1Department of Psychology, The University of Connecticut, Storrs, Connecticut 06269; and 2Moscow Brain Research Institute, Russian Academy of Medical Sciences, Moscow 103064, Russia
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Swadlow, Harvey A. and
Alexander G. Gusev.
The Influence of Single VB Thalamocortical Impulses on Barrel
Columns of Rabbit Somatosensory Cortex.
J. Neurophysiol. 83: 2802-2813, 2000.
Extracellular recordings were
obtained from single neurons in ventrobasal (VB) thalamus of awake
rabbits while field potentials were recorded at various depths within
topographically aligned and nonaligned barrel columns of somatosensory
cortex (S1). Spike-triggered averages of cortical field potentials were
obtained following action potentials in thalamic neurons. Action
potentials in a VB neuron elicited a cortical response within layer 4 with three distinct components. 1) A biphasic, initially
positive response (latency <1 ms) was interpreted to reflect
activation of the VB axon terminals (the AxTP). This response was not
affected by infusion of an
-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate receptor antagonist within the barrel. In contrast, later components of
the response were completely eliminated and were interpreted to reflect
focal synaptic potentials. 2) A negative potential [focal synaptic negativity (FSN)] occurred at a mean latency of 1.65 ms and lasted ~4 ms. This response had a rapid rise time (~0.7 ms)
and was interpreted to reflect monosynaptic excitation. 3) The third component was a positive potential (the
FSP), with a slow rise time and a half-amplitude duration of ~30 ms.
The FSP showed a weak reversal in superficial cortical layers and was
interpreted to reflect di/polysynaptic inhibition. The amplitudes of
the AxTP, the FSN, and the FSP reached a peak near layer 4 and were
highly attenuated in both superficial and deep cortical layers. All
components were attenuated or absent when the cortical electrode was
missaligned from the thalamic electrode by a single cortical barrel.
Deconvolution procedures revealed that the autocorrelogram of the
presynaptic VB neuron had very little influence on either the amplitude
or duration of the AxTP or the FSN, and only a minor influence (mean,
11%) on the amplitude of the FSP. We conclude that individual VB
thalamic impulses entering a cortical barrel engage both monosynaptic
excitatory and di/polysynaptic inhibitory mechanisms. Putative
inhibitory interneurons of an S1 barrel receive a highly
divergent/convergent monosynaptic input from the topographically aligned VB barreloid, and this results in sharp synchrony among these
interneurons. We suggest that single-fiber access to disynaptic inhibition is facilitated by this sharp synchrony, and that the FSP
reflects a consequent synchronous wave of feed-forward inhibition within the S1 barrel.
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Barrel columns of somatosensory cortex receive a
dense, topographically precise input from a few hundred neurons of the
corresponding ventrobasal thalamic barreloid (Jensen and
Killackey 1987; Land et al. 1995
; Van der
Loos 1976
). This thalamic input exerts a powerful influence on
many cortical neurons, mediating salient receptive field properties and
short-latency responses to peripheral stimulation
(Armstrong-James et al. 1992
; Simons and Carvell
1989
; Swadlow 1995
). Extracellular
cross-correlational methods have been the major tools used to study
unitary thalamocortical interactions, and these methods have allowed
identification of the classes of cortical neurons that respond
monosynaptically to thalamocortical input (Reid and Alonso
1995
; Swadlow 1995
; Tanaka 1983
).
Such methods are well-suited for the detection of monosynaptic,
excitatory, suprathreshold effects on individual postsynaptic neurons.
However, they are less well suited for the detection of inhibitory or
disynaptic events (cf., Aertsen and Gerstein 1985
), and
they do not readily yield information concerning the spatial
distribution of events mediated by single thalamocortical afferents.
Spike-triggered averaging of postsynaptic potentials has been used to
examine the time course and amplitude of unitary postsynaptic potentials elicited by single presynaptic fibers (e.g., Kirkwood and Sears 1982a; Mendell and Henneman 1971
;
Munson and Sypert 1979
). Spike-triggered averaging can
also be used to examine extracellular fields elicited by single
presynaptic elements. This method has been predominantly used with
large-diameter axons entering the spinal cord, to trace the influence
of sensory afferent fibers (Munson and Sypert 1979
),
descending central axons (Kirkwood 1995
; Taylor
et al. 1977
), or spinal interneurons (Kirkwood et al.
1993
; Schmid et al. 1993
). In the present work
we apply this method to awake sensory neocortex, to trace spatial and
temporal synaptic influences of individual ventrobasal thalamocortical
axons on topographically aligned barrel columns. We present evidence
that, in addition to providing a powerful monosynaptic excitatory
input, individual impulses of single thalamic afferents gain access to potent inhibitory mechanisms of sensory neocortex.
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Recordings were obtained from S1 cortical barrel columns and
from ventrobasal (VB) thalamic barreloids of adult, fully awake Dutch-belted rabbits. All recordings were obtained from the
representation of medium-to-large central or posterior vibrissae. Some
of the methods have been recently described (Swadlow
1995; Swadlow et al. 1998
) and are reported
briefly here.
General surgical methods
Initial surgery was performed under pentobarbital sodium anesthesia (initial dose, 25-35 mg/kg) using aseptic procedures. After removal of skin and fascia above the skull, the bones of the dorsal surface of the skull were fused together using stainless steel screws and acrylic cement. A stainless steel rod (6 mm diam, thinned to 2 mm in places to conserve space on the skull) was oriented in a rostrocaudal direction, positioned over the midsagittal suture, and cemented to the acrylic mass. The rabbit was rigidly held by this bar during later surgery and recording sessions. A layer of silicone rubber always covered the exposed skull between recording sessions and was also used to buffer the wound margins from the acrylic cement on the skull. After each recording session the entrance in the skull was filled with gelfoam that was soaked in an antibiotic ointment and capped with acrylic cement. The remaining exposed skull was then treated with antibiotic ointment and covered with a thick layer of the silicone rubber.
Extracellular recording procedures
Methods ensuring the humane treatment of our subjects during
immobilization of the head and single-unit recording have been described (Swadlow 1989, 1995
). After the
above initial surgery, all subsequent recordings were obtained from
fully awake rabbits. Rabbits were held snugly within a body stocking
from which the head protruded and were placed on a foam rubber pad. The
steel bar on the head was then fastened to a restraining devise in a manner that minimized stress on the neck when the head was immobilized. After each recording session the small entrance in the skull (see Achieving topographical alignment between VB barreloids and S1 barrels) was resealed with moist gelfoam that was impregnated with
antibiotic ointment and covered with acrylic cement. It is noteworthy
that, although rabbits may respond to gentle tactile stimulation of the
facial region with movements of the nose or vibrissae, no such
responses are observed during the above procedures. Rabbits did not
appear disturbed by these manipulations.
Recording from VB thalamocortical neurons
Well-isolated extracellular action potentials were recorded from
VB thalamic neurons using quartz-filament microelectrodes of 40 µm OD
(Reitboeck 1983) (Thomas Recording). Filaments had an
inner core (12 µm diam) of Platinum (90%)/Tungsten (10%) and were
pulled to a fine taper in the helium-filled chamber of a vertical
puller. Electrodes were either ground to a sharp tip using a diamond
grinding wheel, or metal tips were exposed by crushing the very tip of
the tapered quartz glass under manual control. Electrode impedance (at
1,000 Hz) was generally 1-2 M
.
Five or seven of these quartz-filament electrodes were positioned in a
concentric array above the vibrissa representation of VB thalamus. Each
electrode was guided within a stainless steel cannula (160-200 µm
OD, 100-125 µm ID). The guide tubes were bound together in a
concentric organization (total overall OD at the tips, 480-600 µm),
and a miniature screw-driven microdrive was attached to each electrode,
allowing full independence of vertical motion. The seven cannulae were
chronically implanted so that the tips were 2-3 mm above the VB
representation of the medium-large central or posterior vibrissae after
initial physiological mapping of this structure (Swadlow
1995). On the day of implantation, each cannula contained a
stylus (a tungsten rod, 80-100 µm diam), and the tips were sealed
with a small amount of melted bone wax. Once the cannula were cemented
into place above VB thalamus, styluses were removed and microelectrodes
were introduced into each cannula after being lightly coated with
antibiotic ointment. Microdrives were then attached to the
microelectrodes, and they were driven into the very superficial portion
of VB thalamus. This nucleus was physiologically identified by
1) the presence of reliable, short-latency (<6 ms)
responses to stimulation of a single, dominant vibrissa (the
"principal vibrissa") (Simons 1978
), and weaker or
no responses to stimulation of surrounding vibrissae, and 2) at least some neurons that showed a strong preference for direction of
whisker displacement. After this identification procedure, microelectrodes were then retracted ~200 µm and remained in this position for several days until cortical mapping was achieved.
After preamplification and filtering (low- and high-frequency 1/2 band-pass, 300 and 10,000 Hz, respectively), VB spike data were led to a digital oscilloscope for display and analysis of synaptic or antidromic latencies. Spike data were also digitized at 25-33 kHz, and action potentials were discriminated on-line, using a commercially available software package (DataWave Technologies). This package allows spikes to be sorted based on a number of parameters derived from voltage amplitude and time measures. Spike waveforms were saved, along with stimulus information, and were subsequently reexamined and discriminated off-line. Only well-discriminated neurons were selected for analysis in this study.
In some cases stimulating electrodes were inserted into VB through two of the cannulae. This allowed antidromic identification of layer-6 corticothalamic neurons recorded via the cortical recording electrode.
Cortical recordings of field potentials
Cortical recordings were obtained using several different procedures.
1) Quartz-Platinum fiber electrodes were similar to those
used in thalamus. However, these electrodes were fabricated from larger
diameter fibers so that the dura could be readily penetrated (80 µm
OD, 25 µm metal core) and were tapered and sharpened so that
impedances were usually 0.6-1 M. In some cases, one to seven of
these electrodes were acutely placed within barrel cortex using a
motor-driven, seven-channel microdrive (tip separations, 200-300 µm)
(Eckhorn and Thomas 1993
). In other cases, they were
chronically implanted within the cortex, being guided by a three- to
seven-barrel cannula system cemented to the skull and positioned just
above the dura. Independent vertical movements within the cortex were achieved using the same screw-driven microdrive system described above
for thalamic recordings.
2) Silicone
probes1 with six
independent recording surfaces (vertically separated by 200 µm) were
placed within the cortex and were lowered until the deepest recording
site reached the approximate depth of layer 6. Stimulation pulses were
then delivered via VB stimulating electrodes (above) until a
long-latency (>10 ms) antidromically activated neuron was detected
that showed a supernormal period of increased conduction velocity
(decreased antidromic latency) at intervals of 8-12 ms after a prior
impulse (Swadlow 1989). Antidromic identification was
based on 1) a refractory period of <2 ms and 2)
a constant latency (<0.2 ms variability) to the second of two pulses
delivered at an interval of 10 ms (Swadlow et al. 1978
).
Previous work has shown that such neurons are found exclusively in
layer 6, and that descending corticofugal neurons of layer 5 have more
rapidly conducting axons and do not show supernormal conduction
(Swadlow 1989
, 1990
).
Field activity was continuously recorded from cortical sites at
sampling frequencies of 7,000-8,000 Hz. After amplification and
filtering (half-amplitude low- and high-frequency responses, 10 and
9,000 Hz, respectively), field data were digitized and led to DataWave
data acquisition system. Spike-triggered averages of field data with VB
triggering spikes were created off-line using our own software. These
were generated by averaging the cortical responses to 3,000 VB
spikes. No stimuli were delivered, and animals sat in a quiescent
state. We rejected field data in which high-frequency, high-voltage
transients occurred.
Achieving topographical alignment between VB barreloids and S1 barrels
Our general experimental strategy was to 1) record from one or more superficial barreloids of VB thalamus and identify the principal vibrissa providing dominant input, 2) identify the topographically aligned cortical barrels, and 3) localize layer 4 using physiological criteria and record spike-triggered averages. 4) Next, the VB electrodes would be lowered further into a new set of more ventral VB barreloids, topographic alignment with the appropriate new set of cortical barrels would be reestablished, and new averages would be generated. Several steps were necessary to achieve these aims.
IDENTIFYING THE PRINCIPAL VIBRISSA PROVIDING INPUT TO VB BARRELOID NEURONS UNDER STUDY. Initial assessment of the principal vibrissa for each VB neuron was done qualitatively, using hand-held stimuli. In most cases, deflection of only a single vibrissa resulted in a strong response, and other whiskers yielded considerably weaker or no responses. For those neurons showing strong responses to deflection of >1 whisker, latency measures were obtained (below), and the whisker yielding the shorter latency was considered the principal vibrissa.
IDENTIFYING THE TOPOGRAPHICALLY ALIGNED S1 BARREL.
To minimize the amount of dura exposed during the search for
topographical alignment (because exposed dura gradually thickens and
microelectrode penetrations become difficult), it was necessary to make
microelectrode penetrations through very small openings in the skull.
This was achieved by using a dental drill to thin a region of skull
(~3 × 3 mm) above the "barrel field" of S1 to a thickness
of ~200 µm (from an initial thickness of 1.5-2 mm). At this
thickness the skull was transparent, and the larger superficial vessels
could be clearly visualized. Thereafter, very small entrances in the
skull could be made using a dental drill with a tip diameter of 0.24 mm. In the initial entrance, a finely tapered, low-impedance (<1 M)
tungsten-in-glass microelectrode was then inserted to the approximate
position of layer 4. This layer was physiologically identified by the
presence of a strong, short-latency (<7.0 ms), multiple-unit response
to very brief (<4 ms) air-puff stimulation of the vibrissae
(Swadlow 1995
; Swadlow et al. 1998
).
Next, the principal vibrissa for this electrode site was ascertained as described above for VB neurons. Because of the orderly representation of the vibrissae on the cortical surface, only a few such penetrations were usually necessary to locate the S1 barrel that was topographically aligned with the VB neuron under study. In some cases, strong, short-latency (<7 ms) responses were observed from each of two vibrissae, probably reflecting a transition region between two barrels.
In these cases, the electrode was moved to a position resulting in a
dominant, short-latency input from single vibrissa (presumably nearer
the center of the barrel).
FIELD RECORDING FROM IN OR NEAR LAYER 4. After topographical alignment, the tungsten microelectrode used for initial mapping was replaced with a quartz-platinum electrode (or silicone probe electrode) used to record spike-triggered averages. The identity of the principal vibrissa was re-ascertained and the depth of this electrode was adjusted to maximize short-latency (<7 ms) multiple-unit responses to deflection of the principal vibrissa (above). In some cases additional recording electrodes were placed in neighboring barrels.
NEW VB NEURONS WERE ISOLATED. During subsequent recording sessions thalamic electrodes were lowered until new VB neurons were isolated. Principal vibrissae were ascertained and topographic alignment was reestablished with cortical recording sites.
Antagonizing -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid
(AMPA)/kainate receptors near the recording site
To block postsynaptic components of the unitary field potential 6,7-dinitroquinoxaline-2,3-dione (DNQX) was pressure injected into the cortex at a distance of 350 µm from cortical recording sites. DNQX (1 µl, 5 mM) was dissolved in 165 mM NaCl and injected within or near layer 4 over a period of 6 min, using a 36-gauge (112 µm OD, 60 µm ID) sharpened cannula.2
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In three rabbits, 25 VB neurons were studied that were in precise topographic alignment with a cortical field electrode. Twenty of these VB neurons elicited an average unitary field potential with a clear short-latency (<1 ms) response within the topographically aligned barrel column. These 20 neurons provide the basis for most of the present analyses. Another data set obtained from two additional rabbits consisted of 13 VB neurons that each elicited a short-latency response within the topographically aligned S1 barrel column. These cells were used to study the cortical depth profiles of the elicited field potentials.
Figure 1 presents the spike-triggered average obtained from an S1 recording site following spontaneous impulse activity in a topographically aligned VB neuron. For this recording site, and for each of the topographically aligned sites studied here, the response profile within and near layer 4 could be readily divided into three distinct components: 1) an initially positive, biphasic or triphasic response thought to represent the invasion of the presynaptic impulse into the axon terminal arborization within layer-4 (the axon terminal potential, AxTP), 2) a longer-latency, negative-going potential believed to reflect a local, focal synaptic depolarization (the focal synaptic negative potential, FSN), and 3) a subsequent long-lasting positive potential, believed to reflect focal synaptic hyperpolarization (the focal synaptic positive potential, FSP). Table 1 presents the latency, amplitude, and temporal characteristics of these responses. Values given in the table (and, with noted exceptions, those presented below) are for only those 17 topographically aligned thalamocortical sites that yielded both AxTPs and FSNs of >1 µV.
|
|
AXTP
The AxTPs had latencies of 0.54-0.92 ms (mean, 0.71 ms).
These latencies were measured at the onset of the initial positive deflection, at a depth producing the maximal amplitude cortical response (left arrow in Fig. 1A). Maximum amplitudes of the
AxTP were always found at cortical depths in which minimal latencies (<7.0 ms) were observed in multiple-unit cortical responses after stimulation of the receptor periphery. These maximal amplitude responses were therefore believed to be in or very near to layer 4 (METHODS) (Swadlow 1995; Swadlow et
al. 1998
). The maximal amplitudes of the AxTP (measured from
the peak of the initial positive response to the peak of the secondary
negative response) had a mean value of 3.25 µV, and the duration of
the initial positive peak (taken at half amplitude) was 0.23 ms.
Eight of the VB neurons generating spike-triggered averages were tested with microstimulation pulses (0.2 ms, <20 µA cathodal) delivered to the cortical recording site. In each case this resulted in an antidromic response. Figure 1C shows that the antidromic latency was correlated with the latency of the AxTP (R2 = 0.82, P = 0.002).
In three cases, 1 µl of DNQX were injected into the cortex at distances of 300-400 µm from the recording site.
Figure 2 shows the results of injecting 1 µl of DNQX within the cortex at a distance of 350 µm from the recording site. AxTPs were little affected by antagonism of AMPA/kainate receptors within the cortical barrel (Fig. 2A). In contrast, the FSN and FSP were nearly abolished by this manipulation (Fig. 2B) and took several hours to reach partial recovery (Fig. 2C). Virtually identical experiments were performed in two additional cases (separation between recording microelectrode and injection site, 300-400 µm), with nearly identical results. In each of these two additional cases, the amplitude of the AxTP following the injection remained very close (85-105%) to the control amplitude However, after the injection of DNQX, the amplitudes of the FSNs and FSPs were reduced to <20% of control values.
|
FSN
The FSN was sharply rising (10-90% rise time, 0.68 ms) and emerged out of the trailing edge of the AxTP (arrow at right in Fig. 1A). The latency of this response was, in some cases, difficult to ascertain precisely because the FSN arose out of the AxTP, and, as noted above, this latter response varied in both amplitude and waveform with cortical depth. Nevertheless, we estimated the onset of the FSN to have a mean latency of 1.67 ms and, for individual pairs, to be roughly correlated with the latency of the AxTP (R2 = 0.24, P = 0.04).3 The maximal amplitude of the FSN (measured from the baseline to the peak of the negative potential) had a mean value of 3.86 µV. Generally, the FSN decreased to the baseline level within several milliseconds. The duration of the half-amplitude response had a mean value of 2.25 ms.
FSP
The FSN was followed by a prolonged positive response, the FSP (Fig. 1B). The onset of this response appeared to emerge from the trailing end of the FSN. Because of this, and because of the relatively long time course of this response, we make no attempt to measure the latency of onset. The maximal amplitude of the FSP (measured from the baseline just before the thalamic spike to the peak of the positive potential) had a mean value of 5.42 µV. The FSP had a total duration of 30-70 ms and a mean half-amplitude duration of 31.8 ms.
Depth distribution of field components
The maximal amplitudes of the AxTP, FSN, and FSP were all found within or very near to layer 4. Figure 3 presents a case in which a single stationary probe with six recording sites was used to generate spike-triggered averages at vertical site separations of 200 µm. The maximal response (Max) for all unitary field components was seen at site 3 (from top), and at sites 3 and 4, multiple-unit responses to peripheral stimulation were seen at latencies of <7 ms, indicating that these sites were within or very near to layer 4. The deepest site (bottom) was located within layer 6. This assertion is based on the presence of a neuron recorded at this site that showed 1) a very long latency (>10 ms) antidromic response to thalamic stimulation and 2) a decrease in antidromic latency (the supernormal period) after a single prior impulse (METHODS). Although the latency of the AxTP seen in Fig. 2 is approximately the same at each of the six sites, the amplitude reaches a clear peak at site 3 and is diminished both above and below this depth. The amplitude of the FSN and FSP also reaches a peak at the same site. Figure 4 presents the relationship between cortical depth and amplitude of the AxTP, the FSN, and the FSP for six VB neurons studied with the same six-channel stationary silicone probe described above. Note that for each of these VB neurons, the amplitude of the AxTP, the FSN, and the FSP reached a peak at or next to site 3 (Max) and was considerably diminished both superficially and deeper within the cortex.
|
|
No signs of reversal of any components of the unitary field
potentials were revealed by the data presented above. However, the
above recordings extended only ~ 400 µm superficial to the site of the maximal response, and this may have been insufficient to
reveal the reversal. Therefore 13 additional VB neurons were studied,
each with a well-defined AxTP, FSN, and FSP within layer 4 of the
topographically aligned barrel column. We used a traditional, finely
tapered microelectrode with a single low-impedance (<1 M) recording
site to sequentially record spike-triggered averages from different
depths (200-300 µm vertical separation of recording sites) within
the barrel column. In no case did the AxTP or the FSN show
any sign of reversal in more superficial cortex (up to 900 µm
superficial to the maximal response). However, in each of the 13 cases
the FSP showed a weak reversal at distances of 400-900 µm
superficial to the maximal response. Figure
5 shows responses obtained from one such
case. Here, the maximal response of all unitary field components
occurred at a depth of 1,115 µm beneath the surface of the dura
(bottom), and a mild reversal of the FSP was seen at a depth
of 415 µm. Note that for this case (and for each of the 13 cases) the
maximum amplitude of the reversed FSP was much less (17 ± 7%,
mean ± SD) than the maximum value of the FSP seen near layer 4.
|
Responses elicited by VB neurons of nonaligned barreloids
Responses from nonaligned barreloid-barrel pairs were generally studied by identifying a cortical site that yielded a clear AxTP and FSN to spikes of a topographically aligned VB barreloid neuron and, in addition, recording fields from this site that were simultaneously generated by neurons in one or more neighboring VB barreloids. In this way we could be confident that a weak unitary field potential generated by the nonaligned VB neuron was not due to a nonoptimal depth placement of the cortical recording site. Figure 3C shows one such case. Here, the same six-channel silicone probe electrode that generated the average unitary waveforms shown in Fig. 3, A and B (from an aligned VB barreloid neuron), also, during the same recording session, generated the average unitary waveforms from a VB barreloid neuron that was missaligned by two barreloids. Note that virtually no AxTP or FSN is seen, and only a small hint of an FSP is detectable. Figure 6 shows a histogram of the amplitude of the AxTP, FSN, and FSP when triggering VB neurons are in topographically aligned barreloids,4 or when they are missaligned by one, or by or two to three barreloids. Note that all components of the spike-triggered average are strongly attenuated when missaligned by a single barreloid.
|
In one fortuitous case, three thalamic neurons (a1, a2, and b1) in two neighboring VB barreloids (a and b) were recorded simultaneously with fields in two neighboring cortical barrels (A and B), each of which was in topographic alignment with one of the VB barreloids under study, and missaligned with the other. Figure 7 (top) presents this case schematically, and spike-triggered averages obtained from sites A and B are shown below. Note that VB neurons a1 and a2 elicited clear AxTPs and FSNs in S1 barrel A (aligned), but not in barrel B (neighboring barrel, nonaligned). Conversely, the VB neuron "b" elicited clear responses in S1 barrel B, but not in barrel A.
|
Contribution of the autocorrelogram to spike-triggered averages
In all methods of spike-triggered averaging, the
autocorrelogram of the triggering spike can influence the average
waveform obtained (Kirkwood 1979; Moore et al.
1970
) (see DISCUSSION). We assessed the
contribution of the autocorrelogram by deconvolution of each of the
spike-triggered averages that contributed to Table 1 with the
associated autocorrelogram of the triggering VB spike (Brillinger et al. 1976
; also see
Eggermont et al.
1993
).5 For each
of the 17 cases, the deconvolved AxTP and FSN were very similar
to the original waveforms (within ±10% in amplitude and duration).
The effect of the deconvolution on the FSP was somewhat greater
and depended on the shape of the autocorrelogram. Whereas sharp side
bands in the autocorrelogram resulted in a decrease in the
amplitude of the deconvolved FSP, a central trough resulted in an
increase in the amplitude of the FSP. For most of the 17 cases, the
change in the amplitude of the FSP was <15% (mean change, 13 ± 7.4%), but in a few cases, greater changes were seen (up to 28%).
Figure 8 shows 3 of the 17 cases. Extreme
cases are shown on the left (very sharp side-bands in the
autocorrelogram) and the right (deep trough in the
autocorrelogram), where the magnitude of the FSP varied by >20%. To
validate the Fourier-based deconvolution procedure, we convolved the
results of the three deconvolutions shown in Fig. 8 with their
associated autocorrelograms. In each case, the original waveform was
reconstituted and was virtually indistinguishable from the original
spike-triggered average waveform.
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The present work has identified three components in the average unitary field potential elicited in an S1 barrel by a single, topographically aligned VB barreloid neuron.
AxTP
Previous studies of extracellular spike-triggered averages in
spinal cord and brain stem have differentiated the presynaptic component of the potential into "axonal" and "terminal"
components (Kirkwood 1995; Munson and Sypert
1979
; Schmid et al. 1993
). Although we observed
a low-amplitude axonal potential throughout the depth of the cortex, we
have focused here on the larger amplitude component found in the
vicinity of layer 4. The maximal amplitude response at this depth, as
well as the restriction of the AxTP to a single barrel column were
expected, based on morphological studies of individual VB axon terminal
arborizations (Jensen and Killackey 1987
).
The initially positive-going response of the AxTP is consistent both
with previous analyses of axon terminal responses and with theoretical
analyses of recordings from near the sealed ends of axons (cf.,
discussion in Munson and Sypert 1979). The total duration of the initial positive component was brief (mean, 0.37 ms).
It should be noted, however, that the timing of the triggering VB spike
was accurate only to the nearest 100 µs (because of software limitations), and this will result in an overestimation of the duration
of any resultant waveforms by a similar value. By testing our system
using calibrated square waves, this was found to be the case. It should
also be noted that the duration of the AxTP was measured at a depth
that yielded a maximal amplitude waveform, and this was thought to
correspond to layer 4. In some cases, the duration of the AxTP at this
depth was somewhat greater than observed in axonal potentials of
reduced amplitude seen 200-500 µm deeper in cortex (e.g., Fig.
3A). The increased amplitude and duration of the AxTP within
layer 4 is consistent with increased membrane area and temporal delays
occurring in a highly branched terminal axon.
The AxTP occurred at a mean onset latency of 0.71 ms after the onset of
the thalamic action potential. This value is in agreement with a mean
antidromic latency 0.91 ms reported for VB neurons that were
antidromically activated via S1 stimulation (Swadlow 1995), indicating a mean axonal conduction time of ~ 0.71 ms [axonal conduction time was estimated by subtracting 0.2 ms
from the antidromic latency (e.g., Fuller and Schlag
1976
)].
Antidromic responses were seen in each of eight tested VB neurons that
were in topographic alignment with a cortical recording site, and
antidromic latency was highly correlated with the latency of the AxTP
(Fig. 1C). For these cells, however, antidromic latencies were a mean of 0.38 ms greater than were AxTP latencies. Approximately 0.2 ms of this discrepancy is expected, due to utilization time after
the antidromic stimulus and delay of invasion near the soma (e.g.,
Fuller and Schlag 1978), and 50 µs can be attributed
to the time stamp resolution of the VB triggering spike (100 µs, above). The reason for the remaining discrepancy (0.13 ms) is unknown.
FSN
Several lines of evidence favor the hypothesis that the FSN is largely due to local active inward currents associated with monosynaptic excitation via the VB thalamocortical neuron under study: 1) the amplitude of the negative potential was sharply delimited in depth; 2) this optimal depth was the same as that which yielded a) maximal amplitude AxTPs, b) very-short latency (<7 ms), multiunit spike activity elicited by peripheral air-puff stimulation, and c) the maximal amplitude short-latency evoked response to peripheral air-puff stimulation; 3) the FSN was sharply limited in the horizontal domain, with little response being seen in the neighboring barrel; 4) the latency of onset of the FSN is consistent with monosynaptic thalamocortical activation and was significantly correlated with the latency of the AxTP; and 5) the FSN was completely blocked by antagonism of AMPA/kainate receptors, which left the AxTP intact. Together, these data strongly suggest that the FSN reflects local inward currents associated with monosynaptic thalamocortical excitation, and that the focus of this activity is within layer 4.
The FSN had a sharp rise time and a total duration of ~4 ms. Although
Kirkwood (1995) reports similarly brief extracellular focal synaptic potentials in bulbospinal neurons of brain stem, Munson and Sypert (1979)
found that extracellular
potentials elicited by 1a afferent fibers mirror the time course of
motoneuron EPSPs. As noted above, it was not possible to differentiate
the end of the FSN from the beginning of the FSP. It is likely that the
falling phase of the FSN (which was often abrupt, e.g., Fig. 1)
reflects the onset of disynaptic inhibition (Agmon and Connors
1992
; Swadlow 1995
), and this may have truncated
the FSN.
Cross-correlation studies of VB thalamocortical neurons and putative
feed-forward inhibitory interneurons of S1 (Swadlow
1995) revealed that a sharp increase in cortical spike
probability occurs from 1.4 to 2.0 ms following VB action potentials
(Swadlow 1995
). This peak in the cross-correlogram is
very brief, having a half-amplitude response of ~1 ms. These values
closely coincide with the rising phase and peak of the FSN reported in
the present work. The temporal correspondence between the rising phase
of the FSN and the time course of spikes elicited by VB thalamocortical
neurons supports our contention that the FSN is due to excitatory
synaptic currents and also supports the hypothesis that spikes
resulting from monosynaptic input are largely generated during the rise
time or derivative of the excitatory postsynaptic potential (EPSP)
(Cope et al. 1987
; Kirkwood and Sears
1982b
). Although synaptic currents are generally thought to be
the main contributors to cortical field potentials (e.g.,
Mitzdorf 1985
), correspondence between the rising phase of the FSN and the sharp increase in cortical spike probability raises
the possibility that local action potentials may make some contribution
to the FSN.
FSP
Our data suggest that the FSP reflects an outward current
generated by local inhibitory postsynaptic potentials (IPSPs). Two lines of evidence support this notion and argue against the possibility that the FSP is a passive reflection of a distant inward current. 1) The cortical depth yielding the maximal-amplitude FSP was
the same as that yielding the maximal amplitude of the AxTP and FSN. As
noted above, the short-latency synaptic responses to air-puff stimulation seen at this depth are indicative of layer 4. This layer is
rich with GABAergic interneurons (Harris and Woolsey 1983; Simons and Woolsey 1984
; White
1978
, 1989
) that have an axonal projection that
is largely restricted to the same layer 4 barrel (Harris and
Woolsey 1983
). One would therefore expect IPSPs generated by
these neurons to have a similar depth distribution to the EPSPs
generated by thalamocortical afferents. 2) The amplitude of
the FSP reached a sharp peak within and near layer 4, and only a weak
reversal of the FSP was seen in superficial cortex. This is a
predictable consequence of an active outward current in layer 4, and a
passive inward current in superficial cortex. This result is expected
because inhibitory interneurons form many of their synapses on
pyramidal neurons (White 1989
), which have apical dendrites that extend into superficial cortex. It is important to note
that a positive-going waveform in layer 4 could result from excitatory
input to superficial cortical layers. However, in such a case a
stronger negativity would be expected to occur in superficial layers,
and a weaker and more spatially dispersed positivity would be expected
to occur in deeper layers. For these reasons, we believe that the FSP
reflects outward currents associated with local IPSPs, and that the
focus of this activity is within layer 4. However, this conclusion must
remain tentative until confirmed or disconfirmed by intracellular analyses.
The FSP was somewhat more broadly distributed both horizontally and in depth than either the AxTP or the FSN. This could result if some of the inhibitory interneurons that receive monosynaptic VB input have an axon terminal distribution that is more widespread than that of the thalamocortical axons. Alternatively, because the FSP can be influenced by broadly synchronous activity of presynaptic elements (below), this could reflect a weak, broad synchrony between VB thalamocortical neurons of the same and neighboring barreloids.
Assessing the influence of the autocorrelogram on the obtained spike-triggered averages
A confounding problem for all methods using spike-triggered
averaging is that the obtained waveform (the intracellular or extracellular average unitary waveform or the cross-correlogram function) may receive contributions in addition to those initiated by
the "triggering" event under direct observation. In this respect, the arguments of Moore et al. (1970) concerning
"primary effects" versus "secondary effects" on
cross-correlogram functions apply equally well to methods of
spike-triggered averages of intracellular and extracellular potentials
(also see, e.g., Kirkwood 1979
). Thus any
cross-correlogram function or spike-triggered average waveform may be
considered to be a compound waveform that results from the convolution
of the idealized, unitary waveform (the "kernel" waveform) (cf.
Kirkwood 1979
; Knox
1974
)6 with
1) the autocorrelogram of the triggering neuron under study, 2) the cross-correlogram of the triggering neuron with other
neurons that project to the neuron or tissue under study (Hamm
et al. 1985
; Moore et al. 1970
), and
3) the poststimulus histogram of any variations in
conduction time that result from axonal conduction or synaptic delays.
We used the Fourier-based method of deconvolution
(Brillinger et al. 1976; Eggermont et al.
1993
) to assess the contribution of the autocorrelogram to the
spike-triggered average waveform. This analysis indicated that the
autocorrelogram contributed little to either the amplitude or time
course of the AxTP or the FSN. In addition, in most of the 17 cases
analyzed, we saw only modest changes in the amplitude and time course
of the FSP. However, when the autocorrelogram showed either strong
central side bands (Fig. 8, left) or a deep central trough
(Fig. 8, right) the FSP was significantly influenced.
We did not analyze the influence of variations in conduction along VB
thalamic axons on spike-triggered averages because of the uniformly
rapid conduction time of this axonal system (Swadlow 1995), and we did not analyze the influence of presynaptic
synchrony (Hamm et al. 1985
; Moore et al.
1970
). The most obvious potential source of such synchrony is
between the presynaptic VB neuron under study and other VB neurons in
the same barreloid. Whereas sharply synchronous activity (cf.,
Alonso et al. 1996
; Swadlow et al. 1998
)
among VB neurons could influence the faster components of the
spike-triggered average (i.e., the AxTP and the FSN), more broadly
synchronous activity could influence the slower components (the FSP).
Preliminary studies of VB neurons within the same barreloid have
revealed little sharply synchronous activity (Swadlow et al.
1998
; present work). However, a definitive analysis of the influence of presynaptic synchrony on thalamocortical unitary field
potentials must await further analyses of the degree of synchrony among
thalamocortical neurons of the same VB barreloid.
Comparisons of unitary thalamocortical field potentials with previous studies of sensory neocortex
Extracellular field potentials elicited by single thalamocortical
afferents have not been previously studied in any region of cortex. In
principle, however, they should bear a resemblance to the extracellular
fields and underlying postsynaptic potentials generated by synchronous
barrages of thalamocortical impulses (i.e., those resulting from
sensory or gross electrical stimulation). The earliest synaptic
responses in primary sensory cortex following electrical stimulation of
thalamic afferents are monosynaptic EPSPs seen predominantly within and
near layer 4 (e.g., Agmon and Connors 1992;
Ferster and Lindstrom 1983
). In the extracellular record, this early response consists of a negative potential that reflects active current sinks (Agmon and Connors 1991
;
Mitzdorf 1985
). This potential usually reverses in very
superficial layers, reflecting the passive return of currents (current
sources) through the apical dendrites of pyramidal neurons. In the
present work, the spatial distribution and other characteristics of the
FSN are highly consistent with excitatory synaptic currents being generated within and near to layer 4. However, we found no evidence for
the expected reversal of this potential in superficial cortex. This
could be due to a domination of the excitatory extracellular synaptic
currents by nonpyramidal elements with relatively local dendritic
fields. Although thalamocortical afferents to layer 4 of barrel cortex
are known to synapse onto spiny stellate, nonspiny stellate, and
pyramidal neurons (White 1989
), only the latter population would contribute to a reversal of this potential in upper
layers. Recent work in the thalamocortical slice preparation shows that
fast-spike GABAergic interneurons of layer 4 show large unitary EPSPs
with rise times to thalamocortical synapses that are much faster than
those of pyramidal neurons (Gibson and Connors 1998
).
The summation of the extracellular currents generated by the
synchronous generation of such EPSPs could contribute heavily to the
extracellular field potential elicited by single thalamocortical afferents. Thus extracellular currents generated by these elements and
other nonpyramidal neurons may dominate the extracellular record.
Alternatively, the lack of reversal in the FSN in superficial cortex
could be due to limitations in the methods used here to analyze unitary
field potentials. Current source-density analysis (e.g.,
Mitzdorf 1985) has not yet been used to study currents generated by unitary presynaptic elements. This method may be well
suited to localize such minute currents, but it requires a larger
number of vertically spaced sampling points than was used here.
Intracellular studies in a wide range of species and sensory systems
have shown that an ubiquitous disynaptic IPSP follows the monosynaptic
EPSP that is generated by a synchronous thalamocortical barrage (e.g.,
Creutzfeldt and Ito 1968; Ferster and Lindstrom 1983
; Hellweg et al. 1977
). One might suspect
that the occurrence of prominent di/polysynaptic IPSPs would require
synchronous generation of action potentials, such as those resulting
from potent sensory or electrical stimulation. Indeed, previous work
examining spike-triggered averages of intracellular postsynaptic
potentials or extracellular fields have shown little or no evidence of
disynaptic inhibition. In the present work, however, the FSP resulting
from single "spontaneous" VB thalamocortical impulses was of
greater amplitude and duration than any other component of the
spike-triggered average. Unlike the FSN, this response did show a
clear-cut reversal in superficial cortex. This is expected because
pyramidal neurons are thought to provide the synaptic targets of most
inhibitory interneurons of sensory cortex (White 1989
).
One explanation for the large magnitude of the FSP in barrel cortex is
given below.
Current source-density analysis of responses generated by gross sensory
or electrical stimulation of afferent pathways has revealed a vertical
cascade of excitatory events following initial activation of layer 4 (Mitzdorf 1985). Our single-fiber analysis yields no
evidence of such a vertical cascade, with all presumed active current
sinks and sources reaching a peak within or near to layer 4. This
result is, perhaps, not surprising. A vertical cascade of
excitation would require the generation of action potentials in pyramidal or spiny stellate neurons, and single thalamocortical impulses may be rarely sufficient for this task. In contrast, the
inhibition reflected by the FSP is thought to result from IPSPs generated by the activation of fast-spike GABAergic interneurons within and near layer 4 (see below). These cells have very low thresholds to excitatory input and do discharge at reasonably high
"efficacy" following VB thalamocortical impulses (Swadlow 1995
). As noted above, the axons of these neurons are largely restricted to the layer-4 barrel in which they are found. The resulting
outward synaptic currents would therefore be expected to reach a peak
near layer 4.
Conclusions: effects of individual VB thalamocortical neurons on an S1 barrel
Our results indicate that a single action potential in a single VB thalamocortical neuron results in a clear sequence of events that is largely limited to a single S1 barrel: 1) the arrival of the VB impulse in layer 4 (the AxTP), 2) monosynaptic excitation generated by this axonal input (the FSN), and 3) a long-duration positive wave (the FSP) that we interpret to reflect local di- or poly-synaptic inhibition.
Previous work in rabbit barrel cortex (Swadlow 1995) has
identified a class of fast-spiking (Amitai and Connors
1995
; Kawaguchi and Kubota 1993
;
McCormick et al. 1985
; Simons 1978
),
putative feed-forward inhibitory interneurons (suspected
inhibitory interneurons, SINs) within and near layer 4. Results
derived from cross-correlation and microstimulation indicated that most
SINs within a layer-4 barrel receive monosynaptic input from a large
number of the VB thalamocortical neurons of the topographically aligned
VB barreloid. They are thus engaged in a secure, highly
divergent/convergent relationship with neurons of the aligned VB
barreloid (a "complete network") (cf. Abeles 1991
;
Griffith 1963
).
A predictable consequence of such divergent/convergent networks is a
degree of "sharp" synchrony among the recipient elements (Moore et al. 1970; Sears and Stagg
1976
), and SINs within an S1 barrel indeed demonstrate sharp
synchrony (Swadlow et al. 1998
). Thus when recording the
spike trains of two SINs randomly selected within the same barrel,
~4% of the spikes of the two SINs are coincident (±1 ms), creating
a sharp peak in the cross-correlogram. Given ~350 SINs in an S1
barrel,7 one may
conclude that sharply synchronous events between SINs are not limited
to the two cells under study, but will include other SINs as well (the
number will depend on the degree of dependence or independence of
synchronous events). We have previously argued that the sharply
synchronous activity among these elements will result in a
corresponding synchronous, feed-forward release of GABA onto
postsynaptic targets, and that such summated IPSPs may be especially
effective when excitatory drive to target cells is weak and
asynchronous (Swadlow et al. 1998
). We suggest that the
FSPs observed here reflect this sharply synchronous, feed-forward inhibitory process, and that a functional consequence of sharp synchrony among feed-forward GABAergic interneurons is to amplify the
impact of individual VB impulses on inhibitory mechanisms of the neocortex.
![]() |
ACKNOWLEDGMENTS |
---|
This work was supported by National Science Foundation Grant IBN-9723967 and National Institute of Mental Health Grant MH-59322 to H. A. Swadlow and Russian Foundation of Basic Research Grant 98-04-48208 to A. Gusev.
![]() |
FOOTNOTES |
---|
1 These electrodes were generously provided by the University of Michigan Center for Integrated Sensors and Circuits. The "Schmidt 2" model was used in these experiments.
2
We made several unsuccessful attempts to
antagonize postsynaptic components of the spike-triggered average
(Kirkwood et al. 1991) by iontophoresis of DNQX (e.g.,
Swadlow and Hicks 1997
). This may be due to the large
size of the barrels studied (~300-400 µm diam) and the relatively
limited spread of agents by this method.
3 In three cases our estimates of FSN latency were supported by subsequently eliminating the FSN using DNQX (e.g., Fig. 2, A and B).
4 Mean amplitude values of topographically aligned components of the STA are somewhat less in Fig. 6 than in Table 1. This is because Table 1 is derived from only those 17 spike-triggered averages in which the AxTP and FSN exceeded an amplitude of 1 µV. Values here are from all 25 spike-triggered averages obtained from topographically aligned VB-S1 pairs.
5
This is achieved by dividing the Fourier
transform of the spike-triggered average waveform by the Fourier
transform of the autocorrelogram of the presynaptic VB neuron, followed
by an inverse Fourier transform (Brillinger et al. 1976;
Eggermont et al. 1993
).
6 This idealized kernel can be thought of as the waveform that might result from a single, temporally isolated spike in a single VB thalamocortical neuron that was uncorrelated with spikes of any other neurons that project to the recording site.
7
This is based on an estimated barrel diameter
and thickness of 400 µm (McMullen et al. 1994a) and a
density of parvalbumen-positive GABAergic interneurons of
7,100/mm3 (McMullen et al. 1994b
).
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.
Received 23 September 1999; accepted in final form 17 December 1999.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|