Experimental Brain Research
Exp. Brain Res, 37, 253-263 (1979)
@ Springer-Verlag 1979
Anisotropic Receptive Field Structure of Cat Horizontal Cells J. Molenaar 1 and W.A. van de Grind 1'2 1 PsychophysiologyLaboratory, Universityof Amsterdam, Jan SwammerdamInstitute, le Const. Huygensstr. 20, NL-1054 BW Amsterdam, The Netherlands 2 PhysiologischesInstitut der Freien Universit/itBerlin, Arnimallee22, D-i000 Berlin
Summary. Horizontal (H-)cells were recorded intracellularly in the retinae of optically intact feline eyes in vivo. A small light spot orbiting slowly around the receptive field centers was used to quantify the fine structure and diameter of the receptive fields. Receptive field diameters measured in this way were larger than those measured with centered spots of increasing diameter. All H-units studied showed clearly anisotropic receptive field structures. These results are summarized in polar plots representing the local response generating sites with their corresponding "response plus transport" latencies. It is shown that the anisotropic receptive field properties are not incompatible with the approximately homogenous spatial distribution of H-cell somata reported by W/issle and Rieman (1978) for the axonless type of horizontal cell. Finally it is concluded that each H-cell might be involved in many different locally specialized signal processing activities. words: Horizontal cells - Cat retina microcircuits - Local anisotropy Key
-
Local circuit neuron
-
Retinal
In the retina of the cat two morphologically distinct types of horizontal (H-) cells have been found. Cajal (1894) described H-cells with a long thin axon and Gallego (1964, 1971) discovered a second type of H-cell lacking an axon. Later, Fisher and Boycott (1974) called the Cajal-type and Gallego-type H-cells B-type and A-type, respectively, and we will abbreviate this as "HB- and HA"-cell. A detailed morphological description of these two horizontal cell types, supplemented with invaluable quantitative data on the patterns of terminal aggregations and on the cell density and dendritic field size as a function of retinal eccentricity has recently been presented by Boycott et al. (1978), W~issle et al. (1978a, b). The axonless HA-cells seem to form a syncytoid layer, with extensive gap junctions between their dendrites (Sobrino and Gallego, 1970; Gallego, 1976; Kolb, 1977). The HB-cell dendritic tree is rather homogeneously filled with its many fine dendritic processes, whereas the HA-cell dendritic tree is much coarser and
0014-4819/79/0037/0253/$ 2.20
254
J. Molenaar and W.A. van de Grind
inhomogeneously filled. The long thin axon of the HB-cell ends in a complex proliferating axon terminal system ( H B A T ) which appears to be functionally independent from the soma (Nelson et al., 1975). The dendrites of the H A - and HB-cells form the postsynaptic elements in the cone pedicles and do not contact the rods. On the other hand, the H B A T ' s only contact the rods. Nelson et al. (1975) showed that cones contribute some 2 0 % of the intracellularly recorded light response of the HBAT, despite the apparent lack of direct connections between the H B A T and the cones. Similarly, the rods are responsible for a substantial part ( 3 0 - 4 0 % ) of the light response of both the H A - and the HB-cell soma. These findings might be explained by the existence of gap junctions between rods and cones (Raviola and Gilula, 1973) or of contacts of cone basal filaments within the synaptic complex of rods (Gallego, 1971, 1976). Because of the strong rod dominance of H B A T ' s they probably respond slower to dynamic light modulations than either H A - or HB-cells. Thus, it seems safe to assume that the H B A T ' s are in fact identical with the narrow bandwidth H-units described by Foerster et al. (1977a, 1977b) and that they can be reliably identified in electrophysiological studies as H-units with a relatively low flicker fusion frequency (25-35 Hz under photopic conditions, op.cit.). From their morphological characteristics one might expect HB-cells to have reasonably isotropic receptive fields with sizes comparable to the size of their dendritic tree. The HA-cells form a syncytoid layer and might therefore be expected to have receptive field sizes exceeding the sizes of their individual dendritic trees. This suggests the possibility to discriminate between HA-cells, HB-cells and H B A T ' s during single unit recordings on the basis of a quantification of the size and anisotropy of the receptive field. To study this possibility we recorded intracellular responses from H-cells in the cat's retina to light spots of variable diameter moving with variable speed along a variable diameter circular path centered on the midpoint of the region of maximum sensitivity of the unit's receptive field. Our results show that almost all H-units have receptive field sizes that far exceed the size of the dendritic trees reported in the histological literature for H A as well as HB-cells. Interestingly, this also holds for most other H-cell receptive field sizes reported in the literature on the cat's retina (e.g., Brown and Murakami, 1968; Steinberg, 1969; Foerster et al., 1977a, 1977b; Nelson, 1977). It is conceivable (Gallego, 1976) that most, if not all, of these results were obtained from HA-cells only. Nevertheless, from our results as well as from those of Foerster et al. (1977a, 1977b) on functionally different H-units, we are more inclined to conclude that other types of H-cells also have RF-sizes in excess of the size of their dendritic trees. The most intriguing result of our present study is, however, that all the H-cell receptive fields showed clear anisotropies not only within the small territory corresponding to their dendritic tree, but also within the rest of their receptive fields up to the farthest boundaries. Even on the assumption that all our recordings have been from HA-cells, this result is remarkable and complicates the interpretation of the functional role of H-cells in vision.
Auisotropic RF Structure of Cat Horizontal Cells
255
Material and Methods Preparation The surgical preparation was performed under pentobarbital anesthesia (40 mg/kg i.p.). After additional local anesthesia of the infraorbital trigeminal nerve endings as well as all wounds and pressure points with Depot-Impletol, muscle relaxation was initiated with 80 mg gallamine triethiodide. A Loosco Amsterdam Infant Ventilator MK2 was used for artificial respiration and end-tidal COz was kept between 3.8 and 4.8% by adjusting respiratory volume. Temperature was kept between 37 and 39 ~ C. During the experiment muscle relaxation was maintained with a continuous infusion of a mixture of 20 mg gallamine triethiodide, 0.75 mg d-tubocurarine, and 2 ml 5 % glucose in Ringer solution per hour. Additional i.v. injections of 6-12 mg pentobarbital were given whenever necessary to maintain a relatively constant level of anesthesia. Except for the fact that we used a new type of stereotaxic frame and a newly developed technique for intraocular stereotaxic target localization (Molenaar and van de Grind, in prep.) the methods of preparation, refraction, care of eye optics, and recording were identical to those described by Foerster et al. (1977a, 1977b). In addition, refraction was checked regularly by estimating the diameter of ganglion cell receptive fields (range 0.3-i deg).
The Study of Receptive Field Anisotropies One can study the organization of visual receptive fields by turning a light spot on and off at various retinal positions and comparing the local responses. This classical sampling principle (Kuffler, 1953) has several disadvantages when large anisotropic receptive fields have to be characterized during brief intracellular recording periods. To resolve the local sensitivity changes one has to use relatively small spots stimulating sequentially many local regions (e.g., 100 or more). Since small spots evoke weak local responses, averaging of (e.g., 20 or more) responses at each sampling position will be necessary. Even if the required minimum number of local responses (e.g., 20 • 100) could be obtained during the brief intraeellular recording periods, this would not guarantee the detection of all local sensitivity differences in the studied region. (Unless some form of overlapping sampling is used, but that would require an even more prohibitive number of sampling positions). For these reasons we developed a different procedure using an orbiting light spot stimulus generated with the help of a device described in the companion paper (Molenaar and van de Grind, 1979). Figure 1 serves to illustrate the reasoning that led to the choice of this stimulation method. Suppose that most of the contacts between receptors and H-cells are found near the main dendrites of the H-cells. Then a local response maximum might be expected whenever the light spot covers part of the retinal area overlying such a main dendrite. For the example of a cell with three main dendrites sketched in Fig. 1, three local response maxima can therefore be expected when a light spot circulates about the cell soma. This expectation is indicated schematically in Fig. IB. Since cat retinal H-cells hyperpolarize in response to light, the expressions "response maximum" or "response peak" in this paper refer to the hyperpolarizing direction, which is downwards in all our illustrations. In Fig. 1B the abscissa represents the spot's angular position along the orbit, measured from the (arbitrarily chosen) 12 o'clock reference position in the direction of rotation. The spot's position a is expressed in percent of a full 360 ~ revolution. Thus, one revolution always represents the same length along the abscissa, regardless of the spot's angular velocity co. The response obtained during one complete orbit of the light spot is called one response period. Figure 1B shows one (hypothetical) response period for each of a series of angular velocity values. These individual response periods are shifted vertically in linear proportion to their corresponding co-value. In this way a very informative diagram is obtained and for ease of reference we will call it an "a-co diagram". Let us call the time delay between the moment of stimulation of a given sensitive locus and the arrival of its response at the electrode tip "c. Then the light spot will have moved cot during the time needed for the local response to reach the electrode. Thus, the positions of the local responses will shift away from the origin of the a-axis in proportion to oJ. A simple way to quantify these shifts and use them to measure ~ is proposed in Fig. lB. Lines connecting the peak positions along the c~-axes for different co-values - the "peak-position lines" - have slopes proportional to the values of 9 for the corresponding response loci. Furthermore, the peak-position lines can be
256
J. Molenaar and W. A. van de Grind a(
A
B 1" in m s :
50
1000
500
1 c~ i n % o f 0
20
one
40
rev 60
80
100
.0
o')
.1
\
.2
r~__/
:>
/
,3
.4
I. I
e%oo ,o
line
Fig. 1. A A light spot (symbolized as a black disc) is supposed to follow a circular path (the "orbit") centered on the microelectrode tip. The angular velocity 02 is expressed in rps and the spot's position is given with reference to the 12 o'clock position (a = 0) in terms of angle a. The hatched area represents a hypothetical anisotropic H-unit receptive field. B A simulated a-co diagram for the hypothetical H-unit of part (A). The upper row indicates the positions along the a-axis of the local response generation sites with their corresponding, arbitrarily chosen, delays ~. The H-unit responses are simulated for five values of co, viz., 0.1-0.5 rps
extrapolated to the (9 = 0 axis to obtain an estimate of the actual positions of the local response generating sites along the orbit: Fig. lB. If the latency ~, which includes generation and transport delay, depends on the angular velocity co of the spot, the peak-position lines will be curved. (If the luminance of the spot is constant, the stimulus intensity in terms of the number of photons per local response region per revolution decreases with increasing angular velocities and this could cause an increase of 9 with co). The curvature of the peak-position lines can be used to quantify the relation between co and ~. A set of or-co diagrams for different values of the orbit diameter enables one to deduce the position of the cell's main sensitive regions with their corresponding time delays, Such results can then be summarized most conveniently in the form of polar plots like those presented in Fig. 3 below. For small spots and small orbit diameters, the polar plots can be expected to correlate well
Anisotropic RF Structure of Cat Horizontal Cells
257
with the unit's main dendritic pattern. For larger orbit diameters the polar plots are presumed to show the distribution of the receptive fields or local response generation sites of those neighbouring H-cells connected with the recorded unit. In studying a large anisotropic receptive field (RF) it is difficult to define the position of its "center point". Obviously the exclusive use of small stimuli cannot lead to conclusive results. Perhaps it is best to view the "center" of an anisotropic RF as the region where the product of density and sensitivity of local response regions is highest. Accordingly, a relatively large spot covering many local response sites is better suited to localize anisotropic receptive fields. With this in mind, we used the following localization procedure. Sinusoidally flickering light spots (frequency 1 Hz, modulation depth 90%, average retinal illuminance of 100 cat photopic troland, diameters ranging from 15 to 0.3 degrees) were, one at a time and starting with the largest, positioned on the screen so as to give a maximum response. The search for the optimum position of ever smaller spots was restricted to the immediate vicinity of the center of the area previously stimulated by the optimally positioned larger spots. The resulting RF-localization for the smaller spots was accepted only if it proved to give optimum responses for the whole range of diameters used. The diameter of the spot that evoked a response of the order of 90 % of the maximum obtainable response was used as a rough estimate of the RF-diameter. The second step of our procedure consisted of the measurement of the unit's response to flickering spots in the frequency range from 1-100 Hz and for at least 2 different spot sizes. This was done to be able to relate our results to those of Foerster et al. (1977a, 1977b) who classified their recordings according to the response bandwidth in narrow bandwidth or Hn-units (photopic flicker fusion frequency or "FFF" of 2 5 4 0 Hz), medium bandwidth or Hm-units (photopic FFF 55-70 Hz) and wide bandwidth or Hw-units (photopic FFF 95-110 Hz). Finally, in the third step of the procedure the circular orbit of the moving light spot stimulus was centered on the midpoint of the RF. The first orbit diameter was usually made approximately equal to the estimated RF diameter. At least 20 response periods were recorded at each angular velocity and at least 5 different angular velocities were selected before the analysis for one orbit diameter was regarded to be complete. Materials
From the retinae of the left eyes of 10 cats we obtained 21 intracellular H-cell recordings of sufficient stability and duration (5 to 45 minutes) to analyze the receptive field anisotropies with the above methods. Only 4 H-units could be studied for more than 3 orbit diameter values, but the results from the other 17 stable recordings fully confirmed the general picture (see below) emerging from the results of the most extensively studied units.
Results T h e m a i n a d v a n t a g e o f t h e o r b i t i n g l i g h t s p o t m e t h o d is t h a t it a l l o w s a q u i c k s u r v e y o f a n i s o t r o p i e s . I f t h e r e s p o n s e to o n e c o m p l e t e r e v o l u t i o n o f t h e light s p o t s h o w s o n l y o n e p e a k , this m i g h t b e d u e to b a d c e n t e r i n g o f t h e o r b i t a n d if it shows two peaks the RF might be elongated rather than circular symmetric. H o w e v e r , as s o o n as m o r e t h a n 2 p e a k s a r e f o u n d p e r r e s p o n s e p e r i o d t h e R F m u s t b e s p a t i a l l y a n i s o t r o p i c . I n o u r s t u d y o n H - u n i t s all 21 r e c o r d e d units ( a n d also all t e s t e d b r i e f l y p e n e t r a t e d units) s h o w e d at l e a s t t h r e e p e a k s . T h i s is s t r o n g e v i d e n c e t h a t H - u n i t r e c e p t i v e fields a r e a n i s o t r o p i c . S i n c e t h e p r e s e n t s t u d y o f R F - a n i s o t r o p y was t h e first o f its k i n d w e c o u l d n o t y e t p r o f i t f r o m t h e p o t e n t i a l s p e e d ( s e e b e l o w ) o f t h e o r b i t i n g light s p o t m e t h o d . F i r s t o n e has to k n o w m o r e a b o u t t h e r e l a t i o n - if a n y - b e t w e e n co a n d and about the reproducibility of the single orbit responses. Therefore, we had to r e s o r t to t h e t i m e c o n s u m i n g v e r s i o n o f t h e m e t h o d o u t l i n e d a b o v e . F i g u r e 2
258
J. Molenaar and W.A. van de Grind
A | | degrees
1.7 rev/sec
8 degrees 0.4 r e v/see
2my
B
in % r e v
0 .0 .5
1.0
1.5
50
100
Fig. 2A and B. Intracellular H-unit responses measured with orbiting light spots. A illustrates the reproducibility of the gross characteristics as well as the fine structure of the individual responses to single spot revolutions (no averaging was used!). The angular velocities were 1.7 and 0.4 rps along orbits with a diameter of 11 and 8 deg, respectively. The downward direction indicates hyperpolarization. B shows part of an a-c0 diagram measured with a spot of 500 cat photopic trolands moving along an orbit with a diameter of 11 deg at angular velocities of 0.58, 0.85, and 1.11 rps. The dashed lines indicate how some of the more prominent response peaks are projected on the (z-axes to determine the peak positions (black dots). The oblique lines drawn through each of the four sets of peak positions are called the "peak position lines" and from their slopes one can calculate the value of the delay x for each of the local response generation sites as explained in the text. Horizontal bars represent 100 ms
presents some samples from an intracellular H-unit recording in an experiment where the spot diameter was 1.5 deg and the orbit diameter 11 deg. Since the area of the orbiting spot was relatively small compared to the total area of the H-cell's RF, the spot evoked a relatively small local hyperpolarizing response whenever it passed over a sensitive region. As illustrated in Fig. 2, the cell's response varied systematically and reproducibly with the position of the spot along its orbit as well as with the spot's angular velocity. The narrowest peaks in the responses in Fig. 2 have a width of about 5 % of a revolution and this corresponds well with the part of the orbit covered at any moment by the spot (4.5 %). For low angular velocities of the orbiting spot the amplitudes of the local hyperpolarizing response peaks can be roughly predicted with the help of amplitude-diameter plots such as those published by Foerster et al. (1977b, Fig. 6A). At increasing angular velocities, subsequent peaks tend to fuse and the response therefore becomes smoother. This smoothing phenomenon might be quantitatively predictable from the pulse response of the unit as measured (e.g., Steinberg, 1969) with flashes covering the whole RF, but this has not yet been attempted. In all e-co diagrams obtained from the present series of measurements the peak position lines proved to be straight. The deviations of the actual peak positions from the axis-crossings of the best fitting st{aight line were non-systematic and did not exceed 3 % of a complete revolution. This is an important result, showing that within the range of co-values used, does not depend on co. Such a result suggests that the use of two co-values (e.g.,
Anisotropic RF Structure of Cat Horizontal Cells
259
B
2 1 degree=
t
223 ,lain
Fig. 3A and B. Plots representing the fine structure of the RF of two H-units with different dynamic characteristics. Plot A was determined from c~-eodiagrams measured for a unit with a "medium bandwidth" (Hm-unit according to the criteria of Foerster et al., 1977a, 1977b). This unit had a flicker fusion frequency of 60 Hz. Plot B summarizes some results for a "narrow bandwidth unit" ("H~-unit", op. cit.), with a flicker fusion frequency of 30 Hz. The small circles have a diameter equal to that of the stimulating spot (1.5 deg) and are centered on the calculated positions of the local response generating sites. The numbers in the circles represent the corresponding delay times in ms as computed from the slopes of the peak position lines. The larger circles indicate the spot trajectories used in these experiments. The calculated projection of the microelectrode on the retina is included in these polar plots
0.5 rps and 2 rps) per orbit d i a m e t e r value might prove to be sufficient in future studies to estimate both ~ (from the slope of the p e a k position lines) and the positions of the local response regions (from linear extrapolation of the peak position lines to the e0 = 0 axis). A n o t h e r i m p o r t a n t finding, illustrated in Fig. 2 A was that the fine structure of the single orbit response is highly reproducible and in fact even the differences b e t w e e n single r e s p o n s e periods (1 orbit) and a v e r a g e d responses (20 orbits) were slight. Thus, it might prove sufficient in future studies to characterize the anisotropies at a given orbit diameter value f r o m the responses to two stimulus orbits completed, say, in 0.5 s (w = 2 rps) and 2 s (c0 = 0.5 rps), respectively. A u t o m a t i c control of the orbit g e n e r a t o r would then allow the study o f anisotropies along 20 circular paths in the R F within 1 min. This is an attractive aspect of the orbiting light spot method. Figure 3 presents typical results obtained in the way illustrated in Figs. 1 and 2B. T h e results in Fig. 3 A refer to a unit with a relatively small R F (-- 2,000 ~m d i a m e t e r on the retina) and a " m e d i u m " FFF, viz., 60 Hz. T h e results of Fig. 3B were o b t a i n e d f r o m a unit with a larger R F (--~ 3,000 ~m) and a lower F F F (30
260
J. Molenaar and W. A. van de Grind
Hz). These units were classified according to the criteria outlined by Foerster et al. (1977a, 1977b) as an Hm- and an Hn-unit, respectively. In partial confirmation of the results of the latter authors we found that all our H-unit recordings could be classified unambiguously as either narrow-bandwidth (FFF of 25-40 Hz) or medium-bandwidth (FFF of 55-70 Hz) units. However, we did not register any units with FFF's as high as those of the Hw-units of Foerster et al. (1977a, 1977b). The optical projections of the microelectrodes onto the retina calculated with the nodal point of the eye as the center of projection, are included in Fig. 3. This clearly demonstrates that the positions of the local response generating sites are not simply correlated with the electrode position. Thus, it seems highly unlikely that the measured response fluctuations are somehow caused by light reflections or shadows of the electrode. The blank areas near the electrode tips in Fig. 3 signify that no response fluctuations were found for small orbit diameters. This does not necessarily mean that the structure of the central RF region is isotropic, however. Most probably the light spots used were too large to reveal any fine structure in this central region. An intriguing aspect of the results presented in Fig. 3 is the rather irregular distribution of the delay times. We assume that this is a direct reflection of the underlying connectivity pattern. Generation sites with larger delay times are probably connected to the recording site via other more peripherally located H-units and/or via smaller dendrites or axon branches.
Discussion
As mentioned at the beginning of this paper, the H-unit RF-sizes far exceed the size of H-unit dendritic or axonal trees. Therefore, the anisotropy of the large H-unit receptive fields reported above reflects network rather than single unit properties. Wfissle and Rieman (1978) recently presented data on the spatial distribution of HA-cell somata. They interpreted their findings as suggestive of a reasonably isotropic network structure. However, in the absence of quantitative data on connectivity patterns a homogenous distribution of somata cannot be assumed to imply isotropic network properties. We illustrate this point with the help of a mosaic of HA-perikarya published by W/issle and Rieman (1978, their Fig. 8a). The formulation of realistic connectivity rules is a major problem, but since we only require an "existence proof", a simple rule might do. Therefore, we applied the simplest of all rules, viz., that all neighbours in the mosaic nearer than a fixed distance (here 80 ~tm) from each other are interconnected. The result is shown in Fig. 4 and it illustrates the following general rule: Even in a network of homogeneously distributed cells, simple, relatively indifferent, connectivity rules can lead to anisotropic RF-properties of the units in the network (e.g., the one marked with an arrow in Fig. 4). Simulation studies for networks of this type (Fig. 4) based on the assumptions (1) that network segments can be viewed as leaky cables (half length of 150-200 ~tm), (2) that the nodes (cells) have a fixed attenuation factor, and (3) that segment delay is constant (50-100 ~s/~m) led to results (to be presented elsewhere) that were
Anisotropic RF Structure of Cat Horizontal Cells Fig. 4. A simple connectivity rule applied to a mosaic of HA-perikarya (dots) published by Wfissle and Rieman (1978, their Fig. 8a) leads to the presented HA-syncytium. The rule was that all pairs of perikarya separated less than 80 p~were interconnected. The arrow indicates a unit for which the response was simulated (see the text). The thick lines indicate the main pathways leading from the area under the light spot's orbital path to the studied unit
261
1.5
1.0 E E
-~ .5
.0
I
.0
.5
|
1.0 X in
i
!
1.5
2.0
mm
qualitatively very similar to the experimental results, especially as far as the n u m b e r and amplitude distribution of the peaks were concerned. Quantitative agreement would require realistic connectivity rules, which have not yet been formulated. It seems reasonable (see also Foerster et al., 1977a, 1977b) to assume that the results of Fig. 3A and 3B are obtained from morphologically different structures such as HA-cell and a H B A T , respectively. However, in that case the above ideas about network anisotropy would only hold for the results of Fig. 3B if the H B A T ' s would prove to be interconnected as well. Similarly, one can only safely ascribe results like those of Fig. 3A to HA-cells after it has been proved that HB-cells are not extensively interconnected. In general, however, the large RF-size of H-units in most species studied so far seems to imply that extensive interconnectivity is much more universal than has been hitherto reported. It would be interesting to learn more about these possibilities from future histological work. Our present results even suggest larger H-unit RF-sizes than estimates from area-response functions indicate. Local response generating sites with a more peripheral position in the R F can hardly contribute to the light response, when the H-unit is already hyperpolarized to near saturation values by a large spot also covering all of the unit's more centrally located local response generating sites. With single, relatively small (1~176 orbiting light spots we even found clear local responses for the largest orbit diameters tried (25~ The present results, even though they are of a preliminary nature, throw an interesting new light on the possible function of H-units in vision. For example, even though the RF-sizes of these local circuit neurons (Rakic, 1975; Schmitt et al., 1976) are large, the H-units can respond to relatively small spots moving along specific tracks within their R F ' s and, judging from our latency data (Fig. 3), there might even be regions in the R F ' s " t u n e d " to specific velocities and/or directions of movement. Such local specializations certainly deserve further study. All this suggests that one H-unit could be involved in several, possibly different, local information processing activities simultaneously. The advantage
262
J. Molenaar and W.A. van de Grind
o f s u c h a n o r g a n i z a t i o n w o u l d b e a n e c o n o m y in t e r m s o f t h e n e c e s s a r y n u m b e r o f cells, b u t a d i s a d v a n t a g e w o u l d s e e m to b e t h e p o s s i b i l i t y o f r e l a t i v e l y s t r o n g interference between different local processes. Recent morphological studies of t h e o u t e r p l e x i f o r m l a y e r o f t h e r a b b i t r e t i n a b y S j 6 s t r a n d ( 1 9 7 8 , 1 9 7 6 ) also p o i n t in t h e d i r e c t i o n o f l o c a l s p e c i a l i z a t i o n s , a p t l y c a l l e d " m i c r o c i r c u i t s " by Sj6strand. We think that the further analysis of the possible functional equivalents of these morphological "microcircuits" of local circuit neurons d e s e r v e s m u c h m o r e a t t e n t i o n t h a n it has r e c e i v e d u n t i l n o w .
Acknowledgements. This work was supported by the Netherlands Organization for the Advancement of Pure Research (Z.W.O.). We gratefully acknowledge the help of A.W. Schreurs with mechanical, R. Voorhorst, J. Nivard, J. Janse with electronic, R. Swart and Mrs. A. Thiele (Berlin) with histological problems. Furthermore, we thank H. de Ridder for his help in analyzing part of the data and Mrs. W. van der Pol for her accurate typework.
References Brown, K.T., Murakami, M.: Rapid effects of dark and light adaptation upon the receptive field organization of S-potentials and late receptor potentials. Vision Res. 8, 1145-1171 (1968) Cajal, S. R. y : Die Retina der Wirbeltiere. Wiesbaden: Bergmann 1894 Boycott, B.B., Peichl, L., W/issle, H.: Morphological types of horizontal cell in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203,229-245 (1978) Fisher, S.K., Boycott, B.B.: Synaptic connections made by horizontal cells within the outer plexiform layer of the cat and the rabbit. Proc. R. Soc. (Lond.) B 186, 317-331 (1974) Foerster, M.H., Grind, W. A. van de, Griisser, O.-J.: Frequency transfer properties of three distinct types of cat horizontal cells. Exp. Brain Res. 29, 347-366 (1977a) Foerster, M. H., Grind, W.A. van de, Grtisser, O.-J.: The response of cat horizontal cells to flicker stimuli of different area, intensity and frequency. Exp. Brain Res. 29, 367-385 (1977b) Gallego, A.: Description d'une nouvelle couche cellulaire dans la r6tine des mammif6res et son r61e functionnel possible. Bull. Ass. Anat. 49, 624-631 (1964) Gallego, A.: Horizontal and amacrine cells in the mammal's retina. Vision Res. Suppl. 3, 33-50 (1971) Gallego, A.: Comparative study of the horizontal ceils in the vertebrate retina: Mammals and birds. In: Neural principles in vision. F. Zettler, R. Weiler (eds.), pp. 26-62. Berlin, Heidelberg, New York: Springer 1976 Kolb, H.: The organization of the outer plexiform layer in the retina of the cat. Electron microscopic observations. J. Neurocytol. 6, 131-153 (1977) Kuffler, S.W.: Discharge patterns and functional organization of the mammalian retina. J. Neurophysiol. 17, 558-574 (1953) Molenaar, J., Grind, W.A. van de: A visual stimulator for orbiting light spots and equal energy annuli with continuously variable parameters. Exp. Brain Res. 37, 65-71 (1979) Nelson, R.: Cat cones have rod input: A comparison of the response properties of cones and horizontal cell bodies in the retina of the cat. J. Comp. Neurol. 172, 109-135 (1977) Nelson, R., Liitzow, A. von, Kolb, H., Gouras, P.: Horizontal cells in the cat retina with independent dendritic systems. Science 189, 137-139 (1975) Rakic, P.: Local circuit neurons. Neurosci. Res. Progr. Bull. 13 (1975) Raviola, E., Gilula, N.B.: Gap junctions between photoreceptor cells in the vertebrate retina. Proc. Nat. Acad. Sci. USA 70, 1677-1681 (1973) Schmitt, F.O., Dev, P., Smith, B.H.: Electrotonic processing of information by brain cells. Science 193, 114-120 (1976) Sj6strand, F.S.: The outer plexiform layer of the rabbit retina, an important data processing center. Vision Res. 16, 1-14 (1976)
Anisotropic RF Structure of Cat Horizontal Cells
263
Sj6strand, F.S.: Circuitry analysis of the outer plexiform layer in the rabbit retina (I). J. Ultrastruct. Res. 62, 54-81 (1978) Sobrina, J.A., Gallego, A.: C61ulas amacrinas de la capa plexiforme de la retina. Actas Soc. Esp. Cienc. Fysiol. 12, 373-375 (1970) Steinberg, R.H.: Rod and cone contributions to S-potentials from the cat retina. Vision Res. 9, 1319-1329 (1969) W~issle, H., Rieman, H.J.: The mosaic of nerve cells in the mammalian retina. Proc. R. Soc. (Lond.) B 200, 441-461 (1978) W/issle, H., Boycott, B.B., Peichl, L.: Receptor contacts of horizontal cells in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203, 247-267 (1978a) W~issle, H., Peichl, L., Boycott, B.B.: Topography of horizontal ceils in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203, 269-291 (1978b)
Received February 5, 1979
Exp. Brain Res, 37, 253-263 (1979)
@ Springer-Verlag 1979
Anisotropic Receptive Field Structure of Cat Horizontal Cells J. Molenaar 1 and W.A. van de Grind 1'2 1 PsychophysiologyLaboratory, Universityof Amsterdam, Jan SwammerdamInstitute, le Const. Huygensstr. 20, NL-1054 BW Amsterdam, The Netherlands 2 PhysiologischesInstitut der Freien Universit/itBerlin, Arnimallee22, D-i000 Berlin
Summary. Horizontal (H-)cells were recorded intracellularly in the retinae of optically intact feline eyes in vivo. A small light spot orbiting slowly around the receptive field centers was used to quantify the fine structure and diameter of the receptive fields. Receptive field diameters measured in this way were larger than those measured with centered spots of increasing diameter. All H-units studied showed clearly anisotropic receptive field structures. These results are summarized in polar plots representing the local response generating sites with their corresponding "response plus transport" latencies. It is shown that the anisotropic receptive field properties are not incompatible with the approximately homogenous spatial distribution of H-cell somata reported by W/issle and Rieman (1978) for the axonless type of horizontal cell. Finally it is concluded that each H-cell might be involved in many different locally specialized signal processing activities. words: Horizontal cells - Cat retina microcircuits - Local anisotropy Key
-
Local circuit neuron
-
Retinal
In the retina of the cat two morphologically distinct types of horizontal (H-) cells have been found. Cajal (1894) described H-cells with a long thin axon and Gallego (1964, 1971) discovered a second type of H-cell lacking an axon. Later, Fisher and Boycott (1974) called the Cajal-type and Gallego-type H-cells B-type and A-type, respectively, and we will abbreviate this as "HB- and HA"-cell. A detailed morphological description of these two horizontal cell types, supplemented with invaluable quantitative data on the patterns of terminal aggregations and on the cell density and dendritic field size as a function of retinal eccentricity has recently been presented by Boycott et al. (1978), W~issle et al. (1978a, b). The axonless HA-cells seem to form a syncytoid layer, with extensive gap junctions between their dendrites (Sobrino and Gallego, 1970; Gallego, 1976; Kolb, 1977). The HB-cell dendritic tree is rather homogeneously filled with its many fine dendritic processes, whereas the HA-cell dendritic tree is much coarser and
0014-4819/79/0037/0253/$ 2.20
254
J. Molenaar and W.A. van de Grind
inhomogeneously filled. The long thin axon of the HB-cell ends in a complex proliferating axon terminal system ( H B A T ) which appears to be functionally independent from the soma (Nelson et al., 1975). The dendrites of the H A - and HB-cells form the postsynaptic elements in the cone pedicles and do not contact the rods. On the other hand, the H B A T ' s only contact the rods. Nelson et al. (1975) showed that cones contribute some 2 0 % of the intracellularly recorded light response of the HBAT, despite the apparent lack of direct connections between the H B A T and the cones. Similarly, the rods are responsible for a substantial part ( 3 0 - 4 0 % ) of the light response of both the H A - and the HB-cell soma. These findings might be explained by the existence of gap junctions between rods and cones (Raviola and Gilula, 1973) or of contacts of cone basal filaments within the synaptic complex of rods (Gallego, 1971, 1976). Because of the strong rod dominance of H B A T ' s they probably respond slower to dynamic light modulations than either H A - or HB-cells. Thus, it seems safe to assume that the H B A T ' s are in fact identical with the narrow bandwidth H-units described by Foerster et al. (1977a, 1977b) and that they can be reliably identified in electrophysiological studies as H-units with a relatively low flicker fusion frequency (25-35 Hz under photopic conditions, op.cit.). From their morphological characteristics one might expect HB-cells to have reasonably isotropic receptive fields with sizes comparable to the size of their dendritic tree. The HA-cells form a syncytoid layer and might therefore be expected to have receptive field sizes exceeding the sizes of their individual dendritic trees. This suggests the possibility to discriminate between HA-cells, HB-cells and H B A T ' s during single unit recordings on the basis of a quantification of the size and anisotropy of the receptive field. To study this possibility we recorded intracellular responses from H-cells in the cat's retina to light spots of variable diameter moving with variable speed along a variable diameter circular path centered on the midpoint of the region of maximum sensitivity of the unit's receptive field. Our results show that almost all H-units have receptive field sizes that far exceed the size of the dendritic trees reported in the histological literature for H A as well as HB-cells. Interestingly, this also holds for most other H-cell receptive field sizes reported in the literature on the cat's retina (e.g., Brown and Murakami, 1968; Steinberg, 1969; Foerster et al., 1977a, 1977b; Nelson, 1977). It is conceivable (Gallego, 1976) that most, if not all, of these results were obtained from HA-cells only. Nevertheless, from our results as well as from those of Foerster et al. (1977a, 1977b) on functionally different H-units, we are more inclined to conclude that other types of H-cells also have RF-sizes in excess of the size of their dendritic trees. The most intriguing result of our present study is, however, that all the H-cell receptive fields showed clear anisotropies not only within the small territory corresponding to their dendritic tree, but also within the rest of their receptive fields up to the farthest boundaries. Even on the assumption that all our recordings have been from HA-cells, this result is remarkable and complicates the interpretation of the functional role of H-cells in vision.
Auisotropic RF Structure of Cat Horizontal Cells
255
Material and Methods Preparation The surgical preparation was performed under pentobarbital anesthesia (40 mg/kg i.p.). After additional local anesthesia of the infraorbital trigeminal nerve endings as well as all wounds and pressure points with Depot-Impletol, muscle relaxation was initiated with 80 mg gallamine triethiodide. A Loosco Amsterdam Infant Ventilator MK2 was used for artificial respiration and end-tidal COz was kept between 3.8 and 4.8% by adjusting respiratory volume. Temperature was kept between 37 and 39 ~ C. During the experiment muscle relaxation was maintained with a continuous infusion of a mixture of 20 mg gallamine triethiodide, 0.75 mg d-tubocurarine, and 2 ml 5 % glucose in Ringer solution per hour. Additional i.v. injections of 6-12 mg pentobarbital were given whenever necessary to maintain a relatively constant level of anesthesia. Except for the fact that we used a new type of stereotaxic frame and a newly developed technique for intraocular stereotaxic target localization (Molenaar and van de Grind, in prep.) the methods of preparation, refraction, care of eye optics, and recording were identical to those described by Foerster et al. (1977a, 1977b). In addition, refraction was checked regularly by estimating the diameter of ganglion cell receptive fields (range 0.3-i deg).
The Study of Receptive Field Anisotropies One can study the organization of visual receptive fields by turning a light spot on and off at various retinal positions and comparing the local responses. This classical sampling principle (Kuffler, 1953) has several disadvantages when large anisotropic receptive fields have to be characterized during brief intracellular recording periods. To resolve the local sensitivity changes one has to use relatively small spots stimulating sequentially many local regions (e.g., 100 or more). Since small spots evoke weak local responses, averaging of (e.g., 20 or more) responses at each sampling position will be necessary. Even if the required minimum number of local responses (e.g., 20 • 100) could be obtained during the brief intraeellular recording periods, this would not guarantee the detection of all local sensitivity differences in the studied region. (Unless some form of overlapping sampling is used, but that would require an even more prohibitive number of sampling positions). For these reasons we developed a different procedure using an orbiting light spot stimulus generated with the help of a device described in the companion paper (Molenaar and van de Grind, 1979). Figure 1 serves to illustrate the reasoning that led to the choice of this stimulation method. Suppose that most of the contacts between receptors and H-cells are found near the main dendrites of the H-cells. Then a local response maximum might be expected whenever the light spot covers part of the retinal area overlying such a main dendrite. For the example of a cell with three main dendrites sketched in Fig. 1, three local response maxima can therefore be expected when a light spot circulates about the cell soma. This expectation is indicated schematically in Fig. IB. Since cat retinal H-cells hyperpolarize in response to light, the expressions "response maximum" or "response peak" in this paper refer to the hyperpolarizing direction, which is downwards in all our illustrations. In Fig. 1B the abscissa represents the spot's angular position along the orbit, measured from the (arbitrarily chosen) 12 o'clock reference position in the direction of rotation. The spot's position a is expressed in percent of a full 360 ~ revolution. Thus, one revolution always represents the same length along the abscissa, regardless of the spot's angular velocity co. The response obtained during one complete orbit of the light spot is called one response period. Figure 1B shows one (hypothetical) response period for each of a series of angular velocity values. These individual response periods are shifted vertically in linear proportion to their corresponding co-value. In this way a very informative diagram is obtained and for ease of reference we will call it an "a-co diagram". Let us call the time delay between the moment of stimulation of a given sensitive locus and the arrival of its response at the electrode tip "c. Then the light spot will have moved cot during the time needed for the local response to reach the electrode. Thus, the positions of the local responses will shift away from the origin of the a-axis in proportion to oJ. A simple way to quantify these shifts and use them to measure ~ is proposed in Fig. lB. Lines connecting the peak positions along the c~-axes for different co-values - the "peak-position lines" - have slopes proportional to the values of 9 for the corresponding response loci. Furthermore, the peak-position lines can be
256
J. Molenaar and W. A. van de Grind a(
A
B 1" in m s :
50
1000
500
1 c~ i n % o f 0
20
one
40
rev 60
80
100
.0
o')
.1
\
.2
r~__/
:>
/
,3
.4
I. I
e%oo ,o
line
Fig. 1. A A light spot (symbolized as a black disc) is supposed to follow a circular path (the "orbit") centered on the microelectrode tip. The angular velocity 02 is expressed in rps and the spot's position is given with reference to the 12 o'clock position (a = 0) in terms of angle a. The hatched area represents a hypothetical anisotropic H-unit receptive field. B A simulated a-co diagram for the hypothetical H-unit of part (A). The upper row indicates the positions along the a-axis of the local response generation sites with their corresponding, arbitrarily chosen, delays ~. The H-unit responses are simulated for five values of co, viz., 0.1-0.5 rps
extrapolated to the (9 = 0 axis to obtain an estimate of the actual positions of the local response generating sites along the orbit: Fig. lB. If the latency ~, which includes generation and transport delay, depends on the angular velocity co of the spot, the peak-position lines will be curved. (If the luminance of the spot is constant, the stimulus intensity in terms of the number of photons per local response region per revolution decreases with increasing angular velocities and this could cause an increase of 9 with co). The curvature of the peak-position lines can be used to quantify the relation between co and ~. A set of or-co diagrams for different values of the orbit diameter enables one to deduce the position of the cell's main sensitive regions with their corresponding time delays, Such results can then be summarized most conveniently in the form of polar plots like those presented in Fig. 3 below. For small spots and small orbit diameters, the polar plots can be expected to correlate well
Anisotropic RF Structure of Cat Horizontal Cells
257
with the unit's main dendritic pattern. For larger orbit diameters the polar plots are presumed to show the distribution of the receptive fields or local response generation sites of those neighbouring H-cells connected with the recorded unit. In studying a large anisotropic receptive field (RF) it is difficult to define the position of its "center point". Obviously the exclusive use of small stimuli cannot lead to conclusive results. Perhaps it is best to view the "center" of an anisotropic RF as the region where the product of density and sensitivity of local response regions is highest. Accordingly, a relatively large spot covering many local response sites is better suited to localize anisotropic receptive fields. With this in mind, we used the following localization procedure. Sinusoidally flickering light spots (frequency 1 Hz, modulation depth 90%, average retinal illuminance of 100 cat photopic troland, diameters ranging from 15 to 0.3 degrees) were, one at a time and starting with the largest, positioned on the screen so as to give a maximum response. The search for the optimum position of ever smaller spots was restricted to the immediate vicinity of the center of the area previously stimulated by the optimally positioned larger spots. The resulting RF-localization for the smaller spots was accepted only if it proved to give optimum responses for the whole range of diameters used. The diameter of the spot that evoked a response of the order of 90 % of the maximum obtainable response was used as a rough estimate of the RF-diameter. The second step of our procedure consisted of the measurement of the unit's response to flickering spots in the frequency range from 1-100 Hz and for at least 2 different spot sizes. This was done to be able to relate our results to those of Foerster et al. (1977a, 1977b) who classified their recordings according to the response bandwidth in narrow bandwidth or Hn-units (photopic flicker fusion frequency or "FFF" of 2 5 4 0 Hz), medium bandwidth or Hm-units (photopic FFF 55-70 Hz) and wide bandwidth or Hw-units (photopic FFF 95-110 Hz). Finally, in the third step of the procedure the circular orbit of the moving light spot stimulus was centered on the midpoint of the RF. The first orbit diameter was usually made approximately equal to the estimated RF diameter. At least 20 response periods were recorded at each angular velocity and at least 5 different angular velocities were selected before the analysis for one orbit diameter was regarded to be complete. Materials
From the retinae of the left eyes of 10 cats we obtained 21 intracellular H-cell recordings of sufficient stability and duration (5 to 45 minutes) to analyze the receptive field anisotropies with the above methods. Only 4 H-units could be studied for more than 3 orbit diameter values, but the results from the other 17 stable recordings fully confirmed the general picture (see below) emerging from the results of the most extensively studied units.
Results T h e m a i n a d v a n t a g e o f t h e o r b i t i n g l i g h t s p o t m e t h o d is t h a t it a l l o w s a q u i c k s u r v e y o f a n i s o t r o p i e s . I f t h e r e s p o n s e to o n e c o m p l e t e r e v o l u t i o n o f t h e light s p o t s h o w s o n l y o n e p e a k , this m i g h t b e d u e to b a d c e n t e r i n g o f t h e o r b i t a n d if it shows two peaks the RF might be elongated rather than circular symmetric. H o w e v e r , as s o o n as m o r e t h a n 2 p e a k s a r e f o u n d p e r r e s p o n s e p e r i o d t h e R F m u s t b e s p a t i a l l y a n i s o t r o p i c . I n o u r s t u d y o n H - u n i t s all 21 r e c o r d e d units ( a n d also all t e s t e d b r i e f l y p e n e t r a t e d units) s h o w e d at l e a s t t h r e e p e a k s . T h i s is s t r o n g e v i d e n c e t h a t H - u n i t r e c e p t i v e fields a r e a n i s o t r o p i c . S i n c e t h e p r e s e n t s t u d y o f R F - a n i s o t r o p y was t h e first o f its k i n d w e c o u l d n o t y e t p r o f i t f r o m t h e p o t e n t i a l s p e e d ( s e e b e l o w ) o f t h e o r b i t i n g light s p o t m e t h o d . F i r s t o n e has to k n o w m o r e a b o u t t h e r e l a t i o n - if a n y - b e t w e e n co a n d and about the reproducibility of the single orbit responses. Therefore, we had to r e s o r t to t h e t i m e c o n s u m i n g v e r s i o n o f t h e m e t h o d o u t l i n e d a b o v e . F i g u r e 2
258
J. Molenaar and W.A. van de Grind
A | | degrees
1.7 rev/sec
8 degrees 0.4 r e v/see
2my
B
in % r e v
0 .0 .5
1.0
1.5
50
100
Fig. 2A and B. Intracellular H-unit responses measured with orbiting light spots. A illustrates the reproducibility of the gross characteristics as well as the fine structure of the individual responses to single spot revolutions (no averaging was used!). The angular velocities were 1.7 and 0.4 rps along orbits with a diameter of 11 and 8 deg, respectively. The downward direction indicates hyperpolarization. B shows part of an a-c0 diagram measured with a spot of 500 cat photopic trolands moving along an orbit with a diameter of 11 deg at angular velocities of 0.58, 0.85, and 1.11 rps. The dashed lines indicate how some of the more prominent response peaks are projected on the (z-axes to determine the peak positions (black dots). The oblique lines drawn through each of the four sets of peak positions are called the "peak position lines" and from their slopes one can calculate the value of the delay x for each of the local response generation sites as explained in the text. Horizontal bars represent 100 ms
presents some samples from an intracellular H-unit recording in an experiment where the spot diameter was 1.5 deg and the orbit diameter 11 deg. Since the area of the orbiting spot was relatively small compared to the total area of the H-cell's RF, the spot evoked a relatively small local hyperpolarizing response whenever it passed over a sensitive region. As illustrated in Fig. 2, the cell's response varied systematically and reproducibly with the position of the spot along its orbit as well as with the spot's angular velocity. The narrowest peaks in the responses in Fig. 2 have a width of about 5 % of a revolution and this corresponds well with the part of the orbit covered at any moment by the spot (4.5 %). For low angular velocities of the orbiting spot the amplitudes of the local hyperpolarizing response peaks can be roughly predicted with the help of amplitude-diameter plots such as those published by Foerster et al. (1977b, Fig. 6A). At increasing angular velocities, subsequent peaks tend to fuse and the response therefore becomes smoother. This smoothing phenomenon might be quantitatively predictable from the pulse response of the unit as measured (e.g., Steinberg, 1969) with flashes covering the whole RF, but this has not yet been attempted. In all e-co diagrams obtained from the present series of measurements the peak position lines proved to be straight. The deviations of the actual peak positions from the axis-crossings of the best fitting st{aight line were non-systematic and did not exceed 3 % of a complete revolution. This is an important result, showing that within the range of co-values used, does not depend on co. Such a result suggests that the use of two co-values (e.g.,
Anisotropic RF Structure of Cat Horizontal Cells
259
B
2 1 degree=
t
223 ,lain
Fig. 3A and B. Plots representing the fine structure of the RF of two H-units with different dynamic characteristics. Plot A was determined from c~-eodiagrams measured for a unit with a "medium bandwidth" (Hm-unit according to the criteria of Foerster et al., 1977a, 1977b). This unit had a flicker fusion frequency of 60 Hz. Plot B summarizes some results for a "narrow bandwidth unit" ("H~-unit", op. cit.), with a flicker fusion frequency of 30 Hz. The small circles have a diameter equal to that of the stimulating spot (1.5 deg) and are centered on the calculated positions of the local response generating sites. The numbers in the circles represent the corresponding delay times in ms as computed from the slopes of the peak position lines. The larger circles indicate the spot trajectories used in these experiments. The calculated projection of the microelectrode on the retina is included in these polar plots
0.5 rps and 2 rps) per orbit d i a m e t e r value might prove to be sufficient in future studies to estimate both ~ (from the slope of the p e a k position lines) and the positions of the local response regions (from linear extrapolation of the peak position lines to the e0 = 0 axis). A n o t h e r i m p o r t a n t finding, illustrated in Fig. 2 A was that the fine structure of the single orbit response is highly reproducible and in fact even the differences b e t w e e n single r e s p o n s e periods (1 orbit) and a v e r a g e d responses (20 orbits) were slight. Thus, it might prove sufficient in future studies to characterize the anisotropies at a given orbit diameter value f r o m the responses to two stimulus orbits completed, say, in 0.5 s (w = 2 rps) and 2 s (c0 = 0.5 rps), respectively. A u t o m a t i c control of the orbit g e n e r a t o r would then allow the study o f anisotropies along 20 circular paths in the R F within 1 min. This is an attractive aspect of the orbiting light spot method. Figure 3 presents typical results obtained in the way illustrated in Figs. 1 and 2B. T h e results in Fig. 3 A refer to a unit with a relatively small R F (-- 2,000 ~m d i a m e t e r on the retina) and a " m e d i u m " FFF, viz., 60 Hz. T h e results of Fig. 3B were o b t a i n e d f r o m a unit with a larger R F (--~ 3,000 ~m) and a lower F F F (30
260
J. Molenaar and W. A. van de Grind
Hz). These units were classified according to the criteria outlined by Foerster et al. (1977a, 1977b) as an Hm- and an Hn-unit, respectively. In partial confirmation of the results of the latter authors we found that all our H-unit recordings could be classified unambiguously as either narrow-bandwidth (FFF of 25-40 Hz) or medium-bandwidth (FFF of 55-70 Hz) units. However, we did not register any units with FFF's as high as those of the Hw-units of Foerster et al. (1977a, 1977b). The optical projections of the microelectrodes onto the retina calculated with the nodal point of the eye as the center of projection, are included in Fig. 3. This clearly demonstrates that the positions of the local response generating sites are not simply correlated with the electrode position. Thus, it seems highly unlikely that the measured response fluctuations are somehow caused by light reflections or shadows of the electrode. The blank areas near the electrode tips in Fig. 3 signify that no response fluctuations were found for small orbit diameters. This does not necessarily mean that the structure of the central RF region is isotropic, however. Most probably the light spots used were too large to reveal any fine structure in this central region. An intriguing aspect of the results presented in Fig. 3 is the rather irregular distribution of the delay times. We assume that this is a direct reflection of the underlying connectivity pattern. Generation sites with larger delay times are probably connected to the recording site via other more peripherally located H-units and/or via smaller dendrites or axon branches.
Discussion
As mentioned at the beginning of this paper, the H-unit RF-sizes far exceed the size of H-unit dendritic or axonal trees. Therefore, the anisotropy of the large H-unit receptive fields reported above reflects network rather than single unit properties. Wfissle and Rieman (1978) recently presented data on the spatial distribution of HA-cell somata. They interpreted their findings as suggestive of a reasonably isotropic network structure. However, in the absence of quantitative data on connectivity patterns a homogenous distribution of somata cannot be assumed to imply isotropic network properties. We illustrate this point with the help of a mosaic of HA-perikarya published by W/issle and Rieman (1978, their Fig. 8a). The formulation of realistic connectivity rules is a major problem, but since we only require an "existence proof", a simple rule might do. Therefore, we applied the simplest of all rules, viz., that all neighbours in the mosaic nearer than a fixed distance (here 80 ~tm) from each other are interconnected. The result is shown in Fig. 4 and it illustrates the following general rule: Even in a network of homogeneously distributed cells, simple, relatively indifferent, connectivity rules can lead to anisotropic RF-properties of the units in the network (e.g., the one marked with an arrow in Fig. 4). Simulation studies for networks of this type (Fig. 4) based on the assumptions (1) that network segments can be viewed as leaky cables (half length of 150-200 ~tm), (2) that the nodes (cells) have a fixed attenuation factor, and (3) that segment delay is constant (50-100 ~s/~m) led to results (to be presented elsewhere) that were
Anisotropic RF Structure of Cat Horizontal Cells Fig. 4. A simple connectivity rule applied to a mosaic of HA-perikarya (dots) published by Wfissle and Rieman (1978, their Fig. 8a) leads to the presented HA-syncytium. The rule was that all pairs of perikarya separated less than 80 p~were interconnected. The arrow indicates a unit for which the response was simulated (see the text). The thick lines indicate the main pathways leading from the area under the light spot's orbital path to the studied unit
261
1.5
1.0 E E
-~ .5
.0
I
.0
.5
|
1.0 X in
i
!
1.5
2.0
mm
qualitatively very similar to the experimental results, especially as far as the n u m b e r and amplitude distribution of the peaks were concerned. Quantitative agreement would require realistic connectivity rules, which have not yet been formulated. It seems reasonable (see also Foerster et al., 1977a, 1977b) to assume that the results of Fig. 3A and 3B are obtained from morphologically different structures such as HA-cell and a H B A T , respectively. However, in that case the above ideas about network anisotropy would only hold for the results of Fig. 3B if the H B A T ' s would prove to be interconnected as well. Similarly, one can only safely ascribe results like those of Fig. 3A to HA-cells after it has been proved that HB-cells are not extensively interconnected. In general, however, the large RF-size of H-units in most species studied so far seems to imply that extensive interconnectivity is much more universal than has been hitherto reported. It would be interesting to learn more about these possibilities from future histological work. Our present results even suggest larger H-unit RF-sizes than estimates from area-response functions indicate. Local response generating sites with a more peripheral position in the R F can hardly contribute to the light response, when the H-unit is already hyperpolarized to near saturation values by a large spot also covering all of the unit's more centrally located local response generating sites. With single, relatively small (1~176 orbiting light spots we even found clear local responses for the largest orbit diameters tried (25~ The present results, even though they are of a preliminary nature, throw an interesting new light on the possible function of H-units in vision. For example, even though the RF-sizes of these local circuit neurons (Rakic, 1975; Schmitt et al., 1976) are large, the H-units can respond to relatively small spots moving along specific tracks within their R F ' s and, judging from our latency data (Fig. 3), there might even be regions in the R F ' s " t u n e d " to specific velocities and/or directions of movement. Such local specializations certainly deserve further study. All this suggests that one H-unit could be involved in several, possibly different, local information processing activities simultaneously. The advantage
262
J. Molenaar and W.A. van de Grind
o f s u c h a n o r g a n i z a t i o n w o u l d b e a n e c o n o m y in t e r m s o f t h e n e c e s s a r y n u m b e r o f cells, b u t a d i s a d v a n t a g e w o u l d s e e m to b e t h e p o s s i b i l i t y o f r e l a t i v e l y s t r o n g interference between different local processes. Recent morphological studies of t h e o u t e r p l e x i f o r m l a y e r o f t h e r a b b i t r e t i n a b y S j 6 s t r a n d ( 1 9 7 8 , 1 9 7 6 ) also p o i n t in t h e d i r e c t i o n o f l o c a l s p e c i a l i z a t i o n s , a p t l y c a l l e d " m i c r o c i r c u i t s " by Sj6strand. We think that the further analysis of the possible functional equivalents of these morphological "microcircuits" of local circuit neurons d e s e r v e s m u c h m o r e a t t e n t i o n t h a n it has r e c e i v e d u n t i l n o w .
Acknowledgements. This work was supported by the Netherlands Organization for the Advancement of Pure Research (Z.W.O.). We gratefully acknowledge the help of A.W. Schreurs with mechanical, R. Voorhorst, J. Nivard, J. Janse with electronic, R. Swart and Mrs. A. Thiele (Berlin) with histological problems. Furthermore, we thank H. de Ridder for his help in analyzing part of the data and Mrs. W. van der Pol for her accurate typework.
References Brown, K.T., Murakami, M.: Rapid effects of dark and light adaptation upon the receptive field organization of S-potentials and late receptor potentials. Vision Res. 8, 1145-1171 (1968) Cajal, S. R. y : Die Retina der Wirbeltiere. Wiesbaden: Bergmann 1894 Boycott, B.B., Peichl, L., W/issle, H.: Morphological types of horizontal cell in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203,229-245 (1978) Fisher, S.K., Boycott, B.B.: Synaptic connections made by horizontal cells within the outer plexiform layer of the cat and the rabbit. Proc. R. Soc. (Lond.) B 186, 317-331 (1974) Foerster, M.H., Grind, W. A. van de, Griisser, O.-J.: Frequency transfer properties of three distinct types of cat horizontal cells. Exp. Brain Res. 29, 347-366 (1977a) Foerster, M. H., Grind, W.A. van de, Grtisser, O.-J.: The response of cat horizontal cells to flicker stimuli of different area, intensity and frequency. Exp. Brain Res. 29, 367-385 (1977b) Gallego, A.: Description d'une nouvelle couche cellulaire dans la r6tine des mammif6res et son r61e functionnel possible. Bull. Ass. Anat. 49, 624-631 (1964) Gallego, A.: Horizontal and amacrine cells in the mammal's retina. Vision Res. Suppl. 3, 33-50 (1971) Gallego, A.: Comparative study of the horizontal ceils in the vertebrate retina: Mammals and birds. In: Neural principles in vision. F. Zettler, R. Weiler (eds.), pp. 26-62. Berlin, Heidelberg, New York: Springer 1976 Kolb, H.: The organization of the outer plexiform layer in the retina of the cat. Electron microscopic observations. J. Neurocytol. 6, 131-153 (1977) Kuffler, S.W.: Discharge patterns and functional organization of the mammalian retina. J. Neurophysiol. 17, 558-574 (1953) Molenaar, J., Grind, W.A. van de: A visual stimulator for orbiting light spots and equal energy annuli with continuously variable parameters. Exp. Brain Res. 37, 65-71 (1979) Nelson, R.: Cat cones have rod input: A comparison of the response properties of cones and horizontal cell bodies in the retina of the cat. J. Comp. Neurol. 172, 109-135 (1977) Nelson, R., Liitzow, A. von, Kolb, H., Gouras, P.: Horizontal cells in the cat retina with independent dendritic systems. Science 189, 137-139 (1975) Rakic, P.: Local circuit neurons. Neurosci. Res. Progr. Bull. 13 (1975) Raviola, E., Gilula, N.B.: Gap junctions between photoreceptor cells in the vertebrate retina. Proc. Nat. Acad. Sci. USA 70, 1677-1681 (1973) Schmitt, F.O., Dev, P., Smith, B.H.: Electrotonic processing of information by brain cells. Science 193, 114-120 (1976) Sj6strand, F.S.: The outer plexiform layer of the rabbit retina, an important data processing center. Vision Res. 16, 1-14 (1976)
Anisotropic RF Structure of Cat Horizontal Cells
263
Sj6strand, F.S.: Circuitry analysis of the outer plexiform layer in the rabbit retina (I). J. Ultrastruct. Res. 62, 54-81 (1978) Sobrina, J.A., Gallego, A.: C61ulas amacrinas de la capa plexiforme de la retina. Actas Soc. Esp. Cienc. Fysiol. 12, 373-375 (1970) Steinberg, R.H.: Rod and cone contributions to S-potentials from the cat retina. Vision Res. 9, 1319-1329 (1969) W~issle, H., Rieman, H.J.: The mosaic of nerve cells in the mammalian retina. Proc. R. Soc. (Lond.) B 200, 441-461 (1978) W/issle, H., Boycott, B.B., Peichl, L.: Receptor contacts of horizontal cells in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203, 247-267 (1978a) W~issle, H., Peichl, L., Boycott, B.B.: Topography of horizontal ceils in the retina of the domestic cat. Proc. R. Soc. (Lond.) B 203, 269-291 (1978b)
Received February 5, 1979
