A 64-channel electrode array was used to study the spatial and temporal characteristics of fast (>200 Hz) electrical oscillations recorded from the surface of rat cortex in both awake and anesthetized animals. Transient vibrissal displacements were effective in evoking oscillatory responses in the vibrissa/barrel field and were tightly time-locked to stimulus onset, coinciding with the earliest temporal components of the coincident slow-wave response. Vibrissa-evoked fast oscillations exhibited modality specificity and were earliest and of largest amplitude over the cortical barrel, which corresponded to the vibrissa stimulated, spreading to sequentially engage neighboring barrels over subsequent oscillatory cycles. The response was enhanced after paired-vibrissal stimulation and was sensitive to time delays between movement of separate vibrissae. These data suggest that spatiotemporal interactions between fast oscillatory bursts in the barrel field may play a role in rapidly integrating information from the vibrissal array during the earliest cortical response to somatosensory stimulation.
Rodents exhibit a collection of large facial vibrissae organized into a regular array on the snout which are under voluntary control of the mystacial musculature. The vibrissae play a central role in exploratory behavior, during which they are “whisked” to transiently contact objects in the environment rapidly and repeatedly (Welker 1964). The vibrissae are not merely a crude probing apparatus but rather appear to define a functionally continuous sensory organ with sensitivity similar to that of the primate fingertip (Carvell and Simons 1990). This is reflected in the disproportionately large region of rodent somatosensory cortex devoted to the face, in which the representation of the vibrissae is segregated into distinct cellular aggregates or “barrels,” which collectively mirror the topology of the vibrissae on the contralateral mystacial pad (Woolsey and Van der Loos 1970) (see also Fig. 1). The distinctive cellular organization of barrel cortex, and punctate nature of its associated receptive surface, facilitates the study of spatiotemporal interactions of activity within the barrel field arising from controlled displacement of the vibrissae and has made the rodent somatosensory system a popular model of integrative sensory information processing.
Some of the earliest electrophysiological studies of barrel cortex examined the response of individual units to controlled displacement of the vibrissae (Armstrong-James 1975; Simons 1978; Welker 1971) and a focus on unit activity persists in most studies published to-date (for recent reviews seeJones and Diamond 1995). Yet, transient stimulation of the vibrissae also results in summed extracellular postsynaptic potentials recordable at the cortical surface as a series of slow waves comprising the somatosensory-evoked potential (SEP) complex (Di and Barth 1991; Di et al. 1990). Epicortical mapping, performed with multielectrode arrays spanning the extent of the barrel field, indicates that the SEP evoked by single-vibrissa stimulation is localized sufficiently to identify activity in separate barrels (Di and Barth 1991) and may be used to study spatiotemporal interactions between barrels (Jones and Barth 1997). The vibrissa-evoked SEP consists of an initial positive/negative slow wave (∼25-ms duration) followed by a slower biphasic wave (∼100- to 300-ms duration), which have been designated P1/N1 and P2/N2, respectively, to indicate their polarity and sequence of occurrence (see Fig.1). This response reflects a stereotyped sequence of excitation and inhibition within sensory cortex (Steriade 1984), and the SEP exhibits a similar morphology in a number of species, including man (Allison et al. 1989a,b).
Studies using wide bandwidth recording and digital high-pass filtering of the SEP waveform have revealed a series of previously unrecognized fast (>200 Hz) small-amplitude deflections superimposed on the P1/N1 component (Eisen et al. 1984; Green et al. 1986; Maccabee et al. 1983). Such oscillations are of interest because they occur at the earliest stages of information processing in somatosensory cortex and reflect highly synchronized interactions between cell populations that proceed at a rate that is an order of magnitude faster than that reflected by SEP slow waves. Fast oscillations have been observed extracranially in human somatosensory cortex in response to stimulation of the median and ulnar nerves (Curio et al. 1994a,b; Emori et al. 1991; Gobbelé et al. 1998;Hashimoto et al. 1996; Yamada et al. 1988) and intracranially in rat cortex after electrical stimulation of somatosensory thalamus (Kandel and Buzsaki 1997). The functional anatomy of the phenomenon, however, is still poorly understood.
Our laboratory has developed methods for high-resolution mapping of surface potentials that we have used to study the spatial and temporal organization of both spontaneous and stimulus-evoked electrical activity in somatosensory cortex (Di and Barth 1991;Jones and Barth 1997; MacDonald and Barth 1995). The purpose of the current study was to apply these techniques to investigate fast oscillations. Specifically we sought to determine an adequate sensory stimulus for generation of fast oscillations in barrel cortex, determine whether the phenomenon exhibits modality specificity and, within somatosensory cortex, a somatotopic organization, and examine the horizontal spread of fast oscillations within the barrel field arising from controlled stimulation of individual vibrissa, which might be suggestive of their role in the integrative processing of afferent information in the rat somatosensory system.
Seven adult male Sprague-Dawley rats (375–400 g) were studied. Two animals were prepared for chronic recording, in which a unilateral craniectomy was performed under ketamine HCl (100 mg/kg) and xylazine (15 mg/kg) anesthesia to expose a large region of cortex in the right hemisphere. Stainless steel reference and ground screws were placed in the contralateral frontal bone and just posterior to lambda, respectively, and a short length of aluminum tube was positioned over the craniectomy site and secured to the skull using dental cement. The tube was sealed with a tight-fitting plug which provided back-pressure against the cortex, the scalp incision was closed and covered with a broad-spectrum antibiotic ointment, and the animals were allowed to recover for 24 h.
In the remaining five animals, a craniectomy was performed as above with the exception that the dura was reflected at the craniectomy site. This allowed for improved spatial resolution of subsequent surface mapping not possible in the chronic preparations because removal of the dura degrades the postsurgical viability of exposed cortex. After surgical manipulation, anesthetic levels were maintained throughout the remainder of experiment such that the corneal reflex could barely be elicited. The cortex was moistened regularly using warm physiological saline, and body temperature was maintained using a regulated heating pad.
Vibrissa were stimulated using a laboratory-built apparatus that transformed computer generated waveforms into silent dorsal displacements of a short length of 18-gauge hypodermic tubing. Vibrissae were tied together and collectively attached to the stimulator for whole-field stimulation. In paradigms that required single-vibrissa stimulation, the vibrissae were clipped to ∼2 cm, and 1 cm of the length was inserted into the stimulator arm where it could slide freely to avoid pulling. Stimulation was monitored using a dissecting microscope, and the magnitude of displacement was kept sufficiently small so that the stimulus did not disturb the mystacial pad or adjacent vibrissae. Auditory-evoked potentials (AEP) were obtained using click stimuli (0.3-ms monophasic pulses) presented contralaterally using a piezo-electric speaker positioned ∼10 cm lateral to the animal’s head.
A flat array of 64 silver electrodes configured in an 8 × 8 matrix (tip diameter, 100 μm; interelectrode spacing, 500 μm) was placed over parietotemporal cortex, and a silver ball reference electrode was secured in a burr hole in the contralateral frontal bone. SEPs evoked by single-vibrissa stimulation were used to align the electrode array consistently across animals (seeresults and Fig. 1 for additional details). Field potentials were analogue filtered (band-pass cutoff = −6 dB at 3–300 Hz, roll-off = 5 dB/octave) and digitized at 1 kHz in recording conditions in which all 64 channels were used. In conditions in which a subset of array channels was used, analogue filters were set to a wide bandwidth (band-pass cutoff = −6 dB at 3–2000 Hz, roll-off = 5 dB/octave) and the sampling speed increased to 5 kHz (memory limitations of the data-acquisition system did not permit sampling of the full complement of electrode channels at rates >1 kHz).
Postrecovery recording in unanesthetized animals was carried out under light restraint. Animals were placed in an acrylic tube to which they had been previously habituated, and their head immobilized using a steel post that had been secured to the skull during surgery. The plug was removed from the aluminum tube covering the craniectomy site, and the recording array introduced into the tube, brought into contact with the dura, and secured. Twenty 1-s trials were recorded during each of the following conditions: calm immobility (to be used as baseline) whisking without contacting an object, and passive stimulation of the contralateral or ipsilateral vibrissae and mystacial pad, in which a nonconducting rod was brought into contact with the mystacial pad by the experimenter and swept continuously through the vibrissae in a rostral-caudal direction. Originally, a stimulation condition was planned in which the animal would repeatedly contact an object using the vibrissae on command; however, this proved impractical and passive stimulation (condition 3) was substituted. A final stimulation paradigm then was used that was designed to approximate in a controlled fashion the afferent conditions produced by exploratory whisking. This consisted of a 10-Hz large-amplitude (∼2 mm) sinusoidal displacement of the vibrissae [10 Hz was chosen as an approximate upper limit of average whisking frequencies observed in behavioral studies (e.g.,Carvell and Simons 1990; Welker 1964)]. Transient contact of the vibrissae with an object then was modeled by adding brief monophasic pulses to the background sinusoidal displacement (duration, 250 μs; amplitude, approximately one-half of sinusoid), which were timed to coincide with the peaks of the displacement. Ten 1-s trials of each stimulus condition were recorded for both contralateral and ipsilateral vibrissae.
The sinusoidal and sinusoidal + transient whole-field stimulation paradigm described in the preceding text was repeated in the anesthetized animals. In addition, to evaluate possible interactions of fast oscillations within the barrel field, eight channels of the array spanning row C of the barrel field were selected for 5-kHz sampling after pulse displacement of the most caudal vibrissa (C1), a more rostral vibrissa (C4), and a combination of both (C1 + C4). One hundred trials of each stimulus condition were recorded in random order. Finally, to investigate the modality specificity of fast oscillations, the array was moved laterally and caudally to allow simultaneous recording from auditory and posterior-lateral somatosensory cortex. Fast oscillations were averaged (n = 100) from electrodes located over the A1 barrel and over the center of primary auditory cortex after stimulation of the A1 vibrissa and in response to click stimuli.
Power spectral density (PSD) of the response to the three stimulus conditions in the unanesthetized animals (baseline, whisking, and vibrissal stimulation) was estimated as the average of a 128-point overlapping fast Fourier transform (FFT) window applied to digitally high-pass filtered data (100–2,000 Hz) of the 20 trials of a given condition (overlap, 64 points; frequency resolution, 39 Hz/coefficient). Power was normalized to the largest spectral peak present across the three conditions. Peak-locked averages of the response after whole-field sinusoidal and combined sinusoidal + transient displacement in both unanesthetized and anesthetized animals were constructed using 500 points of wideband data centered at the positive peaks of the sinusoid over all trials in a given animal (10 1-s trials at 10 Hz; n = 100). Averages of individual- and paired-vibrissa displacement in anesthetized animals were computed off-line from 100 trials of a given stimulus condition that had been spooled to disk. Records were edited visually, and trials containing ketamine spikes or other spontaneous discharges were omitted from the average.
Figure 1 summarizes the procedure used to align the recording array consistently across animals. The P1/N1 component of the SEP evoked by single-vibrissa stimulation (Fig. 1 C) was of largest amplitude at electrodes located in close proximity to the corresponding cortical barrel, and stimulation of the C1 and C4 vibrissae provided two amplitude foci which were used to position the array over somatosensory cortex (Fig. 1, B and D,respectively). Traces in these figures are superimposed over a template adapted from previous cytochrome oxidase (CO) histology in animals of similar weight (Jones and Barth 1997) and depict the approximate areal locations of the barrel field and primary auditory cortex in relation to the 64 electrode locations of the recording array (Fig. 1 A, dots).
Figure 2 depicts typical single trial data (Fig. 2 A) and averaged power spectra (Fig. 2,B and C) of baseline, whisking, and passive stimulation conditions from an electrode positioned at the center of the barrel field in unanesthetized animals. Here it may be seen that there was a general increase in power at all frequencies during whisking without object contact and during passive vibrissal stimulation. However, a spectral peak in the 300- to 400-Hz bandwidth was observed during manual stimulation (Fig. 2, B andC, ↓) that was not evident in either the baseline or free whisking conditions. This peak was absent when the ipsilateral vibrissae were stimulated (Fig. 2 C, · · · ), suggesting that the phenomenon did not merely reflect a change in general arousal produced by the stimulation. Note that the apparent spectral peaks at ∼150 Hz in these figures result from high-pass filtering of the typically large low-frequency power observed in behaving animals (Jones and Barth 1997).
Although the above results suggested an association of high-frequency activity with stimulation of the vibrissae, the use of poorly controlled manual stimulation made the phenomenon difficult to study in detail. To more clearly delineate the source of fast oscillations, the unanesthetized animals were restrained and a second stimulation paradigm used that was intended to approximate in a controlled fashion the afferent conditions of free whisking and those that occur during the palpitation of an object. This was accomplished by collectively attaching the vibrissae to a stimulator and impressing on them either a 10-Hz sinusoidal movement as a model of free whisking or a combination of the sinusoidal motion and a transient pulse displacement added at the forward peaks of the motion as a model of transient contact of the vibrissae with an external object. As shown in Fig.3, sinusoidal displacements of the vibrissae in the unanesthetized animals were ineffective at generating either high-frequency activity or a stimulus-locked slow-wave response (Fig. 3 A, left). This was confirmed in peaked-locked averages in which slow activity consisted of a poorly defined, low-amplitude biphasic wave, with an absence of stimulus-related fast activity. (Fig. 3 B, left). In contrast, combined sinusoidal and pulse displacement resulted in large-amplitude slow waves that were time-locked to the onset of the transients accompanied by a burst of high-frequency oscillations (Fig. 3, A and B, right). Fast oscillations were evident in peak-locked averages, both in the high-passed traces and as small-amplitude ripples superimposed on the rising phase of the slow-wave activity (Fig.3 B, →). Frequency analysis of the averaged oscillatory activity indicated that its center frequency was similar to that of the spectral peaks observed in Fig. 2. Thus while the continuous motion of the experimenter-held stimulator used in the unrestrained animal was ineffective in generating distinct and stimulus-locked slow-wave activity (Fig. 2 A, bottom), transient contact occurring as the stimulator brushed through the vibrissal array apparently produced sufficient oscillatory activity to give rise to identifiable peaks in the spectral data (Fig. 2, B and C).
Results similar to those described in the preceding text also were obtained in the anesthetized animal. Combined sinusoidal and pulse displacement of the vibrissae resulted in stimulus-locked bursts of fast oscillations, whereas sinusoidal displacement alone was ineffective (Fig. 4 A). The phenomenon repeated consistently across all four animals. Spectral analysis of the response indicated that the average center frequency of the oscillatory response was ∼300 Hz. In one animal, the surface array was removed and a stainless steel monopolar microelectrode (World Precision Instruments, PTM23B05KT; exposed tip diameter, 2–3 μM; impedance, 0.5 MΩ) was lowered perpendicular to the cortical surface in 150-μm increments, with averages (n = 50) of the evoked fast oscillations computed at each depth (Fig. 4 B). These laminar recordings revealed that fast oscillations were of cortical origin and exhibited a polarity reversal at a depth of ∼600 μm and are consistent with results obtained in a previous study byKandel and Buzsaki (1997).
To investigate the modality specificity of evoked fast oscillations, the electrode array was repositioned to a more lateral and posterior location for simultaneous recording from both somatosensory and auditory cortex. Consistent array positioning was achieved by monitoring the spatial distribution of SEPs evoked by stimulation of the A1 vibrissa and AEPs produced by clicks. Two electrode locations were chosen for subsequent wide bandwidth recording of fast oscillations, one positioned over the A1 barrel and a second centered over primary auditory cortex. Transient stimulation was sufficient to elicit fast oscillatory activity in either modality with a high degree of specificity (Fig. 5). Pulse displacement of vibrissa A1 evoked large-amplitude fast oscillations from its corresponding barrel in all animals (Fig. 5 A, top) and no response in auditory cortex (Fig. 5 A, bottom). Clicks were also effective in evoking large-amplitude fast oscillations. However, these were constrained to auditory cortex (Fig. 5 B, bottom) and generated no response in the barrel field (Fig.5 B, top). In both modalities, the poststimulus latencies of evoked oscillatory bursts varied by ∼1–2 ms between animals but were sufficiently stable within each animal to permit averaging.
With the array returned to its original position over the barrel field, a subset of eight electrodes in line with the barrels of row C were chosen to examine the somatotopic organization of fast oscillations in somatosensory cortex (Fig.6 A). Pulse displacement of vibrissa C4 or C1 produced fast oscillations of largest amplitude over the corresponding cortical barrel (Fig. 6, B andC, respectively). However, the oscillatory response also spread rapidly within the parent row. This was particularly apparent with stimulation of C1 (Fig. 6 C), with spread most evident in the rostral direction, as indicated by a shift in latency of the maximum peak to peak amplitude (represented by dark vertical bars) of the oscillatory burst at more rostral recording sites. Simultaneous stimulation of both C1 and C4 evoked fast oscillations along the entire barrel row (Fig. 6 D), which appeared larger in amplitude than would be expected from the sum of the single vibrissa responses. To examine this issue in more detail, a linear model of the combined response was constructed as the sum of the responses to C1 and C4 stimulation (Fig. 6 E, shown superimposed on the observed response in Fig. 6 F). This revealed that the multivibrissa response was indeed larger than the sum of single vibrissa responses and, furthermore, that the differences between model and data evolved over time. During the initial cycles (3–6 ms), the data and model were nearly identical (Fig. 6 F, ↓). However, the oscillatory response increased disproportionately over subsequent cycles, with the amplitude of the combined response at the end of the burst approximately twice what would be expected if it were the linear sum of the single vibrissa responses (Fig. 6 F, dashed box).
To condense the large amount of graphic data required to summarize the results of paired-vibrissae stimulation in the remaining animals, pseudocolor topographical maps were constructed from oscillatory data, as demonstrated in Fig. 7. Here, the line traces of Fig. 6 C are reproduced at top, from which a 20-ms data window is extracted and used to generate a topographic map with an expanded horizontal (time) scale, and the color range of which is normalized to the maximum peak amplitude of the response (Fig. 7 B). The caudal-to-rostral spread of the response is reflected by the upward extension of the colored areas of the map as the response develops. This same map is shown in Fig.8 A, left, with a superimposed line reflecting a rostral spread of the fast oscillations from the C1 barrel at a speed of ∼125 μm/ms. Similar propagation of the C1 response was observed in all four animals (Fig. 8 A; R1–R4; row 2). Caudal propagation of fast oscillations evoked by C4 stimulation was also detectable in these animals but was less pronounced (Fig. 8 A, row 1). The response to combined stimulation of C4 + C1 (Fig. 8 A, row 3) was consistently enhanced over a linear model computed from the responses to single-vibrissa stimulation (Fig. 8 A, row 4). This was particularly pronounced in R4, where individual vibrissa stimulation barely elicited a response.
The response enhancement observed after multivibrissal stimulation suggested the possibility that neuronal interactions within intervening barrels served to reinforce oscillatory bursts when neighboring vibrissae were contacted simultaneously, raising the question of whether the interaction pattern could be manipulated by asynchronously stimulating two vibrissae. This was investigated in a fifth animal, in which the response to paired C1–C3 stimulation was recorded using inter-stimulus intervals of 0.0, 1.6, and 3.3 ms (Fig. 8 B). The latter values of the interstimulus interval correspond to 180 and 360° phase shifts of a 300-Hz oscillation and were chosen on the assumption that two propagating sources of fast oscillations might exhibit effects that were additive when aligned in phase and tend to cancel when they were not. As in previous results, simultaneous displacement of the vibrissae resulted in an enhanced response (Fig.8 B, left). However, when the interstimulus interval was increased to 1.6 ms, the response was attenuated and showed little enhancement over the linear model (Fig. 8 B, center). Here, the response centered at the C1 barrel was shifted ∼180° out of phase with respect to the that of the C3 barrel. When the interstimulus interval was further increased to 3.3 ms, the C1 and C3 responses were again in phase and resulted in a recovery of the response (Fig.8 B, right) with enhancement over the linear model similar to that observed after simultaneous stimulation.
These results indicate that fast oscillations are a characteristic feature of the evoked response in sensory cortex of both the awake and the ketamine-anesthetized rat. Fast oscillations are specific to the sensory modality activated and time-locked to stimulus onset, coinciding with the earliest P1 and N1 components of the slow-wave response. In the barrel field, fast oscillations exhibit a somatotopic organization and are earliest and of largest amplitude over the cortical barrel, which corresponds to the vibrissa stimulated, spreading to sequentially engage neighboring barrels over subsequent oscillatory cycles. This rapid spread within somatosensory cortex results in an enhanced response during multivibrissal stimulation that is sensitive to time delays between movement of the separate vibrissae.
Field potentials recorded using the relatively large (100 μm) and low-impedance (∼1 kΩ) electrodes as performed here reflect the sum of volume conducted currents from both nearby and distant cellular populations. However, it may be assumed that surface potentials recorded intracranially are dominated by postsynaptic activity on apical dendrites of populations of pyramidal cells directly beneath the electrode, which share a common orientation and coherent temporal activation (Creutzfeldt and Houchin 1974). This is particularly true in preparations where the dura mater has been removed and the array is in direct contact with the cortical surface. Previous laminar recordings of evoked field potentials in the rat barrel field (Barth and Di 1990; Di et al. 1990) have established that the slow wave components of the surface SEP complex are produced by synchronized activation of both supra- and infragranular pyramidal cells. The polarity of the surface slow waves reverses in the depth, as would be expected from a vertically oriented core conductor approximating the geometry of parallel apical dendrites. The laminar profile of fast oscillations evoked by vibrissa stimulation in the present study is similar to that of the slow wave complex, reversing polarity at a depth of ∼600 μm, indicating that they also are produced intracortically and involve synchronous postsynaptic potentials in local populations of pyramidal cells. This conclusion is consistent with recent laminar recordings performed in somatosensory cortex of the unanesthetized rat during electrical stimulation of the thalamic ventrolateral and ventral posterolateral nuclei (VL and VPL) (see Fig. 11 in Kandel and Buzsaki 1997). Thalamic stimulation evokes small-amplitude fast oscillations (>200 Hz) that superimpose on the early slow wave components of the cortical response, reverse polarity in the granular layer, and are phase-locked to bursts of multiunit activity localized to layer V pyramidal cells.
The similarity in frequency and laminar distribution of sensory-evoked fast oscillations to those evoked by thalamic stimulation suggest a common cortical generator for both phenomena. However, a cellular correlate of the vibrissa-evoked fast oscillations presented here has not been reported in intracellular or extracellular studies of unit activity in barrel cortex. This may be due to several reasons. The combined high-frequency and slight phase variability of the phenomenon may tend to obscure its rhythmicity in cumulative peristimulus spike histograms, which typically are used in extracellular unit studies. In addition, fast oscillations are affected by anesthesia regime and in preliminary studies were found to be attenuated or extinguished by many commonly used anesthesias including halothane, pentabarbitol, and urethan. Unanesthetized or lightly ketamine/xylazine anesthetized preparations were required for observation of fast oscillations reported here. Finally, fast oscillations may reflect primarily subthreshold activity. In the hippocampus, simultaneous intracellular and field potential recording has demonstrated a class of interneurons that fire bursts of action potentials time-locked to the peaks of fast (∼200 Hz) oscillations in local field potentials (Buzsaki et al. 1992). Action potentials recorded from pyramidal cells postsynaptic to the interneurons also are locked to the oscillations, but these cells usually fire only one spike per oscillatory cycle and often not at all. If similar circuitry were involved in barrel cortex, the phenomenon might not be apparent in extracellular unit recordings, which largely reflect the activity of pyramidal cells as interneurons typically comprise only a small fraction of units sampled in vivo (Simons et al. 1989). Fast oscillations also may escape detection in the increasing number of intracellular results based on in vitro techniques, as it is known that some types of oscillatory activity observed in vivo do not persist in slice preparations (see comments in Gray and McCormick 1996).
The spatiotemporal organization of high-frequency oscillations recorded epicortically, and the stimulus parameters required to evoke them, suggests that they reflect activity of circuitry related to the processing of afferent information carried by specific thalamocortical projections and are not merely an index of arousal. This is supported by several lines of evidence. First, the phenomenon is modality specific. As shown in Fig. 5, simultaneous monitoring of electrode positions over somatosensory and auditory cortex demonstrate that an adequate stimulus generates robust oscillations that are confined to the appropriate cortical area. This would not be expected if the phenomenon represented activation of brain areas associated with general arousal. Second, within the somatosensory modality, stimulation of the ipsilateral vibrissae fails to produce high-frequency oscillations (Fig. 2), implicating participation of the lateralized lemniscal pathway. Furthermore individual vibrissa stimulation demonstrates that the oscillations have a somatotopic organization that mirrors that of barrel cortex (Fig. 6). Afferents from the vibrissae pass though the ventrobasal thalamus, which is itself somatotopically organized (Van der Loos 1976), devoid of interneurons (Barbaresi et al. 1986), and the cortical projections of which terminate in the overlying barrels in a homologous fashion (Land et al. 1986). Thus the onset of fast oscillations at a barrel after displacement of its principal whisker represents the arrival of information that has remained largely segregated as it passed through the ascending pathway. Last, the adequate stimulus for the generation of fast oscillations in barrel cortex is not merely movement of the vibrissae, but rapid and transient displacement (Figs.3 and 4). Fast oscillations are not observed during whisking (Fig. 2) nor during an imposed sinusoidal displacement of the vibrissae (Figs. 3and 4), whereas pulse displacement of the vibrissae, either in isolation in combination with sinusoidal background motion, results in robust bursts of fast oscillations. Such abrupt stimulation mimics the transient contact of vibrissae with objects in the environment that occurs during exploratory behavior, and through which the rodent rapidly extracts behaviorally relevant features of an environmental cue (Carvell and Simons 1990).
Although fast oscillations originate in a focal area of the barrel field after single vibrissa contact, they spread to several adjacent barrels at speeds of ∼125 μm/ms. Assuming an average center to center distance of ∼500 μm between the major barrels, the time required for propagation of fast oscillations from the principal barrel to its nearest neighbor is 4 ms. This interbarrel latency shift is in accord with measurements of peak unit responses recorded in unanesthetized animals after principal versus adjacent whisker deflections (3.6-ms difference) (Simons et al. 1992). Given that the conduction velocity of horizontal fibers in barrel cortex has been estimated at ∼250 μm/ms (Mason et al. 1991; Simons 1995), a rate of 125 μm/ms suggests that the spread of fast oscillations reflects polysynaptic activity. Anatomic evidence indicates that direct projections between barrels do not occur (Hoeflinger et al. 1995) but rather interbarrel communication is mediated by polysynaptic pathways in supra- and infragranular layers, the latter of which contains the putative pyramidal cell population participating in the fast oscillatory response. Furthermore horseradish peroxidase studies in mouse barrel have demonstrated that labeling is preferential along the row of the barrel injected (Bernardo et al. 1990a,b), providing at least an anatomic substrate for the rostral and caudal spread of fast oscillations observed in the present study.
The cellular mechanism underlying facilitation of fast oscillations in response to paired-vibrissae stimulation is not known. When two vibrissae within a row are stimulated simultaneously, the beginning of the fast oscillatory burst (∼3–6 ms) is similar to what would be predicted from the sum of the responses to each vibrissa stimulated independently (Fig. 6 and 8). At burst initiation, fast oscillations are constrained to the respective barrels of the stimulated vibrissae. Within the next several milliseconds, fast oscillations spread beyond these borders and begin to interact within the barrel field. The effect of simultaneous vibrissa stimulation is to produce an oscillatory response in the interaction zone that is consistently larger than the sum of the two individual vibrissa responses. The increase develops over the time course of the response, and extends over multiple barrels within the activated row. Enhancement is also phase sensitive, with incremental 1.6-ms increases in the interstimulus interval resulting in attenuation then recovery of fast oscillations.
The simplest explanation for this phenomenon may be adapted from the work in the hippocampus by Busaki, who has proposed that fast oscillations are generated by the activation of bursting interneurons which project inhibitory synapses on the apical dendrites of local pyramidal cells (Buzsaki et al. 1992; Ylinen et al. 1995) If similar circuitry is involved in barrel cortex, the nonlinear response increase may reflect redundant horizontal convergence of excitatory inputs arising from each of the stimulated vibrissa barrels onto the dendritic fields of these inhibitory interneurons within the interaction zone. Multivibrissal input that results in the arrival of excitatory volleys that are in phase and thus temporally summate would bring to threshold an increased number of redundantly targeted inhibitory interneurons compared with the afferent volley resulting from single-vibrissal or out-out-phase multivibrissa stimulation. In this way, facilitation during multivibrissa stimulation may be similar to subliminal fringe phenomena typically associated with synchronized efferent projections to motorneuron pools. Because of their closed field geometry, extracellular potentials produced by the interneurons would not be recorded by surface electrodes used in the present study. Yet their increased burst firing would be expected to produce a marked increase in the amplitude of rhythmic postsynaptic potentials on targeted pyramidal cells, which is reflected in the surface record, with a concomitant decrease in their overall unit activity. This hypothesis is consistent with results observed bySimons (1985), who found that the principal effect of simultaneous displacement of two adjacent vibrissae is a decrease in unit firing in pyramidal cells, and furthermore that the firing rate is less than what would be expected by adding the unit responses when the two vibrissae were stimulated independently.
The emerging functional role of barrel cortex suggested by recent behavioral studies is in the integration of multivibrissae afferent information required by complex tasks such as active touch (Guic-Robels et al. 1992; Hutson and Masterson 1986) and orienting and object recognition (Brecht et al. 1997). Multibarrel integration occurring >15 ms poststimulus is mediated almost entirely by slow inhibitory processes (Simons 1985) and is associated with the falling phase of the N1 component of the SEP and subsequent P2/N2 components extending to several hundred milliseconds (Barth and Di 1990). In contrast, fast oscillations coincide with the earliest cortical response to vibrissa displacement. They are superimposed on the P1 and rising phase of the N1 components of the SEP, which typically are associated with the rapid spread of intra- and interbarrel excitation (Barth and Di 1990), and possibly, as noted earlier, brief and rhythmic bursts of inhibition. Fast oscillations are most sensitive to transient displacement of the vibrissae and thus conceivably could provide a mechanism of accurately marking stimulus onset and rapidly propagating this information within the barrel field as a phase-encoded oscillatory burst of 10- to 15-ms duration. Phase-sensitive interactions between bursts might reflect the processing of spatiotemporal displacement patterns of multiple vibrissae which, in the context of exploratory whisking, may extract behaviorally relevant features of the object under exploration. As the participation of oscillatory phenomenon in sensory information processing is currently a topic of some debate, the validity of these claims deserves further exploration and experimental verification.
This research was supported by Whitehall Foundation Grant S-97-06 and National Institute of Neurological Disorders and Stroke Grant 1 R01 NS-36981-01A1.
Address for reprint requests: D. S. Barth, Dept. of Psychology, University of Colorado, Campus Box 345, Boulder, CO 80309-0345.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 1999 The American Physiological Society