Neuronal Basis of A Sensory Analyser, the Acridid Movement Detector System: III. Control of Response Amplitude by Tonic Lateral Inhibition

Rowell & O’Shea (1976) have described the response of the locust Lobular Giant Movement Detector neurone (LGMD: O’Shea & Williams, 1974) to simple ON and OFF stimuli presented to small areas of the retina by either illuminating or darkening a 5° or 10° spot. The response is phasic, and the response to OFF stimuli exceeds that to ON. For any given area of stimulus, the response rises rapidly to a stable plateau with increasing stimulus amplitude (ΔI). The system adapts rapidly and efficiently over at least 61og₁₀ units of luminance, and except at the extremes of this range, there is little change in the response of the cell.

In the experiments of Rowell & O’Shea, the luminance of the test spot was initially the same as that of the surround. In this paper it is shown that the magnitude of the response to the test spot is determined primarily by the relationship between the initial luminance of the spot and that of the surround. The results suggest the existence of a tonic lateral inhibitory network, sited peripherally in the optic lobe before the divergence of separate ON and OFF channels, probably in the lamina. This finding explains a number of previously puzzling observations on the MD system.

Experiments were performed on Schistocerca nitens (Thunberg) and S. americana gregaria (Forskâl) from culture. (Prior to Dirsch’s (1974) revision, these species were known as S. vaga and S. gregaria respectively.) Action potentials of the Descending Contralateral Movement Detector neurone (DCMD: Rowell, 1971; O’Shea, Rowell & Williams, 1974) were recorded extracellularly from the ventral nerve cord. This neurone is connected to the LGMD by a spike-transmitting electrical synapse (O’Shea & Rowell, 1975 a) and normally follows it one to one. The spikes were amplitude-discriminated by a variable window discriminator and counted electronically.

Visual stimuli were produced by a modification of the apparatus described by Rowell & O’Shea (1976). The animal saw a screen (> 90° solid angle) illuminated by a diffuse light source; this screen was painted either matt white or matt black, depending upon the experiment. In the centre of the screen, a solenoid opened a shutter to allow the animal to see, through 35° hole, a white diffuser. The choice of the value 5°is justified in the first section of the Discussion. The shutter took about 15 ms to open. The diffuser was illuminated from behind at a known intensity, and neutral density filters could be inserted between the hole and diffuser, allowing change of intensity without change in colour temperature. Opening or closing the shutter thus caused a known increment or decrement of illuminance over 5°. This change in lumiance of the centre of the screen is referred to as the test stimulus, and its amplitude (delta-I_test) was ± 1 log₁₀ unit in all experiments. This degree of change is enough to produce a saturated response in the MD neurones (Rowell & O’Shea, 1976). Differences between the initial luminance of the test area (I_s) or that of the surrounding screen (I_t) were produced by painting white or black either the surround or the displayed surface of the shutter; the difference in reflectivity between the white and black surfaces was approximately 2 log₁₀ units. High-quality matt paints were used throughout to minimize unwanted reflexions. Using this apparatus, it was possible to present increments or decrements of 1 log₁₀ unit over the test area, coupled with any of four different surround intensities, each differing by 1 log₁₀ unit (see Fig. 1). The brightest luminance used was 350 cd/m², and the lowest 0·35 cd/m².

The animal was mounted ventral surface up, and recording electrodes placed on one pro/mesothoracic connective. The ocelli and the unused eye were covered with an opaque wax. The stimulating device was mounted on a perimeter, and was moved 10°between each presentation ; after one traverse of the retina had been made, the device was moved first back to its original position and then advanced 5° before repeating the procedure. In this way, a strip of points across the retina, each 5°apart, was tested. When this was complete, the head was rotated a few degrees, and a new strip tested. This procedure minimized response decrement by using new area of the retina for virtually every trial. Variation due to the sensitivity gradient of the retina (Palka, 1967; Rowell, 1971 ; O’Shea & Rowell, 1976) was compensated for by averaging a large number of independent readings. Trials were separated by 4 min. During the first minute the necessary movements and illumination changes to prepare for the next trial were carried out, and for the subsequent 3 min before the trial the animal adapted to an unvarying pattern of test area and surround. In pilot experiments the general room lighting was increased by approximately 1 log unit for 1 min of each rest period, in an attempt to keep the eye from adapting completely over the course of the experiment to the lower mean level of illumination which this entailed, relative to conditions when no experiments were being made. This procedure produced no significant differences, and was later abandoned.

In some experiments a bright annulus was projected on to the surround screen, either concentric with or eccentric to the test area. This annulus had an inside diameter of 9° and an outside diameter of 22·6° in some experiments, and of 10·5° and 25·8° in others. When the annulus was projected concentrically with the test area, the latter was surrounded by (in turn) a 2° or 2·75° annulus at surround intensity, a 6·8° or 7·6° annulus 0·51og₁₀ units above surround intensity, and finally by the remainder of the screen at surround intensity (Fig. 3 A).

The various conditions diagrammed in Fig. 1 produced the responses (spike frequencies) shown in Fig. 2 and Table 1. In Fig. 2 the straight lines corresponding to a least squares fit to the mean values have been drawn. It is probable, however, that the function underlying the relationship is sigmoid rather than linear ; it is likely that the curve would flatten out at steady values at both extremes, were the range of surround intensities to be extended. The results are compatible with those of other experiments employing the same conditions (Rowell & O’Shea, 1976; O’Shea & Rowell, 1976). As the luminance of the surround falls below that of the initial luminance of the test area so the response to the luminance change in the test area increases. Over the range of surround luminances investigated (3 log₁₀ units), both ON and OFF responses approximately double in size.

The results suggest that a bright annulus surrounding the test area would depress both ON and OFF responses. This is shown to be true of OFF responses in the experiment shown in Fig. 3. In this experiment, it was impracticable to move the test area relative to the retina for each trial, due to the difficulty of recentering the annulus without stimulating the eye. For this reason, Fig. 3 B shows slightly decrementing response curves, rather than mean values as in Fig. 2. As a control, the annulus was then projected on to the screen 30° distant from the test area, and Fig. 3 C shows that this separation abolishes the inhibitory effect. A similar depression of ON responses was also demonstrated by the same means. For both test and control experiments, interstimulus interval was 60 s, and conditions were alternated. The use of an annulus, instead of modifying the whole surround brightness, has the advantage that no ommatidia are simultaneously stimulated by the test area and the inhibitory annulus (see Discussion).

In the compound eye of Limulus, inhibition develops almost linearly with time until its maximum efficiency is reached approximately 400 ms after the onset of a transitory inhibitory illumination lasting 200 ms (Ratliff, Hartline & Lange, 1966), and is thereafter maintained tonically. As the morphological basis of lateral inhibition in Limulus (Miller, 1966) is very different from that in the locust (Horridge, 1966; Shaw, 1969; Horridge & Meinertzhagen, 1970; Laughlin, 1975) it would be of interest to know the time course of lateral inhibition in the latter. In principle, this can be achieved by the present experimental design, as the onset of the annulus can be arranged to precede that of the test stimulus by any desired amount. In practice, however, short intervals are ambiguous, because the annulus, as a large-field stimulus, tends to elicit both a short-latency EPSP and a longer latency IPSP in the LGMD (O’Shea & Rowell, 1975 b; C. H. F. Rowell, M. O’Shea & J. L. D. Williams, in preparation). These effects sum with the response to the test stimulus ; accordingly, the minimum time for the lateral inhibitory effect to develop cannot readily be ascertained. As soon as the post-synaptic potentials have ended (about 200 ms with an annulus of this size and intensity), however, the lateral inhibitory effect is fully developed, and shows no decrement after 4 min. This indicates a tonic, non-fatiguing inhibitory system.

To assess the possible role of lateral inhibition in a system it is of course necessary to consider the size of the stimulus in relation to that of the receptor elements. The ommatidia of a compound eye overlap in their receptive fields − that is, the angle of acceptance is greater than the interommatidial angle. It is therefore impossible to design a stimulus situation in which only light or dark are seen by each ommatidium. For any boundary, there are inevitably some ommatidia which see both sides. When assessing the results of an experiment in which the surround intensity is altered, it is therefore important to know what proportion of the receptors which are viewing the test area are also seeing the surround, and, in the case of those receptors which are seeing both, what proportion of the input to the individual receptor is derived from each.

The experiments reported here used light adapted animals (mean luminance of the field of view was approximately 30 cd/m²). The average 50% acceptance angle of the light-adapted ommatidium of Locusta (and we presume of Schistocerca) was found by Wilson (1975) to be 1·4°. Previous determinations, which gave much larger angles, were shown to be erroneous due to mechanical damage caused to the receptor. Wilson’s data shows some scatter, and the individual receptor curve he figures has a still smaller half angle (1·3°). Probably the smaller angles of his sample are the nearest to those of the totally undamaged cell. At 1° from the axis, sensitivity is down to 25 % of the axial level, and is zero at 2°.

The compound eye of a male S. nitens has been found to have 9126 ommatidia (C. H. F. Rowell & C. J. Platt, unpublished). Assuming that the hexagonal facets are evenly distributed to cover a 180° hemispherical receptive field, the expected interommatidial angle can be calculated to be 1· 17°. The diameter of a 5° circle superimposed on such an array spans almost exactly 5 facets. Measurements of the retina with a goniometer stage in a scanning EM (by courtesy of Dr C. J. Platt) show that a 5° angle actually spans from 5− 6 facets, using that part of the retina (about 30° below the horizontal meridian) used in the experiments. The slight increase over the theoretically derived figure is presumably because the curvature of the eye is not absolutely regular, and is rather flatter in this region than towards the edges of the eye.

Taking the theoretical figure, which is the worse figure for our argument, it can be calculated that a 5° disc covers 19 facets in their entirety, and partially extends over a further peripheral 12. Of these 12, about 6 are affected almost as much by the test stimulus as the surround, and are therefore ambiguous. Of the 19 ‘core’ facets, the central 7 are beyond influence (⩽20 % sensitivity) of the surround. The surround extends in to the 20 % response zone of 6 more, and into the 40% response zone of the final 6. Thus of the 25 facets which contribute a significant response to the test stimulus, 12 (48%) have some surround within their 40% response zone. For a 10° test area, the comparable figures are 12 ambiguous peripheral facets, and 86 core facets. Of the 86, 60 are unaffected by surround, and in 14 the surround extends into their 40 % response zone, making a total of 26 of the 98 responding facets (26 %) for which this is the case. Turning now to the observations upon the retina, a count on a sample 5° area gave 24 core facets, and 13 shared ones; for a 10° area, 107 core facets, and 31 shared ones. Thus the actual figures are a significant improvement over the theoretically derived ones. These figures indicate that the influence of the surround on the final response to a 10° test area will be small, and suggest that the same is possibly true of a 5⁰ test area. This uncertainty was resolved in pilot experiments by comparing the effects of 5° and 10° stimuli. The stimuli had identical results, except that the overall response (not the changes brought about by the surround) was smaller for the larger area stimulus. This is to be expected, since the MD system discriminates against large-field stimuli (Palka, 1967; C. H. F. Rowell, M. O’Shea & J. L. D. Williams, in preparation). Experiments were therefore standardized on the 5° stimulus.

Confirmation of the validity of the argument is provided by the experiment with a concentric annulus surround. The inner boarder of the annulus was here up to 2· 75° distant from the outer edge of the test area, so that no ommatidium could have been stimulated simultaneously by both annulus and test spot at greater than the 3 % level, and none of the ‘core’ ommatidia (those whose facets were covered by the stimulus) could have seen the surround at all. The homogeneity of these results causes us to dismiss as a possible explanation of our findings the contribution of ommatidia which include both test area and surround within their field of view.

The results obtained in these experiments can be explained by postulating a peripheral tonic lateral inhibitory network between the afferents of the retinotopic projection, such as has been described in flies (Zettler & Järvilehto, 1972; Zettler & Autrum, 1975) and Limulus (reviewed by Ratliff et al. 1966), but not previously demonstrated in the Orthoptera. In such a system, the retinal elements, or their post-synaptic continuations, inhibit their neighbours in proportion to the intensity of their illuminance. The effect of such a system can be understood by considering the responses to some of the stimuli shown in Fig. 1. In the first OFF response (where I_s= I_test) both test area and surround inhibit the responses to each other equally, as both are equally illuminated. The response to test area is thus moderately inhibited at the time it begins. In the first ON response, however, I_s is greater than I_test. The surround therefore inhibits the response to the test area, and in turn the inhibitory effect of the test area on the response to the surround is weakened. Consequently, the response to the test area is strongly inhibited by the surround at the moment it commences. By contrast, in the later ON and OFF responses, I_s is much less than I_test. The surround therefore exerts little inhibitory effect on the response to the test area, and furthermore it is inhibited by the response to the test area, thus weakening its effect still further. When the response to the test area begins, it is subject to little inhibition. Thus when the surround is darker than the initial luminance of the test area, the response is large, being effectively disinhibited ; when the surround is lighter than the initial luminance of the test area, the response is weak, being inhibited ; and when the surround and test area intensities are equal, an intermediate response is seen.

Rowell & O’Shea (1976) obtained their results under the last of the above conditions where test area and surround are initially of the same luminance. Palka’s (1967) experiments, however, used mainly OFF stimuli obtained by quenching brightly lit sources in an otherwise darkened room. Until now, the present authors have been puzzled by the large numbers of action potentials obtained per response in those experiments. The relationship shown in Fig. 2, however, explains this as the result of extreme disinhibition of the response as a consequence of the dark surround.

All points on the retina which have been tested have produced similar results, and there is no indication of any heterogeneity over the retina. Especially, there is no indication of any ‘centre/surround’ organization such as that described from the vertebrate visual system, and for this reason we have throughout this account used the term ‘test area’ rather than ‘centre’, as contrasted with ‘surround’. Another characteristic of a simple lateral inhibition network is that distant areas of the retina do not affect each other measurably, and contiguous areas affect each other maximally. Our data indicate that this is the case in the present experiments. A bright annulus projected around the test area is inhibitory, even though its inner edge is separated from the test area itself by some degrees of unlit surface, whereas the same annulus, projected on to the retina some 30° distant from the test area, has no discernible inhibitory effect.

The lateral inhibition described here is tonic, causes the same percentage inhibition in both ON and OFF responses, and is unimodal (that is, both ON and OFF responses are inhibited by an increase of illuminance of the surround areas of the retina). These observations suggest that it is sited between tonic afferents, at a peripheral stage in the optic lobe prior to the differentiation of separate ON and OFF channels, and preclude the possibility that it is situated between the phasic, ON/OF medullary afferents of the second chiasma (O’Shea & Rowell, 1976). It probably corresponds to the network of lateral inhibitory connexions seen in the lamina of the fly (Strausfeld & Braitenberg, 1970; Zettler & Järvilehto, 1972; Strausfeld, 1976), though we have no evidence that precludes its being situated in the distal portion of the medulla. The situation described from the fly lamina would be expected, a priori, to produce the described effects in the LGMD, given the circuit so far specified in our previous papers in this series. As far as we know, this is the first time that the consequences of the peripheral lateral inhibition network have been reported in the response characteristics of a high-order visual interneurone. If we are correct in attributing this lateral interaction to the lamina (i.e. before any marked divergence of the visual input has taken place) then one would expect that similar effects will be found in most components of the visual system, and indeed D. Edwards (personal communication, 1976) has recently found analogous effects in a visual unit in the cockroach nerve cord.

This work was supported by USPHS grant 5 RO1 NS 90404 to C.H.F.R.