Download Orientation topography of layer 4 lateral networks revealed by

European Journal of Neuroscience, Vol. 11, pp. 4291±4308, 1999
ã European Neuroscience Association
Orientation topography of layer 4 lateral networks
revealed by optical imaging in cat visual cortex (area 18)
 va ToÂth and ZoltaÂn F. KisvaÂrday
Tagrid Yousef, Tobias Bonhoeffer,1 Dae-Shik Kim,2 Ulf T. Eysel, E
Abteilung fuÈr Neurophysiologie, Ruhr-UniversitaÈt Bochum, UniversitaÈtsstrasse 150, 44801, Bochum, Germany
1
Max-Planck Institute fuÈr Neurobiologie, Am Klopferspitz 18A, 82152 MuÈnchen-Martinsried, Germany
2
Center for Magnetic Resonance Research, University of Minnesota Medical School, 2021 6th Street, SE, Minneapolis,
MN 55455, USA
Keywords: anatomical tracing, lateral network, layer 4, optical imaging, orientation selectivity
Abstract
The functional speci®city of corticocortical connections with respect to the topography of orientation selectivity was studied by optical
imaging of intrinsic signals and bulk injections of ¯uorescent latex beads (green and red) and biocytin into layer 4. The distributions of
retrogradely labelled cells and anterogradely labelled axon terminals were histologically reconstructed from all cortical laminae, and
the resulting anatomical maps compared with the optically imaged functional maps. Layer 4 injections produced extensive horizontal
labelling up to 2±3 mm from the injection centres albeit without the clear patchy pattern described after layer 2/3 injections (Gilbert &
Wiesel, 1989, J. Neurosci., 9, 2432±2442; KisvaÂrday et al., 1997, Cerebral Cortex, 7, 605±618). The functional (orientation)
distribution of the labelled projections was analysed according to laminar location and lateral spread. With regard to the former, no
major difference in the orientation topography between supragranular- (upper tier), granular- (middle tier) and infragranular (lower
tier) layers was seen. Laterally, proximal and distal projections were distinguished and further dissected into three orientation
categories, iso- (6 30°), oblique- (6 30±60°) and cross-orientations (6 60±90°) with respect to the orientation preference at the
injection sites. The majority of distal connections (retrograde and anterograde) was equally distributed across orientations (35.4%
iso-, 33.7% oblique-, and 30.9% cross-orientations) that are equivalent with a preponderance to dissimilar orientations (oblique- and
cross-orientations, 64.6%). In one case, distal excitatory and inhibitory connections could be morphologically distinguished. For both
categories, a marked bias to dissimilar orientations was found (excitatory, 63.7%; inhibitory, 86.6%). Taken together, these results
suggest that the long-range layer 4 circuitry has a different functional role from that of the iso-orientation biased (52.9%, KisvaÂrday
et al., 1997, Cerebral Cortex, 7, 605±618) layer 2/3 circuitry, and is perhaps involved in feature difference-based mechanisms, e.g.
®gure ground segregation.
Introduction
Recent electrophysiological and psychophysical experiments suggest
that perceptual discrimination and grouping begin at the early level of
visual cortical processing (Allman et al., 1985; von der Heydt &
Peterhans, 1989; Nothdurft, 1993; Lamme, 1995; Kapadia et al.,
1995; Baumann et al., 1997; Kastner et al., 1997). Although the
underlying mechanisms of visual perception are not yet clear, the
anatomical substrate of the above processing has been tentatively
assigned to the long-range lateral network that could mediate direct
effects between visuotopic locations where the receptive ®elds no
longer overlap (Albus, 1975a; Tusa et al., 1979). On the basis of the
diversity of the interactions seen in the electrophysiological and
psychophysical experiments it is tempting to speculate that the longrange lateral networks should link cortical locations representing a
broad spectrum of the cue elements, e.g. orientation. Indeed, our
recent results obtained with focal tracer injections in layer 2/3 of the
visual cortex of the cat showed only a moderate iso-orientation bias
of long-range lateral connections implying that a considerable
proportion of connections indeed link non-iso-orientations, including
Correspondence: Dr Z. F. KisvaÂrday, as above.
E-mail: [email protected]
Received 24 May 1999, revised 19 August 1999, accepted 24 August 1999
cross-orientations (KisvaÂrday et al., 1997). Similar results were
obtained in the primate and tree-shrew visual cortex (Malach et al.,
1993, 1994; Bosking et al., 1997), casting doubt on former
suggestions as to the strict speci®city of lateral connections to either
iso- (Michalski et al., 1983; Nelson & Frost, 1985; Ts'o et al., 1986;
Gilbert & Wiesel, 1989) or cross-orientations (Matsubara et al., 1985,
1987).
Previous work on the functional speci®city of lateral networks has
focused exclusively on layer 2/3. Because different cortical layers are
composed of different neuron types, the question arises whether the
layer 2/3 ®ndings could also be applied to deeper laminae. In the
present study, we focused our interest on layer 4 connections.
Electrophysiological studies showed the continuity of similar
orientation preferences from layer 2/3 into layer 4 (Hubel &
Wiesel, 1962, 1963; Albus, 1975b; Bauer, 1982; Murphy & Sillito,
1986). Anatomical studies have shown that there are spiny stellate
(Martin & Whitteridge, 1984) and inhibitory basket cells (Somogyi
et al., 1983) in layer 4 which send long, lateral axons over several mm
similar to layer 2/3 pyramidal cells. Because layer 4 is the very ®rst
stage in the cortical processing of the arriving retino-thalamic signals
it would be important to know whether the lateral connections here
are distributed in a functionally, e.g. orientation, speci®c manner. To
this end, the retrograde tracer, ¯uorescent latex beads, and the
4292 T. Yousef et al.
anterograde tracer, biocytin, were injected into layer 4 and the
distributions of the labelled connections were reconstructed and
analysed with respect to functional orientation maps obtained with
the optical imaging technique based on intrinsic signal detection
(Bonhoeffer & Grinvald, 1991, 1993). The results show that the longrange component of the layer 4 lateral network establishes an overall
non-iso-orientation-biased topography suggesting a different role
from its layer 2/3 counterpart.
Materials and methods
Surgical procedures
Normal adult cats were acutely prepared for surgery using
standard procedures in accordance with institutional and Federal
Guidelines for the Care and Use of Laboratory Animals; initial
surgical anaesthesia was induced with intramuscular injection of a
mixture of ketamin, 7 mg/kg (Ketanest, Parke-Davis, Berlin,
Germany) and xylazin, 1 mg/kg (Rompun, Bayer Belgium, SintTruiden, Belgium). All wounds and pressure points were treated
with the local anaesthetic, xylocain gel (Astra Chemicals, Wedel/
Holstein, Germany) in order to minimize pain and discomfort to
the animal. During cranial surgery and throughout the entire
optical imaging experiments, the animals were lightly anaesthetized (N2O : O2, 70 : 30% plus Na-pentothal, 2 mg/kg/h) and
paralysed (alcuronium chloride, 0.06 mg/kg/h, i.a.), and their
physiological conditions (blood pressure, 95±120 mmHg; heart
rate; body temperature, 38.5 °C; expiratory CO2, 3.5±4%; and
EEG) continuously monitored.
From the beginning of the experiments, the corneas were protected
from drying out using neutral contact lenses and eye drops of 1.5%
saline. Thirty minutes prior to optical imaging the nictitating
membranes were retracted and the pupils dilated with 5% phenylephrinhydrochloride (Neosynephrin-POS, Ursapharm, SaarbruÈcken,
Germany) and 1% atropinsulphate (Atropin-POS, Ursapharm,
SaarbruÈcken, Germany), respectively. At this phase of the experiment, correction lenses were chosen on the basis of tapetal re¯ection.
A craniotomy (14 3 7 mm2) was made on the dorsal surface of the
skull centred to Horsley±Clarke co-ordinates A1 and L4 to expose the
quasi-central visual representation of areas 17 and 18 (Fig. 1). A
metal chamber of 30 mm inner diameter was cemented onto the skull
using dental cement (Paladur, Heraeus Kulzer GmbH, Wehrheim,
Germany). The dura mater was cut along the wall of the craniotomy
and removed. Damaged dura vessels were either cauterized (Fine
Science Tools GmbH, Heidelberg, Germany) or squeezed with a ®netip forceps until bleeding entirely stopped. The surface of the exposed
cortex was cleaned with arti®cial cerebrospinal ¯uid or Ringer
solution, and all tissue debris was carefully removed. The chamber
was ®lled with silicone oil (50 cSt viscosity, Aldrich, Milwaukee, WI,
USA) and sealed with a round glass-cover (Fig. 1A).
Visual stimuli and optical recording
Full-®eld visual stimuli of sinusoidal drifting (10±15 °/s) square-wave
gratings (0.1±0.4 cycles per degree) of four different orientations
were presented binocularly using a video screen (SONY) in 120 Hz
non-interlaced mode and the Neuroseq software (Vision Research
Graphics, Durham, NH, USA). For obtaining orientation maps the
stimulus grating was moved in opposite directions at each presentation. Individual stimuli were presented 25±50 times for 4.5 s followed
by an inter-stimulus interval for 10 s when the animals viewed a blank
screen. The cortex was illuminated with light at 605 6 5 nm using a
circular ®bre optic slit lamp (Schott, Mainz, Germany). Video images
were recorded with a slow-scan CCD camera (Bischke CCD 6012P,
FIG. 1. Photographs of the exposed cortical region subjected to the present
study. (A) Part of area 18 of the left hemisphere is seen through the glass
window of the imaging chamber. (B) A block of the same cortical region is
shown after histological ®xation. Asterisk indicates the location of an injection
site where green ¯uorescent latex beads were delivered by pressure. Dots
denote, approximately, the border between areas 17 and 18 (Tusa et al., 1979)
and the arrow points to the mm scale at the zero position of the ear-bar. Same
scale applies in A.
6 3 8 mm sensor) attached to a `tandem-lens' of two objectives (SMC
Pentax 1 : 1.2, 50 mm) resulting in a picture resolution of 21.28 mm
per camera pixel. The images were collected and processed using the
system Imager 2001 (Optical Imaging, Germantown, NY, USA) and
Matrox IM-640 imaging board controlled by the software
VDAQ218k.
Image analysis
Typically 10 blocks were collected and combined, each containing
®ve stimulus trials of individual stimuli. The images were analysed in
several steps. To correct for uneven illumination each image acquired
for a particular orientation and direction was divided by the sum of
images for all orientations and directions resulting in single condition
maps using the TVMIX program of the system Imager 2001 (Optical
Imaging). Differential maps were generated by dividing single
condition maps of opposite polarity (i.e. single condition maps
obtained with gratings of perpendicular orientations). Further
evaluation of the images was done in two steps using a custom
software written in IDL (Research Systems, Boulder, CO, USA). A
band-pass ®ltering was used to reduce noise (DC shift in the images
and small blood vessels) but retaining the dominant structure of the
orientation maps. Before the differential maps were vectorially
summed, high-pass ®ltering with a pixel-kernel equivalent of
1064 mm had been applied followed by a low-pass ®ltering. The
high-pass ®ltering parameter was chosen on the basis of the expected
periodicity of the orientation domains (~ 1.25 mm, Swindale et al.,
1987) and the low pass-®ltering parameter on the basis of the smallest
orientation domains, in the range of 200±300 mm in diameter.
Orientation (angle) maps were produced by summing the vector
components of differential maps on a pixel-by-pixel basis and
displayed with colour coding.
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4293
FIG. 2. Columnar arrangement of orientation selectivity in part of an orientation map revealed by electrophysiological recordings. (A) Zoomed-in orientation angle
map coded by the right-hand scheme in which the white spot indicates the locations of a recording site. Black bars represent averaged orientation vectors at each
location. (B) Polar-plots show orientation and direction responses of single sites at known distances from the cortical surface. (C) The average orientation
preference (thick bar) along the recording track was determined from the orientation preferences calculated by Fourier analysis (WoÈrgoÈtter & Eysel, 1987) of the
individual responses (thin bars). There is a clear match between the orientation preference shown by optical imaging and electrophysiological recordings including
layer 4. Scale bar, 100 mm.
Electrophysiological recordings from orientation columns
It was important to verify that the optically imaged orientation
preferences re¯ect the orientation preferences of the underlying
cortical tissue at any given image pixel position. Therefore, recordings
of multiunit, or occasionally single unit, activity were made from
several locations penetrating through several layers (Fig. 2). All
penetrations were made perpendicular to the cortical surface. Their
locations were chosen on the basis of single condition and angle maps
and, eventually, determined with the help of vascular images.
Electrical responses of single cells and multiunit clusters were
recorded with glass pipettes (inner tip diameter 5±10 mm) ®lled with
0.5 M Na-acetate. For visual stimulation, computer-controlled moving
light bars (Cambridge Electronic Design, CED-1401 laboratory
interface) were used generated with a `Picasso' image synthesizer
(Innisfree, Cambridge, MA, USA) and presented on a Tektronix 608
monitor 28.5 cm in front of the animal's eye. The stimulus orientation
was varied in pseudo-random sequence by multiples of 22.5° and
drifted with optimal velocity along the orthogonal axis of the orientation across the receptive ®eld area. Peristimulus time histograms (not
shown) and polar-plots (Fig. 2B) were generated from ®ve stimulus
sweeps in opposite directions at each orientation. The preferred
orientation at each recording site was calculated using Fourier analysis
of the raw data (SDO-analysis, WoÈrgoÈtter & Eysel, 1987).
Tracer injections and histology
The surface vascular pattern of the imaged regions served as
reference for guiding the injection pipette into particular locations of
the orientation preference maps. Extracellular pressure injections of
¯uorescent latex beads (10±15 mm tip diameter, 10±36 nL, Molecular
Probes, Eugene, OR, USA) or iontophoretic delivery of 2% biocytin
dissolved in 0.5 M Na-acetate (6±8 mm tip diameter, positive 0.6±
0.7 mA, 1-s ON/1-s OFF duty cycle for 20 min; Sigma, Deisenhofen,
Germany) were made into the middle cortical layer. In order to align
the histological images with optical images four to eight additional
reference penetrations were made around each injection using empty
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4294 T. Yousef et al.
FIG. 3. Matching between optical images and histological sections. (A and B) Superimposed images of surface vascular maps and reconstructions of sections
contours. Cross-hairs mark the location of reference penetrations made by empty glass electrodes in the vascular maps which caused small, recognisable damage to
the tissue along their courses. The light microscopic image of one of these such penetrations is seen at the upper (C) and lower surface (D) of an osmicated section.
Reference penetrations served to match the optical images with the histological reconstructions (see Materials and methods). (E and F) Zoomed-in vascular images
of (A and B), respectively, showing the goodness of matching of reference penetration positions as viewed in the in vivo optical image (cross-hairs) and as found in
the ®xed tissue (small dots). Notice that none of the corresponding locations differs by more than ~ 40 mm. Asterisks mark tracer injection sites. Scale bars, 1 mm
(A and B); 50 mm (C and D); 1 mm (E and F).
glass-pipettes (10±15 mm tip diameter). Their precise positions in the
vascular maps were marked on enlarged photographic images as well
as their stereotaxic locations documented (Fig. 3). After 30±45 h the
animals received an overdose of anaesthetics and were perfused
transcardially with a mixture of 2±4% paraformaldehyde and 0.5%
glutaraldehyde in 0.1 M phosphate buffer (pH 7.6). For the bead-
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4295
injected animals only paraformaldehyde was used. Large blocks of
area of interest were dissected, and frozen sections (30 mm) from the
bead-injected, and vibratome sections from the biocytin-injected
tissue were cut parallel to the cortical surface. In the case of bead
injections, alternate series of sections were collected. One series was
mounted onto chrome-gelatine-coated slides in a 1 : 1 mixture of
toluol and DePeX (SERVA Finebiochemica GmbH KG, Heidelberg,
Germany) for reconstructing the distribution of labelled cells. Onethird (OI17RHr) or one-quarter (OI20LHg and OI20LHr) of the
section series was stained for Nissl substance for determining laminar
boundaries, and another third (OI17RHr) or quarter (OI20LHg and
OI20LHr) of the series underwent osmication (1% OsO4 for 20 min,
dehydration and embedding in resin, Durcupan ACM, Fluka Chemie
AG, Buchs, Switzerland). The latter series served to recover the
location of reference penetrations. In the case of biocytin injections,
all sections were treated with osmium followed by embedding in
resin.
Reconstruction of the labelling
Fluorescent labelling is known to fade with time and under intense
illumination ¯uorescence rapidly bleaches. Therefore, great care was
taken to preserve the ¯uorescent signal partly by storing the sections
at 4 °C and partly by illuminating only the inspected region during
reconstruction. No such precaution with the resin-embedded,
biocytin-labelled specimens was necessary.
Labelled neuronal somata and axon terminals were reconstructed
from the entire depth of the cortex using 3 100 objective and the
reconstruction procedure reported earlier (KisvaÂrday et al., 1997).
Brie¯y, the position of each labelled structure was registered according
to x- and y-coordinates with the help of a personal computer and the
reconstruction system Neurolucida (MicroBrightField, Colchester,
VT, USA). Reference penetrations were determined from osmicated
sections (Fig. 3C and D) and their exact positions reconstructed in a
similar manner to labelled structures. Data of single sections were
assigned a z-value according to their distances from the cortical
surface. Finally, all individual sections of the same tissue block were
aligned. In this procedure, small blood vessels (< 20 mm diameter) that
run from one section into the adjoining section (bead-injected cases) or
labelled ®bres with corresponding cut ends (biocytin-injected cases)
served to ®nd the best match. In this manner, the data of many
individual sections were compiled into a single data ®le resulting in a
semi-three-dimensional reconstruction of the labelling. Concerning
the bead injections it needs to be added that the Nissl-stained and
osmium-treated sections could not be used for cell mapping.
Therefore, the distributions shown for the bead injections represent
only part of the entire labelling.
FIG. 4. Light microscopy of ¯uorescent bead labelling in 30-mm horizontal sections. (A) The injection site is shown at the level of upper layer IV at a depth of
630 mm from the pial surface. The core region of the injection site is characterized by the intense ¯uorescence of the tracer (diameter, < 320 mm) around which
labelled neurons are readily visible. The labelling density rapidly falls off over 100±150 mm from the injection centre indicating an uptake zone con®ned mainly to
the core region. The framed area is enlarged in B, showing bead-labelled cells in the halo zone (arrows). At distal locations the labelling intensity of neurons was
somewhat lower than for cells at the injection site. Nonetheless strongly labelled cells either alone (C) or in clusters (D) could be found up to several mm from the
injection site. In the present analysis, only cells with unequivocal labelling were taken into account. Scale bars, 100 mm (A); 10 mm (B,C); 50 mm (D).
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4296 T. Yousef et al.
FIG. 5. Light microscopy of a layer 4 biocytin injection. (A) The injection site and the resulting proximal axon terminal and soma labelling are shown. In remote
regions, two types of labelled axons could be distinguished. (B and C) Inhibitory axons of the basket cell type which characteristically formed en passant terminals
and often contacted perikarya (asterisks). (D and E) Putative excitatory axons gave off a mixture of en passant terminals (arrows in D) and terminals with short
stalks (arrowheads in D). Frequently these axon segments could be traced back to main, myelinated axons radiating out from the injection site. The arrow in E
indicates the direction of a thick, myelinated axon that branched into a cluster of thin collaterals. Scale bars, 100 mm (A); 10 mm (B and C); 20 mm (D and E).
FIG. 6. Orientation topography of layer 4 projections revealed by a combination of optical imaging and biocytin injection. (A±D) Single condition maps each
recorded during visual stimulation by moving gratings of one of four orientations (insets). Dark zones represent highly active regions whereas white zones are
less active. (E) The orientation angle map was generated by vector summation of the four single condition maps shown in (A±D). The colour-coded scheme
indicates which region prefers which orientation. (F±K) Topography of biocytin-labelled connections in the same, optically imaged region. The axon terminal
(F) and soma (G) labelling are displayed in a density matrix form where every pixel corresponds to a particular orientation image pixel. The colour-coded
logarithmic scale indicates the number of labelled structures per pixel. For a better comparison of the anatomical and optical images, a cross-hair is provided
in each panel centred onto the injection site. Accordingly, the injected zone is found in a yellow-orange zone of the angle map (E) representing close to horizontal stimulus orientation. Notice that the axon terminal labelling shows signs of clustering from ~ 800 mm (circle) from the injection centre. No clear clustering in the soma labelling is visible. (H) The reconstructed labelled terminals (red dots) and somata (white dots) are viewed in a frontal plane. The injection
centre corresponds to the highest density of labelling in the middle of layer 4. From this plane the labelling is smoothly distributed around the injection site.
The bouton cluster that occurs in the medial bank of the gyrus, probably corresponding to area 17, lays outside the optically imaged region (broken line) and,
therefore, is not part of the analysis. (I±K) Horizontal views of the labelling in the three main tiers reveal no marked differences in the distribution patterns.
The outlined areas refer to the region of the optical images. At each level, clustering is present, however, the clusters strongly differ in size, form and density
from one another. Most of the somata were found in regions of high bouton density indicating reciprocity in the projections. L, lateral; V, ventral. Scale bars,
1 mm (A±D); 1 mm (E±G); 1 mm (H±K).
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4297
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4298 T. Yousef et al.
Aligning procedure
The anatomical reconstructions were carefully aligned with the
optical images using reference penetrations (see Fig. 3). First, the
anatomical data were corrected for shrinkage caused by the ®xative
and subsequent histological treatment. In the following, all ®gures
have been corrected for tissue shrinkage. Then the same data were
tilted and rotated into the plane of the optical images. In practice, this
plane was found when the reconstructed reference penetrations
running perpendicular to that plane were viewed as spots. At this
stage, the anatomical reconstructions can be superimposed onto the
optical images using the corresponding reference penetration marks
(Fig. 3A and B). As shown in Fig. 3E and F, our aligning method had
an estimated error of < 50 mm.
Quantitative calculations
Optical maps are composed of pixels each covering a known area,
whereas the anatomical data represent coordinate points. In order to
obtain a quantitative view of the orientation distribution of the
labelled connections the aligned anatomical data set was converted
into a density distribution map using a pixel resolution (21.28 mm grid
constant) identical to that of the orientation maps. Orientation tuning
curves were calculated by counting the number of labelled structures
in each image pixel and the results were binned at every 22.5°. For a
direct comparison with previous quantitative results on the orientation distribution of lateral projections (KisvaÂrday et al., 1997;
Bosking et al., 1997) the labelling was grouped according to iso(6 30°), oblique- (6 30±60°) and cross-orientation (6 60±90°)
categories with reference to orientation preferences of the injection
sites (Table 2). Furthermore, the anatomical data were dissected into
proximal and distal regions by drawing a circle around the injection
core in the centre. Instead of using a ®xed value, we applied a
qualitative de®nition whereby the radius of the circle varied from
case to case. Because the clustering phenomenon is generally
associated with the long-range system, projections shorter than the
®rst occurring clusters were considered proximal and projections
longer than that distal. Accordingly, the radii dividing proximal and
distal projections ranged between 460 and 850 mm, providing an
average of 740 mm for the seven cases.
Results
Electrophysiological ®ndings
We tested whether the average orientation preferences detected by the
optical imaging technique differed from the orientation preferences
recorded with electrodes. To this end the orientation tuning of cortical
cells was determined at 19 domain locations of four animals. In all
cases, the optically imaged orientation preferences matched the
electrophysiologically determined preferred orientations. Secondly,
we carried out a laminar analysis for orientation selectivity to ®nd out
whether optical imaging could be applied for determining orientation
preferences below layer 3. Therefore, at 12 of the 19 recording sites,
orientation tunings were determined from several layers including
layer 4. An example for the results of such penetrations and the
orientation map of the same region is shown in Fig. 2. The recording
site (white spot in A) was in an orientation domain and had a
preference for near-horizontal orientations (orientation coding is in
grey-scale). Individual recordings at 12 different cortical depths of
the same site revealed similar orientation preferences as shown by the
generated polar plots (Fig. 2B). In a more condensed form, the same
set of data is represented by the bar-cluster which showed a scatter
< 20° around the average orientation preference of this cortical
column (Fig. 2C). Importantly, at all other penetrations a similarly
small orientation scatter was found when layers 2±4 were considered.
Anatomical ®ndings
Injection sites
In order to explore the functional speci®city of lateral connections of
layer 4, small injections of ¯uorescent latex beads (red and green) and
biocytin were made into regions of orientation domains. We selected
these locations for our injections because here the orientation
preferences between neighbouring columns change in small steps,
hence our injections could be restricted to a reasonably narrow range
of orientations showing an average standard deviation of 8.5° for the
seven injection sites. The locations of the injections were chosen to
be in the middle of the images so that much of the resulting labelling
could be con®ned to a large part of the imaged regions. In some cases,
however, the injections could only be made 50±200 mm lateral from
their desired locations in order to avoid damage to surface blood
vessels. The overall lateral extent of the injection sites was 150±
320 mm in diameter as marked by the dense tracer deposit and the
surrounding halo region. In Figs 4A and B, and 5A, light micrographs
show representative examples of the core and halo regions of the
injection sites taken at a low magni®cation.
The laminar position of the injections and the surrounding labelling
was determined in horizontal Nissl-stained and osmicated sections
using qualitative criteria, e.g. differences in cell and ®bre density,
soma size and the presence of large neurons, for example large
pyramidal cells in the border region of layers 3/4 and in layer 5. In
addition to these landmarks, rotation of the three-dimensional
reconstructions into a plane perpendicular to the cortical surface
was also helpful (Fig. 6H). We used a simpli®ed scheme of the sixlayered cortex whereby the super®cial layers (1±3) represented the
upper tier, layer 4 the middle tier and the deep layers (5±6) the lower
tier of the cortex. In the case of OI25LH, the border region between
the middle and lower tiers could not be determined unequivocally.
Therefore, the labelling was analysed by pooling the data of these two
cortical tiers together.
All seven injections were con®ned primarily to layer 4.
Nonetheless, in two bead-injected cases the involvement of other
layers cannot be excluded. Case OI20LHr comprised the lowermost
division of layer 3 (inset to Fig. 10C) and case OI20LHg overlapped
partly with the uppermost division of layer 5 (inset to Fig. 10E). In
the core regions, not all labelled cells and axon terminals could be
traced due to the strong tracer deposit. Around the core region,
however, both types of tracer injection had a halo zone where the
labelled structures could already be distinguished. In case of the bead
injections directly above the core, labelled cells were numerous
probably as a result of back¯ow of the tracer and, below the core,
labelled cells could take up the tracer via their ascending dendrites
(see insets of Fig. 10). The latter phenomenon was also observed in
the biocytin-injected cases (Fig. 6H). Here, many layer 6 cells below
the layer 4 injection were labelled probably by taking up the tracer
via their profusely branching apical dendrites.
General topography of the labelling
Fluorescent latex beads are known to be transported retrogradely. We
used this tracer to reveal neuronal somata with projection to the
injection site. Conversely, biocytin, commonly used as anterograde
tracer, was selected to determine the termination of axons of cells
located at the injection sites. While biocytin is indeed a potent
anterograde tracer, to some degree, it can also be transported
retrogradely as indicated by labelled somata at remote locations from
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4299
TABLE 1. Summary of the major experimental parameters
Injection site
Case
Tracer
Core radius
(mm)
Laminar
extent
Preferred
orientation (°)
OI20LHr
OI20LHg
OI17RHr
OI25LH
Red latex beads (10 nL)
Green latex beads (10 nL)
Red latex beads (36 nL)
Biocytin
75
120
100
95
Upper 4
Mid 4
Mid 4
Upper 4
OI26RH
Biocytin
106
F081292
F261093
Biocytin
Biocytin
160
92
Sections
(n)
(mm)
101
116
82
67
23*
23*
15*
21
30
30
30
70
Upper 4
122
17
70
Entire 4
Mid 4
76
87
28
21
70
70
Labelled somata (S)
and boutons (B) (n)
7184(S)
26519(S)
4488(S)
650(S)
38005(B)
141(S)
35739(B)
260525(B)
551(S)
59778(B)
Maximum lateral
extent of labelling
(mm)
3.93 3 3.05²
5.87 3 3.54²
3.77 3 3.90²
4.40 3 3.48
3.74 3 3.66²
3.05 3 2.76²
5.26 3 3.55²
4.94 3 3.90²
*Numbers represent alternate sections used in the reconstructions. ²Refers to anterior±posterior and medial±lateral directions (A±P 3 M±L).
the injections. We utilized this dual characteristic of biocytin for
simultaneous detection of reciprocal projections.
Labelled somata and axon terminals were found up to several mm
from the injection centres. In Fig. 4C, a light microscopic example is
shown for a cell labelled retrogradely with red-¯uorescent beads.
Such neurons were occasionally found in clusters up to several mm
from the injections (Fig. 4D). It is readily seen that the intensity of the
¯uorescent label of these remote cells is similar to the labelling
intensity of those lying proximal to the injection site (Fig. 4B).
Interestingly, biocytin-labelled somata were always less numerous
than bead-labelled somata although the injections were similar in
size. This indicates that the axon terminals have a lower af®nity in
taking up biocytin compared with latex beads, whereas the opposite
applies for somatic and dendritic uptake when the two tracers are
compared.
Distinction between excitatory and inhibitory cells
It is known that biocytin reveals the ®ne morphological details of
cells in a Golgi-stain manner, a characteristic that could be utilized
for distinguishing between excitatory and inhibitory components of
the labelled structures. We reported previously (KisvaÂrday et al.,
1997) that in addition to spiny neurons, biocytin stains basket cells of
the GABAergic cell population, some of which provide long-range
lateral inhibitory connections. In one of the four biocytin cases
(OI26RH), the quality of the staining allowed us to distinguish
between presumed excitatory and inhibitory elements on the basis of
morphological characteristics. In Fig. 5, examples for basket (Fig. 5B
and C) and presumed excitatory (Fig. 5D and E) axon collaterals are
illustrated. Typically, the basket axons gave off large en passant
boutons which often terminated on somata. Excitatory axons emitted
en passant and club-like boutons of various sizes. The global
branching pattern also differed between the two types of axon. Basket
axons rami®ed more frequently and did not form distinct patches
although small clustering could occur. Long-range excitatory axons
emitted thin collaterals some of which could form patches. The few
labelled basket somata found proximal to the injection site (< 300 mm,
Fig. 8E) gave off smooth, beaded dendrites. Contrary to this, spiny
neurons were quite numerous and could be found up to 3 mm (e.g. in
Fig. 6G).
Three-dimensional analysis of the labelling
After registering the three-dimensional position of each labelled cell
body and axon terminal, their laminar position was determined using
histological information of Nissl-stained sections and anatomical
landmarks visible in the osmicated sections. The reconstructed
distributions were divided into three main tiers using computergenerated planes running quasi-parallel to the cortical surface and
approximating laminar divisions. It needs to be added here that area 18
is not an entirely ¯at structure, hence some large sections contained
more than one lamina. By the same token, the planes used for dividing
cortical tiers could not exactly follow the depth ¯uctuation of the
laminar boundaries in the tissue. Thus, a small proportion of the
labelling was probably assigned to neighbouring tiers.
In general, all layer 4 injections resulted in labelling that spanned
virtually the entire cortical depth. There was no major difference in
the pattern of labelling between different laminae, proximal to the
column of injection the labelling was diffuse while more distally
clustering occurred. Labelled somata and axon terminals showed a
tendency to be in close neighbourhood indicating that the projections
are highly reciprocal. An example is provided in Fig. 6H±K for a
biocytin injection showing a similar, quasi-clustered labelling pattern
of somata (white dots) and axon terminals (red dots) in each of the
three main tiers. When the entire labelling is viewed from the same
plane in a form of density distribution, it is obvious that the overall
pattern strongly resembles the pattern of the single tiers indicating a
columnar-like arrangement for the most part of the densely labelled
locations (Fig. 6F and G). A similar trend was found for all other
injections including the bead injections illustrated in Fig. 10.
Another feature that is worth note concerns the clustering
pattern of the layer 4 lateral projections. The examples illustrated
in Figs 6, 8 and 10 revealed a different clustering pattern from
those reported for the labelling after layer 2/3 injections (Gilbert
& Wiesel, 1989; KisvaÂrday et al., 1997). The layer 4 injections
resulted in only a modest number of clusters possessing irregular
size and shape (Figs 6F, 8A, and 10A, C and E) and, in general,
were less obvious than in the case of the layer 2/3 injections
mentioned above. Furthermore, the average lateral extent of layer
4 projections measured here was 4.3 mm (Table 1) compared with
the 6.5±8 mm of the layer 2/3 circuitry (Gilbert & Wiesel, 1989;
KisvaÂrday et al., 1997). There was, however, an interesting
common characteristic between the layer 2/3 results and the
seven layer 4 cases shown here. Namely, the cell and axon
terminal density values between the labelled clusters seldom
showed zero values. This observation indicates that for the layer 4
projections, too, there is a kind of `baseline' between the injection
site and many locations especially at distances shorter than 2 mm.
All layer 4 injections gave rise to labelling up to 2±3 mm from
the injection site mainly in the same layer. The only exception
was found in the case of OI20LHg in which the injection site
included part of upper layer 5. Here the lateral extent of the
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4300 T. Yousef et al.
FIG. 7. Quantitative orientation distribution of biocytin-labelled axon terminals and somata shown in Fig. 6. (A) The orientation preference of the injection site was
determined by counting the number of orientation pixels corresponding to the core and immediate halo region. (B and C) The results, respectively, for the entire
bouton and soma labelling. (D±I) The same data with respect to different tiers. Proximal (broken lines) and distal (solid lines) projections differ in their
distributions. The former occupy mainly similar orientations and the latter display a strong tendency to link many different orientations.
labelling was up to 4 mm mainly in the lower cortical tier,
suggesting that the most distal projections of this case are the
product of layer 5 pyramidal neurons rather than projections of
cells in layers 2/3 and 4 (Gilbert & Wiesel, 1983; Martin &
Whitteridge, 1984; Gabbott et al., 1987).
comparing anatomical density maps and orientation maps are
shown in Figs 6A±G and 8A±E for biocytin injections, and Fig.
10A±F for bead injections. In order to foster a better comparison
of corresponding parts of the maps, a cross-hair centred on the
injection site was overlaid in each ®gure panel.
Comparison between the distribution of layer 4 lateral
projections and orientation maps
Quantitative results
We were interested to know the speci®city of layer 4 lateral
projections with respect to functional orientation preference maps.
To this end, the number of labelled structures (axon terminals and
somata, see Table 2) were counted in each image pixel and
displayed as density distribution matrices. This format of the
anatomical data was used to carry out a quantitative analysis of
their orientation speci®city at the spatial resolution of a single
image pixel (see Materials and methods for details). Examples for
A quantitative measure on the orientation speci®city of lateral
connections in layer 4 was obtained on entire cell distributions
rather than taking only individual clusters into account. As a ®rst
step, the orientation preferences of the injection sites were
determined in a circle representing the core and halo regions.
The results of two biocytin cases are shown, respectively, in Figs
7A and 8F, and three bead injection cases in Fig. 10G±I.
Interestingly, three of the seven injection sites showed skewed
orientation distributions over a relatively broad range (e.g. Fig.
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4301
FIG. 8. Topography of labelled excitatory and inhibitory connections in superposition with the vascular image of the part of area 18 under study. (A±D) Density
distributions of excitatory (A) and inhibitory (B) axon terminals and somata possessing spiny (C) and smooth (D) dendrites. Excitatory connections (A and C) show
larger lateral extent than inhibitory connections (B and D). Distal excitatory boutons display weak clustering while distal inhibitory boutons provided only two
small clumps. No clustering in the soma labelling was seen. (F) The orientation distribution of the injection site is skewed because it was not in the middle of an
orientation domain, some 250 mm away from the nearest orientation centre as shown in E. Scale bars, 1 mm (A±E).
8F), while the others were symmetric to a narrow orientation
range (e.g. Fig. 10I). With a close look at the orientation maps,
one sees that the injections of the former category were not in the
middle of an orientation domain, hence they occupied a few other
orientations in addition to the dominating one.
A laminar analysis of the projections was carried out using the
categories upper, middle and lower tiers instead of individual
laminae. Examples of this analysis are shown in Figs 7 and 9. The
orientation distribution curves differed only in minor details and the
main tendencies remained the same throughout the different tiers.
Therefore, for the sake of simplicity, the data of individual tiers were
pulled together in each case and used for analysing the orientation
speci®city of the projections as a function of lateral distance from the
injection site. In this analysis, proximal and distal projections were
separated on the basis of qualitative criteria (see Materials and
methods). The results showed that proximal projections always had a
major peak at similar orientation preferences with respect to that of
the injection sites (broken lines in graphs of Figs 7, 9 and 10). In
addition to this, in the cases where one or two nearby orientation
centres ¯anked the injection site, other high peaks could also occur at
different orientations (Fig. 10K and L). An interesting feature was
that none of the distribution curves including those of distal
projections was symmetric to the average orientation preference of
the injection site. Most of the distributions were skewed even in the
cases where the injection site had a narrow, symmetrically distributed
orientation range (e.g. Fig. 10G). This ®nding suggests that in¯uences
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4302 T. Yousef et al.
FIG. 9. Orientation speci®city of lateral excitatory and inhibitory connections according to the main tiers of cortical lamination. There is no major difference between
excitatory and inhibitory components in either tier. Proximal projections prefer similar orientations and distal projections have the tendency to link dissimilar
orientations. In the upper tier, the nine labelled somata at the injection site were found at iso-orientations (not shown). Same conventions apply as in Fig. 7.
other than solely orientation selectivity could also play a role in the
organization of the projections. Concerning the orientation preference
of distal projections, a different picture emerged from that seen for
the proximal projections. They commonly represented the entire
orientation range in a manner that resulted in an overall preference to
non-iso-orientations. When the curves of the distal projections are
viewed in detail it is clear that the peaks can occur up to 90° around
the orientation preference of the injection sites. This trend was found
in all other instances in addition to those shown in Figs 7, 9 and 10.
In one case, the excitatory and inhibitory structures could be
analysed separately. Figure 9 shows the quantitative results for the
three tiers of this case. When the middle tier projections are
viewed, there is no clear difference between the excitatory (Fig.
9C) and inhibitory (Fig. 9D) distribution curves. Concerning the
upper tier, however, there was a marked difference between the
distal excitatory and inhibitory curves (Fig. 9B). While this
situation could be typical for upper tier projections, the overall
distribution, which was clearly dominated by the middle-tier
projections, showed no marked differences between excitatory and
inhibitory connections. A similar picture emerged when the
labelled excitatory (P-type) somata were taken into account
(Fig. 9E). Proximally, they chie¯y occupied similar orientations
and distally they preferred a broad range of orientations.
Inhibitory (B-type) somata were found only proximal to the
injection site at similar orientations (Fig. 9E).
Comparison between the labelling methods
In order to summarize the orientation distributions according to the
applied methods, the following three labelling categories were
established; ¯uorescent bead-labelled somata (Fig. 11A and B),
biocytin-labelled somata [including excitatory and inhibitory
types (Fig. 11C and D) representing retrograde projections] and
biocytin-labelled axon terminals [(Fig. 11E and F) representing
anterograde projections]. The average `orientation tuning' curve in
each category is indicated by a thick line and the data of individual
cases by thin grey lines. With regard to proximal projections (Fig.
11A, C and E), there are some notable differences between the
different methods. In the case of bead-labelled cells (Fig. 11A) and
anterogradely labelled terminals (Fig. 11E), strong non-iso-orientation components were present around the dominant iso-orientation
peak that were not seen for the biocytin-labelled somata (Fig. 11C).
Concerning distal projections (Fig. 11B, D and F), a common feature
of all labelling categories was that the distribution curves were more
¯at than their proximal counterparts and showed no clear-cut
preference to any particular orientation range. Interestingly, in the
case of biocytin-labelled somata (Fig. 11D) where a strong ¯uctuation
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4303
in the curve was present the lowest intensity was found around
similar orientations.
In order to provide an overall view of the quantitative results, the
data presented in Fig. 11 are shown in a cumulative form in Fig. 12A
where all cases (anterograde and retrograde) are pulled together. The
main message of the resulting graphs is that distal connections show a
preference towards dissimilar orientations whereas proximal projections are engaged with similar orientations.
For a direct comparison of the present results with the known
orientation preference of the layer 2/3 long-range circuitry
(KisvaÂrday et al., 1997), the distribution values were expressed in a
simple form as shown in Fig. 12B. We used a previously established
method for dissecting the orientation map into categories of iso(6 30°), oblique- (6 30±60°) and cross-orientations (6 60±90°) with
respect to the orientation preferences at the injection sites, and
analysed the pooled data of the layer 4 connections accordingly.
There are two important facts to note. First, there is no appreciable
difference between the three orientation categories within any of the
labelling groups corresponding to anterograde and retrograde
labelling (Fig. 12B). Secondly, each group provided very similar
distributions that indicate their strong reciprocal relationship. Taken
together, the above quantitative data for the distal connections show
that all orientation categories are rather equally represented. As a
consequence of such an orientation topography, non-iso-orientation
connections of oblique- and cross-orientations exceed iso-orientation
connections by a ratio of two to one.
Discussion
The present study was undertaken to explore the orientation
selectivity of lateral connections with respect to layer 4. The main
®nding is that lateral connections link all orientations in a rather
balanced manner. Because by de®nition the iso-orientation category
represents only one-third of the entire orientation range, the ®ndings
are equivalent with a bias of the layer 4 lateral connections to noniso-orientations. These data together with previous ®ndings on the
iso-orientation bias of the layer 2/3 circuitry (Gilbert & Wiesel, 1989;
Malach et al., 1993; Bosking et al., 1997; KisvaÂrday et al., 1997)
extend our view as to the heterogeneity of lateral connection to
different layers, in this case to layer 4.
Technical considerations
The major conclusion of the present study on the orientation speci®city
of layer 4 projections relies largely on data where no distinction
between inhibitory and excitatory components of the labelling could
be made. In a previous study, using a combination of electrophysiological mapping of orientation selectivity and anatomical tracing of
layer 2/3 neuronal connections, we showed that inhibitory projections
tend to surround the termination sites of the excitatory projections
(KisvaÂrday et al., 1997). As a result of such an arrangement, the
inhibitory elements provided a smaller contribution to iso-orientation
locations than the iso-orientation biased excitatory ones. On the basis
of these ®ndings a logical assumption is that the percentage values
provided here for the oblique- and cross-orientation categories (in Fig.
12B) re¯ect largely inhibitory rather than excitatory connections. We
think that this is not the case, because in the single case, where
inhibitory and excitatory components could be distinguished, strong
non-iso-orientation components were provided both by the excitatory
and inhibitory projections at distal locations (Figs 8 and 9). Moreover,
the known limited lateral extent of inhibitory projections (1±1.5 mm,
for review see KisvaÂrday, 1992) indicates that the most distal
projections must represent purely excitatory links.
Another point that should be mentioned regards the orientation
distribution curves in Fig. 11. As is always the case when different
tracers are used, the present study showed some noticeable
differences between the soma labelling by latex beads and biocytin
(Fig. 11A±D). Although we do not know exactly the underlying
factors for the observed differences, we think that the different uptake
mechanisms involved and the mode of tracer delivery, pressure
injection for the beads and iontophoresis for biocytin played an
important role. A similar reasoning may apply when comparing the
different data in the literature (Matsubara et al., 1987; Gilbert &
Wiesel, 1989).
A third point that needs to be taken into account is whether the
®ltering properties of the optical images can signi®cantly affect the
orientation distribution of the analysed connections. In order to know
this we carried out calculations for three cases (OI20LHr, OI20LHg,
OI17RHr). In each case, the optical data were processed with nine
different combinations of high- and low-pass ®lters (Fig. 13A), and
then the resulting images used for calculating the orientation
distribution of the anatomical connections. As one can see in
Fig. 13B and C, the orientation distributions of the connections were
nearly identical, there was no statistically signi®cant difference (nonparametric test, Pearson correlation) between the curves. Therefore,
we are sure that changing the ®ltering parameters of the images does
not affect the statistics of the connections.
Finally, it is well known that orientation selectivity in the same
cortical columns varies slightly from cell to cell (present study of area
18, < 20°; area 17, 5±15° in Albus, 1975b; and 9±18° in Murphy &
Sillito, 1986). This scatter of orientation selectivity has to be taken
into account for estimating the speci®city of connections between
different locations. Using the above values the exactness of our
calculations, between any two cortical points, is estimated to be in the
range of 6 30±40°, corresponding to a maximum error of 30±45%.
Although such a potential error seems to be quite high, it is very
much in line with the broad orientation tuning width of individual
cells in area 18 (70°, see ®gure 7/6B in Orban, 1984).
Orientation speci®city of layer 4 lateral connections
Previous studies on the topographic relationship between functional
maps and anatomical connections have chie¯y been restricted to the
super®cial layers. One of the reasons for this stems from the notion
that optical imaging signals derive largely from super®cial layers and
upper layer 4 (Bonhoeffer & Grinvald, 1995). In this regard, our
optical imaging camera was focused 750 mm below the cortical
surface, instead of the commonly used 600 mm, to lower the focal
plane well into layer 4. When consecutive images were taken at 750
and 600 mm from the surface there was no appreciable difference
between the resulting orientation maps. There are two possibilities in
this regard. First, our camera cannot view structures deeper than layer
3, and the images simply re¯ect activity changes from the super®cial
layers. Alternatively, our camera can detect deep layer signals, and
the correspondence between upper and lower layer images shows that
orientation selectivity is organized in columns. At this stage we do
not know whether our camera did indeed image layer 4.
The second reason that most studies concentrate on super®cial
layer connections is due to the well-known presence of pyramidal
cells which provide extensive patchy projections (Rockland & Lund,
1982; Gilbert & Wiesel, 1983; KisvaÂrday & Eysel, 1992). However,
there is anatomical evidence that other layers could also play an
important role in horizontal interactions. Indeed, some layer 4 spiny
stellate cells give rise to clustered axons in layer 3 (Martin &
Whitteridge, 1984), and some layer 5 pyramidal cells also have
clustered projections in their own layer and in overlying layers 1±3
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4304 T. Yousef et al.
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4305
(Gilbert & Wiesel, 1983; Martin & Whitteridge, 1984; Gabbott et al.,
1987). These anatomical ®ndings strongly suggest that lateral
connections originating in the supragranular layers represent only
part of the horizontal system in the visual cortex. Then the question
arises whether the orientation preferences of layer 4 lateral
connections differ from those of the more super®cial and deep
layers. Our quantitative results clearly demonstrated that layer 4
distal projections possess an almost equal preference to iso-, obliqueand cross-orientations (see Fig. 12) that is not the case for the layer 2/
3 projections of area 18 (KisvaÂrday et al., 1997). Concerning the
orientation topography of the deep layers, there are no available data
yet.
FIG. 11. Quantitative evaluation of the
orientation speci®city of layer 4 lateral
projections as distinguished by different
labelling methods: (A and B) with ¯uorescent
latex beads; (C and D) retrograde biocytin
labelling; (E and F) anterograde biocytin
labelling. Left-hand panels show graphs of
proximal, right-hand panels distal projections.
Individual cases are drawn by grey lines and
averages shown by black lines. Generally,
proximal projections are engaged with similar
orientations although other orientations are
also represented for the bead labelling. A
different distribution is seen for distal
projections, they are distributed to many
orientations with a bias to dissimilar
orientations.
FIG. 10. Orientation topography of ¯uorescent bead-labelled cells after layer 4 injections. (A, C and E) The three cases of bead labelling in superimposition
with vascular maps. (B, D and F) Their corresponding orientation angle maps. Insets on the left-hand side indicate the laminar location of the injection cores
(in red), the halo regions (in black) and sites of possible back¯ow (in grey) of the tracer along the injection track. In all three cases, the main body of the injection was con®ned to layer 4 (see insets), although for case OI20LHg with some additional contribution to upper layer 5. The labelling was densest at proximal locations (within the circles). More distally, only a few and less densely labelled clusters emerged providing an overall pattern that resembles the
biocytin cases. The orientation preferences at the injection sites shown in (G±I) were calculated from respective angle maps shown in (B, D and F). Of the
three injections, case OI20LHr was made closest to the middle of an orientation domain; this resulted in the narrowest orientation range, a single orientation
bin in (G). Concerning the labelled projections (J±L), in none of the cases could a clear segregation according to a narrow orientation range be observed.
Importantly, distal projections (solid lines) show an overall preference to dissimilar orientations. The same conventions apply as in Fig. 7. Scale bar, 1 mm
(A±F).
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4306 T. Yousef et al.
FIG. 12. (A) Cumulative orientation distribution of layer 4 projections as revealed by the present study, including retrograde and anterograde labelling. Proximal
projections, grey line; distal projections, black line. Broken line indicates the percentage of distal projections if they were evenly distributed to orientations. (B) Bar
graphs show the cumulative results for anterograde (biocytin-labelled boutons), retrograde (bead + biocytin-labelled somata) and all (anterograde + retrograde)
cases according to categories of iso- (6 30°), oblique- (6 30±60°) and cross-orientations (6 60±90°) with respect to the orientation preferences calculated for the
injections. Non-iso-orientation (oblique- and cross-orientations) connections account for 64% of all distal projections.
TABLE 2. Orientation distribution of proximal and distal projections
Case
Label
OI20LHr
Somata
OI20LHg
Somata
OI17RHr
Somata
OI25LH
Boutons
Somata
OI26RH
Boutons
Somata
F081292
Boutons
F261093
Boutons
Somata
Average
Distance
from initial
segment
Orientation preference
Iso
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
Proximal
Distal
48.9
41.3
45.9
36.0
33.8
30.7
60.8
41.2
89.7
37.0
66.4
32.6
73.6
40
29.2
38.9
68.9
30.3
82.7
26.0
Proximal
Distal
60.0
35.4
(2531)
(829)
(7931)
(3336)
(1147)
(338)
(18325)
(3245)
(559)
(10)
(17950)
(2840)
(89)
(8)
(38547)
(54842)
(23719)
(7685)
(354)
(32)
Oblique
Cross
28.0
19.4
16.0
25.3
32.0
28.5
15.8
35.1
4.0
37
24.4
42.8
19.0
50
32.7
35.5
30.9
35.5
17.3
27.6
23.1
39.3
38.1
38.7
34.2
40.8
23.4
23.7
6.3
26.0
9.2
24.6
7.4
10
38.1
25.6
0.2
34.2
0.0
46.4
22.0
33.7
(1449)
(389)
(2759)
(2341)
(1083)
(313)
(4763)
(2763)
(25)
(10)
(6581)
(3724)
(23)
(10)
(43156)
(41592)
(10640)
(8995)
(74)
(34)
Total (100%)
(1199)
(787)
(6571)
(3581)
(1158)
(449)
(7038)
(1871)
(39)
(7)
(2498)
(2146)
(9)
(2)
(50337)
(32051)
(92)
(8647)
(0)
(57)
(5179)
(2005)
(17261)
(9258)
(3388)
(1100)
(30126)
(7879)
(623)
(27)
(27029)
(8710)
(121)
(20)
(132040)
(128485)
(34451)
(25327)
(428)
(123)
18.0
30.9
Data are given as a percentages, with absolute values in parentheses.
Previous estimates on the orientation speci®city of lateral
connections
For the aforementioned reasons, virtually all available information on
the speci®city of lateral projections with respect to orientation are
concerned with layer 2/3. Brie¯y, the results obtained in cat could be
divided into two major groups as to whether they support the
connectivity concepts of linking similar (Michalski et al., 1983;
Nelson & Frost, 1985; Ts'o et al., 1986; Gilbert & Wiesel, 1989) or
dissimilar orientations (Matsubara et al., 1985, 1987). Interestingly,
recent cross-correlation data indicate that lateral connections are
neither highly speci®c nor random for orientation (Hata et al., 1991;
Tamura et al., 1996). Functional±anatomical results strongly support
such a view. In the study of KisvaÂrday et al. (1997), 53±59% of
excitatory and 46±48% of inhibitory connections were shown to
prefer similar orientations, whereas a considerable proportion of both
functional types connected to dissimilar orientations including crossorientations.
Concerning other species, the data on the orientation speci®city of
lateral connections have also been restricted to the super®cial layers. In
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
Functional topography of orientation selectivity in layer 4 4307
their measurements did not include projections outside the patches,
where non-iso-orientations are likely to be encountered, the actual
numerical ®gure given in their report probably represents an underestimate of the total proportion of non-iso-orientation connections. In
V1 of the tree shrew, the orientation analysis of corticocortical
connections was carried out by taking into account patch as well as
interpatch regions (Bosking et al., 1997). The result obtained for
distances longer than 500 mm from the injection centres showed that an
average of 57.6% of the connections occupied sites with orientation
preferences within 6 35° of the average preferred orientations of the
injection sites. This ®gure is in concordance with our recent
calculation obtained in layer 2/3 of the cat visual cortex (average,
52.9%, > 650 mm using 6 30° con®dence range in KisvaÂrday et al.,
1997) but it differs from the present layer 4 data.
Functional implications
FIG. 13. An example of testing whether different ®ltering parameters can
signi®cantly change the distribution of orientation domains and, therefore, the
orientation statistics of lateral connections. (A) The nine combinations of highand low-pass ®ltering values applied in the test. (B and C) The resulting
orientation distribution graphs are shown for the proximal and distal
connections of case OI20LHr, respectively. Clearly, in both panels, the nine
graphs have an almost identical course indicating that ®ltering of optical
images plays little if any role in the orientation statistics of the connections
studied.
the primate (Malach et al., 1993, 1994) and the tree shrew visual cortex
(Bosking et al., 1997), optical imaging of orientation selectivity was
combined with anterograde tracing of biocytin. In the study of Malach
et al. (1993), approximately half of the patchy projections were found
at orientations which were not present at the injection sites. Because
The present ®ndings show that the layer 4 horizontal network has a
complex relationship with orientation maps. We believe that such a
circuitry could account for electrophysiological and psychophysical
results that are dif®cult to interpret by simple network rules, e.g. `like
connects to like'. Layer 4, being the main recipient layer of the primary
afferents, is assumed to be involved in the generation of novel
receptive ®eld features like orientation selectivity. In this regard,
recent data from our laboratory using the remote inactivation paradigm
indicate that in layer 4 lateral cross-orientation inhibitory interactions
have an in¯uence on the sharpening of orientation selectivity (Crook
et al., 1998). The present study provides anatomical support for the
observed effects. Furthermore, our results are compatible with the
view in which the above inhibitory effects could be mediated either
directly using cross-orientation inhibition or indirectly, via crossoriented excitatory connections terminating on inhibitory cells.
At this stage, we can only anticipate that a similar circuitry to the
one shown here underlies a range of other electrophysiological
observations. For example, in primate V1 some neurons prefer
complex visual stimuli consisting of radically different orientation
components, respectively, within and outside of their `classical'
receptive ®elds (Sillito et al., 1995). In cat visual cortex, there are socalled double orientation tuned neurons in most layers including layer
4 which are best driven by cross-like ®gures (Shevelev et al., 1994,
1995). Conceivably, the `input±output' sites of these neurons ought to
be engaged with many different orientations as in the network
presented here.
Finally, a potential implication of lateral connections associated
with a broad range of functional speci®city is that the individual
neuron is subjected to a host of modulatory effects (Gilbert &
Wiesel, 1990; Kapadia et al., 1995). Following this line of thought
onto large networks, such interactions might be crucial in
accounting for some psychophysical effects. For example, ®gureground segmentation tasks and perceptual grouping of distributed
feature elements probably involve mechanisms based on feature
dissimilarity (Nothdurft, 1992; Lamme, 1995; Kastner et al., 1997;
Baumann et al., 1997) or feature similarity (for review see Singer
& Gray, 1995), respectively. The present study suggests that for
orientation cues such interactions could be mediated at a ®rst
stage already by the layer 4 circuitry.
Acknowledgements
The authors would like to thank Ms D. Strehler and Ms P. Hentrich for
photography, Mr RoÂbert Magyar for his computer expertize, and Drs PeÂter
BuzaÂs, Maxim Volgushev and Bashir Ahmed for helpful comments on earlier
version of the manuscript. This work was supported by the Deutsche
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308
4308 T. Yousef et al.
Forschungsgemeinschaft (SFB509, TPA6 to Z.F.K. and Ey 8/23) and the
European Communities (SC1 0329-C).
References
Albus, K. (1975a) A quantitative study of the projection area of the central and
paracentral visual ®eld in area 17 of the cat. I. The precision of the
topography. Exp. Brain Res., 24, 159±179.
Albus, K. (1975b) Quantitative study of the projection area of the central and
paracentral visual ®eld in area 17 of the cat. II. The spatial organization of
the orientation domain. Exp. Brain Res., 24, 181±202.
Allman, J., Miezin, F. & McGuinness, E. (1985) Stimulus speci®c responses
from beyond the classical receptive ®eld: neurophysiological mechanisms
for local-global comparisons in visual neurons. Annu. Rev. Neurosci., 8,
407±430.
Bauer, R. (1982) A high probability of an orientation shift between layers 4
and 5 in the central parts of the cat striate cortex. Exp. Brain Res., 48, 245±
255.
Baumann, R., van der Zwan, R. & Peterhans, E. (1997) Figure-ground
segregation at contours: a neural mechanism in the visual cortex of the alert
monkey. Eur. J. Neurosci., 9, 1290±1303.
Bonhoeffer, T. & Grinvald, A. (1991) Iso-orientation domains in cat visual
cortex are arranged in pinwheel-like patterns. Nature, 353, 429±431.
Bonhoeffer, T. & Grinvald, A. (1993) The layout of iso-orientation domains in
area 18 of cat visual cortex: optical imaging reveals a pinwheel-like
organization. J. Neurosci., 13, 4157±4180.
Bonhoeffer, T. & Grinvald, A. (1995) Optical imaging based on intrinsic
signals: the methodology. In Toga, A. & Mazziotta, J.C. (eds), Brain
Mapping: The Methods. Academic Press, San Diego, pp. 55±97.
Bosking, W.H., Zhang, Y., Scho®eld, B. & Fitzpatrick, D. (1997) Orientation
selectivity and the arrangement of horizontal connections in tree shrew
striate cortex. J. Neurosci., 17, 2112±2127.
Crook, J.M., KisvaÂrday, Z.F. & Eysel, U.T. (1998) Evidence for a contribution
of lateral inhibition to orientation tuning and direction selectivity in cat
visual cortex: reversible inactivation of functionally characterized sites
combined with neuroanatomical tracing techniques. Eur. J. Neurosci., 10,
2056±2075.
Gabbott, P.L.A., Martin, K.A.C. & Whitteridge, D. (1987) Connections
between pyramidal neurons in layer 5 of cat visual cortex (Area 17). J.
Comp. Neurol., 259, 364±381.
Gilbert, C.D. & Wiesel, T.N. (1983) Clustered intrinsic connections in cat
visual cortex. J. Neurosci., 3, 1116±1133.
Gilbert, C.D. & Wiesel, T.N. (1989) Columnar speci®city of intrinsic
horizontal and corticocortical connections in cat visual cortex. J.
Neurosci., 9, 2432±2442.
Gilbert, C.D. & Wiesel, T.N. (1990) The in¯uence of contextual stimuli on the
orientation selectivity of cells in primary visual cortex of the cat. Vision
Res., 30, 1689±1701.
Hata, Y., Tsumoto, T., Sato, H. & Tamura, H. (1991) Horizontal interactions
between visual cortical neurones studied by cross-correlation analysis in the
cat. J. Physiol. (Lond.), 441, 593±614.
von der Heydt, R. & Peterhans, E. (1989) Mechanisms of contour perception
in monkey visual cortex. I. Lines of pattern discontinuity. J. Neurosci., 9,
1731±1748.
Hubel, D.H. & Wiesel, T.N. (1962) Receptive ®elds, binocular interaction and
functional architecture in the cat's visual cortex. J. Physiol. (Lond.), 160,
105±154.
Hubel, D.H. & Wiesel, T.N. (1963) Shape and arrangement of columns in cat's
striate cortex. J. Physiol. (Lond.), 165, 559±568.
Kapadia, M.K., Ito, M. & Gilbert, C.D. (1995) Improvement in visual
sensitivity by changes in local context: parallel studies in human observers
and V1 of alert monkeys. Neuron, 15, 843±856.
Kastner, S., Nothdurft, H.-C. & Pigarev, I.N. (1997) Neuronal correlates of
pop-out in cat striate cortex. Vision Res., 37, 371±376.
KisvaÂrday, Z.F. (1992) GABAergic networks of basket cells in the visual
cortex. In Mize, R.R., Marc, R. & Sillito, A.M. (eds), Progress in Brain
Research, Vol. 90, Mechanisms of GABA in the Visual System. Elsevier
Science, Amsterdam, pp. 385±405.
KisvaÂrday, Z.F. & Eysel, U.T. (1992) Cellular organization of reciprocal
patchy networks in layer III of cat visual cortex (area 17). Neuroscience, 46,
275±286.
KisvaÂrday, Z.F., ToÂth, EÂ., Rausch, M. & Eysel, U.T. (1997) Orientationspeci®c relationship between populations of excitatory and inhibitory lateral
connections in the visual cortex of the cat. Cerebral Cortex, 7, 605±618.
Lamme, V.A.F. (1995) The neurophysiology of ®gure-ground segregation in
primary visual cortex. J. Neurosci., 15, 1605±1615.
Malach, R., Amir, Y., Harel, M. & Grinvald, A. (1993) Relationship between
intrinsic connections and functional architecture revealed by optical
imaging and in vivo targeted biocytin injections in primate striate cortex.
Proc. Natl Acad. Sci. USA, 90, 10 469±10 473.
Malach, R., Tootell, R.B.H. & Malonek, D. (1994) Relationship between
orientation domains, cytochrome oxidase stripes, and intrinsic horizontal
connections in squirrel monkey area V2. Cerebral Cortex, 4, 151±165.
Martin, K.A.C. & Whitteridge, D. (1984) Form, function and intracortical
projections of spiny neurones in the striate visual cortex of the cat. J.
Physiol. (Lond.), 353, 463±504.
Matsubara, J.A., Cynader, M. & Swindale, N.V. (1987) Anatomical properties
and physiological correlates of the intrinsic connections in cat area 18. J.
Neurosci., 7, 1428±1446.
Matsubara, J.A., Cynader, M., Swindale, N.V. & Stryker, M.P. (1985) Intrinsic
projections with visual cortex: evidence for orientation speci®c local
connections. Proc. Natl Acad. Sci. USA, 82, 935±939.
Michalski, A., Gerstein, G.L., Czarkowska, J. & Tarnecki, R. (1983)
Interactions between cat striate cortex neurons. Exp. Brain Res., 51, 97±107.
Murphy, P.C. & Sillito, A.M. (1986) Continuity of orientation columns
between super®cial and deep laminae of the cat primary visual cortex. J.
Physiol. (Lond.), 381, 95±110.
Nelson, J.I. & Frost, B.J. (1985) Intracortical facilitation among co-oriented,
co-axially aligned simple cells in cat striate cortex. Exp. Brain Res., 61, 54±
61.
Nothdurft, H.-C. (1992) Feature analysis and the role of similarity in preattentive vision. Percept. Psychophys., 52, 355±375.
Nothdurft, H.-C. (1993) The role of features in preattentive vision: comparison
of orientation, motion and color cues. Vision Res., 33, 1937±1958.
Orban, G.A. (1984) Neuronal operations in the visual cortex. Springer-Verlag,
Berlin, pp. 1±367.
Rockland, K.S. & Lund, J.S. (1982) Widespread periodic intrinsic connections
in the tree shrew visual cortex. Science, 215, 1532±1534.
Shevelev, I.A., Lazareva, N.A., Novikova, B.V., Tikhomorov, A.S. & Sharaev,
G.A. (1994) Double orientation tuning in the cat visual cortex units.
Neuroscience, 61, 965±973.
Shevelev, I.A., Lazareva, N.A., Novikova, B.V., Tikhomorov, A.S. & Sharaev,
G.A. (1995) Sensitivity to cross-like ®gures in the cat striate neurons.
Neuroscience, 69, 51±57.
Sillito, A.M., Grieve, K.L., Jones, H.E., Cudeiro, J. & Davis, J. (1995) Visual
cortical mechanisms detecting focal orientation discontinuities. Nature, 378,
492±496.
Singer, W. & Gray, C.M. (1995) Visual feature integration and the temporal
correlation hypothesis. Annu. Rev. Neurosci., 18, 555±586.
Somogyi, P., KisvaÂrday, Z.F., Martin, K.A.C. & Whitteridge, D. (1983)
Synaptic connections of morphologically identi®ed and physiologically
characterized large basket cells in the striate cortex of the cat. Neuroscience,
10, 261±294.
Swindale, N.V., Matsubara, J.A. & Cynader, M.S. (1987) Surface organization
of orientation and direction selectivity in cat area 18. J. Neurosci., 7, 1414±
1427.
Tamura, H., Sato, H., Katsuyama, N., Hata, Y. & Tsumoto, T. (1996) Less
segregated processing of visual information in V2 than in V1 of the monkey
visual cortex. Eur. J. Neurosci., 8, 300±309.
Ts'o, D.Y., Gilbert, C.D. & Wiesel, T.N. (1986) Relationships between
horizontal interactions and functional architecture in cat striate cortex as
revealed by cross-correlation analysis. J. Neurosci., 6, 1160±1170.
Tusa, R.J., Rosenquist, A.C. & Palmer, L.A. (1979) Retinotopic organization
of areas 18 and 19 in the cat. J. Comp. Neurol., 185, 657±678.
WoÈrgoÈtter, F. & Eysel, U.T. (1987) Quantitative determination of orientational
and directional components in the response of visual cortical cells to
moving stimuli. Biol. Cybern., 57, 349±355.
Ó 1999 European Neuroscience Association, European Journal of Neuroscience, 11, 4291±4308