|Home | About | Journals | Submit | Contact Us | Français|
One potential mechanism of temporal lobe epilepsy is recurrent excitation of dentate granule cells through aberrant sprouting of their axons (mossy fibers), which is found in many patients and animal models. However, correlations between the extent of mossy fiber sprouting and seizure frequency are weak. Additional potential sources of granule cell recurrent excitation that would not have been detected by markers of mossy fiber sprouting in previous studies include surviving mossy cells and proximal CA3 pyramidal cells. To test those possibilities in hippocampal slices from epileptic pilocarpine-treated rats, laser scanning glutamate uncaging was used to randomly and focally activate neurons in the granule cell layer, hilus, and proximal CA3 pyramidal cell layer while measuring evoked excitatory postsynaptic currents (EPSCs) in normotopic granule cells. Consistent with mossy fiber sprouting, a higher proportion of glutamate-uncaging spots in the granule cell layer evoked EPSCs in epileptic rats compared to controls. In addition, stimulation spots in the hilus and proximal CA3 pyramidal cell layer were more likely to evoke EPSCs in epileptic rats, despite significant neuron loss in those regions. Furthermore, synaptic strength of recurrent excitatory inputs to granule cells from CA3 pyramidal cells and other granule cells was increased in epileptic rats. These findings reveal substantial levels of excessive, recurrent, excitatory synaptic input to granule cells from neurons in the hilus and proximal CA3 field. The aberrant development of these additional positive-feedback circuits might contribute to epileptogenesis in temporal lobe epilepsy.
Recurrent, excitatory circuits can generate excessive positive-feedback and seizure activity (Wong et al., 1986), and their aberrant development might be epileptogenic. In many patients with temporal lobe epilepsy (Sutula et al., 1989; de Lanerolle et al., 1989; Houser et al., 1990) and after epileptogenic injuries in animal models (Nadler et al., 1980; Lemos and Cavalheiro, 1995; Golarai et al., 2001; Santhakumar et al., 2001) granule cell axons (mossy fibers) grow from their normal location in the hilus into the molecular layer where they form synapses (Babb et al., 1991; Represa et al., 1993; Zhang and Houser, 1999; Buckmaster et al., 2002) and excite neighboring granule cells (Wuarin and Dudek, 1996; Molnár and Nadler, 1999; Lynch and Sutula, 2000; Scharfman et al., 2003). Some studies found positive correlations between anatomical measures of mossy fiber sprouting and seizure frequency (Mathern et al., 1993, 1997; Lemos and Cavalheiro, 1995; Wenzel et al., 2000b; Pitkänen et al., 2005; Kharatishvili et al., 2006), but most have not (Cronin and Dudek, 1988; Sloviter, 1992; Masukawa et al., 1992; Mello et al., 1993; Buckmaster and Dudek, 1997; Spencer et al., 1999; Timofeeva and Peterson, 1999; Gorter et al., 2001; Nissinen et al., 2001; Lynd-Balta et al., 2004; Rao et al., 2006; Pitkänen et al., 2000; Wenzel et al., 2000a; Lehmann et al., 2001; Zhang et al., 2002; Raol et al., 2003; Jung et al., 2004; Williams et al., 2004; Harvey and Sloviter, 2005; Kadam and Dudek, 2007; Buckmaster and Lew, 2011). Lack of correlation might be attributable to other aberrant, recurrent, excitatory circuits that activate granule cells but are not detected by mossy fiber labeling techniques.
Additional possible sources of positive-feedback to granule cells include surviving mossy cells and proximal CA3 pyramidal cells. Normally, mossy cells receive excitatory synaptic input from nearby granule cells (Scharfman et al., 1990; Sik et al., 2006), but most of their axon projections are to distant septotemporal levels (Buckmaster et al., 1996) leaving relatively few recurrent synapses with granule cells in hippocampal slices (Buckmaster et al., 1992; Scharfman, 1995). However, after lesioning entorhinal input to the dentate gyrus, mossy cell axons sprout and synapse with granule cells (Del Turco et al., 2003; Prang et al., 2003), which raises the possibility of synaptic reorganization under other conditions, including temporal lobe epilepsy. Proximal CA3 pyramidal cells in control animals receive excitatory synaptic input from granule cells and occasionally extend axon collaterals retrogradely into the inner molecular layer of the dentate gyrus where they could synapse with granule cell dendrites (Li et al., 1994). In epileptic rats, tracer injections reveal more axon projections from proximal CA3 to the dentate molecular layer (Siddiqui and Joseph, 2005). Together, these findings suggest that in temporal lobe epilepsy granule cells might receive excessive recurrent excitation through surviving mossy cells and proximal CA3 pyramidal cells, but functional evidence is lacking. We used laser scanning photo-uncaging of glutamate to evaluate potential contributions of hilar and CA3 neurons to recurrent excitation of granule cells in epileptic pilocarpine-treated rats.
All experiments were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Stanford University Institutional Animal Care and Use Committee. Male Sprague-Dawley rats (Harlan) were treated with pilocarpine when they were 27–41 d old, as described previously (Buckmaster, 2004). Briefly, pilocarpine (380 mg/kg, i.p.) was administered 20 min after atropine methyl bromide (5 mg/kg, i.p.). Diazepam (Hospira) was administered (10 mg/kg, i.p.) 2 h after the onset of stage 3 or greater seizures (Racine, 1972) and repeated as necessary over the next 24 h to suppress convulsions. Control rats included animals that were treated identically but did not develop status epilepticus (n = 13) and naïve rats (n = 10). There were no significant differences in naive and pilocarpine-treated control rats, so data were combined. Control rats were 65 ± 3 d old (mean ± sem, range 47–100 d) when slices were prepared. Epileptic rats (n = 32) were used for slice experiments when they were 72 ± 2 d old (range 51–91 d), 38 ± 2 d (range 17–57 d) after pilocarpine-induced status epilepticus. Video monitoring, which began at least 10 d after pilocarpine treatment, verified spontaneous motor convulsions in all epileptic rats. No controls were observed to have spontaneous seizures.
Animals were deeply anesthetized with urethane (1.5 g/kg, i.p.) and decapitated. Tissue blocks including the dentate gyrus were removed rapidly and stored for 3 min in modified ice-cold artificial CSF (mACSF) containing (in mM): 230 sucrose, 2.5 KCl, 10 MgSO4, 1.25 NaH2PO4, 26 NaHCO3, 2.5 CaCl2, and 10 glucose. Horizontal slices (400 µm) were prepared with a microslicer (Leica, VT1000S). Slices were incubated at 32°C for 30 min in a submersion-type holding chamber that contained 50% mACSF and 50% normal ACSF, which consisted of (in mM): 126 NaCl, 3 KCl, 2 MgSO4, 1.25 NaH2PO4, 26 NaHCO3, 2 CaCl2, and 10 glucose. After that, slices were transferred to normal ACSF at 32°C for 1 h. ACSF was aerated continuously with a mixture of 95% O2 and 5% CO2. Slices were maintained at room temperature until used for recording.
Slices were transferred to a recording chamber where they were minimally submerged in high divalent cation ACSF (oxygenated in 95% O2 and 5% CO2) containing (in mM): 121 NaCl, 2.5 KCl, 1.25 NaH2PO4, 4 CaCl2, 4 MgSO4, 26 NaHCO3, and 10 glucose. High concentrations of divalent cations were used to suppress polysynaptic recurrent excitation (Shepherd et al., 2003; Jin et al., 2006), and 10 µM 2-amino-5-phosphonovaleric acid (d-APV) was added to block N-methyl-d-aspartate receptor-dependent events and further reduce the possibility of polysynaptic responses (Jin et al., 2006; Kumar et al., 2007). Recordings were obtained at room temperature (22–23°C) from cells identified by Nomarski optics with a 63X water-immersion lens and infrared video microscopy (Zeiss). Recordings were obtained with an Axopatch 700B amplifier and pClamp software, filtered at 2 kHz for voltage-clamp or 10 kHz for current-clamp, digitized at 20 kHz, and stored digitally. Series resistance was monitored continuously, and those cells in which this parameter changed by > 20% were rejected. Recording electrodes were pulled from borosilicate glass tubing (1.5 mm outer diameter) and had a resistance of 3–5 MΩ when filled with internal solution, which contained (in mM): 100 potassium gluconate, 40 HEPES, 10 EGTA, 5 MgCl2, 2 Na2ATP, 0.5 NaGTP, and 20 biocytin. Osmolarity was adjusted to 300 mOsm and pH to 7.3 with 1 M KOH. Access resistance was < 20 MΩ.
An established glutamate uncaging protocol was used (Jin et al., 2006; Kumar et al., 2007; Brill et al., 2009). A frequency-tripled Nd:YVO4 laser (Series 3500 pulsed laser, 100 kHz repetition rate; DPSS Lasers) was interfaced with an upright microscope through its epifluorescence port via several mirrors and lenses. Movement of the laser beam was controlled with mirror galvanometers and triggered by scanning and data acquisition software (developed by J.R. Huguenard), which also registered recorded soma position with respect to stimulation sites. Caged glutamate (200 µM, (S)-α-amino-2,3-dihydro-4-methoxy-7-nitro-δ-oxo-1H-indole-1-pentanoic acid, Tocris) was added to 20 ml of recirculating high divalent cation ACSF at the beginning of each experiment. Focal photolysis of caged glutamate was accomplished by a 400-µs 170-mW UV light stimulus through a 5X objective to activate neurons. Half-width of the light beam was ~50 µm. All neurons tested exhibited at least one action potential when photostimulated at the soma.
Researchers were blind to experimental groups during data analysis but not during data acquisition. Intrinsic electrophysiological properties were measured. Input resistance was the slope of regression lines of plots of voltage responses to current steps ≤± 10 pA. Membrane time constants were calculated from single-exponential fits to voltage responses to hyperpolarizing currents. Action potential amplitude was measured from resting membrane potential to the peak.
Spontaneous and evoked excitatory postsynaptic current (EPSC) data were evaluated using Mini Analysis (Synaptosoft). Threshold for event detection was set at 3 times root mean square noise level. EPSC recordings were obtained at the inhibitory reversal potential (−70 mV) without pharmacologically blocking GABA receptor-mediated events. However, at the concentration used in the present study (200 µM) the caged glutamate compound substantially but incompletely blocks GABAA receptors (Fino et al., 2009). Before collecting photostimulus-evoked responses, spontaneous EPSCs were recorded for at least 2 min. Laser photostimuli randomly scanned the granule cell layer, hilus, and proximal CA3 pyramidal cell layer in a grid pattern with 60 µm spacing and 1 s interstimulus intervals. Averages of 241 ± 6 and 247 ± 7 spots/slice were scanned for connectivity mapping of granule cells in control and epileptic rats, respectively.
Several parameters were evaluated to quantify synaptic connectivity. We identified spots that when stimulated evoked an EPSC in a recorded granule cell (see hotspot criteria in Results). Percentage of hotspots was defined as the fraction of tested spots within a region that were identified as hotspots. Composite amplitude was defined as the sum of peak amplitudes of all synaptic events during the detection window (5–40 ms after the laser stimulus). Composite amplitude was used to construct connectivity maps. The number of EPSCs that occurred during detection windows was measured. Mean individual EPSC amplitude was calculated by dividing composite amplitude by number of EPSCs.
Connectivity maps were divided into four regions: granule cell layer (62 ± 3 spots in control, 73 ± 4 spots in epileptic), border (40 ± 1 spots in control, 41 ± 1 spots in epileptic), hilus (110 ± 5 spots in control, 104 ± 5 spots in epileptic), and CA3 pyramidal cell layer (26 ± 1 spots in control, 28 ± 1 spots in epileptic) (Figure 1B3). After electrophysiological data had been obtained and slices had been processed for Neuronal-specific nuclear protein (NeuN)-labeling (see below), contours were drawn around the granule cell layer and proximal CA3 pyramidal cell layer. Neurons at the border of the granule cell layer and hilus include granule cells and hilar neurons. Therefore, spots touching the contour line between the granule cell layer and hilus were given special consideration and designated as “border” spots. Conversely “granule cell layer” spots were within the granule cell layer contour and did not touch the border with the hilus. The hilar region was limited by the border region and by straight lines drawn from the ends of the granule cell layer to the proximal end of the CA3 pyramidal cell layer. Only spots within stratum pyramidale and not other strata of CA3 were included as “CA3 pyramidal cell layer” spots.
After electrophysiological recording, hippocampal slices were placed in 4% formaldehyde in 0.1 M phosphate buffer (PB; pH 7.4) at 4°C for at least 24 h. After fixation, slices containing biocytin-labeled cells were stored in 30% ethylene glycol and 25% glycerol in 50 mM PB at −20°C until they were processed using a whole-mount protocol with counterstaining by NeuN-immunocytochemistry. Slices were rinsed in 0.5% Triton X-100 and 0.1 M glycine in 0.1 M PB and then placed in a blocking solution containing 0.5% Triton X-100, 2% goat serum (Vector), and 2% bovine serum albumin in 0.1 M PB for 4 h. Slices were incubated in mouse anti-NeuN serum (1:100, MAB377, Chemicon) in blocking solution overnight. Slices were then rinsed in 0.5% Triton X-100 and 0.1 M glycine in 0.1 M PB and incubated with Alexa 594 streptavidin (5 µg/ml) and Alexa 488 goat anti-mouse (10 µg/ml, Molecular Probes) in blocking solution overnight. Slices were rinsed in 0.5% Triton X-100 and 0.1 M glycine in 0.1 M PB, mounted on slides, coverslipped with Vectashield (Vector), and imaged with a confocal microscope (Zeiss).
After immuno-processing and collecting images of biocytin-labeled recorded neurons, slices were cryoprotected in 30% sucrose in 0.1 M PB and sectioned with a microtome set at 40 µm. A section from each slice was mounted on a glass slide and coverslipped in Vectashield. NeuN-immunoreactivity was imaged throughout the section thickness with a 10X objective on a confocal microscope. Neurolucida (MBF Bioscience) was used to draw a contour around the hilus (as defined above), and the area was recorded. The number of hilar NeuN-positive profiles was counted, and density was calculated by dividing by hilar area. Nissl staining was superior to NeuN-immunoreactivity for visualizing individual neuronal somata in cell layers. A section from each slice was stained with 0.25% thionin. Using Neurolucida and a 10X objective, a contour was drawn around the CA3 pyramidal cell layer from the proximal tip to a level determined by a straight line drawn across the tips of the granule cell layer, and the area was recorded. Using a 100X objective, all neuron profiles within the proximal CA3 pyramidal cell contour were counted. Proximal CA3 pyramidal cell density was calculated by dividing the number of neuron profiles by area. To estimate granule cell profile density, a contour was drawn around the granule cell layer. Stereo Investigator (MBF Bioscience) was used systematically sample within the granule cell layer (counting frame = 20 × 20 µm; counting grid = 200 × 200 µm). Using a 100X objective, granule cell soma profiles were counted if they fell at least partially within the counting frame but did not touch upper or left borders. Granule cell profile density was measured by dividing number of counted profiles by sampled area.
All chemicals were from Sigma unless specified otherwise. Statistical values are presented as mean ± standard error of the mean (sem). The level of p < 0.05 was considered statistically significant.
Laser scanning photostimulation is a powerful approach for efficiently evaluating neuronal circuits, but it is not necessarily a direct measure of monosynaptic connectivity. Therefore, several experiments were conducted to establish validity before comparing connectivity of control and epileptic animals. Direct versus synaptic responses to glutamate uncaging were identified as described previously (Kumar et al., 2007). In voltage-clamp mode, direct responses evoked at sites close to recorded granule cell somata or proximal dendrites consistently had short onset latencies (< 5 ms) and persisted after bath application of tetrodotoxin (1 µM, n = 5, data not shown). In current clamp recordings, an action potential could be triggered shortly after the onset of a direct response. An excitatory postsynaptic current (EPSC) would be generated in the synaptic targets of the directly activated neuron after a delay for action potential propagation and synaptic transmission. Therefore, responses with onset latencies > 5 ms after photostimulation were considered to be synaptic. A relatively short detection window (5–40 ms, 35 ms duration) reduced the probability of confounding effects through polysynaptic activation.
Spontaneous EPSC frequency increases in granule cells in epileptic pilocarpine-treated rats (Simmons et al., 1997; Wuarin and Dudek, 2001). Spontaneous events that occur within the detection window could be mistaken as evoked EPSCs. To reduce that possibility, slices in the present study were bathed in the presence of d-APV and high concentrations of divalent cations. Under those conditions frequencies of sEPSCs were similar in granule cells from control (4.1 ± 0.6 Hz, n = 33) and epileptic rats (4.2 ± 0.4 Hz, n = 38), suggesting any effect of spontaneous events would be equal in both groups. To further decrease contamination of spontaneous events, photostimulation scanning was completed three times for each granule cell, and the three resulting maps were merged into an average map by a “two successes” filtering criterion (Figure 1). That is, only sites associated with EPSCs in their detection windows during at least two of three trials were considered to be sites of presynaptic neurons (“hotspots”). At those sites, responses of all non-failure trials were averaged. Together, the “two successes” criterion and similar sEPSC frequencies in control and epileptic rats reduced possible confounding effects of sEPSCs and polysynaptic activation on connectivity maps.
To test the reliability of this method for locating presynaptic neurons, individual hotspots were stimulated repetitively at 30 s intervals for 7–14 trials (Figure 2). Several parameters were measured to evaluate consistency of evoked EPSCs. Onset latency was measured as the interval between the onset of laser stimulation and the onset of the first EPSC in the detection window. Latency jitter was calculated as the standard deviation of the Gaussian distribution of onset latencies (bin size = 0.2 ms). Success rate was measured as the number of trials evoking at least one EPSC in detection windows divided by the total number of trials. All tested hotspots had short average onset latency (8–14 ms), low average latency jitter (0.49–0.55 ms), and high average success rate (> 75%) (Figure 2B). There was no correlation between onset latency and latency jitter for evoked EPSCs (R2 = 0.02–0.08). Such a correlation would be expected if evoked EPSCs contained both monosynaptic and longer latency polysynaptic responses with increased jitter (Larimer et al., 2008).
Intrinsic excitability of principle neurons may increase in epilepsy (Prince and Tseng, 1993; Jin et al., 2005). Changes in intrinsic physiology could enhance responses of granule cells to synaptic inputs. To test this possibility, intrinsic physiology of granule cells was measured and compared in control and epileptic rats. There were no significant differences in resting membrane potential, input resistance, membrane time constant, action potential amplitude, or action potential frequency in response to a 200 pA, 500 ms depolarizing current step (Table 1).
Glutamate receptor density of principal neurons can increase in epilepsy (Mody and Heinemann, 1987; Babb et al., 1996; Mathern et al., 1998), which could enhance direct responses to uncaged glutamate, resulting in more action potential firing of neurons in epileptic rats compared to controls. Hyperexcitability could make hotspots more likely to be detected with equivalent photostimulation in epileptic animals, which would be a confounding factor when evaluating connectivity. To test this possibility, neurons in control and epileptic rats were evaluated to determine whether their direct responses to glutamate uncaging were similar under our stimulation and recording conditions (Figure 3). Whole-cell or cell-attached current-clamp recording was used to obtain responses of granule cells (n = 14 in control, n = 20 in epileptic), mossy cells (n = 11 in control, n = 9 in epileptic), CA3 pyramidal cells (n = 11 in control, n = 12 in epileptic), and hilar ectopic granule cells (n = 7 in epileptic). Maps (540 × 540 µm, centered on recorded neurons) were scanned with 60 µm spacing, 400 µsec stimulus durations, and 170 mW stimulus intensity in normal ACSF with 200 µM caged glutamate. All neurons responded with at least one short-latency action potential when photostimuli were located close to the soma or proximal dendrites (latency to peak = 6.2 ± 0.2 ms, n = 488 action potentials). The number of action potential-evoking spots per cell was smaller for normotopic granule cells and hilar ectopic granule cells compared to those of mossy cells and CA3 pyramidal cells (Figure 5B), which is consistent with size differences of the neuron types. However, for each cell type, there was no significant difference between control and epileptic rats in the number of action potential-evoking spots per cell or number of action potentials per action potential-evoking spot.
Maps constructed to compare direct responses to glutamate uncaging also were used to test whether or not polysynaptic activation occurred. Each map tested 81 spots in a 540 × 540 µm area. In a subset of cells (controls: 8 mossy cells and 6 CA3 pyramidal cells; epileptics: 3 granule cells, 5 mossy cells, 5 CA3 pyramidal cells, and 6 hilar ectopic granule cells), maps were expanded to include the entire granule cell layer, hilus, and proximal CA3 region (425 ± 9 stimulation sites per map). If polysynaptic activation occurred, postsynaptic target neurons should receive EPSPs that trigger an action potential. Despite the large number of sites tested (>18,000), photostimulation only evoked action potentials by direct activation at sites close to the soma or proximal dendrites, never at more distant locations. These findings suggest that under our stimulation and recording conditions, directly activated neurons failed to bring their postsynaptic targets to threshold for firing an action potential. Therefore, recorded EPSCs were rarely if ever generated by polysynaptic activation and instead were likely to be monosynaptic.
Neuron loss occurs in the dentate gyrus of epileptic pilocarpine-treated rats (Mello et al., 1993; Kobayashi and Buckmaster, 2003). All other things being equal, slices with lower neuron density would be expected to display fewer hotspots. Densities of neuronal profiles were measured in the proximal CA3 pyramidal cell layer, hilus, and granule cell layer (Figure 4). All regions demonstrated reduced neuron density in epileptic rats. The most severe loss was in the CA3 pyramidal cell layer where profile density decreased almost by half (0.0041 ± 0.0003 versus 0.0022 ± 0.0001 / µm2, p < 0.001). Neuron profile density in the hilus of epileptic rats, which is likely a balance of neuron loss during status epilepticus and subsequent addition of newly generated ectopic granule cells, was reduced to 70% of the control value (0.00082 ± 0.00006 versus 0.00057 ± 0.00005 / µm2, p < 0.005). The granule cell layer demonstrated moderate but significant reductions in profile densities in epileptic rats (0.0194 ± 0.0010 versus 0.0150 ± 0.0007 / µm2, p < 0.001). Therefore, based on neuron density alone, one would expect reduced connectivity in epileptic animals compared to controls
Experiments described above established the validity of the glutamate photouncaging approach for comparing connectivity maps of normotopic granule cells in slices from control and epileptic rats. Furthermore, neuron profile density measurements suggested epileptic rats would have reduced connectivity unless synaptic reorganization had occurred. To evaluate recurrent excitatory circuits of granule cells, whole-cell voltage-clamp recordings were obtained from 33 and 38 granule cells in slices from control and epileptic rats, respectively. Laser stimuli (400 µs stimulus duration, 170 mW stimulus intensity, 200 µM caged-glutamate in ACSF) were applied to the granule cell layer, hilus, and proximal CA3 pyramidal cell layer in a grid pattern with 60 µm spacing. Maps were constructed of recurrent excitatory connectivity based on composite amplitudes of evoked EPSCs at each spot (Figure 5). Composite amplitude is the summed amplitude of EPSCs that occur during the detection window (5–40 ms after laser stimulation). Most maps of control rat slices contained relatively few if any hotspots. Some maps of epileptic rat slices were similar to controls, but most contained more hotspots in the granule cell layer, hilus, and/or proximal CA3 pyramidal cell layer. Maps of slices from epileptic rats were more likely than those of controls to have at least one hotspot in the granule cell layer [33/38 (87%) versus 11/33 (33%), p < 0.005, Chi-squared], border [28/38 (74%) versus 12/33 (36%), p < 0.005], and hilus [27/38 (71%) versus 12/33 (36%), p < 0.005]. A larger proportion of tested sites in the CA3 pyramidal cell layer were hotspots in epileptic rats compared to controls [10/38 (26%) versus 5/33 (15%)], but the difference was not statistically significant.
To quantify and compare connectivity between control and epileptic groups, hotspot ratios were calculated (Figure 6A). Hotspot ratios in the granule cell layer were higher in epileptic rats compared to controls (6.7 ± 1.2% versus 1.4 ± 0.5%, p < 0.0005), suggesting more recurrent excitatory inputs from normotopic granule cells. Hotspot ratios of the hilus (6.0 ± 1.6% in epileptic versus 1.9 ± 0.7% in control, p < 0.05) and proximal CA3 pyramidal cell layer (4.2 ± 1.3% in epileptic versus 0.6 ± 0.3% in control, p < 0.05) also were higher in epileptic rats. For all regions combined, hotspot ratios were 4 times higher in epileptic rats (7.6 ± 1.8% of stimulation sites) than in controls (1.9 ± 0.5%, p < 0.01) (Figure 6B). The border region showed the smallest increase (2.7 times controls), and the CA3 pyramidal cell layer increased most (6.6 times controls). Therefore, contrary to expectations based on reduced neuron density, these findings suggest that compared to controls epileptic rats have more neurons in the hilus and proximal CA3 pyramidal cell layer that synapse with and excite granule cells.
Higher hotspot ratios in slices from epileptic animals suggest granule cells, hilar neurons, and CA3 pyramidal cells sprouted axons and formed new synapses with normotopic granule cells. If so, one might expect presynaptic neurons to have longer axon trajectories in epileptic animals. To begin testing that possibility, distances from hotspots to recorded granule cells were measured. Mean distance to recorded granule cells was farther for hotspots in the granule cell layer (352 ± 20 µm versus 248 ± 39 µm, p < 0.05), hilus (487 ± 18 µm versus 310 ± 28 µm, p < 0.005) and CA3 pyramidal cell layer (823 ± 28 µm versus 574 ± 36 µm, p < 0.005) in epileptic rats compared to controls, respectively (Figure 6C).
To measure strength of excitatory synaptic input to granule cells evoked by glutamate uncaging, composite EPSC amplitude within detection windows was measured. Average composite amplitude per hotspot of epileptic rats were larger than that of controls for responses evoked by stimulating in the granule cell layer, CA3 pyramidal cell layer, and border regions, but not in the hilus (Figure 6D). Composite amplitude is determined by the number of EPSCs during the detection window and the amplitudes of individual EPSCs. Average numbers of EPSCs evoked per hot spot always were close to one (Figure 6E). Numbers of EPSCs tended to be larger in all regions in epileptic rats compared to controls and reached statistical significance in the granule cell layer and hilus. Increased EPSC number could be attributable to more action potentials discharged per neuron or more neurons activated per site. In control and epileptic rats granule cells and mossy cells each had similar numbers of action potentials evoked by direct stimulation (Figure 3B). Therefore, increased EPSC numbers probably reflect more presynaptic neurons activated per hotspot and more recurrent excitatory connectivity in epileptic rats. Mean amplitudes of individual EPSCs were calculated by dividing composite amplitudes by the number of EPSCs during detection windows. Mean amplitudes of individual EPSCs were larger in epileptic animals compared to controls for hotspots in the granule cell layer, CA3 pyramidal cell layer, and border regions, but not in the hilus (Figure 6F). These findings suggest recurrent excitatory synaptic circuits to granule cells from CA3 pyramidal cells and from other granule cells are not only more frequent but also stronger in epileptic animals compared to controls, which is consistent with larger excitatory synapses found on proximal dendrites of granule cells in epileptic animals (Thind et al., 2010).
Hilar ectopic granule cells normally are rare but develop after epileptogenic injuries (Parent et al., 1997; Scharfman et al., 2000), and occur in patients with temporal lobe epilepsy (Houser, 1990; Parent et al., 2006). They have been proposed to contribute to epileptogenesis in part because their soma and much of their dendrites are located in the hilus among granule cell axons, suggesting they receive abnormally high levels of excitatory synaptic input from normotopic granule cells (Pierce et al., 2005). Their hilar position also might increase their probability of receiving synaptic input from other excitatory neurons in the area, including proximal CA3 pyramidal cells. The frequency of mEPSCs is higher in ectopic compared to normotopic granule cells in epileptic pilocarpine-treated rats (Zhan et al., 2010), which is consistent with more excitatory synapses and/or higher probability of glutamate release. To further evaluate excitatory synaptic input to hilar ectopic granule cells, small, round hilar neurons were targeted for recording (n = 18). Biocytin-labeling confirmed that their morphological characteristics matched those of hilar ectopic granule cells (Cameron et al., 2011; Pierce et al., 2011), which included a small soma positioned in the hilus. All biocytin-labeled ectopic granule cells in the present study had a spiny dendrite that extended through the granule cell layer and into the molecular layer (Figure 7). Mossy fiber boutons were evident on ectopic granule cell axons that extended in the hilus with a primary branch projecting toward, and in some cases into, stratum lucidum of CA3 (Figure 7D).
To evaluate synaptic inputs to hilar ectopic granule cells we laser scanning glutamate uncaging was used as described above to construct excitatory connectivity maps (Figure 8). Results of hilar ectopic granule cells were largely similar to those of normotopic granule cells in epileptic rats (Figure 6). However, there were some differences. For example, as expected because of their position in the hilus instead of the granule cell layer, distances from hilar and CA3 hotspots were shorter for ectopic granule cells compared to normotopic granule cells in epileptic rats (Figure 6C). Unlike normotopic granule cells in epileptic rats, hilar stimulation did not evoke a higher hotspot ratio in hilar ectopic granule cells compared to granule cells in controls. And, stimulation sites in the granule cell layer tended to evoke more EPSCs in ectopic granule cells compared to normotopic granule cells in epileptic animals (Figure 6E), but individual EPSC amplitudes were not different (Figure 6F).
Glutamate photouncaging in the hilus and CA3 pyramidal cell layer rarely synaptically excited granule cells in control rats. This study’s principal finding was that despite reduced neuron densities in epileptic rats, photostimulation evoked increased recurrent excitatory responses. These results suggest that in addition to mossy fiber sprouting other sources of positive-feedback to granule cells develop from hilar and CA3 neurons, which might be epileptogenic.
Normally, some CA3 pyramidal cells make retrograde axon projections into the hilus (Ishizuka et al., 1990). Within the hilus, pyramidal cell axons synapse with mossy cells (Scharfman, 1994), which raises the possibility that photostimulated pyramidal cells excited granule cells indirectly through mossy cells in the present study. This seems unlikely. Photouncaging in CA3 evoked responses in granule cells with low failure rates and little jitter in onset latencies, which is more consistent with monosynaptic than disynaptic transmission. Furthermore, photostimulating CA3 never evoked an action potential in recorded mossy cells. CA3 pyramidal cell axons might synapse with granule cell basal dendrites that develop in patients (Franck et al., 1995; von Campe et al., 1997) and animal models of temporal lobe epilepsy (Spigelman et al., 1998) including pilocarpine-treated rats (Thind et al., 2008). However, 39 of 71 recorded granule cells were biocytin-labeled sufficiently to determine whether they had a basal dendrite and none did. CA3 pyramidal cells normally synapse with dentate interneurons (Kneisler and Dingledine, 1995). CA3 input to at least one type of dentate interneuron (GFP-positive somatostatin hilar neurons in GIN mice) increases in epileptic pilocarpine-treated mice (Hunt et al., 2011), which supports the hypothesis of epilepsy-related axon sprouting by CA3 pyramidal cells.
In control rats CA3 pyramidal cells occasionally project axons into the molecular layer of the dentate gyrus (Li et al., 1994). This occurs most frequently in the temporal hippocampus, from which slices were obtained in the present study, and is consistent with the low but non-zero hotspot ratios in CA3 of control animals. In epileptic rats despite substantially reduced neuron density, the average hotspot ratio in CA3 increased over six times that of controls, which is consistent with many more fibers projecting into the dentate gyrus molecular layer after Phaseolus vulgaris-leucoagglutinin was injected into CA3 (Siddiqui and Joseph, 2005). Together, these findings suggest that after epileptogenic injuries surviving CA3 pyramidal cells sprout axons into the molecular layer that synapse with dendrites of normotopic granule cells.
Hilar photostimulation evoked EPSCs in granule cells in controls and even more frequently in epileptic animals. Previous glutamate uncaging and microdrop studies reported no excitatory inputs from the hilus (Wuarin et al., 1996; Molnar and Nadler, 1999). In the present study, success rate for evoking an EPSC by photostimulation appeared to be high (> 75%, Figure 2B), which facilitated detection of functional connectivity, and the entire hilus was scanned with 60 × 60 µm resolution. Other potential contributing factors include animal age, which can affect hilar neuron survival in slices, and duration and severity of status epilepticus, which affects hilar neuron loss.
Results of the present study do not distinguish whether increased hotspot probability in the hilus of epileptic rats is attributable to mossy cells, ectopic granule cells, or both. Normally, because of their preferentially distant axon projections, mossy cell-to-granule cell circuits are mostly but not completely amputated in hippocampal slices (Buckmaster et al., 1992; Scharfman, 1995). Hilar hotspots in slices from control rats probably are attributable to mossy cells. In patients with temporal lobe epilepsy (Blümcke et al., 2000; Seress et al., 2009) and in animal models (Buckmaster and Jongen-Rêlo, 1999; Jiao and Nadler, 2007), many but usually not all mossy cells die. Reduced hilar neuron profile density in the present study is consistent. Surviving mossy cells might sprout axons and form new synapses with nearby granule cells, which could account for increased hotspot probabilities in epileptic animals. Another major hilar cell type, somatostatin interneurons, sprout axons and form new synapses with granule cells in epileptic pilocarpine-treated mice (Zhang et al., 2009). Mossy cells have the capacity to sprout axons and form new synapses even in adult animals (Benowitz and Routtenberg, 1997; Namgung et al., 1997; Del Turco et al., 2003; Prang et al., 2003; Pierce et al., 2007). Mossy cell feedback to granule cells, which normally is distributed distantly, might spread more locally to include nearby granule cells and thereby increase preservation of mossy cell-to-granule cell circuits in slices from epileptic rats.
There is at least one other potential hilar source of excitatory synaptic input to granule cells. Hilar ectopic granule cells normally are rare but develop after epileptogenic treatments (Parent et al., 1997; Scharfman et al., 2000) and occur in patients with temporal lobe epilepsy (Houser, 1990; Parent et al., 2006). They receive excitatory synaptic input from normotopic granule cells (Pierce et al., 2005) and extend axon collaterals into the dentate gyrus molecular layer (Cameron et al., 2011; Pierce et al., 2011) where they could synapse with normotopic granule cells. Ectopic granule cells account for 34% of excitatory neurons in the hilus of epileptic pilocarpine-treated rats, mossy cells account for the rest (Jiao and Nadler, 2007). More work is needed to determine the respective contributions of ectopic granule cells and surviving mossy cells to aberrant positive-feedback circuits.
Highly connected network hubs might contribute to seizure generation in the dentate gyrus (Morgan and Soltesz, 2008). Somata and proximal dendrites of ectopic granule cells are in the hilus where axons of mossy cells, CA3 pyramidal cells, and especially granule cells extend. Hilar ectopic granule cells have been proposed to be aberrant network hubs in part based on their expected high numbers of excitatory synaptic inputs (Cameron et al., 2011; Pierce et al., 2011). If so, one would predict higher hotspot ratios in ectopic versus normotopic granule cells in epileptic rats. However, photostimulation revealed hotspot ratios that were similar when stimulating the granule cell layer and CA3 pyramidal cell layer and were reduced when stimulating the hilus. Fewer hilar hotspots are consistent with smaller dendritic arbors in the inner molecular layer where mossy cell axons concentrate (Zhan et al., 2010). Similar granule cell layer hotspot ratios are consistent with similar EPSCs in normotopic and ectopic granule cells evoked by mossy fiber stimulation in epileptic rats (Zhan et al., 2010). Photostimulation results, therefore, do not support the hypothesis that hilar ectopic granule cells receive more excitatory synaptic input than normotopic granule cells in epileptic animals. However, we cannot exclude the possibility of additional sources of excitatory synaptic input from outside of the areas stimulated within 400 µm-thick slices used in the present study.
Recurrent excitatory circuits can generate seizure-like activity (Wong et al., 1986), so it is reasonable to hypothesize that increased development of recurrent excitatory circuits might be epileptogenic. Results of the present study suggest CA3-to-granule cell projections are developed excessively in epileptic pilocarpine-treated rats, which might convert a largely unidirectional circuit from the dentate gyrus to CA3 to one with potential to amplify signals from the dentate gyrus and to route activity from CA3 into the dentate gyrus (Scharfman, 2007). Excessive retrograde axonal projections might allow burst-firing CA3 pyramidal cells more direct access to the aberrant granule cell associative network, which might trigger seizure events among synaptically reorganized granule cells. In addition, findings of the present study are consistent with the possibility of increased mossy cell feedback to nearby granule cells with similar potential consequences. Electrical stimulation in the proximal CA3 region or hilus can evoke seizure-like events in hippocampal slices from epileptic patients (Gabriel et al., 2004) and from laboratory animals (Patrylo and Dudek, 1998; Winokur et al., 2004), which has been attributed to mossy fiber sprouting. However, axon sprouting by CA3 pyramidal cells and mossy cells also could contribute. Indirect evidence suggests axon sprouting by mossy cells and CA3 pyramidal cells might compete with sprouting mossy fibers for postsynaptic sites on granule cell dendrites (Shetty et al., 2005). These findings might help explain the weak correlation between anatomically measured mossy fiber sprouting and seizure frequency (see Introduction).
Increasing evidence reveals widespread reorganization of excitatory circuits in epileptic brains. Mossy fiber sprouting by granule cells was detected over 30 years ago (Nadler et al., 1980) and might be the most dramatic example in part because of the retained potential for generating new granule cells in adults (Kron et al., 2010). Excitatory axon sprouting occurs in other hippocampal regions of adult animals after epileptogenic injuries (Lehmann et al., 2001), including layer II entorhinal cortical input to dentate granule cells (Thind et al., 2010), CA3 associational connections (McKinney et al., 1997), CA3 projections to CA1, subiculum, and entorhinal cortex (Siddiqui and Joseph, 2005), CA1 input to other CA1 pyramidal cells (Perez et al., 1996; Esclapez et al., 1999; Smith and Dudek, 2001), and CA1 input to subiculum (Cavazos et al., 2004). Epileptogenic injuries in neocortex also cause excitatory axon sprouting (Salin et al., 1995; Jin et al., 2006). Clearly, development of excessive, recurrent, excitatory circuits is a common epilepsy-related abnormality.
We are grateful to Dr. David Prince for providing access to equipment. Supported by NINDS/NIH.