| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
J Neurophysiol (January 1, 2003). 10.1152/jn.00868.2001
Submitted on Submitted 22 October 2001; accepted in final form 27 September 2002
1Department of Neurobiology; and 2Howard Hughes Medical Institute, Duke University Medical Center, Durham, North Carolina 27710
 |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
Tucker, Thomas R. and Lawrence C. Katz. Recruitment of Local Inhibitory Networks by Horizontal Connections in Layer 2/3 of Ferret Visual Cortex. J. Neurophysiol. 89: 501-512, 2003. To investigate how neurons in cortical layer 2/3 integrate horizontal inputs arising from widely distributed sites, we combined intracellular recording and voltage-sensitive dye imaging to visualize the spatiotemporal dynamics of neuronal activity evoked by electrical stimulation of multiple sites in visual cortex. Individual stimuli evoked characteristic patterns of optical activity, while delivering stimuli at multiple sites generated interacting patterns in the regions of overlap. We observed that neurons in overlapping regions received convergent horizontal activation that generated nonlinear responses due to the emergence of large inhibitory potentials. The results indicate that co-activation of multiple sets of horizontal connections recruit strong inhibition from local inhibitory networks, causing marked deviations from simple linear integration.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
In primary visual
cortex, neurons with similar receptive field properties are grouped
into functional modules, and within each module, inhibitory and
pyramidal cells are densely interconnected in a local neuronal circuit.
In addition to local connections, the axons of pyramidal cells in layer
2/3 extend up to several millimeters and form clusters interconnecting
iso-orientation domains (Bosking et al. 1997
;
Gilbert 1992; Gilbert and Wiesel 1989),
providing the potential for both local and long-range neuronal interactions, but how these interactions may affect the spatial and
temporal distribution of neuronal activity in layer 2/3 is uncertain.
Long-range horizontal connections contact both inhibitory and pyramidal
cells (Hirsch and Gilbert 1991; McGuire et al.
1991; Weliky et al. 1995), evoking a balance of
excitation and inhibition that is poorly understood. Thus it is unknown
whether the excitation delivered by multiple horizontal connections
increases pyramidal cell excitability or recruits local inhibitory
circuits and suppresses pyramidal cells. Following electrical
stimulation in brain slices, horizontal connections generate
subthreshold excitatory postsynaptic potentials in pyramidal cells
(Yoshimura et al. 2000), but the convergence of a
sufficient number of horizontal inputs may enable pyramidal cells to
generate action potentials. Excitatory postsynaptic potentials (EPSPs)
delivered by horizontal connections may be augmented by temporal
summation (Thomson and Deuchars 1994), persistent sodium
currents (Stafstrom et al. 1985), and
N-methyl-D-aspartate (NMDA) receptors
(Artola and Singer 1990; Sutor and Hablitz
1989), supporting the idea that convergent horizontal pathways
may increase pyramidal cell excitability. However, an alternative
possibility is that convergent horizontal connections may reduce
pyramidal cell activation by recruiting local inhibition. While
horizontal connections evoke purely excitatory events after weak
electrical stimulation, they generate compound EPSP/inhibitory
postsynaptic potentials (IPSPs) following strong stimulation. Thus
convergent horizontal connections, which generate purely excitatory
events when stimulated individually, may evoke inhibition when
stimulated in combination. This idea is supported by evidence that
excitation and inhibition are inseparable (Douglas and Martin
1991; Somers et al. 1995, 1998).
To determine how local neuronal populations and pyramidal cells in layer 2/3 integrate convergent horizontal connections, we stimulated two sets of horizontal connections in tangential slices of layer 2/3 of ferret visual cortex, and assayed the spatiotemporal distribution of neuronal activity with intracellular recording and voltage-sensitive dye imaging. We found that responses generated by horizontal connections were integrated nonlinearly due to the emergence of strong inhibitory synaptic potentials.
| |
METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
Slice preparation and optical recording
Adult ferrets (P40-P60, Marshall Farms, North Rose, NY) were
killed under pentobarbital sodium anesthesia (100 mg/kg, ip), and
tangential brain slices of layer 2/3 were cut to 350 µm using a
vibratome. Slices were incubated in an interface chamber with voltage-sensitive dye (0.1 mg/ml, RH461, Grinvald et al.
1987, Molecular Probes, Eugene, OR) in normal artificial
cerebrospinal fluid (ACSF; in mM: 125 NaCl, 1.3 MgSO4, 2.8 CaCl2, 4 KCl, 1 KH2PO4, 10 dextrose, and 26 NaHCO3) for approximately 90 min at 28°C, and
subsequently transferred to a submersion chamber perfused with warmed,
oxygenated ACSF at 30°C on a Zeiss Axiovert 100 TV microscope for
recording. Electrical stimuli were delivered with concentric bipolar
electrodes (FHC, Bowdoinham, ME) driven with isolated current sources
(A360, WPI, Sarasota, FL) and triggered by a pulse generator (Master-8,
AMPI, Jerusalem, Israel). A 250-W lamp driven by a stable power supply
(ATM75-15M, Kepco, Flushing, NY) delivered light through a 546 ± 20-nm filter (Chroma Technology, Brattleboro, VT), the epifluorescence
port, and a 10× objective lens (Zeiss Fluar, 0.5NA) to the slice.
Emitted light was filtered at 590 nm longpass, and directed to the
bottom port to a photodiode array with 256 elements collecting images
at 2 kHz (16 × 16 square array, C5897, Hamamatsu, Bridgewater,
NJ). In each trial, slices were exposed for 325 ms with intertrial
intervals of 10 s, using an electromechanical shutter (VS35,
Vincent Associates, Rochester, NY). Images were averaged over 25-50
trials to improve the signal/noise ratio, and then processed digitally
by tripling the number of pixels in each dimension (yielding images
48 × 48 pixels), temporally filtered at 500 Hz and spatially
filtered with a Gaussian kernel (5 × 5). For displaying the
images as three-dimensional movies in pseudocolor, additional
processing was performed using IP-Lab (Scanalytics, Fairfax, NJ),
and IDL (Research Systems, Boulder, CO).
|
Electrophysiology
Sharp electrodes were fabricated from borosilicate glass tubing
(WPI, Sarasota, FL) to have resistances of 90-130 M
and filled with
3M KAc. Intracellular recordings were obtained in current-clamp mode
(AxoClamp-2B, Axon Instruments), low-pass filtered at 3 kHz (Model 410, Brownlee Precision, San Jose, CA), and digitized at 20 kHz (AT-MIO-16E,
National Instruments, Austin, TX). Optical and electrical recordings
were simultaneously acquired on PC computer running custom software
written in C++. Following a recording, cells were filled with
neurobiotin (Vector Labs, Burlingame, CA) by current injection (+0.5
nA, 500-ms pulses at 1 Hz), and the slices were fixed for subsequent
processing. Neurons included in the analysis had pyramidal cell
morphologies and resting potentials more negative than 
65 mV with
overshooting action potentials. Using a programmable micromanipulator
(SM1, Luigs and Neumann, Ratingen, Germany), intracellular recordings
were targeted to selected neuronal populations during image
acquisition, enabling simultaneous acquisition of optical and
electrical recordings from the same cortical locus. Following an
experiment, the exact locations of the stimulating and recording
electrodes were determined by acquiring a brightfield image with a CCD
camera (C2400, Hamamatsu) and aligning it with images acquired from the
photodiode array.
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
In previous experiments, we imaged tangential slices of visual cortex stained with voltage-sensitive dyes to observe patterns of optical activity evoked by horizontal connections, delivering electrical stimuli at single cortical sites. We found that these optical patterns included two features: a diffuse zone of activation surrounding the stimulation site, and numerous ovoid domains of activity, or optical clusters, corresponding to axonal clusters of horizontal connections (see companion paper). Here, we use a similar approach to study interactions between patterns of activity driven by horizontal connections, delivering electrical stimuli at two cortical sites. In the first part, we investigate how integration of horizontal inputs is represented in the spatiotemporal distribution of population-based neuronal activity, imaging with voltage-sensitive dyes. In the second part, we use intracellular recording to investigate how horizontal inputs are integrated by individual pyramidal cells.
Spatial distribution of emergent inhibition
To investigate the integration of horizontal inputs by neuronal populations, we positioned two stimulating electrodes up to several hundred microns apart in tangential slices of ferret visual cortex layer 2/3, enabling the stimulation of two predominantly distinct sets of horizontal connections. As the resulting patterns extended over 1.2 mm in radius, there was substantial overlap in the two patterns of activity, enabling interaction between the neural populations driven by each set of connections. The experiments were conducted by first stimulating each electrode separately to determine the pattern of responses evoked by each individual set of connections (Fig. 1, A and B), and then stimulating both electrodes simultaneously to observe interactions between the two patterns. As the simplest outcome would have been linear summation, we added the two patterns evoked by separate stimulations to obtain an "expected response" (Fig. 1C) for comparison with the "actual response" of stimulating both electrodes together (Fig. 1D).
The actual spatiotemporal distribution of responses was substantially different from expected, indicating that the interaction could not be described by simple addition. To quantify these differences, we subtracted the expected from actual response and obtained a "difference image" (Fig. 1E) that facilitated the identification of regions that deviated most strongly from the expected. These difference images revealed large ovoid regions of negative-going responses approximately 500 µm in diameter [full-width at half-maximum (FWHM)], indicating that the interaction between the two patterns of connections yielded domains of strong suppression.
As described in the companion paper, optical responses evoked by electrical stimulation of horizontal connections consist of a diffuse zone and optical clusters. We considered three different types of interactions, including those between 1) two optical clusters, 2) an optical cluster and a diffuse zone, and 3) two diffuse zones. We found that suppression domains with similar characteristics were evoked by all types of interaction, suggesting that they depended primarily on the overlap in the activity patterns, rather than the specificity of horizontal connections formed by optical clusters or diffuse zones.
In general, suppression domains were centered on optical clusters when two discrete sets of horizontal connections evoked patterns having overlapping optical clusters (Fig. 1). Next, there were cases in which an optical cluster evoked by one horizontal pathway overlapped the diffuse zone generated by a different pathway. In this situation, suppression domains were centered on optical clusters residing within the diffuse zone, and optical clusters outside of this perimeter were unaffected (Fig. 2). Finally, when the stimulation sites were separated by several hundred microns or less, there was substantial overlap between two diffuse zones, and the resulting interaction yielded suppression domains coincident with the overlapping area (Fig. 3).
|
|
To quantify these responses, we identified the center of suppression domains and used this location to compare the time courses of the actual and expected responses. At this site, the duration of the actual response was significantly shorter than expected, having a peak width (FWHM) of approximately 13 versus 25 ms. The amplitude of the actual response was also substantially smaller than expected throughout its entire time course, but the strongest suppression occurred during the decay phase, 4.9 ± 2.2 ms after the peak. At this point the amplitude of the actual response was typically about 70% of the expected response (Table 1).
|
The suppression domain's combined effect on amplitude and time course bears the hallmarks of synaptic inhibition. To investigate the cellular mechanism of suppression, we obtained targeted intracellular recordings from neurons within suppression domains, monitoring both optical and electrical responses simultaneously. Following two-site stimulation, the intracellular recordings invariably revealed the emergence of large inhibitory postsynaptic potentials, events that were absent following single-site stimulation. Given the strong correlation between optical and electrical recordings, and the evidence that suppression domains rely solely on the convergence of horizontal connections, we focused on the mechanism of this "emergent inhibition" using intracellular recording.
Intracellular recordings of emergent inhibition
In visual cortex, horizontal connections elicit spikes in
inhibitory, but not pyramidal cells (Hirsch and Gilbert
1991). Therefore EPSPs recorded in pyramidal cells following
horizontal activation are strictly monosynaptic, whereas IPSPs are
mediated by disynaptic connections through local inhibitory
interneurons. Due to the absence of polysynaptic excitation,
intracellular recordings from pyramidal cells provide a direct
measurement of both excitation and inhibition driven by horizontal
connections. We used this framework to determine how the interaction
between two discrete sets of horizontal connections recruited the
activity of local inhibitory networks.
Pyramidal cells had resting potentials near the chloride reversal potential, making chloride-mediated IPSPs small and difficult to quantify; therefore cells were moderately depolarized (10 ± 2 mV) via current injection (200 ± 50 pA) to facilitate detection and measurement of IPSPs. Using the same stimulation protocol described above, electrodes were stimulated separately to determine individual contributions to response; these were added to determine the expected response, and simultaneous stimulation of electrodes yielded the actual response. The actual responses consistently included large inhibitory PSPs, much more hyperpolarizing than responses to separate stimulations or the expected response (Fig. 4, A and B). The emergence of large IPSPs was particularly striking when separate stimulations generated a pair of depolarizing events, but simultaneous stimulation evoked a large hyperpolarization (Figs. 5, A and B, and 6, A and B).
|
|
|
To quantify this disparity between responses evoked by separate and
simultaneous stimulations, we subtracted the expected from the actual
responses, which yielded large hyperpolarizing potentials up to 10 mV
in amplitude (3.5 ± 2.2 mV, n = 46). As these
inhibitory PSPs emerged during simultaneous stimulations, but were
absent from the sum of separate stimulations, we termed these
"emergent IPSPs." Thus the emergent IPSP is the additional inhibition evoked by simultaneous stimulation, which was not
anticipated from the responses evoked by separate stimulations. Through
the remainder of the paper, we use this term to refer to the difference between the actual and expected responses, a difference which in every
case indicated the recruitment of hyperpolarizing inhibition.
Several features of the emergent IPSP match those of GABAA
receptor-mediated IPSPs. First, the emergent IPSP had an onset latency
consistent with disynaptic activation, occurring 3.2 ± 0.8 ms
after the onset of the EPSP. Because EPSPs are delivered directly to
pyramidal cells, but IPSPs are delivered through intervening inhibitory
cells, the onset of IPSPs is delayed relative to EPSPs. Second, the
time course of the hyperpolarization is consistent with IPSPs mediated
by GABAA receptors. For emergent IPSPs, the rise time
(7.4 ± 4.2 ms), the FWHM (30.7 ± 7.9 ms), and the decay time constant (36.0 ± 18 ms), were all in the range of values reported for GABAA receptor-mediated IPSPs
(Buhl et al. 1994; Thomson and Deuchars
1997; Thomson et al. 1996; Tamas et al.
1997, 1998).
Because hyperpolarizing responses may be driven by intrinsic
conductances as well as GABAergic inhibitory postsynaptic potentials, we tested directly the identity of the hyperpolarizing response by
partially blocking GABAA receptors with low
concentrations of bicuculline (3 µM, Fig. 5). This eliminated all
hyperpolarizing events evoked by both separate and simultaneous
stimulations, abolished the emergent IPSP (4.1 ± 2.0 vs.
0.02 ± 0.49 mV, n = 3), and made actual and
expected responses nearly identical.
The behavior of the emergent IPSP was investigated over a broad range of stimulation strengths. Using strong stimuli, large IPSPs were evoked in the recorded pyramidal cell when either of the two sites was stimulated separately, and the sum of these two responses yielded potentials that were more negative than the chloride reversal potential, a physiologically unrealistic prediction that confounded assessment of the emergent IPSP. Weak stimuli produced purely excitatory PSPs that summed linearly, indicating the absence of emergent IPSPs (Fig. 3, C and D). Thus emergent IPSPs were evoked using a range of moderate stimulation strengths.
Temporal properties of the emergent IPSP
As the emergent IPSP relied on integration of synaptic potentials
generated from both stimulation sites, we next investigated the
temporal properties of this integration by varying the interval between
the two stimuli (Fig. 6). The emergent IPSP was largest when stimuli
were delivered simultaneously, and decayed progressively as the
interval was increased, becoming small or absent when stimuli were
delivered 10 ms apart. Plotting the amplitude of the emergent IPSP
against the time interval between stimuli showed that the emergent IPSP
arises from a process of integration having a time constant of 6.2 ± 0.8 ms (n = 7). This time constant is very similar to that of inhibitory cells (Tamas et al. 1997, 1998),
supporting the idea that emergent IPSPs arise from a local inhibitory
network which integrates the excitatory PSPs delivered by convergent
horizontal inputs.
While the previous experiments indicated that emergent IPSPs arise from
two sets of horizontal connections driving a local neuronal population,
they did not reveal whether these connections drive the local
population equally. To test the possibility that horizontal connections
make different contributions to the emergent IPSP, we introduced a
temporal delay between the two stimuli and alternated the order of
stimulus delivery (Fig. 7). By
stimulating one electrode 5 ms both before and after the other
electrode, we found in 86% of cases that the emergent IPSP was larger
using one protocol rather than the other, having an average difference of 1.3 ± 0.4 mV (n = 7), indicating that the
order of stimulation was significant.
|
This asymmetry is unlikely to arise as an artifact of preparing tangential slices. First, the ability of both pathways to generate EPSPs and IPSPs indicated that the local neuronal circuit of inhibitory cells and pyramidal cells was intact, and that long-range horizontal connections were uninjured by slicing. Second, emergent IPSPs required inhibitory cells to receive connections from both pathways. Thus even if the slicing process selectively damaged only the synapses made by one set of horizontal connections onto local inhibitory cells, a highly unlikely event, then none of the affected inhibitory cells could contribute to the emergent IPSP, and both pathways would be impacted equally. Finally, both simultaneous and temporally-offset stimulations deliver to the local circuit identical amounts of excitation distributed in spatially identical patterns. Thus it is solely a difference in temporal distribution of synaptic activation that evokes asymmetrical emergent IPSPs.
Model of emergent inhibition
In principle, emergent IPSPs may be generated by inhibitory cells that integrate synaptic inputs from multiple horizontal connections. To test this idea, we constructed two model neuronal circuits to probe the subthreshold and superthreshold activation of inhibitory cells and to determine how synaptic potentials arising from horizontal activation were integrated by pyramidal cells and their local inhibitory network.
Both models A and B contain two pyramidal cells (P1 and P3, "projection cells") whose long-range horizontal connections make excitatory synapses with a local population of inhibitory cells (I1 and I2) and a pyramidal cell (P2, "recipient cells"). The difference between the two models is that each inhibitory neuron in model A is driven by only one projection cell (P1 or P3), whereas each is driven by both cells (P1 and P3) in model B (Fig. 8, A and B). The difference in circuit behavior between the two models is illustrated in Fig. 8, C and D. Stimulating projection cells separately evoked purely subthreshold potentials in the recipient cells of both models. In contrast, simultaneous stimulation yielded subthreshold activation of inhibitory cells for model A and superthreshold activation of inhibitory cells in model B, evoking an emergent IPSP in the recipient pyramidal cell (Fig. 8, C and D, bottom row). Thus model B is capable of generating emergent IPSPs, but model A is not.
|
Models A and B were compared with a third model,
linear summation, by varying stimulation strengths and measuring
response amplitudes evoked by simultaneous stimulation at a single time point near the peak amplitude of IPSPs (Fig. 8E, inset,
vertical line). For each model, as stimulus strength was increased,
responses initially grew more depolarizing as the amplitude of EPSPs
increased, but then they became hyperpolarizing as inhibition
intensified. However, due to the neuronal circuitry, each model
generated different amounts of inhibition. In model A,
simultaneous stimulation generated responses which were largely
identical to those of linear summation, but responses more negative
than about 1 mV were less hyperpolarizing due to inhibitory shunting.
In model B, responses diverged dramatically from linear
summation as small increases in stimulus strength evoked large,
stepwise increases in hyperpolarization due to generation of emergent
IPSPs. The steep, linear progression in IPSP amplitude was enabled by
the recruitment of increasingly larger numbers of inhibitory cells,
which offset the effect of shunting.
Next, we made a direct comparison of models A and B and linear summation by plotting the actual against expected responses (Fig. 8F). On this graph, linear summation yields a straight line having unity slope. Model A responses diverged from the linear model at negative potentials due to the effect of inhibitory shunting. In contrast, model B responses were disparate from those of linear summation, because the actual responses were much more hyperpolarizing than expected due to emergent IPSPs.
Using model B, we were able to assess the number of
inhibitory cells contributing to the emergent IPSPs in our experimental results (Fig. 8G). Plotting our experimental data in the
same format, actual versus expected responses, and generating a series of curves from the model to indicate how many inhibitory cells must be
recruited by simultaneous stimulation to obtain various levels of
hyperpolarization, we estimated that 
20 inhibitory cells contribute
to the intracellularly recorded emergent IPSPs.
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
By combining intracellular recording and voltage-sensitive dye imaging in tangential slices of ferret visual cortex layer 2/3, we investigated properties of integration in local cortical circuits driven by different sets of horizontal connections. At cortical sites where horizontal connections converged onto the same neuronal populations, neuronal responses deviated from linear integration due to the emergence of large inhibitory postsynaptic potentials, or emergent inhibition. These inhibitory PSPs reduced the amplitude and time course of membrane potential fluctuations in pyramidal cell recordings, and suppressed population-based neuronal activation in optical recordings. The results indicate that horizontal connections, by converging onto neuronal populations, recruit strong inhibition from local inhibitory networks. As the recruitment of inhibition was not specific to the different types of connections formed by horizontal collaterals, axonal clusters or diffuse zones, emergent inhibition reflects a general property of local inhibitory networks. Furthermore, these results indicate that the balance of excitation and inhibition driven by horizontal connections is strongly biased toward inhibition.
Physiological properties of emergent inhibition
The large amplitude of emergent IPSPs in pyramidal cells indicates
that the convergence of horizontal connections strongly activates local
inhibitory networks. The present experiments indicate that the largest
emergent IPSPs would require the concerted activation of 20
inhibitory cells. Previous experiments have shown that single
inhibitory cells elicit rather small changes in conductance and weak
IPSPs (<1 mV in pyramidal cells depolarized by 20 mV; Gupta
et al. 2000; Thomson and Deuchars 1997). In
contrast, in vivo intracellular recordings reveal much larger changes
in inhibitory conductance during visual processing (Anderson et
al. 2000; Borg-Graham et al. 1998; Hirsch
et al. 1998). Taken together, these data indicate that
inhibitory cells recruited by horizontal connections do not perform in
isolation during cortical processing, but as part of coordinated
inhibitory networks.
There are two factors contributing to the ability of inhibitory neurons
to perform as a local network. First, the synapses formed by axonal
clusters are distributed over many hundreds of neurons in a cortical
area of a few hundred microns in diameter. By delivering EPSPs to
numerous inhibitory cells in a local population, horizontal connections
evoke coordinated depolarization of inhibitory cells, priming these
cells for co-activation. Second, cortical inhibitory cells are coupled
by gap junctions, facilitating synchronous activation (Galarreta
and Hestrin 1999; Gibson et al. 1999). In combination, these factors may coordinate the activity of local inhibitory networks, yielding large emergent IPSPs during convergent activation by multiple sets of horizontal connections.
Role of emergent inhibition in visual processing
The potency of this effect made it possible to visualize the
spatiotemporal dynamics of emergent inhibition by imaging with voltage-sensitive dyes. The images revealed large domains of
suppression extending up to several hundred microns, sculpting
the distribution of population-based neuronal activity. The behavior of
suppression domains and emergent inhibition suggests a number of ways
that they may contribute to visual processing. First, neuronal
responses in vivo have a complicated dependence on stimulus contrast
which might be explained in terms of emergent inhibition (Levitt
and Lund 1997; Sceniak et al. 1999). For
example, simply by varying contrast it is possible for combinations of
visual stimuli to evoke either facilitation or suppression
(Polat et al. 1998), suggesting that discrete
populations of neurons in cortex may deliver either EPSPs or IPSPs.
Although this seems paradoxical, such opposing effects can be explained
by local circuit integration yielding emergent inhibition. Our
experimental results suggest that high contrast stimuli may evoke
emergent inhibition yielding suppression, while low contrast stimuli
may fail to recruit inhibitory networks and yield facilitation,
consistent with a previous model (Somers et al. 1998).
Second, delivering temporally offset electrical stimulation to
spatially distant horizontal connections yielded emergent IPSPs with
different amplitudes, suggesting that emergent inhibition may
participate in direction selectivity. When a visual stimulus is moved
between two points in visual space, the locus of neuronal activity is
shifted between two areas of cortex. By analogy to this situation, we
delivered stimuli at two separate cortical sites in layer 2/3 at
slightly different times, simulating the movement of visually-evoked
neuronal activity in cortex. Delivering these temporally offset stimuli
revealed that the amplitude of the emergent IPSP depended on the order
of stimulation, indicating that different sets of horizontal
connections do not drive inhibitory networks equally. By analogy to the
visual stimulation paradigm, reversing the order of electrical
stimulation corresponds to reversing the direction of motion of a
visual stimulus, implying that visual stimuli drifting in opposite
directions would generate different amounts of inhibition in the local
cortical circuit, yielding direction selective responses. Such
asymmetries in the spatial or temporal distribution of inhibition are a
component of many models of direction selectivity in cortical neurons
(Maex and Orban 1996; Mineiro and Zipser
1998; Suarez et al. 1995).
The emergence of large IPSPs during coincident activation of a neuronal
population by horizontal connections implicates local inhibitory
networks in detecting the coincidence of neuronal signals. Because
emergent IPSPs were generated only when horizontal connections were
stimulated within about 10 ms of each other, the local inhibitory network acted as a coincidence detector of horizontal activation. The
result of this coincidence detection was a significant reduction in the
time course of neuronal activity, because emergent IPSPs transformed
long duration depolarizations into brief, transient events. As
transient depolarizations generate temporal precision in neuronal
action potentials (Mainen and Sejnowski 1995), and inhibitory cells generate synchronous firing of pyramidal cells (Cobb et al. 1995), coincident activation by horizontal
connections may lead to synchronous activation of local pyramidal
cells. Because emergent IPSPs are generated by horizontal inputs having
temporal disparities 10 ms, neurons receiving slightly temporally
offset horizontal inputs would be expected to generate more temporally synchronized output.
| |
APPENDIX: CIRCUIT MODELS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
Circuit models were constructed with a modified version of
Neuron software (Hines 1989; Moore and Stuart
2000) using idealized integrate-and-fire neurons. The local
inhibitory neuronal circuit consisted of 20 inhibitory cells, each
making a single connection on pyramidal cell P2 with a synaptic
conductance of 10 nS and reversal potential of 75 mV
(Einh), enabling unitary IPSPs of 1 mV
during depolarizing current injection (Fig. 8). In model A,
inhibitory cells were contacted by either of two pyramidal cells, P1 or
P3, establishing two distinct inhibitory cell populations, I1 and I2,
having 10 cells each. In model B, each inhibitory cell received two synaptic connections, one from each pyramidal cell, P1 and
P3. P2 received excitatory synaptic connections from P1 and P3.
For simplicity, the conductance of excitatory synapses made by P1 and
P3 were assigned as a function of synaptic strength. Synapses made by
P1 and P3 with P2 had conductance calculated by
Gep = Sp, where
Gep is the conductance of excitatory
synapses on pyramidal cells in nanosiemens (nS), and
Sp is the stimulation strength for
pyramidal cells that varied from 0 to 35 units, enabling P2's EPSPs to
range from 0 to 6 mV with reversal potential of 0 mV
(Eex). Synapses made by P1 and P3 with
inhibitory cells had conductance calculated by the formula
Gei(Sp,J) = SiN, where
Gei is the conductance of excitatory
synapses on inhibitory cells, J is the ordinal number of the
inhibitory cell, Si is a modified stimulation strength incorporating a threshold and scaling factor for
inhibitory cell activation, and N is a formula for varying synaptic strengths across a distributed population of inhibitory cells
in the network. The inhibitory cell-adjusted synaptic strength, Si, was calculated by
Si = 2(Sp 20). Thus inhibitory cells had a higher threshold for activation than pyramidal cells, but above this
threshold, the inhibitory drive increased twice as fast as excitation
delivered to P2. The network factor for distributing synaptic weights
across the population of inhibitory cells was calculated by
N = exp(J/10), where J is the
ordinal number of inhibitory cell, enabling synaptic strength to
decrease systematically across the population. For model A,
ordinal numbers ranged from 1 to 10 for each population, I1 and I2. In
model B, P1 and P3 contacted all 20 inhibitory cells, and
ordinal numbers ranged from 1 to 20 for assigning P1 synapses; for
assigning P3 synapses, cells were numbered in reverse order (from 20 to
1) so that an inhibitory cell receiving a strong synapse from one
population received a weak synapse from the other. Thus this factor
accomplished two aims: it made the number of recruited inhibitory cells
dependent on stimulation strength, and by assigning ordinal numbers in
a crisscrossing pattern, it yielded a heterogeneous distribution of
inhibitory cells, including those driven primarily by one population and those driven relatively equally by both. To study the behavior of
these models, P1 and P3 were driven to spike by 200-µs electrical stimuli, while P2 was depolarized by 12 mV via current injection of 500 pA.
Idealized neurons were constructed from cylindrical compartments.
Pyramidal cells had two dendritic compartments (5 × 500 µm,
diam-by-length), an axon (0.5 µm × 2 mm), and a soma (30 × 30 µm). Inhibitory cells had two dendrites (5 × 500 µm),
an axon (0.5 × 100 µm), and a soma (20 × 20 µm).
Inhibitory synapses were located on the soma, and excitatory synapses
were placed on the dendrites, and their conductances were modeled by an
alpha function having a 5-ms time constant for excitatory synapses and
6-ms time constant for inhibitory synapses. Axons contained
Hodgkin-Huxley style sodium and potassium channels with
GNa = 0.5 S/cm2
and GK = 0.1 S/cm2. Additional parameters were
ECl = 75 mV,
ENa = 50 mV,
EK = 77 mV,
Cm = 1 µF/cm2,
m = 2 ms, Ri = 25 M, Ra = 200 -cm,
Gleak = 300 µS/cm2, and
Eleak = 75 mV.
Using model B, we estimated the number of inhibitory cells that were responsible for generating emergent IPSPs in the experimental data (Fig. 8G). Across the full range of stimulation strengths, we controlled the number of inhibitory cells recruited by simultaneous stimulation of S1 and S2 and generated curves for actual and expected responses (each inhibitory cell activating a synaptic conductance of 10 nS). Assigning the recruitment by simultaneous stimulation to zero inhibitory cells yielded responses identical to model A. The majority of experimental data points were located among curves obtained by varying the recruitment of inhibitory cells in model B from 2 to 20, allowing an estimate of the number of inhibitory cells contributing to the emergent IPSP of each pyramidal cell response.
| |
ACKNOWLEDGMENTS |
|---|
We thank D. Fitzpatrick and R. Mooney for helpful discussions and critical comments on the manuscript. We are grateful to R. Timberlake and the physics shop at Duke University for support.
This study was supported by National Institutes of Health Grants EY-07960 to L. C. Katz and NS-10952 to T. R. Tucker. L. C. Katz is an investigator in the Howard Hughes Medical Institute.
| |
FOOTNOTES |
|---|
Address for reprint requests: T. R. Tucker, Box 3209 Duke University Medical Center, Dept. of Neurobiology, Durham, NC 27710 (E-mail: ttucker{at}neuro.duke.edu).
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
METHODS
RESULTS
DISCUSSION
APPENDIX
REFERENCES
|
|---|
V) and the
optical response from a single photodiode (
4
) and model
B (
) following simultaneous stimulation of S1
and S2. In addition, responses are shown for linear summation
(
) from adding responses obtained by stimulating S1 and
S2 separately. Inset: responses generated by models A
and B and linear summation for stimulation strengths 20, 25, 26.5, and
28.5 units. For graphing, responses were measured at a single time
point near the peak of the IPSP (dotted line). Scale bar is 2 mV and 10 ms. Responses were purely excitatory for stimulation strengths