Address correspondence to Edward Bertram, Department of Neurology, PO Box 800394, Medical School, University of Virginia, Charlottesville, VA, U.S.A. E-mail: email@example.com
Purpose: Mesial temporal lobe epilepsy (MTLE) is a common form of epilepsy that affects the limbic system and is associated with decreases in memory and cognitive performance. The medial prefrontal cortex (PC) in rats, which has a role in memory, is associated with and linked anatomically to the limbic system, but it is unknown if and how MTLE affects the PC.
Methods: We evoked responses in vivo in the PC by electrical stimulation of the mediodorsal (MD) and reuniens (RE) nuclei of the thalamus at several time points following status epilepticus, before and after onset of spontaneous seizures. Kindled animals were used as additional controls for the effect of seizures that were independent of epilepsy.
Results: Epileptic animals had decreased response amplitudes and significantly reduced recruiting compared to controls, whereas kindled animals showed an increase in both measures. These changes were not associated with neuronal loss in the PC, although there was significant loss in both the MD and RE in the epileptic animals.
Conclusions: There is a significant reduction in the thalamically induced evoked responses in the PCs of epileptic animals. This finding suggests that physiologic dysfunction in MTLE extends beyond primary limbic circuits into areas without overt neuronal injury.
To study the thalamic connection to the PC in epilepsy, we used trains of stimuli to induce the thalamic recruiting response, which is a progressive augmentation of the amplitude of the response wave that plateaus by the fourth stimulation in the train (Fig. 1). This response was first described in the early 1940s (Dempsey & Morison, 1942; Morison & Dempsey, 1943), and is strongly linked to midline thalamic stimulation and cortical excitation (Verzeano et al., 1953; Herkenham, 1986; Bazhenov et al., 1998). Because the PC shares connections with a variety of structures that are also networked to the midline thalamus, it was important to examine a cortical response that was strongly linked to thalamic input. The recruiting response is also a useful tool for examining the PC–midline thalamic connection because the maximal amplitude of the response in the recruiting rhythm is much larger than the response to single stimuli.
Materials and Methods
All animals were used under a protocol approved by the animal care and use committee (ACUC) of the University of Virginia.
Groups for study
We grouped our animals using predetermined time points, based on a known progression of alterations in physiology (Mangan & Bertram, 1998). We used six groups in this study:
1Naive, unstimulated controls (n = 12).
2Groups experiencing status epilepticus (SE), used at time points before evolving spontaneous seizures:
(a) 2 days (n = 5), for an acute measure following SE.
The method for creating chronic seizure models by electrical stimulation of the hippocampus, for both epileptic and kindling models, has been described previously (Lothman et al., 1990). Briefly, adult Sprague–Dawley rats (250–300 g) were implanted with a single, twisted-pair, stainless steel, bipolar electrode while under ketamine/xylazine anesthesia (50 mg/kg ket.,10 mg/kg xyl.). During all surgeries, the bite bar was set at −3.3 mm. The electrode was placed in CA1 of the mid-ventral hippocampus, (from Bregma, −5.6 mm anteroposterior (AP), +4.9 mediolateral (ML), −5 mm dorsoventral (DV)) and secured with dental acrylic and skull screws (Paxinos & Watson 1998).
Status epilepticus animals
Five days after the surgery, the animals were stimulated with a train (50 Hz peak to peak, 400 μA, 1 ms biphasic square-wave for 10 s, delivered every 11 s), after which the rats went into a period of status epilepticus for 10–14 h. They were then allowed to recover. Headsets were removed under anesthesia prior to the physiology experiments. In epileptic rats, spontaneous seizures were confirmed using 24-h video surveillance simultaneously with EEG recording. Rats having a minimum of two spontaneous seizures were considered epileptic. Data were recorded using Stellate HARMONIE-E v. 6 software (Stellate Co., Montreal, PQ, Canada). Identification of seizure activity has been described (Bertram, 1997).
Kindled rats received stimulations (50 Hz peak to peak, 400 μA, 1 ms biphasic square-wave, lasting 10 s) every hour, six times daily, every other day. Animals were considered fully kindled when they demonstrated rearing and falling behaviors (Racine Class V seizures). This protocol has been outlined in detail previously (Lothman & Williamson, 1993). Kindling until the first motor seizure required an average of 15 ± 2 stimulations, after which they received an average of 20 ± 3 additional stimulations to ensure that all animals were equally and adequately kindled. Animals were used no sooner than 48 h after their most recent kindling stimulation.
Rats were put under urethane anesthesia (1.2 g/kg, i.p.) and placed in a stereotactic frame in a Faraday cage. A twisted-pair, bipolar stainless steel electrode with a diameter of 0.56 mm, with the tips separated by 0.5 mm at an angle, was placed in the midline thalamus on a trajectory aimed to pass through the MD and RE thalamic nuclei (from Bregma: −2 mm AP, 0.5 mm ML, 4.5–7.5 mm DV, 5° angle). A glass recording pipette of approximately 1 MΩ resistance was filled with 0.9% saline solution and Fast-Green dye and connected to the amplifier through a silver–silver chloride wire. It was placed in the medial PC (from Bregma: 3.5 mm AP, 0.4 mm ML, 1.5–4 mm DV, 2° angle from vertical). This trajectory includes the anterior cingulate areas (Paxinos & Watson 1986).
Stimulations were delivered using a Winston pulse train timer (Winston Electronics, St. Louis, MO, U.S.A.), which drove a separate constant-voltage stimulus isolation unit (SIU). Stimulation was started at 15 V and then increased by 10 V until the response amplitudes plateaued, as defined by increased stimulation intensity without increased response amplitude (measured resistance across stimulating electrodes 30 kΩ, maximum calculated maximum current 2.3 mA). Only maximal responses were used for analysis. Recordings were obtained using an Axon Smartprobe, which was connected to a Cyberamp 380 amplifier and digitized with an Axon Digidata 1200 (MDS Analytical Technologies, Toronto, ON, Canada). Recordings were viewed and measured using AxoScope 8.2 software. Filtering during recordings was set to 0.5 kHz lowpass and 2 Hz highpass, with a digitizer gain of 2.
The recruiting rhythm stimulation protocol consisted of trains of stimuli of five 0.2-ms pulses with an interstimulus interval of 120 ms. Responses to 10 trains (given at 10-s intervals) were averaged for each stimulation position. The response amplitudes and latencies were measured for the first and fourth stimulation responses, as measured from the baseline at stimulus onset. Responses were measured from the baseline value occurring immediately before the stimulation to the highest-amplitude positive peak following stimulation (Fig. 1C).
We used two stimulation sites: the MD (5–5.5 mm below surface) and the RE (6.5–7 mm below surface). For each stimulation site, we dropped the recording electrode in increments of 0.5 mm in a range of 1–4 mm below the cortical surface into the medial PC. Responses of maximal amplitude, as measured at the fourth response wave in the train, were obtained for each stimulation site for each animal and used for our analysis, and the depths at which those responses were obtained were noted.
Following the experiment, the rats were perfused with 4% paraformaldehyde. Their brains were removed and placed in fixative with 20% glucose for at least 48 h. They were then frozen, sectioned, and examined to confirm the position of the electrode tracts. In cases where one or more electrodes were found to be outside the regions of interest, the data from those animals were not included in the analysis.
Because the presence or absence of seizure activity in the recording site could influence the responses and cause differences between groups, we determined if there was a difference in seizure involvement in the PCs of kindled (n = 4) and epileptic (n = 2) animals. To record seizures, twisted-pair, stainless steel, bipolar electrodes were placed in the standard position in CA1 of the hippocampus, as well as in the medial PC (from Bregma: +3.5 mm AP, +0.5 mm ML, 2–4.0 mm DV, 2° angle). Animals were allowed to recover for 5 days before recording. Recordings were made from the start of the kindling process in order to observe the evolution and progression of seizures in the PC with repeated stimulations. A minimum of two seizures were obtained for each epileptic animal, whereas a minimum of 20 were obtained for each kindled animal. The principal purpose of this data collection was to determine if PC was involved in seizure activity, and, therefore, we did not further quantify aspects of the seizures that we recorded.
The method for EEG data acquisition has been described earlier (Bertram et al., 1997). EEG recording was accompanied by simultaneous video recording. Epileptic rats were monitored continuously 24 h/day. Data from kindled animals were recorded during the times in which they were stimulated, with recordings ending 10 min after the stimulation. After recordings were obtained, brains were fixed and sectioned to confirm the correct placement of the electrodes.
To determine if cell death in the PC following SE may have a role in the changes found in our experiments, we performed a silver-staining procedure on a group of 7 day post-SE rats (n = 5) and controls (n = 4). Silver stain is believed to stain dying or injured neurons preferentially. Animals used for silver staining were not involved in electrophysiologic experiments. Seven days following SE, their brains were perfused, stored, and sectioned as described. All silver staining was done using the Neurosilver Kit by FD Neurotechnologies (Catonsville, MD, U.S.A.), following their protocol. Sections were kept from light during and after the process. Stained slides were dehydrated and coverslipped for examination.
Slides were examined qualitatively with a standard light microscope. At low magnification, regions of positive-staining neurons were identified as grayish-black areas against a golden-brown background. At higher magnification, individual neurons were identified as being positively stained if they were blackish in color, as opposed to brown. Previous studies have found no evidence for neuronal damage in our kindled animals (Bertram et al., 2001; Racine et al., 2002); therefore, we did not stain kindled brains for comparison.
Recruiting responses were elicited in the PC by stimulation of both the MD and RE nuclei. The qualitative nature of the response waveforms varied with the recording depth, but there was a consistent anatomic region in which the responses of maximal amplitudes were obtained (Table 1), which corresponded to layers III–V of the anterior cingulate gyrus, also known as cingulate areas 1 and 3 (Paxinos & Watson, 1989). These maximal electrode placements remained consistent among the groups that were subsequently studied.
Table 1. Maximum response depths and latencies in control, epileptic, and kindled animals. (A) Average depths for recording and stimulating electrodes at which maximal recruiting responses were obtained, in millimeters below surface. (B) Latencies for both the onset and peak of the response waveforms at maximum amplitude, in ms following the stimulation
Control (n = 12)
Epileptic (n = 6)
Kindled (n = 6)
Differences between groups were not statistically significant in either A or B (p > 0.05 by t-test).
A. Electrode depths (in mm)
5.1 ± 0.1
5.2 ± 0.1
5.3 ± 0.1
2.6 ± 0.2
2.2 ± 0.2
2.7 ± 0.2
6.5 ± 0.2
7.0 ± 0.1
6.7 ± 0.1
3.1 ± 0.2
2.8 ± 0.4
3.1 ± 0.2
B. Latencies (in ms)
11.6 ± 2.1
9 ± 1.4
11.1 ± 1.9
11.3 ± 1.7
8.5 ± 1.7
7.6 ± 1.0
27.11 ± 3.1
21.72 ± 1.1
25.29 ± 2.8
26.36 ± 3.4
21.13 ± 1.5
22.23 ± 1.0
Typical responses included a positive wave that increased with each subsequent stimulation until the fourth stimulation, at which point response waves plateaued at an amplitude that was, on average, three to four times greater than the amplitude after the first stimulation. The latency to the onset of this wave, as well as the latency to its peak, was consistent (Table 1). The recruiting wave was usually followed by a prolonged, lower-amplitude negative wave (see Fig. 1). Outside of our selected regional boundaries, waves of the same latency as the recruiting wave analyzed were often negative and nonrecruiting (Fig. 1A). Maximal amplitudes were obtained, normally at a stimulation intensity between 50 and 60 V (Fig. 1B).
Comparison of responses
We compared the recruiting responses of control animals to those of epileptic animals (Fig. 2A). Responses qualitatively similar to those seen in control animals were elicited at similar electrode depths, and the characteristic response latencies were also unchanged (Table 1, Fig. 1). However, first response waves from the epileptic animals were smaller than those in control responses, with the difference significant following RE stimulations (Table 2). Fourth response waves were also significantly smaller at both stimulation sites. Recruiting responses were significantly attenuated but not eliminated in epileptic animals; fourth responses were two to three times larger than first waves in that group. A comparison in the amplitude differences between the first and fourth responses, a measure of the strength of the recruiting response, was found to be significantly different for both MD and RE stimulation sites (Fig. 2B and 2C). Morphology and latency measurements were similar to controls.
Table 2. Comparison of max responses
Absolute amplitudes (mV)
Control (n = 12)
2 Day (n = 5)
1 Week (n = 6)
2 Week (n = 6)
Epileptic (n = 6)
Kindling (n = 6)
Average amplitudes (in mV) of the first and fourth response waves to MD and RE stimulation for each group of animals studied, with their standard errors. Differences across groups were significant (p < 0.001 by ANOVA, d.f. = 40, F = 6.95 for MD, 5.57 for RE). Post hoc SNK method revealed pairwise differences.
ap < 0.05 compared to control.
bp < 0.05 compared to epileptic.
3.81 ± 2.23
3.68 ± 1.26
3.38 ± 1.69
4.62 ± 1.27
2.60 ± 0.54
6.43 ± 1.26a
13.39 ± 6.72
9.18 ± 2.17
18.13 ± 3.52b
8.73 ± 1.84
5.25 ± 0.50a
24.28 ± 2.81ab
5.12 ± 0.827
4.58 ± 1.185
5.52 ± 2.37
5.15 ± 8.71
1.93 ± 0. 61a
8.85 ± 1.60b
15.94 ± 1.83
10.45 ± 2.94
18.8 ± 4.50b
10.38 ± 13.73
6.52 ± 1.02a
29.35 ± 5.26ab
To examine the evolution of this attenuation following SE, we examined recruiting responses at 2, 7, and 14 days following SE, all points prior to the onset of spontaneous seizures. At 2 and 14 days following SE, we saw a decline in the response amplitudes, whereas we saw an increase at 7 days, (Table 2, Fig. 2). All responses in those groups were similar in both morphology and latency measurements.
In contrast to the decrease in response found in epileptic animals, recruiting responses in the kindled animals showed an increase in amplitude, producing the largest recruiting responses seen in this experiment. The change was significantly different compared to both epileptic and control animals (Table 2, Fig. 2). Again, latency measurements and morphology were similar to controls (Table 1).
In order to determine if differential involvement of the PC in seizure activity could have influenced group differences, specifically between kindled and epileptic animals, we recorded spontaneous seizures in epileptic animals as well as stimulated seizures in kindled animals in the PC. In both groups we observed, although the exact seizure onset point was unclear, that there was a build-up in activity in the PC within several seconds following hippocampal onset (Fig. 3). In kindled animals, recruitment of the PC in seizure activity occurred very early in the kindling process, usually by the first or second stimulation. The recording shown in Fig. 3B,C was recorded after five stimulations, prior to the development of a motor component in the animal’s seizure, although the PC was involved by the second stimulation.
The observation of PC involvement in both kindled and epileptic animals suggests that the presence or absence of seizure activity in the PC in either model did not play a role in the differences seen between PC responses in epileptic and kindled animals, as this area was consistently involved in hippocampal seizures in both groups.
It has been shown in previous studies that cell loss occurs in the MD and RE after status epilepticus, but not in our kindling model (Bertram et al., 2001). We used silver staining to determine if there was comparable cell loss in the PC, as cell loss in this region could contribute to our results. We found that there was significant cell death specific to the MD and RE nuclei in the 7 day animals, as expected (Fig. 4). Positive staining was observed to be specific to the MD and RE nuclei within the midline thalamus, indicating that they were the nuclei most strongly affected by SE (see Fig. 4E and 4H). For each animal, cells unstained by silver were also seen in those nuclei, suggesting that cell loss in these nuclei was incomplete, and that there were still cells that could project to destinations such as the PC.
By contrast, positive neurons in the PC were rarely observed after SE (see Fig. 4A–4D). Within the same coronal slices of SE rats, positive staining was observed in the piriform and olfactory nuclei, which are areas already known to be damaged by SE (Brandt et al., 2003). We found positive scattered staining in the PC in only one of the five, 7-day post-SE animals, seen in layers I, III, and VI.
In this study, we found that there was a significant attenuation of thalamically evoked responses in the PC of epileptic rats. The attenuation evolved over a period of weeks during the post-SE period. Differences in seizure activity in the PC are not thought to be responsible for the change, as seizure activity was found in the PC in both kindled and epileptic animals. Neuronal loss in the PC was not found, although neuronal loss in the thalamus was confirmed. These findings have potential implications for prefrontal cortical function in MTLE.
The recruiting response is a thalamically induced cortical response that is specific to stimulation of the midline thalamic nuclei. Although the precise mechanisms of the recruiting response are largely unknown, the current theory is that the response is a result of the overlapping, or “nonspecific” connectivity of the midline thalamus (Morison & Dempsey, 1943; Verzeano et al., 1953; Schlag & Chaillet, 1962). The midline nuclei send monosynaptic projections to the entire medial PC, with each nucleus overlapping in their termination fields, forming a diffuse and interconnected network. These projections terminate in layers I, III, and VI of the PC (Barbas et al., 1991; Negyessy & Goldman-Rakic, 2005; Hoover & Vertes, 2007). Reciprocal connections return from the PC from layers II and V (Freedman & Cassell, 1991; Vertes, 2004). The theory is that the recruiting response is generated as this overlapping, reciprocal connectivity is able to incorporate and synchronize additional thalamic nuclei with each additional stimulation in a train. This is not to be confused with the similar “augmenting” response described in sensory and motor pathways, which is thought to be primarily cortically based (Herkenham, 1986; Castro-Alamancos & Connors, 1996). In addition, it is possible that other structures in the limbic system that share connections with both the PC and midline thalamus (such as the amygdala and hippocampus) may contribute to recruiting responses, but that possibility is, as yet, untested.
The changes we found in the PC responses are likely related to the thalamic neuronal loss, but the mechanism is unclear at this time. The altered responses changed over time, which parallels the many changes that occur following an acute neuronal injury. Cell death and recovery is an evolutionary process that occurs in several brain regions in our epilepsy model (Ben-Ari et al., 1981; Cavalheiro et al., 1982; Brandt et al., 2003; Dinocourt et al., 2003). The process involves the loss of neurons, loss of synaptic input, local reinnervation, and a spectrum of pre- and postsynaptic changes that alter individual and population excitability. The evolution of these changes have been studied extensively and in great detail in hippocampal and amygdalar circuits, among others, in epilepsy (Mangan & Bertram, 1998; Lehmann et al., 2000; Kumar & Buckmaster, 2006; Sloviter et al., 2006; Sun et al., 2007). PC neurons do not appear to die in our model of MTLE, but cell death in the midline nuclei likely causes a loss of first-line synaptic input, which, in turn, could lead to a level of denervation hypersensitivity (Koyama et al., 1993; Iizuka et al., 2002; Carlson et al., 2007; Viisanen & Perrovaara, 2007), and alterations in postsynaptic channel and receptor expression. During the subsequent weeks, reinnervation takes place and postsynaptic properties stabilize and return to an equilibrium state, causing the response to decrease. This evolution timeline would explain the rise in recruiting response that takes place after 7 days, and the subsequent decline. The exact mechanisms that cause these shifts in excitation are unknown.
The basis for the change following kindling is also unclear. As mentioned, the process of cell death that occurs in our epileptic model does not occur in our kindling model (Represa et al., 1993; Racine et al., 2002). Changesin kindled animals may include postsynaptic alterations as well as other mechanisms that allow for a better recruitment of the other thalamic nuclei that contribute to the recruiting response.
The finding that the peak latencies were not significantly different among groups in this study reflects some minor variance in the geometry of the peak from rat to rat, the slight variation in rise time to high and low amplitude peaks, and may also reflect the differences in postsynaptic properties to which we have alluded. The lack of any significant difference between onset latencies among groups suggests that latency is not a truly variable factor in observations.
The effect that the alteration in responses has on the normal function of the PC, if any, is also unclear. The PC is involved in various memory functions, including autobiographical, spatiotemporal, and working memory (Macrae et al., 2004), as well as emotional processing, executive functioning, and decision making (Barbas, 1995; Price, 2005). Lesions in the PC have given mixed but interesting results depending on the animal and the method used (Lacroix et al., 1998; Deacon et al., 2003), and abnormalities in the PC have been implicated as underlying factors in some psychiatric disorders, such as schizophrenia (Mitelman et al., 2005). Although the deficits in learning and memory that have been observed in patients with TLE (Glowinski, 1973; Viskontas et al., 2000; Voltzenlogel et al., 2006) have been rightly attributed to cellular damage in the hippocampus and other limbic structures, our results suggest that the effects of that damage may extend into the physiology, and perhaps functionality, of structures like the PC. This may even be the case in the kindled animals, in which some memory deficits have been suggested (Hanneson et al., 2001). More-detailed studies on this subject will be difficult, as it will not be easy to attribute a specific memory deficiency to the PC or any other limbic structure.
In conclusion, we have found an attenuation in PC responses to midline thalamic stimulation in epileptic rats, suggesting an impairment in their connection. This attenuation evolves following thalamic cell injury incurred by status epilepticus and may be attributable to cell loss. Other factors, including inputs to the PC from other structures and possible pre- and postsynaptic modifications may also contribute to the change in response. It is possible that this physiologic attenuation may be an indication of some form of functional deficiency in the PC, with implications for its involvement with learning and memory processes. These data reinforce our current understanding that a variety of higher-order functions require a complex circuitry, incorporating diverse groups of brain structures.
This study was funded by NIH Grant NS25605. We thank John Williamson for his excellent technical support. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
Conflict of interest: We have no conflict of interest in this study.