It is a matter of ongoing debate whether newly generated granule cells contribute to epileptic activity in the hippocampus. To address this question, we investigated neurogenesis and epileptiform activity (EA) along the hippocampal septotemporal axis in the intrahippocampal kainate (KA) mouse model for temporal lobe epilepsy. Multisite intrahippocampal in vivo recordings and immunolabeling for c-Fos showed that the KA-induced status epilepticus (SE) extended along the septotemporal axis of both hippocampi with stronger intensity at ipsilateral temporal and contralateral sites. Accordingly, we found a position-dependent increase in proliferation (incorporation of bromodeoxyuridine) and neurogenesis (immunolabeling for doublecortin): Both were selectively increased in the ipsilateral temporal and entire contralateral subgranular zone, sparing the septal region close to the injection site. The newborn neurons were hyperexcitable and functionally integrated into the hippocampal network as revealed by patch-clamp recordings. Analysis of chronic EA also showed a differential intensity pattern along the hippocampal axis: EA was low in the septal portion with prominent sclerosis and granule cell dispersion but most pronounced in the transition zone where neurogenesis reappeared. In conclusion, SE stimulates neurogenesis in a position-dependent manner and coincidence of neurogenesis and stronger EA distal to the injection site suggests a proepileptogenic effect of increased neurogenesis.
In mesial temporal lobe epilepsy (MTLE), patients suffer from focal epileptic seizures that often originate from the hippocampal formation. MTLE is frequently associated with hippocampal sclerosis, which includes cell death in CA1 and CA3, gliosis, and granule cell dispersion (GCD) (Houser 1990; Thom et al. 2002). GCD represents an abnormal broadening of the granule cell layer that has been suggested being due to wrong migration of granule cells born after status epilepticus (SE) (Parent et al. 1997; Haas et al. 2002; Jessberger et al. 2005). However, we and others have shown that GCD results from displacement of adult, instead of newly generated, neurons since in the intrahippocampal kainate (KA) mouse model for MTLE, GCD formation is accompanied by a loss of neurogenesis (Kralic et al. 2005; Heinrich et al. 2006; Nitta et al. 2008). This observation was astonishing since SE and/or epileptic seizures have been shown to stimulate neurogenesis in the hippocampus of rodents, in particular in systemic models for MTLE (Parent et al. 1997; Gray and Sundstrom 1998; Jessberger et al. 2005).
Addressing this controversy might be of particular relevance for the understanding of mechanisms underlying the generation of epileptiform activity (EA) since immature granule cells are hyperexcitable compared with adult granule cells (Schmidt-Hieber et al. 2004). Indeed, a proepileptic effect of increased neurogenesis has been suggested (Scharfman et al. 2000; Jung et al. 2004; Kwak et al. 2008), but there is also evidence that neurons born after SE integrate into the network in an antiepileptic fashion (Jakubs et al. 2006). Thus, the role of neurogenesis in epilepsy remains to be clarified. A particular problem is that the rodent epilepsy models that show increased neurogenesis often do not develop GCD, which is characteristic for many human cases of MTLE (Thom et al. 2002; Fahrner et al. 2007).
Here, we investigated in the whole hippocampal formation how neurogenesis spatially relates to the extension of SE and recurrent EA in the intrahippocampal KA mouse MTLE model. KA-injected mice show prominent GCD close to the injection site (Suzuki et al. 1995; Bouilleret et al. 1999; Heinrich et al. 2006) and EA, which involves the ipsi- and contralateral hippocampus (Meier et al. 2007). We show that SE triggers indeed a spatially selective increase in neurogenesis—mutually exclusive with GCD. Functional integration and hyperexcitability of the newborn cells, together with an overlap with increased EA at corresponding positions suggest a proepileptogenic effect of neurogenesis.
Materials and Methods
Experiments were carried out with young-adult (7–10 weeks) male C57Bl/6 mice (Biomed Center, University of Freiburg) or with transgenic doublecortin (DCX)-promoter-DsRed2 mice (Couillard-Despres et al. 2006), in accordance with the guidelines of the European Community's Council Directive of 24 November 1986 (86/609/EEC) and approved by the regional council.
KA Injection and Electrode Implantation
Unilateral intrahippocampal KA injections (50 nL, 20 mM) were performed as previously described (Müller et al. 2009) (coordinates relative to bregma: anterio-posterior [AP] = −2.0 mm, mediolateral [ML] = −1.4 mm, dorsoventral [DV] = −1.8 mm). Controls were injected with 0.9% saline. A subgroup of mice (KA: n = 9, saline: n = 4) were immediately implanted with platinum–iridium wire electrodes (Ø 125 μm, Teflon insulated; World Precision Instruments) at four positions along the septotemporal axis of the hippocampus (1) AP = −2.0, ML = −1.4, DV = −1.9; 2) AP = −2.8, ML = −2.0, DV = −2.0; 3) AP = −3.4, ML = −2.75, DV = −2.75; 4) AP = −3.8, ML = −2.5, DV = −4.0). In some of these mice (KA: n = 5), an additional electrode was implanted into the septal contralateral hippocampus (AP = −2.0, ML = +1.4, DV = −1.9). A stainless steel jeweler's screw above the frontal cortex served as the reference. Electrodes were fixed to the skull with cyanoacrylate and dental cement and soldered to a permanently mounted connector. Mice destined for later immunohistochemical or patch-clamp analysis were not implanted with electrodes but kept under observation for several hours after surgery to verify behavioral SE, characterized by mild convulsive movements, chewing, rotations, or immobility, as previously described (Riban et al. 2002).
In Vivo Intrahippocampal Recordings
For in vivo recordings of hippocampal local field potentials (LFPs), mice were connected to a miniature preamplifier (Multi Channel Systems). Signals were amplified (1000-fold, bandpass 1 Hz–5 kHz), digitized (sampling rate 10 kHz, Power1401 analog-to-digital (A/D) converter, Spike2 software; Cambridge Electronic Design), and stored on hard disk. Mice were recorded for 2–3 h during SE and at 3, 7, 14, and 21 days after KA injection. Only recordings from electrodes located in the molecular or granule cell layer, verified in Nissl-stained sections (see below), were accepted.
Patch-Clamp Analysis and Single-Cell Reconstructions
Transgenic DCX-DsRed2 mice (3–4 weeks after KA injection, n = 4) were anesthetized with isoflurane, decapitated, and their brains were immersed in ice-cold artificial cerebrospinal fluid (ACSF) containing (in mM): 87 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2, 7 MgCl2, 75 sucrose, and 10 glucose (equilibrated with carbogen: 95% O2/5% CO2). Transverse slices (350 μm) from the temporal hippocampus were collected as described in Bischofberger et al. (2006) and incubated (30 min, 36 °C). Patch-clamp recordings were performed at room temperature in ACSF containing (in mM): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 1 MgCl2, and 25 glucose (equilibrated with carbogen). Patch pipettes pulled from borosilicate glass (Hilgenberg) were filled with a solution containing (in mM): 135 Kgluconate, 20 KCl, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 0.1 ethyleneglycol-bis(2-aminoethylether)-N,N,N′,N′-tetra acetic acid (EGTA), 2 MgCl2, 2 Na2ATP, and 0.2% biocytin (pH = 7.28). A liquid junction potential of 10 mV was subtracted. Immature granule cells were visualized by detection of DsRed2 fluorescence (filter set 14, Axioskop2 FS; Zeiss) and patched under Dodt-IR contrast video microscopy. Recordings were low-pass filtered (10 kHz) and digitized (20 kHz; SEC05LX amplifier, NPI; software: PatchMaster, Heka). Pipette capacitance and series resistances (13–30 MΩ) were compensated in bridge mode via the amplifier. Seal resistances (Rseal) were >14.9 GΩ (16.3 ± 0.4 GΩ). To avoid underestimation of input resistance (Rin), only recordings with Rseal/Rin ratios of >5.3 (8.0 ± 1.3) were accepted. For fiber stimulation (bipolar 10-μs pulses of ±50–90 V), a theta glass capillary with bipolar Ag-/AgCl-coated wires and filled with recording ACSF was positioned on the outer molecular layer. When mentioned, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors were blocked with 1,2,3,4-tetrahydro-7-nitro-2,3-dioxoquinoxaline-6-carbonitrile disodium (CNQX; 50 μM), N-methyl-D-aspartate (NMDA) receptors with D(-)-2-amino-5-phosphonopentanoic acid (D-AP5; 50 μM), and chloride channels (γ-aminobutyric acid [GABA]A receptors) with picrotoxin (PTX; 100 μM). Recordings were analyzed using Igor Pro (WaveMetrcis) and FitMaster (Heka). The Rin values were calculated from slopes of steady-state I/V relations with voltage responses <10 mV from resting membrane potential (Vrest). Slices were fixed in paraformaldehyde (4% PFA in 0.1 M phosphate buffer (PB), pH 7.4), and posthoc reconstruction of recorded cells was performed as described below.
Bromodeoxyuridine (BrdU) was injected at 6 days after KA injection since proliferation has been reported to be strong at this time point (Kralic et al. 2005). Five consecutive intraperitoneal BrdU injections (50 mg/kg body weight in 0.9% saline; Sigma-Aldrich) were made every 3 h during a period of 12 h to acquire a saturating amount of BrdU.
Perfusion and Tissue Preparation
Mice were deeply anesthetized and perfused at 24 h (KA: n = 6; saline: n = 4; naive: n = 4), 7 days (KA: n = 9; saline: n = 5), or 14 days (KA: n = 5; saline: n = 2) after BrdU injections and at 6 h after KA/saline injection for c-Fos immunolabeling (KA: n = 4; saline: n = 4; naive: n = 2) with 4% PFA for 10 min followed by postfixation (4 h, 4 °C). Subsequently, brains were rinsed in PB (overnight, 4 °C) and were cut coronally (50 μm) on a vibratome (VT1000S; Leica).
Immunohistochemistry and Histology
A free-floating protocol was used for immunohistochemistry as described earlier (Heinrich et al. 2006). All tissue sections were treated equally and tissue sections of the same experiment were processed in parallel. For BrdU detection, sections were preincubated in 2 N HCl (30 min, 37 °C), followed by neutralization in 0.1 M Tris-buffered saline (pH 8.5, 10 min). The following primary antibodies were used: goat polyclonal anti-DCX (1:500), rabbit polyclonal anti-c-Fos (1:1000; both Santa Cruz Biotechnology), rat monoclonal anti-BrdU (1:500; Oxford Biotechnology), and mouse anti-polysialylated neural cell adhesion molecule (PSA-NCAM; 1:400, IgM; Chemicon). For detection, Cy™2-, Cy™3-, or Cy™5-conjugated secondary antibodies were used (1:200; Jackson ImmunoResearch Laboratories). Counterstaining was performed with 4′,6-diamidino-2-phenylindole (DAPI; 1:10 000); biocytin-filled recorded cells were detected with FITC-AvidinD (1:500; Vector Laboratories). Sections were coverslipped with antifading reagent (ProLong gold; Invitrogen) or IMMU-Mount (Thermo Fisher Scientific).
For counting of BrdU-positive cells, immunoperoxidase detection was performed using a biotinylated secondary antibody (1:200; Vector Laboratories) and the avidin-biotin complex in PB (Vectastain Elite Kit; Vector Laboratories). Detection was achieved with 0.05% 3,′3′-diaminobenzidine tetrahydrochloride (DAB; Sigma-Aldrich) and 0.002% H2O2. For Nissl staining, tissue sections were mounted on glass slides and immersed in 0.1% cresyl violet solution for 15 min. Following both methods, sections were dehydrated in ethanol, cleared in xylene, and coverslipped with Hypermount (Thermo Fisher Scientific).
Microscopic Analysis and Counting Procedures
Histological sections were analyzed with an Axioplan 2 microscope (Zeiss); photomicrographs were taken with a digital camera and processed with Axiovision software (Zeiss). Single-cell reconstructions were performed with Apotome technology. Identical exposure times were used for immunofluorescence-labeled sections of one experiment to allow comparison.
To quantify the degree of GCD at the electrode implantation sites, Nissl-stained coronal sections containing the electrode tracks were selected. Due to longitudinal shape of the posterior hippocampus in coronal slices including septal and temporal hippocampal portions, we measured the width of the granule cell layer only at the precise tip sites of the electrodes, at 50 and 100 μm next to them (Fig. 5H) and at corresponding contralateral positions.
Quantification of BrdU-positive cells was performed at 7 days after KA (1 day after BrdU) in DAB-stained sections with a Leica DMR microscope (Leica Microsystems) and Stereoinvestigator Software (MicroBrightField). Quantification of BrdU–/DCX–double-labeled cells was carried out at 14 days after KA in fluorescence-stained sections with the Axioplan 2 microscope. Numbers of BrdU-positive or BrdU-/DCX-positive cells were determined in the subgranular zone (SGZ) or in the hilus in every fifth section (50-μm sections, resulting distance 250 μm) along the whole longitudinal axis of the hippocampus. Due to their unequal distribution, all BrdU-positive cells within the region of interest (SGZ or hilus) were counted (Popken and Farel 1996). Cell numbers were normalized relative to the length of the SGZ (in mm) or area of the hilus (in mm2), respectively. Sections were put in four groups corresponding to the positions of the implanted electrodes.
For calculation of power spectral density (PSD), continuous artifact-free traces of 900-s duration were selected. Data were extracted for all recording positions and imported to Matlab (The Mathworks). The PSD estimate was computed using Welch's averaged periodogram spectral estimation method (Matlab Signal Processing Toolbox). The 900-s period was cut into 2-s segments with 50% overlap, followed by application of a Hamming window and a 215-point Fast-Fourier-Transformation. Periodograms were calculated for each segment, averaged, and scaled with the sampling frequency, resulting in the PSD. Broad-band PSD and PSD in selected frequency bands (broad theta: 3-8 Hz, gamma: 40–80 Hz, fast ripples: 250–500 Hz) were calculated for each channel and each mouse during SE and at 21 days after injection and averaged across KA-injected or control mice, respectively.
For comparison of mean EA for all positions and time points, we manually marked 10 EA episodes/mouse/time point (population spikes exceeding 3-fold the baseline, interspike intervals <1 s) and cut with a 25-s window starting 5 s before the onset. Data were squared, convolved with a 1-s rectangular window (amplitude = 1), and averaged. To monitor changes, smoothed EA was set relative to the mean between 4 and 1 s before EA onset at position 4, where activity in low frequency bands was least affected by the KA injection.
All data are presented as mean ± standard error of the mean (SEM). Statistical comparisons between groups were made with a one-way analysis of variance (ANOVA) followed by a Tukey Multiple Comparison test or with a two-way ANOVA with a Bonferroni posthoc test (GraphPad Software, version 4.02). Significance thresholds were set at *P < 0.05, **P < 0.01, and ***P < 0.001.
The Initial SE Extends along the Entire Hippocampus
To address the controversy between increased neurogenesis in systemic models for MTLE and the loss of neurogenesis in the ipsilateral septal hippocampus of KA-injected mice, we quantified the septotemporal extension of the initial SE instead of only regarding the injection site. To this end, the SE, induced by a focal KA injection, was recorded at four positions along the ipsilateral septotemporal axis and in some mice in the contralateral hippocampus (Fig. 1C).
Concomitant with the behavioral SE (see Materials and Methods), periodic epileptiform population spikes occurred in the ipsi- and contralateral hippocampus (Fig. 1A,B). SE always extended along the whole septotemporal axis of the ipsilateral and in the septal contralateral hippocampus (the temporal contralateral hippocampus was not recorded). Quantitative evaluation revealed that PSD in the theta (3–8 Hz) (Fig. 1D) and gamma band (40–80 Hz) (Fig. 1E) was significantly higher in the temporal and contralateral hippocampus than at the injection site.
On the cellular level, SE induced a strong expression of the neuronal activity marker c-Fos at all positions of both hippocampi and in the overlying cortex, except for CA1 and CA3 of the ipsilateral septal hippocampus (Fig. 1F–I). In contrast, saline-injected mice displayed only weak c-Fos labeling in the same areas (Fig. 1J–M) indicating that surgery per se stimulated c-Fos expression only slightly. In naive mice, only a few scattered cells were c-Fos positive (Fig. 1N,O).
Together, our intrahippocampal LFP recordings and c-Fos immunolabeling show that unilateral KA injection induces a SE in both hippocampi, which displays higher intensity in temporal versus septal areas of the ipsilateral side.
Cell Proliferation Is Strongly Increased in Both Hippocampi after SE
Given the extension of SE, we quantified cell proliferation in BrdU-labeled sections along the whole septotemporal axis of both hippocampi, instead of only at the injection site.
In saline-injected and naive mice, a few BrdU-positive cells were aligned in the SGZ at all septotemporal positions of the hippocampus (Fig. 2A–F). Cell numbers were comparable in both groups at all positions (Fig. 2M), indicating that the injection per se did not strongly stimulate cell proliferation in the SGZ.
In contrast, BrdU incorporation was massively enhanced in the SGZ of KA-injected mice at all positions along the septotemporal axis of both hippocampi (Fig. 2G,I–L), except for the ipsilateral septal hippocampus close to the injection site (Fig. 2H). In this region, only a few proliferating cells were counted in the SGZ and their number was comparable with control groups (Fig. 2M).
More importantly, we show that in the ipsilateral intermediate and temporal hippocampus, many closely aligned BrdU-positive cells were visible in the SGZ (Fig. 2J,L), and their numbers were significantly increased compared with controls (Fig. 2M). Furthermore, there was a significant increase in cell proliferation in the entire contralateral SGZ (Fig. 2G,I,K,M).
In the hilus of saline-injected and naive mice, only a few BrdU-positive cells were observed (Fig. 2A–F), and cell numbers were comparable in both groups (Fig. 2N). In contrast, in the ipsilateral septal hilus of KA-injected mice, there was a clustering of BrdU-positive cells (Fig. 2H), which was not observed at any other position (Fig. 2G–L). Quantification revealed that this increase was significant relative to the contralateral side and to controls (Fig. 2N).
In summary, we demonstrate that KA injection triggers a strong increase of proliferative activity in both the ipsi- and contralateral SGZ, except for the injection site where proliferation is strongly increased in the hilus.
Position-Dependent Upregulation of Neurogenesis in the SGZ
To determine the fate of proliferating cells in the SGZ, we injected BrdU at 6 days after KA followed by double immunolabeling for BrdU and DCX (immature granule cells; Brown et al. 2003) 7 or 14 days later.
In saline-injected mice, a single row of DCX-positive cells, with dendrites extending into the molecular layer, was observed at all septotemporal positions (Fig. 3A,B). Only a few DCX-positive cells were double-labeled with BrdU indicating a slow proliferative turnover under control conditions (Fig. 3K). In contrast, almost all BrdU-positive cells (80–90%) were also DCX-positive, showing that proliferative activity gave mainly rise to neurons (Fig. 3L).
In KA-injected mice, the pattern of DCX staining appeared dramatically changed: Within 3 weeks after KA injection, DCX labeling in the ipsilateral septal hippocampus gradually disappeared, in parallel with GCD formation (Fig. 3E). Accordingly, a significant reduction of DCX–/BrdU–double-labeled cells (Fig. 3K) and a small fraction of DCX-positive among BrdU-labeled cells (∼6%) (Fig. 3L) confirmed the reduction of neurogenesis at this position. Instead, proliferating cells close to the injection site became either microglia or astrocytes (data not shown). At more temporal positions, DCX-labeled cells reappeared in the SGZ, and the number of DCX–/BrdU–double-labeled cells was even increased compared with controls (Fig. 3F,K). Interestingly, this reappearance always occurred at the transition zone of GCD to normal dense granule cell layering (Fig. 3G). Moreover, the newly born neurons were arranged in several layers and extended basal dendrites into the hilus, a feature not seen in controls (Fig. 3H,I).
On the contralateral side, a significant increase of DCX-/BrdU-positive cells was observed in the SGZ at all septotemporal positions when compared with saline-injected controls (Fig. 3C,D,J,K). Furthermore, 80–90% of BrdU-positive nuclei colocalized with DCX-positive cells (Fig. 3J,L), reflecting significantly increased neurogenesis in the whole contralateral hippocampus.
In summary, neurogenesis is increased in both hippocampi except for the area surrounding the injection site, indicating that GCD and neurogenesis are mutually exclusive.
Morphological and Physiological Characteristics of Newly Generated Neurons
To test whether newly generated granule cells possess proepileptic intrinsic properties and whether they integrate into the hippocampal circuitry, we performed patch-clamp recordings of DCX-DsRed2–positive granule cells in the temporal hippocampus at 3–4 weeks after KA injection. Patched cells were biocytin-labeled and morphologically reconstructed. Their immature phenotype was confirmed by combined DCX-DsRed2 and PSA-NCAM detection (Fig. 4).
We found different morphological phenotypes ranging from “newborn” with short horizontal dendrites parallel to the hilar border of the granule cell layer (Fig. 4A) to mature-like dendritic trees reaching the hippocampal fissure (Fig. 4A,I,J). The majority of recorded DsRed2-positive cells had an intermediate immature morphology with less elaborated dendrites entering barely the outer molecular layer (Fig. 4AC).
First, we investigated the intrinsic properties of DCX-DsRed2–positive cells: All recorded cells had Rin values of >1 GΩ (mean: 2.127 ± 0.287 GΩ, n = 7) (Fig. 4D,K). All but one cell (with very immature morphology) showed overshooting action potentials, indicating that they were fully functional and excitable (Fig. 4F,L). The rheobase, that is, the current threshold for action potential induction within 250 ms, was very low (19.9 ± 6.5 pA, n = 7). In addition, subthreshold depolarization peaks, reminiscent of T-type calcium spikes in immature cells (Schmidt-Hieber et al. 2004), were observed in some cells (Fig. 4E,F); others showed more mature action potential firing (Fig. 4L). Altogether, recorded cells showed the typical characteristics of adult-generated hyperexcitable young granule cells (Schmidt-Hieber et al. 2004).
To test whether newly generated granule cells already receive glutamatergic input, spontaneous excitatory postsynaptic currents (sEPSCs) were recorded in DCX-DsRed2–positive cells while blocking GABAA and NMDA receptors. sEPSCs with kinetics of fast glutamatergic currents were observed in 5 of 7 cells (Fig. 4G). Extracellular stimulation in the outer molecular layer evoked excitatory postsynaptic currents (EPSCs), which showed paired pulse facilitation in some cases (Fig. 4M,N). Involvement of AMPA receptors in sEPSCs was confirmed by inhibition with CNQX (n = 3) (Fig. 4H). Only in two cells with very immature phenotype (Fig. 4A, arrowhead), neither spontaneous nor evoked EPSCs were detected.
In summary, newly generated granule cells in the KA-injected temporal hippocampus appear to integrate functionally into the synaptic network, and they produce action potential output. Furthermore, their intrinsic excitability is very high compared with mature cells.
Chronic EA Is Strongest in the Intermediate Hippocampus
To investigate whether the differential distribution of newly generated granule cells affects hippocampal activity in vivo, we analyzed the extension of EA along the septotemporal axis by intrahippocampal recordings at 3 (Fig. 5A1), 7 (Fig. 5A2), 14, and 21 days after KA (Fig. 5A3). EA occurred in all mice that had experienced SE. Single epileptiform population spikes or short periods of EA were visible already at 3 days after KA injection (Fig. 5A1) and increased in duration and amplitude up to 21 days after KA injection at all positions (Fig. 5A3). At all time points, the amplitudes of epileptiform population spikes were larger at distance to the injection site (position 2 or 3) than directly at the injection site or in the temporal hippocampus as indicated by smoothed averaged EA (Fig. 5B).
To take into account that raw signals were dominated by the highly synchronized large-amplitude EA population spikes, we calculated the PSD for all channels in control and epileptic mice (Fig. 5C–E). In controls, PSD was comparable at all recording sites, while in KA-injected mice there were substantial differences in PSD among the four positions (Fig. 5C): In the gamma band (40–80 Hz), the average PSD was significantly increased in the temporal compared with the septal hippocampus of KA-injected mice (Fig. 5D). In addition, in the intermediate hippocampus (position 3), PSD was significantly increased compared with controls (Fig. 5D). In the fast ripple range (250–500 Hz), which is characteristic for epileptogenic regions (Bragin et al. 2010), the PSD was significantly increased in the intermediate hippocampus (position 2 and 3) of KA-injected mice compared with corresponding sites in controls (Fig. 5E). Taken together, EA was maximal in the intermediate hippocampus (Fig. 5BE), where cell proliferation was increased (Fig. 5F) and neurogenesis reappeared (Fig. 5G).
Finally, we measured granule cell layer width at all implantation sites and at corresponding contralateral locations at 3 weeks after KA injection to investigate a potential relation between GCD and EA occurrence (Fig. 5H). Close to the injection site, granule cell layer width was significantly increased compared with the contralateral hippocampus and gradually declined toward intermediate and temporal positions (Fig. 5I).
In summary, we show that in the septal hippocampus, maximal GCD coincides with low EA. Instead EA is maximal at intermediate sites, where neurogenesis recovers.
In the present study, we provide a detailed spatial characterization of acute SE and chronic EA in relation to cell proliferation and neurogenesis along the septotemporal axis of the hippocampus in the intrahippocampal KA mouse model of MTLE. We show that following a focal KA injection into the septal hippocampus, SE and EA extend along the septotemporal axis of both hippocampi. Moreover, SE and EA display even stronger intensity in the intermediate and temporal hippocampus than at the injection site, where GCD is most prominent. Likewise, proliferative activity in the SGZ is strongly increased in the ipsilateral temporal and the entire contralateral hippocampus. The proliferating cells in the SGZ give rise to hyperexcitable and functionally integrated young granule cells. We conclude that 1) SE stimulates neurogenesis in a position-dependent manner; 2) the epilepsy-related increase in young granule cells is proepileptogenic; and 3) a transition zone, located temporal to the strongly sclerotic focus, might be crucial for EA generation.
Strength of SE along the Septotemporal Axis Mirrors the Neurogenesis Pattern
Our in vivo analysis of acute SE at ∼3–6 h after KA injection revealed that SE is stronger in the contralateral and ipsilateral temporal hippocampus than at the injection site. This pattern may have two reasons: 1) Absence of c-Fos expression in CA3 and CA1 of the ipsilateral septal hippocampus during SE suggests that principal cells in these regions are inactive or already dying due to the acute neurotoxic effects of KA (in accordance with Le Duigou et al. 2005); consequently, a diminished number of active cells may cause reduced amplitudes of the highly synchronized population spikes during SE. 2) An immediate strong activation of neurons at the injection site could result in a persistent depolarization block, as described for acute hippocampal slices prepared from mice within hours after the KA injection (Le Duigou et al. 2005), and thus in reduced activity in our recordings. An early strong activation is, however, difficult to prove in vivo due to masking effects of anesthesia on SE, yet preliminary results obtained in mice recorded during and immediately after KA injection support this hypothesis (U Häussler, unpublished observations).
In line with the septotemporal SE pattern, we find a position-dependent stimulation of cell proliferation and neurogenesis in the SGZ, both being substantially increased at all ipsi- and contralateral sites, except for the injection site. This striking spatial overlap indicates a stimulating effect of SE on neurogenesis, which is in accordance with a number of studies showing that short strong neuronal activation evokes increased neurogenesis (Bengzon et al. 1997; Madsen et al. 2000; Nakagawa et al. 2000). In fact, a dependency of neurogenesis on the strength of SE has been shown: After mild seizures, a high percentage of precursor cells survive and develop into neurons (Mohapel et al. 2004; Yang et al. 2008), while strong seizures induce a glial fate (Yang et al. 2008). It is therefore conceivable that intrahippocampal KA injection causes an immediate excess activation at the injection site followed by a depolarization block (Le Duigou et al. 2005) and gliogenesis (Heinrich et al. 2006) and a milder but more sustained activation in the ipsilateral temporal and entire contralateral hippocampus, stimulating neurogenesis.
Does Increased Neurogenesis Contribute to Persistent Hippocampal Hyperexcitability?
Our finding that neurogenesis is lost at the sclerotic focus (in accordance with Kralic et al. 2005; Heinrich et al. 2006; Ledergerber et al. 2006), but is increased everywhere else in the hippocampus, clarifies the discrepancy to other animal models, where an epilepsy-related increase in neurogenesis has been shown (Parent et al. 1997; Gray and Sundstrom 1998; Jessberger et al. 2005). Furthermore, we provide evidence that GCD, induced by the loss of reelin (Heinrich et al. 2006; Müller et al. 2009), and the activity-triggered increase in neurogenesis are mutually exclusive but occur in close apposition within one hippocampus. This is of particular relevance for the interpretation of contradicting results in human MTLE patients, where either increased (Crespel et al. 2005; Thom et al. 2005), unchanged (Fahrner et al. 2007), or reduced proliferation and neurogenesis have been reported (Mathern et al. 2002; Pirttila et al. 2005): The investigated human tissue was most likely taken from different hippocampal levels with varying degrees of sclerosis and GCD.
Moreover, since we found increased neurogenesis outside the sclerotic region, it was possible to address the question whether increased neurogenesis contributes to hyperexcitability in this focal model, comparable with what has been proposed in the systemic epilepsy models (Scharfman et al. 2000; Jung et al. 2004; Kwak et al. 2008). Our patch-clamp recordings revealed a high Rin and a very low rheobase in newly generated compared with mature granule cells (Young et al. 2009), indicating that the newborn neurons in our model are indeed hyperexcitable (Schmidt-Hieber et al. 2004; Couillard-Despres et al. 2006; Ge et al. 2008). Furthermore, we show that the newly generated granule cells are functionally integrated into the synaptic network and receive glutamatergic input. Whether this network integration happens in an aberrant and/or accelerated fashion, as reported for other epilepsy models (Scharfman et al. 2000; Esposito et al. 2005; Overstreet-Wadiche et al. 2006; Shapiro et al. 2007), remains to be investigated.
On the network level, we have to be cautious when estimating the contribution of these newly generated granule cells to EA since the spatial resolution of LFP recordings limits interpretations on a cellular level. Intriguingly, EA was strongest in the intermediate hippocampus, that is, the transition zone between GCD and increased neurogenesis indicating that the patterns of neurogenesis and EA overlap only partially. This suggests that in the chronic phase of epilepsy, the addition of young granule cells results in increased network excitability in a position-dependent manner but is not per se proconvulsive. This is supported by the observation that after long survival times (8 months after KA), the number of newborn neurons in KA-injected animals was back to normal levels, but EA persisted (U Häussler, unpublished observations), consistent with Hattiangady et al. (2004). Instead, it is likely that the SE-induced neurogenesis plays a proepileptogenic role particularly during early stages of the disease through the sheer number of hyperexcitable newborn granule cells that integrate into the dentate gyrus network. In addition, their basal dendrites might contribute to the formation of a pathological circuitry, by offering targets for sprouted mossy fibers or contributing themselves to sprouting, as described in other models (Spigelman et al. 1998; Jessberger et al. 2007; Shapiro et al. 2008; Kron et al. 2010), all together promoting recurrent connectivity and feed-forward excitation (Morgan and Soltesz 2008). Differential integration mechanisms of newly generated granule cells (Murphy et al. 2011), with respect to their septotemporal position, might lead to a functionally heterogeneous granule cell network reflected by the lack of a precise spatial overlap of increased neurogenesis and strong EA.
A further important result of our study is that EA is low at the injection site, and, thus, the hippocampal region with strongest sclerosis and prominent GCD does most likely not contribute to hyperexcitability. Instead, we suggest that the network in the transition zone closely apposed to the sclerotic focus where neurogenesis recovers seems to be strongly proconvulsive. Analyzing the feed-forward connectivity and integration mechanisms in this transition zone is therefore the next step toward understanding the processes of EA generation.
Deutsche Forschungsgemeinschaft (SFB TR3 and SFB 780); German Ministry of Education and Research (FKZ 01GQ0420, 01GQ0830).
We thank J. Bischofberger for providing DCX-DsRed2 mice and for helpful advice. We are grateful to S. Huber and F. Moos for excellent technical assistance. Conflict of Interest: None declared.