A group of neurons with the characteristics of dentate gyrus granule cells was found at the hilar/CA3 border several weeks after pilocarpine- or kainic acid-induced status epilepticus. Intracellular recordings from pilocarpine-treated rats showed that these “granule-like” neurons were similar to normal granule cells (i.e., those in the granule cell layer) in membrane properties, firing behavior, morphology, and their mossy fiber axon. However, in contrast to normal granule cells, they were synchronized with spontaneous, rhythmic bursts of area CA3 pyramidal cells that survived status epilepticus. Saline-treated controls lacked the population of granule-like cells at the hilar/CA3 border and CA3 bursts.
In rats that were injected after status epilepticus with bromodeoxyuridine (BrdU) to label newly born cells, and also labeled for calbindin D28K (because it normally stains granule cells), many double-labeled neurons were located at the hilar/CA3 border. Many BrdU-labeled cells at the hilar/CA3 border also were double-labeled with a neuronal marker (NeuN). Taken together with the recent evidence that granule cells that are born after seizures can migrate into the hilus, the results suggest that some newly born granule cells migrate as far as the CA3 cell layer, where they become integrated abnormally into the CA3 network, yet they retain granule cell intrinsic properties. The results provide insight into the physiological properties of newly born granule cells in the adult brain and suggest that relatively rigid developmental programs set the membrane properties of newly born cells, but substantial plasticity is present to influence their place in pre-existing circuitry.
The pilocarpine and kainic acid animal models of epilepsy are two of the most common experimental tools in epilepsy research (Nadler, 1981; Ben-Ari, 1985; Turski et al., 1989; Mello et al., 1993; Dudek et al., 1994; Sperk, 1994). In adult rats, systemic administration of either pilocarpine, a muscarinic agonist, or kainic acid, an agonist of the kainic acid subtype of glutamate receptor, leads to status epilepticus, which lasts for several hours. Several days to weeks after status, spontaneous limbic seizures often occur and continue intermittently thereafter (Nadler, 1981; Ben-Ari, 1985; Tauck and Nadler, 1985; Turski et al., 1989; Mello et al., 1993; Liu et al., 1994; Sperk 1994; Hellier et al., 1998; Motte et al., 1998).
Brain damage after pilocarpine- and kainic acid-induced status epilepticus is usually severe and widespread. In the hippocampus, neuronal loss in the pyramidal cell layers and hilar region is common, whereas the dentate gyrus granule cells are relatively resistant (Nadler, 1981; Mello et al., 1993; Obenaus et al., 1993; Liu et al., 1994; Buckmaster and Dudek, 1997; Okazaki et al., 1999). A common finding is mossy fiber “sprouting,” which refers to the abnormal growth of dentate granule cell axons (the mossy fibers) into the inner molecular layer of the dentate gyrus. Similar pathology has been reported in human temporal lobe epilepsy (TLE; Margerison and Corsellis, 1966; Babb and Brown, 1987; Meldrum and Bruton, 1992;Sutula et al., 1989; Babb et al., 1991; Mathern et al., 1997).
As a result of the studies of pathology in animal models and TLE, hilar and pyramidal cell loss has often been associated with limbic epileptogenesis. However, previous studies of pilocarpine- and kainic acid-treated rats showed that there can be some preservation of hilar and CA3 pyramidal neurons (Buckmaster and Dudek, 1997; Okazaki et al., 1999). Therefore, it was of interest to examine which neurons survived. To address this issue, immunocytochemical experiments were performed after pilocarpine or kainic acid treatment, using antibodies to a number of neuronal markers. To our surprise, immunocytochemistry using an antibody to the calcium-binding protein calbindin D28K (CaBP) demonstrated a cluster of immunoreactive cells at the hilar/CA3 border that, to our knowledge, has not been previously reported, either in normal or epileptic conditions. CaBP normally stains granule cells and a few scattered GABA neurons (“interneurons”; Baimbridge and Miller, 1982; Sloviter, 1989; Tóth and Freund, 1992; Seress et al., 1993). After seizures, CaBP expression can decrease in granule cells (Maglóczky et al., 1997; Sloviter et al., 1991; Tønder et al., 1994; Shetty and Turner, 1995; Yang et al., 1997; Nagerl et al., 2000) (but see Lowenstein et al., 1991), but novel cell types expressing CaBP in the hilus or CA3 have not been described.
This report presents morphological, immunohistochemical, and electrophysiological evidence that the CaBP-immunoreactive cells at the hilar/CA3 border are related to granule cells. The hypothesis is raised that they represent newly born granule cells that migrated from their site of origin to the hilar/CA3 border. This hypothesis is based on a number of studies that illustrate the plasticity in the dentate gyrus after seizures. For example, seizures can lead to altered location of granule cells, as demonstrated by dispersion of the granule cell layer (Houser, 1990). Seizures can also lead to the extension of “basal” dendrites of granule cells into the hilar region (Spigelman et al., 1998; Buckmaster and Dudek, 1999). Finally, seizures increase granule cell neurogenesis, and it has been shown that the newly born cells can migrate into the hilus (Parent et al., 1997).
Dentate granule cell neurogenesis is a process that normally occurs throughout life (Gould and Cameron, 1996; Gage et al., 1998). Many newly born cells arise from the subgranular proliferative zone, a lamina between the granule cell layer and the hilus (Gould and Cameron, 1996). Seizure-induced neurogenesis occurs after brief seizures (Bengzon et al., 1997; Nakagawa et al. 2000) or those that are more prolonged, such as pilocarpine- or kainic acid-induced status epilepticus (Parent et al., 1997, 1998; Gray and Sundstrom, 1998;Nakagawa et al., 2000). Kindling also increases granule cell neurogenesis (Parent et al., 1998; Scott et al., 1998). Of particular relevance to our hypothesis was a study by Parent et al. (1997), which used the mitotic marker bromodeoxyuridine (BrdU) to label newly born cells. They showed that some granule cells that are born after seizures can migrate from the subgranular zone into the hilus.
Our results suggest that a large number of newly born granule-like cells can migrate out of the granule cell layer after pilocarpine-induced status epilepticus, and they can migrate quite far from their presumed site of origin, the subgranular zone. The results also suggest functional implications.
MATERIALS AND METHODS
Animal care and use met the guidelines set by the National Institutes of Health and the New York State Department of Health. All chemicals were purchased from Sigma (St. Louis, MO) unless otherwise noted.
Pilocarpine and kainic acid treatment. Adult male Sprague Dawley rats (180–240 gm) were injected with either (1) atropine methylbromide (1 mg/kg, s.c.) and 30 min later with pilocarpine hydrochloride (380 mg/kg, i.p.), or (2) kainic acid (12 mg/kg, i.p.). Diazepam (5 mg/kg, i.p.; Wyeth-Ayerst) was injected after 1 hr of status. The onset of status was defined as the first stage 5 seizure (Racine, 1972) that did not abate after several minutes. After ∼5 hr, animals were injected with 2.5 ml 5% dextrose in lactate–Ringer's solution subcutaneously. For ∼7 d, diet was supplemented with apples that were cut open and left at the bottom of the cage. Saline controls were the same age group as drug-treated rats and received identical treatment (atropine, diazepam, and apple), but were injected with saline instead of convulsant. Rats were observed for spontaneous behavioral (motor) seizures (stage 5) at random times between 7:00 A.M. and 8:00 P.M. Observation duration and frequency ranged from 5 min (1–5 times per day) to 3 hr (1–2 times per week).
Hippocampal slice preparation and maintenance. Hippocampal slices (400-μm-thick) were prepared from ether-anesthetized rats after decapitation. Slices were cut in ice cold buffer (“sucrose-buffer”) containing (in mm): 126 sucrose, 5 KCl, 2.0 CaCl2, 2.0 MgSO4, 26 NaHCO3, 1.25 NaH2PO4, and 10d-glucose, using a Vibroslice (Campden Instruments). Slices were immediately placed on a nylon net at an interface of sucrose buffer and warm (32–33°C), humidified (95%O2, 5% CO2) air using a slice chamber (Fine Science Tools) that was modified in two ways: (1) buffer approached the slices from their undersurface and was directed up and over them and then to a distant exterior port; (2) more air vents were made to allow more humidified air to the area where slices were located. All slices from a given animal were placed in the recording chamber immediately after the dissection. Thirty minutes after slices were placed in the chamber, buffer was switched to one containing NaCl substituted equimolar for sucrose (“NaCl-buffer”). Recordings began 30 min thereafter until ∼7 hr after the dissection. Flow rate was ∼1 ml/min.
Electrophysiological recording and analysis. Intracellular and extracellular recordings were made as previously described (Scharfman 1995a,b). Recordings were made with intracellular glass electrodes (0.75 mm inner diameter, 1.0 mm outer diameter) filled with 4% Neurobiotin (Vector Laboratories, Burlingame, CA) in 1m potassium acetate (60–80 MΩ). Intracellular data were collected using an intracellular amplifier with a bridge circuit (Axoclamp 2B; Axon Instruments, Foster City, CA), and the bridge was balanced whenever current was passed. Extracellular electrodes were filled with NaCl buffer (5–10 MΩ). Data were collected using a digital oscilloscope (Nicolet Instruments) and analyzed with accompanying software. Data were also digitized and saved on tape (Neurocorder; Neurodata Instruments) for analysis offline.
Cells that were impaled were first screened to ensure that they were healthy (stable resting potential more than −55 mV, overshooting action potential; Table 1). Their intrinsic (membrane) properties were then characterized using intracellularly injected current steps (0.05–1.5 nA, 150 or 500 msec). Analysis of intrinsic properties was made as previously described (Scharfman 1995a,b). In brief, resting potential was defined as the difference between the potential while intracellular and that recorded after withdrawing the microelectrode from the cell. Input resistance was defined by the steepest slope of an I–V curve based on steady-state responses to a family of current pulses (0.05–1.0 nA, 150 msec). Time constant was defined as the time to reach 63% of the steady-state response to a minimal current step (0.1 nA), i.e., one that did not activate rectifying currents. Action potential (AP) characteristics were based on a single AP at threshold evoked by current injected intracellularly (a 0.1–0.5 nA, 150 msec pulse) at resting potential. AP “total” amplitude was measured from resting potential to peak, and “threshold” amplitude was from the membrane potential at which the AP started to peak (i.e., between thearrowheads shown in Fig. 6 A). Total AP duration was the time between the start of the AP until the point when it repolarized to the starting membrane potential. Half-width was the period from the start of the AP to the time when the AP reached half of the threshold amplitude. AP maximum rising and decay slopes were defined by the maximum dv/dt using a resolution of 50 kHz. Dv/dt ratio was defined as the ratio of slope to decay. Afterhyperpolarization (AHP) amplitude was measured from the membrane potential at which the AP started to the peak of the AHP. AHP half-duration was measured from the point when the AHP started to the point when it had decreased to half its peak amplitude. Statistical comparisons were made using Student's t tests (PSI-plot version 4.5; Poly Software International) or χ2 analysis. Significance was set at 0.05 before all experiments.
Intracellular dye injection and processing. Neurobiotin was injected with repetitive depolarizing current pulses (+0.3–1.0 nA, 20 msec, 30 Hz, 5–15 min) after electrophysiological data were collected. Immediately after the experiment, slices were immersed in fixative (4% paraformaldehyde) and refrigerated for up to 2 weeks. They were resectioned at 50 μm using a vibratome (Ted Pella). After incubation overnight in 0.4% Triton X-100, sections were washed in Tris buffer (3 times for 5 min each), incubated in 0.3% H2O2 for 45 min, washed, incubated in ABC (Vector Laboratories), washed, incubated for 20 min in 0.1% NiNH3SO4, followed by 20 min in diaminobenzidine (Polysciences, Warrington, PA; 50 mg/100 ml Tris), washed, dehydrated in a series of graded alcohols (10 min each: 70, 90, and 95%, then 10 min in 100% twice), cleared in xylene, and coverslipped in Permount (Fisher Scientific, Pittsburgh, PA). Slides were examined using an Olympus BH-2 light microscope and photographed with 35 mm camera attachment using Tmax film (100ASA; Eastman Kodak, Rochester, NY).
Histology and immunocytochemistry. Standard procedures for cresyl and Timm stain were used, as previously described (Scharfman et al., 1999a). For immunocytochemistry (Sloviter 1991; Scharfman 1999a), perfusion-fixed (4% paraformaldehyde, pH 7.4) tissue was sectioned (50 μm) with a vibratome (Ted Pella) and processed with an antibody to CaBP (monoclonal, 1:100,000, Sigma or polyclonal, 1:5000, Chemicon, Temecula, CA), calretinin (polyclonal, 1:20,000; Chemicon), Glial fibrillary acidic protein (GFAP; monoclonal, 1:1000; Boehringer Mannheim, Indianapolis, IN), neuropeptide Y (polyclonal, 1:2000; Peninsula), parvalbumin (polyclonal, 1:3000; Swant, Bellinzona, Switzerland), somatostatin (polyclonal, 1:8000; Peninsula), GABA (polyclonal, 1:10,000; Chemicon), glutamic acid decarboxylase (GAD 67; polyclonal, 1:3000; Chemicon), or NeuN (an antibody to a neuronal nuclear protein; monoclonal, 1:5000; Chemicon). For GABA or GAD, fixative contained 2% paraformaldehyde and 1.5% glutaraldehyde (Polysciences).
Studies of neurogenesis. Bromodeoxyuridine (BrdU; 50 mg/kg, i.p.; Boehringer Mannheim) was injected twice per day, ∼8 hr apart, on days 4–11 and 26–30 after pilocarpine injection. After perfusion with 4% paraformaldehyde, sections were processed with an antibody to BrdU (monoclonal, 1:1000, Boehringer Mannheim) as described (Parent et al., 1998), using diaminobenzidine as a chromagen and nickel intensification (50 mm NiCl2). Sections labeled for BrdU were double-labeled for CaBP and NeuN by incubating with an avidin–biotin blocking solution followed by incubation in primary antibody for CaBP or NeuN. Visualization of the second marker was made with a water-soluble chromagen that binds to the avidin–biotin peroxidase complex (NovaRed; Vector).
Rats were examined 1–21 months after pilocarpine- (n = 72) or kainic acid- (n = 10) induced status and compared to 10 age-matched saline-treated controls. Of the 72 pilocarpine-treated rats, 49 were dedicated solely to anatomical studies, and the other 23 were studied both physiologically as well as anatomically. The 10 rats treated with kainic acid were only studied anatomically. Of the saline controls, all 10 were studied anatomically, and 4 of the 10 were also studied physiologically. For animals that were studied both physiologically and anatomically, one hemisphere was dedicated for hippocampal slices, and the other hemisphere was immediately fixed by immersion. Anatomical findings from immersion-fixed tissue were equivalent to those from perfusion-fixed tissue, so the results from immersion- and perfusion-fixed tissue are pooled.
General behavioral and anatomical observations
Pilocarpine- and kainic acid-treated rats demonstrated substantial hilar and CA3 neuronal survival (Fig. 1). These rats had spontaneous seizures and mossy fiber sprouting, similar to previous studies of pilocarpine-treated rats that had brief periods of status epilepticus (i.e., 1 hr; Gibbs et al., 1997; Bausch and Chavkin, 1997),or longer periods of status (i.e., over 2 hr; Tauck and Nadler, 1985; Mello et al., 1993; Obenaus et al., 1993; Liu et al., 1994; Okazaki et al., 1995, 1999; Parent et al., 1997; Hellier et al., 1998; Mathern et al., 1998; Motte et al., 1998; Sutula et al., 1998).
Spontaneous limbic seizures (stage 5 of Racine, 1972) were observed in all rats that had convulsant-induced status, except for 9 (n = 3 of 72 pilocarpine, n = 6 of 10 kainic acid). This value is likely to underestimate the number of seizures because the animals were not continuously monitored. In addition, seizures without a behavioral component were not monitored. The frequency of observed stage 5 seizures was 2.47 ± 0.37 seizures per month (mean ± SEM, n = 47; range, 0.25–14.2) for the pilocarpine group and 0.68 ± 0.15 (n = 4; range, 0.25–1.0) for the kainic acid group. Some seizures were greater than stage 5, including wild running and vocalizations.
Anatomical evaluation along the entire septotemporal axis of the hippocampus was conducted in 29 rats after pilocarpine-induced status and recurrent seizures. In these cases, the ventral pole was cut horizontally until reaching the middle of the hippocampus, and then the brain was rotated to section the septal half of the hippocampus in the coronal plane. The horizontal sections facilitated evaluation of the entorhinal cortex, where substantial neuronal loss in the superficial layers was almost always present (28 of 29 rats; 96.5%; Fig. 1). Thinning of the cortex overlying the hippocampus (the perirhinal and temporal cortices) and ventricular enlargement occurred in 18 of 29 rats (62.1%). Sprouting in the dentate gyrus, usually evaluated by neuropeptide Y immunoreactivity, was common (i.e., 28 of 29, 96.5%; Fig. 1). Four of the pilocarpine-treated rats were also examined using Timm stain to mark mossy fibers, and these results demonstrated similar mossy fiber sprouting as shown by neuropeptide Y staining. In the four kainic acid-treated rats with recurrent seizures, all had entorhinal loss and sprouting, but there was no apparent shrinkage of the perirhinal/temporal cortex or ventricular enlargement. Saline-treated controls had no detectable hippocampal or entorhinal cell loss, perirhinal/temporal cortical thinning, ventricular enlargement, or sprouting (Fig. 1).
Calbindin-immunoreactive cells at the hilar/CA3 border in rats with status
Immunocytochemistry was used to identify characteristics of surviving hilar neurons (Figs.2-4). CaBP immunocytochemistry showed, surprisingly, a novel plexus of immunoreactive cells at the hilar/CA3 border in rats with recurrent seizures, but not in saline controls (Fig. 2, A vsD). The hilar/CA3 cells resembled granule cells in their CaBP immunoreactivity, and also in the small size and oval shape of their somata (Figs. 2 B, 3 D, 10, 11). For these reasons and others described below, the hilar/CA3 cells are referred to as “granule-like.”
The granule-like cells were present in all rats that were examined throughout the septohippocampal axis (n = 18). The borders of the hilus were defined as in Amaral (1978). Two different fixatives and antibodies to CaBP provided the same results. The plexus was present in animals that were killed within a broad temporal window (1–21 months after status). It was present in animals that had a high frequency of observed seizures as well as those with no observed seizures. Some of the animals with a plexus had mossy fiber sprouting, entorhinal cell loss, and ventricular enlargement (Figs. 1, 4), whereas in other cases these characteristics were absent. Thus, the granule-like CaBP-stained plexus at the hilar/CA3 border was present after two pharmacologically distinct convulsants and appeared to be unrelated to the number of observed spontaneous seizures, survival time after status, or other changes in hippocampal and extrahippocampal structure that might be developing concurrently.
To quantify the CaBP-labeled granule-like cells, 50 μm sections were examined at 200–400 μm intervals throughout the septal half of the hippocampus in 18 rats with spontaneous seizures (n = 14 pilocarpine, n = 4 kainic acid). The septal half was chosen because preliminary studies showed that it contained the most CaBP-immunoreactive hilar neurons. Only tissue that was stained using the same antibody (Sigma) was used. For a given animal, all CaBP-labeled hilar cells were counted in each 50 μm section, and the greatest value of CaBP-immunoreactive hilar cells/50 μm section was then averaged across all animals. Thus, the maximum number of CaBP-labeled cells/50 μm section was 55.4 ± 7.4 (n = 18; range, 16–129). For pilocarpine-treated rats, the mean was 55.9 ± 8.5 (n = 14; range, 16–129). For the kainic acid group, there were 54.0 ± 16.8 cells (n = 4; range, 23–99). In saline-injected controls that were studied in an analogous manner, the number of CaBP-immunoreactive hilar cells per section was lower (6.0 ± 1.48; n = 5; range, 1–9; Student's t test,p < 0.05), and none of the CaBP-immunoreactive cells were located at the hilar/CA3 border.
The hilar/CA3 cells could vary in their distribution within the hilar/CA3 area, and in their number from section-to-section along the septotemporal axis. Variation was slight within a range of several hundred micrometers (Fig. 4 B,C), but could be substantial over large distances. An example of variation within a single animal is shown in Figure 3. The midhippocampal section had many more CaBP-immunoreactive cells in the hilus than a section from the septal pole of the same animal (Fig. 3 A vs C). The midhippocampal section also showed immunoreactive cells throughout the hilus, whereas in the septal section the cells were clustered at the hilar/CA3 border near the dorsal blade of the dentate gyrus (Fig. 3 A vs C).
Further immunocytochemical analysis of hilar/CA3 granule-like cells
It was possible that the CaBP-immunoreactive cells at the hilar/CA3 border were actually a modified GABA neuron or glial cell. To address this possibility, immunocytochemistry was conducted using antibodies to several markers of GABA neurons and a marker of glia. GABA neuron markers included GABA (n = 10; Fig.3 B), glutamic acid decarboxylase (GAD67; n = 4; data not shown), parvalbumin (n = 7; Fig. 4), neuropeptide Y (n = 50; Fig. 4), calretinin (n = 14; Fig. 4), and somatostatin (n = 6; Fig. 4). GFAP (n = 4; Fig. 4) was used as a glial marker. Although the granule-like cells at the hilar/CA3 border were labeled by CaBP, there were few or no cells in this area that were stained using markers of GABA neurons or GFAP (Fig. 4). Thus, the hilar/CA3 cells had neurochemical similarities to granule cells because they stained for CaBP, but did not stain like GABAergic interneurons or normal adult glia. They also contrasted with other hilar/CA3 cell types, such as CA3 pyramidal cells and mossy cells, because pyramidal cells and mossy cells are not CaBP-immunoreactive, either before or after seizures (Figs. 2, 3; Freund et al., 1991; Sloviter, 1989; Seress et al., 1993; Shetty and Turner, 1995).
Hippocampal slice electrophysiology and intracellular labeling
To determine the morphological and physiological characteristics of neurons at the hilar/CA3 border, neurons in this location were recorded intracellularly in slices from pilocarpine-treated animals that had status and subsequent spontaneous seizures (referred to below as the pilocarpine group, n = 55 slices from 23 rats). All slices from a given animal were placed in the recording chamber at the same time so that they would be exposed to the same conditions. Four saline-treated controls (14 slices) and four pilocarpine-treated animals that did not have status or subsequent spontaneous seizures (10 slices), were examined in an analogous manner. Data from the last two groups did not show differences, so they are pooled below and referred to as controls.
Neurons from the hilar/CA3 border were recorded (pilocarpine,n = 62; controls, n = 14), as well as granule cells in the granule cell layer (pilocarpine, n= 68; controls, n = 26). In control rats, granule cells located 50–100 μm from the granule cell layer were used for additional comparisons (“ectopic” granule cells; n= 6; Gaarskjaer and Laurberg, 1983; Table 1).
Morphology of granule-like neurons at the hilar/CA3 border
In the pilocarpine group, many neurons at the hilar/CA3 border had, surprisingly, the characteristics of normal adult granule cells in the granule cell layer. No cells with granule-like characteristics were found at the hilar/CA3 border in control tissue.
Twenty-one hilar/CA3 neurons from the pilocarpine group were recorded that had granule cell-like intrinsic properties (see below), and 12 were filled with Neurobiotin. All cells had small, oval cell bodies and spiny dendrites, like granule cells (Fig.5). There were no thorny excrescences, also like granule cells. Three had a polarized dendritic tree, like a normal adult granule cell (i.e., dendrites arising from one side of the cell body; Fig. 5 A). Other cells were bipolar, having dendrites emerging from opposite sides of the cell body (Fig.5 B). The bipolar cells resembled granule cells after seizures with “basal” dendrites (Spigelman et al., 1998; Buckmaster and Dudek, 1999).
An extensive axon was labeled in three granule-like cells, and mossy fiber boutons were evident in the hilar region and in stratum lucidum of area CA3 (Fig. 5 B). Figure 5 B3 shows an example in which the major branch of the axon could be traced from the hilus to area CA3b. There were periodic mossy fiber boutons, and some had filamentous extensions, as has been described for classic mossy fiber boutons of normal adult granule cells of the granule cell layer (Fig. 5 B4–B5; Blackstad and Kjaerheim, 1961; Hamlyn, 1962;Claiborne et al., 1986; Frotscher, 1985, 1989; Chicurel and Harris, 1992; Acsády et al., 1998). Axons of two of the three neurons also had collaterals in the inner molecular layer (Fig.5 B3). The inner molecular layer axon collaterals suggest that granule-like cells can contribute to mossy fiber sprouting (but see Parent et al., 1999).
Electrophysiological characteristics of granule-like cells at the hilar/CA3 border
As shown in Table 1, the membrane properties of granule-like cells at the hilar/CA3 border could not be distinguished from granule cells that were located in the granule cell layer in the same slices. Membrane properties also were similar to normal adult granule cells recorded from saline controls or “ectopic” granule cells in normal rats (i.e., granule cells that were located 50–100 μm from the granule cell layer/hilar border; Table 1).
Other types of cells at the hilar/CA3 border were distinct from granule-like cells. In both control and epileptic tissue, cells with morphological and physiological similarities to interneurons or CA3c pyramidal cells were encountered at the hilar/CA3 border (Figs.6-7; interneurons: n = 10 pilocarpine, n = 2 control; pyramidal cells: n = 26 pilocarpine,n = 6 control). Hilar mossy cells were not encountered at the hilar/CA3 border, but were recorded in the areas closer to the granule cell layer (data not shown).
APs of granule-like hilar/CA3 neurons and granule cells located in the granule cell layer were similar (Fig. 6), and were unlike the APs of interneurons and CA3c pyramidal cells. Single APs (evoked at threshold by injected current, see Materials and Methods) of granule cells and granule-like cells were characterized by a relatively fast rising phase and slow decay, and a triphasic AHP (Fig. 6, Table 1), as has been described for granule cells in normal tissue (Scharfman, 1992, 1995a;Spruston and Johnston, 1992; Staley et al., 1992; Williamson et al., 1993; Mott et al., 1997; Lübke et al., 1998). Interneuron APs were characterized by relatively similar AP rising and decay slopes (i.e., a dv/dt ratio closer to 1.0; Table 1), and a large, rapid AHP, as previously described (Fig. 6; Lacaille et al., 1989; Scharfman, 1992, 1993, 1995a; Buhl et al., 1994; Buckmaster and Schwartzkroin, 1995; Smith et al., 1995; Freund and Buzsáki, 1996; Mott et al., 1997; Lübke et al., 1998; Parra et al., 1998). CA3c pyramidal cell firing behavior at threshold demonstrated intrinsic bursts of three or four APs, as has been demonstrated in past studies of these neurons in normal tissue (Fig. 6; Wong and Prince, 1981; Scharfman, 1993; Smith et al., 1995). Pyramidal cell AP slopes were similar to granule cells, but there was no appreciable fast AHP (Fig. 6; Wong and Prince, 1981; Scharfman, 1993; Smith et al., 1995).
In granule cells and granule-like neurons, multiple APs (evoked by currents more than threshold, injected intracellularly) occurred in trains with strong spike frequency adaptation. Strong spike frequency adaptation has been described previously as a characteristic of normal adult granule cells (Scharfman, 1992; Williamson et al., 1993; Mott et al., 1997). Adaptation was quantified in two ways: (1) by analyzing the first three interspike intervals of a train with exactly four APs in 150 msec (Fig. 7 E; the amount of injected current was manipulated so that exactly four APs were evoked in each cell), and (2) by analyzing the first interstimulus interval as a function of injected current (f–I curves; Fig. 7 F). Both interspike interval and f–I curves yielded comparable values between granule-like cells and granule cells located in the cell layer (Fig. 7 A,B,E,F). However, hilar interneurons recorded in the same slices were quite different, as has been described in normal rats (Fig. 7 C; Scharfman, 1992, 1995; Buckmaster and Schwartzkroin, 1995; Mott et al., 1997; Lübke et al., 1998). Thus, the f–I curve of interneurons had a shallow slope compared to granule cells and granule-like cells (Fig.7 F). Relatively weak adaptation of interneurons was evident in the similarity of interspike intervals within a train of four APs (Fig. 7 C,E). CA3c pyramidal cells were also different from granule cells and granule-like cells in that intrinsic bursts of APs were evoked rather than trains of adapting APs, as has been described in normal rats (Fig. 7 D; Wong and Prince, 1981; Scharfman, 1993; Smith et al., 1995). Thus, in membrane properties and firing behavior, granule-like cells were similar to granule cells and distinct from interneurons and pyramidal cells.
Spontaneous epileptiform bursts in area CA3 of slices
Interestingly, there were abnormalities in the CA3 region in slices from pilocarpine-treated rats with spontaneous seizures, in addition to abnormalities at the hilar/CA3 border. CA3 pyramidal cells often had periodic spontaneous bursts of APs riding on large depolarizations (presumably EPSPs). Furthermore, all granule-like cells impaled in slices with CA3 population bursts (n = 16) had bursts of APs that were synchronous with CA3 (Fig.8). The CA3 bursts resembled those that occur in normal rat hippocampal slices exposed to convulsants (Wong and Traub, 1983; Swann and Brady, 1984; Ault et al., 1986; Chestnut and Swann, 1988; Müller and Misgeld, 1991; Perreault and Avoli, 1991), including pilocarpine (Nagao et al., 1994; Rutecki and Yang, 1998) or animal models of epilepsy (Lee et al., 1995b; Smith et al., 1998).
CA3 spontaneous population bursts were 75–500 msec in length, occurred at a 0.1–0.33 Hz, and only occurred in animals that had spontaneous seizures. Bursts could be recorded extracellularly, but this was difficult because population spikes within each burst were only 1–2 mV in amplitude (Figs. 8-9), much less than population spikes recorded using normal slices, similar recording conditions, and exposed to convulsants (Scharfman, 1994).
CA3 epileptiform bursts were recorded preferentially in animals that were examined at least 5 months after status. Thus, in the 18 rats in which extracellular recordings were made over 5 months after status, 90% (19 of 21 slices; n = 9 rats) of slices had spontaneous activity in at least one of the 3 CA3 subfields (i.e., a, b, and c). For tissue <5 months after status, 47% of slices (15 of 32; n = 9 rats) demonstrated spontaneous CA3 bursts, which was significantly different (χ2 analysis, χ2 = 10.38).
All CA3 pyramidal cells that were recorded intracellularly had bursts of APs that were synchronous with bursts recorded extracellularly from the CA3 cell layer (n = 20), but CA3 cells did not burst if there was no extracellularly recorded CA3 burst (n = 6) or in saline controls (n = 6). Figure 9 shows a simultaneous extracellular recording from the CA3b pyramidal cell layer and an intracellular recording from a CA3c pyramidal cell. These records illustrate synchronous spontaneous bursts in pyramidal cells of areas CA3c and CA3b. Field potentials in CA3a and CA3c were usually smaller than those in CA3b, and in some cases could not be detected, even when epileptiform bursts were recorded extracellularly in CA3b. However, in those cases, CA3 a or c neurons had bursts that were synchronized with the CA3b population burst recorded extracellularly.
The onset of bursts in granule-like cells relative to the onset of the CA3 population burst (defined as the onset of the first population spike within a burst) varied. In some cases bursts of APs in granule-like cells sometimes began at the onset of the population burst in CA3 (Fig. 8). This suggests tight synchrony. However, in other slices the bursts of granule-like cells could occur up to 48 msec after the onset of CA3 population bursts. The onset of bursts in pyramidal cells also varied relative to the onset of the CA3 population burst. Pyramidal cell bursts could start up to 10 msec before, or up to 55 msec after the onset of the population event in CA3b. However, despite variability among different neurons, a given cell maintained a relatively fixed latency relative to the onset of the CA3 population burst.
Are granule-like hilar/CA3 neurons newly born granule cells?
Given the similarity of the cells at the hilar/CA3 border to granule cells immunocytochemically (i.e., immunoreactive for CaBP but not parvalbumin, somatostatin, calretinin, etc.), morphologically (i.e., small soma, spiny dendrites, mossy fiber axon), and electrophysiologically (membrane properties, firing behavior), and given the evidence that granule cells born after seizures migrate into the hilar region (Parent et al., 1997), we hypothesized that the CaBP-labeled cells at the hilar/CA3 border were granule cells that were born after status and migrated abnormally to the hilar/CA3 border.
To address this hypothesis, three pilocarpine-treated rats and two saline controls were injected with BrdU (after pilocarpine or saline injection) and were examined several weeks later. Double-labeling of neurons at the hilar/CA3 border was apparent in all pilocarpine-treated animals, but not in saline-treated controls. Figure10 shows double-labeling of CaBP- and BrdU-immunoreactive cells at the hilar/CA3 border from one of the pilocarpine-treated animals with recurrent seizures, examined 4 weeks after BrdU injections stopped. Figure11 shows BrdU-labeled cells were also immunoreactive for NeuN, and therefore are likely to be neurons rather than glia. These data are consistent with the CaBP-positive nature of newly born granule cells in adult rats described elsewhere (Markakis and Gage, 1999). Thus, cells at the hilar/CA3 border included newly born cells with granule cell characteristics. Importantly, other types of cells at the hilar/CA3 border may also be newly born.
The number of CaBP/BrdU- or NeuN/BrdU-immunoreactive neurons we observed does not necessarily reflect the number that were newly born. BrdU labeling is likely to underestimate the number of newly born cells, given that BrdU was only administered for a short period, and granule cells could have been born at many other times, i.e., whenever animals had spontaneous seizures. Furthermore, CaBP immunoreactivity is not present in immature granule cells (Rami et al., 1987; Goodman et al., 1993), and some of the newly born cells could have been immature at the time of perfusion-fixation. CaBP expression also decreases in adult granule cells after seizures (Baimbridge and Miller, 1984;Baimbridge et al., 1985; Shetty and Turner, 1995; Sloviter et al., 1991; Yang et al., 1997; Nagerl et al. 2000) (but see Lowenstein et al., 1991).
In summary, a novel population of granule-like neurons has been demonstrated in the hippocampus in two animal models of epilepsy. The granule-like neurons were similar morphologically, electrophysiologically, and neurochemically to the granule cells in the granule cell layer. Results of BrdU-labeling experiments were consistent with the hypothesis that the granule-like cells represent granule cells that were born after status and migrated incorrectly to the hilar/CA3 border. In addition, it was shown that the granule-like neurons had spontaneous, regular epileptiform bursts that were synchronized with area CA3 pyramidal cells. Epileptiform activity occurred preferentially in animals that had been examined long after status. The epileptiform bursts of newly born granule-like cells and CA3 pyramidal cells suggest new mechanisms that could contribute to abnormal excitability.
Comparison of granule-like neurons and normal adult granule cells in the granule cell layer
Granule cells in the granule cell layer and granule-like neurons were similar in several ways. They were similar in the morphology because they had small, oval somata, mossy fiber axons with giant boutons, spines on dendrites, and lacked thorny excrescences. They were similar in immunoreactivity to CaBP, but not somatostatin, parvalbumin, or calretinin. They were similar to granule cells in pilocarpine-treated rats with mossy fiber sprouting because they had axon collaterals in the inner molecular layer. To our knowledge, there is no other hippocampal cell type besides granule cells that has all of these characteristics.
Similarities were also apparent physiologically. Membrane properties and firing behavior of granule-like cells were similar to normal granule cells (e.g., high resting potential, triphasic AHP, and strong spike frequency adaptation). The characteristics were distinct from other cell types in the hilar/CA3 region, such as interneurons and CA3 pyramidal cells. They were also unlike mossy cells in normal rats (Scharfman and Schwartzkroin, 1988; Scharfman, 1999).
Granule-like cells differed from normal granule cells in three ways. First, granule-like cells were not immunoreactive using antibodies to GABA or GAD. This is a characteristic of granule cells located in the cell layer, particularly in the time just after seizures (Sandler and Smith, 1991; Schwarzer and Sperk, 1995;Sloviter et al., 1996; Makiura et al., 1999). However, neither the granule cells in the granule cell layer nor the granule-like cells at the hilar/CA3 border appeared to be strongly immunoreactive in our tissue. One reason might be that the animals that were examined did not have a recent behavioral seizure.
Second, neuropeptide Y is also expressed in granule cells after seizures (Sperk et al., 1992, 1996; Lurton and Cavalheiro, 1997), but was not a characteristic of granule-like neurons (i.e., there was no evidence of small, oval neuropeptide Y-immunoreactive somata at the hilar/CA3 border). However, there could have been axon terminal expression; this could not be evaluated because the locations of axon terminals of granule-like cells overlap the sites of terminals of other neurons that express neuropeptide Y (e.g., granule cells, interneurons).
Third, most granule-like cells had bipolar dendritic trees rather than polarized dendrites, the latter being the norm for granule cells in the cell layer. However, in rodent epileptic tissue some granule cells have “basal” dendrites, making them bipolar (Spigelman et al., 1998; Buckmaster and Dudek, 1999). Furthermore, other species or other nonepileptic preparations have demonstrated that granule cells can develop basal dendrites (Seress and Pokorny, 1981; Lübbers and Frotscher, 1987; Seress and Mrzljak, 1987; Isokawa et al., 1993). Therefore, the presence of basal dendrites on granule-like cells does not distinguish them greatly from granule cells in the cell layer.
Thus, there were few clear differences between granule-like cells and granule cells. Perhaps the one obvious difference was the regular epileptiform bursts of granule-like cells (see below).
Spontaneous bursts of area CA3 pyramidal cells in pilocarpine-treated rats with seizures
One of the surprising findings in this study was that area CA3 pyramidal cells had periodic synchronous bursts in slices from pilocarpine-treated rats after status and spontaneous seizures. This was surprising because previously studies of pilocarpine-treated rats have not described this activity (Isokawa, 1996a,b; Bausch and Chavkin, 1997; Molnár and Nadler, 1999 Okazaki et al., 1999), probably because they focused on the dentate gyrus only or there was more CA3 damage. However, such activity is a common finding in acute hippocampal slices that are exposed to convulsants, including pilocarpine (Wong and Traub, 1983; Swann and Brady, 1984; Ault et al., 1986; Chestnut and Swann, 1988; Müller and Misgeld, 1991; Perreault and Avoli, 1991;Nagao et al., 1994; Lee et al., 1995b; Rutecki and Yang, 1998;Smith et al., 1998). It also has been noted that CA3 sharp waves (Buzsáki, 1986), a form of synchronized CA3 activity, occurs in epileptic tissue (Buzsáki et al., 1989).
Granule-like neurons were synchronized with area CA3 epileptiform bursts
As mentioned above, one clear difference between granule-like cells and granule cells in the cell layer was spontaneous bursts of granule-like neurons that were synchronized with area CA3 population bursts. In some cases the synchrony was close enough so that ephaptic interactions or gap junctions may have been the underlying substrate. However, in other cases synchrony was not tight, suggesting polysynaptic mechanisms.
The synchronized activity could contribute mechanistically to the generation of spontaneous limbic seizures that occur in these animals. This might occur if granule-like cells innervate granule cells of the granule cell layer, which is possible given their axon collaterals in the inner molecular layer. Thus, a spontaneous burst in a granule-like cell might trigger reverberatory excitatory activity in the sprouted network of granule cells. That could in turn lead to excitatory activity in CA3, via the mossy fiber innervation of CA3. Seizure-like activity could exit the hippocampus via CA1. However, both CA3 and granule-like neurons could also innervate interneurons and have indirect inhibitory effects, dampening seizure-like activity (Ribak and Peterson, 1991; Sloviter, 1992; Kotti et al., 1997).
At the present time it is too soon to say whether the granule-like cells, and their periodic bursts with CA3 pyramidal cells, play a role in the recurrent seizures of pilocarpine-treated rats. The net effect of the synchronous activity in CA3 and the hilar/CA3 border is likely to depend on several factors, including the extent of cell loss, sprouting of any of the residual cell types onto other neurons, changes in excitatory/inhibitory receptors after seizures (Rice et al., 1996;Kapur and Macdonald, 1997; Brooks-Kayal et al., 1998) and the myriad of other effects that seizures produce (Sheng and Greenberg, 1990; Gall, 1993; Lynch et al., 1996).
One argument against a critical role in epileptogenesis comes from a study of seizure-induced neurogenesis in pilocarpine-treated rats (Parent et al., 1999). It was shown that recurrent seizures occurred after irradiation stopped seizure-induced neurogenesis in the subgranular zone (Parent et al., 1999). Although it was noted that some newly born cells in the hilus occurred despite irradiation, it is likely they were glia (Parent et al., 1999). These data suggest that granule-like cells at the hilar/CA3 border are not critical to limbic seizures in pilocarpine-treated rats. However, they could be a contributing factor, when present. They may also be important to other aspects of the epileptic condition, such as behavioral/cognitive deficits associated with seizures (Leite et al., 1990; Hermann et al., 1992; McNamara et al., 1992; Stafstrom et al., 1993; Sutula et al., 1995; Stubley-Weatherly et al., 1996). Granule-like cells may play a role in these abnormalities because of their ability to disrupt normal hippocampal signal processing in the trisynaptic pathway.
The origin of granule-like neurons
There are several hypotheses that could explain the origins of the granule-like cells at the hilar/CA3 border. Because some of the granule-like cells were labeled by BrdU, they could represent some of the new granule neurons that are born after seizure-induced neurogenesis (Bengzon et al., 1997; Parent et al., 1997, 1998; Gray and Sundstrom, 1998; Scott et al., 1998; Nakagawa et al. 2000).
Another hypothesis, which does not involve neurogenesis, is that granule-like cells are caused by abnormal migration of adult granule cells from the granule cell layer. This possibility seems unlikely given the lack of precedent for adult neurons to move to new locations in the mature brain. A third possibility would be that the normal hilar/CA3 cells (interneurons, mossy cells, CA3c pyramidal cells, and glia) develop a granule cell phenotype (morphologically and electrophysiologically) after seizures. This seems unlikely given the extensive physiological and morphological changes that would be involved.
Relevance to other epileptic or nonepileptic conditions
Other animal models of epilepsy besides those considered in the present study may also involve granule-like neurons, because several types of seizures or seizure-inducing stimuli can lead to increased granule cell neurogenesis and possible migration into the hilus. However, in other conditions, the appropriate milieu (including glia, growth factors, etc.; Lowenstein and Arsenault, 1996) may be different and not foster the development of the same cluster of granule-like cells at the hilar/CA3 border or the epileptiform bursts.
It will be valuable to examine human epileptic tissue for CaBP immunoreactivity, because the one study of CaBP immunoreactivity in human epileptics that has been published demonstrated CaBP-immunoreactive neurons in the hilar region with morphological characteristics of granule cells, although it was pointed out that they could be interneurons (Sloviter et al., 1991).
Because a variety of events besides seizures can lead to increased dentate granule cell neurogenesis, granule-like neurons may also develop after other events besides seizures. Thus, changes in such diverse factors as stress (Cameron and Gould, 1994; Gould et al., 1997,1998), learning (Gould et al., 1999), exercise (van Praag et al., 1999), environmental stimulation (Kempermann et al., 1998a,b), neurotransmitters and hormones (Cameron et al., 1995; Brezun and Dazsuta, 1999; Tanapat et al., 1999), lesions (Gould and Tanapat, 1997), or hypoxia/ischemia (Liu et al., 1998;Bossenmeyer-Pourié et al., 1999) can result in alterations in neurogenesis in the dentate gyrus. A few hilar cells that have anatomical similarities to granule cells have been demonstrated in studies of ischemia (Hsu and Buzsáki, 1993) and mouse mutants (Robbins et al., 1999; Woodhams and Terashima, 1999), and many CaBP-immunoreactive hilar neurons have been reported after tetanus toxin injection (Lee et al., 1995a). Thus, granule-like neurons may develop after a variety of normal and pathological events. Exactly which life events (normal or abnormal) trigger the development of granule-like neurons, how many are necessary to alter hippocampal function, and when this contributes to or counteracts neuropathology, are questions for future experiments.
This work was supported by National Institutes of Health Grant 38285 to H.E.S. We thank Drs. Gyuri Buzsáki, Sam Deadwyler, Fred Gage, Dan Lowenstein, Timothy Pedley, Philip Schwartzkroin, and Karen Smith for discussions during the course of this study and comments on an earlier form of this manuscript, and Annmarie Curcio and Ruth Marshall for technical and secretarial assistance.
Correspondence should be addressed to Helen Scharfman, Center for Neural Recovery and Rehabilitation Research, Helen Hayes Hospital, Route 9W,West Haverstraw, NY 10993-1195. E-mail:.