Pervasive synchronization of local neural networks in the secondary ...

Exp Brain Res (2002) 147:227–242 DOI 10.1007/s00221-002-1233-3

RESEARCH ARTICLE

Kevin D. Alloway · Mengliang Zhang · Susan H. Dick · Stephane A. Roy

Pervasive synchronization of local neural networks in the secondary somatosensory cortex of cats during focal cutaneous stimulation Received: 22 April 2002 / Accepted: 25 July 2002 / Published online: 1 October 2002
Springer-Verlag 2002

Abstract Extracellular discharges were recorded from 205 neurons in the secondary somatosensory (SII) cortex of isoflurane-anesthetized cats. Cross-correlation analysis was used to characterize the temporal coordination of SII neurons recorded during cutaneous stimulation with a focal air jet that moved back-and-forth across the distal forelimb. Over two-thirds of the recorded neuron pairs (n=357) displayed significant levels of synchronized activity during one or both directions of air-jet movement. The probability of detecting correlated activity varied according to the distance separating the neurons. Whereas synchronized responses were observed in 82.3% of the pairs in which the neurons were separated by 200–300 m, the incidence of synchronization declined to 52.3% for neurons that were separated by 600–800 m. The distance between neurons also had a significant effect on the temporal precision of correlated activity. For neurons that were separated by 200–300 m, synchronized responses in the cross-correlograms (CCGs) were characterized by narrow (0.5–1.0 ms) peaks at time zero. For SII neurons that were more widely separated, the peak half-widths were substantially broader and more likely to be displaced from time zero. Analysis of directional sensitivity indicated that only 14.2% of the correlated neurons displayed K.D. Alloway ()) · M. Zhang · S.H. Dick · S.A. Roy Department of Neuroscience and Anatomy, Milton S. Hershey Medical Center, Penn State University College of Medicine, Hershey, PA 17033-2255, USA e-mail: [email protected] Tel.: +1-717-5316413 Fax: +1-717-5315184 S.H. Dick Neurosource, Inc., 515 N. State Street, Chicago, IL 60610, USA S.A. Roy Northwestern University School of Medicine, Department of Physiology, Chicago, IL 60638, USA K.D. Alloway Department of Neuroscience and Anatomy, H109, Hershey Medical Center, 500 University Drive, Hershey, PA 17033, USA

a directional preference for synchronized activity. By comparison, 63.4% of the neurons displayed a directional preference in their discharge rate. These results indicate that stimulus-induced synchronization is a prominent feature among local populations of SII neurons, but synchronization does not appear to play a critical role in coding the direction of stimulus movement. A comparison of these results with those obtained from similar experiments conducted in primary somatosensory (SI) cortex indicates that neuronal synchronization is more likely in SII cortex. This finding is discussed with respect to the known functional differences between the SI and SII cortical areas. Keywords Binding · Cross-correlation · Directional sensitivity · Sensory coding · Tactile

Introduction Substantial evidence indicates that tactile stimuli produce time-locked responses throughout the somatosensory system. Thus, stimulus-induced activity in the periphery is highly correlated with subsequent neuronal events in the brainstem, thalamus, and primary somatosensory (SI) cortex (Lee and Ebner 1992; Alloway et al. 1993, 1994; Johnson and Alloway 1994, 1996; Vickery et al. 1994; Gynther et al. 1995; Swadlow 1995; Nicolelis et al. 1998; Zachariah et al. 2001). The time-locked sequence of neuronal responses in these brain regions is not surprising, however, because the underlying anatomical substrate consists of topographically organized projections that transmit information rapidly and reliably from one level of the neuraxis to the next. Although tactile stimulation should cause neuronal responses to be coordinated in ways that are consistent with the hierarchical organization of the somatosensory system, the pattern of neuronal coordination within each brain region is often difficult to predict from the organization of local circuit connections. While some somatosensory brain regions exhibit distinct patterns of

228

time-locked correlations that denote serial interactions among local connections (Shosaku 1986), responses in most areas are dominated by substantial amounts of stimulus-induced synchronization. Simultaneous recordings of multiple neurons in the brainstem, ventrobasal thalamus, or SI cortex have shown that neighboring neurons with similar place and submodality properties are likely to discharge at precisely the same time when a discrete cutaneous stimulus is administered to the periphery (Metherate and Dykes 1985; Alloway et al. 1995; Swadlow et al. 1998; Roy and Alloway 1999, 2001; Nunez et al. 2000). To understand more fully how neuronal activity is coordinated in the somatosensory system during cutaneous stimulation, we recently characterized neuronal responses in the secondary somatosensory (SII) cortex with respect to neuronal responses in the thalamus or SI cortex. Our studies, which were conducted in anesthetized cats, indicate that focal cutaneous stimulation evokes synchronized responses in corresponding somatotopic representations of SI and SII (Roy et al. 2001). Furthermore, SII neuronal responses are tightly coordinated with respect to antecedent events in the ventrobasal thalamus (Roy and Alloway 2001). Indeed, synchronized events occurring among local pairs of thalamic neurons were more likely than asynchronous thalamic events to evoke a subsequent response in SII cortex (Roy and Alloway 2001). Consistent with our previous work examining stimulus-induced interactions between the thalamus and SI cortex (Alloway et al. 1993; Johnson and Alloway 1994, 1996), coordinated responses between thalamus and SII cortex were detected only in those neurons having overlapping receptive fields (RFs). Given the relationship between somatotopic organization and stimulus-induced coordination, a major aim of the present study was to characterize the incidence and strength of correlated activity in SII cortex as a function of neuronal proximity and RF overlap. Moreover, because SII neurons have larger RFs than SI neurons (Alloway and Burton 1985a, 1985b), another aim was to compare the incidence and strength of neuronal synchronization in SII with that observed in SI (Roy and Alloway 1999). Finally, given evidence suggesting that SII is concerned with the global, spatially distributed features of tactile stimuli (Burton 1986), it is plausible that a major function of SII cortex is to encode stimuli that move across the skin. For this reason, we used a repetitive, back-and-forth moving stimulus to enable an analysis of the incidence and strength of synchronized activity as a function of stimulus direction. As a further test of this hypothesis, we compared the directional preferences of synchronized responses in SII with the directional preferences of the underlying neuronal discharge rates.

Materials and methods All procedures followed guidelines established by the National Institutes of Health on the use and care of laboratory animals and

were similar to those used in our previous reports (Johnson and Alloway 1994, 1996; Roy and Alloway 1999; Roy et al. 2001). A total of 18 recording sessions were conducted on two domestic cats in which sterile operating procedures were used to expose SII and implant a stainless steel recording chamber onto the surrounding cranium. A stainless steel bolt attached to the occipital ridge was used to immobilize the animal’s head during all recording sessions. During each recording session, the animal was intubated through the oral cavity and ventilated with a 2:1 gaseous mixture of nitrous oxide and oxygen containing 0.5% isoflurane to prevent reflexive movements. Because the head was not held in a stereotaxic instrument, the concentration of isoflurane was lower than that needed to prevent reflexive movements when the soft tissues surrounding the ears, eyes, and mouth are contacted by ear bars or other stereotaxic devices. Heart rate and end-tidal CO2 were monitored continuously, and body temperature was maintained at 37C by a thermostatically controlled heating pad. The final recording session for each animal was terminated by deeply anesthetizing the animal with an intravenous injection of pentobarbital sodium followed by an injection of 20 mg lidocaine and 1,000 USP units of heparin. The animal was transcardially perfused with 0.9% saline, neutral formalin, and 10% sucrose in formalin. The brain was removed and placed into 30% sucrose in formalin until it sank. The SII cortex was blocked, frozen, and cut into 50-m coronal sections that were mounted onto chrome-alumcoated slides and stained with thionin. Cortical electrophysiology During each recording session an array of three to eight tungsten electrodes (1–5 MW at 1 kHz; Frederick Haer, New Brunswick, ME) was advanced into the forelimb representation of SII cortex, which is located in the anterior ectosylvian gyrus. The electrodes were arranged in a linear configuration (13 or 14) or in a pair of rows (23 or 24). In all cases, the tips of adjacent electrodes were separated by 200–300 m. Electrodes penetrated the cortex orthogonal to the dural surface and were advanced by means of a hydraulic microdrive while the skin was gently manipulated with brushes and other hand-held stimuli. Once single neurons were isolated on two or more electrodes, a hand-held air jet was used to map the RFs and confirm that the neurons were sensitive to hair movements. Extracellular neuronal discharges for each electrode channel were displayed continuously on an oscilloscope during data acquisition. Analog voltage signals for each electrode channel were sampled at a frequency of 18.45 kHz or greater depending on the number of electrodes recorded simultaneously by the A/D board (Data Translation 2839, Marlboro, MA). The voltage potentials were subsequently converted into digital signals and the moment at which the positive wave of each spike discharge exceeded a preset threshold was time stamped to a resolution of 0.1 ms (DataWave Technologies, Broomfield, CO). A computer software program was used during offline analysis to display clusters of neuronal discharges on a Cartesian grid with respect to several parameters including spike width, spike amplitude, and time of maximum spike peak (Autocut 3.0, DataWave Technologies, Broomfield, CO). To improve neuronal isolation, parameter boundaries used to sort individual spikes were adjusted to remove instances in which interspike intervals were less than 1.0 ms. Time stamps from each group of sorted waveforms were then used to generate peristimulus timed histograms (PSTHs), cross-correlograms (CCGs), and joint peristimulus timed histograms (JPSTHs). Cutaneous stimulation Moving air jets activate mechanoreceptors in the skin without producing the lateral distortions caused by dragging a probe across the skin surface (Ray et al. 1985), and we have consistently shown that neurons sensitive to hair movements are activated by computer-controlled air jets (Johnson and Alloway 1994, 1996; Roy and Alloway 1999, 2001; Roy et al. 2001). In the present

229 study, computer-controlled moving air jets were administered in blocks of 200 trials. Each trial was subdivided into three periods: a prestimulus period for recording spontaneous activity (3 s), a stimulus period consisting of a repetitive moving air jet (3 s), and a poststimulus period (2 s). Neuronal activity was recorded onto hard disk during all three of these periods, but was not recorded during the intertrial intervals (4–6 s). We used a modified Grass polygraph module to produce smooth movement of the air jets. The ink pen in the polygraph module was replaced by a hollow tube (1 mm ID) in which the end of the tube was curved so that airflow from the tube was delivered orthogonal to the hairy skin as the tube moved back-and-forth across the hairy skin. Airflow was controlled by an electronic valve that was gated by the data acquisition system (Experimenters Workbench 6.0, DataWave Technologies, Broomfield, CO). Air pressure (20 psi) to the tube was regulated by a needle valve in series with a pressure gauge. Air-jet motion followed a curvilinear trajectory that spanned a distance of 3–7 cm, which corresponds to a velocity range of 6– 14 cm/s. The motion of the air-jet tube was controlled by a 1-Hz sawtooth wave signal from a function generator that lasted for 3 s so that the skin was stimulated 3 times in each direction. Analysis of neuronal responses Cross-correlation analysis was used to characterize the coordination of discharges across pairs of SII neurons recorded by different electrodes. With this analysis, CCGs were constructed to portray changes in the probability of a target neuron discharge given that the reference neuron discharged at time zero. In a stimulus-based paradigm, correlated neuronal activity can be produced by stimulus coordination or may occur randomly according to the rate of neuronal discharges. We minimized these effects by subtracting a shift predictor from the raw CCG (Gerstein and Perkel 1972; Alloway et al. 1993; Johnson and Alloway 1996; Roy and Alloway 1999; Roy et al. 2001). The shift predictor was also used to calculate 99% confidence limits. Because shift-corrected CCGs frequently contain multiple small peaks that barely exceed the 99% confidence limits, the shift-corrected CCGs were also subjected to a smoothing procedure in which the average of each set of three consecutive bins in the unsmoothed CCG was used to determine the height of the middle bin in the smoothed CCG (Aertsen et al. 1989; Gochin et al. 1989). Only those peaks that exceeded the 99% confidence limits in both the smoothed and unsmoothed shiftedcorrected CCGs were considered statistically significant. To characterize the strength of neuronal coordination, we measured the following parameters: Correlation coefficient. The correlation coefficient, p(t), indicates the proportion of discharges in a pair of neurons that are correlated. As in previous studies (Eggermont 1992; Roy and Alloway 1999), it was calculated as: pðtÞ ¼ ½CE
=½ðNA  ðNA Þ2 =TÞðNB  ðNB Þ2 =TÞ
1=2 where CE is the number of correlated events in the tallest 2-ms period of a significant peak in the raw or shift-corrected CCG, T is the time interval over which the CCG was calculated, and NA and NB represent the number of neuronal discharges recorded from cortical neurons A and B during time T. Neurons A and B were always recorded by separate electrodes and one of them was arbitrarily selected to serve as the reference neuron, discharges of which are represented at time zero in the CCG. The correlation coefficient was always measured from CCGs in which each bin was 1 ms in duration. Peak time. This is represented by the time of the tallest bin in a shiftcorrected CCG. Hence, peak time indicates the time lag that occurs most often among correlated discharges in a pair of neurons.

Peak half-width. This was obtained by measuring the width of the shift-corrected CCG peak at half the height of its tallest bin. Thus, the peak half-width provides a measure of variability in the relative timing of correlated discharges recorded from a pair of neurons. For CCGs that had relatively broad peaks, we ignored those instances in which a single bin (1.0 ms) dipped into the peak. Directional preference index. To determine if neuronal synchronization or mean discharge rates varied systematically according to the direction of the moving stimulus, a directional preference index was calculated for the neuronal discharge rates and the correlation coefficient on each trial: Direction Preference Index ¼ ðbackward response  forward responseÞ=ðmean responseÞ where “response” is either firing rate or the correlation coefficient of the raw CCG for each direction. By definition, the direction preference index varies between –2.0 and +2.0 for each trial. We plotted both trial-by-trial and cumulative changes in the direction preference index as a function of trial number. If a directional preference is not present, then both the trial-by-trial and cumulative changes should fluctuate around a value of zero. For statistical purposes, a directional preference was considered significant if a matched-sample t-test (comparing the trial-by-trial result against an equal number of zero values) indicated that differences due to chance were below a probability level of 1%. Receptive field overlap. The outlines of the RFs were determined by using a hand-held air jet to activate the peripheral skin. While listening to the output of a single electrode channel over the audio monitor, the air jet was manually moved across the skin and the edges of the RF were noted. A pair of calipers was then extended across the width or length of the RF and the air jet was then moved through the RF again, thereby confirming that the tips of the calipers were located at the RF edges (length or width). The calipers were then moved to a life-size outline drawing of the forelimb and were used to guide where the RF boundaries were drawn onto the outline with a colored pen. The process was repeated for each electrode channel using a different colored pen to code the electrode channel number. To calculate RF overlap for a pair of neurons recorded from separate electrodes, the RF boundaries of each electrode channel were traced onto a graph paper subdivided into 1-mm2 bins. The number of bins in the overlap area was divided by the number of bins representing the combined RFs for both channels. The RF overlap was calculated only for those neuron pairs that displayed significant levels of stimulus-induced synchronization.

Joint peristimulus timed histograms To visualize changes in synchronization as a function of stimulus position, joint peristimulus timed histograms (JPSTHs) were constructed for neuron pairs displaying significant amounts of correlated activity in the shift-corrected CCGs. Raw JPSTHs, which illustrate the incidence of correlated events produced by the combination of stimulus-induced coordination and neuronal interactions, were constructed first (Aertsen et al. 1989). A predictor JPSTH, which was calculated from the cross product of each neuronal PSTH (PSTHprod), was used to estimate the expectation density of the JPSTH matrix. The difference between the raw and predicted JPSTHs (JPSTH–PSTHprod) estimated the number of joint events that occurred in excess of the expectation density. To obtain a normalized measure of departure from the null hypothesis, a normalized JPSTH was constructed by dividing the difference matrix by the cross product of the standard deviations (SDprod) of the PSTHs. These calculations were implemented by importing the DataWave Autocut files into NeuroExplorer (Nex Technologies,

230 Littleton, MA). The components of the normalized JPSTH (PSTHs, JPSTH matrix, coincident histogram, and CCG) were then imported into a graphics program to construct the final figures (Canvas 7.0, Deneba Systems, Inc., Miami, FL).

Results Stimulus-induced responses averaging at least 3 discharges/s above spontaneous levels were recorded from 205 neurons across 18 recording sessions. We characterized neuronal coordination at 3 or 4 different cortical depths within each recording session and, thus, a total of 56 experiments were conducted. From these experiments, a total of 357 neuron pairs were available for characterizing the relative timing of discharges across SII cortex. In all pairs the neurons were recorded from separate electrodes; instances in which two or more isolated neurons were recorded from the same electrode were not considered as a neuron pair for our analysis. For each experiment, the stimulus trajectory was usually arranged so that the air jet moved through the longest possible axis of the combined RFs to evoke the maximum response from all neurons that were recorded simultaneously. Because most neurons had elongated RFs extending along the wrist and distal forepaw, proximaldistal stimulus movements were used to activate the majority of neurons (n=163), which formed 295 neuron pairs. Another set of neurons (n=28) was activated exclusively by air jets moving in the ulnar-radial directions, and these neurons provided a total of 52 neuron pairs. Only 14 neurons, yielding 10 neuron pairs, were studied with both proximal-distal and ulnar-radial stimulus movements in sequential blocks of 200 trials. Synchronization of SII cortical neurons Using 99% confidence limits, cross-correlation analysis revealed that 69.2% of the neuron pairs displayed synchronized responses to air-jet stimulation moving in either the distal-proximal or ulnar-radial directions. As shown in Table 1, the incidence of synchronization declined as the distance between the recorded neurons increased. For neurons separated by 200–300 m, 82.3% of the pairs displayed significant levels of synchronization in one or more stimulus directions. By comparison, among neuron pairs spanning a distance of 600–800 m, only 52.3% of the pairs were synchronized during stimulus motion in one or more directions.

Table 1 Incidence of neuronal synchronization in SII cortex

A representative example of neuronal synchronization in neighboring parts of SII cortex is shown in Fig. 1. As indicated by the RF drawings, neurons A174a and A174b had largely overlapping RFs on the radial side of the ventral wrist. This is consistent with the fact that a distance of only 250 m separated the electrodes that recorded these neuronal responses. The summed PSTHs indicated that air-jet movements in both the distal (from site 1 to site 2) and proximal (from site 2 to site 1) directions evoked similar responses from each neuron (Fig. 1B). Cross-correlation analysis revealed the presence of synchronized activity during air-jet stimulation in both the distal and proximal directions. Regardless of stimulus direction, the shift-corrected CCGs contained a narrow peak at time zero, which indicates that both neurons tended to discharge at precisely the same time. Examination of the mean waveforms recorded by each electrode revealed clear differences in the shape and duration of the action potentials, and this indicates that the temporal precision of these correlated discharges was not due to instances in which one neuron’s discharge was recorded by both electrodes. The correlation coefficients from these shift-corrected CCGs indicated that the proportion of activity that was synchronized varied, on average, from 4.0% to 2.9% as the stimulus direction changed. For this and all other neuron pairs, directional comparisons were based only on responses recorded during the second, third, fourth, and fifth sweeps of the moving air jet (see bottom of Fig. 1B). Responses during the first and last sweeps were not included in our analysis of directional preferences because any imbalance in the magnitude of the air-jet onset and offset responses, which are independent of stimulus direction, could easily bias the direction preference index. This procedure also insured that we analyzed only those air-jet responses that were preceded by an air jet moving in the opposite direction. An example of neuronal synchronization across a broader expanse of SII cortex is shown in Fig. 2. As indicated by the RF drawings, both neurons represented multiple digits on the ventral surface of the distal forepaw. Even though the electrode tips were separated by 750 m, the RFs for these neurons contained substantial amounts of overlap. The summed PSTHs indicated that air jet movements in both the distal and proximal directions evoked vigorous responses from each neuron, but the shapes of these responses were distinct in their appearance. Consistent with these differences in the timing of discharges relative to the stimulus, the shiftcorrected CCGs had relatively broad peaks that were slightly displaced from time zero (Fig. 2C). These

Cortical distance (m)

Recorded pairs Synchronized pairs Proportion

200–300

301–450

451–599

600–800

Total

181 149 82.3%

60 37 61.7%

72 38 52.8%

44 23 52.3%

357 247 69.2%

231

Fig. 1A–C Example of synchronized responses among a pair of neurons that were separated by 250 m. A Trajectory of the moving air jet shown with respect to the receptive fields (RFs) of each neuron. B Left Peristimulus timed histograms (PSTHs) illustrate summed responses to 200 trials of repetitive air jet movements. Each trial contained three stimulus cycles in which the moving air jet alternated between points 1 and 2 (see A). Solid lines below the PSTHs indicate the time intervals used for measuring the influence of stimulus direction on neuronal synchronization. Bin widths,

25 ms. Right Mean waveforms of discharges recorded at the beginning or end of the block of 200 trials; each waveform represents the mean of 500 discharges. Scale bars are valid for all waveforms. C Shift-corrected cross-correlograms (CCGs) comparing the pattern of all correlated discharges with those obtained during specific stimulus directions. The proportion of activity correlated during the CCG peak is indicated by the correlation coefficient, p(t). Dotted lines indicate 99% confidence limits. Bin widths, 0.5 ms

correlation patterns indicate that the discharges for these neurons were coordinated in time, but one neuron had a consistent tendency to discharge 5–15 ms before the other neuron and this relative timing pattern was characterized by substantial amounts of variation. The correlation coefficients for the shift-corrected CCGs were slightly different in the two stimulus directions (2.8% vs 3.0%), and this suggests that stimulus direction had a weak effect, if any, on the proportion of activity that was synchronized.

population CCGs revealed an increase in temporal variability as cortical separation increased. As illustrated by Fig. 3, the population response across relatively short distances (i.e., 200–300 m) was characterized by a narrow CCG peak centered on time zero. By comparison, the population of correlated responses detected across longer distances (i.e., 600–800 m) contained a peak that was substantially broader and shallower. To assess the effect of cortical distance and RF overlap on the relative timing of correlated SII responses, we measured the peak times and peak half-widths of the shift-corrected CCGs (see “Materials and methods”). The time of the tallest bin in the CCG peak (i.e., peak time) indicates the most frequent interspike interval that occurs among discharges across two neurons. By comparison, the width of the CCG peak at half its height (i.e., peak half-width) provides an indication of the amount of variability among the interneuronal interspike intervals. As illustrated by Fig. 4A, the average peak time was only 0.90€0.16 ms (mean € SEM) for synchronized neurons located within 300 m of each other, but increased to 3.57€0.81 ms for pairs of neurons separated by 600– 800 m. A one-way analysis of variance indicated that a significant portion of the variability in peak time could be

Effect of cortical distance and receptive field overlap on SII synchronization The main goal of this study was to determine how SII neuronal responses are coordinated as a function of their cortical proximity and RF overlap. To visualize the impact of cortical distance on neuronal synchronization, individual shift-corrected CCGs that displayed significant interactions were summed together to produce a population CCG for each range of cortical distances appearing in Table 1. Consistent with the coordination patterns seen in the representative neuron pairs (see Figs. 1, 2), the

232 Fig. 2A–C Example of synchronized responses recorded from a pair of neurons separated by 750 m. Stimulus trajectory, RFs, and PSTHs are illustrated as in Fig. 1. Bin widths for the CCGs are 5 ms

Fig. 3 Comparison of population CCGs as a function of electrode separation. Each population CCG represents the sum of individual shift-corrected CCGs that were classified as displaying synchronized (top panels) or unsynchronized responses for each range of

electrode separations listed in Table 1. The number of neuron pairs used to illustrate the population response is shown for each panel. Bin widths, 2 ms

attributed to neuronal separation (F=8.89, P