Frequency-dependent response of SI RA-class neurons to vibrotactile stimulation of the receptive field

Transcription

1 Somatosensory & Motor Research 2001; 18(4): Frequency-dependent response of SI RA-class neurons to vibrotactile stimulation of the receptive field B. L. WHITSEL 1,2, E. F. KELLY 3, M. XU 2,3, M. TOMMERDAHL 2 and M. QUIBRERA 1,4 1 Department of Cell and Molecular Physiology; 2 Department of Biomedical Engineering, UNC School of Medicine; 3 Department of Diagnostic Sciences, UNC School of Dentistry; 4 Department of Statistics, University of North Carolina, Chapel Hill, NC 27599, USA Abstract Three types of experiment were carried out on anesthetized monkeys and cats. In the first, spike discharge activity of rapidly adapting (RA) SI neurons was recorded extracellularly during the application of different frequencies of vibrotactile stimulation to the receptive field (RF). The second used the same stimulus conditions to study the response of RA-I (RA) cutaneous mechanoreceptive afferents. The third used optical intrinsic signal (OIS) imaging and extracellular neurophysiological recording methods together, in the same sessions, to evaluate the relationship between the SI optical and RA neuron spike train responses to low- vs high-frequency stimulation of the same skin site. RA afferent entrainment was high at all frequencies of stimulation. In contrast, SI RA neuron entrainment was much lower on average, and was strongly frequency-dependent, declining in near-linear fashion from 6 to 200 Hz. Even at 200 Hz, however, unambiguous frequencyfollowing responses were present in the spike train activity of some SI RA neurons. These entrainment results support the periodicity hypothesis of Mountcastle et al. (J Neurophysiol 32: , 1969) that the capacity to discriminate stimulus frequency over the range 5 50 Hz is attributable to the ability of SI RA pyramidal neurons to discharge action potentials in consistent temporal relationship to stimulus motion, and raise the possibility that perceptual frequency discriminative capacity at frequencies between 50 and 200 Hz might be accounted for in the same way. An increase in vibrotactile stimulus frequency within the range Hz consistently resulted in an increase in RA afferent mean spike firing rate (MFR). SI RA neuron MFR also increased as frequency increased between 6 and 50 Hz, but declined as stimulus frequency was increased over the range Hz. At stimulus frequencies > 100 Hz, and at positions in the RF other than the receptive field center (RF cen ter ), SI RA neuron MFR declined sharply within s of stimulus onset and rebounded transiently upon stimulus termination. In contrast, when the stimulus was applied to the RF c e n te r, MFR increased with increasing frequency and tended to remain well maintained throughout the period of high-frequency stimulation. The evidence obtained in combined OIS imaging and extracellular microelectrode recording experiments suggests that SI RA neurons with an RF c en te r that corresponds to the stimulated skin site occupy small foci within the much larger SI region activated by same-site cutaneous flutter stimulation, while for the RA neurons located elsewhere in the large SI region activated by a flutter stimulus, the stimulus site and RF c en te r are different. Key words: somatosensory, cerebral cortex, vibration, flutter, spike discharge entrainment, periodicity coding, frequency discrimination Introduction Optical intrinsic signal (OIS) imaging studies (Tommerdahl & Whitsel, 1996; Tommerdahl et al., 1998, 1999a, b) have demonstrated that SI cortex responds differentially to same-site cutaneous flutter (25 Hz) vs vibration (200 Hz). In cats and monkeys 25 Hz stimulation reliably produced, as expected, an increase in optical absorbance in the topographically appropriate region of SI. An increase in the optical absorbance of sensory cortex is known to reflect increased concentration of K + in the extracellular space, glial swelling, and perhaps other sequelae of local neuronal activity (Grinvald et al., 1991, 1994, 1999; Holthoff and Witte, 1996; Kohn et al., 2000). The time course of the absorbance increase evoked in SI by 25 Hz flutter stimulation was generally consistent with published descriptions (Mountcastle et al., 1969, 1990) of the spike discharge response of SI RA neurons to such stimulation it began within milliseconds of stimulus onset, was relatively well maintained throughout the period of stimulation, and decayed to prestimulus levels shortly after stimulus termination. A very different sequence of optical changes occurred in SI when the skin was stimulated at 200 Hz (Tommerdahl et al., 1999a, b). During the initial 1 3 s of 200 Hz stimulation there was an increase in SI absorbance indistinguishable in magnitude and spatial extent from the absorbance changes detected at the same time after onset of 25 Hz stimulation of the same skin site. With continuing 200 Hz stimulation, however, absorbance decreased in most of the SI region that had responded to 25 Hz stimulation, except at one or, more frequently, several relatively small loci where an increase in absorbance persisted until stimulus termination. Two further observations indicated that the prominent changes in both the spatial profile and magnitude of the SI OIS that occurred during an exposure Correspondence: B. L. Whitsel, Ph.D., 155 Medical Research, CB#7545, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA. Tel.: ; ; Fax: ; bwhitsel@med.unc.edu ISSN (print)/issn (on line)/01/ Taylor & Francis Ltd DOI: /

2 264 B. L. Whitsel et al. to continuous high-frequency skin stimulation were attributable at least in part to active inhibition, rather than just an activity-dependent decrease in mechanoreceptor afferent responsivity (Leung, 1994; Leung et al., 1994). First, within several seconds after the onset of 200 Hz vibration, absorbance in the topographically appropriate regions of SI not only declined sharply from the values reached shortly after stimulus onset, but often fell to levels below those observed prior to stimulus onset, even as absorbance continued to increase in the topographically appropriate sector of the neighboring SII (Tommerdahl et al., 1999a, b). Second, the normally robust and well-maintained absorbance increase evoked in SI by a 2 5 s exposure to 25 Hz flutter underwent a rapid and substantial reduction in areal extent and intensity when a 200 Hz component was superimposed on the much larger amplitude flutter stimulus (Tommerdahl et al., 1999b). These OIS imaging results were tentatively interpreted to indicate (1) that the spike discharge activity of most RA-type SI pyramidal neurons increases only transiently in response to the vigorous activity a suprathreshold high-frequency stimulus evokes in the central projections of RA-I (RA) afferents, and (2) that within 1 3 s after stimulus onset the effect on these same SI RA neurons of continuing highfrequency afferent drive becomes predominantly inhibitory (Tommerdahl et al., 1999a, b). This interpretation is consistent with the strong positive correlation consistently reported between optical absorbance and sensory cortical neuron spike discharge activity (Grinvald, 1985; Grinvald et al., 1991, 1994, 1999, 2001; Tommerdahl and Whitsel, 1996; Tommerdahl et al., 1999a, b), the scarcity of SI pyramidal neurons that receive their principal input from RA-II (Pacinian, PC) afferents (Mountcastle et al., 1969, 1990), and the exquisite responsivity of SI GABAergic intrinsic neurons to high-frequency input drive (McCormick et al., 1985). However, it has not yet been directly verified by a detailed analysis of the behavior of RA-type SI neurons during high-frequency skin stimulation. The present paper fills this gap by reporting results of analyses of the spike discharge activity of RA-class SI pyramidal neurons in response to stimulation of the RF at frequencies of vibrotactile stimulation between 6 and 200 Hz. The main results were obtained by systematically applying analytical techniques which enable measurement of entrainment quantitatively, and independently of responsivity, across a range of vibrotactile stimulus amplitudes representative of those encountered during everyday life (Whitsel et al., 2000). To address the possibility that frequency dependencies in the responses of SI RA neurons might merely reflect the effects of stimulus frequency at the sensory periphery, we also evaluated the effects of same-site flutter vs vibratory stimuli on the spike discharge activity of RA afferents. Methods Subjects/general procedures All procedures are consistent with USPHS and NIH policies and guidelines on animal care and welfare in biomedical research, and were reviewed and approved by an institutional committee prior to initiation of the experiments. Afferent recording experiments Cutaneous mechanoreceptor afferent spike discharge activity was recorded in three monkeys (Macaca nemistrina) and in two cats (Felis domestica). A surgical level of general anesthesia was achieved with pentobarbital (25 30 mg/kg, i.v.). Skeletal neuromuscular transmission was blocked by Norcuron (vecuronium bromide mg/kg, i.v.). Positive pressure respiration was provided via a tracheal tube. End-tidal CO 2 was maintained between 3.5 and 5.0%. Glucose (5%) and 0.9% saline were administered intermittently to maintain normal metabolism, hydration, and electrolyte balance. Supplemental neuromuscular blocking drug and anesthetic were administered i.v. on regular schedules (Norcuron mg/ kg/h; pentobarbital 2 5 mg/kg/h). The median (forelimb) or tibial (hindlimb) nerve was exposed, freed from surrounding tissues, covered with mineral oil, and transected central to the recording site. Filaments were dissected using fine forceps, and placed one-at-a-time onto a Ag AgCl recording electrode connected to conventional neurophysiological monitoring instrumentation. Filaments from which vigorous multi-unit spike discharge activity could be evoked by hand-held mechanical stimulation (using fine brushes and/or probes) of either the volar surface of the hand or foot (monkeys), or the pads of the distal forepaw (cats) were subdivided until the action potentials of individual afferents were evident. Afferents were classified as rapidly adapting (RA; more precisely, RA-I; Vallbo et al., 1984) if spike discharge activity (1) could be evoked by application of gentle mechanical stimuli to a localized skin site, (2) was absent or rare in the absence of such stimulation, and (3) occurred only transiently in response to maintained mechanical contact with the RF. Spike discharge activity of RA afferents was recorded from 26 different fibers/filaments. Eight recordings of monkey RA spike discharge activity met the standard criteria (consistency of action potential amplitude and waveform) for activity deriving from a single fiber (SUR), while for four additional recordings the activity was judged to derive from two or three RA afferents (MURs) each of which had an RF that included the skin site contacted by the stimulator probe. Fourteen RA afferents were studied in cats, all meeting SUR criteria. Euthanasia was accomplished by i.v. administration of pentobarbital (50 mg/kg). Cortical recording experiments Extracellular microelectrode recordings of SI neuron spike discharge activity were obtained in five monkeys (three macaques Macaca nemistrina and two squirrel monkeys Saimiri sciureus) and in five cats (Felis domestica). The experiments on macaques yielded recordings from both cortical neurons and mechanoreceptive afferents, with the cortical recordings always carried out first. General anesthesia was induced by supplying 4% halothane in a 50/50 mixture of oxygen and nitrous oxide to a light- and air-tight enclosure housing the subject. The trachea was intubated and the tracheal tube connected to an anesthesia machine (Forreger Compac-75). The anesthetic gas mix was adjusted (typically % halothane in 50/50 N 2 O/oxygen) to maintain a stable level of surgical anesthesia. Methylprednisolone sodium succinate (20 mg/kg) and gentamicin sulfate (2.5 mg/kg) were injected intramuscularly to lessen the probability of halothane-induced cerebral edema and prevent bacterial septicemia, respectively. A valved catheter in a superficial hindlimb vein enabled administration of drugs, glucose (5%), and electrolytes (0.9% saline). A 1.5 cm opening was trephined in the skull overlying SI cortex. A recording chamber (25 mm i.d.) was positioned over the opening and cemented to the skull with dental acrylic. Wound margins were infiltrated with local anesthetic, closed with sutures and bandaged, and the dura overlying SI incised and removed. After completion of surgical procedures subjects were immobilized with i.v. Norcuron (loading dose mg/kg; maintenance dose mg/kg/h). From this point on, the 50/50 mix of N 2 O and oxygen was provided via a positive pressure ventilator and the concentration of halothane adjusted (typically between 0.5 and 1.0%) to maintain heart rate, blood pressure, and

3 Vibrotactile frequency coding 265 the EEG at values consistent with general anesthesia. Rate and depth of ventilation were modified to maintain end-tidal CO 2 between 3.0 and 4.5%. After obtaining a high-resolution photograph of the exposed cortical surface the recording chamber was filled with artificial cerebrospinal fluid and closed with a clear glass plate containing an o -ring. The o -ring permitted a microelectrode to be advanced while maintaining hydraulically sealed closed-chamber conditions, thus minimizing cortical and vascular movements associated with the cardiac and respiratory cycles. The glass plate enabled (via an operation macroscope) determination of the precise site and micrometer position at which the microelectrode made initial contact with SI cortex. Extracellular recordings of neuronal spike discharge activity were obtained using glassinsulated tungsten wires (impedance kv at 10 khz). At the maximal depth of a penetration, and also at sites where recordings of particular interest had been obtained, a microlesion was created by passing 1 10 m A of d.c. current through the microelectrode. Euthanasia was by i.v. administration of pentobarbital (50 mg/ kg) followed by intracardial perfusion with 0.9% saline and 10% formalin. The region of SI studied was removed and serially sectioned at 30 m m; monkey SI was sectioned in the sagittal plane, cat SI in the coronal plane. Sections were mounted on glass slides, Nissl-stained, coverslipped, and inspected microscopically. Areas 3a, 3b, 1 and 2 were distinguished on the basis of established cytoarchitectonic criteria (monkey Powell and Mountcastle, 1959; Jones and Porter, 1980; Sur et al., 1982; cat Hassler and Muhs-Clement, 1964; McKenna et al., 1981). Microelectrode tracks were plotted using a microscope with a drawing tube attachment. The sites at which recordings were made along each track were reconstructed using the three types of micrometer readings recorded for each microelectrode penetration: (1) position where the microelectrode made contact with the pial surface; (2) locus at which spike discharge activity was recorded; and (3) depth at which each microlesion was placed. In four experiments (two squirrel monkeys, two cats) the OIS responses evoked in SI by 25 Hz flutter and by 200 Hz vibration were recorded prior to microelectrode recording, using an imaging system employed in previous studies (for methodological details see Tommerdahl et al., 1999a, b). The system consisted of a computer-interfaced CCD camera, the light source, guide, and filters required for near-infrared (833 nm) illumination of the cortical surface, a focusing device, and a recording chamber with an optical window. Near-infrared illumination was used because it minimizes the contributions to OIS images of the changes in blood flow and flow/volume that normally accompany cortical neuronal activation, and also because the intrinsic signal obtained at 833 nm exhibits substantially higher spatial and temporal resolution than the signals recorded at lower wavelengths. Once OIS images of the SI response to 25 and 200 Hz stimulation of the same skin site had been obtained, the optical recording system was disassembled and replaced with the apparatus for performing extracellular microelectrode recordings. The goal of the microelectrode recording component of each combined OIS imaging/neurophysiological recording experiment was to characterize the spike train responses of SI RA neurons at known locations within the optical response pattern produced under the same stimulus conditions. Vibrotactile stimulation A mechanical stimulator (Chubbuck, 1966) was used to deliver sinusoidal vertical skin displacement stimuli to a skin locus from which hand-held gentle mechanical stimuli evoked vigorous SI neuron or RA afferent spike discharge activity. The stimulator made contact with the skin via the flat end of a plastic cylindrical probe (2 or 5 mm in diameter) threaded to the stimulator shaft. Sinusoidal probe motion began at phase zero (at 1.0 mm skin indentation), and initially advanced further into the skin. Since peak-to-peak amplitude of the sinusoid never was greater than 0.6 mm, the probe remained in contact with the skin during stimulation. In most studies the probe continued to indent the skin during the interval between successive stimuli ( trials ). In several experiments the stimuli were superimposed on an intermittent pedestal i.e., the probe was maintained in a position above the skin surface prior to stimulation, advanced rapidly (10 ms) to produce 1.0 mm of skin indentation at ms before the onset of sinusoidal stimulation, and retracted to the offthe-skin rest position ms after stimulus termination. Similar stimulus conditions and protocols were used to study afferents and cortical neurons. Each neuron and afferent was studied with at least one frequency selected from the range of frequencies humans experience as flutter (5 50 Hz), and another from the range humans experience as vibration ( Hz). Most afferents and neurons were studied using 5 7 frequencies within the range Hz, presented in either fully randomized or interleaved order. Neural data collection/analysis Spike discharge activity and an analog signal of stimulus position were digitized at 20 khz. Software allowed post-experimental display and review of both neuroelectrical and stimulator events. A high-resolution monitor was used to inspect the relationship between spike firing and stimulator events, evaluate action potential waveforms, and discriminate (using voltage windows) spikes attributable to different units. For each recording 1 3 nonoverlapping voltage windows were selected so that each contained action potentials attributable either to a single unit or to a small grouping of units. An electronic file was generated for every run, containing the times of occurrence of the action potentials falling within each voltage window, and the times of specific stimulator events (onset and termination of the pedestal, onset of each individual stimulus cycle, termination of each trial, etc.). Two aspects of spike discharge activity are especially relevant to the perception of vibrotactile stimuli, and these were measured separately. The first, responsivity, was measured by counting the number of spikes in a designated time period and then dividing that count by the number of stimulus cycles. Incomplete stimulus cycles were ignored in this calculation. For SI RA neurons the measure of responsivity was adjusted in two ways to correct for the significant amounts of spontaneous or background activity typically present: the first method subtracted background mean firing rate (MFR) from the MFR during the stimulus period before converting to spikes/cycle (adjusted and unadjusted MFRs were correlated, < 0.95). The second used background mean firing rate as a covariate in analysis of variance (ANOVA) of the unadjusted responsivity. Corresponding results from these two approaches were invariably highly similar. Responsivity was measured for two overlapping time periods (0 0.5 and 0 2 s after stimulus onset), in order to detect and characterize, at least crudely, possible rapid changes in unitary activity following stimulus onset. For some runs stimulus duration was 0.8 s, and for these responsivity was computed for the periods and s after stimulus onset. Entrainment, the organization of spike discharge activity into orderly temporal patterns coupled to the sinusoidal motion of the stimulus, was assessed using three measures which collectively enable entrainment to be measured quantitatively under a wide variety of stimulus conditions (Whitsel et al., 2000). All three measures take values between 0 and 1, with 1 indicating perfect entrainment. The first, r 1, derives from the theory of circular statistics (Batschelet, 1981), and measures the tendency of spikes to cluster near a single modal or most-favored position in the stimulus cycle. This behavior corresponds to the classical view of neuronal entrainment as developed by Talbot et al. (1968) for skin mechanoreceptive afferents. The second, r 2, extends the approach embodied by r 1 to the specific and physiologically common situation in which spikes are generated approximately 180 apart in the stimulus cycle. This often occurs, for example, for Pacinian (PC or RA-II) afferents even at low amplitudes of sinusoidal stimulation, and for RA-I afferents exposed to moderate-to-large stimulus amplitudes (Johansson et al., 1982; Whitsel et al., 2000). The third measure, r s, takes a more general approach based on spectral analysis of spike trains. Each record to be analyzed consists initially of a sequence of events of unit amplitude (the spikes) occurring at precisely known but unequally spaced times. This record is first transformed into a series of samples spaced equally in time, using an algorithm which has been demonstrated (French and Holden, 1971a, b) to yield unbiased and alias-free estimates of the power spectrum. This resampling is carried out at a rate 10 times the stimulus frequency, in order to capture a minimum of five harmonics in the power spectrum. The transformed record, in turn, is centered, detrended, and converted to standard scores (mean zero, variance one) in order to remove variation due to the overall response level. Its power spectrum then is calculated using conventional FFT-based methods (Marple, 1987). The resulting measure of entrainment, r s, reflects the

4 266 B. L. Whitsel et al. TA B L E 1. Experimental variables Variables/levels RA afferents (N = 26) No. of runs = 103 RA neurons (N = 81) No. of runs = 326 Stimulus amplitudes (m m, p-p) Stimulus duration (s) > Contactor sizes (mm) Pedestal type Continuous Interrupted Discharge type SUR MUR Species Cat Monkey proportion of the normalized signal amplitude appearing at the stimulus frequency and its integral harmonics. For full details, see Whitsel et al. (2000), who demonstrated that r s approximates r 1 for monopolar patterns of phase-locked spike discharge activity, approximates r 2 for bipolar patterns, and yields high values for many additional, more complex patterns that neither r 1 nor r 2 can effectively measure. In this paper we use the highest of the three measures calculated for a given record as the final, best measure of the degree of entrainment for that record denoted r b. Table 3 gives the number and proportion of SI RA neurons (total sample = 324) for which r 1, r 2, or r s yielded the highest measure of entrainment. The central aim was to investigate effects of stimulus frequency on unit responsivity and entrainment. To that end, for each afferent or cortical neuron the data generated in trials in which the same frequency was applied were grouped to form a single run, and measures of responsivity and entrainment were calculated for that run as a whole. The entrainment measures r 1 and r 2 were calculated based on all available spikes, and r s was calculated from the time average of the corresponding resampled trial-by-trial records. The results of each run were entered into a database containing, in addition to the measures of neuroelectrical activity, descriptors of the associated experimental conditions for example, experiment and unit ID, and stimulus properties such as frequency, amplitude, duration, and contactor size. Both the spike discharge activity of individual SI RA neurons ( single-unit recordings SURs) and the activity of small RA neuron groupings (consisting of 2 5 neurons; multiunit recordings MURs) were studied. In all, 21 RA single neurons and 9 RA MURs were studied in monkeys; and 38 RA single neurons and 13 RA MURs in cats. Consideration of the possibility that the measures of entrainment and responsivity obtained from a MUR might deviate from those obtained from a SUR led us to (1) label each digitized record of neuronal activity (file) in a way that identified it as a record of single- or multiunit spike discharge activity, and (2) evaluate the two types of data (single unit vs MUR) separately. Separate databases were constructed for the records of RA afferent and cortical neuron activity. The afferent database contained information derived from 103 runs carried out on the 22 single RA afferents and 4 small groupings of RA afferents; the cortical neuron database summarized the results of 326 runs carried out on the 59 single SI RA neurons and 22 small SI RA neuron groupings. Table 1 summarizes the experimental conditions associated with both databases. Quantitative treatment of the information in the databases consisted mainly of analyses of variance using the neuroelectrical response measures as dependent variables, and stimulus frequency as the primary independent variable. Statistical analyses and plots were generated using SYSTAT and MATLAB under Windows 98. Results Representative observations The results shown in Figures 1 4 were obtained from a single SI RA neuron and a single RA afferent recorded in the same subject. Each unit was exposed to 15 repetitions of continuous (2 s duration), 100 m m amplitude sinusoidal stimulation at five different frequencies (12, 25, , and 150 Hz; see figure legends for additional details). The raster plots in Figure 1 show the time series of spikes recorded during each stimulus, and also during the initial 1 s of the 15 s interstimulus interval (ISI). Substantial differences between the spike firing behaviors of the SI RA neuron and the RA afferent are apparent. First, the neuron, but not the affer-

5 Vibrotactile frequency coding 267 FIG U RE 1. Spike trains recorded from an exemplary SI RA neuron (five raster plots on left) and RA afferent (right), both studied in the same subject. Each raster plot shows the responses to 15 consecutive presentations of a given frequency of stimulation to the RF c e n te r. At all stimulus frequencies amplitude = 100 m m, duration = 2 s, ISI = 15 s. Probe diameter was 2 mm for the SI neuron, 5 mm for the afferent. Downward arrows indicate stimulus termination. The RF of the neuron (area 1, lamina III) included the proximal pad of digit 2 and the neighboring part of the medial interdigital pad on the foot. The RF of the afferent was confined to the distal pad of digit 2 of the foot.

6 268 B. L. Whitsel et al. FIG U RE 2. Circular histograms and r b values for the spike train data of Figure 1. Line lengths are scaled relative to the most frequently occurring angular position. ent, exhibited a level of spontaneous firing sufficiently high and variable to obscure the boundary between the stimulus and no-stimulus periods at every stimulus frequency. Second, while the neuron s mean spike firing rate (MFR), like the afferent s, increased with increasing stimulus frequency within the range Hz, the magnitude of the increase was much smaller than that of the afferent. Third, although the afferent displayed a metronome-like capacity to follow the sinusoidal stimulus that is visually obvious up to at least 50 Hz (it is less apparent at the higher frequencies only because of the time scale of the raster plots in Fig. 1), entrainment of the neuron s spike discharge activity is apparent only at the lowest frequency. Specifically, at 12 Hz, but not at any of the higher frequencies, the vibrotactile stimulus caused the neuron to discharge bursts of spikes at roughly consistent positions in the stimulus cycle. Fourth, the afferent showed a much stronger tendency, with increasing stimulus frequency, for MFR to decline both within and across trials. The contrast between the entrainment behaviors of the same RA afferent and SI RA neuron is captured more completely and quantitatively by circular histogram analysis (Fig. 2; note that a circular histogram is in essence a cycle histogram, the ends of which have been joined together). For the afferent the pattern of action potential phase locking to the sinusoidal stimulus is of the classic monopolar (r 1 ) type, and it is apparent that the high degree of entrainment visually evident in the spike raster plots for frequencies between 10 and 50 Hz was maintained all the way to the highest frequency (150 Hz). By contrast, the circular histograms on the left reveal that the lesser entrainment exhibited by the SI RA neuron at 12 Hz degraded substantially and progressively as stimulus frequency increased. At the two highest frequencies, in fact, the timing of spike firing appears essentially unrelated to the stimulus cycle (i.e., r b < 0). Further contrasts between the same SI RA neuron and RA afferent are made evident by Figure 3, which for both units plots responsivity and entrainment as functions of trial number (top four plots) and temporal position within the stimulus period (bottom four plots). For this neuron, but not the afferent, responsivity decreased drastically with increasing frequency, especially within the range Hz. In addition, for the afferent entrainment was uniformly high and stable both across and within trials at all stimulus frequencies, whereas for the SI neuron entrainment was not only frequency-dependent, but considerably more variable with some suggestion of temporal trends most evident within the flutter range of stimulus frequencies.

7 Vibrotactile frequency coding 269 FIG U RE 3. Responsivity (in terms of average spikes/cycle) and entrainment (in terms of r b ) for the same neuron and afferent (Figs. 1 and 2) plotted as functions of trial number and time in seconds after stimulus onset. FIG U RE 4. Relationship between mean firing rate and stimulus frequency for the same neuron (top) and afferent (bottom) whose data are shown in Figures 1 3.

8 270 B. L. Whitsel et al. FIG U RE 5. Spike trains (rasters at left), PSTs (middle), and circular histograms (right) for RA SI neuron in lamina III of area 3b (cat). The RF included digit pads 2 4 as well as neighboring regions of hairy skin on the ventral surfaces of digits 2 4. Three frequencies of stimulation of the digital pad of digit 3 were delivered (10 Hz, top; 50 Hz, middle; and 100 Hz, bottom). At all frequencies amplitude = 500 m m; duration = 800 ms; ISI = 2 s; probe diameter = 5 mm. The responses to presentations of each frequency were recorded. Mean firing rate (calculated over the entire 800 ms period of stimulation) at 10 Hz was spikes/s, SD = 7.00; at 50 Hz, spikes/s, SD = 6.40; and at 100 Hz, spikes/s, SD = Average rate of spike discharge in the absence of stimulation was 7.66 spikes/s (arrows on PST ordinates). A property common to the great majority of the recordings of SI RA neuronal activity was that MFR did not reliably increase as stimulus frequency increased over the range Hz. For many such recordings, in fact, a Hz increase in stimulus frequency at frequencies > 50 Hz was accompanied by a substantial decrease in MFR. In addition, in most cases the initial rate of stimulusevoked spike firing failed to be maintained over even relatively brief (0.5 2 sec) periods of constant high-frequency stimulation. The SI neuron whose data are shown in Figures 1 4 is atypical in this regard, and the phenomenon is considerably more obvious in the results obtained from other neurons. For example, the single RA neuron whose data are shown in Figure 5 emitted well-maintained and well-entrained spike discharge in response to 10 Hz stimulation (plots at top), but its responses to samesite 50 Hz (middle) or 100 Hz stimulation (bottom) were neither well-maintained nor well-entrained. At the two higher frequencies of stimulation this neuron s response in the initial 100 msec after stimulus onset either approximates (at 50 Hz) or exceeds (at 100 Hz) its response to 10 Hz. However, with continuing stimulation at 50 Hz MFR declined progressively from its initial values, and at 100 Hz firing rate fell below background for a brief period, and then resumed at a rate much lower than that evoked by 10 Hz stimulation. Figure 6 shows the pronounced frequency-dependency of another SI RA neuron. This neuron, like the SI neurons shown in Figures 1 5, responded vigorously

9 Vibrotactile frequency coding 271 FIG U RE 6. PST (top) and circular histograms (middle) generated from the spike train responses of area 3b RA neuron (lamina IV) to seven different frequencies. All stimuli were applied to the ulnar edge of the digital pad of contralateral forepaw digit 5 (cat). At all frequencies amplitude = 200 m m, duration = 800 ms, ISI = 5 s, probe diameter = 5 mm. The responses to at least 40 trials were recorded at each stimulus frequency. Plots at bottom show MFR and r b as functions of frequency. and in a sustained manner to each of the six mid-range frequencies of flutter stimulation that were delivered (only the PSTs for the responses to 7 and 20 Hz stimulation are shown at the top of Fig. 6 due to space contraints). In contrast, the response of this same neuron to each of the five higher stimulus frequencies (50, 60, 70, 95, and 105 Hz) was quite different: at each of these frequencies there was a burst of spike discharge activity within the first 50 ms of stimulation that was followed by either a precipitous slowing or complete elimination of spike firing lasting for ms. Moreover, at every frequency between 50 and 105 Hz spike firing resumed with continuing stimulation, but at a rate substantially lower than that recorded at the same time after onset of flutter stimulation. Figure 6 also shows that (1) at frequencies of 50 Hz or higher, MFR and entrainment were substantially lower than at any frequency between 7 and 25 Hz, and (2) entrainment dropped progressively as frequency was increased above 50 Hz (see circular histograms, MFR vs frequency, and entrainment vs frequency plots in Fig. 6). OIS/neural recording experiments Combined OIS imaging and extracellular single neuron recording experiments yielded information about why some SI RA neurons (e.g., the neuron whose data are illustrated in Figs. 1 4) continued to discharge vigorously throughout the application of a high frequency stimulus to the RF, while for others (e.g., the neurons in Figs. 5 and 6) MFR fell rapidly towards or even below background shortly after the onset of high-frequency stimulation. Figure 7 shows OIS difference images obtained from SI cortex of a cat at two different times (1 and 5 s) after the onset of 25 Hz vs 200 Hz stimulation of the same skin site. Consistent with the findings reported in earlier studies (Tommerdahl et al., 1999a, b), the absorbance changes evoked by flutter vs

10 272 B. L. Whitsel et al. FIG U RE 7. Bottom left panel photograph showing superficial vasculature in pericoronal cortex (cat), entry points of three microelectrode penetrations (3 ), and boundaries of cytoarchitectonic areas 3b and 3a (heavy black lines). CS = coronal sulcus. Top row, two panels at left difference images showing absorbance increase at 1 s after onset of 25 Hz (left) or 200 Hz (2nd from left) stimulation, respectively, of the contralateral pisiform pad. At both frequencies stimulus amplitude = 100 m m, ISI = 10 s, probe diameter = 5 mm. Pixel darkness indicates magnitude of the stimulus-evoked increase in absorbance. Middle row, two panels at left images showing absorbance increase at 5 s after onset of 25 Hz (left) and 200 Hz (2nd from left) stimulation. Image orientation indicated by axes at bottom (L = lateral; A = anterior; M = medial; P = posterior). Figurine of forelimb at bottom right indicates (filled dot, arrow) stimulated skin site. Bottom row, 2nd panel from left plot at same magnification as OIS images that identifies regions in which absorbance values were largest (upper 5%) at 5 s after onset of 25 Hz (gray regions) or 200 Hz (black regions) stimulation. Panel at right: same contour plot described above, at higher magnification (in this panel the regions of maximal absorbance increase at 5 s after onset of 25 Hz and 200 Hz stimulation are outlined by gray and black lines, respectively). Top right drawing of coronal section depicting entry points (arrows with numbers) of microelectrode penetrations 1 3, and sites (lines perpendicular to each track) at which RA neuron spike discharge activity was recorded. Line to right of a track indicates a site where both 25 Hz and 200 Hz vibrotactile stimulation of the pisiform pad evoked above-background activity throughout the 5 s stimulus period. Line to left of a track indicates a site where 25 Hz, but not 200 Hz stimulation, evoked abovebackground activity throughout the 5 s period of stimulation. Gray shading shows approximate locus of region in which RA neurons exhibited a sustained increase in absorbance in response to both 25 Hz and 200 Hz stimulation of the pisiform pad. Sustained, abovebackground RA neuron spike discharge activity was detected consistently in this region during both 25 Hz and 200 Hz stimulation of the pisiform pad, whereas only 25 Hz stimulation yielded consistent responses in the regions traversed by penetrations 2 and 3. vibratory stimulation were quite different: whereas absorbance at all loci within the topographically appropriate region increased rapidly and tended to remain at levels well above background throughout the full 5 s period of 25 Hz stimulation, with 200 Hz stimulation absorbance increased at most of the same loci during the initial 1 s of stimulation, but then declined rapidly towards background. Several spatially discrete zones located within the relatively extensive region that responded to 25 Hz were exceptional, however, in that absorbance in these regions remained at above-background values throughout the full 5 s period of 200 Hz stimulation. Since absorbance and neuronal spike discharge activity are highly correlated, we anticipated that RA neurons in the regions of SI that exhibited a sustained increase in absorbance in response to both 25 and 200 Hz stimulation would respond with a sustained increase in MFR to both frequencies, and the responses of RA neurons in regions of SI which yielded a sustained increase in absorbance to 25 Hz, but not to 200 Hz, would be correspondingly different. For the experiment illustrated in Figure 7, the specific expectation was that RA neurons encountered in Penetration #1 (circle with crosshairs labeled 1 in panel at right) would be vigorously

11 Vibrotactile frequency coding 273 FIG U RE 8. PST and circular histogram pairs generated from the spike train data recorded at seven sites in penetration 2 of Figure 7. The top bottom sequence of PST/circular histogram pairs preserves the superficial-to-deep order of the sites at which RA neuron activity was studied. The first five recordings were made at successively deeper locations within layer III (between 0.25 and 0.6 mm from the pial surface); the last two in layer IV (between 0.65 and 0.85 mm). activated throughout the full period of either 25 or 200 Hz stimulation of the RF, whereas the RA neurons recorded in either Penetration #2 or Penetration #3 would exhibit well-sustained, abovebackground spike discharge activity in response to 25 Hz stimulation, but not to 200 Hz stimulation. The PST histograms in Figures 8 10 show the detailed form of the MFR responses to 25 Hz and to 200 Hz stimulation at each site at which SI RA neuron activity was studied during the three microelectrode penetrations shown in Figure 7; and the circular histograms in the same figures reveal the degree to which each response was entrained. These results are viewed as generally consistent with the prediction: that is, the response of almost every SI RA neuron/neuron cluster sampled in these three penetrations was robust, well maintained, and well entrained when the frequency of stimulation was 25 Hz, but vigorous, sustained, whereas entrained spike discharge activity to 200 Hz stimulation was especially evident when the microelectrode encountered neurons within a SI region that displayed a sustained increase in absorbance during 200 Hz (e.g., Penetration #1). The findings in Figures 8 10 (and the similar findings obtained in the three other experiments of this type) demonstrate that although RA neurons located in nearby SI cell columns may react in a very similar way to flutter stimulation, their responses to high-frequency stimulation of the same skin site can be very different. Specifically, these results support the idea that the focal, and usually multiple SI regions that exhibit increased absorbance throughout the full period of 200 Hz stimulation are occupied by RA neurons whose MFR remains substantially above background during 200 Hz stimulation (e.g., PSTs

12 274 B. L. Whitsel et al. FIG U RE 9. PST and circular histogram pairs generated from the data recorded at five sites in penetration 1 of Figure 7. Format as in Figure 8. The first two recordings were made in layer III (between 0.7 and 0.75 mm from the surface); the next in layer IV (at 0.8 mm); and the last two recordings in layer V (between 0.85 and 0.95 mm). in Penetration #1; Fig. 9), and that for the RA neurons in the much larger region that undergoes a transient increase in absorbance during 200 Hz stimulation of the same skin site MFR (1) increases transiently, (2) collapses to a near- or belowbackground level, (3) remains either at belowbackground or at near-background levels for as long as the 200 Hz stimulus continues, and then (4) elevates ( rebounds ) transiently upon stimulus termination (e.g., PSTs obtained in Penetration #2; Fig. 8). We sought to evaluate the above interpretation by studying the effects of a change in stimulus position within the RF on the spike train responses evoked from the same RA neuron by flutter vs vibratory stimulation. Our rationale was that a shift in the place of stimulation relative to the RF center should be accompanied by a systematic modification of the frequency dependency of the RA neuron response to vibrotactile stimulation. That is, it was expected that when a vibrotactile stimulus is delivered to the RF center the spike discharge activity that it evokes will be relatively well maintained and entrained to both flutter (6 50 Hz) and vibration ( Hz), and although the activity evoked by flutter stimulation will remain relatively well maintained and entrained as distance between the stimulus site and the RF center is increased, that evoked by vibration will not. The representative results shown in Figure 11 were obtained from an RA neuron in area 3b (lamina III). This neuron responded with vigorous spike discharge activity when each frequency of stimulation was applied to its RF center, and at each of the five stimulus frequencies (12, 15, , and 150 Hz) the rate of spike firing declined only modestly over the 2 s stimulus period. In contrast, when the

13 Vibrotactile frequency coding 275 FIG U RE 10. PST and circular histogram pairs generated from the data recorded at five sites in penetration 3 of Figure 7. Format as in Figure 8. The first recording was obtained in layer IV (0.96 mm from the surface); the others in layer V (between 1.0 and 1.3 mm). Interestingly, at two sites in layer III in this penetration (not shown) RA neuron MFR was observed to increase during 25 Hz stimulation and drop to below-background levels during 200 Hz stimulation, but substantial entrainment of spike discharge activity occurred in response to both stimulus frequencies. stimulator probe (5 mm diameter) was repositioned so that it was centered on a skin site 8 mm away from the RF center, only frequencies of stimulation within the flutter range (compare PST histograms on left and right) evoked vigorous and relatively wellsustained spike discharge activity. The data in Figure 11 (and the data from four other SI RA neurons studied in the same way) support our interpretation of the findings obtained in the OIS imaging/neurophysiological recording experiments. Taken together, the observations obtained in combined imaging/neurophysiological recording experiments and in studies of the effects of changes in the position of vibrotactile stimulation within the RF strongly suggest that: (1) 200 Hz stimulation evokes a vigorous and relatively wellsustained response from only a small subset of the RA neurons within the large SI region that responds vigorously and in a sustained manner to same-site 25 Hz flutter; and (2) the RA neurons that do respond in this way to 200 Hz stimulation are those in which the RF center coincided with the stimulated skin site. Results of statistical analyses Statistical analyses performed on the entire RA SI neuron and RA afferent databases confirmed and extend the exemplary findings illustrated in Figures Figure 12A summarizes results regarding the relationship between responsivity and stimulus frequency for both the RA afferents and SI neurons, using measures of responsivity for the and 0 2 s periods after stimulus onset. For the

14 276 B. L. Whitsel et al. FIG U RE 11. Effects of stimulus frequency on the activity evoked at two different sites in the RF of the same area 3b RA neuron. Responses were evoked using stimuli 2 s in duration applied using a 5 mm probe. The RF included contiguous portions of the first and second interdigital pads on the volar foot (macaque monkey). PSTs on left were obtained by stimulating at the RF cen ter (the distal tip of the first interdigital pad), PSTs on right were obtained by stimulating at an off-rf cen te r site (at the center of the second interdigital pad). Center-tocenter distance between the two sites was 8 mm. afferents, average responsivity is less than one spike per cycle throughout the entire range of stimulus frequencies, and although there appears to be some decline of responsivity at the higher frequencies, these differences are small and statistically insignificant. Compared to RA afferents, SI neurons showed generally much higher responsivities, and a sharp decline in responsivity concentrated primarily in the flutter range of frequencies. In Figure 12A, which was constructed using only the data obtained in runs in which stimulus amplitude was 100 m m, this effect is highly significant (for s, 6 F 168 = 23.3, p < 10 6 ; and for 0 2 s, 6 F 168 = 21.2, p < 10 6 ). Furthermore, orthogonal-polynomials analysis reveals that the overall effect is dominated by the linear component of the downward trend. The same pattern appears in the complete RA SI neuron database, and also in separate analyses of its major subgroupings, in particular in cats vs monkeys, and in SUR vs MUR. Another noteworthy feature of Figure 12A is that mean responsivity of both RA afferents and SI neurons over the period 0 2 s after stimulus onset is consistently lower than mean responsivity measured during the period s. This rapid drop in responsivity is, in fact, independently significant, overall, for both afferents and SI neurons. For the RA afferents the mean difference is spikes/cycle, t 102 = 7.43, p < 10 6, while for SI neurons it is 0.462, t 323 = 14.1, p < Moreover, as shown in Figure 12B, for SI neurons but not RA skin afferents the rapid drop in responsivity is itself highly frequencydependent, again being more pronounced at lower stimulus frequencies. The overall ANOVA is highly

15 TA B L E 2. Association between non-maintenance of MFR ( collapse ) and stimulus frequency Vibrotactile frequency coding 277 Stimulus frequency (Hz) Collapse No Yes TA B L E 3. Number and proportion of runs in which the r 1, r 2, or r s measure yielded the highest value of RA neuron entrainment r 1 r 2 r s Total n * % *For 2 of the 326 runs r 1 = r s significant ( 6 F 168 = 19.5, p < 10 6 ), and this is again driven primarily by the linear component of the monotonic decrease with increasing stimulus frequency. To characterize the overall statistical association between vibrotactile stimulus frequency and the tendency (illustrated in Figs. 5, 6, and 9) for the mean firing rate of many, but not all SI neurons to either approach or fall below background levels ( collapse ) shortly after the onset of stimulation, every run of SI neuron activity was categorized by eye as displaying or not displaying that phenomenon, and the results cross-tabulated against stimulus frequency. As shown in Table 2, the overall association is highly significant 2 (x 6 = 141.2, p < 10 6 ), and it takes the specific form of a progressive increase in the proportion of units that exhibit mean firing rate collapse with increasing stimulus frequency (Cochran s test of linear trend = 109.3, df = 1, p < 10 6 ). Figure 12C shows that while average RA afferent MFR increases significantly and nearly linearly with stimulus frequency (as implied by the periodicity hypothesis Talbot et al., 1968), average SI RA neuron MFR increases only up to about 25 Hz, with high variability, and either remains level or declines at higher frequencies. The average effects of vibrotactile frequency on SI RA neuron and RA afferent entrainment are summarized in Figure 12D. RA afferent entrainment is uniformly high and stable to at least 100 Hz, and declines only slightly at 150 Hz. The overall ANOVA for frequency is significant ( 4 F 83 = 5.3, p < 0.001), a result due entirely to the slight droop in r b at 150 Hz, and to the extremely low variabilities associated with RA afferent r b at all stimulus frequencies. Variances for SI RA neurons are almost ten times larger than the variances for RA afferents (itself a statistically significant difference), but even this relatively high variability cannot mask the large and systematic dependence of SI RA neuron entrainment on stimulus frequency. For the SI neuron entrainment vs frequency data shown in Figure 12D, 6 F 168 = 43.4, p < 10 6, and this result is again due primarily to the linear component of the downward trend. As in the case of the effects of stimulus frequency on responsivity, the same basic patterns appear both in the complete database and in separate analyses of its major subgroups such as cats vs monkeys and SUR vs MUR. In particular, the relationship is of the same form for the SUR data by itself, but statistically it is even stronger because the variance associated with SURs is typically less than half the variance associated with MURs. Figure 12D also makes it apparent that for SI RA neurons mean r b approaches zero at high stimulus frequencies, and the relatively shapeless circular histograms typically obtained at such frequencies (Fig. 2, left; Figs. 5 and 6) likewise provide little assurance that meaningful degrees of entrainment are actually present. Surprisingly, however, even at 150 Hz average SI RA neuron r b is significantly greater than zero (t 1 8 = 7.71, p < ). Moreover, the power spectra of the resampled and time-averaged spike trains provide clear, direct and compelling evidence that partial frequency-following responses do occur in some SI RA neurons at frequencies of 100 Hz and higher. Figure 13, for example, shows the spectra of MUR recordings obtained during two microelectrode penetrations (#1 and #2) that traversed area 3b in monkey SI cortex. In both cases sinusoidal stimuli of 100 m m peak-to-peak amplitude were presented at five different frequencies in randomized order. At every frequency a sharp peak in power is evident at the stimulus frequency and at its low harmonics (multiple harmonics are especially apparent at the lower frequencies), while the total harmonic response becomes generally smaller at higher frequencies, in parallel with the aggregate statistical behavior shown in Figure 12D. These peaks in power, which are well above the noise background, can derive only from subsets of spikes that occur at consistent positions in the stimulus cycle, and therefore they reflect corresponding degrees of entrainment in the neuronal response. 1 Extracellular microelectrode recording observations paralleling the OIS imaging observations described in the Introduction were made in five different experiments, yielding 62 recordings of SI RA neuron activity evoked by 25 Hz vs 200 Hz