Intracranial potentials correlated with an event-related potential, P300, in the cat.

Intracranial recordings of long-latency, event-related potentials were obtained from paralyzed, artificially respirated cats. A modified oddball paradigm was employed in which cats were presented with a randomized series of two tones, a 'frequent' 4 kHz stimulus and a 'rare' 1 kHz stimulus. A tail shock was administered 700 ms after onset of the rare tone. Under these circumstances the stimulus elicited a positive component at the vertex similar to the human P300. Intracranial potentials associated with the rare tone usually manifested components of greater amplitude than did potentials associated with the frequent tone. A positive component occurring in latency between 200 and 350 ms only accompanied the presentation of the rare stimulus. The P300 component, which was positive at the dura, appeared as a negative component within a few millimeters of the surface over a wide area of the marginal and suprasylvian gyri. Changing the probability of the rare stimulus resulted in a reduction in the amplitudes of both the intracranial negative component and the P300 recorded at the skull. Components of large amplitude associated with the rare stimulus were obtained from the region of the hippocampus. These components reversed polarity, sometimes more than once, as the electrode was advanced. Substantial latency differences were often observed between the P300 recorded at the skull and P300-like intracranial components associated with the rare stimulus. These results suggest that the cortices of the marginal and suprasylvian gyri and the hippocampal region contribute to the generation of the cat P300.


INTRODUCTION
The P300 is a late positive component of the averaged event-related potential recorded from the scalp of humans that reflects the expectancies of the subject rather than the physical parameters of the stimulus 6,28. These contingencies have led to the designation of the P300 as an 'endogenous component' of event-related potentials29. The amplitude of the P300 is inversely proportional to the global probability of the stimulus 5.14. 22 24,26 whereas its latency is related, in part, to task difficulty 9AI,25,26 and subject age 12.
Typically, subjects are asked to count rare stimuli, which are randomly intermixed with a series of frequent stinmli. The rare stimulus evokes the P30(I.
Since this component is best elicited when the subject is engaged in the task, the P300 is believed to reflect cognitive events of information processing. Very little, however, is known of the neural structures involved in generating the P300 since intracranial recordings and neuroanatomical identification of the recording sites have only infrequently been employed in humans 13,15,27.32,34. Models of the P300 with animals conditoned to respond to the rare stimulus have been developed 2,3,7A°,16AS,21` but there are few reports of P300like potentials recorded from within the brain 10A~. In this study a cat model, developed earlier in our laboratory 7`s, was employed. Cats exposed to a classical adversive conditioning paradigm showed a late positive component which had an average latency at the midline of the skull of 260 ms. This component's amplitude was large when the signal was task relevant, its amplitude varied inversely with stimulus probability, and it could be evoked by stimuli of different modalities 30, characteristics similar to the P300 in humans. The present study in the cat measures the potentials at several intracranial locations at the time of the skull P300 event to help locate and characterize neural generators of this animal analogue of the human P300.

Surgi~'al preparation
Four female cats weighing from 3.3 to 4.3 kg were used in this study. Otoscopic examination verified that each animal was free from ear lice and that the tympanum was clearly visible. Surgical preparation occurred in two stages. Initially under general anesthesia a headholder was implanted, and a stainless steel screw recording electrode placed either 1 or 11 mm behind the bregma to serve as a fixed control recording site. After the development of a P3011 component during training a second surgery was performed. The animals were re-anesthetized and from one to five cannulas were implanted in the right side of the skull. The cannulas were constructed from 14 gauge stainless steel tubing. The dura was not penetrated in this procedure, and sterile bone wax was packed into the cannulas to reduce refection. Fig. I shows the placement of cannulas relative to the cortical surface. All surgery was performed under sodium pentobarbital anesthesia (40 mg/kg). Cats were allowed at least one week to recover before training or electrophysiological data recording connnenccd.

General procedures
The cats were paralyzed with galleminc triethiodide (10 mg/kg), intubated, and artificially respirated, and then supported on a piece of foam rubber. Expired carbon dioxide was maintained near 4%. Body temperature was monitored by a rectal thermometer and maintained by a circulating water heating pad, Experiments were carried out in a doublewalled sound attenuating chamber with the cat's head held in place by the implanted headholder. The eyelids of the left eve were kept opell i~ ~i wife k~(,!~ and a pupillometer positioned m tronl ~,~ the e~c t~ measure pupillary dilation. The pupithmletcr, described elsewhere ~, measured the ;_ltll()tti~ of light reflected from the iris. To prevent drying ni the ext, thin layer ofIerramycm eve ointment x~ applied Auditory stimuli consisted of I kliz told 4 kHz tones, 50 ms in duration, with a rise tim~:~ o( 5 ms and an intensity of 65 dB SPI~. Stimuli were delivered through Beyer earphones coupled to the cat's left ear bv a hollow plastic tube. Behavioral reinforcement was produced by a constant current eleclrical stimulus administered to the tail 7t)(I ms after onset of the 1 kHz tone. The electrical stimulus consisted of a 300 ms train of 15 pulses (2 mA, 2 ms) delivered through a pair of ring electrodes placed 2 cm apart on a shaved portion of the cat's tail.

Behavioral paradigm
A variation of the 'oddball paradigm" employed m human P300 studies was utilized ( Fig. 2A). Cats were presented with a random series of two tones, a 'frequent" 4 kHz stimulus and a 'rare" l kHz stimulus at two-second intervals. The electrical stimulus to the tail was administered 700 ms after onset of the 4 kHz tone. This procedure led to a conditioning of a pupillary dilation, which began about 300 ms after onset of the 4 kHz tone. When conditioning is established, the rare stimulus elicits a positive potential at the vertex similar to the human P3007,~°. The pupil response served as a 'behavioral' measure of the animal's conditioning to the Fare 4 kHz stimulus.
Training began with an habituation phase during which no shock was employed. During habituation approximately 4(10 stimuli, l(lC/c of which were 'rare'~ were presented. Usually at the end of such a stimulus series, no conditioned pupil response was apparent to either stimulus as well as no late positive skull potentials. In all cases conditioning was not begun until the averaged pupil response showed stimulus related dilation to be absent. During conditioning stimuli were presented in blocks of 400 and within three or four such blocks the conditioned pupillary response and P300 attained maximum values. Fig. 2 shows the development of the P300 recorded at the skull (B) and the pupillary dilation response (C) m one of the cats. The uppermost records on each side represent the averaged event-related potential and pupillary dila-  A random series of 'frequent' 4 kHz and 'rare' 1 kHz tones were presented at 2-second intervals. A tail shock occurred 700 ms after onset of the rare tone. Averages of eventrelated potentials recorded from the skull (B) and pupillary dilation (C) from one of the cats. Stimuli were presented in blocks of 400, 10% of which were rare tones. The interrupted lines represent potentials evoked by the frequent tones. Continuous lines are potentials evoked by the rare tones. The uppermost records represent the initial 400 trials without tail shock. Subsequent records represent sequential blocks of 400 trials with conditioning. In the first set of trials with conditioning a P300 component became discernable (arrow), which increases in amplitude over the next two series along with pupillary dilation. tion response associated with the rare (continuous line) and frequent (interrupted line) stimuli. Subsequent records show the averaged potentials and pupillary responses to successive blocks of stimuli during conditioning. The arrows in the evoked potential records indicate the latency of the conditioned P300 at approximately 240 ms.
To facilitate training, the duration of the rare stimulus initially persisted until the onset of the electrical stimulus to the tail, that is, 700 ms. Once the P300 appeared, the tone duration was reduced to 50 ms. During data collection the 1 kHz stimulus was presented approximately 10% of the time, except when the effect of increasing the probability to 40% was studied.

29
The paradigm was controlled by a PDP 11/40 computer including randomization of the signal sequence, tone duration, and electric stimulus delivery.

Neural recording
Intracranial electrodes were constructed from 0.005 inch diameter stainless steel wire insulated with epoxylite. The insulation was removed from the terminal 200-300 um of each electrode, lntracranial electrodes were held in stereotaxic carriers and advanced into the brain through implanted cannulae.
The potential recorded from a skull electrode (steel screw) near the vertex was monitored during intracranial electrode advancement. Generally no more than two intracranial electrodes were simultaneously employed. Electrode recordings were obtained relative to an indifferent needle electrode in the back of the neck. A ground electrode was located approximately 1 inch posterior to the indifferent electrode.
Brain potentials were amplified 1000 times with a band pass of 0.1 to 300 Hz (3 dB points, 6 dB/octave slope) and along with the pupillometer output were digitized on-line by the PDP 11/40, sorted with respect to which stimulus had occurred, and averaged. A 625 ms analysis interval beginning at stimulus onset was used with the initial cat. A 650 ms interval beginning 40 ms before stimulus onset was used with the other three animals. During either time base 512 samples of the incoming data were obtained from each channel. Data was not obtained from all recording sites in all four cats. Data was collected at all five locations in cats T-land T-2, at sites 2 and 5 only in cat T-3, and at site 2 only in cat T-4.
During neural recording electrolytic lesions were made at selected intracranial locations for subsequent histological verification of recording sites.
When sacrificed, cats were deeply anesthetized and perfused through the heart with saline and 10C,~ buffered formalin and the brains removed. Frozen sections were cut at 50um and stained with cresyl violet.

Animal care
All procedures were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals (DHEW Publication No. 80-23). Due to the use of paralyzed preparations in a classical condi-tioning paradigm, care was taken to insure that the cats were as comfortable as possible. Typically, the position of the cat was changed at least once during the experimental period, which did not last beyond four hours. Experiments on the same animal were conducted at weekly intervals. The level of electric stimulus employed was not associated with prolonged pupillary dilation. The cats did not resist removal from their normal quarters in the vivarium. All four cats either maintained or slightly increased their body weight during the investigation.

RESULTS
A series of averaged potentials to both the fre-quent and rare tones are shown in Fig. 3B representing results from several penetrations through the marginal gyrus at site 2 in the same cat on different days. There is a significant amplitude disparity between the potentials associated with the rare and frequent stimuli with the former being of considerably larger amplitude. At the dura there is a negative-positive component occurring at 50-100 ms followed by a late positivity peaking at 200 ms, particularly to the rare tone. An interrupted line has been set at this latter latency. To the rare stimulus there is a reversal of polarity of the late positivity (henceforth termed the P300) recorded at the dura immediately below the cortical surface (4,6, and 8 mm). Additional polarity changes occur at deeper levels. A negative-to-posi-  The reversal of polarity of the P300 within the marginal gyrus at site 2 was a consistent observation. Of ten such penetrations in four cats, in which records were obtained within 6 mm of the dura, all manifested the polarity reversal. Fig. 4 shows one example from each cat of event-related potentials associated with the rare stimulus at the dura and within the marginal gyrus. With the exception of the 4 mm record from cat T-3, depth records in Fig. 4 represent the most superficial data obtained below the dura in those penetrations. The interrupted lines are at the latency of the P300 recorded at the dura. The 31 solid line represents the latency of the P300 recorded at the same time at the skull. In all four cases an obvious reversal of polarity of the P300 recorded at the dura is apparent in the depth records. The two skull records for each cat replicate well, indicating the stability of the P300 evoked potential latency during movement of the depth electrode.
It is clear from Fig. 4 that latencies of the P300 recorded at the dura and the corresponding negativity in the depth records agree closely with the latency of the P300 recorded at the skull in cats T-3 and T-4.
However, the latencies of P300s recorded at the skull for cats T-1 and T-2 occur almost 75 and 100 ms later, respectively, than their dural or depth counterparts.   Tables I and 11.
The second column in Table I  within the brain to that recorded from the dura appears in the last column. Table II differs from Table 1 only in that the latency of the P300 recorded at the dura is the point at which the amplitude ratios were determined. In the last column of both tables a negative value for the amplitude ratio means that the potential within the brain is negative relative to the potential at the dura at the latency of measurement, be it skull (Table I) or dura (Table II). Tables I and II agree Fig. 5. A series of closely spaced (depthwise) intracranial records from a site 2 penetration in cat T-2. The timing of the polarity reversal in the depth records corresponds to the latency of the P300 recorded at the dura (vertical interrupted line) rather than the latency of the P300 recorded at the skull (arrows indicate latencies of corresponding P300s at the skull).

For cats T-3 and T-4 the amplitude ratios from
tude is associated with the rare stimulus, but which does not necessarily occur at the same latency as the P300 recorded at the skull. In this study an emphasis was placed on collecting data from the middle of the marginal gyrus, site 2. However, an attempt was made to determine the distribution of the P300 recorded over a portion of the cortex by sampling electrical activity at several other locations (see Fig. 1). In T-1 the amplitudes of P300s recorded at the dura averaged between 28 and 33/~V for sites 2 through 5 while the amplitude rostrally in the marginal gyrus at site 1 was 20 ~V. In cat T-2 amplitudes of P300s recorded at the dura were also, for the most part, closely grouped, although the magnitudes were substantially larger than in T-1. For sites 1 through 4, P300 amplitudes averaged 60 to 70 ~V. In contrast, the P300 amplitude at site 5 averaged only 35 gV. In the third cat the amplitude at site 5 was again smaller, averaging 37 ~V compared to 63 ~V from site 2. It is obvious that P300 components can be recorded from the dura over a wide area of cortex without any consistent amplitude difference. Moreover, a reversal of polarity of the dura positivity was seen at all recording sites as the electrode penetrated through the cortex.
With respect to latency, a wide range of values was observed across dura recording sites, especially in cat 33 T-1. In this animal latencies ranged from as low as 135 ms at site 4 to 268 ms at site 3, whereas the values at the skull were between 238 and 265 ms. In cat T-2 the latencies at the dura varied from 235 to 315 ms. whereas at the skull the latencies were longer, ranging from 310 to 370 ms. In contrast in both cats T-3 and T-4 the latencies of P300s recorded at the dura at sites 2 and 5 corresponded to those recorded from the skull.
In three of four cats the amplitudes of P300s recorded at the skull were about the same as amplitudes recorded at the dura. In cats T-l, T-2 and T-4 the amplitudes at the skull/dura were 34/28, 60/61, and 28/38~V, respectively. Only in cat T-3 was the difference between skull, 100 uV, and dura, 63/~V, substantial.
Latency as the sole criterion for defining a P300like potential recorded from the brain is insufficient. It is necessary to demonstrate that potentials represented in the depth records, as in the marginal gyrus in these experiments, reflect neural events underlying the P300 phenomenon. Factors such as stimulus probability should have similar effects on skull and depth records if the correspondence is to be made. This is an important criterion for examples like the data from cat T-2, in which the latency of the polarity reversal differs substantially from the latency of the P300 recorded at the skull. That such an amplitude change can be obtained by manipulating the probability of the rare stimulus is shown in Fig. 6 for two depth records from cat T-2, each representing a different experiment. Also shown are the associated P300s recorded from the skull and the pupil responses.
Increasing the probability of the rare tone from 0.10 to 0.40 was associated with an attenuation in the amplitudes of both the P300 recorded at the skull and the negative potential in the depth record. When measured at the latency of the P300 recorded at the dura (vertical interrupted line), the reduction in amplitude at a depth of 3 mm (Fig. 6A) was 83%. The reduction in amplitude of the P300 recorded at the skull (arrows) was 63%. On another day (Fig. 6B) the amplitude reductions at a depth of 5 mm and at the skull were 57% and 48%, respectively. Table III summarizes the results of changing stimulus probability. The amplitudes of the negative P300-1ike events recorded intracranially were meas- The results in Table Ill show the anticipated reduction in amplitude at both the skull and in the depth records as probability of the rare stimulus is increased. There was no consistent change in the pupil response, that is, both increases and decreases in amplitude were observed. Results of human studies have been interpreted as indicating that the hippocampus might be a generator  It is obvious from Fig. 7 that the peak amplitude in the depth records, whether negative or positive, occurs, for the most part, earlier than the latency of the P300 recorded at the dura. The latencies of P300s re, corded at the skull rather closely approximate peaks in the depth records at several levels (8, 12, 16 and 18 mm), but not all. In particular, the peaks in the more superficial records occur much earlier than the corresponding P300 recorded at the skull. Table IV Tables I and II, here it is the P300 recorded at the skull which seems to be more closely related temporally to the peaks in the depth records at the level of the hippocampus. Although the ratios agree in sign except for the first two measures, the ratio determined at the latency of the P300 recorded at 35 the skull is larger than the ratio determined at the latency of the P300 recorded at the dura at every level.
Therefore, it is the former which more clearly represents the degree of polarity and amplitude change which occurred in the depth records. The results from T-2 were similar to T-3 in that P300s recorded from the skull were more closely related temporally to peaks in the depth records than were P300s recorded from the dura. The data from cat T-1 was ambiguous. It was not clear which of the P300s corresponded more closely to the peaks in the depth records. Like the data depicted in Tables 1 and II. these results underscore the extent to which latencies of P300-1ike components recorded within the brain may or may not correspond closely to the latencies of P300s recorded at the dura or skull.

DISCUSSION
The major finding to emerge from this study is the reversal of polarity of a P300-1ike potential recorded at the dura when the recording electrode was positioned within 6 mm of the surface. This occurred over a broad strip of cat cortex, including the marginal and suprasylvian gyri. The fact that P300s recorded from the skull replicated well while the electrode was advanced into the brain indicates that the polarity reversal reflected a change in the position of the electrode with respect to the location of a generator rather than to a major change in the activity of the generator itself. The reversal of polarity suggests that the electrode passed through a dipole generator oriented roughly parallel to the electrode track. Considering the relatively superficial depths at which polarity reversals were obtained, the logical site for such a P3(X) generator would be the cortex itself.
In an earlier study Farley and StarrT, s recorded the cat P300 from an array of fixed skull electrodes, The largest P300 amplitude was obtained from the center of the array at the vertex. Although slightly off the midline site 2 in the present study corresponded closely to Farley and Starr's central recording site. Our attention was focused on this site initially because it seemed to offer the highest probability of obtaining an intracranial polarity reversal. However, Farley and Starr's results also showed that the decrease in amplitude associated with recording from electrodes immediately surrounding the central site was not great. Their electrodes were spaced 11 mm apart. Recording sites 1, 3, 4 and 5 in the present study were no more than 11 mm from site 2. It is not surprising, therefore, that the amplitude of P300s recorded from the dura at these locations did not consistently differ much from the amplitude of the P300 recorded from the central site 2 position. The occurrence of a P300 at the dura over a wide area of cat cortex and the relative similarity of amplitude across dura recording sites are both consistent with Farley's results.
Human studies have shown that increasing the probability of the rare stimulus causes a decrease in the amplitude of the P300 (refs. 5,14,22-24,26). Earlier work 30 in our laboratory produced the same result for the cat P300 recorded from the skull and is substantiated by the present study. In addition we have presented evidence that the amplitude of the negative potential obtained superficially within the brain at sites 2 and 3 in the marginal gyrus and site 5 in the suprasylvian gyrus is also reduced by increasing stimulus probability. No probability tests were conducted at sites 1 and 4. In our view it is important that a reduction in amplitude of an intracranial component occur when probability of the rare stimulus is increased if the component does in fact reflect activity of a P300 generator.
Depth records from the region of the hippocampus contained components which occurred at the latency of the P300 recorded at the skull and manifested a large amplitude in response to the rare stimulus. The multiple polarity reversals of these components observed as the electrode traversed the limbic structures are consistent with the location of a P300 generator in the region of the hippocampus. Moreover, the complex arrangement of dendate gyrus, hippocampus, and parahippocampal gyrus provide an anatomical substrate consistent with our observations. Depending upon the angle of penetration an electrode could easily traverse successive layers of cortex, If a dipole generator were distributed across these cortical layers, the result would be multiple polarity reversals, as we have observed. No attempt was made to vary stimulus probability while records were being obtained from this area.
Depth recording in human subjects suggest that the hippocampus, parahippocampal gyrus, and amygdala could be involved in generating the P300: Large P300-1ike potentials manifesting steep voltage gradients and polarity reversals across recording sites have been reported 13,~5.27. In one study single unit act tivity from these structures also correlated with the behavior of the potentials 13. The hippocampus has also been implicated using the analysis of magnetic fields associated with ionic current flow in the brain 20. In contrast substantial excisions involving these structures failed to reduce the P300 recorded from the scalp 33. Also, on the basis of intracranial records other investigators 34 argue for a thalamic origin of the human P300. In an animal study j~ P300-1ike potentials have been recorded from the cingulate cortex and anterior thalamus of the rabbit. Clearly more effort is needed in animal subjects to identify possible subcortical generators of the P300.
If the cat P300 originated from a single source, it would be expected that the latencies of P300-1ike potentials recorded from the skull and various intracranial locations would be very nearly identical because of the instantaneous transmission of potentials by volume conduction 17. However, in the present study P300-1ike potentials from the skull, dura, and depth records often differed substantially in latency. We suggest that these latency disparities reflect the existence of multiple generators of the cat P300. The same point has been made with respect to human data 32. Evidence has been presented in this study for at least two possible generators, a large area of cortex including the marginal and suprasylvian gyri and the region of the hippocampus. If multiple sources contribute to the cat P300, summation and cancellation of activity from different generators would be expected at each recording site and complex changes in the evoked potentials could occur with relatively small movements of the electrode 31. This could explain the substantial latency and polarity differences among records from different recording sites often observed in this study and underscores the importance of using functional criteria. as opposed to simply latency and polarity criteria, to determine whether an intracranial component is related to the P300. What is essential is that the intracranial component be differentially sensitive to rare, task relevant stimuli, as is the P300 recorded at the skull. In this regard it is useful to increase the probability of the rare stimulus and look for a diminished amplitude in the intracranial P300-1ike component.
To the best of our knowledge a reversal of polarity of a P300-1ike potential at the cortical level in either human or animal studies has not been reported elsewhere. In part this may reflect the limited extent of P300 studies in animals, Species differences may also be involved. In addition the P300 from a classically conditioned, paralyzed preparation may not be directly comparable to P300s obtained under other circumstances 7.
While the results of this study suggest that a large area of cat cortex and the hippocampal region contribute to the generation of the cat P300, certain considerations limit this interpretation. Conceivably, an electrode could pass outside of the region of an active generator containing a dipole layer, yet penetrate at an angle more or less parallel to the orientation of the dipole and still record a reversal of polarity of a P300like component. Presumably the amplitude of the component would be less than if the electrode had penetrated the active region. This suggests the importance of closer spacing of recording sites to specify more exactly the locations of the generators of P300-1ike components. Also, it should be noted that even if a subcortical component closely resembles a component recorded at the surface (dura or skull), it may actually contribute little or nothing to the surface potential TM. Finally, the recording of P300-1ike single unit activity from suspected generator sites as well as ablation studies involving those structures are needed to corroborate the results of the present study.