|Home | About | Journals | Submit | Contact Us | Français|
The goal of this study was to examine morpho-physiological changes in the dorsal subiculum network in the mouse model of temporal lobe epilepsy using extracellular recording, juxtacellular and immunofluorescence double labeling, and anterograde tracing methods. A significant loss of total dorsal subicular neurons, particularly calbindin, parvalbumin (PV), and immunopositive interneurons, was found at 2 months after pilocarpine-induced status epilepticus (SE). However, the sprouting of axons from lateral entorhinal cortex (LEnt) was observed to contact with surviving subicular neurons. These neurons had two predominant discharge patterns: bursting and fast irregular discharges. The bursting neurons were mainly pyramidal cells, and their dendritic spine density and bursting discharge rates were increased significantly in SE mice compared to the control group. Fast irregular discharge neurons were PV-immunopositive interneurons, and had less dendritic spines in SE mice when compared to control mice. When LEnt was stimulated, bursting and fast irregular discharge neurons had much shorter latency and stronger excitatory response in SE mice compared to the control group. Our results illustrate that morpho-physiological changes in the dorsal subiculum could be part of a multilevel pathological network that occurs simultaneously in many brain areas to contribute to the generation of epileptiform activity.
Neuronal loss and synaptic reorganization of intra- and extra-hippocampal connections have been well-known to be associated with temporal lobe epilepsy (TLE) (3, 4, 6, 7, 9, 14, 17, 18, 27, 29, 33, 55, 56). As a major output of the hippocampus and the first brain region encountered by neural activity emanating from the hippocampus (1, 41, 42), subiculum is also found to play an important role in the initiation and maintenance of epileptic discharges in TLE (5, 16). Subiculum propagates abnormal neuronal firing to cortex, including entorhinal and perirhinal cortex, and subcortical regions, such as amygdala and thalamus (42, 49, 60,). In patients with TLE, Cohen et al (2002) described a spontaneous, rhythmic activity initiated in the subiculum of slices, and revealed that depolarizing GABAergic responses in neurons downstream to the sclerotic CA1 region contribute to human interictal activity (10). In the rat pilocarpine model, Knopp et al (2005) observed the loss of subicular neurons and reduced arborization and spine density in the proximal part of the apical dendrites of pyramidal cells, and suggested a partial deafferentation from CA1 (26). However, most previous electrophysiological studies were done in the slices of subiculum, and the lack of in vivo data has hindered the extrapolation of the results from in vitro approaches. In the present study in the mouse pilocarpine model of temporal lobe epilepsy, we aimed to 1) investigate cytoarchitectonic and dendroarchitectonic changes of neurons in the dorsal subiculum (Dsub) in general, and to define the subtypes of lost neurons and changes in dendritic spines in particular; 2) reveal the discharge patterns of different subtypes of surviving subicular neurons and their relation to interictal activity; and 3) evaluate the role of the entorhinal-subiculum pathway in the generation of interictal epileptiform-like activity by combining anterograde tracing and neurophysiological recording.
Male and female mice (B6: 129-Grm5m I Rod/J) weighing 20–32g were used for this study. Status epilepticus (SE) was induced according to our established procedures (28, 29, 53–56). Briefly, mice were given a single subcutaneous injection of methyl-scopolamine nitrate (1mg/kg) 30 min before the injection of either saline in the control or pilocarpine in the experimental groups. In the latter groups, mice were given a single i.p. injection of pilocarpine (300 mg/kg) and experienced status epilepticus for about 4 hrs. Pilocarpine-induced behavioral changes, including hypoactivity, tremor, head bobbing, and myoclonic movements of the limbs, progressing to recurrent myoclonic convulsions with rearing, falling, and status epilepticus, were similar [similar to what in previous reports?] to our previous reports (28, 53–56).
All experiments were approved by the Tan Tock Seng Hospital – National Neuroscience Institutional Animal Care & Use Committee. The mice were kept in a specific pathogen-free room and maintained with free access to food and water on a 12 hr light-dark cycle. Efforts were made throughout the study to minimize animal suffering and to use the minimum number of animals.
For NeuN, parvalbumin (PV), calbindin (CB), and calretinin (CR) immunocytochemical studies, 7 experimental mice at 2 months after SE and 6 age-matched control mice were used. They were deeply anesthetized with chloral hydrate (0.40 g/kg), perfused transcardially with 10 ml of saline initially, and followed by 100 ml of 4% paraformaldehyde in 0.1 M phosphate buffer (PB) (pH 7.4) for 20 min. After perfusion, the brains were removed and kept overnight in 30% sucrose in 0.1 M PB. Coronal sections at 40 µm thicknesses were cut in a cryostat (HM505E, Microm, Zeiss, Oberkochen, Germany). A set of five serial sections was prepared from each brain and placed in different wells of a 24-well tissue culture dish for the control, NeuN, CB, CR and PV immunocytochemical reactions. For immunocytochemical studies, freely-floating sections were treated in 4% normal goat serum for 2 hrs at room temperature, washed in 0.1 M phosphate-buffered saline (PBS) containing 0.1% Triton-X 100, and placed overnight in primary rabbit antibodies for NeuN (1:1000) (Chemicon International, Inc., CA, USA), PV (1:3000), CB (1:3000) and CR (1:3000) (Swant Inc, Switzerland). After incubation, sections were washed in PBS and placed in biotinylated goat anti-rabbit IgG (Vector Laboratories, Burlingame, CA, USA) diluted 1:2000 in PBS/Triton X-100 for 1 hr. They were placed in avidin–biotin complex (ABC) reagent (Vector Laboratories) in PBS/Triton X-100 for 1 hr, washed in PBS, and reacted in a solution of 0.12% H2O2 and 0.05% 3,3-diaminobenzidine (DAB) (Sigma-Aldrich, Missouri, USA) in Tris buffer (TB) for 15 min. Sections were then mounted, dehydrated, covered, and photographed by using an image analysis system.
Mice (30 SE and 22 controls) were anesthetized with 20% urethane (0.75g/kg, i.p.) and fixed in a stereotactic apparatus. A concentric stainless steel stimulating electrode (World Precision Instruments Inc, USA) with tip diameter of 0.1mm was implanted into the lateral entorhinal cortex (LEnt) with coordinates (mean value ± standard deviation) posterior to the bregma: 2.7 ± 0.1 mm; lateral to the midline: 4.3 ± 0.05mm; and ventral to the dura: 2.8 ± 0.10mm (44). Extracellular recordings of neurons from the dorsal subiculum were performed using single-barrel glass micropipettes (external diameter of the tip: 2–3µm, impedance from 5 to 15MΩ.) filled with biotinylated dextran amine (BDA, Molecular Probes Inc, USA) solution (5%) for juxtacellular labeling of neurons. The stereotaxic coordinates (mean value ± standard deviation) for all recording sites were: posterior to bregma: 3.4 ± 0.2 mm; lateral to midline: 2.0 ± 0.15mm; and ventral from the dura1.35 ± 0.20mm (44). Signals were amplified (P16; Grass Instruments; Astro-Med, Inc, West Warwick, U.S.A.) and fed to storage oscilloscopes. Single-cell signals were filtered at a range of 30Hz-3k Hz. Local field activity was filtered at a range of 0.1Hz-1k Hz. Single-cell and field activities were stored simultaneously on a computer via a Cambridge Electronic Design (Cambridge, UK) interface using Spike 2 software. The spontaneous activity of a single neuron was recorded for 10–20 min after stable firing. The lateral entorhinal cortex was electrically stimulated with single pulses at 0.2mA, 0.2ms, 0.1–0.2Hz and delivered with a S88K stimulator (Grass Instrument, Astro-Med Inc, West Warwick, USA). To confirm the stimulation site, positive electric current was administered (1~2mA) for 30~60 sec through the stimulating electrode after the experiment to produce a local lesion of the LEnt.
After the electrophysiological recording, the neurons were labeled by juxtacellular injection of biotinylated dextran amine (BDA) with a modification of the protocol reported previously (24, 45). Briefly, positive current at 10–30nA, 500ms was applied at a frequency of 1 Hz through an iontophoresis pump (Kation Scientific, Minneeapolis, MN, USA) for 20 min. One day after BDA delivery, mice were deeply anaesthetized, and perfused with 10ml of saline, followed by 100ml of 4% paraformaldehyde and 0.2% picric acid in 0.1 M PB (pH 7.4) for 20 min. Coronal sections of the brain were cut at 40 µm thickness, serial sections were transferred to different wells of a 24-well tissue culture dish, and sections were placed overnight in avidin-horseradish peroxidase (HRP) conjugate (1:1000) and reacted in a solution of 0.12% H2O2 and 0.05 DAB in TB for 20 min for single-labeling of recorded cells according to the previous reports (40, 58).
To correlate electrophysiological properties to different subtypes of interneurons, immunofluorescence double-labeling for BDA with PV, CB, or CR was done. To efficiently label recorded neurons, we developed a new method to double-amplify BDA immunopositive product using avidin-horseradish peroxidase conjugate and primary mouse antibody for HRP. Sections were incubated overnight in avidin-horseradish peroxidase conjugate (1:1000), washed in 0.1% TBS-TX (TBS containing 0.05% Triton X-100), and blocked in 4% normal goat serum for 2 hrs, and then placed in a primary mouse antibody for HRP (1:300) (ABcam, Cambridge, UK) and rabbit antibodies for PV (1:150), CB (1:200) or CR (1:150) for 24 h. After washing in 0.1% TBS-TX, sections were incubated in Cy3-conjugated goat anti-mouse IgG (1:100) for 4 hrs and followed in FITC-conjugated goat anti-rabbit IgG (1:100) for 4 hrs. The sections were mounted, dried, and covered with FluorSave TM Reagent (Calbiochem-Novabiochem, CA, USA) to retard fading, and then examined with an Olympus Fluoview FV500 Confocal Laser Scanning Biological Microscope.
Twenty mice (8 for control and 12 at 2 months after pilocarpine injection) were used. Mice were anesthetized with chloral hydrate (40mg/kg) and fixed in a Stoelting stereotaxic apparatus. A small hole was drilled into the skull above the intended injection sites, and a glass micropipette (with a diameter of 20–30 µm) containing a 2.5% solution of PHA-L (Vector Laboratories, Burlingame, CA, USA) in 0.1M phosphate buffered saline (pH 7.4) was lowered into the superficial layer (mainly in layers II and III) of the lateral entorhinal cortex at 2.7 mm posterior to the bregma, 4.3 mm lateral to the midline, and 2.8 mm ventral to the dura (44). PHA-L was delivered iontophoretically with positive current (5µA: 7 seconds on, 7 seconds off) for 10 minutes. Seven days after PHA-L delivery, animals were deeply anaesthetized, and perfused transcardially with 50 ml of saline, followed by 100 ml of 4% paraformaldehyde and 0.15% picric acid in 0.1M PB (pH 7.4) for 30 minutes. Coronal sections at 40 µm thickness were cut in a cryostat, and serial sections were transferred to different wells of a 24-well tissue culture dish for control, PHA-L single staining, and PHA-L with PV, CB, or CR double-labeling.
For the PHA-L immunocytochemical study, freely-floating sections were placed overnight in primary goat antibody for PHA-L (1: 5,000) (Vector Laboratories, Burlingame, CA, USA), and for 2 hours in biotinylated horse anti-goat IgG for PHA-L diluted 1: 500 in TBS/Triton X-100. Sections were then incubated in an avidin-biotin complex (ABC) reagent in TBS/Triton X-100 for 2 hours, reacted in a solution of 0.12% H2O2 and 0.05% 3,3’-diaminobenzidine (DAB) (Sigma-Aldrich, Missouri, USA) in TB for 20 minutes, and mounted. Alternative PHA-L-stained sections were counterstained with cresyl fast violet (CFV), coverslipped, and photographed using image analysis system.
For double-labeling of PHA-L with CB, CR, and PV, sections were incubated in primary goat antibody for PHA-L (1:1000) and rabbit antibodies for CB (1:500), CR (1:500) and PV (1:500) (Swant, Switzerland) for 48 hours, washed in TBS/Triton X-100, and placed in biotinylated horse anti-goat IgG (1:200) and swine anti-rabbit IgG (1:100) for 4 hours. They were then incubated in ABC solution for 2 hours, and reacted in DAB-Nickel solution for 20 minutes. Sections were then incubated in rabbit peroxidase anti-peroxidase (PAP) (1:100) solution overnight, and reacted with DAB alone.
To further confirm possible synaptic contacts between PHA-L immunopositive en passant or terminal boutons and CB, CR, or PV immunopositive neurons observed from DAB immunostaining, immunofluorescence double-labeling was performed. Sections were incubated in a primary goat antibody for PHA-L (1:1000) and rabbit antibodies for CB (1:500), CR (1:500), and PV (1:500) (Swant, Switzerland) for 48 hours, washed in TBS/Triton X-100, and placed for 4 hrs in FITC-conjugated goat anti-rabbit IgG against CB, CR, PV and Cy3-conjugated donkey anti-goat IgG for PHA-L. The sections were then mounted, dried, and coverslipped by using FluorSaveTM Reagent (Calbiochem-Novabiochem, CA) to retard fading. The tissue preparations were digitized and reconstructed three-dimensionally using a Confocal Laser Scanning Biological Microscope.
Electrophysiological data were analyzed off-line using Spike2 and DataPac software. The original data were separated on three-frequency bands by digital filtering (FIR 555 Rolloff): 1) 2 Hz low pass band for slow-wave activity; 2) 20–60 Hz band pass for beta-gamma activity; and 3) 500 Hz high pass for unit activity. The sharp increase in the amplitude of beta-gamma activity was considered a beginning of the Up phase. The flattening of beta-gamma activity was considered the Down phase.
Discriminated unit discharges were acquired and further analyzed for interspike interval and firing rate. In previous studies (2, 46), burst analysis was performed using the bursts v1.25.s2s script of Spike2. Burst events were identified by the following parameters: the maximum interval between the first two spikes (30 ms), and the minimum number of spikes within a burst (n=2). Burst rate, percentage of burst discharges /total discharges, and the duration of each burst were acquired. To analyze field activity, Fast Fourier transformation (FFT) of waveform data into power spectrums was performed. FFT size was set to 128 Hz, as determined by Nyquist criteria, with hamming windowing of root mean square data. Evoked responses of neurons (latent period and duration) were analyzed on the basis of peristimulus time histograms (PSTH). The latent period was defined as the time period from stimulation to the beginning of evoked discharges. The duration was indicated as the time period for evoked discharges.
For cell counting of NeuN, CB, CR, and PV immunopositive neuronal profiles, three sections from the dorsal subiculum of each animal were selected. Neuronal number in the dorsal subiclum was counted manually and the respective area for those counted neurons was measured using the KS 100 Imaging System (Carl Zeiss Vision, Germany). The cell number was then indicated as the number per square millimeter (no/mm2). Counting of neuronal profiles was done by an investigator blinded to the experimental conditions to which the mice were subjected. The density of cells was indicated as the mean value ± standard deviation (SD). In the present study, in the same manner as that employed in a comparative study between the control and SE mice, we measured the relative number of NeuN, CB, CR and PV immunopositive neurons, but not an absolute number of neurons as measured by stereological techniques. No calibration was made because the chances of neuronal counting errors might occur equally in both groups of the control and SE mice. Statistical significance for two groups of control and experimental mice was determined by a Student’s t-Test, whereas for those with more than two [more than two what?], One-way ANOVA followed by post-hoc test was used. A P-value of less than 0.05 was considered statistically significant.
Juxtacellularly-labeled neurons with BDA were examined under light microscope with high magnification of ×1000. The number of dendritic spines in the 2nd and 3rd order dendrites was counted and indicated as the number per millimeter (No/mm) and compared between normal and SE groups. In the present study, since a comparison study was made between the control and SE mice, no absolute number was needed. Therefore the calibration of the data was not done.
In the present study, behavioral changes, including hypoactivity, tremor, head bobbing, clonic movements of the limbs progressing to recurrent clonic convulsions with rearing, and falling, were observed during pilocarpine-induced status epilepticus (SE), which were similar to that described in our previous reports (28, 29, 55, 56, 61). According to our long-term EEG telemetry and video monitoring study, almost all pilocarpine-induced status epilepticus mice developed epilepsy (Zhang et al., 2008)
Quantitative analysis showed that the total area of the dorsal subiculum (Dsub) in SE (0.166±0.049 mm2) and the control (0.174±0.059 mm2) mice did not change significantly (P>0.05). However, a significant loss of NeuN immunopositive neurons (14.7% ±3.6%) occurred in the Dsub of SE mice (Fig.1A1-A3). The density of NeuN immunopositive neurons was reduced to 2614±279/mm2 compared to the control value of 3068±206/mm2 (Student’s t-test, P<0.05). The population of PV and CB immunopositive neurons decreased 34.4% ± 5.1% (Student’s t-test, P<0.05) (Fig.1B1-B3) and 46.5% ± 6.3% respectively (Student’s t-test,P<0.05) (Fig.1C1-C3), whereas CR immunopositive neurons decreased 19.4% ± 4.0% (Student’s t-test, P<0.05) (Fig.1D1-D3). The percentage change of PV or CB immunopositive neurons was significantly greater than that of NeuN positive neurons (ANOVA post-hoc tests, P<0.05). However, there was no significant percentage change for CR immunopositive neurons compared to NeuN immunopositive neurons (ANOVA post-hoc tests, P>0.05).
Electrophysiological experiments were performed on 22 control and 30 SE mice under urethane anesthesia. Local field potentials in the control group displayed an Up-Down pattern comparable to that described by Isomura and co-authors (18) (Fig. 2A). In the SE group, a similar Up-Down pattern was observed. It was characterized as 0.3–1.0 Hz waves with an increase in the amplitude of gamma activity (30–60 Hz) during the Up-phase that coincided with local negative waves (Fig. 3A, B). In contrast to the control group, sharp field spikes were superimposed on these slow waves (Fig. 2 B, Fig. 3B, downward arrows). Fast Fourier transform analysis showed that spectral power electrical activity in the frequency band of less than 1 Hz was increased in the SE group (Fig. 2 C), which was reflected by higher amplitude of Up-Down state waves (Fig. 2 A, B). High-frequency oscillations were not a subject of these experiments, because these studies require simultaneous recordings of local field potentials and unit activity by different microelectrodes.
Spontaneous discharges of 101 and 79 neurons were recorded in 30 SE and 22 control mice. According to the pattern of spontaneous firing (13, 36), these neurons were divided into 3 groups: bursting neurons, irregular discharging neurons, and regular discharging neurons. A neuron was defined as bursting when it discharged with 1) a minimum of two spikes within a burst; 2) an interval between the first two spikes of less than 30 ms; 3) an interspike interval histogram with a bimodal shape; and 4) a percentage of bursting spikes among total spikes of more than 10%. A neuron was defined as irregular discharging when its interspike histogram had a single bell-shape form and there was no sign of regular activity on autocorrelogram. A neuron was defined as regular discharging when its interspike histogram had a single bell-shape form and there was rhythmicity on its autocorrelogram. The percentages of classified neurons in SE and control mice are shown in Table 1. The percentage of bursting neurons was 56.9% in the control group, but 65.3% in the SE group. The percentage of irregular discharging neurons was respectively 40.5% in the control group and 29.7% in SE group. The percentage of regular discharging neurons in the total number of recording Dsub neurons did not change noticeably.
Spontaneous activity of 66 bursting neurons in the SE group and 45 in the control group were taken for analysis. Firing rate of bursting neurons decreased significantly to 3.42±0.68Hz in SE mice from 4.97±0.79Hz in the control (t-test, P<0.05, Table 1). The bursts were more frequent in SE mice compared to those in the control mice (Fig. 4A1, A2, B). Bursting discharges were scattered irregularly in control mice (Fig. 4A1), but aggregated in SE mice (Fig. 4A2). Statistical analysis showed that both mean burst rate and percentage of discharges in bursts were increased significantly in SE mice (Fig. 4B, C). There was no significant difference in mean-burst length between SE and control mice (Fig. 4D). The spike number per burst was 2.68±0.56 and 2.86±0.45 in the control and SE mice respectively (t-test, P>0.05).
All bursting neurons increased their discharge frequency during the Up-phase. An example of a bursting neuron in SE mice is shown in Fig. 3. This neuron discharged predominantly during the Up-phase with increased amplitude of gamma activity, generating 2–4 bursts (Fig 3 A). Burst script analysis showed that 90.1% of spikes in this neuron appeared within bursting state. The mean burst duration was 26.6 ms. The mean number of spikes per burst was 3.2. Within a burst discharge, the amplitude of sequential action potentials reduced gradually (insets in Fig. 3 A). Autocorrelograms showed that this neuron displayed rhythmic activity with multiple frequencies (Fig. 3 B). The discharges of this neuron at regular intervals of 1/3 seconds were modulated by the periodicity of occurrence of Up and Down phases. During the Up-phase, this neuron generated burst action potentials with durations of 200–300 ms (Fig. 3 A and b1). Within each burst, action potentials occurred regularly at intervals of 5–15 ms (Fig. 3B, b2). This neuron generated a burst of spikes during gamma oscillations; bursting spikes occurred during both ascending and descending phases. Juxtacellular labeling showed that the soma of this neuron had a pyramidal shape with a typical apical dendrite (Fig. 3 C).
In the present study, the average firing rate of irregular discharging neurons in Dsub of SE mice did not change significantly when compared to that of the control mice (Table 1), but it was much faster than the firing rate of bursting neurons from the same region. This result was consistent with that obtained from previous in vitro studies (20, 34, 48). The firing rate of irregular discharge neurons under urethane anesthesia observed in our experiments was similar to that of the freely moving state (2), but was lower than that in slice preparation (19, 37). An example of an irregular discharging neuron in SE mice was shown in Fig 5. The mean firing rate of this neuron was 7.12Hz. The local field activity (low pass) showed that the Up and Down state of interictal-like activity was correlated with the discharges of this neuron (Fig. 5A). Discharges increased during the Up-phase of interictal-like spikes, and decreased during the Down state (Fig. 5B). BDA labeling showed that this irregular discharging neuron had multiple dendrites oriented in different directions (Fig. 5C).
The number of regular discharging neurons was less than 5% of all neurons recorded (Table. 1). The periodicity of occurrence of Up and Down phases (Fig. 6 A1) showed the network activity. The action potentials demonstrated a consistently similar waveform (Fig. 6 A2-A3). The interval between spikes was close to a normal distribution (Fig. 6B). These neurons had soma with a pyramidal shape with a single apical dendrite (Fig. 6C).
In order to identify the immunochemical characteristics of Dsub neurons with different patterns of spontaneous activity, fluorescent double-labeling was carried out in 25 control mice after electrophysiological recording. Sixteen Dsub neurons were successfully labeled by juxtacellular injection of BDA. Bursting activity was observed in 8 Dsub neurons with pyramidal shape and these neurons were never double-labeled with PV (Fig. 7A1-3), CB (Fig. 7B1-3), or CR (Fig. 7C1-3). However, in 8 BDA-labeled neurons with irregular spontaneous activity, 5 were co-localized with PV (Fig. 7D1-3), and none of them was double-labeled with CB (Fig. 7E1-3), or CR (Fig. 7F1-3). The other 3 were not labeled by any of these 3 neuropeptides.
The responses of bursting neurons to a single pulse of LEnt stimulation in the control group (n=17) consisted of single discharges, and none of them showed multiple spikes (Fig. 8A1). However, in the SE group, 16 out of 25 bursting neurons responded with 2–3 spikes (Fig. 8A2), and 9 neurons responded with single spikes. The latency of evoked discharges (Fig. 9A1) was significantly decreased from 9.6±1.0ms in the control group to 7.1±1.2ms in the SE group (Student’s t-test, P<0.05). After evoked spikes, a period of suppression of unit discharges was observed in 14 neurons from control (82.4%), and 17 neurons from SE (68.0%) mice. The duration of suppression was significantly less in SE mice compared to the control group (239.7±78.6ms vs. 331.4±75.7ms, Student’s t-test, p<0.05, see Fig. 8A2, Fig. 9A2).
Out of 12 irregular discharging neurons in the control group, 2 (16.6%) displayed a single spike followed by suppression, and 10 responded to LEnt electrical stimulation with only an initial suppression of spontaneous discharges (Fig. 8B1). In the SE group, in 9 out of 13 irregularly discharging cells (69.2%), the LEnt stimulation evoked a single spike followed by suppression of unit discharges (Fig. 8B2). The remaining 4 cells demonstrated only suppression of spontaneous discharges after LEnt stimulation. The duration of the excitatory response was increased significantly in SE mice compared to the control group (4.0±0.7 ms vs 1.5±0.5 ms, Student’s t-test, p<0.05). However no significant difference was observed for the latencies between the two groups (Fig. 9B1). There was also no significant changes in the period of suppression in the irregular discharging neurons in SE mice compared to that of the control group (264.0± 76.9ms vs 339.3±98.1ms, p=0.059) (Fig. 9B2).
Two out of 8 control and 4 out of 16 SE mice with PHA-L injection were excluded from data analysis due to incorrect injection sites. In the remaining animals, PHA-L was iontophoretically injected into layers II-III of LEnt. Quantitative study showed that the areas of PHA-L injection site were 0.144 ± 0.009 mm2 in the control, 0.135 ± 0.011 mm2 and 0.133 ± 0.01 mm2 in the experiments mice at 2 months after PISE. The rostra-caudal extents of the PHA-L injection site were 684 ± 25.6 µm in the control, 670 ± 27.6 µm and 640 ± 30.3 µm in the experimental mice at 2 months after PISE. No significant difference was shown among the two groups of mice in the rostra-caudal extents and the areas of PHA-L injection site by Student’s t-test (P > 0.05). In the dorsal subiculum of the control and SE groups, PHLA-L immunopositive en passant and terminal boutons were observed mainly in the ipsilateral stratum pyramidal, and their density was 1110± 115 mm2 in the control (Fig. 10A1, A3) and 1740±170/mm2 in SE (Fig. 10 A2, A3) mice. It was increased significantly in SE mice (Student’s t-test, P < 0.05). Double immunostaining showed PV and CB, and CR-immunopositive neurons contacted by PHA-L-immunopositive en passant and terminal boutons in the control (Fig. 10, B1, C1, and D1) and SE (Fig. 10, B2, 3; C2,3; D2,3) mice. The size of PHLA-L immunopositive en passant and terminal boutons in the SE group was much larger than that in the control group (3.3±0.32µm2 in SE and 1. 41±0.07 µm2 in the control mice) (Student’s t-test, P< 0.05). Some terminal boutons aggregated to form bundles of grape-like structures (Fig. 10D3). Further three-dimensional reconstruction of immunofluorescence-labeled PHA-L immunopositive en passant and terminal boutons with PV, CB, and CR immunopositive neurons confirmed their close contacts in Dsub. An example of 3D reconstruction of a contact between a PHLA-L immunopositive en passant bouton and a CR immunopositive interneuron from SE mice was shown in Fig. 10E, E1-3. Quantitative assessment showed (Fig. 10 B4) that the number of PHLA-L immunopositive en passant or terminal boutons contacting PV immunopositive neurons increased significantly (152±10.1/mm2 in SE and 114± 10.1 mm2 in the control mice) (Student’s t-test, P < 0.05), but no obvious change was observed for those contacting CB or CR immunopositive neurons (Student’s t-test, P > 0.05) (Fig. 10 C4, D4).
Of 16 juxtacellularly-labeled neurons in SE and 13 in control mice after electrophysiological recording, 9 neurons from SE and 8 from control mice had bursting discharges and typical morphological characteristics of pyramidal neurons (Fig. 11A1, B1). Quantitative analysis showed that the density of dendritic spines in the 2nd and 3rd order dendrites was significantly higher in SE (Fig. 11A2, C) than in control (Fig. 11 B2, C) mice (Student’s t-test, P<0.05). However, it did not change significantly in the 4th order dendrites (Fig. 11 C).
Seven BDA-labeled neurons in SE and 5 in control mice were interneurons (Fig. 12A1, B1). They discharged irregularly and had granular or round soma, and were relatively smaller than pyramidal neurons. Quantitative analysis demonstrated that the density of dendritic spines in the 3rd and 4th order dendrites in SE mice (Fig. 12A2, C) was significantly reduced when compared to that in control mice (Fig. 12B2, C), (Student’s t-test, P<0.05), but no obvious change was observed in the 2nd order dendrites (Fig. 12C).
The present study provides the first comparative description of morpho-physiological characteristics of neurons in the dorsal subiculum in control and SE mice in vivo. We showed that: 1) significant neuronal loss (in particular for CB and PV immunopositive neurons) occurred in SE mice; 2) PV-immunopositive interneurons demonstrated fast irregular discharges; and 3) significant increase of en passant and terminal boutons in PHA-L immunopositive axons from the LEnt occurred in SE mice, especially for those contacting PV immunopositive interneurons, which was associated with increased dendritic spines on pyramidal neurons.
These morphological changes were accompanied by the following changes in the electrical activity of the subicular neuronal network: 1) the power of low-band EEG was increased in SE mice; 2) high amplitude EEG spikes occurred during the Up-phase; 3) the mean burst rate and percentage of burst discharges increased in the pyramidal neurons of SE mice; 4) the latency of evoked potentials to LEnt electrical stimulation was shorter and the duration was longer in bursting (pyramidal) neurons in SE mice; and 5) the duration of the suppression period after electrical stimulation of the LEnt was significantly shorter in the SE group.
Neuronal loss in the hippocampus of animal models (6, 50, 52) and of patients (15, 31, 32, 51) with MTLE is well known. However, so far, microanatomical reorganization of neural circuits in the subicular in epilepsy remains elusive. Compared to neurons in the hippocampus, subicular neurons were assumed to be resistant to epileptic injury (14, 22, 49). However, recent studies showed that neuronal loss occurred in the subiculum of the rat model (9, 26) and of patients with MTLE (14). In the present study, we did not demonstrate the reduction of the area of subiculum, but we showed that about 14% of subicular neurons were lost, which is similar to the data from the rat model of MTLE (26). The loss of higher percentages of CB (46.5%) and PV (34.4%) immunopositive interneurons suggests that there is a disproportionate loss of pyramidal neurons and interneurons, i.e., loss of the higher percentage of interneurons may result in the unbalance of interaction among pyramidal neurons and interneurons, leading to the generation of interictal epileptiform activity in the dorsal subiculum. Of course, further study is needed to exclude the possibility that loss of higher percentages of CB and PV is not caused by down-regulation of CB or PV per se but not death of these interneurons.
In the present study, we demonstrated a higher density of en passant and terminal boutons of PHA-L immunopositive axons from the LEnt in the pyramidal cell layer of the dorsal subiculum. It strongly suggests that axon sprouting from LEnt to subiculum may occur. Furthermore, we observed that bursting neurons had more dendritic spines in SE compared to the control groups. It may provide a neuroanatomical basis to support the present neurophysiological theory that the latency of evoked potentials to LEnt electrical stimulation is shorter and the duration is longer in bursting (pyramidal) neurons in SE mice compared to the control groups. Enhanced synaptic contacts have also been reported in CA1-subiculum-CA1 circuit (8). Our result is inconsistent with that of a previous study by Knopp et al. (2005)(26). This may be due to the fact that we investigated the bursting discharge neurons in the in vivo state, but Knopp’s study is based on an in vitro model. In the present study, enhanced local synaptic contacts of bursting pyramidal neurons and reduced dendritic spines in subicular interneurons in SE mice may be related to the amplification and synchronization of epileptic discharges.
The disturbance in subiculum local network electrical activity 2–3 months after SE was obvious in freely-moving mice. These animals displayed interictal spikes and pathological high-frequency oscillations, as well as electrographic and behavioral seizures (Bragin et al. unpublished observations). These disturbances were not so obvious under urethane anesthesia. However, we found an increase in the power of low-band EEG activity and the occurrence of EEG interictal spikes during the Up-phase under our experimental conditions.
It is known that pyramidal neurons in subiculum are burst-firing cells, and GABAergic interneurons are fast irregular-firing cells (19, 20, 25, 35, 38). However, it is unknown which subtypes of interneurons are fast irregular-firing cells. In the present study, we developed a novel method to double-labeled BDA-injected neurons with markers for interneurons such as CB, CR, and PV. Using this method, by double amplification with avidin-horseradish peroxidase conjugate and primary mouse antibody for HRP, neurons with juxtacellular injection of BDA were well labeled. The co-localization of BDA with PV in 5 out of 8 BDA-labeled neurons strongly suggests that PV immunopositive interneurons may be excited in SE mice when the entorhinal-subiculum pathway is stimulated. Whether these interneurons could produce a depolarizing GABAergic effect on the postsynaptic neurons as shown in a previous study in patients with MTLE, or whether their effect is inhibitory, remains to be elucidated (11, 43).
In the present study, synchronous interictal spikes were observed in the epileptic subiculum (Fig. 2), as they were in previous reports on human epileptic subiculum (10, 11, 61). To the best of our knowledge, there are no publications describing the neuronal correlates of this in vivo interictal epileptiform activity. Our results show that on the basis of patterns of electrical activity, subicular neurons can be divided into three groups: bursting, regular discharging, and irregular discharging cells, which is similar to the classification used in freely-moving rats (2, 47). A similar classification of subicular neurons into bursting neurons, regular-spiking neurons, and fast irregular-spiking neurons was adopted in in vitro experiments (20, 34, 35, 38, 48, 50, 57, 59). However, there is a large discrepancy in the percentages of these three types of neurons between this report and previous reports. Previous in vitro studies showed that bursting activity is a vital characteristic of subicular pyramidal neurons (30, 35, 37, 48). Our results showed that in vivo spontaneous patterns are correlated with the immunochemical properties of Dsub neurons. We have demonstrated that pyramidal neurons display bursting or regular discharges, and that none of the bursting neurons were PV, CB, or CR immunopositive. Irregular fast-discharging cells are interneurons and they are mostly PV-immunopositive.
The number of bursting neurons was increased in the epileptic Dsub by about 10%, which is in accordance with the data obtained in epileptic rats (59), although the percentage increase is much lower in the mice. This may be explained by differences in experimental conditions, i.e., the in vitro study by Welmer et al (59) and the in vivo approach by our group. In the present study, we observed increased mean-burst rate and percentage of burst discharges in the bursting neurons of SE mice, as were observed in the previous study by Richard et al. (2002) (46). However, the reduction of firing rate of bursting neurons was inconsistent with the hypothesis of hyperexcitability. One possible explanation is that hyperexcitability in pyramidal neurons of the subiculum may be related to enhanced bursting discharges instead of the high firing rate in the in vivo state.
The responses of Dsub neurons to LEnt stimulation included evoked discharges (excitatory) followed by suppression (inhibitory) of discharges. All bursting (pyramidal) neurons in SE mice responded with excitatory discharges, but in some of them a period of suppression could not be detected due to the low rate of spike discharges. The latency of evoked discharges was significantly (~2.5 ms) shorter in the SE group, which is consistent with our histological data that show an increased density of synaptic terminals from the LEnt. The percentage of irregular firing neurons that responded to LEnt stimulation with spike discharge also increased from 16.6% in the control group to 69.2%. Morphological data showed an increased density of PHA-L boutons specifically on the PV type of interneurons but not on the CB and CR type. These data indicate that the targets of reorganized axons from LEnt in the mouse model are pyramidal cells and PV immunopositive interneurons. Functionally, this reorganization would lead to much stronger feed-forward excitatory and inhibitory connections. However, the decrease in post-spike inhibition after electrical stimulation of the LEnt may indicate that feed-back inhibition in the SE group is weaker.
The present study suggests that epileptogenesis in the dorsal subiculum is associated with considerable cell death and synaptic reorganization. These morphological changes are accompanied by significant changes in the pattern of electrical activity of surviving neurons. Given that the subiculum is a major output of the hippocampus, the increase in the density of synaptic connections from the LEnt in the present study, and from CA1 area as reported previously (8), may make it a very important region for the generation of interictal and ictal epileptiform activity, which is integrated into a very complex neural network for epileptogenesis, including activity of hippocampus and entorhinal cortex. Of course, further studies are still needed to establish the causal relationship between neurophysiological changes in the subiculum and onset of epilepsy after pilocarpine-induced status epilepticus.
This study was supported by research grants (Nos: NMRC/0777/2003, NMRC/0960/2005 from the National Medical Research Council, SHF/FG217P/2005 from Singhealth Research Fund) to FR Tang and NIH, USA grants (NS-08208, NS-33310) to Dr. Engel.