well nutrition medium containing 25% heat-inactivated horse serum, 25% Hank's balanced salt solution (Invitrogen), 50% minimal essential medium, and 2 mM glutamine (pH 7.2). Slices were incubated as static cultures (Stoppini et al. 1991) in 5% CO2 at 37 °C for at least 10 days in vitro (DIV) before experiments started; the medium was changed every second day. Then, the cultures were incubated in medium containing KA (100 nM, if not mentioned otherwise) or control medium and continuously imaged for 24 h (n = 12 cultures for each group).
Control and KA-treated cultures were kept in incubation medium and imaged by an inverted confocal fluorescence microscope (LSM 510, Carl Zeiss MicroImaging) equipped with an Argon 488-nm scanning laser. Images were captured with a Zeiss Axiocam digital camera (10× objective with an additional electronic zoom factor of 1.7) for 24–26 h. A closed chamber with a heated cover (37 °C and a humidified 5% CO2 atmosphere) was used (Tokai hit, Spectra Services, Inc.). To capture rapid dynamic events in the slice cultures, stacks of 5–8 optical sections (512 × 512 pixel array) were collected every 15 min with 7–8 µm intervals of z-step size. The lowest laser power setting was used to avoid photodamage. Acquired time-lapse movie stacks were z-projected (summation) and converted to a maximum projection image by using Zen2010 software (LSM 510, Carl Zeiss MicroImaging). The moved distances of 19 single GCs each from control cultures and KA-treated cultures during a time period of 24 h were directly measured in digital images of time-lapse sequences, taking the middle of the cell body and surrounding structures as reference points. Only neurons were selected that were clearly visible in the cultures and displayed the majority of their dendrites allowing for their identification as differentiated GCs. In contrast, GCs that showed fragmented dendrites and rounding-up of their cell somata as obvious signs of neuronal degeneration were not used for these measurements of GC motility. It should be pointed out, however, that these measurements could not be performed with the investigator blind to the experimental condition since the differences in GC motility were dramatic between KA-treated slice cultures and control cultures (see Supplementary Videos S1 and S2). From these data, the speed of cell nucleus movement was calculated (mean ± SEM). For Figure 4, cells from each group were selected, and single optical sections from different time points were shown. Pictures were processed using Photoshop CS3 (Adobe).
Whole-Cell Patch-Clamp Recordings
Slice cultures were transferred to a recording chamber and continuously superfused at near physiological temperature (30–31 °C) with artificial cerebrospinal fluid containing (in mM): 125 NaCl, 25 NaHCO3, 25 glucose, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, and 1 MgCl2 (equilibrated with 95% O2 and 5% CO2). Patch pipettes were pulled from borosilicate glass tubing (2 mm outer diameter, 1 mm inner diameter) on a Sutter P-87 puller (Sutter Instrument Company). The intracellular pipette solution contained (in mM): 135 K-gluconate, 20 KCl, 2 MgCl2, 0.1 EGTA, 10 HEPES, 2 Na2ATP. The resistance of the filled electrodes was between 4.5 and 5.6 MΩ. Patch-clamp recordings were obtained from visually identified GCs using infrared differential interference contrast video microscopy on an upright microscope equipped with a × 40 water immersion objective (Zeiss) and a video camera (C2400-07, Hamamatsu). All measurements were done in current-clamp mode at resting membrane potential. The series resistance was fully compensated. Recordings were performed using an Axopatch 200B amplifier (Axon Instruments). Resting membrane potential, action potential threshold, and input resistance of each cell were measured immediately after break-in. Recordings were obtained from 10 control cultures and 10 cultures to which KA was applied during the experiment.
Development of GCD In Vitro
Our in vivo studies have shown that GCD develops over several weeks. Thus, following the initial incubation for 10 days the cultures for these studies were incubated on 3 consecutive days for 8 h each with KA (either 3, 5, 10, or 15 µM) followed by washout with incubation medium. After an additional incubation period of 7 days, the cultures were fixed and immunolabeled for Prox1 and Reelin (n = 10 animals for each group; 1 culture was used from each animal). In KA-treated cultures (15 µM) and in control cultures incubated in vitro for the same period of time, the width of the GCL was measured as described (Haas et al. 2002; Tinnes et al. 2011). Measurements were taken in 100 µm intervals along the borders of the GCL using Axiovision software. The mean widths were calculated and used to determine the average layer width for each condition.
In Situ Hybridization
These experiments were aimed at analyzing rapid KA effects on reelin mRNA expression. Therefore, following the initial incubation period the slice cultures were incubated for 45 min in KA (either 3, 5, 10, or 15 µM) followed by washout in fresh medium (6 h). For in situ hybridization (ISH), the slices were fixed in buffered 4% paraformaldehyde for 24 h at 4 °C. ISH for reelin mRNA was performed on whole slices with digoxigenin-labeled riboprobes as described previously (Heinrich et al. 2006). All reelin mRNA-positive cells in the hilar region between the 2 blades of the GCL were counted (1 slice/animal; 5 animals per condition).
Propidium iodide (PI) is not taken up by living cells and is, therefore, commonly used to stain degenerating neurons (e.g. Bruce et al. 1996). In the present study, PI staining was used to identify degenerating neurons following KA application. PI (Fluka, Cat. No.: 81845) was diluted in culture medium (5 µg/mL).
Results are presented as mean ± SEM. Differences between groups were tested for statistical significance (Student's t-test or 1-way ANOVA with Tukey's multiple comparison test). Significance level was set at P < 0.05.
KA-Induced GCD In Vivo Involves Differentiated Neurons
The dentate gyrus of the hippocampal formation is one of the few brain regions in which adult neurogenesis is known to occur (Altman and Das 1965; Kaplan and Hinds 1977; Eriksson et al. 1998; Zhao et al. 2008; Mongiat and Schinder 2011). The broadening of the GCL observed in GCD might therefore result from enhanced adult neurogenesis that is associated with epilepsy (Parent et al. 1997) and aberrant migration of the newly generated GCs or, alternatively, from displacement of more mature GCs located in the superficial portions of the GCL. To address this point, we first aimed to determine the extent to which differentiated GCs were involved in GCD formation in vivo and took advantage of transgenic mice expressing eGFP in differentiated GCs (Feng et al. 2000; Vuksic et al. 2008). As eGFP is expressed in these animals in a mosaic, Golgi stain-like manner, the sections were counterstained by immunolabeling for Prox1, a marker of the GCs, to provide an overview of the GCL as a whole.
Over a time period of 21 days after KA injection, GCD developed in the ipsilateral (injected) hippocampus, but not in the contralateral hippocampus (Fig. 1A–D) and involved the majority of GCs in the dentate gyrus. This contrasted to the severely reduced number of newly generated GCs at sites of GCD (Heinrich et al. 2006; Nitta et al. 2008; Häussler et al. 2012), clearly indicating that the few adult-born GCs in these mature (9–12 weeks old) mice cannot account for the massive broadening of the GCL. At all time points examined after KA injection differentiated Thy1-eGFP-labeled GCs were observed throughout the entire, increasing thickness of the GCL, and there was no evidence for newly generated GCs in the subgranular zone displacing these differentiated neurons toward more superficial positions in the GCL. When compared with GCs in the contralateral dentate gyrus or in control animals, the angle between apical GC dendrites appeared larger as also observed in Golgi-impregnated GCs in tissue samples from epileptic patients (Freiman et al. 2011). Like in patient tissue, we also found GCs with basal dendrites (Fig. 2A,B).
Together with our previous studies in which we quantified the numbers of bromodeoxyuridine (BrdU)-positive and BrdU/doublecortin (DCX)-positive cells (Häussler et al. 2012) as well as sex-determining region Y-box 2 (Sox2)-positive progenitors (Sibbe et al. 2012), the present results confirm that GCD on the side of KA injection is not resulting from increased neurogenesis and aberrant migration of these newly generated neurons. Our findings rather suggest that fully differentiated GCs participate in GCD formation.
KA Application In Vitro Results in GCD Formation
As observed in vivo (Figs 1 and 2), individual differentiated GCs in control cultures and cultures exposed to KA were eGFP-labeled in a Golgi stain-like manner, showing their axonal processes, dendritic branches, and spines (Fig. 3A).
With the concept that repeated application of the glutamate receptor agonist KA might mimic excessive glutamate release during repeated seizures, we incubated the cultures with KA for 8 h per day over 3 consecutive days. As observed under in vivo conditions, the width of the GCL increased significantly in the KA-treated slice cultures (control: 78.8 ± 2.4 µm; KA: 123.6 ± 5.0 µm; mean ± SEM; P < 0.001; n = 10 slice cultures per group; Student's t-test; Fig. 3B). In addition, we studied GCD development using different concentrations of KA and observed a concentration-dependent formation of GCD in vitro (Supplementary Fig. S1). Collectively, these results reveal slice cultures as an appropriate model to study the cellular and molecular mechanisms underlying KA-induced GCD formation.
KA-Treated Slice Cultures Show Burst Activity of GCs
Unilateral injection of KA into the hippocampus of rodents results in recurrent seizures and epileptiform GC activity, respectively (Bouilleret et al. 1999; Riban et al. 2002; Häussler et al. 2012). As a next step, we studied whether similar GC activity was induced in GCs of hippocampal slice cultures that were incubated in the presence of KA.
Patch-clamp recordings of GCs in slice cultures under control incubation conditions revealed similar passive intrinsic properties as GCs in acute hippocampal slices of mice (Vuksic et al. 2008; Young et al. 2009). In contrast to GCs in control cultures, GCs of KA-treated cultures had a more hyperpolarized resting membrane potential (control: −78.3 ± 1.8 mV, n = 10; KA: 84.5 ± 2.7 mV, n = 8, P < 0.05) and a significantly reduced input resistance (controls: 507.7 ± 69.6 MΩ, n = 10; KA: 259.9 ± 28.1 MΩ, n = 8, P < 0.01). Both changes point to a decrease in intrinsic membrane excitability. However, while control GCs showed excitatory postsynaptic potentials (EPSPs) and rare action potentials, GCs exposed to KA gave rise to burst activity (Fig. 3C). Repeatedly occurring bursts of action potentials with massive and long-lasting depolarizations (several hundreds of milliseconds) were recorded in these cultures (on average, 11.4 ± 2.2 bursts per minute). In contrast, no bursts were recorded in GCs of control cultures (P < 0.01) as similarly observed by other investigators (Franck et al. 1995; Routbort et al. 1999). The effects of KA application were particularly impressive when GCs were first recorded under control conditions for extended periods of time (up to 2 h) and then exposed to KA for 3 min. Together, the results show that application of KA to hippocampal slice cultures leads to changes in GC activity that mimic those observed after unilateral KA injection into the hippocampus in vivo.
Real-Time Microscopy of KA-Treated Slice Cultures
With the concept that an increased motility of differentiated GCs would result in a broadening of the GCL and eventually in GCD, slice cultures were exposed to KA and then continuously imaged for a time period of 24 h (see Materials and Methods section). We took advantage of the mosaic expression of eGFP in single differentiated GCs and studied the movement of individual neurons by real-time microscopy. While GCs under control conditions remained almost unchanged during a recording period of 24 h with only minor dendritic movements (Fig. 4; Supplementary Video S1), there was a significant increase in GC motility in the KA-treated cultures (Fig. 4; Supplementary Video S2). The nuclei of KA-treated GCs moved into dendritic processes (speed of nuclear movement following KA application: 0.238 ± 0.095 µm/h; control: 0.062 ± 0.02 µm/h; mean ± SEM; P < 0.01; 19 cells per condition, recording time: 24 h). The highest speed of all GC nuclei analyzed following KA application was 0.67 µm/h, compared with 0.15 µm/h under control conditions. This nuclear movement is reminiscent of nuclear translocation as observed during neuronal migration in development (Rakic 1971, 1972; Nadarajah et al. 2001; Nadarajah and Parnavelas 2002; Kriegstein and Noctor 2004). Nuclear translocation into a dendritic process resulted in a more dispersed GCL and in a reorientation of dendrites that now often originated from the basal pole of the cell body (Fig. 4). Such recurrent basal dendrites are a characteristic feature of GCs in TLE and in experimental epilepsy models (Spigelman et al. 1998; Ribak et al. 2000; Freiman et al. 2011; Murphy et al. 2012). They develop by nuclear translocation and are not the result of an abnormal growth process (Murphy and Danzer 2011). In addition to nuclear translocation into one of the dendritic processes, we also observed that some eGFP-labeled GCs in the KA-treated cultures showed dendritic fragmentation and rounding up of their cell bodies, thus indicating degenerative changes (Supplementary Video S2). Such obviously degenerating GCs were not included in the quantitative studies of GC motility; however, these degenerative changes might contribute to neuronal repositioning and GCD development.
GCD is Associated with Decreased Expression of Reelin
Since previous studies had shown that GCD in human TLE and KA-induced epilepsy in mice is associated with decreased expression of the extracellular matrix protein Reelin (see Introduction section), we studied Reelin expression in slice cultures exposed to KA. We observed a significant loss of reelin mRNA-synthesizing neurons in the hilar region after 45 min of incubation in the presence of KA (Fig. 5A). Moreover, the sections used for GCD measurements were double-labeled for Prox1 and Reelin and showed a dramatic loss of Reelin-immunoreactive cells in the subgranular zone and hilus (Fig. 5B). Of note, both reelin mRNA-expressing and Reelin-immunoreactive cells in the outer molecular layer, the marginal zone of the dentate gyrus, were less affected (Fig. 5A,B). It has been reported that in addition to Cajal-Retzius cells in the marginal zone, GABAergic interneurons in the subgranular zone and hilus express Reelin in the postnatal period (Alcántara et al. 1998; Drakew et al. 1998; Pesold et al. 1998; Ramos-Moreno et al. 2006). This KA-induced loss of Reelin-immunoreactive cells thus mainly appeared in a region nearby the GCL, where we observed an increased GC motility after KA application.
The loss of reelin mRNA-expressing and Reelin-immunoreactive cells in the hilar region raised the issue of neurodegenerative changes following KA application. Moreover, it could not be excluded that rapid degeneration of many GCs contributed to the increased motility of eGFP-positive GCs as observed in our real-time microscopy studies. Indeed, we noticed that some eGFP-stained GCs rounded up, accompanied by the fragmentation of their dendritic processes. Therefore, we studied neuronal degeneration in the hilar region and in the GCL in control cultures and cultures exposed to different concentrations of KA. With 100 nM KA, the concentration used in our imaging studies, no statistically significant increase in the number of degenerating, PI-positive neurons was observed in the KA-treated cultures following the 24-h period used for imaging (Fig. 6A,B,E,F), making it unlikely that the KA-induced changes in GC motility were caused by massive degenerative changes. With 1.5 µM KA, we observed a significant increase in the number of PI-positive cells (Fig. 6E,F); however, this increase was much more pronounced in the hilus than in the GCL. Interestingly enough, even with a KA concentration of 15 µM we observed relatively few PI-positive, degenerating GCs, which was in stark contrast to abundant PI-positive CA3 pyramidal cells and hilar neurons, including some GAD67-immunoreactive, putative GABAergic inhibitory neurons (Fig. 6C,D,E,F). These findings are in line with previous studies that showed relative vulnerabilities of dentate neurons in epilepsy (Sloviter et al. 2003). Comparable with the results of the present in vitro experiments, epilepsy-induced neuronal degeneration in vivo mainly involved CA3 neurons, hilar mossy cells and various types of GABAergic interneuron but largely spared the GCs. However, as described previously, many GCs express GAD in response to heavy stimulation (Sloviter et al. 1996; Schwarzer and Sperk 1995; Fig. 6D).
A characteristic feature of the mammalian dentate gyrus is its laminated organization (Förster et al. 2006). In the present study, we have shown that epilepsy-induced GCD, a spread of the normally tightly packed cell bodies of the GCL, involves increased motility of fully differentiated neurons. Following application of the glutamate receptor agonist KA, the nuclei of GCs moved into dendritic processes resulting in a broadening of the GCL and in dendritic reorientation, including the formation of recurrent basal dendrites. We hypothesize that GCD formation in human TLE, which is associated with excessive glutamate release, similarly involves somal translocation in differentiated GCs. Repeated seizures and KA application, respectively, might eventually lead to an increasing spread of GC somata, appearing as GCD in sections of tissue samples from epileptic patients or Prox1-immunolabeled slice cultures.
We cannot exclude that neuronal degeneration in the hilar region and GCL contributes to GCD, particularly following recurrent seizures and repeated KA applications. In our real-time microscopy studies, we observed in fact that some eGFP-labeled GCs in the KA-treated cultures rounded up and showed dendritic fragmentation as signs of neuronal degeneration. Moreover, incubation of the slice cultures for 24 h in a high concentration of KA (15 µM) resulted in significant neuronal degeneration, mainly in the CA3 region and hilus, as visualized by staining for PI. Of note, even with this concentration (the highest concentration of KA used in the present study) neuronal degeneration in the GCL only affected ∼4% of the GCs (compared with >30% of hilar neurons; Fig. 6E,F). Neuronal degeneration is followed by glia hypertrophy that eventually might contribute to GCD by displacing GC somata and processes (Heinrich et al. 2006). We would like to point out, however, that with 100 nM KA as used for the imaging experiments, no statistically significant differences in the number of PI-stained degenerating neurons neither in the GCL nor hilus were found between KA-exposed cultures and control cultures after the 24-h imaging period. We conclude that the massive GCD found in epilepsy is likely to be the result of several processes. While nuclear translocation and dendritic reorientation are likely to predominate in an initial phase following strong glutamate receptor activation, secondary changes such as neuronal degeneration and glial hypertrophy may contribute to GCD mainly during later stages.
Like GCD in human TLE and in KA-induced epilepsy in vivo, KA-induced GCD in slice culture was associated with decreased Reelin expression, particularly by interneurons in the hilar region. Following incubation in the presence of KA the numbers of hilar reelin mRNA expressing and Reelin-immunoreactive neurons were significantly decreased. As discussed below, we hypothesize that this Reelin deficiency underlies the increased GC motility observed following KA application.
Granule Cell Dispersion: Cause or Consequence of Epilepsy?
Postmitotic neurons stop migrating when they have arrived at their destinations and formed synaptic connections both with incoming afferents and target cells (however, see Morozov et al. 2006, for the migration of synaptically connected interneurons). Hence, altered neuronal positioning associated with epilepsy has been deemed to be the result of a migration defect during development (Flint and Kriegstein 1997; Chevassus-au-Louis et al. 1999; Jacobs et al. 1999; Guerrini and Parrini 2010). It appeared plausible that aberrant GC migration during the development of the dentate gyrus resulted in GCD and was associated with aberrant neuronal connections and altered network activity, eventually leading to epileptic activity (Koyama et al. 2012). Indeed, dispersion of the GCL would enable axonal collaterals of superficial GCs to get in contact with the dendrites of deeply located GCs. Under normal conditions, the molecular layer containing GC dendrites corresponds to the input side and is clearly separated from the output side, which comprises the GC axons (mossy fibers), and connections with other GCs are rare when compared with hilar ectopic GCs (Dashtipour et al. 2001).
It was therefore important to demonstrate that GCD does not necessarily cause epilepsy but might result from epileptic activity. While this was, of course, difficult to determine in human TLE, it could be shown in experimental epilepsy in rodents induced by application of the glutamate receptor agonist KA. Unilateral injection of KA into the hippocampus induced epileptiform GC discharges and recurrent seizures followed by the development of GCD on the injected side, but not on the contralateral side (Bouilleret et al. 1999; Riban et al. 2002; Heinrich et al. 2006; Häussler et al. 2012). Of note, adult neurogenesis was severely decreased adjacent to the injection site, whereas GCD turned out to be maximal near the site of excitotoxin injection (Häussler et al. 2012). Application of BrdU, a marker of newly generated cells, labeled newly formed microglia and astrocytes, mainly in the hilar region of the dentate gyrus near the site of KA injection, but very few new GCs (Kralic et al. 2005; Heinrich et al. 2006; Nitta et al. 2008; Häussler et al. 2012). These results were supported by significantly decreased Notch signaling following KA injection, pointing to a disruption of the stem cell niche (Sibbe et al., 2012). The results of the present study, which revealed an increased motility of GC nuclei, have confirmed that the development of GCD involves changes in differentiated GCs.
Somal Translocation in Differentiated GCs Results in GCD
Somal translocation is a characteristic feature of migrating neurons during brain development, irrespective whether the leading process is guided by glial fibers or aligned with other neuronal processes (Rakic 1971, 1972; Nadarajah et al. 2001; Nadarajah and Parnavelas 2002; Kriegstein and Noctor 2004; Borrell et al. 2006; Cooper 2008). Recent studies have shown that the leading tip of migrating neurons pulls the soma forward through a myosin II-dependent forward flow of F-actin within the leading process (He et al. 2010). SUN-domain proteins, SUN1 and SUN2, and the KASH-domain proteins Syne-1/Nesprin-1 and Syne-2/Nesprin-2 seem to be involved by connecting the nuclear envelope with microtubule-based motor proteins (Zhang et al. 2009). The results of the present study revealed that nuclear translocation does take place in differentiated GCs, thus confirming an unexpected potential of mature neurons for changes in motility (Morozov et al. 2006; Murphy and Danzer 2011). It remains to be shown whether similar molecular mechanisms are involved in epilepsy-induced nuclear translocation in differentiated neurons as described for migrating neurons during brain development.
Somal translocation in differentiated GCs in vitro has been described before (Murphy and Danzer 2011). We here show that this process is significantly enhanced in GCs by KA-induced epileptiform activity, which, in turn, resulted in a significant broadening of the GCL in vitro. It is reasonable to conclude that the translocation of the soma into one of the cell's dendrites is similarly enhanced following KA application in vivo and in human TLE and contributes to the dispersed GCL found on the side of KA injection and in tissue biopsies from TLE patients, respectively. As we induced GCD by applying the glutamate receptor agonist KA, we have reason to assume that GCD in human TLE, associated with increased glutamate release during seizure activity, develops in a similar way. The results of the present study do not exclude that newly generated, Thy1-eGFP-negative, neurons migrate abnormally and thus participate in the formation of GCD.
Decreased Expression and Altered Processing of Reelin Might be Involved in GC Motility
While the precise signaling cascades following excessive glutamate receptor activation in epilepsy remain to be elucidated, we confirm and extend previous findings indicating that a Reelin deficiency is involved in the development of GCD. Thus, it was previously shown that infusion of Reelin-blocking antibodies (CR-50), but not of unspecific IgG, into the hippocampus of normal mice induced GCD (Heinrich et al. 2006). Moreover, infusion of recombinant Reelin on the same side as the KA injection significantly reduced the formation of GCD (Müller et al. 2009). These findings established a role for Reelin in the stabilization of the mature laminated cortical architecture (Haas et al. 2002; Heinrich et al. 2006; Gong et al. 2007; Frotscher 2010; Duveau et al. 2011; Tinnes et al. 2011).
In the present experiments, ISH for reelin mRNA as well as immunostaining for Reelin protein revealed a KA-induced loss of labeled interneurons in the subgranular zone and hilus, that is, close to the granular layer where an increase in GC motility was observed. Degeneration of GABAergic, Reelin-containing interneurons might contribute to this cell loss because we observed an increased number of PI-positive, degenerating hilar neurons in the KA-treated cultures, including GAD67-immunoreactive cells. However, it remains to be shown to what extent degeneration of Reelin-synthesizing cells or changes in Reelin expression predominate. Previous studies have provided evidence for changes in reelin gene expression since a hypermethylation of the reelin promoter region was found in tissue samples from epileptic patients (Kobow et al. 2009). Finally, altered Reelin processing in response to KA application (Duveau et al. 2011; Tinnes et al. 2011) might also be involved.
What are the consequences of decreased Reelin availability? In support of a role for Reelin in the stabilization of hippocampal architecture, it was recently demonstrated that Reelin induces serine3 phosphorylation of cofilin via Apolipoprotein E receptor 2 (ApoER 2), Disabled1 (Dab1) and LIM kinase1 (Chai et al. 2009). Cofilin is an actin-associated protein involved in the depolymerization of F-actin. Depolymerization of F-actin is required for the reorganization of actin filaments in various processes associated with changes in cell shape such as growth and migration (Bamburg 1999; Pollard and Borisy 2003; Jovceva et al. 2007; Kiuchi et al. 2007). Serine3 phosphorylation of cofilin renders it unable to depolymerize F-actin, thereby stabilizing the actin cytoskeleton and terminating migratory activity (Arber et al. 1998; Yang et al. 1998). Reelin-induced cofilin phosphorylation is therefore likely to be involved in the “stop signal” function of Reelin that prevents the invasion of neurons into the Reelin-rich marginal zone during the development of the cortex. In reeler, the marginal zone is densely populated reminiscent of the invasion of GCs into the dentate molecular layer and hilus in epilepsy-induced GCD associated with deficient reelin expression. Collectively, these results suggest that Reelin-induced cofilin phosphorylation controls neuronal migration during cortical development and prevents abnormal neuronal motility in the mature CNS.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (TR-3: D6, D7 to C.A.H. and M.F., and FR 620/12-1 to M.F.).
M.F. is Senior Research Professor of the Hertie Foundation. The authors thank Guoping Feng for providing the Thy1-eGFP mice, Huan Long and Shaobo Wang for genotyping, Marie Follo for her help with real-time microscopy, Susanne Huber for excellent technical assistance, and Sandra Dieni for helpful comments on the manuscript.
Conflict of Interest: None declared.