Rats perform texture discrimination during tactile exploration with their whiskers with high spatial and temporal precision. Although the peripheral mechanoreceptors provide tactile information with exquisite temporal resolution, physiological studies have suggested that this information might be lost at the cortical level. To address this discrepancy, multiunit and single-unit recordings were performed in the barrel cortex of isoflurane-anesthetized rats using continuous sinusoidal vibration of single whiskers at 15–700 Hz. In multiunit recordings, sustained phase-locked responses occurred up to vibration frequencies of 700 Hz, and 1:1 stimulus locking was observed up to 320 Hz. Wide-band responses of multiunits showed frequency encoding between 20 and 320 Hz. The discharge rates were not different for stimuli in the low- and high-frequency ranges, but they were significantly lower for non-phase-locked responses to high-frequency vibration. Response adaptation was present in only 25% of the cases, whereas in the majority of cases, entrainment to the vibratory frequency remained constant or even increased with stimulus duration. Increased entrainment to high-frequency stimulation was accompanied by the emergence of induced activity in the gamma-band range. Analysis of single-unit activity revealed that phase locking to vibratory stimuli was more precise than that observed for the multiunit responses. The results show that whisker vibrations at frequencies above 100 Hz are faithfully encoded by sustained phase-locked responses of cortical neurons under isoflurane anesthesia. It is conceivable that the awake animal makes use of the tactile signals at even much higher frequencies, which can be provided by the peripheral mechanoreceptors during texture discrimination.
- somatosensory system
- barrel cortex
- gamma oscillations
- tactile discrimination
- isoflurane anesthesia
Rats use the whiskers arranged around their snout to probe objects and to discriminate among very fine textures (Guic-Robles et al., 1989, 1992; Carvell and Simons, 1990, 1995; Brecht et al., 1997; Krupa et al., 2001; Stüttgen et al., 2006). Whiskers touching rough surfaces are estimated to vibrate at frequencies in the order of 1000 Hz (Mehta and Kleinfeld, 2004) and were recently demonstrated to engage in high-velocity micromotions (Ritt et al., 2008). They originate in sinus hair follicles that house several types of mechanoreceptors (Andres, 1966; Rice et al., 1997; Ebara et al., 2002) exquisitely sensitive to high-frequency, low-amplitude vibrissal movements. The primary afferent fibers discharge with high precision to stimulus transients (Gibson and Welker, 1983; Shoykhet et al., 2000; Jones et al., 2004a,b) and phase locked to vibrations up to 1 kHz (Deschênes et al., 2003). Mechanoreceptors from vibrissa follicles of cats have been shown to respond precisely phase locked even to much higher frequencies (Gottschaldt and Vahle-Hinz, 1981). Thus, mechanoreceptors in vibrissa follicles from different species seem to have similar response properties, although some animals whisk and others do not.
The ascending transmission of signals appears to preserve precision of timing and high-frequency components of vibrissa movements in the principal trigeminal nucleus and the ventral posteromedial nucleus (VPM) of the thalamus (Vahle-Hinz and Gottschaldt, 1983; Deschênes et al., 2003; Vahle-Hinz et al., 2007). The picture seems completely different at the level of the barrel cortex [primary somatosensory cortex (S1)]. Here, studies in anesthetized rats have reported phase locking only up to ∼40 Hz whisker vibration (Garabedian et al., 2003; Khatri et al., 2004; Kleinfeld et al., 2006), and coding strategies other than temporal patterning or phase locking have been put forward for high-frequency signals (Arabzadeh et al., 2003, 2004; Hartmann et al., 2003; Neimark et al., 2003; Andermann et al., 2004; Moore, 2004; Kleinfeld et al., 2006). Other studies have demonstrated oscillations at hundreds of hertz evoked by short pulse vibrissal stimuli in intracellular recordings and local field potentials (Jones and Barth, 1999; Jones et al., 2000; Barth, 2003), which indicate that the cortical network does possess the prerequisites for generating high-frequency activity. In awake rats, only relatively coarse stimuli, air puffs, or variable-width apertures have been used, which preclude an analysis of timing precision in the submillisecond range (Krupa et al., 2001; Melzer et al., 2006b).
Thus, the available data suggest an intriguing puzzle. Although high-frequency components of vibrissa movements, reflecting subtle surface irregularities, are precisely encoded by the peripheral mechanoreceptors and transmitted by the ascending pathway, there is no convincing evidence that these temporal patterns are used by the cortical circuitry. The data of the present study may contribute to resolving this issue. Using light carefully controlled isoflurane anesthesia, we show that S1 neurons of rats are readily able to phase lock to continuous high-frequency whisker vibrations above 100 Hz. In previous studies, such high-frequency phase locking may have been masked by suppressive effects of the anesthetics used (Detsch et al., 1999; Friedberg et al., 1999; Sceniak and MacIver, 2006; Vahle-Hinz et al., 2007).
Materials and Methods
This study was performed after approval of the Hamburg Administration of Health and Consumer Protection on 18 adult Wistar rats (400 g body weight) using methods reported previously in detail (Detsch et al., 1999; Vahle-Hinz et al., 2007). Anesthesia was induced by inhalation of isoflurane via a mask, a cannula was inserted into the trachea, and the lungs were mechanically ventilated with 1.5–2.0 vol% end-tidal isoflurane in 100% oxygen during the surgical preparation. Inspired and end-tidal isoflurane concentrations as well as end-tidal CO2 concentration were monitored continuously (Capnomac; Datex, Helsinki, Finland). The femoral vein was cannulated for administration of vecuronium bromide (4 mg · kg−1 · h−1) for muscle relaxation. The heart rate was monitored continuously via an electrocardiogram, and the body temperature was maintained at 37.0 ± 0.5°C by means of a feedback-controlled heating pad coupled to a rectal probe. A parieto-occipital electrocorticogram (ECoG) was displayed throughout the experiment to aid assessment of anesthetic depth in addition to tests of absence of cardiovascular responses when noxious pinching of a paw was applied. The external auditory meati were treated with lidocaine, and the animal's head was mounted in a stereotaxic holder with blunt ear bars in the flat skull position. A unilateral 3 × 4 mm craniotomy was performed above the barrel cortex, the dura was removed, and the exposed cortical surface was covered with agar. At the end of the experiments, the animals were killed under 4.0 vol% isoflurane with potassium chloride intravenously, and the brain was removed and fixed by immersion in 4% glutaraldehyde in phosphate buffer. After dehydration, 50 μm frozen sections were made and stained with cresyl violet for histological verification of the recording sites.
Neuronal recording and stimulation.
Concentrations of isoflurane (0.9–1.5 vol%) were adjusted during the recording session to the individual rat's susceptibility to the anesthetic as determined by the absence of pinch-induced movement, ECoG, and cardiovascular responses. Four individually movable glass-covered tungsten microelectrodes (1 MΩ impedance at 1 kHz) were used for simultaneous recordings of extracellular action potential (spike) discharges of multiunit and single-unit activities at four sites within the barrel cortex. The electrodes were inserted perpendicularly at 1–2 mm distances, and the principal whisker of the receptive fields was determined by manual movement of the whiskers while slowly lowering the electrodes. Most recordings were made at depths of 50–550 μm, corresponding to cortical layers 2/3 and 4, as measured from the cortical surface and verified histologically by small electrolytic lesions (−4 μA for 30 s) applied through the tip of the recording electrode in some cases. The neuronal activity was amplified, filtered, displayed on a personal computer, and stored with 25–35 kHz sampling rate on a hard disk using the Alpha-Map Data Acquisition System (Alpha Omega Engineering, Nazareth, Israel). Cortical neurons responding with sustained spike discharges to vibration of whiskers were selected. As sustained responses, we considered those in which spike discharges occur continuously during the entire duration of the stimulus and with a rate >15 spikes/s above that outside periods of stimulation (ongoing activity). A feedback-controlled electromechanical stimulator (Somedic, Stockholm, Sweden) was used to produce whisker movement with constant displacement amplitudes. Single whiskers were attached to the stimulator probe ∼5 mm from the skin and were moved by controlled sinusoidal-shaped stimuli (10–700 Hz, 20–600 μm amplitude, 0.5 or 1 s duration delivered at 5 s intervals) in rostrocaudal direction. For each unit, the principal whisker was chosen and the stimulus parameters (amplitude and frequency of vibration) were adjusted, allowing for optimal and robust responses. To allow pooling of the data, three ranges of vibratory frequencies were used, i.e., high (>100 Hz), medium (60–90 Hz), and low (15–30 Hz) frequencies.
Off-line analyses were performed with several programs for single-unit sorting (Offline-Sorter; Plexon, Dallas, TX) to generate time stamps, peristimulus time histograms (PSTHs), and phase histograms (Spike2 software; Cambridge Electronic Design, Cambridge, UK), as well as time–frequency analyses and statistics (custom-written Matlab routines). Fifty consecutive responses to each kind of stimulus were used in all analyses. PSTHs (bin width 1 ms) were generated from the neuronal responses, and the response activity per stimulus was determined as mean discharge frequency (spikes per second) above ongoing activity, calculated from the discharge rates measured during stimulus application and the 200 ms period preceding stimulus onset, respectively. The units were assigned to a group (high-, medium-, and low-frequency responders) according to their best responses. Data from units responding equally well to stimuli in two or three frequency ranges were assigned to these different response groups.
Phase histograms were generated by plotting the spikes against the phase of the vibratory stimulus cycle. For this analysis, the ON responses were excluded, and only parts of the responses beyond 200 ms after stimulus onset were used. Thereby, 300–28,000 cycles were analyzed for each set of 50 responses, depending on the stimulus frequency and duration. The temporal consistency of spike timing across stimulus cycles was analyzed by using vector strength (VS) as a measure. VS is calculated by fast Fourier transformation (FFT) and quantifies the degree of coupling of the spikes to the same phase of a vibratory cycle during each repetitive cycle of stimulation (Goldberg and Brown, 1969; Khatri et al., 2004). A VS value of 1 indicates perfect phase locking to the stimulus, whereas a VS of 0 indicates absence of a consistent phase relationship of the spikes to the stimulus. Statistical significance was tested according to the method used by Durand and Greenwood (1958), and a p < 0.01 was considered significant to classify phase-locked (PLR) and non-phase-locked (NLR) responses. The phase in itself at which the spikes occurred was not analyzed, because elicitation of spikes in the mechanoreceptor axon depends on the position of the receptor alongside the hair shaft in the follicle with respect to the propagation of the wave along the whisker, which in turn depends on the position of the stimulator probe and the frequency and amplitude of the vibration.
A frequency–domain analysis of the spike responses was obtained by applying FFT (160 ms window and 157 ms overlap) to the time stamps of the spikes transformed into continuous data with 3 kHz sampling rate. Time–frequency plots showing absolute power of the frequency components were obtained by applying FFT analysis to the mean of 50 consecutive responses, resulting in a measure of the evoked activity, which is by definition strictly locked to stimulus onset. Oscillatory activity that occurs with a jitter in onset latency or phase from trial to trial is lost during the averaging process. This activity, however, is recovered when the FFT analysis is applied to each single response before averaging, reflecting the total activity that comprises both stimulus-locked (evoked) and non-locked (induced) activities (Tallon-Baudry and Bertrand, 1999). To obtain only the non-locked activity, first the mean evoked response was subtracted from each single response, and then the FFT analysis was applied to each single response, 50 of which were then averaged. Comparisons of different responses and stimulation groups were made with Wilcoxon's tests. A p < 0.01 was considered significant. Data are given as mean ± SEM.
Quantitative data were derived from 57 multiunits of the S1 activated by vibratory stimulation of single whiskers at frequencies of 15–700 Hz. Data from 10 single units sorted from these multiunits or recorded individually were analyzed separately and are reported below. Seventeen multiunits were activated by vibration at several different frequencies; thus, a total number of 79 different vibratory responses were used for the analyses. The mean discharge rate of all vibratory responses (n = 79) increased to 55.5 ± 4.2 spikes/s above ongoing activity during the sustained discharges (response activity) and was 14.6 ± 2.0 spikes/s outside periods of stimulation (ongoing activity). Sixty responses were classified as PLR and had a mean rate of 1.1 ± 0.2 spikes per cycle, whereas the 19 NLRs had a significantly lower rate (0.4 ± 0.1 spikes per cycle, p < 0.005; Table 1). The mean discharge rates of PLRs did not differ for low-, medium-, and high-frequency stimulation (Fig. 1), as well as for NLRs during medium-frequency stimulation, but were lower for NLRs during high-frequency stimulation. The associated ongoing activities were significantly higher for NLRs than for PLRs.
Responses to high-frequency vibration
Responses to high vibratory frequencies (>100 Hz) occurred phase locked to the vibratory cycles (VS of 0.25 ± 0.03) in 36 PLRs (72%), with a discharge rate per cycle approximately twice as high (0.45 ± 0.05 spikes per cycle) than in the 14 NLRs (28%; 0.24 ± 0.03 spikes per cycle; VS of 0.04 ± 0.01). For 30 of the PLRs, time–frequency analysis revealed an increase of evoked power at the stimulation frequency. For six PLRs, bands of evoked power were not clearly discernible, presumably because spikes occurred phase locked only intermittently but not continuously to several consecutive cycles of vibratory stimulation.
Examples of PLRs are shown for four multiunits in Figure 2A–C. A typical response (Fig. 2A,B, PSTHs in left column) consisted of a marked ON burst, followed by a discharge pause of 50–100 ms, presumably of inhibitory origin as evidenced by the subsequent rebound burst. Then, a regular discharge followed that was entrained by the vibration frequency as substantiated by the peak in the phase histograms (Fig. 2A,B, middle column) and the band of increased evoked power in the time–frequency plots (right column). In 39% (14 of 36) of the PLRs, this entrainment increased with stimulus duration (see also Figs. 4, 6) and in 36% (13 of 36), it remained constant, whereas in 25% (9 of 36) of PLRs, a decrease of entrainment and spike rate indicated response adaptation. The highest frequency, to which a unit responded in a phase-locked manner, was 692 Hz (Fig. 2C). The mean response rate in this case was 30 spikes/s, and the phase histogram shows that the spikes were elicited significantly more often by one movement direction; thus, precision of spike elicitation amounted to <1 ms. A band representing stimulus frequency is discernible in the middle part of the stimulus epoch in the time–frequency plot (Fig. 2C). In the case of the recording shown in Figure 2D, some cycles of the 130 Hz vibration elicited three spikes (see original record in inset). This three-to-one response component occurred consistently as reflected by the three significant peaks in the phase histogram and three bands of evoked power at 130, 260, and 390 Hz in the time–frequency plot.
Examples of four single units are shown in Figures 3 and 4. The single unit of Figure 3 was activated best by vibration of the whisker at 220 Hz. The PSTH reflects the sustained response, but only the phase histogram reveals the phase locking of the spike discharges to the vibratory cycles, and the time–frequency plot shows the 1:1 discharges over several consecutive cycles by the emergence of a band at 220 Hz (Fig. 3a–c). The original record of this unit shows the sustained responses elicited by four consecutive stimuli and its low ongoing activity (Fig. 3d). The variability of the individual responses is reflected in Figure 3, e, h, and i, where response entrainment increases (e, i) or decreases (h) during the stimulus epoch. At expanded timescale (Fig. 3f), the coupling of four of six spikes to the half-phase of the downward direction is visible. Thus, the precision of spike generation amounts to less than a half-cycle, i.e., <2.3 ms. Figure 4A–C depicts data from three single units sorted from three multiunit recordings. The ON responses are missing from the PSTHs because individual spike shapes cannot be identified within the bursts of action potentials elicited by all units at stimulus onset. In the case of the single unit shown in Figure 4C, the pronounced discharge pause after stimulus onset is missing, and the spikes occur phase locked to each vibratory cycle (see inset). The overall response pattern of the sustained discharges is similar to that of the multiunits showing an increase in vibratory entrainment during later parts of the stimulus epoch in two units (Fig. 4A,B, PSTHs, left column, and time-frequency plots, right column), whereas the response strength is greatest during the early part of the stimulus epoch in one unit (C). Phase locking appears to be more precise than for multiunits, as reflected by the sharp peak in the phase histograms (Fig. 4, middle column) and a VS of >0.8. The mean response rate of all 10 single units increased to 22.6 ± 4.0 spikes/s above ongoing activity (1.7 ± 0.8 spikes/s), and the discharge rate per cycle was 0.16 ± 0.03 with a VS of 0.47 ± 0.10. Comparison with the data from the multiunit recordings suggests that approximately three single units may have contributed to each of the multiunit recordings.
Responses to low- and medium-frequency vibration
Responses to low vibratory frequencies (15–30 Hz) occurred phase locked to the vibratory cycles in all (14 PLRs; VS of 0.42 ± 0.06) but one unit tested (NLR; VS of 0.01). The discharge rate per cycle was approximately twice as high (3.0 ± 0.6 spikes per cycle) in the PLRs compared with the NLR (1.6 spikes per cycle).
Low-frequency stimulation elicited multiple spikes per cycle that occurred locked to a specific phase of the whisker movement, resulting in one peak per cycle in the PSTH and the phase histogram (significant VS, *p < 0.01) in the multiunits shown in Figure 5, A and B, thereby representing directionally sensitive responses. The time–frequency plots show a band of evoked power representing stimulus frequency and also evidence for higher harmonics. Although the responses of the unit shown in Figure 5A occur at constant strength throughout the stimulus epoch, a marked adaptation is present in the responses of the unit shown in Figure 5B. A 1:1 coding of low-frequency vibrations was present in 71% (10 of 14) of the PLRs. In 29% (4 of 14) of the PLRs, both movement directions elicited spikes, again locked to a specific phase of the cycle. This response pattern is evident in the PSTH and the phase histogram of the example in Figure 5C. Both peaks in the phase histogram are significant, thus resulting in a 2:1 vibratory response that is reflected in the prominent band at 37.6 Hz in the time–frequency plot. Bands at the fundamental stimulus frequency or higher harmonics occurred in 93% (13 of 14) of the PLRs. When the units discharged in response to both movement directions, the highest power always was present in the band of the second harmonic.
Responses to medium vibratory frequencies (60–90 Hz) occurred phase locked to the vibratory cycles (VS of 0.33 ± 0.22) in 9 PLRs (69%), with a discharge rate per cycle (0.94 ± 0.2 spikes per cycle) ∼20% higher than in the four NLRs (0.76 ± 0.60 spikes per cycle; VS of 0.04 ± 0.01). An increase of evoked power at the stimulus frequency was evident in the time–frequency plots of all PLRs. An example is shown in Figure 6, with phase-locked responses of a multiunit to 60 and 90 Hz whisker vibration. The entrainment to the vibration increased during the 1 s epoch of stimulation (see PSTHs and time–frequency plots), and the peak in the phase histograms shows the precise coupling to a specific phase of the vibratory cycles.
Wide-band vibratory responses
In the majority of cases in our dataset, only the best frequency was studied to which the units responded. For some recording sites, it was noted, however, that units responded to a broad range of frequencies. These were studied in detail in five cases, an example of which is shown in Figure 6 with responses of one multiunit to whisker vibration at frequencies between 20 and 560 Hz. The response rates were above 35 spikes/s for all frequencies, and significant phase locking was observed over a broad range of frequencies between 20 and 320 Hz and also at 460 Hz (Fig. 6, phase histograms, middle column). In case of the highest frequencies, the precision of spike timing was better than 1 ms. The vector strength, however, was negatively correlated with frequency. Bands of evoked power representing the stimulus frequency were prominent in the time–frequency plots between 20–320 Hz vibration (Fig. 6, right column). In three other units with wide-band vibratory responses, phase locking was present at all frequencies tested (10–200, 20–135, and 20–165 Hz, respectively). In the remaining unit, phase locking occurred only in the low and medium frequency ranges (20 and 65 Hz but not at 158 Hz).
In another eight cases, stimulation was performed at low and high frequencies but not in the medium frequency range. In all of these cases, signals were phase locked at low frequencies and, in six of eight cases, also at high frequencies. The highest vibration frequency that elicited 1:1 responses in the barrel cortex under isoflurane anesthesia was 320 Hz.
Entrainment of the spike discharges with the vibratory stimuli occurring continuously or even increasing during the stimulation epoch was observed for responses to high-frequency vibrations more often (75% of PLRs) than adaptation (25%). The occurrence of oscillatory response components not strictly locked to vibratory whisker movements was studied further by time–frequency analyses of the total and induced activities (Fig. 7). Because the phase and latency of induced oscillatory activity varies across stimulus presentations, this fraction of the response is cancelled out by the averaging process used to determine the evoked activity. Therefore, the time–frequency power was computed first for each single trial, 50 of which were then averaged (total activity) or which were averaged after subtraction of the evoked activity (induced activity). Multiunit responses to low- and medium-frequency vibration showed bands of evoked power at the stimulation frequency and its harmonics but no indication of additional induced activity (Fig. 7A,B). In contrast, prominent induced oscillations in the gamma-band frequency range were visible in the plots of total and induced activities for multiunits responding to high-frequency vibration (Fig. 7C,D). Interestingly, this induced gamma-band response emerges concomitantly with the spike timing entrainment, either during the later part of the stimulus epoch (see band of evoked 190 Hz in Fig. 7C) or shortly after stimulus onset (see band at 130 Hz evoked activity in Fig. 7D).
When the induced activity was analyzed for the population average of 19 PLRs, a clear increase in power in the gamma band (30–70 Hz) was observed during the stimulus epoch (Fig. 8A). To compare the changes in frequency components during stimulation with those during prestimulus ongoing activity, relative power was computed (Fig. 8B). A 1/f frequency distribution was present during ongoing activity, but a prominent shift to gamma-band frequencies occurred especially during the second half of the stimulus epoch. In the population average, this induced gamma-band response reached significance only for the second half of the responses (Fig. 8C). Moreover, a significant decrease of non-phase-locked oscillatory activity was observed at lower frequencies, ∼20 Hz (Fig. 8C).
This is the first report about spike discharges of S1 neurons entrained by high-frequency vibrations of a whisker in a phase-locked manner for long stimulus durations. In our dataset, phase-locked responses occurred up to vibratory frequencies close to 700 Hz, and 1:1 responses were present up to ∼300 Hz. Also, the discharge rates elicited by whisker vibration were higher than reported before for multiunits with a mean of 60.5 ± 5.2 spikes/s, which occurred continuously for several hundred milliseconds in our sample of 60 PLRs. Adaptation of the responses was rarely seen; rather, entrainment to the vibratory frequency remained constant or increased with stimulus duration. Another significant finding of the present study is that phase locking of the spikes can be accompanied by the emergence of induced (non-phase locked) activity in the gamma-band range. Our finding of high-frequency responses hitherto unobserved in the barrel cortex suggests the possibility that awake animals make use of the information about vibratory signals also at frequencies at hundreds of hertz, which are provided by the vibrissal mechanoreceptors.
The main difference between our study and previous measurements in anesthetized rats lies in the type of anesthesia. In most studies (Arabzadeh et al., 2003, 2004; Andermann et al., 2004; Melzer et al., 2006a), urethane anesthesia has been applied intraperitoneally, resulting in a relatively uncontrolled state of anesthesia over the course of the experiment. Urethane is widely used because a single intraperitoneal injection provides immobility for surgical procedures and electrophysiological recordings for several hours. It is generally thought to only minimally affect circulation and respiration as well as neuronal activity at all stages of the CNS. The mechanisms of action of urethane have only begun to be elucidated recently. An in vitro study of recombinant neurotransmitter receptors in frog oocytes showed that urethane effected small but discernible changes in multiple receptor systems, including GABAA, glycine, NMDA, AMPA, and acetylcholine receptors (Hara and Harris, 2002). A study using slices from rat visual cortex (Sceniak and MacIver, 2006) has recently reported a depression of neuronal spike output attributable to augmentation of a potassium leak conductance. In addition, the effects of urethane on subcortical stages may add to the suppressive effects on signal processing (Friedberg et al., 1999). Comparing three stages of urethane anesthesia by incremental intraperitoneal doses, Erchova et al. (2002) reported bursting activity that became more prominent and increasingly synchronized between barrels with increasing depth of anesthesia. Cortical responses to whisker stimulation were found only during bursts. Very similar observations are made during the burst-suppression mode induced by high concentrations of isoflurane (Detsch et al., 2002). In this respect, isoflurane and urethane anesthesia may be comparable. However, it appears that, at a level of anesthesia adequate for electrophysiological studies, isoflurane induces far less suppression of cortical activity than urethane. This is evidenced by (1) the absence of bursting in ongoing activity under isoflurane as is also seen in awake animals regardless of the well known stimulus-related oscillatory activity (Jones and Barth, 1999; Barth, 2003), (2) a generally higher neuronal activity as reflected in the higher rate of ongoing spike activity (Erchova et al., 2002), and (3) absence of adaptation to long-lasting stimulation (Andermann et al., 2004). Low rates of ongoing and evoked activities as well as prominent response adaptation within a few tens of milliseconds were also reported in studies using other anesthetics, such as pentobarbitone, ketamine–xylazine, halothane, or fentanyl (Deschênes et al., 2003; Garabedian et al., 2003; Khatri et al., 2004). Our data clearly suggest that high-frequency components of stimuli can be faithfully encoded by S1 neuronal discharges and that the absence of fast response components reported in the previous studies (for review, see Moore, 2004; Kleinfeld et al. 2006) is primarily attributable to the usage of anesthetics that block the emergence of fast rhythms in the thalamocortical system.
Despite the more faithful transmission of fast signals to the cortical level, the function of the thalamocortical circuits under light isoflurane still differs in important respects from processing in the awake state. A particularly important difference concerns the initial part of the S1 response. Only rarely (Fig. 4C) did we observe a vibratory response present immediately after stimulus onset. The majority of responses began with an ON burst, followed by a discharge pause and a subsequent rebound burst before the discharges became entrained to the stimulus vibration. This initial non-stimulus-entrained part lasted for 50–100 ms, which is longer than an awake behaving rat would wait for the relevant information. When a surface structure is relevant for a decision, rats usually attain this decision very fast after contact with the stimulus (Harvey et al., 2001; Krupa et al., 2001). The ON burst transmitted by the somatosensory system triggers intracortical feedback inhibition, which under most anesthetics is potentiated and thus more powerful than during wakefulness (Detsch et al., 1999, 2002). The same response pattern is present in VPM neurons under isoflurane anesthesia and can be suspended by blockade of GABAA receptors, the sites of action of isoflurane in the thalamus, thereby reestablishing sustained vibratory responses from stimulus onset (Vahle-Hinz et al., 2001, 2007). It is very likely that, in the awake animal, both thalamic input as well as cortical activity faithfully convey the continuous vibratory structure of the input from stimulus onset.
In addition to evoked high-frequency responses, our data provide evidence for intrinsically generated gamma-band oscillations, which correspond to temporal patterns that are not phase locked to stimulus transients. These induced oscillations in the range of 30–70 Hz only occurred with fast vibratory stimuli, accompanying the emergence of phase locking at the higher frequencies. With low- or medium-frequency vibration, in contrast, phase-locked components seemed to dominate this frequency range, possibly masking intrinsically generated gamma-band oscillations. Our observations complement a number of previous studies that have reported similar oscillatory activity in the somatosensory system. Work in slices from rodent somatosensory cortex has shown that the local circuitry is capable of generating rhythmic activity at gamma-band frequencies, based on an interplay of intrinsically bursting excitatory neurons, fast-spiking inhibitory interneurons, and modulatory transmitter systems (Chagnac-Amitai and Connors, 1989; Buhl et al., 1998). Studies on awake rats have demonstrated the occurrence of gamma oscillations in barrel cortex in relation to whisker stimulation (Jones and Barth, 1997) and tactile exploratory behavior (Hamada et al., 1999). Related evidence for fast oscillatory activity has also been obtained in monkey and human somatosensory cortex. In an early study, Ahissar and Vaadia (1990) reported oscillatory single-cell activity over multiple frequency ranges, including the gamma band that was most prominent in ongoing activity but became less stable in frequency for most neurons during tactile stimulation. Lebedev and Nelson (1995) observed gamma oscillations in single units in S1 of monkeys trained to execution of vibratory-cued movements, both in ongoing activity preceding stimulation onset as well as phase locked to the vibratory cue stimulus. Additional studies in the somatosensory system of rats (Nicolelis et al., 1995; Miyashita and Hamada, 1996), cats (Johnson and Alloway, 1994, 1996), and monkeys (Steinmetz et al., 2000) have provided clear evidence for correlated firing of neural populations, suggesting the possibility of assembly formation by neural coherence (Engel et al., 1992, 2001; Singer and Gray, 1995; Herrmann et al., 2004). In humans, several EEG and magnetoencephalographic studies have reported task-related augmentation of gamma-band activity in paradigms in which subjects had to direct spatial attention to tactile or painful stimuli applied to a target finger (Desmedt and Tomberg, 1994; Bauer et al., 2007; Hauck et al., 2007).
A key question concerns the possible functional relevance of evoked and induced oscillatory activity in the somatosensory system. Based on the presumed absence of high-frequency phase locking in barrel cortex, models of tactile vibrissal processing have suggested a dual coding strategy (Kleinfeld et al., 2006). Although phase locking was assumed to be relevant for encoding of low-frequency transients occurring during exploratory whisking or of low-frequency vibrations underlying flutter sensation (Salinas et al., 2000), processing of high-frequency input signals during texture discrimination was thought to rely on rate coding rather than on temporal patterning (Kleinfeld et al., 2006). Our data suggest the possibility that evoked response components may also serve for processing of high-frequency temporal stimulation patterns during texture discrimination, as proposed for example by models of closed-loop neural processing (Ahissar and Kleinfeld, 2003).
Currently, the possible functional relevance of induced activity in the somatosensory system is unresolved. In contrast to phase-locked activity, which may be relevant for the encoding of temporal stimulus features, intrinsically generated oscillations might have a function for linking activity distributed across different barrels. This possibility is suggested by studies on the visual system that have demonstrated a role of synchronized induced oscillations for the integration of signals into coherent ensemble activity and for modulating the functional impact of such signals on other neural populations (for review, see Engel et al., 1992, 2001; Singer and Gray, 1995; Herrmann et al., 2004). Testing this possibility for the somatosensory system will require correlation studies that investigate the coherence of induced oscillations as a function of the spatiotemporal configuration of stimulus patterns across the whisker array.
This work was supported by European Community Grant IST-2000-28127 and the Volkswagen Foundation.
- Correspondence should be addressed to Dr. Christiane Vahle-Hinz, Department of Neurophysiology and Pathophysiology, University Medical Center Hamburg-Eppendorf, Martinistrasse 52, 20246 Hamburg, Germany.