Evidence for a contribution of lateral inhibition to orientation tuning ...

European Journal of Neuroscience, Vol. 10, pp. 2056–2075, 1998 © European Neuroscience Association

Evidence for a contribution of lateral inhibition toorientation tuning and direction selectivity in cat visualcortex: reversible inactivation of functionally characterizedsites combined with neuroanatomical tracing techniques

John M. Crook, Zoltan F. Kisvarday and Ulf T. EyselDepartment of Neurophysiology, Faculty of Medicine, Ruhr-University of Bochum, D-44780 Bochum, Germany

Keywords: [3H]-nipecotic acid, areas 17 and 18, basket cells, biocytin, cross-orientation inhibition, iso-orientation inhibition

Abstract

We have previously reported that cells in cat areas 17 and 18 can show increases in response to non-optimalorientations or directions, commensurate with a loss of inhibition, during inactivation of laterally remote,visuotopically corresponding sites by iontophoresis of γ-aminobutyric acid (GABA). We now present anatomicalevidence for inhibitory projections from inactivation sites to recording sites where ‘disinhibitory’ effects wereelicited. We made microinjections of [3H]-nipecotic acid, which selectively exploits the GABA re-uptakemechanism, , 100 µm from recording sites where cells had shown either an increase in response to non-optimal orientations during inactivation of a cross-orientation site (n 5 2) or an increase in response to the non-preferred direction during inactivation of an iso-orientation site with opposite direction preference (n 5 5).Retrogradely labelled GABAergic neurons were detected autoradiographically and their distribution wasreconstructed from series of horizontal sections. In every case, radiolabelled cells were found in the vicinity ofthe inactivation site (three to six within 150 µm). The injection and inactivation sites were located in layers II/III–IV and their horizontal separation ranged from 400 to 560 µm. In another experiment, iontophoresis of biocytin atan inactivation site in layer III labelled two large basket cells with terminals in close proximity to cross-orientationrecording sites in layers II/III where disinhibitory effects on orientation tuning had been elicited. We argue thatthe inactivation of inhibitory projections from inactivation to recording sites made a major contribution to theobserved effects by reducing the strength of inhibition during non-optimal stimulation in recurrently connectedexcitatory neurons presynaptic to a recorded cell. The results provide further evidence that cortical orientationtuning and direction selectivity are sharpened, respectively, by cross-orientation inhibition and iso-orientationinhibition between cells with opposite direction preferences.

Introduction

γ-aminobutyric acid (GABA) is the major inhibitory neurotransmitterin mammalian cerebral cortex (for reviews see Curtis & Johnston,1974; Krnjevic, 1984). Every fifth neuron and 15% of synapticboutons in cat visual cortex contain GABA (Gabbott & Somogyi,1986; Beaulieu & Somogyi, 1990), and every cortical cell receives arich GABAergic input (Freundet al., 1983; Somogyi, 1989). Resultsof experiments employing iontophoretic application of GABAA-antagonists close to a recorded cell (Sillito, 1977, 1979; Tsumotoet al.,1979; Sillito et al., 1980) established the importance of GABAergicinhibitory processes for orientation tuning and direction selectivity incat visual cortex. Further evidence for a contribution of intracorticalinhibition to both properties has accumulated from studies employingmultiple visual stimuli (Emerson & Gerstein, 1977; Morroneet al.,1982; Ganz & Felder, 1984; Emersonet al., 1987; Bonds, 1989) orlocal inactivation techniques (Eyselet al., 1988, 1990; Crooket al.,1991, 1996, 1997; Crook & Eysel, 1992; Allison & Bonds, 1994)

Correspondence:Dr J. M. Crook, Leibniz Institute of Neurobiology, Brenneckestrasse 6, D-39118 Magdeburg, Germany.

Received 22 October 1997, revised 4 February 1998, accepted 11 February 1998

and from a number of intracellular recording studies (Creutzfeldtet al., 1974; Innocenti & Fiore, 1974; Satoet al., 1991; Volgushevet al., 1993; Peiet al., 1994; but see Ferster, 1986). However, theway in which GABAergic mechanisms are utilized in the synapticcircuits determining orientation/direction selectivity and the functionalspecificity of the underlying inhibitory connections remain controver-sial issues (for reviews see Chapman & Stryker, 1992; Vidyasagaret al., 1996; Sompolinsky & Shapley, 1997).

Recent studies from our laboratory have provided evidence thatcortical orientation tuning and direction selectivity are sharpened,respectively, via inhibition between cells with radically differentorientation preferences and opposite direction preferences. Thesestudies involved reversibly inactivating functionally characterizedsites in areas 17 or 18 by iontophoresis of GABA while monitoringthe orientation tuning and direction selectivity of single cells recordedat laterally remote, visuotopically corresponding sites in the same

Lateral inhibition and orientation/direction selectivity 2057

area (Crook & Eysel, 1992; Crooket al., 1996, 1997). Remoteinactivation often caused increases in response to non-optimal stimuli,commensurate with a loss of inhibition. Increases in response to non-optimal orientations were elicited almost exclusively from cross-orientation sites and increases in response to non-preferred directionsmainly from iso-orientation sites whose direction preference wasopposite that of a recorded cell. Although these effects may havebeen mediated via a number of different pathways, one obviouspossibility is that they involved the loss of a direct inhibitory inputfrom inactivation sites to recording sites. Consistent with this, lateralinhibitory connections in area 17 (Martinet al., 1983; Somogyiet al.,1983; Albus et al., 1991; Albus & Wahle, 1994) and area 18(Matsubara & Boyd, 1992; Thejomayen & Matsubara, 1993) havebeen shown to link both iso-orientation and cross-orientation sites(Kisvarday & Eysel, 1993; Kisva´rday et al., 1994).

The aim of the present study was to test for the presence of adirect inhibitory pathway from an inactivation site to recording siteswhere the above types of disinhibitory effect had been elicited. Tothis end, we made microinjections close to recording sites of [3H]-nipecotic acid, a highly potent and selective competitor with GABAfor high-affinity GABA uptake (Krogsgaard-Larsen & Johnston, 1975;Johnstonet al., 1976a,b; Larssonet al., 1980; Kovalev & Raevskii,1981). Experiments in monkey visual cortex have shown that injec-tions of [3H]-nipecotic acid can produce selective retrograde labellingof GABAergic neurons (Kritzeret al., 1992), and we verified thatthis was the case with our injections. The present approach thusallowed us to visualize the distribution of GABAergic inhibitoryneurons with projections which terminated in the vicinity of arecording site and to test for their presence at the inactivation site.Some of the results have been presented previously in abstract form(Crook et al., 1992a,b).

Materials and methods

Inactivation experiments

The effects of remote inactivation reported here were elicited duringthe course of experiments described in Crooket al. (1996, 1997).Details of physiological preparation, recording and inactivation pro-cedures, data acquisition and analysis may be found therein. Briefly,experiments were performed on lightly anaesthetized (artificial ventila-tion with 70 : 30% N2O/O2 plus 0.4–0.6% halothane), paralysed(arterial infusion of 0.06 mg/kg per h alcuronium chloride) adultcats (2.5–5.0 kg), which had been prepared acutely using standardprocedures. The electroencephalogram, electrocardiogram, pulse rate,arterial blood pressure (95–140 mmHg), end-tidal CO2 (3.5–4.2%)and body temperature (near 38.5 °C) were monitored continuously asindicators of anaesthetic efficacy. Single-unit recordings were madewith micropipettes filled with 5% biocytin in 0.5M sodium acetate(tip diameter 1–3µm; impedance 1–5 MOhm). An independentlydriven double-barrel pipette (tip diameter 10–20µm) was used forlocal inactivation, with one barrel containing GABA (0.5M, pH 3.0;retaining current – 15 nA) and the other biocytin (5% in 0.5M sodiumacetate) to allow recording of multiunit activity. To minimize corticalpulsations, a 5-mm-diameter chamber whose base was covered withtransparent elastic foil was lowered under micromanipulator controlinto a craniotomy centred on appropriate Horsley-Clarke coordinates(see below) until the foil made contact with the exposed pial surface.Penetrations for recording and inactivation, spacedµ 400–700µmapart, were made through the transparent foil approximately normalto the cortical surface in either area 17 (L1.5–2.0; P3–6) or area 18(A0–2; L2.5–3.0). A multiunit cluster was isolated with the inactiva-

© 1998 European Neuroscience Association,European Journal of Neuroscience, 10, 2056–2075

tion pipette, its orientation/direction preference determined quantitat-ively, and it was verified that iontophoresis of GABA reversiblyabolished driven and spontaneous activity. Thereafter, single cellswere isolated with the recording pipette, their receptive fields mappedand classified (Orban, 1984), and orientation tuning curves for thedominant eye were derived with a moving light bar of optimal length,width and velocity of motion, but suboptimal contrast (0.3–0.5),before, during and after iontophoresis of GABA (100 nA) at theinactivation site. A record was made of the depth from the corticalsurface at which single units were isolated, and the inactivation siteand recording sites where cells showed increases in response to non-optimal orientations or directions were marked by iontophoresis ofbiocytin (positive 0.7–0.9-µA current, 1 Hz, 0.5-s on/off duty cycle,5–10 min). We typically made multiple penetrations with a recordingpipette while the inactivation pipette remainedin situ. The locationsof penetrations for recording and inactivation were documented onan enlarged photograph of the exposed cortical surface, using thebranching pattern of blood vessels as landmarks.

Injections of [3H]-nipecotic acid

At the end of an inactivation experiment, the recording and inactivationpipettes were withdrawn and a glass micropipette (tip diameter10–15µm) containing [3H]-nipecotic acid (Amersham, Bucks, UK;specific activity 25.8 Ci/mmol) was positioned directly above theentry point of a penetration in which cells showing disinhibitoryeffects on orientation tuning or directionality had been recorded. Thestereotaxic coordinates of this penetration and its recorded locationrelative to the superficial vasculature were used to guide placementof the pipette. Using a manually controlled microdrive, the pipettewas lowered in a direction normal to the cortical surface through thetransparent foil covering the exposed pia and into the cortex until thetip was located at a cortical depth corresponding to the centre of agroup of closely spaced recording sites where cells had shown thesame type of effect (on orientation or direction). An injection of[3H]-nipecotic acid (1µCi/µL; 38.8 mM) in artificial cerebrospinalfluid was then made by applying brief pressure pulses to the end ofthe pipette, with each injection being made over a 10–15-min period.Injection volumes ranged from 45 to 320 nL, as assessed by monitoringthe meniscus level of the pipette before and after the injection via astereo operating microscope with an eyepiece graticule. Followingeach injection, the pipette was kept in place for at least 10 min beforebeing withdrawn.

Perfusion and tissue processing

After a postinjection survival time of 60–120 min, cats were givenan overdose of anaesthetic and perfused transcardially with oxygenatedTyrode’s solution followed by a fixative containing 2–4% paraformal-dehyde and 0.5–1% glutaraldehyde in 0.1M phosphate buffer (pH 7.4).Large blocks of the area of interest were dissected and 50–80-µm-thick sections comprising the entire cortical depth were cut on afreezing microtome in a plane parallel to the cortical surface. Sectionswere rinsed in phosphate buffer and reacted for biocytin using theavidin–biotin complexed horseradish peroxidase method (ABC, VectorLaboratories, Burlingame, CA, USA) described in Kisva´rday & Eysel(1993) and Kisva´rday et al. (1994). The most superficial section andevery other third section was osmium-treated and embedded inDurcupan ACM resin (Fluka; Neu-Ulm, Germany) on slides. Autora-diographic detection of [3H]-nipecotic acid labelling was performedon non-osmicated sections mounted on gelatine-coated slides. Theslides were dipped in Ilford K5 nuclear emulsion and stored at 4 °C.After exposure for 10–15 days, they were developed in Kodak D-

2058 J. M. Crooket al.



19B developer, diluted 1 : 1 with distilled water and air-dried. Toverify the selective uptake of [3H]-nipecotic by GABAergic neurons,adjacent semithin (0.5µm) horizontal sections were cut from resin-embedded sections containing injection sites and mounted on separategelatinized slides. Alternate sections were then processed for autora-diography (as above; exposed for 4–8 weeks), or reacted with anti-GABA antiserum (Somogyiet al., 1985) and visualized with aperoxidase conjugated secondary antibody and 3,39-diaminobenzidine.

Histological analyses

Histological reconstructions were made using the software Neuroluc-ida (MicroBrightField Inc., Colchester, VT, USA) on an IBM-compatible computer which was connected to a light microscope(Leitz Diaplan) equipped with a motorized stage. Sections wereexamined at3 100 magnification and charts were drawn of bloodvessels, penetration marks made by the injection, inactivation andrecording pipettes, cell bodies labelled autoradiographically and byiontophoresis of biocytin, and [3H]-nipecotic acid injection sites.These charts were aligned using the profiles of cross-sectionallyrunning small blood vessels and biocytin-labelled fibres with cutends. For alignment of adjacent osmium-treated and autoradiographicsections, the magnification of the computer reconstructed images wasadjusted to correct for the 7–9% greater shrinkage of the osmicatedsections. This value corresponds very closely to the shrinkage factordetermined for biocytin-stained, osmium-treated sections of areas 17and 18 in previous experiments (Kisva´rday & Eysel, 1993; Kisva´rdayet al., 1994) in which the perfusion protocol was the same as the oneused here. We therefore consider shrinkage of autoradiographicsections to have been minimal. Clumps of silver grains whose densitywas clearly higher than that of the surrounding neuropil within theviewed region of the microscope were considered to representradiolabelled cell bodies; neighbouring unlabelled cells could oftenbe identified as clear spaces devoid of silver grains. To avoidmisinterpretation of the results due to the inclusion of neuroglia,clumps of grains whose diameter was, 7 µm were excluded fromthe sample of radiolabelled cells. [3H]-nipecotic acid injection siteswere identified by the dense accumulation of silver grains in autoradio-graphs. In addition, tissue damage caused by the injection procedurewhich was visible in both autoradiographic and osmicated sectionswas used to assess the position of the injection-pipette tip. Slighttissue damage caused by each pipette was readily visible in osmicatedsections and could sometimes be detected in autoradiographs. Underthe light microscope, these penetration marks appeared as dark,slightly oedematous spots, often containing a few red blood cellsand/or darkly stained macrophages. They could be traced up to themost superficial section representing the entry point of the pipette.Thus, each penetration could be identified by comparing its topograph-ical location relative to the superficial vasculature in histologicalsections with its recorded location on the photograph of the corticalsurface taken during the inactivation experiment. The inactivationsite and the recording sites within each penetration were identifiedin autoradiographic and osmicated sections by discrete biocytinlabelling of neuronal somata (diameter of label 30–60µm).

FIG. 1. Selective uptake of [3H]-nipecotic acid. (A–D) Photomicrographs of semithin (0.5µm) horizontal sections from layer IV of area 18 showing GABAergicneurons selectively labelled following an injection of [3H]-nipecotic acid (250 nL). (A,C) Show different parts of the same section processed for autoradiography.(B,D) Show the corresponding parts of an adjacent section which was processed for GABA immunohistochemistry. Injection site marked by asterisk in (C) and(D). Neurons in (A,B) locatedµ 350–500µm from the injection site. Scale bar in (A) is 50µm and applies to all photomicrographs. Note that all radiolabelledneurons (long filled arrows) correspond to cells which are immunoreactive for GABA; also that these cells are interspersed among both GABA-positive andGABA-negative neurons (filled and open short arrows), which were not labelled autoradiographically.


To test for the presence of radiolabelled cells in the vicinity of aninactivation site, the following procedure was adopted. We firstselected for analysis those autoradiographic sections which containedradiolabelled cells within a vertical distance of 150µm from theinactivation site (defined as the centre of biocytin-labelled somata),taking into account section thickness and the fact that autoradiographiclabelling was confined to the upper surface of each section. Chartsof these sections were then aligned and the topographical location ofthe inactivation site was marked on each chart. We then constructeda graph in which for a given cell the vertical distance from theinactivation site to the top of the section in which it was located wasplotted against its lateral distance from the inactivation site. Theradial distance between the cell and the inactivation site was thendefined by the length of a straight line connecting its datum pointwith the origin of the graph. For cells in each section, data pointswere plotted in order of increasing lateral distance from the inactivationsite until a cell was encountered at a radial distance of. 150µm.We then counted the total number of cells located within 100µm and150µm of the inactivation site. A similar graph was constructed todetermine the distance between biocytin-labelled recording sites andthe centre of a [3H]-nipecotic acid injection site.

In the case where iontophoresis of biocytin at an inactivationsite labelled large basket cells with projections to cross-orientationrecording sites, their somata, axonal and dendritic fields were recon-structed from consecutive (autoradiographic and osmicated) sectionsusing a 503 oil objective.

Cortical layering followed the schemes of Lundet al. (1979) forarea 17 and Humphreyet al. (1985) for area 18. Major laminarboundaries were determined by comparing the serial order of sectionswith cortical thickness, and on the basis of cytoarchitectural featuresin osmicated sections such as neuronal density, soma size and layer-specific cell types, particularly the large pyramidal cells at thelayer III/IV border region and in layer V. Additionally, in bothautoradiographic and osmicated sections, the types of neuron labelledby iontophoresis of biocytin at the inactivation site and each recordingsite was taken into account in establishing the identity of corticallayers.

Results

Selectivity of [3H]-nipecotic acid for GABAergic neurons

Nipecotic acid is one of the most potent inhibitors of GABA transportprocesses in the brain (Krogsgaard-Larsen & Johnston, 1975; Larssonet al., 1980; Kovalev & Raevskii, 1981) and it has been shown tocompete selectively with GABA for uptake into slices of rat cerebralcortex with an apparent affinity greater than that of GABA itself(Johnstonet al., 1976a,b). Injections of [3H]-nipecotic acid producedselective labelling of GABA-immunopositive neurons in monkeyvisual cortex (Kritzeret al., 1992), and in cat area 17, GABA-immunopositive neurons were selectively labelled following injectionsof [3H]-GABA (Kisvarday et al., 1987). We therefore expectedthat our injections of [3H]-nipecotic acid would selectively labelGABAergic neurons. To verify this, we compared adjacent, semithin


sections from area 18 processed for autoradiography and GABAimmunohistochemistry (see Fig. 1).

As expected, GABA immunohistochemistry revealed the presenceof strongly immunoreactive cell bodies and terminals, the latteroften surrounding GABA-immunopositive or immunonegative somata(Fig. 1B,D). Following an injection of [3H]-nipecotic acid (250 nL)in two cats, all radiolabelled cells we examined (n 5 75) correspondedto neurons which were strongly GABA immunoreactive (in Fig. 1compare A with B and C with D); even close to the injection site,no GABA-immunonegative cell was radiolabelled. However, withinthe field of radiolabelled cells, a substantial proportion of GABA-immunopositive neurons, including those located close to the injectionsite, were not labelled autoradiographically.

General aspects of [3H]-nipecotic acid labelling

The [3H]-nipecotic acid injection sites consisted of a core region ofdamaged tissue modestly covered with silver grains and a surroundingdense ‘halo’ of silver grains within which few radiolabelled cellscould be identified (injection-site diameterµ 170–250µm; seeFig. 2A,B,D). The paucity of identified radiolabelled cells around theinjection centre may have been caused by the injection procedurepreventing somal accumulation and/or retention of [3H]-nipecoticacid (see Chronwall & Wolff, 1980), although all but the most heavilylabelled cells would have been obscured by the high level of neuropillabelling. Labelling over the neuropil represented [3H]-nipecotic acidarrested within cell processes (for example, axons and dendrites) atthe time of fixation. At the injection site, most of the silver grainsoverlying the neuropil were probably located in nerve terminals. Thiszone, in which the concentration of [3H]-nipecotic acid is highest atthe centre, may well have been larger than that in which terminaluptake resulted in detectable labelling of cell somata. As [3H]-nipecotic acid competes with GABA for high-affinity uptake, thesomatic content of silver grains reflects at any one time the sum ofGABA and [3H]-nipecotic acid arriving from all terminals. Thus, fora soma to be detected as labelled, a high proportion of its terminalsshould be located in regions of high concentration of [3H]-nipecoticacid.

There was a large variation in the density of grains observed overindividual cells considered to be radiolabelled (Fig. 1A,C, 2A–C),which was independent of the distance from the injection site or thetime between injection and fixation. The degree of labelling wasconsidered to reflect the proportion of terminals maintained by eachcell in proximity to injection sites. Particularly in regions where thelevel of neuropil labelling was high, including injection sites, unla-

FIG. 2. Presence of radiolabelled cells in the vicinity of an inactivation site following an injection of [3H]-nipecotic acid (320 nL) close to a recording site.(A,B) Photomicrographs of adjacent, horizontal sections (50µm) from layer III of area 18, processed for autoradiography. Scale bars 100µm. (A) The injectioncentre (asterisk) and the inactivation site (star) were locatedµ 550µm apart in the same section. (B) shows the dorsally adjacent section in which the recordingsite (white asterisk) was located; the black asterisk and star mark the topographical location of the injection centre and the inactivation site. Lateral distancebetween the recording site and injection centreµ 100µm. c labels the same capillaries in (A) and (B). The recording site was identified by biocytin labellingof neuronal somata (diameter of labelµ 60 µm). Outside this zone of biocytin labelling, cells which appear black are those which were heavily labelled by[3H]-nipecotic acid. Some of the more moderately radiolabelled cells, one of which is marked by an arrowhead in (A), can be identified as clumps of grainswhich stand out against the surrounding neuropil labelling. Clear spaces are ‘unlabelled’ neurons whose grain density was well below that of the surroundingneuropil. The [3H]-nipecotic acid injection site consisted of a core region of tissue damage caused by the injection procedure and a surrounding dense ‘halo’ ofsilver grains within which few radiolabelled cells were visible (injection-site diameterµ 250µm). Note in (A) and (B), two darkly radiolabelled cells (arrowed)at a lateral distance ofµ 75–80µm from the inactivation site; and in (A) a more moderately radiolabelled cell (marked by arrowhead) at a lateral distance ofµ 120µm. (C) and (D) show, respectively, higher magnification photomicrographs of parts of the sections in (A) and (B). In (C), long filled arrows andarrowheads point, respectively, to neuronal somata and fibres labelled by iontophoresis of biocytin at the inactivation site. The inset is from a photomicrographtaken at a higher focal plane, showing two lightly radiolabelled neurons (filled arrows), together with neighbouring unlabelled neurons (open arrows), in closeproximity to biocytin-labelled somata at the inactivation site. c marks the capillary adjacent to the star in (A). Scale bar 20µm. (D) Shows neuronal somata(arrowed) labelled by iontophoresis of biocytin at the recording site (white asterisk) close to the core of the [3H]-nipecotic acid injection site (black asterisk);arrowheads point to biocytin-labelled fibres. c, capillary. Scale bar5 50 µm.


belled somata could be identified as clear spaces which were devoidof silver grains or where the grain density was substantially lowerthan that of the surrounding neuropil (Fig. 2A,B; Figs 4–6). Theregion adjacent to an injection site contained a high density ofradiolabelled cells, most of which were heavily invested with silvergrains, and an associated high level of neuropil labelling (Fig. 2A,B).Somal and neuropil labelling was also dense radially above and belowan injection site. In the horizontal plane, the density of radiolabelledsomata decreased more or less smoothly with increasing distancefrom the injection site, with the distribution of radiolabelled cellsalways extending laterally for at least 1 mm from the injection centre(Figs 4–6). Weakly radiolabelled cells predominated at locationslaterally remote from the injection site. In the vertical plane, radiolabel-led cells showed a clearly anisotropic distribution (see insets toFigs 5G and 6J). Together with the fact that, even close to aninjection site, many GABA-immunopositive neurons were not labelledautoradiographically (Fig. 1), this essentially rules out simple diffusionand somatic uptake of [3H]-nipecotic acid. The observed pattern andextent of labelling is consistent with axonal uptake and retrogradetransport of the tracer.

Distribution of neurons labelled by injections of [3H]-nipecoticacid and evidence for inhibitory projections from inactivationsites to recording sites

Area 18

Six injections of [3H]-nipecotic acid were made in area 18, five inlayer III and one in layer IV. One of these was placed remote(300µm) from the nearest biocytin-labelled recording site. The otherfive injections (four in layer III and one in layer IV) were placedwithin µ 100µm of one or two recording sites where disinhibitoryeffects on orientation tuning or directionality had been elicited frominactivation sites in layers III–IV (horizontal distance from injectioncentreµ 400–550µm). Three of these injections targeted sites wherecells had shown an increase in response to the non-preferred direction(but negligible change in orientation tuning) during inactivation ofan iso-orientation site (orientation preference within 22.5°) withopposite direction preference and strong direction bias; the other twotargeted sites where cells had shown increases in response to non-optimal orientations (accompanied in one case by a disinhibitoryeffect on directionality) during inactivation of a cross-orientation site(difference in orientation preference 45–90°). In all five cases,radiolabelled cells were found in the vicinity of the inactivation site.The receptive fields of the recorded cells lay within 6–10° of the areacentralis projection.


FIG. 3. Disinhibitory effect on directionality elicited in a C-cell recorded in layer III (area 18) by iso-orientation inactivation in the same layer. Inactivation andrecording sites (IS and RS) correspond to those of Fig. 2; their topographical location is shown in the inset (top left); A, anterior; L, lateral. Below: orientationtuning of multiunit activity recorded at IS for a moving bar. As in all other cases, the polar diagram was derived from PSTHs in response to five forward andreverse sweeps at each of eight orientations (varied pseudorandomly); PSTHs were smoothed by combining an appropriate number of bins, the peak responsewas plotted as vector length and direction of motion as vector angle. (A–C) Reversible inactivation of IS by GABA iontophoresis. (A) Multiunit response toback-and-forth motion of an optimally oriented bar. (B) Abolition of visually driven and spontaneous activity within 3 min of onset of iontophoresis of GABA(100 nA). (C) Recovery within 3 min of termination of GABA application. (D–F) PSTHs for recorded cell in response to opposite directions of motion of anoptimally oriented bar in the control situation (D), within 5 min of onset of iontophoresis of GABA at IS (E), and within 4 min of termination of GABAapplication. Bar orientation and direction of motion indicated above, stimulus waveform below each PSTH. Direction preference at IS was opposite that ofrecorded cell which showed an increase in response to the non-preferred direction during remote inactivation. PSTHs in (A–C) derived from five consecutivestimulus cycles; those in (D–F) taken from a complete set of PSTHs used for derivation of polar diagrams. Amplitude of motion 20°, cycle duration 1.25 s,stimulus velocity 53°/s throughout. Bin-widths: (A–C) 25 ms (five bins combined); (D–F) 15 ms (three bins combined). Schematic (bottom left) shows positionand size of hand-plotted minimum response field of recorded cell (white rectangle) relative to that for multiunit activity at IS (shaded rectangle); adjoining linesindicate orientation preference. All data derived for contralateral eye.

Injections at iso-orientation recording sites.Figures 2 and 3 illustrateresults from an experiment in which an injection of [3H]-nipecoticacid was made in area 18 close to an iso-orientation recording sitewhere a disinhibitory effect on directionality had been elicited. InFig. 2A and B show photomicrographs of two adjacent, autoradio-graphic sections (50µm) from layer III containing the [3H]-nipecoticacid injection site, the recording and inactivation sites. (C) and (D)show, respectively, higher magnification photomicrographs of partsof the sections in (A) and (B). The injection centre and the inactivationsite (asterisk and star in A) were locatedµ 550µm apart in the samesection. The recording site (white asterisk in B and D), which wasidentified by biocytin labelling of neuronal somata (arrowed in D),was located in the dorsally adjacent section, at a lateral distance ofµ 100µm from the injection centre (black asterisk). The injection of[3H]-nipecotic acid (diameterµ 250µm) produced dense retrogradelabelling of two cells (arrowed in A and B) located within 100µmof the inactivation site and more moderate labelling of another cell,just visible at low power (see arrowhead in A), located within 150µmof the inactivation site. Additionally, as shown in (C), two weaklyradiolabelled cells (short thick arrows) were detected in close proxim-ity to biocytin-labelled neuronal somata (long thin arrows) at theinactivation site. Although both cells showed low grain densities,they could be confirmed as radiolabelled by comparison with twoneighbouring ‘unlabelled’ cells (indicated by open arrows) which


were essentially devoid of silver grains. The effect of remoteinactivation corresponding to Fig. 2 is documented in Fig. 3. Therecorded cell (a C-cell) and the inactivation site showed the sameorientation preference but strong bias for opposite directions ofmotion (compare polar diagram and A with D). Remote inactivation(B) caused a marked reduction in the cell’s direction bias (E), withthe directionality index [(DI): 1-(response to non-preferred/preferreddirection), with mean spontaneous activity subtracted] decreasingfrom 0.76 to 0.16. The effect was due primarily to an increase inresponse to the non-preferred direction. Its time course was compar-able with that for the abolition of multiunit activity at the inactivationsite. Following termination of GABA application, the directionalityof the recorded cell reverted to its original state with a time coursethat closely paralleled that for the recovery of the multiunit responseat the inactivation site (compare F with C).

Comparable results were obtained from another experiment in area18 (data not shown), in which the [3H]-nipecotic acid injection wasalso placed in layer III, close to a recording site where a C-cell hadshown a decrease in DI from 0.91 to 0.53 during iso-orientationinactivation in the same layer. Two very heavily radiolabelled cellswere found within 100µm of the inactivation site and a third, moremoderately labelled cell within 150µm.

Physiological/anatomical data from the other experiment in area18 in which an injection of [3H]-nipecotic acid was placed close to


iso-orientation recording sites are shown in Fig. 4. In this case, theinjection centre was located, 100µm from two sites in lower layerIV (RS1 and RS2) at each of which an S-cell had shown a markedreduction in direction bias [decrease in DI from 0.71 to 0.15 (D,E)and from 0.74 to 0.16 (G,H)] during remote inactivation in the samelayer. As shown in Fig. 4J, the injection (320 nL; diameterµ 200µm),which was placed in lower layer IV but also involved the top of layerV (see inset), gave rise to a field of radiolabelled cells (dots) which,in the tangential plane, was roughly circularly symmetrical about theinjection centre (asterisk). The density of labelling decreased withincreasing lateral distance from the injection site. Radiolabelled cellsshowed a relatively even distribution in the tangential plane and didnot aggregate in clusters. All layers contained radiolabelled cells. Ineach layer, they were concentrated within a lateral distance ofµ 0.3 mm from the injection centre, although in layers II–VI labellingextended laterally for 0.7–1 mm (see shading in inset). Figure 4(K)shows a composite chart of labelling in two autoradiographic sectionsfrom layer IV (see inset) which contained all radiolabelled cells (C1–C6) located within 150µm of the inactivation site (star). Of these sixcells, three (C1–C3; large filled circles) were located, 100µm andthree (C4–C6; small filled circles). 100µm from the inactivationsite. As can be seen in (L), five of these cells (C2–C6) were heavilyinvested with silver grains. C1 was lightly labelled, but its graindensity clearly exceeded that of the surrounding neuropil and wassubstantially higher than that of three neighbouring unlabelled cells(indicated by open arrows).

Injections at cross-orientation recording sites. Both of these injectionswere made in layer III. In one case, which is illustrated in Fig. 5, theinjection (diameterµ 180µm) was placed close to a recording sitewhere a C-cell had shown a marked increase in response to non-optimal orientations, accompanied by a slight increase in response tothe non-preferred direction, during inactivation of a site with ortho-gonal orientation preference (compare top two rows). The effectsresulted in a substantial increase in orientation tuning half-width(measured at half the maximum response from smoothed tuningcurves; see Crooket al., 1991) from 30° to 90° (200%) and a loss ofdirection bias (decrease in DI from 0.30 to 0). The injection producedretrograde labelling of four cells (C1–C4) located within 150µm ofthe inactivation site at the layer III/IV border region (H). Thesecomprised one very heavily radiolabelled cell (C1) and a moderatelylabelled cell (C2) within 100µm of the inactivation site and twodensely labelled cells (C3 and C4) within 150µm (I). The otherinjection was placed, 100µm from two recording sites where anS-cell and a C-cell had shown, respectively, a 52% and a 38%increase in orientation tuning width (with negligible change indirectionality) during remote inactivation in layer III (data not shown).Two densely radiolabelled cells were found within 100µm of theinactivation site and a lightly labelled cell within 150µm.

As illustrated in Fig. 5G, the pattern of labelling produced byinjections in layer III (n 5 5; injection volume 45–320 nL) differedfrom that following the injection in lower layer IV (Fig. 4), both interms of the tangential distribution of radiolabelled cells and thelateral spread of labelling in different layers. Although the overalllateral extent of labelling was about the same in each case, followinglayer-III injections, the density of radiolabelled cells decreased moresmoothly with increasing lateral distance from the injection site.Additionally, in contrast to the uniformly widespread labellingthroughout layers II–VI produced by the injection in lower layer IV,marked laminar differences in the lateral extent of labelling wereseen following layer-III injections. Labelling was widespread in thesuperficial layers, with radiolabelled cells in layer II and layer III


extending laterally for 0.8–0.9 mm and 1.1–1.2 mm from the injectioncentre. However, in layers IV–VI, labelling was more circumscribedand showed a smaller lateral spread than that produced by theinjection in lower layer IV. In layer IV, radiolabelled cells weredistributed within a lateral distance ofµ 0.6 mm from the injectioncentre, and in layer V they were confined mainly within a lateraldistance of 0.3 mm, although a few outlying cells at the top of thelayer strayed as far as 0.7 mm. Radiolabelled cells in layer VI werelocated near the border with layer V and were confined withinµ 0.3 mm of the central axis of the injection.

Area 17

Five injections of [3H]-nipecotic acid were made in area 17 (injectionvolumes 45–60 nL). Three of these were placed remote (. 300µm)from the nearest biocytin-labelled recording site. Each of the othertwo injections was placed, 100µm from two recording sites wherecells had shown an increase in response to the non-preferred direction(but negligible change in orientation tuning) during inactivation ofan iso-orientation site with opposite direction preference and strongdirection bias. In both cases, radiolabelled cells were found in thevicinity of the inactivation site. The recorded cells had receptivefields within 5–8° of the area centralis projection.

Injections in layer IV.Two injections of [3H]-nipecotic acid weremade in layer IV, both in the upper half of the layer. In one case, theinjection was placed remote from biocytin-labelled recording sites.In the other case, which is illustrated in Fig. 6, the injection (diameterµ 170µm) was placed close to two sites in upper layer IV where anS-cell (recorded at RS1) and a C-cell (recorded at RS2) had shown amarked reduction in direction bias during remote inactivation in thesame layer (decrease in DI from 0.94 to 0.45 in D and E and from0.84 to 0.31 in G and H). It produced moderate retrograde labellingof two cells located within 100µm and two others within 150µm ofthe inactivation site (K,L). A surface view of the distribution of cellsradiolabelled following this injection is shown in Fig. 6J. This is nota true reflection of the distribution of radiolabelled cells parallel tothe cortical surface, because injections in area 17 were made on thecrown of the lateral gyrus which is highly curved and the cortex wasnot flattened. For the same reasons, the shading in the inset givesonly an approximate indication of the lateral extent of labelling indifferent layers; radiolabelled cells laterally remote from the injectionsite may have been located in more superficial layers than thoseindicated. Nevertheless, it is clear that labelling was widespread inthe superficial layers, with the distribution of radiolabelled cellsextending laterally up to 1.3 mm from the central axis of the injection.It is also apparent that labelling in layer IV extended laterally forseveral hundred microns from the injection centre. No radiolabelledcells were found in layer VI. Also, superficial sections from layer Iand the upper portion of layer II did not contain radiolabelled cells.This was presumably due to oedema and/or tissue damage, whichwas evident in these sections, preventing perikaryal accumulation ofthe tracer (cf. Chronwall & Wolff, 1980), because radiolabelled cellswere present throughout layers I–II following the other injection inupper layer IV of area 17 which, in all other respects, produced avery similar pattern of labelling to that illustrated.

Injections in layers II/III. Three injections were made in layers II/III.Two of these were confined to layer III and placed remote frombiocytin-labelled recording sites. The other injection spanned thelayer II/III border and was placed close to two recording sites at eachof which a C-cell had shown a marked reduction in direction bias(decrease in DI from 0.94 to 0.45 and from 0.95 to 0.55) during


inactivation of a site in upper layer IV at a horizontal distance ofµ 400µm (data not shown). It produced retrograde labelling of twocells located within 100µm of the inactivation site and a third celllocated within 150µm, all of which were heavily invested with silvergrains. Injections in layers II/III produced a similar pattern of labellingto those made in upper layer IV except that, notably, all injections inlayers II/III gave rise to fields of radiolabelled cells in layer VI,distributed within a lateral distance of 0.3 mm from the injectioncentre.

Inhibitory projections from an inactivation site to cross-orientation recording sites demonstrated by iontophoresis ofbiocytin

In one of the experiments in area 17 in which the injection of [3H]-nipecotic acid was placed remote from targeted recording sites,iontophoresis of biocytin at the inactivation site in layer III fortuitouslylabelled the somata, dendritic and axonal fields of two large basketcells. We are not certain why this injection produced such extensivelabelling, but presume that it was due to the greater amount of tracerdelivered. Although the injection parameters were the same as forother biocytin injections, the injection site was much larger (diameterµ 200µm).

Figure 7A shows a reconstruction of the soma and dendrites(shaded) of each labelled basket cell (BC1 and BC2) and the pooleddistribution of axonal boutons (dots) of both cells, as viewed fromthe cortical surface. BC1 and BC2 were located, respectively, at alateral distance ofµ 40 µm and µ 100µm from the centre ofthe inactivation/injection site (IS; star). Light microscopic featurescharacteristic of both cells are shown for BC1 in (B) and (C). Bothcells showed morphological properties very similar to those ofsuperficial-layer, large basket cells of area 17 described previously(Martin et al., 1983; Somogyiet al., 1983). They had large somata(diameterµ 25–30µm) bearing smooth, beaded dendrites (B) whichwere emitted in many directions from the parent soma. Their mainaxons were heavily myelinated and followed a relatively straightcourse tangential to the cortical surface;en route, they emitted radiallyoriented branches which, in turn, gave off short segments laden withbulbous boutons which often surrounded the perikarya of otherneurons (C). Their axonal fields could be traced only in layers II/III.Both cells had dense axonal arborizations within their dendritic trees;in (A), the paucity of plotted boutons in the vicinity of the inactivationsite reflects the fact that most of the fine axonal collaterals enteringthe core region of biocytin labelling could not be traced. Additionally,the axons of both cells extended horizontally in many directions up

FIG. 4. Disinhibitory effects on directionality recorded in layer IV (area 18) during iso-orientation inactivation in the same layer, and evidence for inhibitoryprojections from the inactivation site to the recording sites. Top row as in Fig. 3. (D–I) PSTHs in response to opposite directions of motion of an optimallyoriented bar for two S-cells (D–F and G–I) recorded at different sites (RS1 and RS2) in lower layer IV during the same penetration, in the control situation(D,G), and 2–4 min after onset (E,H) and offset (F,I) of iontophoresis of GABA at the inactivation site. Direction preference at inactivation site was oppositethat of both recorded cells, and each showed an increase in response to the non-preferred direction during remote inactivation. PSTH bin-widths: (A–C) 90 ms(nine bins combined); (D–F) 30 ms (three bins combined); (G–I) 10 ms. Amplitude of motion 20°, cycle duration 2.5 s, stimulus velocity 27°/s throughout.Spatial relationship between minimum response field of each recorded cell and that for multiunit activity at inactivation site shown on left of appropriate rowof PSTHs. All data derived for ipsilateral eye. (J,K) Horizontal distribution of radiolabelled cells (dots) following an injection of [3H]-nipecotic acid (320 nL), 100µm from RS1 and RS2; asterisk and concentric circle indicate, respectively, the topographical location of the [3H]-nipecotic acid injection and itsapproximate lateral extent. (J) Composite plot derived from aligned charts of all autoradiographic sections. (K) Superimposition of charts of two autoradiographscontaining all radiolabelled cells (C1–C6) located within 150µm of the inactivation site (star); horizontal distance from injection centreµ 460µm. 0.5-mmscale bar and section orientation apply to both (J) and (K); A, anterior; L, lateral. In (K), large and small filled circles highlight radiolabelled cells locatedwithin 100µm (C1–C3) and 150µm (C4–C6) of the inactivation site. In inset to (J), asterisk indicates the laminar location of the [3H]-nipecotic acid injection(diameterµ 200µm), and shading the lateral extent of labelling in different layers, measured from the injection centre; layers II/III were divided into upperand lower halves. Shading in inset to (K) indicates the laminar location of all radiolabelled cells shown. (L) Cells C1–C6 (indicated by white arrows) are shownon photomicrographs of autoradiographic sections. Open arrows point to unlabelled cells. Scale bar5 20 µm. Data in (J–L) derived from 80µm-thickhorizontal sections.


to µ 750µm from the parent soma and innervated only certain sectorsof cortex within their reach. The axonal fields of each cell overlappedto a large extent so that this characteristic axonal projection patternis also evident in the pooled boutonal distribution of both cells.

Boutons of BC1 and BC2 (ringed in Fig. 7A) were present closeto the topographical location of two cross-orientation recording sitessampled in the same penetration in layers II/III (RS1 and RS2) wheredisinhibitory effects on orientation tuning had been elicited fromthe inactivation site. These effects are documented in Fig. 8. Theorientation preference at the inactivation site (top row) differed by67.5° from that of a C-cell recorded in layer II at RS1 (D–F) andwas orthogonal to that of an S-cell recorded in layer III at RS2 (G–I). During remote inactivation, both cells showed an increase inresponse to non-optimal orientations, which resulted in substantialbroadening of orientation tuning [increase in tuning half-width from24° to 58° (142%) in D and E and from 18° to 59° (227%) in G andH]. As shown in (J–L), boutons of BC1 and BC2 were located in closeproximity to RS1 and RS2, respectively. These boutons correspond tothose ringed in Fig. 7A. As the terminals of large basket cells containGABA (Somogyi & Soltesz, 1986; Kisva´rday et al., 1987), this isstrong evidence for a cross-orientation inhibitory projection from theinactivation site to each recording site. Furthermore, as the majorityof synapses made by large basket cells are on the somata and proximaldendrites of their target cells (Somogyiet al., 1983; Kisvardayet al.,1987), BC1 and BC2 must have made synaptic contact with cellslocated close to RS1 and RS2, respectively. It is important to pointout that it was difficult to trace fine axon collaterals in the region ofdense biocytin labelling at both recording sites. Thus, the boutonsplotted in Fig. 8J and L should be regarded as a minimal estimate ofthe total number of boutons supplied by each basket cell in proximityto each recording site. Note also that outside the ringed zone inFig. 7A, strings of boutons distributed at a lateral distance ofµ 70–150µm from RS2 can be seen running towards this recording site.These boutons derived from both BC1 and BC2 and they were locatedin the section containing RS2 and in the ventrally adjacent section.At least some of them probably targeted sites where the orientationpreference was similar to that at RS2. Thus, in addition to theterminals supplied by BC2 in close proximity to RS2, both labelledbasket cells probably had cross-orientation projections which termin-ated in the vicinity of RS2. We were, of course, interested to determinewhether BC1 and BC2 projected to other recording sites wheredisinhibitory effects of remote inactivation had been elicited. Regret-tably, all other such effects were elicited in cells recorded deep tolayer III.


FIG. 5. Disinhibitory effect on orientation tuning elicited in a C-cell recorded in layer III (area 18) during cross-orientation inactivation at the layer III/IV borderregion, and evidence for inhibitory projections from the inactivation site to the recording site. Top row as in Fig. 3. Bin-width in (A–C): 90 ms (nine binscombined); amplitude of motion 20°, cycle duration 2.5 s. (D–F) Polar diagrams derived for the recorded cell, in the control situation (D), and 4–6 min afteronset (E) and offset (F) of iontophoresis of GABA at the inactivation site. Radius of circle in centre of each polar diagram indicates mean spontaneous activity.Stimulus velocity 27°/s throughout. Spatial relationship between minimum response field of recorded cell and that for multiunit activity at inactivation siteshown to left of (D). All data derived for contralateral eye. Recorded cell showed a substantial increase in response to non-optimal orientations duringinactivation of a site with orthogonal orientation preference. (G–I) as in Fig. 4J–L: An injection of [3H]-nipecotic acid (120 nL; diameterµ 180µm) in layerIII, µ 80-µm from the recording site, produced retrograde labelling of four cells (C1–C4) located within 150µm of the inactivation site; horizontal distancefrom injection centreµ 400µm. Data derived from 80-µm-thick horizontal sections. Reconstruction in (H) from two autoradiographs. In (I), open arrows pointto unlabelled cells; scale bar5 20 µm.



FIG. 6. Disinhibitory effects on directionality recorded in layer IV (area 17) during iso-orientation inactivation in the same layer, and evidence for inhibitoryprojections from the inactivation site to the recording sites. Conventions and layout as in Fig. 4. Top three rows: an S-cell recorded near the top of layer IV(RS1) and a C-cell recorded during the same penetration in mid-layer IV (RS2) showed an increase in response to the non-preferred direction during inactivationof an iso-orientation site with opposite direction preference. Amplitude of motion 10°, cycle duration 5 s, stimulus velocity 6.7°/s; data derived for ipsilateraleye. PSTH bin-widths: (A–C) and (G–I) 100 ms (five bins combined); (D–F) 20 ms. (J–L) An injection of [3H]-nipecotic acid (45 nL; diameterµ 170µm) inthe upper half of layer IV,, 100µm from RS1 and RS2, produced retrograde labelling of four cells (C1–C4) located within 150µm of the inactivation site;horizontal distance from injection centreµ 560µm. Data derived from 60-µm-thick horizontal sections. Reconstruction in (K) from a single autoradiograph. In(L), open arrow points to an unlabelled cell; scale bar5 20 µm.



Discussion

In the present study, increases in response to non-optimal orientationsand/or directions were elicited in single cells recorded in layers II/III–IV of cat areas 17 and 18 by localized inactivation in the samearea of a laterally remote, visuotopically corresponding site of definedorientation/direction specificity in layers III–IV. We have providedanatomical evidence for the presence of an inhibitory pathway frominactivation sites to recording sites whose inactivation may havecontributed to the effects.

Retrograde labelling of GABAergic neurons by [3H]-nipecoticacid injections

The selectivity of nipecotic acid as a potent competitor with GABA athigh-affinity GABA uptake sites is well documented (e.g. Krogsgaard-Larsen & Johnston, 1975), and our injections of [3H]-nipecotic acidalmost certainly resulted in selective labelling of GABAergic neurons,as all radiolabelled cells that we tested immunohistochemically werestrongly GABA immunoreactive. That GABAergic neurons, includingthose located in the vicinity of inactivation sites, were labelled byretrograde transport is supported by a number of observations. Theapparent site of [3H]-nipecotic acid injections was always clearlydemarcated by a distinct region of high grain density with a diameterof 170–250µm; radiolabelled cells in the vicinity of inactivation siteswere located well beyond this region and labelling always extendedfor more than 1 mm from it. The precise relationship between theregion of high grain density at injection sites and the area of effectiveuptake is not known. However, injections in layer III of area 18varying in volume from 45 to 320 nL produced essentially the samepattern of labelling, whereas in area 17, injections made in upperlayer IV and layer III separated vertically byµ 200µm produceddifferent results with respect to the presence or absence of radiolabelledcells in layer VI. It therefore appears that the highly efficient uptakeof [3H]-nipecotic acid and the slow delivery of the tracer combinedto produce highly circumscribed effective uptake zones, confinedwithin the apparent site of injections. Diffusion and somatic uptakeof [3H]-nipecotic acid can be discounted, because (i) radiolabelledcells showed a markedly anisotropic distribution in the vertical plane,and (ii) in our control experiments, [3H]-nipecotic acid injectionsfailed to label a substantial proportion of GABA-immunopositiveneurons, even in regions adjacent to the injection site. Dendriticuptake of the tracer by radiolabelled cells in the vicinity of inactivationsites is also highly unlikely considering their lateral distance frominjection sites (µ 300–700µm) in relation to the lateral dendriticspread of putative inhibitory neurons in layers II/III–IV of cat visualcortex (Peters & Regidor, 1981; Somogyiet al., 1983; Fairenet al.,1984; Kisvardayet al., 1985; Somogyi, 1989; Tama´s et al., 1997a,b).Finally, in monkey V1, radiolabelling by [3H]-GABA distal frominjection sites was abolished by prior local injection of colchicine,an inhibitor of axoplasmic transport (DeFelipe & Jones, 1985),demonstrating that cells were retrogradely labelled. Thus, radio-labelling with [3H]-nipecotic acid, which presumably utilizes the

FIG. 7. (A) Surface view of the somata and dendritic fields (shaded) and the axonal boutons (dots) of two large basket cells (BC1 and BC2) labelled byiontophoresis of biocytin at an inactivation site in layer III of area 17. Reconstruction from six consecutive 60µm thick horizontal sections through layers II/III. The injection/inactivation site (IS; star) and BC2 were located in the same section, BC1 in the dorsally adjacent section. Asterisks mark the topographicallocation of two biocytin-labelled sites in layers II/III where the cells of Fig. 8 were recorded (RS1 and RS2). The encircled boutons correspond to those shownin Fig. 8J and L. (B,C) Light microscopic features of BC1. In (B), the large soma (asterisk) and some of the beaded dendrites (arrowed) are visible among otherbiocytin-labelled somata and processes. (C) Axonal boutons of BC1 (arrowed) in apparent contact with the soma of a target neuron (asterisk). Scale bars: (B)50 µm; (C) 10µm.


same uptake and transport mechanisms is almost certainly a productof retrograde transport.

Types of radiolabelled neuron in the vicinity of inactivationsites

Autoradiography alone provides little information on the type of cellbeing labelled. It was generally not possible to infer cell type on thebasis of apparent soma size, assessed from the diameter of clumpsof grains in autoradiograms, because (i) the soma-size distributionsof putative inhibitory neurons overlap extensively, and (ii) apparentsoma size depends on the location of the soma relative to the planeof section and represents a minimal estimate of actual soma size.However, the most likely candidates for radiolabelled cells in thevicinity of inactivation sites are basket cells, which are both GABA-ergic (Somogyi & Solte´sz, 1986; Kisva´rday et al., 1987) and havehorizontally directed axon collaterals in the appropriate range. Themajority were probably large basket cells, which have 0.5–1.5 mm-long tangential projections within and/or between layers II/III and IV(Martin et al., 1983; Somogyiet al., 1983; Naegele & Katz, 1990;Kisvarday, 1992; Thejomayen & Matsubara, 1993; Kisva´rday &Eysel, 1993; Kisva´rday et al., 1994; and see Fig. 7). However, thoselocated closest to injection sites may have been layer-IV clutch cellswhose axon collaterals can extend laterally up toµ 300µm from theparent soma (Kisva´rday et al., 1985; Naegele & Katz, 1990), orsuperficial layer, small basket cells with horizontally oriented axons(DeFelipe & Fairen, 1982; Fairenet al., 1984; Freundet al., 1986a;Meyer & Wahle, 1988). Recentin vitro studies (Tama´s et al.,1997a,b) have revealed a novel class of inhibitory neuron with similarmorphological features to basket cells (dendrite-targeting cells) whoseaxonal fields span lateral distances of up to 1500µm in the superficiallayers and 600µm in layer IV. These cells may also have been amongthose radiolabelled in the vicinity of inactivation sites. A number ofradiolabelled cells had rather large apparent soma sizes, approaching20 µm in major diameter (e.g. Fig. 4: C2 and C4; Fig. 5: C1–C3;Fig. 6: C1 and C4), suggesting that they were large basket cells, clutchcells, or dendrite-targeting cells.

Sphere of influence of GABA iontophoresis and functionalspecificity of the inhibitory pathway from inactivation torecording sites

Although we did not systematically determine the sphere of influenceof GABA iontophoresis, it is possible to make a fairly accurateestimate. Our inactivation pipettes, which readily sampled multiunitactivity, probably had a ‘seeing distance’ of about 50–100µm aroundthe tip (cf. Mountcastle, 1957; Towe, 1975), and the time taken forthe effects on a recorded cell to reach their height was similar to,although typically somewhat longer than, that required for theabolition of multiunit activity at the related inactivation site. It hasbeen shown that GABA ejected iontophoretically in cerebral cortexdiffuses through a volume of tissue significantly larger than that overwhich micropipettes can record action potentials generated by all the


affected cells (Herzet al., 1969). We are therefore confident thatapplications of GABA that were required to elicit effects on arecorded cell inactivated cells located up to 150µm from theinactivation site. However, extensive diffusion of GABA beyond thisdistance can be excluded, because we have previously found thationtophoresis of GABA with ejecting currents and durations ofapplication similar to those used in the present study did not directlyinhibit cells recorded at distances ofµ 250µm.

In areas 17 and 18, cells with similar orientation preferences areorganized in iso-orientation domains, each of which is typicallysubdivided into regions selective for opposite directions of motionextending a few hundred microns in the laminar and columnardimension (Bermanet al., 1987; Swindaleet al., 1987; Bonhoeffer& Grinvald, 1993; Bonhoefferet al., 1995; Shumuel & Grinvald,1996). Thus, given the estimated ‘seeing distance’ of our inactivationpipettes, all labelled cells located within 100µm (including biocytin-labelled large basket cells) can be considered to have had a similarorientation/direction specificity to that shown by multiunit activityrecorded at the inactivation site. This is also likely to have been thecase for labelled cells located more remotely, within 150µm, althoughit is just conceivable that some of these may have had radicallydifferent properties had the inactivation pipette been located in thevicinity of an orientation centre (Bonhoeffer & Grinvald, 1993;Bonhoefferet al., 1995) where orientation preference changes rapidly,or close to the border between iso-direction domains with oppositedirection preference (Bermanet al., 1987; Swindaleet al., 1987;Shumuel & Grinvald, 1996). The effective uptake zone for [3H]-nipecotic acid appears to have been confined within the apparent siteof injections whose diameter wasµ 170–250µm (see above), andour successful injections were placedµ 80–100µm from recordingsites where disinhibitory effects on orientation/direction selectivityhad been elicited. Thus if, as we have argued, radiolabelled neuronsin the vicinity of inactivation sites were basket cells or dendrite-targeting cells, which make synapses primarily with the proximalprocesses of their target neurons (DeFelipe & Fairen, 1982; Somogyiet al., 1983; Fairenet al., 1984; Kisvarday et al., 1985; Tama´s et al.,1997a,b), they must have made synaptic contact with recorded cellsor cells in their vicinity, most of which will have had similarorientation/direction preferences to those of recorded cells. Moredirect evidence on this point was obtained for the biocytin-labelledlarge basket cells of Figs 7 and 8. Each emitted axonal boutons inclose proximity to a recording site where a disinhibitory effect onorientation tuning had been elicited, and both had strings of boutonsdistributed withinµ 70–150µm of one of these sites.

Circuitry underlying effects on orientation/direction selectivity

Loss of inhibition in recurrently connected neurons

Before discussing how inactivation of the inhibitory pathway frominactivation to recording sites may have contributed to the observed

FIG. 8. Top three rows: Iontophoresis of GABA at the inactivation site of Fig. 7 caused reversible broadening of orientation tuning in a C-cell (D–F) recordedin layer II (RS1) and an S-cell (G–I) recorded during the same penetration in layer III (RS2). Conventions and layout as in Fig. 5. Orientation preference atinactivation site differed by 67.5° from that of C-cell and was orthogonal to that of S-cell. Stimulus velocity in top row and in (D–F) 13.3°/s; in (G–I) 3.3°/s.(Inactivation site showed the same preferred orientation and comparable orientation tuning at 3.3°/s and 13.3°/s.). (A–C) Amplitude of motion 10°, cycleduration 2.5 s; PSTH bin-width 60 ms (six bins combined). All data derived for ipsilateral eye. (J) Drawing of a terminal axon segment and boutons (dots)belonging to BC1 of Fig. 7 in close proximity to RS1 (asterisk); reconstruction from a single section of which a photomicrograph is shown in (K). The largeasterisk indicates darkly stained oedematous tissue marking the location of the recording pipette, the small asterisk a pyramidal cell labelled by iontophoresisof biocytin via the recording pipette. Arrows in (J) and (K) point to the same boutons; those which are not arrowed are out of the plane of focus in (K). (L)Reconstruction from a single section of a bouton-ladened terminal axon segment belonging to BC2 of Fig. 7 in close proximity to RS2 (asterisk). Scale bar in(J) is 20µm and also applies to (K and L).


effects, it is important to consider results of a number of intracellularstudies (Creutzfeldtet al., 1974; Innocenti & Fiore, 1974; Ferster,1986; Douglaset al., 1991; Satoet al., 1991; Bermanet al., 1992;Volgushevet al., 1993; Nelsonet al., 1994; Peiet al., 1994), whichhave thrown light on the way excitatory and inhibitory influencesinteract to produce response selectivity. Taken together, these studiescan be considered to have demonstrated the following major points:(i) in most cases, the excitatory and inhibitory input to a cortical cellis tuned to the optimal orientation and direction of motion; (ii) IPSPscan be evoked in response to stimuli presented at extreme non-optimal orientations and non-preferred directions of motion; and(iii) the magnitude of inhibition during non-optimal stimulation isinsufficient to suppress strong excitation comparable with that evokedby optimal stimuli. These findings have led several authors to proposethat intracortical inhibition contributes to cortical orientation/directionselectivity primarily via the rapid suppression of the excitatory inputfrom the thalamus during non-optimal stimulation and that theresultant excitation is amplified via recurrent excitatory connectionsamong cells with similar orientation/direction preferences (Douglas& Martin, 1991; Somerset al., 1995; Suarezet al., 1995; Vidyasagaret al., 1996). In this case, the excitatory load which needs to beopposed during non-optimal stimulation would be small, because thethalamic input is relatively weak, providing only 5–20% of theexcitatory synapses in layer IV (Garey & Powell, 1971; LeVay &Gilbert, 1976; LeVay, 1986; Peters & Payne, 1993; Ahmedet al.,1994), and the intracortical excitatory input would be biased fororientation/direction. Experimental support for the above hypothesisderives from intracellular recordings (Peiet al., 1994) and GABAinactivation experiments (Crooket al., 1996, 1997), which providedevidence for a major contribution of both cross-orientation inhibitionand intracortical iso-orientation excitation to the orientation tuningof cells which receive a monosynaptic input from the thalamus.Therefore, the most likely explanation for the present effects is thatthey were due to the loss of an inhibitory input from cells at theinactivation site to excitatory neurons in the vicinity of recording siteswith radically different orientation preferences (for cross-orientationinactivation) or opposite direction preferences (for iso-orientationinactivation) which, in turn, had recurrent iso-orientation/directionconnections with each other and a recorded cell. It is clear from bothanatomical (Gilbert & Wiesel, 1983; Martin & Whitteridge, 1984;Kisvarday et al., 1986; Gilbert & Wiesel, 1989; Ahmedet al., 1994;Andersonet al., 1994) and cross-correlation (Ts’oet al., 1986; Hataet al., 1991) studies that, although there are long-range excitatoryconnections between iso-orientation domains, the richest excitatoryinterconnections are made between neighbouring cells, and recentanatomical data suggest that lateral iso-orientation excitatory connec-tions are strongest over distances ofµ 200µm (Kisvarday et al.,1996). Thus, recurrent excitation among cells with similar orientation/direction preferences in the vicinity of recording sites could have


mediated substantial response amplification. Loss of inhibition inthese cells during non-optimal stimulation could allow amplificationof responses to non-optimal as well as optimal stimuli and lead to areduction in the response selectivity of a recorded cell. Because inmost cases both the recording and inactivation sites were located inlayers III–IV, which contain a high proportion of cells receiving amonosynaptic thalamic input (Harvey, 1980; Ferster & Lindstro¨m,1983; Martin & Whitteridge, 1984), many of the present effects mayhave reflected the loss of a rapid (disynaptic) inhibitory input torecurrently connected, monosynaptically excited cells. An orientation/direction preference could be conferred upon first-order inhibitory (andexcitatory) neurons via an oriented convergence of thalamocorticalafferents (Hubel & Wiesel, 1962) or an orientation/direction biasedinput from thalamic relay cells (Vidyasagar & Urbas, 1982; Jones &Sillito, 1994; Thompsonet al., 1994a,b). Although Fersteret al.(1996) have recently claimed that the thalamic input is sufficient togenerate orientation tuning in first-order simple cells in area 17, theirresults are actually not inconsistent with a contribution of intracorticalcircuitry to cortical orientation selectivity (Sompolinsky & Shapley,1997; and see discussion in Crooket al., 1997).

Impact of inhibitory projections from inactivation sites to recordingsites

Inactivation of inhibitory projections from cells at the inactivationsites is likely to have had a major impact on the response selectivityof recurrently connected excitatory neurons in the vicinity of recordingsites, for the following reasons.1 The actual number of GABAergic cells within 150µm of inactiva-tion sites with projections to injection sites would have been muchlarger than the three to six cells that were radiolabelled, becauseradiolabelling was confined to the upper surface of each (50–80 µm-thick) autoradiographic section, and in order to facilitate thereconstruction of electrode tracks, every third section was osmium-treated and hence could not be processed for autoradiography.2 Most radiolabelled cells in the vicinity of inactivation sites showedhigh grain densities, suggesting that they had high terminal densitiesand contacted a large number of cells at the injection sites.3 Many GABAergic cells which projected from inactivation sites toinjection sites would also have had terminals in adjacent regions, inthe vicinity of recording sites. This applies, in particular, to largebasket cells whose long-range horizontal axon collaterals extend inmany directions from the soma and distribute boutons in closelyspaced, localized clusters or radial columns (Martinet al., 1983;Somogyiet al., 1983).4 Large basket cells and clutch cells are well-suited to provide rapidfeedforward inhibition which seems necessary for the generationof orientation/direction selectivity; they can be monosynapticallyactivated by thalamic afferents, receive thalamic afferent input directlyon their somata, and have thick, heavily myelinated (presumablyrapidly conducting) axons (Somogyiet al., 1983; Kisvarday et al.,1985; Freundet al., 1986b).5 All types of basket cell have been shown to make synaptic contactprimarily with spiny (excitatory) neurons (DeFelipe & Fairen, 1982;Somogyiet al., 1983; Kisvarday et al., 1985).6 Together, basket cells and dendrite-targeting cells provide a richsynaptic input to the somata and proximal dendritic shafts of spinyneurons (DeFelipe & Fairen, 1982; Somogyiet al., 1983; Kisvardayet al., 1985; Tama´s et al., 1997a). These are very effective locationsfor inhibiting the excitatory input to spiny cells which arrives ondendritic spines and distal dendritic shafts (LeVay, 1973; Ahmedet al., 1994).7 Single basket or dendrite-targeting cells elicit in spiny neurons fast


IPSPs similar to those mediated by GABAA-receptors (Tama´s et al.,1997a), and blockade of GABAA-receptors has deleterious effects onorientation/direction selectivity (Sillito, 1977, 1979; Tsumotoet al.,1979; Sillito et al., 1980).

Impact of single-site inactivation

A remaining issue concerns the potential influence of single-siteinactivation lateral to a recorded cell in the context of the overallspatial organization of inhibitory connections. Vertical inhibitoryconnections, which were a prominent feature of the present results,do not seem to be necessary for the generation of orientation/directionselectivity (Malpeliet al., 1986; Schwarket al., 1986), although theymay make a minor contribution to orientation tuning (Allison &Bonds, 1994). In the tangential plane, radiolabelled cells were moreor less evenly distributed about injection sites with labelling extendinglaterally for 1–1.3 mm from the injection centre. These results are ingeneral agreement with those of previous studies in areas 17 (Albuset al., 1991; Albus & Wahle, 1994) and 18 (Matsubara & Boyd,1992) in which larger injections of retrograde tracers were combinedwith GABA immunohistochemistry, although these studies foundsmall numbers of GABAergic neurons at much greater lateral distances(up to 3 mm) from injection sites. Laterally remote radiolabelled cellswere sparsely distributed, and they had low densities of silver grains,suggesting that they had low terminal densities at the injection site.This accounts well for the failure of cross-correlation (Hataet al.,1991) and electrical stimulation (Welikyet al., 1995) studies to detectinhibitory interactions across lateral distances of more thanµ 700µm.As the periodicity of iso-orientation domains is about 1 mm (Bermanet al., 1987; Swindaleet al., 1987; Shumuel & Grinvald, 1996), thestrongest lateral inhibitory interactions between cells with oppositedirection preferences would seem to occur across iso-direction sub-regions located within the same iso-orientation domain. This wouldexplain why large disinhibitory effects on directionality could beelicited from a single inactivation site. On the other hand, comparisonof the tangential distribution of radiolabelled cells with the layout oforientation preference in areas 17 and 18 (Bermanet al., 1987;Swindale et al., 1987; Bonhoeffer & Grinvald, 1993; Bonhoefferet al., 1995) suggests that a single site receives inhibitory projectionsfrom a number of surrounding cross-orientation sites. This raises thequestion of whether inactivation of the inhibitory input from a singlecross-orientation site could account for the observed effects onorientation tuning. Pertinently, intracellular studies combined withcomputer stimulations (Bermanet al., 1992) have shown that whilefor small excitatory currents inhibition can suppress action potentialdischarge, once threshold is reached the effectiveness of inhibitiondeclines rapidly with increasing excitatory current. In view of thesemarked non-linear interactions between excitatory and inhibitoryinputs, inactivation of the inhibitory input from a single cross-orientation site to recurrently connected excitatory neurons in thevicinity of recording sites would probably have had a disproportion-ately large disinhibitory effect on their orientation selectivity; arecorded cell in contact with these cells might then show substantialbroadening of orientation tuning.

Additional features of the spatial organization of inhibitoryconnections: implications for orientation/direction selectivity

Regions adjacent to injection sites showed a high concentration ofradiolabelled cells, most of which were heavily invested with silvergrains, and the density of labelling decreased with increasing hori-zontal distance. The density of radiolabelled cells may have peakedclose to the injection centre, but the high concentration of silver


grains at injection sites made it extremely difficult to resolve individualcells. The high concentration and heavy labelling of radiolabelledcells close to injection sites is consistent with the fact that the axonsof most putative inhibitory neurons arborize locally, within a fewhundred microns of the soma (Peters & Regidor, 1981; Fairenet al.,1984) and also accounts well for intracellular data showing that theinhibitory input to a cortical cell is maximal for the optimal orientationand direction of motion (Ferster, 1986; Douglaset al., 1991; Satoet al., 1991). These dense, short-range inhibitory connections areprobably necessary to mediate feedback control of recurrent excitationwithin an iso-orientation/direction domain (Douglas & Martin, 1991;and see above).

Injections of [3H]-nipecotic acid which were made in layer III orupper layer IV produced widespread labelling in layers II–IV, whereasin deeper layers radiolabelled cells were confined largely withinµ 0.3 mm of the central axis of the injection. On the other hand,widespread labelling was observed in the deep layers following aninjection in lower layer IV which also involved the top of layer V,and this is consistent with the long-range tangential projections ofdeep-layer large basket cells (Kisva´rdayet al., 1987). Taken togetherwith the results from the present inactivation experiments, thissuggests that cortical orientation/direction selectivity is sharpened bylaterally directed inhibitory connections which are repeated acrosscortical layers or groups of layers. This notion is consistent with thefindings of Malpeli and colleagues (Malpeliet al., 1986; Schwarket al., 1986) which demonstrated a high degree of independence amongcortical layers in the generation of orientation/direction selectivity.

Conclusions

In our previous GABA inactivation experiments (Crook & Eysel,1992; Crooket al., 1996, 1997) we found that increases in responseto non-optimal orientations were elicited almost exclusively fromcross-orientation sites and increases in response to non-preferreddirections mainly from iso-orientation sites whose direction preferencewas opposite that of a recorded cell. In the present study, we havedemonstrated the presence of inhibitory projections from inactivationsites to recording sites in cases where these types of disinhibitoryeffect were elicited. Taken together, the results provide strong evidencethat cortical orientation tuning and direction selectivity are sharpened,respectively, by cross-orientation inhibition (Sillito, 1979; Morroneet al., 1982) and laterally directed iso-orientation inhibition (Emerson& Gerstein, 1977; Ganz & Felder, 1984; Emersonet al., 1987)between cells with opposite direction preferences. We suggest thatbasket cells and possibly also dendrite-targeting cells play a majorrole in the generation of both properties.

AcknowledgementsThis work was supported by the Deutsche Forschungsgemeinschaft.

AbbreviationsDI directionality indexGABA γ-aminobutyric acidIPSP inhibitory postsynaptic potentialIS inactivation sitePSTH peri-stimulus-time histogramRS recording site

ReferencesAhmed, B., Anderson, J.C., Douglas, R.J., Martin, K.A.C. & Nelson, J.C.

(1994) Polyneuronal innervation of spiny stellate neurons in cat visualcortex.J. Comp. Neurol., 341, 39–49.


Albus, K. & Wahle, P. (1994) The topography of tangential inhibitoryconnections in the postnatally developing and mature striate cortex of thecat.Eur. J. Neurosci., 6, 779–792.

Albus, K., Wahle, P., Lu¨bke, J. & Matute, C. (1991) The contribution ofGABAergic neurons to horizontal connections in upper layers of the cat’sstriate cortex.Exp. Brain Res., 85, 235–239.

Allison, J.D. & Bonds, A.B. (1994) Inactivation of the infragranular striatecortex broadens orientation tuning of supragranular visual neurons in thecat.Exp. Brain Res., 101, 415–426.

Anderson, J.C., Douglas, R.J., Martin, K.A.C. & Nelson, J.C. (1994) Synapticoutput of physiologically identified spiny stellate neurons in cat visualcortex.J. Comp. Neurol., 341, 16–24.

Beaulieu, C. & Somogyi, P. (1990) Targets and quantitative distribution ofGABAergic synapses in the visual cortex of the cat.Eur. J. Neurosci., 2,296–303.

Berman, N.J., Douglas, R.J. & Martin, K.A.C. (1992) GABA-mediatedinhibition in the neural networks of visual cortex. In Mize, R.R., Marc,R.E. & Sillito, A.M (eds), Progress in Brain Research, Vol. 90. ElsevierScience Publishers, Amsterdam, pp. 443–476.

Berman, N.E.J., Wilkes, N.E. & Payne, B.R. (1987) Organisation of orientationand direction selectivity in areas 17 and 18 of cat cerebral cortex.J. Neurophysiol., 58, 676–699.

Bonds, A.B. (1989) Role of inhibition in the specification of orientationselectivity of cells in the cat striate cortex.Vis. Neurosci., 2, 41–55.

Bonhoeffer, T. & Grinvald, A. (1993) The layout of iso-orientation domainsin area 18 of cat visual cortex: optical imaging reveals a pinwheel-likeorganization.J. Neurosci., 13, 4157–4180.

Bonhoeffer, T., Kim, D.-S., Malonek, D., Shoham, D. & Grinvald, A. (1995)Optical imaging of the layout of functional domains in area 17 and acrossthe area 17/18 border in cat visual cortex.Eur. J. Neurosci., 7, 1973–1988.

Chapman, B. & Stryker, M.P. (1992) Origin of orientation tuning in the visualcortex.Curr. Opin. Neurobiol., 2, 498–501.

Chronwall, B. & Wolff, J.R. (1980) Prenatal and postnatal development ofGABA-accumulating cells in the occipital neocortex of rat.J. Comp.Neurol., 190, 187–208.

Creutzfeldt, O.D., Kuhnt, U. & Benevento, L.A. (1974) An intracellularanalysis of visual cortical neurones to moving stimuli: responses in a co-operative neuronal network.Exp. Brain Res., 21, 251–274.

Crook, J.M. & Eysel, U.T. (1992) GABA-induced inactivation of functionallycharacterized sites in cat visual cortex (area 18): Effects on orientationtuning.J. Neurosci., 12, 1816–1825.

Crook, J.M., Eysel, U.T. & Machemer, H.F. (1991) Influence of GABA-induced remote inactivation on the orientation tuning of cells in area 18 offeline visual cortex: a comparison with area 17.Neuroscience, 40, 1–12.

Crook, J.M., Kisva´rday, Z.F. & Eysel, U.T. (1992a) Contribution of lateralinhibition to orientation and direction selectivity in cat visual cortex (area18): GABA-inactivation combined with injections of [3H]-nipecotic acid.In Elsner, N. & Richter, D (eds),Rhythmogenesis in Neurons and Networks.Georg Thieme, New York, p. 355.

Crook, J.M., Kisva´rday, Z.F. & Eysel, U.T. (1992b) Local mechanisms ofcortical direction selectivity: GABA-inactivation combined with injectionsof [3H]-nipecotic acid.Soc. Neurosci. Abstr., 18, 1033.

Crook, J.M., Kisva´rday, Z.F. & Eysel, U.T. (1996) GABA-induced inactivationof functionally characterized sites in cat visual cortex (area 18): Effects ondirection selectivity.J. Neurophysiol., 75, 2071–2088.

Crook, J.M., Kisva´rday, Z.F. & Eysel, U.T. (1997) GABA-induced inactivationof functionally characterized sites in cat striate cortex: Effects on orientationtuning and direction selectivity.Vis. Neurosci., 14, 141–158.

Curtis, D.R. & Johnston, G.A.R. (1974) Amino acid transmitters in themammalian central nervous system.Rev. Physiol., 69, 98–188.

DeFelipe, J. & Fairen, A. (1982) A type of basket cell in superficial layers ofthe cat visual cortex. A Golgi-electron microscope study.Brain Res., 244,9–16.

DeFelipe, J. & Jones, E.G. (1985) Vertical organization of gamma-aminobutyricacid-accumulating intrinsic neuronal systems in monkey cerebral cortex.J. Neurosci., 5, 3246–3260.

Douglas, R.J. & Martin, K.A.C. (1991) A functional microcircuit for catvisual cortex.J. Physiol. (Lond.), 440, 735–769.

Douglas, R.J., Martin, K.A.C. & Whitteridge, D. (1991) An intracellularanalysis of the visual response of neurones in cat visual cortex.J. Physiol.(Lond.), 440, 659–696.

Emerson, R.C. & Gerstein, G.L. (1977) Simple striate neurons in the cat. II.Mechanisms underlying directional asymmetry and directional selectivity.J. Neurophysiol., 40, 136–155.

Emerson, R.C., Citron, M.C., Vaughn, W.J. & Klein, S.A. (1987) Nonlinear


directionally selective subunits in complex cells of cat striate cortex.J. Neurophysiol., 58, 33–65.

Eysel, U.T., Muche, T. & Wo¨rgotter, F. (1988) Lateral interactions atdirection-selective striate neurones in the cat demonstrated by local corticalinactivation.J. Physiol. (London), 399, 657–675.

Eysel, U.T., Crook, J.M. & Machemer, H.F. (1990) GABA-induced remoteinactivation reveals cross-orientation inhibition in cat striate cortex.Exp.Brain Res., 80, 626–630.

Fairen, A., DeFelipe, J. & Regidor, J. (1984) Nonpyramidal cells: Generalaccount. In Peters, A. & Jones, E.G (eds),Cerebral Cortex, Vol. 1, CellularComponents of the Cerebral Cortex. Plenum Press, New York, pp. 201–253.

Ferster, D. (1986) Orientation selectivity of synaptic potentials in neurons ofcat primary visual cortex.J. Neurosci., 6, 1284–1301.

Ferster, D. & Lindstro¨m, S. (1983) An intracellular analysis of geniculo-cortical connectivity in area 17 of the cat.J. Physiol. (Lond.), 342, 181–215.

Ferster, D., Chung, S. & Wheat, H. (1996) Orientation selectivity of thalamicinput to simple cells of cat visual cortex.Nature, 380, 249–252.

Freund, T.F., Martin, K.A.C., Smith, A.D. & Somogyi, P. (1983) Glutamatedecarboxylase-immunoreactive terminals of Golgi-impregnated axoaxoniccells and of presumed basket cells in synaptic contact with pyramidalneurons of the cat’s visual cortex.J. Comp. Neurol., 221, 263–278.

Freund, T.F., Magloczky, Z., Solte´sz, I. & Somogyi, P. (1986a) Synapticconnections, axonal and dendritic patterns of neurons immunoreactive forcholecystokinin in the visual cortex of the cat.Neuroscience, 19, 1133–1159.

Freund, T.F., Martin, K.A.C., Somogyi, P. & Whitteridge, D. (1986b)Innervation of cat visual areas 17 and 18 by physiologically identified X-and Y-type thalamic afferents. II. Identification of postsynaptic targets by

GABA immunocytochemistry and Golgi impregnation.J. Comp. Neurol.,242, 275–291.

Gabbott, P.L.A. & Somogyi, P. (1986) Quantitative distribution of GABA-immunoreactive neurons in visual cortex (area 17) of the cat.Exp. BrainRes., 61, 323–331.

Ganz, L. & Felder, R. (1984) Mechanism of directional selectivity in simpleneurons of the cat’s visual cortex analyzed with stationary flash sequences.J. Neurophysiol., 51, 294–323.

Garey, L.J. & Powell, T.P.S. (1971) An experimental study of the terminationof the lateral geniculo-cortical pathway in the cat and monkey.Proc. R.Soc. London, B, 179, 41–63.

Gilbert, C.D. & Wiesel, T.N. (1983) Clustered intrinsic connections in catvisual cortex.J. Neurosci., 3, 1116–1133.

Gilbert, C.D. & Wiesel, T.N. (1989) Columnar specificity of intrinsic horizontaland corticocortical connections in cat visual cortex.J. Neurosci., 9,2432–2442.

Harvey, A.R. (1980) The afferent connections and laminar distribution of cellsin area 18 of the cat.J. Physiol. (London), 302, 483–505.

Hata, Y., Tsumoto, T., Sato, H. & Tamura, H. (1991) Horizontal interactionsbetween visual cortical neurones studied by cross-correlation analysis inthe cat.J. Physiol. (London), 441, 593–614.

Herz, A., Zieglga¨nsberger, W. & Fa¨rber, G. (1969) Microelectrophoreticstudies concerning the spread of glutamic acid and GABA in brain tissue.Exp. Brain Res., 9, 221–235.

Hubel, D.H. & Wiesel, T.N. (1962) Receptive fields, binocular interaction andfunctional architecture in the cat’s visual cortex.J. Physiol. (London), 160,106–154.

Humphrey, A.L., Sur, M., Uhlrich, D.J. & Sherman, S.M. (1985) Terminationpatterns of X- and Y-cell axons in the visual cortex of the cat: projectionsto area 18, to the 17/18 border region, and to both areas 17 and 18.J. Comp.Neurol., 233, 190–212.

Innocenti, G.M. & Fiore, L. (1974) Post-synaptic inhibitory components ofthe responses to moving stimuli in area 17.Brain Res., 80, 122–126.

Johnston, G.A.R., Stephanson, A.L. & Twitchin, B. (1976a) Uptake andrelease of nipecotic acid by rat brain slices.J. Neurochem., 26, 83–87.

Johnston, G.A.R., Krogsgaard-Larsen, P., Stephanson, A.L. & Twitchin, B.(1976b) Inhibition of the uptake of GABA and related amino acids in ratbrain slices by the optical isomers of nipecotic acid.J. Neurochem., 26,1029–1032.

Jones, H.E. & Sillito, A.M. (1994) Directional asymmetries in the length-response profiles of cells in the feline dorsal lateral geniculate nucleus.J. Physiol. (London), 479, 475–486.

Kisvarday, Z.F. (1992) GABAergic networks of basket cells in the visualcortex. In Mize, R.R., Marc, R.E. & Sillito, A.M (eds),Progress in BrainResearch, Vol. 90. Elsevier Science Publishers, Amsterdam, pp. 385–405.

Kisvarday, Z.F. & Eysel, U.T. (1993) Functional and structural topography ofhorizontal inhibitory connections in cat visual cortex.Eur. J. Neurosci., 5,1558–1572.


Kisvarday, Z.F., Martin, K.A.C., Whitteridge, D. & Somogyi, P. (1985)Synaptic connections of intracellularly filled clutch cells: a type of smallbasket cell in the visual cortex.J. Comp. Neurol., 241, 111–137.

Kisvarday, Z.F., Martin, K.A.C., Freund, T.F., Maglocsky, Z.S., Whitteridge,D. & Somogyi, P. (1986) Synaptic targets of HRP-filled layer III pyramidalcells in the cat striate cortex.Exp. Brain Res., 64, 541–552.

Kisvarday, Z.F., Martin, K.A.C., Friedlander, M.J. & Somogyi, P. (1987)Evidence for interlaminar inhibitory circuits in cat striate cortex.J. Comp.Neurol., 260, 1–19.

Kisvarday, Z.F., Kim, D.-S., Eysel, U.T. & Bonhoeffer, T. (1994) Relationshipbetween lateral inhibitory connections and the topography of the orientationmap in cat visual cortex.Eur. J. Neurosci., 6, 1619–1632.

Kisvarday, Z.F., Bonhoeffer, T., Kim, D.-S. & Eysel, U.T. (1996) Functionaltopography of horizontal neuronal networks in cat visual cortex (area 18).In Aertsen, A. & Braitenberg, V. (eds),Brain Theory—Biological Basisand Computational Principles. Elsevier Science Publishers, Amsterdam,pp. 97–122.

Kovalev, G.G. & Raevskii, K.S. (1981) Nipecotic acid, a competitive inhibitorof the net uptake of3H-GABA by rat brain synaptosomes.Bull. Exp. Biol.Med., 91, 692–694.

Kritzer, M.F., Cowey, A. & Somogyi, P. (1992) Patterns of inter- andintralaminar GABAergic connections distinguish striate (V1) and extrastriate(V2, V4) visual cortices and their functionally specialized subdivisions inthe rhesus monkey.J. Neurosci., 12, 4545–4564.

Krnjevic, K. (1984) Neurotransmitters in cerebral cortex: a general account.In Jones, E.G. & Peters, A (eds),Cerebral Cortex, Vol. 2. FunctionalProperties of Cortical Cells. Plenum Press, New York, pp. 39–61.

Krogsgaard-Larsen, P. & Johnston, G.A.R. (1975) Inhibition of GABA uptakein rat brain slices by nipecotic acid, various isoxazoles and relatedcompounds.J. Neurochem., 25, 797–802.

Larsson, O.M., Krogsgaard-Larsen, P. & Schousboe, A. (1980) High-affinityuptake of (RS)-nipecotic acid in astrocytes cultured from mouse brain.Comparison with GABA transport.J. Neurochem., 34, 970–977.

LeVay, S. (1973) Synaptic patterns in the visual cortex of the cat and monkey.Electron microscopy Golgi preparations.J. Comp. Neurol., 150, 53–85.

LeVay, S. (1986) Synaptic organisation of claustral and geniculate afferentsto the visual cortex of the cat.J. Neurosci., 6, 3564–3575.

LeVay, S. & Gilbert, C.D. (1976) Laminar patterns of geniculocorticalprojection in the cat.Brain Res., 113, 1–19.

Lund, J.S., Henry, G.H., Macqueen, G.L. & Harvey, A.R. (1979) Anatomicalorganisation of the primary visual cortex (area 17) of the cat. A comparisonwith area 17 of the monkey.J. Comp. Neurol., 184, 599–618.

Malpeli, J.G., Lee, C., Schwark, H.D. & Weyand, T.G. (1986) Cat area 17.I. Pattern of thalamic control of cortical layers.J. Neurophysiol., 56,1062–1073.

Martin, K.A.C. & Whitteridge, D. (1984) Form, function and intracorticalprojections of spiny neurones in the striate visual cortex of the cat.J. Physiol.(London), 353, 463–504.

Martin, K.A.C., Somogyi, P. & Whitteridge, D. (1983) Physiological andmorphological properties of identified basket cells in the cat’s visual cortex.Exp. Brain Res., 50, 193–200.

Matsubara, J.A. & Boyd, J.D. (1992) Presence of GABA-immunoreactiveneurons within intracortical patches in area 18 of the cat.Brain Res., 583,161–170.

Meyer, G. & Wahle, P. (1988) Early postnatal development of cholecystokinin-immunoreactive structures in the visual cortex of the cat.J. Comp. Neurol.,276, 360–386.

Morrone, M.C., Burr, D.C. & Maffei, L. (1982) Functional implications ofcross-orientation inhibition of cortical visual cells. I. Neurophysiologicalevidence.Proc. R. Soc. London, B, 216, 335–354.

Mountcastle, V.B. (1957) Modality and topographic properties of singleneurons of cat’s somatic sensory cortex.J. Neurophysiol., 20, 408–434.

Naegele, J.R. & Katz, L.C. (1990) Cell surface molecules containingN-acetylgalactosamine are associated with basket cells and neurogliaformcells in cat visual cortex.J. Neurosci., 10, 540–557.

Nelson, S., Toth, L., Sheth, B. & Sur, M. (1994) Orientation selectivity ofcortical neurons during intracellular blockade of inhibition.Science, 265,774–777.

Orban, G.A. (1984)Neuronal Operations in the Visual Cortex. Studies inBrain Function, Vol. 11. Springer-Verlag, Berlin.

Pei, X., Vidyasagar, T.R., Volgushev, M. & Creutzfeldt, O.D. (1994) Receptivefield analysis and orientation selectivity of postsynaptic potentials of simplecells in cat visual cortex.J. Neurosci., 14, 7130–7140.

Peters, A. & Payne, B.R. (1993) Numerical relationships betweengeniculocortical afferents and pyramidal cell modules in cat primary visualcortex.Cerebr. Cortex, 3, 69–78.


Peters, A. & Regidor, J. (1981) A reassessment of the forms of nonpyramidalneurons in area 17 of cat visual cortex.J. Comp. Neurol., 203, 685–716.

Sato, H., Daw, N.W. & Fox, K. (1991) An intracellular recording study ofstimulus specific response properties in cat area 17.Brain Res., 544,156–161.

Schwark, H.D., Malpeli, J.G., Weyand, T.G. & Lee, C. (1986) Cat area 17.II. Response properties of infragranular layer neurons in the absence ofsupragranular layer activity.J. Neurophysiol., 56, 1074–1087.

Shumuel, A. & Grinvald, A. (1996) Functional organization for direction ofmotion and its relationship to orientation maps in cat area 18.J. Neurosci.,16, 6945–6964.

Sillito, A.M. (1977) Inhibitory processes underlying the directional specificityof simple, complex and hypercomplex cells in the cat’s visual cortex.J. Physiol. (London), 271, 699–720.

Sillito, A.M. (1979) Inhibitory mechanisms influencing complex cellorientation selectivity and their modification at high resting discharge levels.J. Physiol. (London), 289, 33–53.

Sillito, A.M., Kemp, J.A., Milson, J.A. & Beradi, N. (1980) A re-evaluationof the mechanisms underlying simple cell orientation selectivity.BrainRes., 194, 517–520.

Somers, D.C., Nelson, S.B. & Sur, M. (1995) An emergent model of orientationselectivity in cat visual cortical simple cells.J. Neurosci., 15, 5448–5465.

Somogyi, P. (1989) Synaptic organization of GABAergic neurons and GABAAreceptors in the lateral geniculate nucleus and visual cortex. In Lam, D.K.-T.& Gilbert, C.D. (eds),Neural Mechanisms of Visual Perception. PortfolioPublishing Company, Houston, pp. 35–62.

Somogyi, P. & Solte´sz, I. (1986) Immunogold demonstration of GABA insynaptic terminals of intracellularly recorded, horseradish peroxidase-filledbasket cells and clutch cells in the cat’s visual cortex.Neuroscience, 19,1051–1065.

Somogyi, P., Kisva´rday, Z.F., Martin, K.A.C. & Whitteridge, D. (1983)Synaptic connections of morphologically identified and physiologicallycharacterized large basket cells in the striate cortex of cat.Neuroscience,10, 261–294.

Somogyi, P., Hodgson, A.J., Chubb, I.W., Penke, B. & Erdei, A. (1985)Antiserum to gamma-aminobutyric acid. II. Immunocytochemicalapplication to the central nervous system.J. Histochem. Cytochem., 33,245–248.

Sompolinsky, H. & Shapley, R. (1997) New perspectives on the mechanismsfor orientation selectivity.Curr. Opin. Neurobiol., 7, 514–522.


Suarez, H., Koch, C. & Douglas, R. (1995) Modeling direction selectivity ofsimple cells in striate visual cortex within the framework of the canonicalmicrocircuit.J. Neurosci., 15, 6700–6719.

Swindale, N.V., Matsubara, J.A. & Cynader, M.S. (1987) Surface organizationof orientation and direction selectivity in cat area 18.J. Neurosci., 7,1414–1427.

Tamas, G., Buhl, E.H. & Somogyi, P. (1997a) Fast IPSPs elicited via multiplesynaptic release sites by different types of GABAergic neurone in the catvisual cortex.J. Physiol. (London), 500, 715–738.

Tamas, G., Buhl, E.H. & Somogyi, P. (1997b) Massive autaptic self-innervationof GABAergic neurons in cat visual cortex.J. Neurosci., 17, 6352–6364.

Thejomayen, D.M. & Matsubara, J.A. (1993) Confocal microscopic study ofthe dendritic organization of patchy, intrinsic neurons in area 18 of the cat.Cerebr. Cortex, 3, 442–453.

Thompson, K.G., Zhou, Y. & Leventhal, A.G. (1994a) Direction-sensitive Xand Y cells within the A laminae of the cat’s LGNd.Vis. Neurosci., 11,927–938.

Thompson, K.G., Leventhal, A.G., Zhou, Y. & Liu, D. (1994b) Stimulusdependence of orientation and direction sensitivity of cat LGNd relay cellswithout cortical inputs: A comparison with area 17 cells.Vis. Neurosci.,11, 939–951.

Towe, A.L. (1975) Notes on the hypothesis of columnar organization insomatosensory cerebral cortex.Brain Behav. Evol., 11, 16–47.

Ts’o, D.Y., Gilbert, C.D. & Wiesel, T.N. (1986) Relationships betweenhorizontal interactions and functional architecture in cat striate cortex asrevealed by cross-correlation analysis.J. Neurosci., 6, 1160–1170.

Tsumoto, T., Eckart, W. & Creutzfeldt, O.D. (1979) Modification of orientationsensitivity of cat visual cortex neurones by removal of GABA-mediatedinhibition. Exp. Brain Res., 34, 351–363.

Vidyasagar, T.R. & Urbas, J.V. (1982) Orientation sensitivity of cat LGNneurones with and without inputs from visual cortical areas 17 and 18.Exp.Brain Res., 46, 157–169.

Vidyasagar, T.R., Pei, X. & Volgushev, M. (1996) Multiple mechanismsunderlying the orientation selectivity of visual cortical neurones.TrendsNeurosci., 19, 272–277.

Volgushev, M., Pei, X., Vidyasagar, T.R. & Creutzfeldt, O.D. (1993) Excitationand inhibition in orientation selectivity of cat visual cortex neurons revealedby whole-cell recordingsin vivo. Vis. Neurosci., 10, 1151–1155.

Weliky, M., Kandler, K., Fitzpatrick, D. & Katz, L.C. (1995) Patterns ofexcitation and inhibition evoked by horizontal connections in visual cortexshare a common relationship to orientation columns.Neuron, 15, 541–552.

Evidence for a contribution of lateral inhibition to orientation tuning ...

Documents

Transcript of Evidence for a contribution of lateral inhibition to orientation tuning ...