the principle of columnar organization (Lorente de No 1938; Mountcastle 1957; Hubel and Wiesel 1962) is fundamental to the effort to understand the functions of the neocortex. The barrel cortex of rats and mice (Woolsey and Van der Loos 1970; Welker 1971; Welker and Woolsey 1974) is an attractive model system for the study of cortical columns, due to its striking anatomical features. A large body of data has been generated characterizing the neurons that make up a barrel column and their synaptic connections (Lubke and Feldmeyer 2007; Schubert et al. 2007; Thomson and Lamy 2007). This ongoing effort informs the development of increasingly detailed circuit models (Helmstaedter et al. 2007). However, to understand the transformations performed by a barrel column, an understanding of the nature of the input to that column is essential. At the cellular level, the receptive fields (RFs) of barrel neurons reflect this input.

The mechanisms underlying RF generation in barrel neurons have been the subject of a long-running controversy. Barrel neurons exhibit suprathreshold RFs dominated by a single, principal whisker (PW), with some surrounding whiskers evoking weaker responses (Simons 1978; Ito 1985; Chapin 1986; Armstrong-James and Fox 1987). Each barrel receives a dense, minimally divergent input from an analogous structure in the ventroposterior medial nucleus of the thalamus (VPM) called a barreloid (Killackey 1973; Van der Loos 1976; White 1978; Land and Simons 1985; Jensen and Killackey 1987; Chmielowska et al. 1989; Land et al. 1995; Varga et al. 2002). Some have argued that PW responses are driven by direct thalamocortical input, while responses to other whiskers are mediated by intracortical mechanisms (Armstrong-James and Callahan 1991; Armstrong-James et al. 1991). Others have argued that multiwhisker RFs are derived from thalamocortical inputs and that the function of the local circuitry within a barrel is to enhance contrast between inputs by damping responses to surround whiskers (Simons and Carvell 1989; Kyriazi and Simons 1993; Brumberg et al. 1996; Pinto et al. 1996; Pinto et al. 2003).

Lesion experiments have produced inconclusive results. Electrolytic lesion or pharmacological inactivation of a single barrel column has been shown to reduce responses to that barrel's PW in adjacent barrels (Armstrong-James et al. 1991; Fox et al. 2003). This suggests some role for corticocortical mechanisms in the generation of barrel RFs, although the use of urethane anesthesia in those studies may have led to an overestimation of the magnitude of this effect (Simons et al. 1992; Goldreich et al. 1999). Lesions of the interpolaris subnucleus of the trigeminal complex reduce responses to surround whiskers in VPM (Timofeeva et al. 2004) and in barrel cortex as well (Kwegyir-Afful et al. 2005). This demonstrates that subcortical mechanisms contribute to the generation of multiwhisker RFs, but because a majority of barrel neurons exhibited multiwhisker RFs even after the lesion, the size of this contribution is unclear.

Here, using current source density (CSD) analysis of cortical field potential recordings combined with single- and multiunit recordings from layer 4 (L4) barrels and from VPM thalamus in isoflurane-anesthetized rats in vivo, we pursued for the first time a comprehensive survey of whisker-evoked population activity in L4 by individually stimulating each of the whiskers on the contralateral mystacial pad. We found that multiple whiskers evoked a detectable response in the average L4 barrel, even after inactivation of the cortex with muscimol. These results provide direct confirmation of the hypothesis that multiwhisker barrel RFs have a significant thalamocortical component and a first quantitative assessment of the spatial characteristics of that component.

METHODS

Surgery and preparation.

Experiments were conducted in accordance with the ethical guidelines of the National Institutes of Health and with the approval of the Institutional Animal Care and Use Committee of the University of Pennsylvania. Adult male Sprague-Dawley rats (300–350 g; n = 35) were anesthetized with isoflurane vaporized in pure oxygen (5% for induction, 2% during surgery) and artificially ventilated (80–100 breaths per minute). Rats used for intracellular recording experiments were paralyzed with gallamine triethiodide. The stroke rate on the ventilator was adjusted to maintain end tidal CO₂ in a range of 3.5–3.7%. EKG was recorded through two subcutaneous thoracic electrodes, allowing heart rate (4–7 Hz) to be continuously monitored. Body temperature was maintained at 37°C via servo-controlled heating blanket and rectal thermometer (Harvard Apparatus, Holliston, MA). The rat was placed in a stereotaxic apparatus and a craniotomy was made directly above both the barrel cortex (1.0–3.0 mm posterior to bregma, 4.0–7.0 mm lateral to midline) and the medial ventroposterior thalamic nucleus (3.0 mm posterior to bregma, 3.0 mm lateral to midline), and the dura was resected. The cisterna magna was drained to improve stability. For intracellular recordings, additional measures were taken to improve stability, including dexamethasone (10 mg/kg ip) to reduce brain swelling, hip suspension, and filling the craniotomy with a solution of 4% agar.

Electrophysiological recordings.

Recordings of local field potentials (LFPs) across the cortical depth were performed with 16-channel silicon probes (Neuronexus, Ann Arbor, MI). Probe recording sites were separated by 100 μm and had impedances of 1.5–2.0 MΩ at 1 kHz. The shank of the probe was 123-μm wide, and each recording site was a circle with area 413 μm² (∼23 μm diameter). The probe was lowered into the brain under visual guidance, oriented normal to the cortical surface, until the most superficial recording site was aligned with the surface (see Fig. 1). LFP signals were amplified and filtered at 0.1 Hz-10 kHz (FHC, Bowdoinham, ME). Extracellular unit recordings from thalamus were obtained using tungsten electrodes with an impedance of 0.5 MΩ at 1 kHz (FHC). Signals were amplified and filtered at 500 Hz-10 kHz. After the extracellular electrodes were positioned, isoflurane was reduced to 1% and the animal was allowed to recover for 30–60 min.

Single-unit recordings were obtained using sharp-tipped multitetrode printed silicon probes (Michigan Probes/NeuroNexus; 2 shanks with 2 tetrodes each, 1–2 MΩ impedance, 150-μm spacing). Tetrodes were lowered normal to the cortical surface. RF mapping was done in layer IV (depths: between 650 and 850 μm). Recordings were filtered (600–6,000 Hz; Neuralynx recording systems), amplified (×5,000), digitized at 33,657 Hz, and stored with stimulus markers. Waveforms crossing set thresholds (100–120 μV) were captured via the A/D card and analyzed off-line. Potential cells were first identified using automated clustering software utilizing peak and trough feature sets (KlustaKwik). These clusters were then examined manually for waveform shape to discard noncell clusters and combine clusters that captured waveforms from the same cell. These clusters were further refined by hand upon examination of the interspike intervals (SpikeSort3D; Neuralynx). Offline reevaluation of clusters was essential to avoid false clusters or extra clusters typically added by the automatic clustering routine. Most tetrodes had one to three separable cells. Clusters that were not easily separable were discarded to avoid contamination of one cell with another.

Intracellular recordings from barrel cortex were performed with glass micropipettes pulled on a P-97 Brown Flaming puller (Sutter Instrument, Novato, CA). Pipettes were filled with 3 M potassium acetate and had DC resistances of 60–80 MΩ. A high-impedance amplifier (low pass filter of 5 kHz) with active bridge circuitry (Neurodata; Cygnus Technology, Delaware Water Gap, PA) was used to record and inject current into cells. Vertical depth was measured by the scale on the micromanipulator. A Power 1401 data acquisition interface and Spike2 software (Cambridge Electronic Design, Cambridge, UK) were used for data acquisition and online averaging. Thalamic recordings were digitized at 50 kHz; all other recordings were digitized at 10 kHz. All recordings were saved to disk for offline analysis.

Whisker stimulation.

Before recording, whiskers were trimmed to ∼10 mm. Individual whiskers were mechanically deflected using a ceramic piezoelectric bimorph stimulator (Piezo Systems, Cambridge, MA) as described previously (Simons 1983; Wilent and Contreras 2004). For each recording, a glass capillary glued to the end of the stimulator was positioned so that a single whisker rested snugly inside it. The tip of the capillary was positioned ∼5 mm from the skin. The whisker was mechanically deflected in the caudal direction (200 μm) by applying a square voltage pulse of a 200-ms duration to the stimulator. Stimuli were delivered at a rate of 0.33 Hz. In extracellular experiments, responses to 30–60 stimuli were recorded for each whisker. Each extracellular mapping consisted of repeating this process for every whisker from Alpha to E6 (29–30 whiskers per animal). Intracellular RF maps were made by first attempting to find the PW [determined by spike output, if any, postsynaptic potential (PSP) amplitude, and onset latency] and then attempting to complete as many concentric circles of whisker averages as possible while the cell remained healthy (stable input resistance and baseline membrane potential, overshooting action potentials). To obtain response measurements for as many whiskers as possible, intracellular averages were based on only 20–30 stimuli.

Cortical inactivation.

A dental acrylic well was built around the craniotomy and filled with buffered normal saline to obtain baseline recordings. To inactivate cortical neurons, control saline was then replaced with a solution of 2.5 mM muscimol (Sigma) in buffered saline, which was allowed to diffuse passively into the cortex. Cortical LFPs and multiunit activity (MUA) were monitored continuously to confirm elimination of spontaneous cortical activity across all cortical depths (see Fig. 8A), usually within 30–40 min.

Data analysis.

Routines for analysis of the data presented in this study were written in Igor Pro (Wavemetrics, Lake Oswego, OR).

The identity of the PW was determined for each animal based on the cortical LFP and thalamic MUA. In both cortex and thalamus, the whisker that evoked the response with the shortest onset latency, largest peak amplitude, and shortest latency to peak was considered to be the PW. For intracellular recordings, the ability of the whisker to evoke spikes when deflected was the primary consideration for determining the PW.

LFPs from the 16-channel probes were used to calculate the CSD of the cortical whisker-evoked responses according to the methods of Swadlow et al. (2002). We chose to carry out detailed analyses on the CSD values due to the inherent lack of spatial resolution across cortical depths provided by monopolar LFP recordings.

Briefly, the one-dimensional CSD was derived from the second spatial derivative of the LFP data as described by Freeman and Nicholson (1975):

where Φ is the LFP, z is the vertical coordinate depth of the probe, and Δz is the inter-recording site distance (100 μm in the present study). Upper and lower boundaries for CSD calculation were obtained by extrapolating recordings from the first and last recording sites. The CSD profiles calculated by this method were inverted, so that current sinks were represented by positive values.

We determined which electrodes were located in L4 based on the onset latency of the averaged PW-evoked CSD responses. The electrode with the shortest PW-evoked sink onset latency (excluding electrodes >1 mm below the pia) was assigned to L4, along with any adjacent electrodes with PW-evoked sink onset latencies within 0.5 ms of the shortest. This ensured that all electrodes assigned to L4 were recording from neuronal populations receiving direct thalamocortical synaptic input. With the use of this method, two to three L4 electrodes were identified for each animal, spanning a region from 680 ± 90 to 810 ± 80 μm below the pia. Total L4 sinks were calculated by half-wave rectifying the averaged CSD signals and taking the integral over the first 20 ms poststimulus for the sum of all CSD signals assigned to L4.

Cortical MUA was obtained by digitally filtering the LFP (bandpass 500 Hz-5 kHz) followed by full-wave rectification of the filtered signal. Total L4 MUA was calculated by taking the integral over the first 20 ms poststimulus for all the averaged MUA signals assigned to L4.

Multiunit recordings in VPM consisted of two to four units of varying amplitude that could not be separated reliably. These recordings were full-wave rectified, averaged and integrated over the first 20 ms to obtain the total whisker-evoked thalamic MUA.

Response values were calculated by the above methods for every whisker deflected. To obtain unbiased counts of the number of whiskers evoking a response at each level of the circuit, an arbitrary significance threshold was applied to the recordings obtained during the stimulation of each whisker, as follows. The 0.5 s of the signal immediately before the stimulus time was divided into twenty-five 20-ms segments. Each of these segments was analyzed in exactly the same manner described above for the first 20 ms poststimulus: rectified, averaged, integrated, and summed across channels where applicable. From this sample of 25 segments, a mean and SD were calculated. Normal probability plots indicated that these values had an approximately normal distribution. On this basis, we chose a detection threshold of 3 SD above the mean of this distribution. If the poststimulus value (which was calculated in exactly the same way as the 25 prestimulus values) were drawn from the same distribution (i.e., if the whisker deflection evoked no response), the response value would be expected to exceed this threshold in ∼1% of cases. Response values exceeding this detection threshold were considered significant and included in the population averages (described below). Values below this threshold were not counted as whisker-evoked responses and were considered to be responses of zero magnitude in the population averages.

For single units, protocols consisted of 60 trials for each whisker that was tested. Peristimulus time histograms were computed for each whisker. The second preceding the stimulus was divided into 25-ms bins, the number of spikes per bin was summed across trials, and the mean number of spikes per prestimulus bin was calculated. The number of spikes in the 25 ms following the stimulus was summed across trials, and this total was compared with a Poisson distribution with λ equal to the mean number of spikes per prestimulus bin. If the probability of observing an equal or greater number of spikes was <1%, the response was considered significant.

Population RF maps.

Population RF (PRF) maps were made for each animal at all levels of the circuit (thalamic MUA, L4 sinks, and L4 MUA) from which recordings were obtained in that animal. Each map was a 9 × 13 matrix, in which each cell contained the value of a single averaged whisker-evoked response, normalized to the value of the PW-evoked averaged response for that animal at that level of the circuit. Each whisker-evoked response was assigned to a cell in the matrix based on the whisker's position on the mystacial pad relative to the PW: the top row of the matrix contained whiskers that were four rows dorsal to the PW; the next row, three rows dorsal, etc.; similarly, the first column contained whiskers that were six arcs caudal to the PW, the next five, etc. The matrix was padded with nonnumeric values at locations where no whisker was stimulated. Mean RF maps were made by summing all numeric values in each cell across all animals and dividing by the number of numeric values (i.e., the number of stimulated whiskers) for that cell.

PRF map vectors.

We calculated a raw vector for each PRF map by representing each whisker-evoked response as a vector with magnitude equal to the response value normalized to the PW-evoked response and direction equal to the direction from the PW to the stimulated whisker and taking the sum of all vectors. These raw vectors were then corrected for the intrinsic anisotropy of the whisker pad (i.e., the tendency of maps with a PW near the edge of the pad to be biased away from that edge) by calculating a correction vector for the map by setting the response to each whisker at a given distance from the PW equal to the mean response for all whiskers at that distance. Corrected vectors were obtained by subtracting the correction vector from the raw vector. Mean vectors were obtained by taking the mean of all the x (rostrocaudal)- and y (dorsoventral)-coordinates. They were considered statistically significant if the 95% confidence interval for the mean coordinate did not include zero.

Histology.

At the end of each experiment, the animal was given a lethal dose of sodium pentobarbital and perfused intracardially with 0.9% saline, followed by cold 4% paraformaldehyde in 0.1 M sodium PBS. The brain was removed and post-fixed overnight in the same paraformaldehyde solution. Brain slices were processed as described previously (Wilent and Contreras 2004) and photographed using an Olympus BX51 microscope (Olympus America, Melville, NY).

RESULTS

Our goals were to 1) generate the first complete maps of whisker-evoked population responses in L4 and 2) to quantify the spatial extent and spatial tuning of the thalamic input to a single barrel. We measured the response evoked by each whisker on the contralateral mystacial pad before and after cortical inactivation with muscimol. Note that a RF is a property of an individual cell. The CSD and MUA data presented here are used to quantify PRFs (PRFs). Throughout this study, we will use RF to refer only to maps obtained from single cells, and PRF otherwise.

PRF mapping.

We recorded the responses to deflections of all contralateral whiskers in the first seven arcs (straddlers to arc 6) across all rows (A–E) (n = 29–30 whiskers per experiment) from the barrel cortex of adult, isoflurane-anesthetized rats (n = 35). The responses to whisker deflections were recorded as LFPs and MUA using a single shank probe with 16 recording sites (Neuronexus) with a 100-μm interelectrode distance (total span = 1,500 μm). The probe was inserted perpendicular to the surface of the cortex under visual guidance so that the most superficial recording site rested just underneath the pia. The placement of the probe with respect to cortical depth was verified histologically by the track left in Nissl stained coronal sections of the barrel cortex (Fig. 1A). For the mapping of thalamic responses, we used a monopolar tungsten electrode to obtain a single site measure of MUA. In a subset of experiments, we obtained intracellular recordings (n = 7) from L4 of barrel cortex using sharp microelectrodes under the same experimental conditions. In another subset of experiments, we recorded single units from L4 (n = 41) using tetrodes.

In all experiments, the recording probe was inserted in a single cortical location. For most recordings (n = 29 of 35), one whisker evoked an LFP response that had both the largest amplitude and the shortest onset latency of all whisker-evoked responses on all channels. This was designated the PW. The PW was usually one of the large whiskers in the middle of the pad (see Table 1). In a few cases the PW was ambiguous (n = 6 of 35). Typically in these cases, one whisker would evoke a larger, shorter latency response on the more superficial channels, and an adjacent whisker would evoke a larger, shorter-latency response on the deeper channels. These recordings were excluded from further analysis. Average LFP responses evoked by 30–60 deflections in the caudal direction showed a smooth change in amplitude and latency as a function of cortical depth (Fig. 1B, the time of whisker deflection indicated by an asterisk). The same electrodes were also used to record MUA (Fig. 1C). The differences in the LFP responses among recording sites reflect differences in the spatiotemporal structure of the synaptic input to the barrel column. These were quantified by CSD analysis (Fig. 1D; see methods). As shown previously (Di et al. 1990; Castro-Alamancos and Oldford 2002; Einevoll et al. 2007; Higley and Contreras 2007; Quairiaux et al. 2007), the response to the PW deflection was characterized by a sink with the shortest latency to onset (5.0 ms, indicated by small square) and to peak (7.4 ms) in the two electrodes corresponding to granular L4 (Fig. 1D, green rectangle). The MUA response also showed the shortest latency to onset in L4 (Fig. 1C, green rectangle). The short latency LFP and MUA responses in L4 were followed by activation of supragranular layers 2–3. In addition, a sink with short latency to onset (8.4 ms) and to peak (9.6 ms) was observed in the infragranular layers (1.1–1.3 mm deep) together with corresponding MUA.

To compare the responses across experiments and simplify the analysis, we summed the sinks from the electrodes in L4 (2 in this case; see Fig. 1, green dotted rectangle). Electrode positions were determined by response onset latency (see methods). This strategy produces a single measure of MUA and CSD for L4 and reduces measurement noise.

An example of PRF measurement in cortical L4 is shown in Fig. 2. The probe was placed in the cortical barrel column associated with the C5 whisker and each major contralateral whisker was stimulated. The position of the probe in the barrel field was verified histologically by the track left in cytochrome oxidase stained tangential sections through L4 in which the barrel field is clearly visible (Fig. 2A, probe track indicated by red arrow). We plotted the response (CSD sinks and MUA) to each whisker deflection in its relative position with respect to the whisker map (Fig. 2B, left). The largest response for both the sinks and the MUA was to the deflection of the PW, C5 (Fig. 2B, red traces). We calculated average responses evoked by 30 whiskers. To obtain the spatial footprint of the PRF in each experiment and to allow the calculation of population averages, we generated PRF maps (Fig. 2B, right). PRF maps represent response magnitude as the integral of the CSD sinks or MUA responses from 0 to 20 ms poststimulus in a color scale normalized to the integral of the PW response. The position of each colored square corresponds to the position of the deflected whisker with respect to the PW, which is placed at the center of the PRF map (see methods). Colored squares represent responses larger than an arbitrary detection threshold defined as 3 SDs of the prestimulus noise, while responses below that threshold were plotted in gray. In Fig. 2, the evoked MUA produced a slightly smaller PRF than the CSD sinks (L4 sinks = 22 whiskers; L4 MUA = 15). Three more examples of whisker-evoked CSD sinks, MUA responses, and the PRF maps obtained from each are shown in Supplemental Fig. S1 (Supplemental Material for this article is available online at the J Neurophysiol website).

PRF size.

To quantify the PRFs, we averaged across all experiments to obtain the mean PRF map for L4 sinks and MUA (Fig. 3A, n indicated below each plot). We considered values below the detection threshold (in gray) as zero. Thus the mean PRF maps show the mean response magnitude normalized to that of the PW in its relative spatial position with respect to the PW (red square at the center of the map). Because the mean PRF maps are centered on the PW for each experiment, their size is not a reflection of any one individual PRF. However, the mean PRF maps do accurately reflect response decay as a function of spatial location. Thus, across all experiments, the mean L4 sink PRF (18 ± 5.7) was significantly larger than the mean L4 MUA PRF (13.4 ± 5.3; P < 0.001; Fig. 3B, left). In no case did we find a whisker that evoked a MUA response exceeding our detection threshold without also evoking a detectable sink. Differences in PRF size between spike output and the underlying synaptic input, represented here by the CSD sinks, are well known to result from the spike threshold nonlinearity (Carandini and Ferster 2000; Cardin et al. 2007) and have been shown in studies of RF size with intracellular recordings (Moore and Nelson 1998; Zhu and Connors 1999; Brecht and Sakmann 2002; Brecht et al. 2003; Higley and Contreras 2003).

Spatial tuning.

The number of whiskers triggering significant responses provides a global measure of PRF size but contains no information about spatial tuning, i.e., how fast the response decays away from the PW. We quantified spatial tuning using two complementary measures as a function of distance from the PW: 1) the response magnitude (Fig. 3B, middle) and 2) the fraction of whiskers generating a detectable response (Fig. 3B, right). 1) We collapsed the 2-dimensional PRF space into one dimension by assuming the distance between any two horizontally or vertically adjacent whiskers was one and calculating the Pythagorean distance from every whisker to the PW. In the plots presented here, distances are rounded to the nearest integer for clarity. The decay of response magnitude as a function of one-dimensional distance was well fit by an exponential in all cases (solid and dashed lines in Fig. 3B, middle; P > 0.99, χ²-test; λ_MUA = 1.13 ± 0.08 and λ_Sinks = 1.62 ± 0.18). 2) The fraction of whiskers generating a detectable response was calculated by dividing the number of whiskers generating a detectable response at a given distance from the PW by the number of whiskers stimulated at that distance. At a distance of one-away from the PW 96% of whiskers evoked a sink in L4, and 86% evoked detectable MUA. That fraction fell to 65 and 45% at a distance of two-away, to 37 and 20% at three-away, and to 22 and 5% four-away from the PW. Thus, despite the large PRFs measured from the CSD sinks and MUAs, the amplitude of the responses as well as the number of whiskers evoking a response decays rapidly with distance. This shows that a L4 barrel integrates inputs over a fairly large portion of the mystacial pad while maintaining sharp spatial tuning.

To demonstrate the rostrocaudal bias in the shape of the PRFs, we plotted the same response parameters as a function of distance separately for whiskers in the same row or arc as the PW (Fig. 3C). Both the response magnitude as a function of distance (Fig. 3C, left) and the fraction of whiskers evoking a significant response (Fig. 3C, right) were significantly larger for whiskers in the same row as the PW than whiskers in the same arc (P < 0.05, two-way ANOVA), providing the first large scale demonstration of a PRF bias, which has previously being suggested from the responses to a small number of whiskers in both cortex and VPM (Simons 1978; Lee et al. 1994b; Derdikman et al. 2003; Petersen et al. 2003).

To quantify the spatial anisotropy of the PRFs we computed a vector for each (see Fig. 4, black lines). These vectors represent the magnitude and direction of each PRF map's deviation from point symmetry in units of PW-evoked response equivalents (see methods). The mean of all vectors is shown in red on each plot. For both the L4 sink PRFs (Fig. 4A) and the L4 MUA PRFs (Fig. 4B), there was a statistically significant (P < 0.05, Student's t test) bias in the caudal direction (caudal indicated as negative values, rostral as positive). The mean thalamic MUA vector (Fig. 4C) was not significantly different from zero in either the rostrocaudal or the dorsoventral direction. The fact that the cortical, but not the thalamic, PRFs exhibit a directional bias makes it unlikely that this bias is an artifact of our stimulation protocol.

To assess the potential contribution of synaptic activity in adjacent barrels to our measures of subthreshold PRF size, we plotted PRF size vs. PW arc for all recordings. Since the barrels associated with more rostral whiskers are smaller in diameter, we reasoned that if activity in adjacent barrels contributed significantly to our measures of PRF size, we should expect that contribution to be larger in smaller barrels. We found no statistically significant effect of arc on PRF size, using a variety of statistical tests [one-way ANOVA, linear regression, t-test (arcs 1–3 vs. 4–5); P > 0.05 for all]. We conclude that influences from other barrels, which we cannot rule out completely, are unlikely to contribute significantly to measures of PRF size (see discussion).

Because our stimulus was a high velocity deflection applied ∼5 mm from the base of the whisker, we considered the possibility that our measurements of PRF size were to some degree influenced by transmission of mechanical energy through the mystacial pad. We reasoned that the thicker caudal whiskers would more effectively transmit vibrations through the skin than the thinner rostral whiskers, so for all PRF maps with the PW D3 (the most frequent PW; n = 8; see Table 1) we compared the responses evoked by the D1 whisker (2-away caudally) and the D5 whisker (2-away rostrally). The mean D1 response was 75 ± 41% of the PW response, while the mean D5 response was 44 ± 26% of the PW. The difference between the two was not statistically significant (P > 0.05, Student's t test), which indicates that the effect of mechanical transmission is likely to be small.

Intracellular recordings.

The LFP mainly reflects extracellular currents associated with postsynaptic potentials and therefore represents summed subthreshold activity, weighted by distance from the electrode, of all the cells in the local network. It is thus possible that PRFs derived from analysis of LFPs are skewed by the presence of a few cells with large subthreshold RFs. To validate our probe measurements of PRF size and spatial tuning, we obtained intracellular recordings with sharp microelectrodes that lasted long enough to allow the mapping of a large portion of the whisker pad (Fig. 5). Intracellular recordings were obtained from L4 (n = 7) in a subset of experiments and under the same conditions as those used for the probe. The RF size was measured by counting the number of whiskers evoking a PSP ≥3 SD above the baseline noise. Postsynaptic potentials in barrel cortex neurons are a well-balanced mixture of excitation and inhibition (Moore and Nelson 1998; Gabernet et al. 2005; Wilent and Contreras 2005; Higley and Contreras 2006; Cruikshank et al. 2007), but at the resting membrane potential of −75 mV, GABAergic responses are reversed (Wilent and Contreras 2004) and all responses appear depolarizing (Fig. 5A). The amplitude and latency of the postsynaptic potentials varied as a function of distance (Fig. 5A); in the example of Fig. 5A, PSPs ranged from 8.4 mV in amplitude and 5.5 ms onset latency in response to the PW (E2) to 1.5 mV and 12 ms in response to a remote whisker three-away (D5). The population values of PSP amplitude (Fig. 5B, left) and latency (Fig. 5B, right) showed a decrease in amplitude and an increase in latency as a function of one-dimensional distance (Fig. 5B) that very closely matched the exponential decay of L4 CSD sink amplitude (solid lines, taken from Fig. 3C; P > 0.99, χ²-test) and the increase in onset latency of the L4 CSD sinks with distance (lines and squares, taken from Fig. 9C; P > 0.99, χ²-test). These observations provide an important validation of the data obtained with the multisite extracellular probe.

Tetrode recordings.

To further validate the PRF maps obtained with LFP and MUA measures we quantified the RF size of single units (n = 41) in a subset of rats (n = 4),with a four tetrode array in L4 (650–850 μm). The proper placement of the probe was confirmed by the short latency (5–6 ms) of the evoked potential recorded on each tetrode in response to the deflection of the PW. Based on the spike waveforms (Fig. 6A), the distribution of the interspike intervals (not shown), and the firing rate (Fig. 6A), all units were classified as regular spiking (RS). One cell with a RF size of 16 whiskers had an action potential waveform similar to those of fast spiking (FS) cells reported in the literature (Bruno and Simons 2002) and was not included in the database.

Neurons recorded simultaneously showed robust responses to the PW (Fig. 6, B and C) but showed large differences in RF size (Fig. 6D). RF sizes of individual neurons ranged from 1 to 12 whiskers in the same recording, 50% of neurons had RFs of 5 whiskers or fewer, but 25% of neurons had RFs of >8 whiskers (Fig. 7A). The largest RF size for a single unit was slightly smaller than the average RF size obtained with MUA (12 vs. 13.4). However, simultaneously recorded neurons with the same PW frequently had RFs that partially overlapped (compare left and middle in Fig. 6B). We observed responses to whisker deflections at distances up to two-away from the PW for 26 units out of 41 and responses up to three-away from the PW for 10 of 41 units (Fig. 7B).

We were concerned that, given the low baseline firing rates we observed (Fig. 7E), some very weak responses achieved statistical significance (e.g., the response to the stimulation of the A1 whisker in the middle of Fig. 6B). To quantify the contribution of these responses to our population metrics we reanalyzed the single-unit RFs, excluding whiskers that evoked responses with a probability <0.1 spikes/stimulus. This approach led to a slight reduction of RF sizes, but they still remained large. RFs ranged from 1 to 10 whiskers in the same recording, 50% of neurons had RFs of 4 whiskers or fewer, and 25% of neurons had RFs of >7 whiskers (Fig. 7C). Responses to whisker deflections at distances up to two-away from the PW were still observed for 17 of 41 units and responses up to three-away from the PW for 6 of 41 units (Fig. 7D).These results are consistent with a substantial contribution of cortical excitatory cell spike activity to the multiwhisker PRFs found in our analysis of L4 MUA.

We found no correlation between a unit's baseline firing rate and its RF size (r = −0.07; P > 0.6; Fig. 7E). However, the correlation between RF size and the response magnitude (in spikes/stimulus) to the PW was significant (r = 0.46; P < 0.003; Fig. 7F). Taken together the findings in Fig. 7, E and F, suggest that the variability in RF size for individual units is due to differences in either the input resistance of the cells or differences in the strength of their synaptic inputs, rather than differences in resting membrane potential.

Cortical inactivation.

To isolate the contribution of subcortical inputs to the synthesis of PRFs in cortical L4, we repeated the PRF mapping after inactivating the cortex with muscimol (2.5 mM) as shown in the example experiment in Fig. 8. Muscimol is a GABA_A agonist that causes cellular hyperpolarization and a large increase in membrane conductance, effectively clamping the membrane potential close to the reversal potential for chloride and preventing neurons from firing at the concentration we used. Thus muscimol application precludes the participation of cortical circuits in sensory responses without eliminating synaptic inputs from the thalamus. However, postsynaptic responses are reduced in amplitude due to the shunting effect of GABA_A conductances. In addition, muscimol has been shown to activate presynaptic GABA_B receptors, which are found on thalamocortical synaptic terminals (Yamauchi et al. 2000; Porter and Nieves 2004). As a result, PRF maps obtained with this approach are likely to underestimate the spatial extent of the thalamocortical contribution. Muscimol was applied to the surface of the cortex and allowed to diffuse through all cortical layers, causing complete suppression of spontaneous activity, typically after 30 min (Fig. 8A). Cortical inactivation remained stable for the duration of the experiment (see Fig. 8A; t = 240 min). As expected, the L4 sinks were reduced in amplitude but not in onset latency, while the sinks in L2–3, which are entirely due to cortico-cortical activation, were completely suppressed (Fig. 8A). In addition, all MUA in cortex was eliminated (Fig. 8B). Strikingly, the PRF map for the example experiment in Fig. 8 was only slightly reduced in size (compare Fig. 8, C vs. D). Only 13 whisker-evoked responses out of 30, located in the periphery of the map, failed to cross the detection threshold (represented in gray).

After blocking cortical activity with muscimol, a mean PRF map calculated from the L4 sinks across all animals (n = 7) showed a reduction in size while maintaining an elongated aspect (Fig. 9A, top row). To control for the possibility that cortical inactivation was altering whisker-evoked spiking of VPM relay cells, we recorded MUA from VPM thalamus before and after muscimol application and constructed PRF maps in a manner similar to that described above (details in methods). Due to the anatomical complexity of the VPM (see discussion), these recordings are not a reliable measure of the spatial extent of thalamic input to a given barrel. Nevertheless, any substantial changes in the relay properties of VPM neurons should be reflected in this measure. In contrast to the effect observed on the L4 sinks, the mean PRF of thalamic MUA showed no change in size with muscimol application in cortex (Fig. 9A, bottom row), indicating that inactivation of the cortex with muscimol did not change the number of whiskers that evoked thalamic spikes. The average number of whiskers eliciting significant L4 sink responses changed from 18 ± 5.7 in control conditions to 9.9 ± 4.5 after muscimol (Fig. 9B, top left; P < 0.001), while the response in the thalamic MUA remained the same (13.1 ± 5.2 in control vs. 14.8 ± 4.3 after muscimol; Fig. 9B, bottom left). There was also a significant reduction in response magnitude as a function of distance as measured by the decay constant of the exponential fits to the L4 sinks (Fig. 9B, top middle; λ_control = 1.86 ± 0.74 and λ_muscimol = 0.90 ± 0.14, control values from paired data). After cortical inactivation with muscimol, significantly fewer whiskers evoked detectable responses (Fig. 9B, top right; 1-away 96% control vs. 74% muscimol, two-away 58 vs. 26%, three-away 32 vs. 9%, control values from paired data). In contrast, there was no significant difference in the decay of the thalamic MUA response (λ_control = 1.28 ± 0.21, λ_muscimol = 1.47 ± 0.33; Fig. 9B, middle bottom) or in the percentage of whisker deflections that evoked a significant response. The absence of any effect on PRF size in thalamus after muscimol application represents an important control to rule out generalized depression triggered by the muscimol. It also indicates, as has been shown before, that suppression of cortical activity does not change RFs in VPM (Diamond et al. 1992b).

Thalamic relay cells are known to exhibit dual firing modes: at relatively depolarized membrane potentials, they exhibit tonic firing, while at more hyperpolarized potentials they exhibit more infrequent burst firing. Because our thalamic data consist of multiunit activity, not isolated single units, it is impossible to unambiguously identify bursts in the record. However, it is unlikely that whisker deflections evoke significantly more bursts because the change in both the amplitude and the spatial distribution of whisker-evoked responses in our thalamic recordings is very small. Note that in Fig. 9B, bottom middle, the grey symbols represent the population mean of the thalamic responses after muscimol normalized to the principal whisker response before muscimol. This illustrates that overall the change in PW-evoked response and in the relative magnitude of other whisker-evoked responses to the PW-evoked response is quite small. Furthermore, Supplemental Fig. S2 shows a scatter plot of the thalamic MUA response evoked by each whisker before vs. after muscimol, with all values normalized to the PW response before muscimol. The fact that these values cluster close to the unity line suggests that it is unlikely that inactivation of the cortex with muscimol significantly alters the relay of signals through VPM under our recording conditions.

We measured the latency to response onset as the time between the whisker deflection and the moment at which the whisker-evoked sink in L4 crossed the significance threshold. To depict the mean population response latencies, we generated PRF maps showing latency instead of amplitude (Fig. 10A). In the latency PRF maps, the values depicted by the color scale are the actual measurements not normalized to the PW. The average onset latency of the PW response in L4 was 5.1 ms and reached values as long as 10–15 ms for some remote whiskers, similar to the PSPs recorded intracellularly (see Fig. 5B, bottom right).

Figure 10B shows a similar latency PRF map based on data obtained after cortical inactivation with muscimol. Surprisingly, although the mean PRF is smaller, the mean onset latencies of the remaining whisker-evoked responses are shorter. The plot of response latency vs. 1-dimensional distance (Fig. 10C) shows that the difference between control and muscimol was not significant for the PW (control = 5.1 ± 0.7 ms; muscimol = 5.4 ± 0.8 ms) or 1 whisker away from the PW (7.2 ± 2.1; 7.0 ± 1.5) but was significantly different two-away (9.0 ± 2.9; 8.0 ± 2.4) and three-away (9.8 ± 2.9; 7.0 ± 0.5; Student's t-test P < 0.05). There were no responses four-away under muscimol (control = 10.5 ± 2.6 ms). We compared the onset latencies of responses evoked by the same whisker before and after muscimol application and found no statistically significant effect of cortical inactivation on individual response latency (P > 0.05, two-way ANOVA). Rather, we found that responses abolished by cortical inactivation exhibited significantly longer onset latencies in control conditions than responses that were not abolished (Fig. 10C; P < 0.05, two-way ANOVA). The mean difference in latency between abolished and nonabolished responses was ∼2 ms (1-away nonabolished 7.4 ± 2.0 ms vs. abolished 9.1 ± 2.0 ms, two-away 7.4 ± 1.8 ms vs. 10.5 ± 3.8 ms, 3-away 7.8 ± 1.9 ms vs. 9.5 ± 2.4 ms). Finally, the mean magnitude of responses abolished by cortical inactivation was not significantly different from those that were not abolished, when controlled for distance from the PW (P > 0.05, two-way ANOVA). These findings are consistent with a corticocortical contribution to RF formation in L4 barrels.

DISCUSSION

Our goal was to estimate the spatial extent of the thalamocortical input to a single barrel in L4. To isolate thalamocortical from corticocortical inputs, we mapped the PRF of the excitatory inputs to L4 before and after inactivating the cortex with muscimol. We measured the PRF by stimulating each contralateral whisker and counting the number of whiskers evoking a significant response. This measure provides the spatial footprint of the PRF but no information about its shape. Therefore, we also obtained two measures of PRF shape or spatial tuning: 1) the magnitude of the response as a function of distance from the PW and 2) the fraction of whiskers evoking a significant response as a function of distance from the PW. Current sinks in L4 reflect excitatory synaptic input to the barrel, while suprathreshold activity was measured from MUA recorded with the same electrode. Both had large PRFs, averaging 18 and 13.4 whiskers, respectively. These observations were validated both by intracellular recordings from L4, which showed large subthreshold RFs with spatial tuning and onset latencies comparable to values obtained from CSD analysis and by extracellular recordings from single neurons showing that a large proportion of RS cells reliably (≥0.1 spikes per stimulus) exhibit suprathreshold responses to whiskers two-away (17 of 41) and three-away (6 of 41) from the PW. When cortical activity was completely suppressed by topical application of the GABA_A agonist muscimol, the average PRF size calculated from the remaining synaptic potentials in L4 remained large (mean ∼10 whiskers). These results provide the first direct confirmation of the hypothesis that a given cortical barrel receives substantial multiwhisker excitatory synaptic input mediated by subcortical mechanisms, independent of any activity in barrel cortex.

Cortical volume.

To properly interpret our MUA and CSD data, it is essential to have some estimate of the volume from which our electrodes were sampling spikes and LFPs. Using paired intracellular and extracellular recordings in the rat hippocampus in vivo, Henze et al. (2000) showed that spikes evoked in cells >140 μm from the extracellular electrode were undetectable. Given that the barrels associated with the large caudal whiskers in the rat are ∼400 μm in diameter (Welker and Woolsey 1974), it is likely that neurons outside the barrel contribute little to our MUA. Thalamic barreloids are tapering, ellipsoid structures, with diameters of 75–225 μm in adult rats (Haidarliu and Ahissar 2001), so we cannot rule out some contribution of neurons in adjacent barreloids to our thalamic recordings. As a result, our PRF maps of thalamic MUA are a poor reflection of VPM RF size but constitute a valid control for changes in thalamic output after cortical inactivation.

That our LFP recordings are unlikely to be heavily contaminated by activity in neighboring barrel columns is suggested by two recent studies in primary visual cortex of macaque (Xing et al. 2009) and cat (Katzner et al. 2009) showing, with two different methods, that LFPs reflect synaptic events within 250 μm of the recording electrode. Furthermore, extracellular potentials attenuate rapidly with distance according to a 1/r law (Bedard et al. 2006), and, finally, CSDs are produced by signal subtraction that greatly attenuates the effect of volume transmission (Mitzdorf 1987).

Indeed, the asymmetry of the PRFs from L4 sinks in our data is inconsistent with an important contribution of synaptic events in adjacent barrels, since passive spread of current tangential to the pia should be uniform in all directions in cortex. For example, barrels B4 and D6 are equidistant from the recording site in Fig. 2A, but the B4 response is 50% of PW, while that of D6 is undetectable. Note that the mean PRF maps (Fig. 3A) are far more symmetrical. These observations suggest that individual PRF maps tend to be asymmetrical because they are dominated by the random subset of thalamocortical terminals nearest to the electrode, while the averages more accurately reflect the total input to the barrel, which is relatively unbiased (with the exception of the row bias mentioned in results).

Barrels vs. septa.

Neurons in the septa between barrels exhibit large RFs with weak spatial tuning, often lacking a clear PW (Brumberg et al. 1999; Brecht and Sakmann 2002; Furuta et al. 2009). Our PRF maps exhibit strong spatial tuning, with PW responses averaging approximately twice those of adjacent whiskers, indicating that the population of neurons contributing to the signals we recorded was dominated by barrel cells. However, given that we cannot rule out some contribution of synaptic events in adjacent barrels, the same is necessarily true of synaptic events in the septa.

PRF size.

Our MUA PRFs are larger than RFs reported previously in studies of single neurons because they reflect the activity of many neurons within the same barrel, which exhibit incompletely overlapping RFs (see Fig. 6D). Our large CSD PRFs are consistent with the large subthreshold RFs in our intracellular recordings and with published data. The thorough mapping of Zhu and Connors (1999), in which an average of 27 whiskers per cell were tested, showed a mean of 10.4 whiskers evoked a subthreshold response, with a range of 7–16 for RS cells. Furthermore, they found that whiskers one-away from the PW elicited PSPs 55% as large as the PW, two-away 27%, and three-away 12% (compare to Fig. 3C). Moore and Nelson (1998) found cells with subthreshold RFs as large as 16 whiskers and showed an example of a L4 cell that exhibited an excitatory PSP in response to 10 of 10 whiskers stimulated. Other studies (Carvell and Simons 1988; Brecht and Sakmann 2002) did not systematically quantify subthreshold RF sizes but showed examples of remote whiskers evoking excitatory PSPs and spikes.

Interpretation of our results in light of previous studies of suprathreshold RFs is complicated by the variety of anesthetic regimes used in those studies and by the extreme sensitivity of suprathreshold responses to these factors. For example, one study (Kwegyir-Afful et al. 2005) showed that a majority of L4 barrel RS units have multiwhisker suprathreshold RFs in rats under light fentanyl sedation, while another (Furuta et al. 2009), doing a very similar experiment while the animals were under deep ketamine/xylazine anesthesia, found that all L4 barrel RS units have monowhisker RFs. Still, there is extensive evidence of multiwhisker suprathreshold RFs in some barrel neurons in various experimental preparations (Simons 1978; Ito 1985; Chapin 1986; Armstrong-James and Fox 1987).

The effects of cortical slow oscillations induced by anesthesia (sometimes called up- and down-states) on whisker-evoked responses have been well documented (Simons et al. 1992; Hasenstaub et al. 2007). In the experiments described here, the LFP exhibited neither spontaneous slow oscillations nor stimulus-evoked “up-states,” so our results are not influenced by these processes.

Contribution of FS cells.

One limitation of this study is that we cannot determine the contribution of FS cells to the PRF maps. FS cells are likely to contribute more to the MUA than to the sinks given their small numbers and high firing rates (∼10% of L4 cells). FS cells have large RFs (Simons 1978; Simons and Carvell 1989; Bruno and Simons 2002; Kwegyir-Afful et al. 2005), and their contribution may lead to overestimation of the MUA PRFs in our sample; however, our own single unit recordings in L4 suggest that the combined RFs of RS cells are sufficient to explain these findings.

Mechanisms of generation of multiwhisker RFs.

The multiwhisker excitatory input to L4 that we observed after cortical inactivation is clearly thalamic in origin. Two thalamic nuclei send projections to barrel cortex L4: VPM and the posterior nucleus. Posterior nucleus neurons respond weakly and with long (>20 ms) latencies to whisker deflections under anesthesia (Diamond et al. 1992a; Lavallee et al. 2005), so their activity is unlikely to contribute to the responses we observed. Early studies emphasized that RFs in VPM were dominated by the PW (Waite 1973b; Shosaku 1985; Rhoades et al. 1987; Sumitomo and Iwama 1987; Chiaia et al. 1991), but there is extensive evidence that many VPM neurons have multiwhisker RFs, particularly under light anesthesia (Waite 1973a; Simons and Carvell 1989; Armstrong-James and Callahan 1991; Lee et al. 1994a,b; Bruno and Simons 2002; Minnery et al. 2003; Timofeeva et al. 2004; Aguilar and Castro-Alamancos 2005). Thorough mapping of VPM RFs in awake animals or those under light anesthesia yields large mean RF sizes and maximal RFs of 10 or more whiskers, consistent with our findings (Ito 1988; Diamond et al. 1992a; Nicolelis et al. 1993; Nicolelis and Chapin 1994; Friedberg et al. 1999, 2004). Recent work by Deschenes and colleagues has identified three functional subdivisions within VPM: the core, where RFs are single-whisker under deep anesthesia, projects to L4 barrels, while the head and the tail maintain large RFs under deep anesthesia, but mostly project to the septa and other cortical areas (Veinante and Deschenes 1999; Pierret et al. 2000; Veinante et al. 2000; Urbain and Deschenes 2007b; Bokor et al. 2008). The responses we observed are likely dominated by thalamic inputs from the core of VPM, because they exhibit strong spatial tuning and that tuning is sensitive to depth of anesthesia (our unpublished data).

Other potential sources of multiwhisker excitatory input to L4 include: divergent thalamocortical projections from nonaligned barreloids, horizontal axonal projections from neurons in other L4 barrels, horizontal axonal projections within L2/3 that contact the apical dendrites of L4 pyramidal cells, and ascending projections from L6 cells with multiwhisker RFs (Zhang and Deschenes 1997; Arnold et al. 2001; Schubert et al. 2003; Staiger et al. 2004; Egger et al. 2008). All cortically mediated inputs are likely abolished by inactivation of the cortex with muscimol, but because of the limitations inherent in that approach, subtraction of the muscimol map from the control does not yield a reliable estimate of their contribution (see Cortical inactivation). Based on the available evidence, it seems reasonable to conclude that these inputs are weak relative to the input from the corresponding VPM barreloid but that they exert some influence on population activity in the barrel.

Functional significance.

Given the one to one mapping of whiskers to barrels it seems more efficient to represent the position of the stimulated whisker using a narrowly delimited input to each barrel, which would yield a very precise cortical map. However, a broad, but sharply tuned input, allows for complex combinatorial activation of many neurons spread over large territories, which may facilitate the coding of other sensory parameters such as the global direction of motion of the whisker pad (Jacob et al. 2008). Furthermore, the differences in response latency across the many whiskers represented in a single barrel increase the spatial discrimination power of L4 neurons at the population level (Foffani et al. 2008).

The range of reported of RF sizes in the whisker-to-barrel pathway under different experimental conditions is likely the product of a regime in which many whiskers evoke some excitatory input to some neurons, and the transformation of this input into spike output is exquisitely sensitive to shifts in membrane potential and input resistance. These observations in anesthetized animals may reflect the ability of the whisker-barrel system in behaving animals to dynamically tune neuronal representations of sensory stimuli to reflect either small details of single-whisker movements or stimulus properties that emerge from multiwhisker patterns of motion, depending on behavioral context (cf. Nicolelis and Fanselow 2002). It has been shown that some behavioral tasks require input from multiple whiskers (Krupa et al. 2001), while others can be performed with only one intact whisker (Knutsen et al. 2006). Inhibitory, neuromodulatory, and corticofugal projections have been shown to influence sensory stimulus encoding (Simons and Carvell 1989; Jacquin et al. 1990; Lee et al. 1994a, b; Fanselow and Nicolelis 1999; Castro-Alamancos 2002; Temereanca and Simons 2004; Timofeeva et al. 2005; Higley and Contreras 2007; Li and Ebner 2007; Urbain and Deschenes 2007a,b; Furuta et al. 2008; Lee et al. 2008; Furuta et al. 2010). All of these processes are potential contributors to this systemic flexibility.

GRANTS

This work was supported by the Conte Center National Institute of Mental Health Grant P50-MH-064045 and a Foundation pour la Recherche Medicale Postdoctoral Fellowship (to T. Bessaih).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).