The effect of electrical kindling, applied twice daily in primary auditory cortex on the neural response properties and tonotopic organization in the lightly ketamine anesthetized cat is presented. Kindling refers to a highly persistent modification of brain functioning in response to repeated application of initially sub-convulsant electrical stimulation, typically in the limbic system but here in auditory cortex, which results in the development of epileptiform activity. Kindling resulted in approximately two-thirds of the animals reaching a fully generalized state in 40 stimulation sessions. Multi-unit recordings were obtained from primary auditory cortex contralateral to the kindled site. Spontaneous activity of single units in fully kindled animals showed a decrease in the mean firing rate compared to sham controls, and a reduction in the rate of burst firing as well as in the mean interspike interval in a burst as compared with normal and sham controls. A 40% enhancement of spontaneous neural synchrony, as measured by spike cross-correlation, was found. Hearing sensitivity, measured by auditory brainstem response, was not affected by the kindling sessions. A profound alteration of the tonotopic map in AI was observed with a large extent becoming tuned to similar high characteristic frequencies. The percentage of double tuned neurons was significantly increased, nevertheless the frequency-tuning curve bandwidth was on average reduced. Thus, electrical kindling resulted in substantial alterations in unit firing characteristics and reorganization of cat auditory cortex.
Kindling refers to a highly persistent modification of brain functioning in response to repeated application of electrical stimulation that results in the development and spread of seizure activity (Goddard et al., 1969) and has been previously described in the visual (Ono et al., 1981; Baba, 1982; Wada et al., 1989; Moneta and Singer, 1986), somatosensory (Majkowski et al., 1981) and motor (Fukushima et al., 1987; Kudo et al., 1997) cortex of the cat. Kindling has been shown to result in long-lasting enhancement in synaptic efficacy (Teskey et al., 1999, 2001), unit activity (Teskey and Racine, 1993; Gernert et al., 2000) and recent evidence suggests that kindling is capable of substantially altering organization in motor cortex of the rat (Teskey et al., 2002). It was reported that neocortical as well as limbic kindling resulted in reorganization in the caudal forelimb area of the motor cortex with a substantial expansion (doubling) of the cortical area capable of eliciting forelimb movement. Although their findings represent the first demonstration of cortical map reorganization experimentally induced with seizure activity they did not record unit activity. Furthermore, the effect of experimentally induced seizure activity on topographic organization has not previously been examined in sensory cortex.
The cat primary auditory cortex (AI) provides an excellent model system for examining how hyper-synchronous (seizure) activity alters neocortical sensory organization and unit firing characteristics. Cat AI exhibits a tonotopic organization in which there is an orderly caudal–rostral (low–high frequency) progression of iso-frequency bands (Merzenich et al., 1975; Reale and Imig, 1980). Cortical tonotopic maps appear to be dynamically regulated, continually altered by ongoing experience even into adulthood. Functional reorganization of the frequency map in AI has been observed in response to discrimination training (Edeline et al., 1993; Recanzone et al., 1993), sensory deafferentation (Robertson and Irvine, 1989; Rajan et al., 1993) and hearing loss induced with systemic application of ototoxic drugs (Harrison et al., 1998) or noise trauma (Eggermont and Komiya, 2000). Cat AI also presents an ideal structure for examining unit firing characteristics following kindling-induced reorganization because the naïve system has previously been described in detail and a great deal is known about it’s underlying mechanisms (Read et al., 2002; Schreiner et al., 2000).
We present the effects of kindling in AI on burst firing, frequency tuning, neural synchrony and tonotopic organization of the AI in the adult cat. This will for the first time examine how hyper-synchronous (seizure) activity alters auditory cortical organization and unit firing characteristics and will allow comparisons with reorganized auditory cortex resulting from insults to the cochlea or auditory nerve. Animals were chronically implanted with bipolar electrodes in the left primary auditory cortex and kindled. The contralateral primary auditory cortex was subsequently mapped (acute recording) under light ketamine anesthesia. We recorded multiple single-unit activity with a rectangular array of eight (4 × 2) microelectrodes positioned across the frequency map in AI. The frequency tuning properties of the individual neurons were determined with the presentation of tone pips. Recordings were also taken during spontaneous activity. We observed shorter interspike intervals during bursting, enhanced neural synchrony, more doubly tuned units and changes in the tonotopic organization of primary auditory cortex. These data have implications for human populations with epilepsy.
Materials and Methods
Chronic Electrode Implantation
The care and use of animals were approved (P88095 and BI2000-035) by the Life and Environmental Sciences Animal Care Committee of the University of Calgary, in accordance with the guidelines set by the Canadian Council for Animal Care. All animals were premedicated with 0.25 ml/kg body wt mixture of 0.1 ml of acepromazine (0.25 mg/ml) and 0.9 ml of atropine methyl nitrate (0.5 mg/ml), administered subcutaneously. Approximately 30 min later, the animals were deeply anesthetized with the administration of 25 mg/kg of ketamine hydrochloride, injected i.m. The incision site was injected subcutaneously with the local anesthetic duracaine (mixture of lidocaine hydrochloride and epinephrine, 10 mg/ml). Animals were chronically implanted with bipolar stimulation electrodes consisting of twisted Teflon-coated stainless steel wire, 178 µm in diameter (A-M Systems). The two poles of each electrode were separated by 1.0–1.5 mm providing differential recordings across cortical layers II–V. The electrodes were targeted at the midline between anterior and posterior ectosylvian sulcus (AES, PES) of the left hemisphere, in the same anterior–posterior plane (Fig. 1) or in the corpus callosum positioned just off midline in the same anterior–posterior plane as the recording electrode. The electrode for recording of evoked potentials was made of similar material as the stimulus electrode and was implanted into AI at ∼9.0 mm posterior and 21.0 mm lateral to bregma. The vertical placement of the recording and stimulating electrodes were adjusted during surgery to provide optimal evoked response amplitudes. Further details are in Teskey et al. (2002).
Following baseline recordings, the afterdischarge thresholds (ADT) were determined as the weakest current to evoke an afterdischarge (AD) of 4 s or longer. The initial current of 50 µA (peak-to-base) was applied, increasing the intensity in steps of 50 µA, and 1 min intervals between current delivery, until an AD was observed in the polygraph record. The kindling stimulation was applied daily and consisted of 1 s long trains of biphasic square wave pulses, 1.0 ms in duration, at a frequency of 60 Hz. The intensity of the stimulation was applied at 100 µA above the ADT on all subsequent kindling sessions. The animals received ∼40 sessions of kindling stimulation prior to the acute recording procedure. The EEG activity was recorded during each kindling session and the behavioral manifestation was scored according to stages 1–5 (see Racine, 1972). Acute recordings were performed at 1–45 days following the completion of kindling stimulation. The sham controls were implanted with chronic electrodes but did not receive stimulation.
Acute Surgical Procedure
Cats were premedicated with the mixture of acepromazine and atropine methyl nitrate as previously described. After ∼30 min they received an i.m. injection of 25 mg/kg of ketamine (100 mg/ml) and, if required, 20 mg/kg of sodium pentobarbital (65 mg/ml, i.m.). Lidocaine (20 mg/ml) was injected s.c. prior to incision. The tissue overlying the right temporal lobe was removed and the dura was resected to expose the area bounded by anterior and posterior ectosylvian sulci. Throughout the experiment, light anesthesia was maintained with ketamine hydrochloride (2–5 mg/kg/h) and the acepromazine/atropine methyl nitrate mixture, administered approximately every 2 h. The body temperature was monitored and maintained around 37°C with a thermostatically controlled blanket. Following the experiment, the animal was sacrificed with an overdose of sodium pentobarbital.
Acoustical Stimulation and Peripheral Threshold Estimation.
Following surgery, the animal was placed in a sound-attenuated room with the head secured to a post on the vibration isolation frame. The sound-treated room was made anechoic for frequencies above 625 Hz with acoustic wedges (Sonex 3″) covering walls and exposed portions of the vibration isolation frame, equipment and floor. Prior to acute recordings, peripheral hearing sensitivity was determined from auditory brainstem response (ABR) thresholds at frequencies of 2, 3, 4, 6, 10, 12, 16 and 20 kHz (for details, see Seki and Eggermont, 2002).
Neural activity was recorded under spontaneous conditions and in response to auditory stimulation with tone pips. The characteristic frequency (CF) and tuning properties of the individual neurons were determined with the presentation of tone pips. The gamma-shaped envelope tone pips were presented randomly at a frequency of once per second and a fixed sound level (Eggermont, 1996). The total tone pip duration was 100 ms and between 1/2 amplitude points was 15 ms. The 81 different frequencies were equally spaced logarithmically between 625 and 20 kHz (or between 1.25 and 40 kHz), such that 16 frequencies were present for each octave.
Recording and Spike Separation Procedure
Multiple single-units were recorded, across the extent of AI with one or two rectangular electrode arrays, with impedances of ∼2 MΩ for each electrode (Frederic Haer Corp.). The electrodes were arranged in a 4 × 2 configuration with inter-electrode distance within rows and columns equal to 0.5 mm. The electrode array was manually advanced using a Narishige M101 hydraulic microdrive to a depth between 600 and 1200 µm, corresponding to layers III/IV. The electrode arrays were oriented approximately perpendicular to the isofrequency contours; recordings started generally at the most rostral end of AI and gradually moving more caudally. The cortical fields were identified by location, reversal of the CF gradients, different response patterns (latency and burst activity) and shape of the frequency tuning curves. Typically, a mapping session lasted for ∼12 h.
The signals were amplified between 300 and 5 kHz and processed by a DataWave multi-channel data acquisition system. Spike sorting was done off line using a semi-automated procedure based on principal component analysis (Eggermont et al., 1983) implemented in MATLAB®. The separated single-unit spike trains extracted from the multi-unit (MU) recording, that all had similar response properties, were added again to form a multiple single-unit spike train. Cross-correlation analysis was only done between neural activities recorded on separate electrodes.
The spontaneous activity of single neurons was obtained from 15 min long recordings during silence and was analyzed in terms of the mean firing rate, coefficient of variation and properties of burst firing, using the Poisson surprise method (Legendy and Salcman, 1985; Valentine and Eggermont, 2001). Only those units that exhibited a minimum spontaneous firing rate of 0.2 spikes/s (180 spikes in 15 min of silence) were considered for the burst analysis.
The Poisson-surprise measure provides an evaluation of how improbable it is that the burst is a chance occurrence. The surprise measure (as implemented in the Stranger® software package; Biographics Inc.) is defined as the negative natural logarithm of the probability of the occurrence of a burst consisting of a given number of spikes in a particular time interval when compared to a Poisson spike train with the same mean firing rate. Only spike sequences having a surprise value of at least 10 were considered as bursts, this correspond to a P < 0.00005, which is very conservative. Bursts were quantified by burst rate (bursts/500 spikes), burst duration (s), number of spikes per burst and mean ISI in burst compared to the overall spike train.
All nonstationary recordings, i.e. those that showed non-random changes in spontaneous firing rate across the 15 min recording sessions, were removed from the analysis, as they may introduce artificial correlations (Brody, 1999). Before cross-correlation functions were calculated, common spikes (occurring within ± 150 µs on different electrode) were deleted from all the records. This is to avoid potential cross-talk from capacitive coupling between electrodes and connecting wires (Zhu et al., 2002). In some cross-correlograms the result of these deletions shows up as a slight dip at or close to lag time zero. We will use upper case R to refer to the peak cross-correlation coefficients and lower case r to indicate correlation coefficients resulting from statistical tests.
Cross-correlation analyses provide an indication of the degree of synchronous activity between single neurons and has been used extensively to examine synaptic interactions among elements of the neural population (Abeles, 1991; Gerstein, 2000). Activity recorded from neurons at different electrode placements were analyzed using cross-correlograms for lead/lag times of 100 ms and a 2 ms bin width. The cross-correlation coefficients were calculated as described in Eggermont (1992).
Evidence from neural modeling suggests that interpretation of neural correlations from unresolved multi-unit recordings may be ambiguous, as it differs from correlations for single-units (Bedenbaugh and Gerstein, 1997; Gerstein, 2000). However, changes in single-unit correlation strengths will be accompanied by comparable changes in the correlation between multi-unit activity (Eggermont, 2000).
The frequency tuning curves were obtained from the time-locked responses to the tone pips as represented by the peak values of the post-stimulus-time histograms (PSTH) as described in Noreña and Eggermont (2002). Due to the considerable level of spontaneous activity and its dependence of stimulus level, the frequency tuning curves were defined at 25% of maximum peak spike counts. The CF was defined as the lowest-frequency sound level combination of the frequency-tuning curve. The frequency-tuning curve bandwidth (octaves) and peak amplitude (spikes/stimuli) were measured at 20 dB above threshold.
Tonotopic Map Estimates
The maps were determined from a series of electrode penetrations made in AI. In a number of animals, recordings included activity from anterior auditory field (AAF), dorsal posterior field (DP) or secondary auditory cortex (AII) neurons. The electrode penetrations of individual animals were plotted on a photograph of the cortical surface, with AES and PES indicated to generally locate AI. The cortical regions covered by recording sites in the kindled and sham control group to a great extent overlapped those made from normal control animals. The electrode positions and CF values were plotted onto a photograph of the exposed cortex surface.
All statistical analyses were performed using Statview® (SAS Institute). The treatment group comparisons were made using between-subjects analysis of variance (ANOVA), with Scheffe post hoc tests (P < 0.05). Only the results of the post hoc tests are reported.
Results were obtained for 15 adult cats. Eight cats were kindled in contralateral AI cortex, relative to the acute recording side, with the stimulus electrode approximately midway between the tips of AES and PES. Two cats were kindled in the corpus callosum, In a sample of 56 control cats, collected over the last few years, the point midway between the tips of PES and AES had a mean characteristic frequency (CF) of 10 kHz with a range from 4.5 to 20 kHz. These CFs were log-normal distributed with a standard deviation of 0.5 octave. Five cats were implanted with the stimulation and EEG recording electrodes but not kindled (sham controls). Normative control recordings represent the pre-treatment data from 25 animals that were used in previous studies investigating the effects of salicylate and quinine (Eggermont and Kenmochi, 1998) and acute noise trauma (Kimura and Eggermont, 1999) on spontaneous firing rates.
As the changes in our dependent measures could vary as a function of delay from kindling completion to the acute mapping, all treatment group comparisons were restricted to delays of 1 day. The data from the partial kindling group were not included in the statistical analysis of the treatment groups, as these animals were all acutely mapped at long delays.
Repeated electrical stimulation of the cat auditory cortex resulted in a progression in electrographic activity (with increases in spike amplitude, frequency and complexity) and the corresponding seizure severity in most animals. The baseline afterdischarge thresholds (ADT) ranged from 100 to 800 µA (mean of 370 µA). Figure 2 illustrates the increase in AD duration with repeated stimulation from the initial stimulation (Fig. 2A) to the 20th stimulation session (Fig. 2C) for a representative individual animal. The increase in mean AD duration (for all kindling animals) and corresponding average seizure severity are summarized in Figure 2B. The behavioral manifestations of the seizure activity became progressively more severe with the repeated occurrence of AD and were classified into stages according to Racine (1972). We report the failure of kindling progression in approximately one-third of the cats, consistent with findings in the rat auditory cortex (Cain, 1982; Seidel and Corcoran, 1986).
Spontaneous activity of single units in fully kindled animals showed a decrease in the mean firing rate compared to sham controls and a reduction in the rate of burst firing as well as in the mean interspike interval in a burst as compared with normal and sham controls. Multiple single-unit activity was recorded simultaneously from 8 or 16 microelectrodes in AI (and occasionally in AAF, AII or DP) from fully kindled animals and sham controls. The analysis of spontaneous activity was restricted to sorted units recorded from AI. The mean baseline rate of spontaneous firing for the 198 single units included in the normative control group was 1.3 ± 1.2 spikes/s. The spontaneous mean firing rate for the fully kindled (n = 249, 2.1 ± 2.4 spikes/s) was significantly lower than for the sham controls (n = 218, 3.2 ± 3.8 spikes/s, P < 0.001). However, relative to normal controls the kindled animals (P < 0.01) and sham controls (P < 0.0001) exhibited an elevated firing rate. In kindled cats, compared to naïve and sham controls, spike bursts did not occur as frequently, but the spikes in a burst did have much shorter mean interspike intervals (Fig. 3). This difference is further emphasized by the median interspike intervals, which are 36, 32 and 16 ms, respectively, for naïve controls, sham controls and kindled cats.
Spontaneous neural synchrony in AI, as measured by spike cross-correlation, was increased by 40% and was more widespread compared to controls. Cross-correlograms were determined for the spontaneous activity of 778 distant electrode pair combinations recorded from kindled, sham control and naïve control cats. All the multiple single-unit pairs included in the analysis had peak-correlation coefficients that were significantly different at the 99% confidence interval, typically with a peak close to zero lag time. Only cross-correlations between AI sites were considered in the analysis. An example of the multiple single-unit cross-correlogram in a kindled subject for a representative electrode array recording is shown in Figure 4. Two time bases are used for the correlograms: above the diagonal indicated by the full line the lag/lead times are 505 ms (10 ms bins) to be able to see long term effects, below the diagonal the lag/lead times are 101 ms (2 ms bins) and those data will be used in all subsequent analyses. The panels along the diagonal show the auto-correlograms for 10 ms bins. In the upper left of each box the numbers indicate the electrode pairs, followed by the peak correlation coefficient R, the peak coincident firing rate r and the lag time of the peak of the correlogram. For easier reference, the numbers at the left and below the panels indicate the electrode positions as in the top right insert of Figure 6.
Unit pairs recorded from the fully kindling group exhibited significantly higher cross-correlation coefficients (n = 296, 0.045 ± 0.033) as compared with the sham (n = 210, 0.031 ± 0.030, P < 0.01) and naïve controls (n = 48, 0.027 ± 0.040, P < 0.005; Fig. 5A). A scatterplot of the cross-correlation coefficients as a function of the geometric mean firing rate (Fig. 5B) indicated that the increased synchrony was partially related to the increase in firing rate for the sham controls (P < 0.0001), but not for the fully kindled animals or naïve controls. One observes that, especially for low firing rate neurons, the increase in R was considerable for the kindled group. R decreased with the number of kindling stimulus sessions but increased with number of stage 5 seizures.
In general, kindling increased not only the average value of R, but also produced a larger number of high correlations in the spike firing recorded on different electrodes in an array. An example (Fig. 6) shows correlation maps in a naïve control and the fully kindled cat. The maps represent, above the diagonal for the normal control and below the diagonal for the kindled cat, the 28 different pair-wise peak cross-correlation coefficients between the firings on eight electrodes. The map for the kindled cat represents the peak values for the correlograms shown in Figure 4 below the diagonal, using the same layout. Along the diagonal, the peak auto-correlation values are set here at 0.05 so as not to interfere with the scale of the color map to visualize the cross-correlation. In the control cat, a high correlation was found between neighboring electrodes 7 and 8 (see insert in top right for numbering), whereas for other pairs the values were mostly <0.07. In the kindled animal, most correlations were >0.07. Thus, although the largest cross-correlation coefficient was not different for these two conditions, a larger number of high correlations was found in the kindled animal, suggesting a more widespread neural synchrony under spontaneous conditions.
Frequency Tuning Properties
The percentage of double tuned neurons was significantly increased in kindled animals; nevertheless, the frequency-tuning curve bandwidth was on average reduced. The sensitivity of peripheral hearing was established using the ABR threshold for all kindled and sham control animals prior to the acute mapping procedure and compared with normative control data that included a total of 14 animals. The ABR threshold shift in the frequency range of 2–32 kHz, compared to naïve controls, was not frequency dependent, and did not differ in kindled (7.8 ± 3.8 dB) or sham (6.0 ± 4.5 dB) animals.
Representative dot rasters in response to tone pips of 25 dB SPL with frequencies between 1.25 and 40 kHz (vertical axis) and for a 100 ms time window after tone pip onset (horizontal axis), are shown in Figure 7 for multiple single-units recorded on one electrode array in AI of a naïve control (top section) and of a kindled cat (bottom section). The electrode arrays were oriented parallel to the midline. Low electrode numbers indicate more anterior positions. One observes that in the kindled cat the units are all tuned to the same broad frequency range and that most electrodes show substantial spontaneous activity. For the control animal, in contrast, higher CF values are found for more anterior positions in AI.
The rate–frequency–sound-level contours for MU clusters recorded from auditory cortex in kindled cats were variable in shape. The 25% contour was used as the frequency-tuning curve and these ranged from very narrow (V-shaped), to very broad, or multiply tuned (Fig. 8). The double tuning for MU clusters was generally also found for the contributing single units. In only one case, the two single units had quite different CFs, resulting in double tuned MU frequency-response areas. In the kindled animals there was a greater tendency for double-tuned multiple single-units (29%), than for the normal controls where the percentage of double tuned neurons in cat AI is nominally only ∼8% (Sutter and Schreiner, 1991; Eggermont, 1998) and for sham controls where it was 14%.
The excitatory frequency-tuning curve bandwidth (BW20dB), taken 20 dB above threshold at CF, is indicated for the tuning curve examples in Figure 8 by a horizontal bar. In the following comparison, all tuning curve types were included. The BW20dB was significantly lower in fully (P < 0.0001) kindled animals in comparison with normal controls but, however, it did not differ from the sham control group (Fig. 9A). The altered frequency tuning of the kindling animals is further shown in the scatterplots of the BW20dB at 20 dB above threshold as a function of CF (Fig. 9B). The distribution of BW20dB on a log scale is shown in Figure 9C–E, for control, sham and kindled animals. The geometric means are indicated by the peak value of the best fitting normal distribution and this was lower for the sham group largely because of some very small values of BW20dB, whereas for the kindled animals the range of values was the same as in normal controls but with a lower mean value. The tendency for greater BW20dB at higher CFs observed in the normal and sham controls was absent in the kindled animals. In the kindling group, the multiple single-units recorded tended to exhibit smaller BW20dBs and lower thresholds at higher CFs than normal controls.
The fully kindled group exhibited significantly (P < 0.005) increased multi-unit thresholds relative to sham controls (Fig. 10A). However, compared with the naïve controls the threshold levels were decreased in the sham controls (P < 0.05) and increased in the fully kindled (P < 0.05) group. The scatterplot of the threshold at CF as a function of CF (Fig. 10B) indicates a significant (P < 0.001) relationship for the fully kindling group, showing a tendency for lower thresholds at higher CFs.
Tonotopic Map Changes
Substantial changes in the tonotopic maps of AI were observed in kindled versus control cats. The cortical regions covered by recording sites in the kindled and sham control group to a great extent overlapped those made from normal control animals. The CF could be unambiguously assigned for 238 multiple single-unit clusters in the kindled group, for which the CFs ranged from 0.65 to 38.30 kHz (mean = 18.38 kHz). For the sham controls, a CF could be assigned to 102 multiple single-neuron clusters, and ranged from 1.42 to 33.64 kHz (mean = 12.75 kHz). The top panels in Figure 11 demonstrate the tonotopic organization of AI in four sham control animals. Shown is a photograph of the cortical surface with the PES and AES indicated when visible. The dots represent recording positions, color indicates the frequency range of the CF. If no value (no color) was indicated the neurons at that site did not respond sufficiently to assess the CF. In these four examples, we see a gradually progression from low to high CFs from the PES toward the AES. In the top-left example, the highest CF was <25 kHz and higher CFs were likely to be found in the banks of the AES which could not be sampled with our array electrode. In the top-right and bottom-left examples the CF range covered the range from 2.5 kHz to ∼35 kHz. In the bottom-right example, the highest CF found between the sulci was 20 kHz.
In the kindled animals, the maps in AI typically exhibited the lack of a tonotopic gradient, with large portions of AI tuned to a relatively narrow range of CFs. The frequency maps for four fully kindled animals are shown in bottom panels of Figure 11. Due to the vastly increased size of the blood vessels, a finding common in all fully kindled cats and likely indicating an increased cortical energy demand during the seizures, the mapping is less dense than in the sham controls. In the top-left example, the sampled sites between the PES and AES had (with one exception) CFs >20 kHz, whereas in the dorsal-posterior part and anterior to the AES the CFs were (with one exception) <5 kHz. In the top-right example, the CFs between the PES and AES are (with one exception) >15 kHz. The animal in the bottom-left shows the most normal map with a 10–20 kHz cluster posterior and a 20–35 kHz cluster more anterior. Two low CFs were found ventrally and likely in AII. The bottom-right example shows a map of an animal kindled in the corpus callosum, most CFs between the PES and AES are in the 10–15 kHz range. These findings suggest either local or global changes in the frequency organization of the AI in kindled animals.
The results of the unit recordings following kindling and, relative to results in the sham controls, can be summarized as follows. Spontaneous activity was reduced in the fully kindled group. The rate of bursting decreased but the ISI within a burst was also significantly reduced. Full kindling resulted in ∼40% enhanced spontaneous neural synchrony as estimated by the cross-correlation coefficient (R). The fully kindled cats exhibited higher unit thresholds at CF. Tonotopic maps in kindled cats were substantially changed and generally showed an absence of a caudo-rostral frequency gradient and tended to show only a limited and narrow range of CFs.
Extent of Stimulation Site Effects
The average current used in the kindling stimulation was 470 µA and this is likely to activate neurons in a large cortical volume. For intra-cortical microstimulation, a current strength of 10 µA combined with the pulse duration of 1.65 ms per phase results in a charge of 16.5 nC/phase. For this charge, one expects that neurons ∼350 µm away from the stimulation site will be at threshold for activation (Ronner and Lee, 1983); this corresponds to an activation volume of ∼0.2 mm3. For the kindling stimulus, the pulse duration per phase is ∼8 ms and for a 470 µA current this corresponds to a charge ∼3.8 µC/phase. This is a 228 times larger charge per phase compared to the intra-cortical microstimulation and one expects the range of the effect on neurons to be increased by the cubic root thereof which equals 6. Consequently, the activated volume would have a radius of 350 µm × 6 = 2.1 mm. This would cover about half of AI.
As shown previously (Khurgel et al., 1995), no degenerative changes specific to the kindling process were observed in brain sections from kindled animals and processed with the degeneration-sensitive cupric silver stain. Thus stimulation and the development of seizures do not depend on nor result in neural degeneration.
Comparison between Normal Controls and Sham Controls
Some differences were found between the naïve and sham controls, specifically the spontaneous firing rate and number of spikes per burst was higher in sham controls. However, burst rate, burst duration and ISI in a burst were not significantly different between the two control groups and neither was the peak cross-correlation coefficient. The only comparison that gave rise to a different interpretation of the effect of kindling was that of spontaneous firing rate as the number of spikes per burst was lower in kindled animals compared to either naïve or sham controls. For spontaneous firing rate, a higher value was found in kindled animals compared to naïve controls and a lower one compared to sham controls. As argued in the next section, implantation may over time produce changes in the surrounding cortex that evidently may have transferred to contralateral cortex.
Spontaneous firing rates were significantly reduced in the fully kindled animals compared to the sham controls. However, they were increased compared to normal controls. Chronic electrode implantation itself is capable of inducing neurochemical and neurophysiological alterations (Loscher et al., 1993, 1995; Niespodziany et al., 1999) that may explain the higher spontaneous activity recorded from sham control animals that underwent chronic electrode implantation but did not receive stimulation. As a consequence it is probably best to conclude that kindling reduces spontaneous activity in auditory cortex. There are few reports of changes in spontaneous activity of sensory cortex resulting from direct activation, but all report increased spontaneous discharge. Increased spontaneous discharge rates (from 2–3 to 17–18 sp/s) have been observed in piriform and perirhinal cortex following amygdala kindling in the rat, utilizing the awake freely moving preparation (Teskey and Racine, 1993). Gernert et al. (2000) provide further evidence that full amygdala kindling results in persistent increases in spontaneous activity of extracellularly recorded single units in the piriform cortex of anesthetized (fentanyl and methohexal) rats (at a delay of at least 5 weeks after the last seizure). Further, neurons recorded in piriform cortex of kindled rats exhibited a significantly lower responsiveness to the excitatory effects of glutamate than naïve controls, as did the sham controls.
Kindling reduced the rate of burst firing compared to controls but the mean ISI within a burst became substantially shorter in kindled cats (mean = 30 ms, median = 16 ms, mode = 5 ms) compared to naïve and sham controls (mean = 52 and 47 ms, median = 36 and 32 ms, mode = 10 and 15 ms). This considerably shorter ISI suggests that activation of the recipient neurons from these cortical neurons would be more reliable than for both groups of controls because of the integration of the resulting EPSP’s (Lisman, 1997) and could have been instrumental in the reorganization of the tonotopic map. So despite the slight reduction in the burst rate by ∼10% compared to controls, the bursting can be considered more effective. Teskey and Racine (1993) found that following kindling, ∼30% of neurons recorded exhibited an apparent increase in burst firing activity, with units often firing in doublets or triplets. The Poisson-surprise burst detection procedure would not consider spike doublets as bursts and would favor burst of more than three spikes so this could explain our different findings in burst rate. However, the considerably shortened ISIs in bursts does predict the occurrence of more spike doublets.
Kindling increased the cross-correlation coefficient (R) between multiple single-unit pairs recorded from AI by ∼40%, in comparison to both naïve and sham controls. For the sham controls, R was significantly related to the geometric mean firing rate of the unit-pairs, but not for the fully kindled animals or the normal controls. Evidence from auditory (Eggermont, 1994; Brosch and Schreiner, 1999) and visual cortex (Engel et al., 1991; T’so , 1991; Lowel and Singer, 1992) indicates that cells with common response properties are more likely to exhibit significantly correlated activity even when anatomically distant. The very similar CFs for large parts of AI in kindled animals may similarly provide the basis for the enhanced correlated activity.
Frequency Tuning Curves and Tonotopic Map Reorganization
Kindling induced alterations in the tonotopic organization of AI, or at least in that part of cortex located between PES and AES (which is normally AI) as the normal procedure to define the extent of AI from the reversal of the frequency gradient fails here. As was shown in Figure 11B, in kindled animals there is no normal tonotopic gradient in this part of auditory cortex. To our knowledge this research represents the first demonstration of representational plasticity induced with kindling in sensory cortex. Generally, a large extent of AI was narrowly tuned around one or two characteristic frequency values. In general those frequencies were in the range of 15–35 kHz, depending on the animal. These frequencies are distinctively higher than would be expected from the estimated stimulation site mean CF point (=10 kHz, range 4.5–20 kHz, 56 cats) or the midway CFs for the sham controls (Fig. 11A; range 5–20 kHz, five cats). The tonotopic organization in the sham controls was normal.
The frequency tuning properties were altered in the kindled animals. There was a tendency for the multiple single-units recorded from AI in the kindled cats to exhibit a reduction in BW20dB at 20 dB above threshold and in the threshold at higher CFs compared to normal controls. In AI there are normally a high percentage of neurons that express narrow V-shaped, tuning curves. In the kindled animals there was a greater tendency for double-tuned multiple single-units (29%), than for the normal controls where the percentage of double tuned neurons in cat AI is nominally only ∼8% (Sutter and Schreiner, 1991; Eggermont, 1998) and for sham controls where it was 14%. This likely represents the result of the competition between thalamocortical afferents that provide the normal map tuning and the horizontal fibers that convey the CF-range imposed by the kindling procedure.
The effect of experimentally induced seizure activity on functional organization has previously been described in motor cortex of the rat (Teskey et al., 2002). Following multiple kindling sessions (to either the corpus callosum or amygdala), the caudal forelimb area of the motor cortex was mapped utilizing intra-cortical microstimulation (ICMS) techniques. They found that representation of the caudal forelimb area was nearly doubled following kindling with specific changes to the wrist and digit area, while the adjacent (elbow/shoulder) region remained unaltered. The topographic reorganization observed in kindled animals was accompanied by increases in synaptic strength, as measured by an increase in area of the late excitatory component of the neocortical evoked potential. They observed no change in the threshold to elicit a motor response and with increasing ICMS intensities more dual site responses were observed in kindled animals. This dual responsiveness observed in motor cortex following kindling may be analogous to the double frequency tuning we observed in cat auditory cortex.
It has been well documented that deprivation, injury and experience in both developing and adult animals can modify sensory topographic maps. Moderate pure tone trauma in juvenile cats results in profound cortical reorganization of frequency map in adulthood (Eggermont and Komiya, 2000). Even mild cochlear high-frequency hearing loss (<20 dB threshold shift) that does not produce changes in cortical tonotopic map, may result in near total loss of surround inhibition, as evidenced by an increase in the BW of excitatory frequency tuning curves and a decrease in percentage of nonmonotonic units (Rajan, 1998). A recent study (Snyder et al., 2000) showed that an acute lesion of the spiral ganglion induces an immediate and dramatic modification of the tonotopic map in the inferior colliculus (IC) similar to that observed in AI after acute cochlear damage by noise exposure (Noreña et al., 2003). These changes suggest that unmasking of new excitatory inputs is at the basis of the observed map changes. Thus, no additional mechanisms beyond the change in the balance between excitation and inhibition need be involved in the tonotopic reorganization after a hearing loss. The changes in tonotopic organization of auditory cortex observed following kindling bear some similarity to those following a peripheral insult by showing increased neural synchrony and increase in the number of multi peaked tuning curves (Noreña et al., 2003). The changes following kindling may thus result from a similar mechanism, i.e. change in the balance between excitation and inhibition, as those following peripheral damage to the auditory system (either by cochlear lesions or noise trauma). However, long-standing cochlear high-frequency hearing losses do not show multi-peaked tuning curves (Rajan et al., 1993; Eggermont and Komiya, 2000), so multipeaked tuning curves may be specific to localized hearing losses (Seki and Eggermont, 2002) or strong gradients in the balance between excitation and inhibition.
The Path of Tetanic Stimulation to Kindling and to Tonotopic Map Changes
Kindling stimulation in auditory cortex is likely to produce epileptiform activity and the associated behavioral changes via its centrifugal connections to subcortical structures, because the auditory thalamus is directly connected to the amygdala (Woodson et al., 2000). Effects on the contralateral side from the stimulation electrodes may also be directly transferred by the corpus callosum. Theoretically, electrical stimulation might, independent of epileptiform activity, also produce changes in the tonotopic map of the auditory cortex and, via centrifugal connections, also in the auditory thalamus and inferior colliculus (Yan and Suga, 1998). In turn, these subcortical changes would affect the tonotopic map in auditory cortex. However, the electrical stimulation in itself does not appear to induce the observed changes in neural synchrony because R decreased with number of kindling stimulus sessions but increased with number of stage 5 seizures. Because neural synchrony, even under spontaneous conditions, depends strongly on the amount of overlap of frequency-tuning, reflecting the common input from thalamic afferents (Eggermont, 1992, 1994, 2000), it is likely that the changes in the tonotopic map are also produced by the convulsions and not by the stimulation itself. Thus, convulsive brain activity that becomes generalized and propagating from neocortex to subcortical and brainstem structures, is potentially also responsible for the observed alteration in neural synchrony and in the tonotopic organization of contralateral auditory cortex.
While the neural mechanisms underlying representational plasticity are not fully known, it is presumed to occur, in part, through enhanced synaptic efficacy (Hess and Donoghue, 1994); however, unmasking of excitatory inputs through down-regulation of inhibition also plays a significant role (Calford, 2002). A Hebbian explanation would suggest that cells, or systems of cells, will tend to become associated while coincidentally active, such that activity in one cell or system facilitates activity in the other. Kindling results in nearly simultaneous activation of all neural elements surrounding the stimulation site and thus generates the type of temporally coincident activity thought necessary for altering neuronal membership of a functional group. Kindling in cat auditory cortex clearly induced activity in neurons across the extent of AI to become highly synchronous — as evidenced by increased R values — and express similar frequency–response properties.
It has been hypothesized that kindling stimulation in cortex produces an accumulation of glutamate. This may trigger altered NMDA channel functions and long-lasting changes in synaptic efficacy of long-range horizontal connections (Racine et al., 1995; Chapman, 1998). Kindling has also been associated with the recruitment of previously ‘silent’ NMDA receptors (Mody and Heinemann, 1987; Mody et al., 1988), which is reflected in prolonged channel openings, bursts and clusters of NMDA activity (Kohr et al., 1993). Increased glutamate agonist potency observed following kindling (Kohr and Mody, 1994), is likely mediated by a reduction in the affinity of channel pore to Mg2+ and enhanced phosphorylation and lasts as long as 28 days post final seizure after kindling cessation (Kohr et al., 1993). Kindling is also associated with a loss in GABAergic inhibition (Lopes da Silva et al., 1995) but this would suggest increased spontaneous activity, which was not observed. Lower spontaneous activity and narrower frequency tuning curves would require increased GABAergic activity. Kindling resulted in the change of CFs, putatively toward that of the stimulation site or its contralateral mirror site, but with enhanced lateral inhibition as suggested by the reduction in frequency–tuning curve bandwidth.
In cat primary auditory cortex horizontal connections generally follow a course parallel to the iso-frequency contours (Read et al., 2001) but others project in directions orthogonal to the iso-frequency contours and may even extend as far as anterior and posterior auditory fields (Wallace et al., 1991). Strengthening of the horizontal connections after kindling may explain the larger percentage of double-tuned frequency tuning curves that we observed. If these horizontal connections become stronger than those of the specific thalamic inputs, they could elicit an outward migration of glutamate hyperactivity and subsequent synaptic changes. This process would likely stop when all the neurons in the stimulated cortical area (and its projection region) were combined into one large neural assembly.
This investigation was supported by grants from the Alberta Heritage Foundation for Medical Research, the Neuroscience Canada Foundation, the Natural Sciences and Engineering Research Council of Canada and by the Campbell McLaurin Chair for Hearing Deficiencies. Makiko Kimura, Hisashi Komiya, Satoshi Seki, Arnaud Noreña and Marie Monfils assisted with the data collection. Greg Shaw provided programming assistance.