Receptive field (RF) structures of neurons in area TE of the monkey inferior temporal cortex were investigated under blockade of inhibition mediated by γ-aminobutyric acid (GABA). Bicuculline methiodide, a GABAA receptor antagonist, was microiontophoretically administered to TE neurons. Blockade of inhibition enhanced responses to a particular range of visual stimuli not only at the RF center, but also at the periphery of or outside RFs where the stimuli originally evoked little or no response, enlarging the RFs. The strongest responses under normal and disinhibited conditions occurred at the RF center in most neurons. The largest increase in responses, reflecting the strongest inhibitory input, usually occurred at the RF center, but in some neurons it occurred in the periphery. A neuron had a silent region within its RF where some stimuli effective at adjacent locations could not elicit responses even under blockade of inhibition. We suggest that (i) afferent information to individual TE neurons originates from a wide retinotopic region beyond their normal RF; (ii) the afferent convergence is not necessarily complete throughout a RF; and (iii) GABAergic inhibition contributes to the generation of RF structures of TE neurons.
The neural processes that lead to visual perception and recognition of objects occur in the occipitotemporal pathway of the primate brain (Ungerleider and Mishkin, 1982; Goodale and Milner, 1992). The response selectivity of neurons for object features becomes gradually complex along this pathway. Meanwhile, the receptive field (RF), a part of the visual field from where a neuron receives information, increases in size. All early visual areas (V1, V2, V4 and TEO) of this pathway are organized in a retinotopic manner and represent the central and contralateral visual fields (Hubel and Wiesel, 1974; Gattass et al., 1981,1988; Boussaoud et al., 1991). At the final stage, TE neurons have a large RF, almost always covering the fovea (the center of gaze), and most of the RFs extend into the ipsilateral field across the vertical meridian (Gross et al., 1972; Desimone and Gross, 1979). TE neurons selectively respond to a complex object feature (Desimone et al., 1984; Tanaka et al., 1991; Fujita et al., 1992). This response selectivity is largely maintained within their RFs (Schwartz et al., 1983; Lueschow et al., 1994; Tovee et al., 1994; Ito et al., 1995). The translation-invariant selectivity has been thought to underlie invariant recognition of visual objects across frontoparallel translation (Gross and Mishkin, 1977; Seacord et al., 1979).
γ-Aminobutyric acid (GABA) is a major inhibitory neurotransmitter of the cerebral cortex. Earlier studies have shown that blockade of intracortical inhibition mediated by GABA reduces orientation, direction and length preferences of area 17 neurons in the cat visual cortex (Sillito, 1975,1977; Sillito and Versiani, 1977; Tsumoto et al., 1979). Recent studies also indicate that GABA-mediated inhibition is involved in the generation of orientation, direction and color selectivities of macaque V1 neurons (Sato et al., 1994,1995,1996). In the cat areas 17 and 18, blockade of the inhibition changes RF subfield structures of neurons. During the blockade, On and Off subfields of simple cells in space and time and those of complex cells in time become less isolated or even completely superimposed (Sillito, 1975; Wolf et al., 1986; Eysel et al., 1994; Eysel and Shevelev, 1998; Pernberg et al., 1998; Murthy and Humphrey, 1999; Frégnac and Shulz, 1999). In area 17, GABAergic inhibition contributes to neuronal sensitivity for cross-like figures (Shevelev et al., 1998). In the rat, cat and macaque somatosensory cortex, blockade of GABAergic inhibition markedly increases the RF size of neurons (Dykes et al., 1984; Alloway et al., 1989; Alloway and Burton, 1991; Kyriazi et al., 1996). In the macaque prefrontal cortex, disinhibition results in a loss of the spatial selectivity for the position of a visual object in most neurons or an enhancement of the selectivity in a small portion of neurons that are not tuned originally. The effects are observed in the sensory, mnemonic and motor phases of the oculomotor delayed response task (Rao et al., 2000). Blockade of GABAergic inhibition also expands the representation fields of the body motor map in the rat primary motor cortex. Inactivation of GABAergic inhibition within the forelimb representation endowed the adjacent cortical area controlling the vibrissa movement with the ability to govern the forelimb movement (Jacobs and Donoghue, 1991). These results indicate that GABAergic inhibition plays important roles in the generation of RF properties of neurons in the primary sensory, motor, and higher-order association cortices.
In the macaque area TE, 25% of neurons are immunoreactive to antibodies raised against GABA or GABA-synthesizing enzymes, and are distributed throughout the cortical layers (Hendry et al., 1987; Tanigawa et al., 1998). Our recent study has shown that GABAergic inhibition is critically involved in the generation of selectivity of TE neurons to stimuli shown at the RF center (Wang et al., 2000). In the present study, we presented stimuli at different locations within and outside RFs and investigated the role of GABAergic inhibition in the construction of the RF structures of TE neurons by microiontophoretically administering bicuculline methiodide, a GABAA receptor antagonist, to TE neurons. Stimulus selectivity of TE neurons was first identified by the ‘reduction process’, in which the stimulus feature critical for activation was determined by stepwise simplification of the most effective stimulus (Tanaka et al., 1991; Fujita et al., 1992; Kobatake and Tanaka, 1994). Then, visual responses of a TE neuron were tested by presenting stimuli at the RF center and four surrounding locations, or at seven locations along the horizontal dimension across its RF under normal and disinhibited conditions. The procedure enabled us to examine the contribution of GABAergic inhibitory input to the RF structures of the functionally characterized TE neurons.
Materials and Methods
Animal Preparation and Surgical Procedure
Two adult monkeys (Macaca fuscata), weighing 5.5 and 7.2 kg, were prepared for repeated recording experiments. The recording experiments were performed once a week on each monkey. All surgical and animal care procedures complied with the guidelines of the National Institutes of Health (1996), and were approved by the animal experiment committee of Osaka University Medical School.
For the preparatory aseptic surgery, the initial anesthesia was induced with ketamine hydrochloride (10 mg/kg, i.m.) after atropine sulfate (0.25 mg, i.m.) had been administered to the monkeys. Anesthesia was maintained with 35 mg/kg pentobarbital sodium (i.p.), supplemented at 5–10 mg/kg/h when necessary. The monkeys were placed on a stereotaxic apparatus. A plastic block for restraining the head was mounted to the skull with stainless steel or plastic screws and acrylic resin. The temporal part of the skull was exposed and covered with resin for subsequent unit recordings. The wound was infiltrated with a local anesthetic (xylocaine) during the surgery and treated with iodine and an antibiotic powder after the surgery. After the monkeys were allowed to recover from the surgery for >2 weeks, the optics of their eyes were checked under anesthesia to select appropriate contact lenses. The curvature of the cornea was first measured using a kerato-refractometer (Topcon, KR-7100, Japan). After a pair of transparent contact lenses with an appropriate curvature was placed on the corneas, the refractive power (diopter) of the eyes was measured. Then, photographs of the fundus of the eyes were taken with a retinal camera (Topcon, TRC-50X, Japan) to determine the positions of the optic disc and the fovea.
In the recording experiments, the monkeys were initially anesthetized with ketamine and atropine. An endotracheal cannula was inserted into the tracheal opening of each monkey. The monkeys were artificially respirated with a mixture of air and 1–2% isoflurane for the subsequent anesthesia. A small hole for electrode penetration was drilled in their skulls over the temporal cortex. During the recording sessions, the monkeys were immobilized with pancuronium bromide (0.02–0.04 mg/kg/h, i.v.), infused with physiological saline which contained 5% glucose and atropine sulfate (0.005 mg/kg/h, i.v.), and anesthesia was maintained with a mixture of N2O and O2 (7:3) and 0.5–1.0% isoflurane. The electrocardiogram and expired CO2 level (4.0–4.5%) were monitored throughout the duration of the experiment. Body temperature was maintained at 37–38°C. The pupils were dilated and the lenses were relaxed by local application of 0.5% tropicamide and 0.5% phenylephrine. The corneas were protected with contact lenses of appropriate refractive power and curvature with an artificial pupil (diameter 3 mm) to focus on a CRT monitor placed 57 cm away from the corneas.
After each recording session, the hole on the skull was cleaned with saline, treated with several drops of 8 mg/ml Decadron (dexamethasone sodium phosphate) on the dura and then covered with small pieces of aseptic medical sponges, antibiotic cream and resin. Within 1 h following the termination of pancuronium bromide, N2O–O2 and isoflurane administration, the monkeys recovered their spontaneous respiration to the normal level. Then, the artificial respiration was stopped. Before returning to their cage, the monkeys were injected with 0.5 mg Vagostigmin (neostigmine methylsulfate) to ensure recovery of their spontaneous respiration and then with an antibiotic. Each recording experiment lasted <16 h from the introduction of anesthesia to the recovery of spontaneous respiration of the monkeys.
Neuronal Recording and Visual Stimuli
Recordings were performed in the dorsal part of area TE (A10–14) in the inferior temporal cortex. The position of each penetration was determined with reference to a marker placed on the resin-coated skull. In one monkey, electrode penetrations were guided by magnetic resonance imaging. Within the hole made in the skull over the temporal cortex, a pinhole was made in the dura using a stainless steel needle with a shaft 0.7 mm in diameter. Then a triple-barreled, glass-coated tungsten microelectrode (see below) was inserted into the cortex through the pinhole using a conventional microdriver. The exposed dura was covered with wax to prevent it from drying and to reduce movement of the cortex caused by pulsation and ventilation. Extracellular action potentials were recorded from single TE neurons and amplified by conventional techniques. Action potentials from single neurons were isolated based on their wave form and amplitude using a window discriminator. Only neurons that were well isolated from other neurons and the background noise were used for the present analysis. The timing of action potentials was recorded using a computer with 1 ms resolution.
After neuronal activity from a single neuron was isolated, we determined its stimulus selectivity. Visual stimuli were presented to the eye contralateral to the recorded hemisphere. Visual responses of a neuron were first tested using >90 hand-held real objects and many two-dimensional paper cutouts to search for effective stimuli. Images of effective stimuli were then obtained with a video camera and modified on a computer. Images of these stimuli were presented on a CRT monitor. The critical stimulus feature essential for activation of a neuron was determined by stepwise simplification of the most effective stimulus, i.e. the ‘reduction process’ (Tanaka et al., 1991; Fujita et al., 1992; Kobatake and Tanaka, 1994). We then selected the most effective and simplest stimulus and two other effective or ineffective stimuli for analysis of RFs.
For each neuron tested, the position and size of the RF were first explored by presenting the selected stimuli at the center and four surrounding locations of a region covering the central part of the visual field. Based on responses obtained at the five locations, we adjusted the central position of the CRT monitor to the visual field where the strongest responses had been evoked. The responses of the neuron were again tested at the five locations to estimate the size and the center position of the RF. We then carried out the following experiments.
In one series of experiments, the stimuli were presented at the center and four surrounding locations covering the visual field of 7.5–17.5° × 7.5–17.5°. The covered area depended on the size of the preliminarily determined RF. When visual stimulation at the surrounding locations could not elicit responses even during application of bicuculline (see below), the surrounding stimuli were presented closer to the RF center and the responses of the neuron were tested again. The central location of the stimulus presentation corresponded to the center of the RF predetermined for each neuron. Thus, the central location of the stimulation was not always at the fovea of the visual field, because the RF centers of TE neurons were usually off from the fovea. Hereafter, when we mention ‘ipsilateral’ or ‘contralateral’ sides of RFs, this is with respect to the RF center and not with respect to the fovea.
In order to evaluate more quantitatively the effects of blocking GABAergic inhibition on the size and structure of RFs, we designed another series of experiments. Visual stimuli were presented at seven positions spanning 40° with a regular horizontal interval (6.7°) and the central stimulus location corresponded to the center of the predetermined RF.
In both series of experiments, stimuli were ~2.5° × 2.5° in size. The darkest luminance of the stimuli was 0.7 cd/m2 and the brightest was 99.2 cd/m2. The stimuli were presented for 1 s at non-overlapping local regions with a small-amplitude motion (up to 0.5°) on a gray background (15.7 cd/m2). The time interval between two stimulus presentations was 1.5 or 2 s and each stimulus was presented 10 times. Because only a small number of the stimuli were used and TE neurons become habituated to a visual stimulus presented repeatedly at a short interstimulus interval (Gross et al., 1972; Miller et al., 1991), successive presentation of the same stimulus, even at different locations, was avoided in these experiments. Otherwise, the sequence of the stimuli was pseudorandom.
Triple-barreled microelectrodes were constructed as described elsewhere (Takao et al., 2000). One of the two glass barrels was filled with 2.5 or 5 mM bicuculline methiodide (pH 3.5, adjusted with HCl; Sigma) and the other with 0.9% NaCl (vehicle, pH 3.5). A low-intensity current (+1 to 5 nA) was used to eject a minimal amount of bicuculline to the immediate vicinity of the recorded neurons. Occasionally, bicuculline application caused epileptiform activity in the recorded neurons. In such cases, we decreased the ejection current until the activity disappeared. Otherwise, the recording was not performed at these sites. In most cases, the low-intensity current used (+1 to 5 nA) did not result in epileptiform activity. When bicuculline was not administered, a retaining current of –20 nA was employed to prevent leakage of bicuculline. Under all conditions, a balancing current with the same intensity and opposite polarity was applied to the vehicle barrel to offset possible current effects on the recorded neurons. When the current polarity of the bicuculline barrel was reversed, the current polarity of the vehicle barrel was automatically reversed. We recorded responses of the neurons to the stimulus sets before, during and after bicuculline application. Usually, the effects of bicuculline on visual responses were tested after bicuculline had been applied for 5 min and recovery data were collected at least 5–10 min after bicuculline application was stopped. The electrodes allowed us to obtain single-unit recordings with a high signal-to-noise ratio and to maintain the single-unit recording for a sufficient time to finish three sessions before, during and after the bicuculline application. Usually, the entire process of examining a neuron, including the determination of its stimulus selectivity by the reduction process, took >2 h.
Firing rates of visual responses during 1 s of stimulus presentation were averaged from 10 presentations. Net visual responses were obtained by subtracting the spontaneous firing rate, which was measured during the 0.5 s prior to the presentation of each stimulus, from the visual response rate. A two-way analysis of variance (ANOVA, stimulated location versus drug condition) was applied to the net firing rates evoked by a stimulus before and during bicuculline application to detect effects of bicuculline on the RF of a neuron. We considered that bicuculline affected the RF if the responses to one of the stimuli used were significantly affected by bicuculline (Drug factor or Drug × Location interaction, P < 0.05). The ANOVA was also applied to the net firing rates before and after bicuculline application to assess whether a neuron recovers its original RF profile from the effects of bicuculline. We considered that a neuron had recovered from the effects of bicuculline if the responses to the stimulus between before and after bicuculline application were not significantly different (both Drug factor and Drug × Location interaction, P > 0.05). In order to examine interaction between bicuculline application and visual stimuli on RF structures, a three-way ANOVA with stimulus, location and drug as factors was applied to the net firing rates for each neuron.
A t-test was used to examine whether a stimulus evoked significant responses at a location relative to the spontaneous firing rate and also whether bicuculline significantly enhanced net responses at a location during bicuculline application in comparison with those before bicuculline application. Net responses at seven locations along the horizontal dimension were plotted against the horizontal positions and interpolated by cubic interpolation (MatLab), in which the interpolated curve passed through each of the original data points. Hereafter, we refer to the interpolated curve as an ‘RF tuning curve’. The width of an RF tuning curve was measured at the half-magnitude level of the peak responses (see below for more details). For all the statistical tests, the significance level was P < 0.05.
We examined the effects of bicuculline on the RFs of 38 single TE neurons recorded from three hemispheres of two adult monkeys. The preferred stimuli of all these 38 neurons were determined by the ‘reduction process’ (see Materials and Methods). Of the 38 neurons, 17 were tested using stimuli presented at the RF center and four surrounding locations, which were at the periphery of the RF for some neurons or outside the RF for others. The remaining 21 were tested at seven locations along the horizontal dimension across their RFs.
RFs under Normal Conditions
The RF size of the TE neurons under normal conditions (i.e. in the absence of bicuculline) exhibited a considerable extent of variation (Fig. 1). Some neurons responded to the most effective stimulus only at one location, while others responded to the stimulus presented at surrounding locations 20° away from the RF center. The neuron shown in Figure 1A, for example, responded strongly to a luminance gradation pattern presented at the center (0°), but showed very weak or no responses to the stimulus presented at locations >6.7° from the center. The width of the RF tuning curve measured at the half-magnitude of the peak response (W0) was 8.1°. Another neuron (Fig. 1B) responded to a different pattern of luminance gradation presented at five locations spanning 26.7°. The W0 of this neuron was 22.7°. The half-peak width of RFs among the 21 neurons tested at seven locations ranged from 7.6° to 31.7° (15.4 ± 6.9°, mean ± SD, Fig. 1C). The number of locations where the most effective stimulus evoked significant responses in a neuron was 3.1 ± 1.4 on the average (± SD, Fig. 1D). Seventy-one per cent of these neurons (15/21) responded to those presented at three or more locations. Only one of the 21 neurons responded at six locations (Fig. 1D); hence, the stimulated range (40°) in the seven-location experiments covered the entire RF of most of the studied neurons in the horizontal dimension. Most RF tuning curves were symmetrical approximately at the RF center under normal conditions (Figs 1A,Band 4). A few neurons (two of 21) exhibited an asymmetric RF tuning profile.
Effects of Bicuculline on RF
Bicuculline affected the visual responses in 16 of the 17 neurons tested at five locations and in 20 of the 21 neurons tested at seven locations (ANOVA, Drug factor and/or Drug × Location interaction, P < 0.05). All of these neurons recovered from the effects after the application of bicuculline was stopped (both Drug factor and Drug × Location interaction, P > 0.05). The following analyses were based on the data of the 16 and 20 neurons.
Bicuculline markedly enhanced responses of neurons to stimuli presented not only at the RF center as we showed in our previous study (Wang et al., 2000), but also at the peripheral locations, where the stimuli originally evoked weak or no visual responses. Figure 2 shows the effects of bicuculline on responses of two neurons. They were recorded from two nearby sites along a vertical penetration and both responded preferentially to faces. The first neuron (Fig. 2A) originally responded to a human face presented only at the RF center. Following bicuculline application, it became responsive (P < 0.01) to the face at the four surrounding locations outside its original RF (8.8° from the center) as well as significantly increasing its responses at the center. This effect was observed only for the face stimulus, not for the other two stimuli, a concentric circle and a horizontal bar (P > 0.05). The second neuron with a large RF (>24.8°) significantly augmented its responses to a monkey face at the center and all four peripheral locations (12.4° from the center, Fig. 2B). This neuron also became responsive to a human face (for image, see Fig. 2A), which was ineffective across its large RF before bicuculline application (P > 0.05), at the left two surrounding locations (not shown; P < 0.05). These effects did not, however, occur in the responses to a horizontal bar (P > 0.05).
Figure 3 summarizes the effects of bicuculline on responses of the 16 neurons to their respective most effective stimulus. For the five neurons with small-size RFs tested at 5.3° (Fig. 3A), the average response at the four surrounding locations during bicuculline application was 458 ± 168% (±SD) of the original responses, while that at the center was 146% of the original response. For the seven neurons with middle-size RFs tested at 8.8° (Fig. 3B), the two values were 318 ± 53% and 248%; for the four neurons with large-size RFs tested at 12.4° (Fig. 3C), the two values were 226 ± 50% and 183%, respectively. The data demonstrate considerable effects of blocking of GABAergic inhibition on responses of TE neurons throughout their RFs (but see Fig. 10).
Data obtained from the 20 neurons in the seven-location experiments provided a quantitative assessment of the effects of bicuculline on the size and structures of the RFs. The effects were evaluated using two parameters: gain index and width of the RF tuning curves. The gain index (GI) is defined as (Ad – Ab)/Ab, where Ad and Ab are the areas (spikes/s × degrees) under the RF tuning curves during and before bicuculline application (see Fig. 5E), respectively. The unit of the index indicates that the area under an RF tuning curve during bicuculline application is increased by 100% from that before bicuculline application. The width of the RF tuning curve was measured at the half-magnitude of peak responses before (W0) or during (W1) bicuculline application, respectively. We also measured the width of the RF tuning curve during bicuculline application at the half-magnitude of the original peak responses (W2, see Fig. 5E).
Figure 4 shows five examples of neurons, the RFs of which were affected by bicuculline in different ways. The first neuron had a small RF to the most effective stimulus, a luminance gradation pattern (W0 = 7.6°, Fig. 4Aa). Bicuculline increased the area under the RF tuning curve by 200% (GI = 2.03). The half-peak width was broadened by 2.4° when the width before bicuculline application was compared with that during bicuculline application at the respective half-peak levels (ΔW1 = W1 – W0, Fig. 5E). When the widths of the RFs before and during bicuculline application were compared at the same original half-peak level, the change in the RF width was 8.7° (ΔW2 = W2 – W0, Fig. 5E). A plot of the normalized responses indicates that the RF profile of this neuron actually became broader following bicuculline application (Fig. 4Ab). The second neuron responded best to a black cross and had a large RF under normal conditions (W0 = 25.8°, Fig. 4Ba). The area under the RF tuning curve was doubled (GI = 1.08). The width was broadened by 13° at the original half-peak level (ΔW2), although the overall shape of the RF profile of the neuron was not changed (ΔW1 = 0.2°, Fig. 4Bb). Figure 4Ca shows that responses of a neuron to a cross of white outline were increased mainly at the RF central region. The GI was 0.71 for this neuron. The neuron had a large RF (W0 = 23.2°) and the width at the original half-peak level became larger by 3.5° (W2), whereas the change at the respective half-peak levels was –5.3° (ΔW1). The overall shape of its RF tuning curve became sharper in comparison with that before bicuculline application (Fig. 4Cb). The responses of the fourth neuron to a luminance contrast pattern were augmented mainly at the contralateral side of the RF (Fig. 4Da,b; for the definition of ‘contralateral’, see Materials and Methods). The half-peak width of the RF was increased by 11.9° (ΔW1) and the width measured at the original half-peak level by 17.3° (ΔW2). The neuron had a GI of 1.35. The fifth neuron had significantly enhanced responses to a red circle at both sides of the peak, but not at the peak (Fig. 4E). The half-peak width was increased by 9.3° (ΔW1) with a GI of 0.75. The overall RF was markedly broadened at both sides (ΔW2 = 9.2°, Fig. 4Eb). The results indicate that bicuculline enlarged the RFs of most neurons and also altered the shape of the RF tuning curve in many of them.
As shown in Figure 4, bicuculline shifted upward the entire or part of the RF tuning curves in individual neurons. Figure 5A shows the distribution of GIs for 47 RFs, which were tested using three stimuli for each of the 20 neurons and were affected by bicuculline (ANOVA, Drug factor and/or Drug × Location interaction, P < 0.05). The minimal and maximal GIs were 0.36 and 5.16, respectively. Eight-five per cent (40/47) of the RFs had a GI > 0.5 (e.g. the RFs in Fig. 4C,E) and 51% of the RFs (24/47) had a GI > 1 (e.g. the RFs in Fig. 4A,B,D). On average, the area under the RF tuning curve during bicuculline application was increased by 150% (GI = 1.47 ± 1.24, mean ± SD).
The change in the size of RFs was first assessed based on a change in the number of locations where a stimulus elicited significant responses of a neuron. The three neurons shown in Figure 4A,D,E, for example, originally responded to their preferred stimuli only at one location (RF center). Following bicuculline application, all of these neurons responded at three locations (P < 0.05). The distribution of the number of effective locations for the 47 RFs analysed in the 20 neurons was shifted by bicuculline toward larger values than that before bicuculline application (Mann–Whitney's U-test, P < 0.01; Fig. 5B). On average, the neurons responded at 2.4 ± 1.5 (mean ± SD) locations before bicuculline application and at 3.3 ± 1.5 locations during bicuculline application.
The change in the size of RFs was also evaluated as a change in the width of RF tuning curves. Figure 5C shows the effects of bicuculline on the width of 43 RF profiles that had a clear response peak both before and during bicuculline application among the 47 RFs. The three RFs in which neurons showed no visual responses (e.g. the RF of Fig. 8B) before bicuculline application and one RF in which a neuron had significant but weak responses (<5 spikes/s) at several locations before bicuculline application were excluded from the analysis. The width at the original half-peak level was increased from 14.1 ± 6.1° before bicuculline application (W0) to 22.4 ± 8.2° during bicuculline application (W2) (mean ± SD, n = 43). The change was significant (U-test, P < 0.0001, Fig. 5C).
We also compared the width of RFs between before (W0) and during bicuculline application (W1) at their respective half-peak levels. In the normalized plots of the RF tuning curves shown in Figure 4Ab–Eb, a difference between the two values (ΔW1 = W0 – W1) reflects a change in the shape of the RF tuning curve. For 35% of the RFs (15/43, Fig. 5D), the difference was >2°. The shape of these RFs became broader during bicuculline application than before bicuculline application (e.g. the RFs in Fig. 4A,D,E), due to a larger amount of disinhibition at the RF periphery. For the other 21% of the RFs (9/43), the difference was <−2° (ΔW1) and their shape became sharper, due to a higher disinhibition at the RF center than at the periphery (e.g. the RF in Fig. 4C). The shape of the remaining 44% (19/43) of the RFs did not obviously change during bicuculline application (−2° ≤ ΔW1 ≤ 2°, e.g. the RF in Fig. 4B). For these RFs, the degree of disinhibition at the RF center and periphery during bicuculline application was approximately in proportion to the original response magnitudes at each location before bicuculline application. Thus the disinhibition altered the shape of the RF tuning curve in 56% of the RFs (24/43), suggesting that structures of these RFs were under the influence of GABAergic inhibition.
The effects of bicuculline on RF structures showed a considerable variation among the neurons (Figs 2–5). The change in the width at the original half-peak level (ΔW2 = W0 – W2, Fig. 5C), for example, ranged from −1.8 to 23.6°. The difference in the effects on the size of RFs may depend on the amount of bicuculline applied. A larger amount of effectively expelled bicuculline would remove a larger amount of inhibition and could affect RFs to a greater extent. To examine this possibility, we analysed the relationship between the change in ΔW2 and the change in the response magnitude at the original peak. We assumed that the latter is indicative of the amount of disinhibition, i.e. the amount of bicuculline applied. The change in the half-peak width was not correlated to the magnitude of increase in peak responses (r = 0.19, P = 0.22, Fig. 6). The difference in the changes of the RF width induced by bicuculline was not likely due to the difference in the amount of applied bicuculline.
Stimulus-dependency of Bicuculline Effects
We have previously shown that the effect of bicuculline on the responses of TE neurons at the RF center depends on stimuli (Wang et al., 2000). In the present study, we show that the effect of bicuculline is stimulus-dependent also at the peripheral region of the RF. Figure 7 shows an example of a neuron. The RF response profile of this neuron to one stimulus, a gradation pattern, was significantly affected by bicuculline (two-ANOVA, P< 0.05, Fig. 7A). In contrast, the RF response profiles to a gradation pattern of the opposite gradient (Fig. 7B) and to another luminance pattern (Fig. 7C) were not affected (two-ANOVA, P > 0.05), although the two stimuli were also effective before bicuculline application. Responses of this neuron at the RF center to the stimulus in Figure 7A during bicuculline application were higher than those to the most effective stimulus before bicuculline application (Fig. 7B). The failure to enhance the responses to the stimuli shown in Figure 7B,C is thus not due to saturation of firing rate in this neuron. Seven of 38 neurons tested in the five-location or seven-location experiments exhibited this kind of stimulus-specific disinhibition. Bicuculline only affected responses of these neurons to a portion of stimuli. A relatively small number of neurons showed this kind of effects compared to our previous results on the responses at the RF center (Wang et al., 2000). This is probably because the number of stimuli tested for each neuron was small (three) in the present experiments.
In order to evaluate the stimulus dependency of bicuculline effects on the RF, a three-way ANOVA with stimulus, location and drug as factors was performed for each neuron tested (Table 1). When effects of bicuculline were examined across all the stimuli used for a neuron, visual responses of 71% (27/38) of neurons were significantly affected by bicuculline application (Drug factor, P < 0.05, third column of Table 1). The percentage was lower than that obtained when a two-way ANOVA was applied to each stimulus (95%, 36/38, see above and Materials and Methods). This is because, for nine neurons of the 36 neurons showing bicuculline effects in the two-way ANOVA test, one or two of the three stimuli used were ineffective even during the application of bicuculline, or one or two stimuli were effective before the bicuculline application but the responses to them were not affected by bicuculline (e.g. the neuron in Fig. 7). A significant interaction between stimulus and bicuculline was observed in 63% (24/38) of neurons (Stimulus × Drug interaction, P < 0.05, sixth column of Table 1), suggesting that bicuculline effects on an RF are dependent on stimuli in these neurons across the stimulated locations. In 58% (22/38) of neurons, the effects of bicuculline were dependent on the stimulated location (Location × Drug interaction, P < 0.05, fifth column of Table 1). The results indicate that these neurons received spatially specific inhibitory inputs; the distribution of inhibitory inputs is different across the stimulus locations but is similar among the stimuli used. Finally, a significant interaction among stimulus, location and bicuculline occurred in 13% of neurons (5/38, Stimulus × Location × Drug interaction, P < 0.05, seventh column of Table 1). In these neurons, bicuculline caused different changes in RF structures when the RFs were tested with different stimuli. This suggests that this population of neurons received spatially specific inhibitory inputs derived from different stimuli. The fractions of the latter three groups of neuron are likely to be an underestimate, because only a small number of stimuli was used in these experiments.
These results indicate that the disihibitory effects of bicuculline on the RF cannot be explained simply by a general increase in excitability of neurons. The effects are dependent on the excitatory inputs of the stimuli that were under the influence of GABAergic inhibition.
Silenced Excitatory Inputs throughout RF
We have previously shown that many TE neurons under blockade of inhibition begin to respond to originally ineffective stimuli presented at the center of RFs (Wang et al., 2000). Similar phenomena occurred also in the peripheral regions of the RFs. The neuron shown in Figure 8A, for instance, originally did not respond to a pattern composed of a yellow circle and a green star (S1) at all the tested locations. Following the application of bicuculline, it responded to the pattern presented at the RF center and three of the four peripheral locations (5.3°). The neuron also began to respond to an image of the bottom view of a tomato (S2) presented at four surrounding locations (5.3°). This neuron originally showed no responses at these locations and only weak responses at the center. By contrast, this neuron responded to another stimulus, the side view of a peach, at all the five locations under control conditions, and enhanced its responses at three surrounding locations during bicuculline application.
Three of the 16 neurons tested in the five-location experiments exhibited a similar behavior to one originally ineffective stimulus, and another neuron to two originally ineffective stimuli. Similar effects were observed in two of the 20 neurons tested at seven locations; an example is shown in Figure 8B. Because the number of stimuli used in the present experiments was small (three stimuli) and we selected effective stimuli whenever possible, the proportion of neurons that showed this behavior was much lower (17%, 6/36) than that reported in our previous study (84%) on the responses at the RF center (Wang et al., 2000). The results indicate that excitatory inputs masked by GABAergic inhibition under normal conditions exist in the RF periphery as well as at the RF center.
Excitatory and Inhibitory Input in the Surrounding Regions
Most RFs (84%, 36/43) had no obvious change in the position of the response peak (≤2°, Fig. 9) following bicuculline application and the strongest responses remained at the original RF center (e.g. Fig. 4A–D). This indicates that excitatory inputs to most neurons were strongest at the center of their RFs with or without GABAergic inhibition.
However, some neurons showed the largest disinhibition at the RF periphery (e.g. Fig. 4D,E). The disinhibition in the surrounding regions shifted the response peak of 16% (7/43) of the RFs from the original location by >2° (Figs 4E and 9). In these neurons, a GABAergic inhibitory input predominant in the surrounding regions generates their RF profiles under normal conditions.
‘Hole’ without Input within RFs
One neuron had a ‘hole’ within the RF where visual inputs of some stimuli effective at adjacent regions did not appear to reach. The neuron shown in Figure 10 did not respond to a stimulus, a half-white and half-black ellipse, presented at position 3 (−6.7°), whereas it exhibited clear visual responses to the same stimulus presented at both sides of this location (positions 2 and 4) under normal conditions (Fig. 10A). Bicuculline application enhanced its responses to the stimulus at positions 2 and 4, but not at position 3 (Fig. 10A). The results suggest that the neuron did not receive input of the stimulus feature at this local region within its RF. Responses of the neuron to the second stimulus, a gradation pattern, under normal conditions were similar to those to the first stimulus (Fig. 10B). Following bicuculline application, however, the stimulus elicited responses of the neuron at position 3. Thus, the neuron received inputs of the second stimulus at this location that were silenced by GABAergic inhibition under normal conditions. Importantly, the third stimulus, another gradation pattern, at the location activated this neuron (Fig. 10C) under normal conditions, indicating that the absence of responses to the first stimulus was not attributable to a deficit in the retina. Indeed, we did not detect any abnormalities in the retina by observing the fundus of both eyes with a retinal camera and an ophthalmoscope. Afferent information on the stimulus feature shown in Figure 10A thus did not completely cover the entire RF in this neuron. The size of this silent ‘hole’ was at least 3° (size of a stimulus and its movement magnitude, 2.5° + 0.5°).
The same neuron did not respond to all three stimuli presented at position 6 under both normal and disinhibited conditions (P > 0.05). The RF center of the neuron was a few degrees contralateral to the fovea. Retinoscopic examination showed that the optic disc of the retina was 15.8° from the fovea on the contralateral side. Given that the distance of the silent region at position 6 was 13.3° from the RF center, the region should correspond to the ‘blind spot’ or the local visual field corresponding to the optic disc of the retina where there are no photoreceptors.
The response behaviors of this neuron to the three different stimuli presented inside the silent ‘hole’ and the ‘blind spot’ indicate that the differences between the responses of neurons to the different stimuli during bicuculline application are related to whether there exist excitatory inputs of the stimulus features at a location of RFs and to whether the excitatory inputs are under GABAergic inhibition, rather than to stochastic fluctuations in responses induced by blockade of GABAergic synapses.
The present experiments showed that: (i) removal of GABAA-receptor-mediated inhibition enhanced visual responses of TE neurons at the periphery of or outside RFs, as well as at the center of RFs, thus resulting in enlargement of RFs of many TE neurons; (ii) the largest disinhibition usually occurred at the RF center, but some neurons had the strongest inhibitory input in the periphery; and (iii) stimulus inputs completely suppressed by GABAergic inhibition existed not only at the RF center, but throughout the RF. The results suggest that GABAergic inhibition plays a crucial role in the construction of normal RF structures in TE neurons.
In addition, the study provided a quantitative measure of the size of normal RFs of TE neurons by presenting stimuli at seven locations along the horizontal dimension across 40°. As a recent study reported (De Beeck and Vogels, 2000), most of the RFs had a bell-shaped profile with the most sensitive region located at the central portion, and the RF size considerably varied among TE neurons. The width measured at the half-peak level of RF tuning curves was 15.4 ± 6.9°. Although an effective stimulus usually evoked responses throughout an RF, we found a neuron that had a ‘hole’ within its RF where inputs of some stimulus features effective at adjacent stimulus locations did not reach. This suggests that the convergence of inputs from earlier areas in the cortical hierarchy is not necessarily complete throughout the RF.
A detailed methodological consideration on the application of bicuculline has been given elsewhere (Fujita and Konishi, 1991). Here we discuss two technical points that we did not address previously. The first concerns the size of the volume affected by iontophoresed bicuculline. Although we did not attempt to determine experimentally the extent of bicuculline diffusion, the effective diffusion distances of iontophoretically applied dopamine, penicillin and d-2-amino-5-phosphonovaleric acid (APV, a glutamate receptor antagonist) in brain tissues have been reported by several research groups. The effective diffusion distance of APV is estimated to be 100–200 μm from the site of iontophoresis in the cat visual cortex (Fox et al., 1989). Diffusion dynamics analysis indicates that the effective action range of dopamine and penicillin is within a radius of ~150 μm from the site of iontophoresis (Lehmenkühler et al., 1991; Nicholson, 1995).
The diffusion distance of bicuculline in the present study is likely to be <100–200 μm, because the mol. wt of bicuculline (509.3) is larger than that of APV (197.1), dopamine (153.1) and penicillin (149.2), and because we used a small pipette tip (2–4 μm in diameter) and a low ejecting current (+1 to 5 nA). The spherical cortical region in the radius of ~150 μm contains 600–800 neurons (Rockel et al., 1980; Lehmenkühler et al., 1991). Most likely bicuculline affected not only the recorded single neuron, but also a group of nearby neurons.
Bicuculline could also exert network effects on the observed neurons from distant neurons within area TE via long-range horizontal axons (Fujita and Fujita, 1996) and from neurons in afferent areas via feedforward and feedback processes. Inactivation of the local inhibitory circuit around the site of iontophoresis could facilitate these mutual excitatory inputs, either by disinhibiting nearby neurons, in turn facilitating the recorded neuron, or by disinhibiting the recorded neuron, then facilitating nearby neurons, distant neurons within area TE or in afferent areas, and affecting the recorded neuron again via a re-entering process. All these effects are, however, due to the removal of the local GABAergic inhibition around the recorded neuron.
The second technical issue concerns the use of balance currents. The balance currents may induce an uptake of bicuculline into the vehicle barrel during the long ejection period and the bicuculline would be released from the barrel once the current was switched to the retaining current with a comparatively high intensity and the opposite polarity. This is theoretically possible, but is unlikely to be a practical problem. In the present experiments, we used low ejection currents of +1 to +5 nA which expelled a minimal amount of bicuculline. The balance currents with the same intensity and opposite polarity may induce an uptake of a small proportion of the ejected bicuculline into the vehicle barrel. Note that the expelled bicuculline would be diluted in extracellular space and would be diluted again with 0.9% NaCl if it was taken up into the vehicle barrel. When the polarity of the current was switched, the high retaining current of +20 nA from the vehicle barrel would quickly drive out this microamount of bicuculline. Usually, we collected the recovery data at least 5–10 min after the polarity of the current was switched. During this period, bicuculline should have been diffused. We believe that what is described above is likely the case, because almost all neurons in the present experiments and 89% of neurons in the previous study (Wang et al., 2000) restored their responses to the original level after the application of bicuculline was stopped.
Construction of RF Structure and Selectivity of TE Neurons
Blockade of GABAergic inhibition enlarged RFs of most of the TE neurons. The shape of the RF tuning curve, however, did not change in the presence of bicuculline in 44% of the tested RFs (Figs 4B and 5D), suggesting a simple gain control of visual responses by GABAergic inhibition in these neurons. In the other 56% of the RFs, the shape of the RF tuning curve was altered by bicuculline (Fig. 5D). The effects could be explained by some specific interactions between neurons. If interconnected neurons have RFs with different sizes but with their RF centers at similar retinotopic positions, the removal of inhibitory inputs from neurons with large RFs would result in the RF shape of postsynaptic neurons becoming broader (35% RFs), whereas the removal of inhibition from neurons with small RFs would result in the RF shape of postsynaptic neurons becoming sharper (21% RFs). Some neurons had strong excitatory inputs at a peripheral region in the RF, which were partially masked by inhibition (Figs 4D,E and 10A,B). These excitatory inputs at sites off from the RF center may arise from other TE neurons having RF centers at different retinotopic locations or from neurons at different retinotopic regions of earlier cortical areas TEO or V4.
A previous study suggested that the selectivity of neurons for object features is generated in areas V4 and TEO, and transferred to area TE (Kobatake and Tanaka, 1994). The response selectivity for a particular feature with invariance of the stimulus position is hypothesized to be generated by convergent projections from neurons having the same stimulus selectivity but with different RF positions in these earlier areas. The hypothesis needs two prerequisites. First, each local retinotopic region in the afferent areas must contain a set of neurons selective for different object features that are represented in area TE. Second, each neuron in area TE must receive convergent projections of the same feature from all different retinotopic sites of its afferent areas.
Our previous study has shown that a substantial fraction of excitatory inputs to TE neurons at the RF center is masked by GABAergic inhibition (Wang et al., 2000). The marked enlargement of RFs of TE neurons induced by bicuculline in the present study indicates that visual information converges upon TE neurons from wider retinotopic regions beyond their normal RFs. If each TE neuron or a population of neurons in a columnar region (Fujita et al., 1992) receives such projections specific for a stimulus feature but converging from a wide visuotopic area, the wiring between the afferent cortical areas and TE would be massive. An anatomical study has, however, shown that axons from a local region in area TEO terminate in only a few multiple columnar regions of 200–300 μm in width in area TE, not in broadly divergent regions of area TE (Saleem et al., 1993).
At the RF center, TE neurons receive multiple, heterogeneous stimulus inputs, some of which are masked by GABAergic inhibition, and the stimulus features of the masked inputs are related to those of the originally effective stimuli for these neurons or for nearby neurons (Wang et al., 2000). The present results indicate that the masked inputs exist not only at the RF center, but also throughout the entire RF. These inputs to the entire RF must also be involved in constructing the RF structures and selectivity of TE neurons.
The distinctive response behaviors of the neuron shown in Figure 10 to the three stimuli at position 3 under normal and disinhibited conditions suggest that inputs of the three stimulus features have different projections to the RF in this local region. In addition, the response patterns of the neuron in Figure 8A indicate that the distributions of the excitatory inputs of two analogous stimuli are shifted with respect to each other and have different gravity centers of projections within the RF. Thus, afferent inputs converging to individual TE neurons from different retinotopic sites do not necessarily possess the same feature and may have different projection distributions within the RF of a TE neuron. The convergence of afferent inputs is also not necessarily complete throughout the RF (Fig. 10). The present results give rise to a scheme different from that proposed by Kobatake and Tanaka (Kobatake and Tanaka, 1994). We propose that the generation of the feature selectivity and the large RF of TE neurons is accomplished in area TE both by the convergent processes of afferent inputs and by excitatory and inhibitory interactions mediated by neuronal circuits within area TE. This scheme is different from that of Kobatake and Tanaka (Kobatake and Tanaka, 1994) in three respects. First, the construction of feature selectivity of TE neurons continues within area TE (Wang et al., 2000). Second, afferent inputs converging to a TE neuron can convey heterogeneous stimulus features (Wang et al., 2000). Third, afferent inputs may have different projection distributions within the RF of a TE neuron.
Source of GABAergic Inhibition
There are three possible sources for the GABAergic inhibition we observed in this study. The first one is a short-range inhibition driven by excitatory, recurrent inputs of local pyramidal neurons. The second is a ‘long-range’ inhibition within area TE. Because axons of most GABAergic interneurons so far identified in area TE (Tanigawa et al., 1998) project horizontally only in a range <1 mm, this type of inhibition is mostly driven by long-range excitatory horizontal axons of pyramidal neurons at distant sites within area TE. GABAergic inhibition driven by excitatory glutamatergic synapses of long-range connections has been demonstrated in the ferret area 17 (Weliky et al., 1995). Large basket cells in cat areas 17 and 18 project their horizontal axons up to a distance of 3.8 mm (Somogyi et al., 1983; Kisvárday and Eysel, 1993; Kisvárday et al., 1993). They receive intracortical inputs from pyramidal cells or other large basket cells and afferent excitatory inputs from early areas (Somogyi et al., 1983; Kisvárday et al., 1993; Kisvárday and Eysel, 1993). Their counterparts in area TE, although not demonstrated yet, may also be engaged in this long-range inhibitory process. The third one is a ‘long-range’ inhibition from afferent areas TEO and V4. This inhibition is driven most likely by long-range excitatory projections from these afferent areas to area TE, as those from area 17 to area 18 of the rat visual cortex (Johnson and Burkhalter, 1996; Shao and Burkhalter, 1996; Gonchar and Burkhalter, 1999). It is unknown whether a ‘direct’ long-range inhibitory projection from areas TEO and V4 to area TE exists, as the direct long-range projection of inhibitory neurons between area 17 and area 18 in the rat visual cortex (McDonald and Burkhalter, 1993).
GABAergic Contribution to Generating RF Structures in Other Cortices
GABAergic inhibition participates in generating RF structures of the primary visual cortical neurons. Blockade of GABAergic inhibition by iontophoretic application of bicuculline changes RF structures of neurons in the cat visual areas 17 and 18 (Sillito, 1975,1977; Sillito and Versiani, 1977; Wolf et al., 1986; Eysel et al., 1994; Eysel and Shevelev, 1998; Pernberg et al., 1998; Murthy and Humphrey, 1999; Frégnac and Shulz, 1999). During bicuculline application, On and Off subfields of RFs of simple cells in space and time and those of complex cells in time become more overlapping with each other. Thus, simple cells, at least outside layer IV, receive both On and Off excitatory inputs across most parts of the RFs, which are similar to the excitatory On and Off inputs of the kind found in complex cells. The spatial specificity of intracortical inhibition is suggested to contribute to spatially isolated On and Off subfields in simple cells (Frégnac and Shulz, 1999). Shunting inhibition mediated by GABAA receptors is involved in the separation of On and Off subfields of simple cells and the mixed On/Off field of complex cells (Borg-Graham et al. 1998). GABAergic inhibition is also involved in response selectivity of cat area 17 neurons for cross-like patterns (Shevelev et al., 1998) and On/Off response characteristics of color-selective neurons in macaque V1 (Sato et al., 1994).
The contribution of GABAergic inhibition to the spatial receptive field is not limited to the visual cortex. Blockade of GABAergic inhibition by bicuculline changes the size, structure and response properties of RFs of neurons in a variety of cortical areas such as the rat barrel cortex (Kyriazi et al., 1996), the cat and macaque primary somatosensory cortex (Dykes et al., 1984; Alloway et al., 1989; Alloway and Burton, 1991), the macaque prefrontal cortex (Rao et al., 2000) and the rat primary motor cortex (Jacobs and Donoghue, 1991).
Neurons in the visual and somatosensory cortex expand their RFs after lesions of afferent areas, deprivation or abnormal experience of sensory inputs, artificial scotoma, and their RFs are also dependent on the synchronized state of the electroencephalogram under anesthesia and contextual stimulus-contrast under awake conditions (Gilbert, 1998; Wörgötter et al., 1998; Kapadia et al., 1999; Polley et al., 1999). RF size of cortical neurons is thus not fixed and can vary in a dynamic manner. GABAergic inhibition may be engaged in some aspects of the dynamic regulation of RFs of cortical neurons.
The RF and stimulus selectivity of TE neurons are generated by integrating information of multiple inputs originating from a wide retinotopic region beyond their normal RF. These inputs are not necessarily similar to each other in the stimulus feature information they convey and may have distinctive projection distributions within the RF of a TE neuron. The afferent convergence is not necessarily complete throughout a RF. Intrinsic GABAergic inhibition is critically involved in the construction of the normal RF structures and feature selectivity of TE neurons from these inputs.
We thank John Maunsell for his comments on the manuscript. Magnetic resonance images were taken at the National Institute for Physiological Sciences. This work was supported by a CREST grant from the Japan Science and Technology Corporation and grants from the Japanese Ministry of Education, Science, Sports and Culture (to I.F.). In the earlier stages of this work, Y.W. was supported by a fellowship from the Uehara Memorial Foundation and Y.M. by a fellowship from the Japan Society for the Promotion of Science for Young Scientists.
|Three-way ANOVA||Factors (P < 0.05)||Interactions (P < 0.05)|
|Stim||Loc||Drug||Stim × Loc||Loc × Drug||Stim × Drug||Stim × Loc × Drug|
|aOf the 38 neurons tested in the five- and seven-location experiments. Stim, stimulus; Loc, location.|
|Five-location experiment (n = 17)||17||16||12||12||9||13||3|
|Seven-location experiment (n = 21)||21||21||15||15||13||11||2|
|Three-way ANOVA||Factors (P < 0.05)||Interactions (P < 0.05)|
|Stim||Loc||Drug||Stim × Loc||Loc × Drug||Stim × Drug||Stim × Loc × Drug|
|aOf the 38 neurons tested in the five- and seven-location experiments. Stim, stimulus; Loc, location.|
|Five-location experiment (n = 17)||17||16||12||12||9||13||3|
|Seven-location experiment (n = 21)||21||21||15||15||13||11||2|