J. Physiol. (1985), 369, pp. 249-268 249 With 12 text-ftgures Printed in Great Britain A COMPARISON OF VISUAL RESPONSES OF CAT LATERAL GENICULATE NUCLEUS NEURONES WITH THOSE OF GANGLION CELLS AFFERENT TO THEM BY B. G. CLELAND AND B. B. LEE* From the Department of Physiology, University of Sydney, N.S. W. 2006, Australia (Received 22 November 1984) SUMMARY 1. We compared visual responses of cat lateral geniculate nucleus (l.g.n.) neurones with those of retinal ganglion cells providing their afferent inputs. Quantitative studies were made on twenty such pairs; eight X on-centre, seven Y on-centre, two X off-centre and three Y off-centre pairs. 2. Receptive field centre locations of cell pairs with correlated activities were very closely superimposed, having a mean centre displacement of 1-6 minutes of arc for X cells and 11 minutes of arc for Y cells. 3. With flashed spots and annuli, responses of l.g.n. cells were almost always smaller than those of their retinal afferents, with peaks and troughs in ganglion cell responses being faithfully followed in the geniculate neurones. This is consistent with almost all impulses from the l.g.n. cell being triggered by the afferent feeding its centre. With spots of different sizes and contrasts, modulation of responses by l.g.n. inhibition was obvious, but effects were complex. 4. With moving bright-bar stimuli, although response histograms were clearly reshaped to some extent in the l.g.n., peak firing rates under different stimulus conditions were often merely attenuated by a constant factor for most l.g.n. cells in comparison with their retinal inputs. For velocity tuning curves, a few cell pairs showed selective attenuation at high speeds, while others showed it at low speeds. All the latter group appeared to have more than one major excitatory afferent. These changes in velocity tuning occurred across the X/Y classification, so that differences in velocity preference of the X and Y systems is more blurred in the l.g.n. than in the retina. INTRODUCTION If an electrode in the retina records from ganglion cells whose receptive fields overlap that of a simultaneously recorded lateral geniculate nucleus (l.g.n.) neurone, it is found that the great majority of a l.g.n. relay cell's impulses are associated with discharges of one or two retinal ganglion cells (Cleland, Dubin & Levick, 1971 a). The delay between ganglion and l.g.n. cell spikes is compatible with the conduction time for the cell type (X, brisk sustained or Y, brisk transient) concerned. Since firing of * Permanent address: Max-Planck-Institut fur biophysikalische Chemie, D-3400 Gottingen, F.R.G.
250 B. G. CLELAND AND B. B. LEE adjacent ganglion cells is relatively uncorrelated (Mastronarde, 1983a), one mayconclude that the major excitatory synaptic input from the retina onto a given relay cell is derived from one or two retinal axons. No convergence of on- and off-centre and little mixing of X and Y pathways could be found (Cleland et al. 1971 a). The excitatory input of l.g.n. relay cells is modulated by inhibitory interactions. mediated by interneurones situated within and around the nucleus (Dubin & Cleland. 1977). These interactions are apparent in the stronger surround inhibition of l.g.n. neurones as compared to ganglion cells (Hubel & Wiesel, 1961). Although the time course and spatial properties of this inhibition can be estimated from intracellular recordings (Singer & Creutzfeldt, 1970; Singer, Poppel & Creutzfeldt, 1972). the wsay in which this inhibition modulates the visual signal with different spatial and temporal stimulus configurations is not entirely clear. A comparison of responses of ganglion and l.g.n. cells demonstrates spatial and temporal filtering takes place (So & Shapley, 1979; Lee, Virsu & Elepfandt, 1981; Troy, 1983), but the comparison is made difficult by variability in response properties with retinal eccentricity and inter-cell variability per se. A direct comparison of visual responses of a l.g.n. relay cell and a ganglion cell afferent to it can provide direct evidence as to the relation between input and output of the l.g.n., and may elucidate the nature of the underlying inhibitory mechanisms. In this paper we make this comparison; the contribution of a ganglion cell to a relacell's activity in response to different stimuli will be discussed elsewhere. METHODS Adult cats were anaesthetized with halothane (1-4%O) in a nitrous oxide and oxygen (70 0 0 N20: 30% 02) gas mixture. After section of the left sympathetic trunk and insertion of tracheal and venous cannulae, the animal was mounted in a stereotaxic frame and a small craniotomv made over onel.g.n. The left eye was then prepared for insertion of a micro-electrode for recording from ganglion cells (Cleland, 1983). Eye movements were prevented with a continuous infusion of 9 mg/kg.h of gallamine triethiodide together with 0-4 mg/kg.h of d-tubocurarine. Depth of artificial ventilation was so adjusted to keep end-tidalpco2 near 40 /. Subscapular temperaturewas held near 37-5 TC with a wrap-around heating blanket. Continuous anaesthesia was maintained with the nitrous oxide-oxygen mixture. Heart rate and the pupil of the unoperated eye were carefully monitored for changes that may be associated with painful stimuli. If such changes occurred. halothane (0-2-0-5%) was added to the gas mixture. The eye was focused on a tangent screen 1-14 m from the corneal surface and an artificial pupil (3 mm) installed. Visual stimuli (moving bars, flashing spots and annuli) were front-projected onto the screen which had a luminance of 1-8 cd/m2. Stimulus shape, contrast. the speed of moving bars and the timing of flashed stimuli were under control of a computer which also collected. averaged and stored unit activity (Cleland, 1983). Contrast was defined as luminance increment. AL. divided by background luminance, L, that is AL/L. After isolation of activity of a singlel.g.n. neurone, its receptive field was plotted and it was identified as on- or off-centre and X or Y using tests described by Cleland et al. (1971 a). Then impulses from a ganglion cell with a receptive field near that of thel.g.n. neurone were recorded. A thin bar (0-1 x 10 deg) was moved back and forth across the cells' receptive fields to determine the separation of the ganglion and l.g.n. cells' field centres. To do this, response histograms for the back and forth directions were slid over one another until they superimposed. This is equivalent to reversing the histogram for the reverse sweep and superimposing it upon the forward sweep so that stimulus position, and not time, is plotted on the x-axis. The point about which the two histograms are symmetrical then indicates the receptive field position. Such a point should be independent of response latency, transmission delay and synaptic delay, which will move the peaks further apart but by equal amounts from this point of symmetry. Thisprocedure. carried out for ganglion and
COMPARISON OF RESPONSES IN RETINA AND L.G.N. l.g.n. cells for horizontal and vertical stimulus movements, enabled centre separation to be precisely determined (to about 1/20 deg). Using this measure, the retinal electrode was then adjusted so as to acquire another cell and the procedure repeated until the receptive fields of the two cells superimposed. Correlation between firing of ganglion and l.g.n. cells was monitored by inspection of their impulses on an oscilloscope, the l.g.n. cell firing reliably after the ganglion cell impulse when a synaptically connected pair was acquired (Cleland et al. 1971 a). These correlations were easily recognized, and could be classified as strong, when practically all geniculate spikes followed a geniculate spike at a given latency, or moderate, when some l.g.n. spikes were not associated with a ganglion cell spike, although often a prepotential was seen. Responses of these neurone pairs to a variety of visual stimuli at different contrasts relative to the background were then recorded. Response histograms were subjected to Gaussian smoothing (usually with of value of 1 bin; histograms usually included 150-200 bins for each direction of movement), peak firing rates measured, and maintained firing rate subtracted to give the values plotted in the Figures. 251 RESULTS We carried out extended searches of the retina for ganglion cells providing afferent input to a l.g.n. cell on about fifty occasions. For twenty cells (eight on-centre X, seven on-centre Y, two off-centre X, three off-centre Y, with eccentricities between 3 deg and 13 deg) a well-isolated ganglion cell spike was obtained which clearly provided a major contribution to the geniculate cell firing; on one occasion two X on-centre ganglion cells were found which were afferent to the same geniculate neurone. In all these instances an extensive quantitative study of responses of the cell pairs was undertaken, using a variety of moving and flashed stimuli at several different contrast levels. On numerous other occasions (ca. twenty), an afferent ganglion cell was identified, but for various reasons quantitative data were unobtainable; either the ganglion cell impulse was too small or too unstable for reliable discrimination, or adjustment of the retinal recording electrode resulted in an injury discharge and presumed death of the ganglion cell (the l.g.n. cell becoming unresponsive). Qualitative observations on these cell pairs indicated that their behaviour was similar to those for which quantitative data were obtained; there was no indication that any bias in the cell sample might have occurred. Whether a ganglion cell was afferent to a l.g.n. cell became immediately apparent as soon as the oscilloscope sweep was triggered on the ganglion cell spike. An example is shown in Cleland et al. (1971 a; Fig. 7); a further example is shown in Fig. 1. Although the cell impulses are not large, they are clearly discriminable. The geniculate recording is of interest since a small triphasic event occurs beginning about 2-5 ms after the ganglion cell fires. These events are sometimes seen in geniculate recordings and are presumably associated with an impulse in the afferent axon (Wang, Cleland & Burke, 1985). There follows an S- or prepotential (Bishop, Burke & Davis, 1962) with a latency of 2-9 ms. Both afferent impulse and prepotential bear a one-to-one relation to the ganglion cell discharge. From the prepotential arises the impulse of the geniculate cell, but not on every occasion, and with a variable latency of about 3-2 ms (Cleland et al. 1971 a). Such recordings as in Fig. 1 were typical, although this was the only instance in which a triphasic potential from the afferent axon was seen. Prepotentials with a one-to-one relationship with ganglion cell impulses were seen for nine of the twenty cell pairs. As found by Cleland et al. (1971 a), the
252 : ~~~~~~~~~~~~~~~~~~~~~~I B. G. CLELAND AND B. B. LEE -,AL- I -~~~~Vo Fig. 1. Simultaneous recording from retina (above) and geniculate (below), both traces being triggered from the ganglion cell spike. The lateral geniculate nucleus (l.g.n.) trace shows a small triphasic potential (arrowed), followed by a small prepotential, from which a l.g.n. cell discharge may arise. Calibration marks are 1 ms and 100 #V. Negative is an upward deflexion of both traces. B Y-on 0 deg 90 deg 90 deg A~~~ ~~~~~~~C A C X-off DY Y-Off 0 deg 0 e 90 deg arm 90 deg Fig. 2. Responses to moving bright bars (01 x 10 deg) of X on-centre (A), Y on-centre (B), X off-centre (C) and Y off-centre (D) cells. Bar velocity was 10 deg/s and sweep amplitude 7.5 deg. Thin trace is ganglion cell response, thick trace l.g.n. cell. Stimulus reverses direction at arrows. In upper histograms of each pair stimulus bar moved horizontally, in lower vertically. Contrast was 33 (AL/L). Responses of cell pairs closely superimpose. Bin width 5 ms. A
COMPARISON OF RESPONSES IN RETINA AND L.G.N. latency from ganglion cell to l.g.n. cell impulse ranged from 2-5 to 3-5 ms for Y cells and 4-3 to 6-5 ms for X cells in our sample. The search for a cell pair was aided by moving a thin bar back and forth across the receptive fields of the cells. The relative locations of receptive field centres could be estimated from the response histograms generated for the two directions of movement, as described in the Methods; since this estimation involved determining the location about which such responses were symmetrical, it was independent of any latency difference between retina and l.g.n. Whenever a correlated pair was observed, their receptive field centres were very closely superimposed. Fig. 2A-D illustrates this for X and Y on- and off-centre cells. In Fig. 2A, responses of an X on-centre cell pair are shown for horizontal and vertical directions of movement. The histograms take the typical form for an on-centre cell and a bright bar, with a peak as the bar passes across the receptive field centre followed by cessation of maintained firing as it moves into the surround (Rodieck & Stone, 1965; Dreher & Sanderson, 1973). Responses of ganglion and l.g.n. cells, indicated by thin and thick lines respectively, line up to within a few minutes of arc. Retinal eccentricity for these cells' fields was 3 deg, close to the vertical meridian, where X cell density is about 2500/mm2 (Hughes, 1981). The mean distance between X on-centre cells would then be about 10 minutes of arc, which is 3 bins in the histogram. The close superimposition of responses suggests that only the one ganglion cell contributed to the l.g.n. cell's response, while from the recordings it also appeared that every geniculate impulse followed a discharge from the ganglion cell. The ganglion cell thus provided by far the major synaptic drive to the geniculate neurone. The efficacy of a ganglion cell in driving its geniculate target varied widely from cell pair to cell pair. Fig. 2B shows a Y on-centre cell pair in which virtually every afferent impulse generated a spike in the l.g.n. cell. At an eccentricity of 8 deg, where this cell pair was found, centres of neighbouring Y on-centre cells would be about 50 minutes of arc distant, about 16 bins in the histogram. As in Fig. 2 A, the responses of this l.g.n. neurone were consistent with its sole excitatory input from the retina coming from the Y on-centre ganglion cell we found. Most searches aimed at finding inputs ofon-centre neurones, so that the transmission of centre-evoked responses to bright stimuli could be studied. We did, however, find afferent cells for three Y and two X off-centre cells. Examples of histograms from one example of each of these cell types are shown in Fig. 2 C and D, horizontally and vertically moving bars being used as before. On the basis of alignment of receptive fields and inspection of recordings, the Y off-centre ganglion cell of Fig. 2 D probably provided the sole contribution to the geniculate response, while with the X off-centre cell of Fig. 2C, another neighbouring ganglion cell was required to account for the l.g.n. cell's responses. The displacement of centres of members of this pair can be seen from the histogram, and inspection of the cells' activities also led to this conclusion. Of the twenty pairs, in twelve of them the recorded ganglion cell appeared to provide the sole synaptic drive to the l.g.n. cell. Correlation between the cells was strong in that practically all l.g.n. spikes were correlated with ganglion cell impulses. For many of these cells, we recorded from neighbouring ganglion cells of the same type (as estimated from cell density) without any correlation in activity being found. In the other eight cases, significant contribution from at least one other ganglion cell 253
254 B. G. CLELAND AND B. B. LEE appeared to be present; correlation was moderate in that a fraction of l.g.n. spikes were not associated with ganglion cell impulses. The shape of response histograms to the thin moving bar was often modified in the l.g.n. in relation to the retinal input. This can be seen in Fig. 2A, in which preceding the peak of the ganglion cell response a slow increase in firing rate can be seen. This is lost in the l.g.n. cell response which has thus a sharper peak. The 8 A X-on 0 deg 90 deg A 0 a O~~0deg 90g d.aeg pne ofxo-etecl o01x1 0~~~9 e e br a eoiyws1 e/ n Fig. 3. A, responses of X on-centre cell to 0-1 x 10 deg bar. Bar velocity was 10 deg/s and sweep amplitude 7-5 deg. Thin trace is ganglion cell response, thick trace l.g.n. cell. Stimulus reverses direction at arrows. In upper histograms ofeach pair stimulus bar moved horizontally, in lower vertically. Contrast was 33. The response of the l.g.n. cell is slightly delayed with respect to the ganglion cell. This is so for both movement directions and for horizontal and vertical movement, indicating that there is no displacement of field centres. B, responses of Y on-centre geniculate cell and Y on-centre ganglion cell. Responses do not exactly superimpose; hence a further ganglion cell input is likely. C, responses of X on-centre ganglion cell and same Y on-centre geniculate cell. Despite very close superimposition of responses, no correlation of activity was observed. maintained firing of the ganglion cell is also not fully transmitted; maintained firing of geniculate neurones was almost always much less than that of their afferent inputs (Cleland et al. 1971 a). However, some maintained firing was present in the l.g.n. cell; the lack of the slow increase in firing is thus likely to be due to inhibitory interaction rather than some kind of threshold effect. Further instances of modelling ofganglion cell responses in the geniculate are shown in Fig. 3. In Fig. 3 A an X on-centre cell pair is shown where the ganglion cell appeared
COMPARISON OF RESPONSES IN RETINA AND L.G.N. to be the sole drive. Responses of the geniculate cell are slightly delayed with relation to the ganglion cell; this is not due to displacement of field centres because the delay is present in both movement directions for horizontal and vertical movement. Several X on-centre pairs displayed this shift, which may be related to the 'lagged X-cells' briefly noted by Mastronarde (1983b). Fig. 3B shows a geniculate cell, clearly classified as Y on-centre with hand-held stimuli, which, although a high proportion of l.g.n. spikes followed those of the ganglion cell, must have received at least one other Y on-centre drive in addition to the one we found, since the geniculate cell firing rate exceeds that of the ganglion cell in some bins for the first movement direction at 0 deg. Such an effect cannot be due to transmission latency, since this took up less than a bin. Thus, since l.g.n. impulses occurred without being preceded by a ganglion cell discharge, an additional input is implied. The peak of the ganglion cell response is much broader than that of the geniculate neurone. For comparison, Fig. 3C shows the same l.g.n. Y cell and an X on-centre cell with almost complete overlap of response peaks. No correlation of the activities of the two cells was present. A priori, when a l.g.n. cell receives more than one excitatory afferent, its maximum firing rate to a particular stimulus might exceed that of either afferent alone. This appeared to be so at certain stimulus velocities, as will be shown later. With l.g.n. cells driven mainly from one afferent, the firing rate of the l.g.n. cell might be expected to be below that of the ganglion cell no matter what stimuli were used. This was generally the case, but in some cells a single afferent impulse triggered two or three l.g.n. discharges, as has also been noted in quasi-intracellular records where a single excitatory post-synaptic potential (e.p.s.p.) may support more than one spike (Singer & Creutzfeldt, 1970). These bursts of impulses were especially common following a suppression of firing, as occurs when the bright bar leaves the centre and enters the surround of an on-centre cell. A response peak may then occur as the bar leaves the surround (Dreher & Sanderson, 1973; Lee, Virsu & Creutzfeldt, 1977), as in Fig. 3B. Doublets or triplets of impulses occurring in this peak had the consequence that the l.g.n. firing rate sometimes exceeded that of the ganglion cell, so that this secondary peak did not bear such a precise relation to ganglion cell activity as the primary peak when the bar moved across the field centre. A further instance is the sharp peak in the off-centre neurone of Fig. 2D when the bar leaves the centre and enters the surround. The cellular basis for this may lie in some kind of inhibitory rebound (Singer & Creutzfeldt, 1970). In conclusion, although modification and attenuation of bar responses occurred between ganglion and l.g.n. cells, the receptive field centres of synaptically connected cell pairs were always very closely superimposed. Judging centre position from bar responses, the mean separation of centres of cell pairs was 1-6 minutes of arc (n = 12) for X and 11 minutes of arc (n = 9) for Y cells. At an eccentricity of 5 deg (typical for our sample), centre separation of on- or off- X cells would be about 12 minutes of arc and of on- or off- Y cells about 50 minutes of arc. 255
256 B. G. CLELAND AND B. B. LEE A comparison of responses to flashed stimuli The more marked surround inhibition of l.g.n. neurones than of ganglion cells, as first described by Hubel & Wiesel (1961), is ascribed to inhibition within the nucleus (Singer & Creutzfeldt, 1970; Singer et al. 1972). When we compared responses to flashed stimuli within a cell pair, the effects of this inhibition were clearly apparent, but the details of the effects seen varied a good deal with different stimuli and contrasts, as well as varying from cell to cell. 0-4 deg 08 deg 2 5 deg 1 0 deg 10.$Y~ 32 100 a'=~~~ (A.w XC1n Fig. 4. Responses of X on-centre cell pair to flashing spots of different diameters and contrasts (the latter is indicated on the left of each row). Thin line indicates ganglion cell response, thick line l.g.n. cell response. Bar indicates stimulus duration (1 s with I s between stimuli). L.g.n. cell responses are always less strong than those of the ganglion cell. Comparison of histograms at different spot sizes reveals action of l.g.n. inhibition, which, however, seems to be a complex function of contrast and spot size. Bin duration 10 Ms. Fig. 4 illustrates responses from an X on-centre cell pair to spots of different diameters and contrasts. With hand-held stimuli centre sizes of ganglion and 1.g.n. cells appeared to be comparable; centre size of the ganglion cell, as estimated from the histogram peak to a narrow bar at low velocity (Cleland, Harding & Tulunay- Keesey, 1983), was about 0-65 deg. With a 0-4 deg spot on the receptive field centre, at the highest contrasts shown the ganglion cell response consists of an initial peak decaying down to a sustained level. The initial peak is also present in the l.g.n. cell response but there is a more rapid decrease in response to a steady level. rthis attenuation may reflect the 'secondary inhibition' seen in intracellular records (Singer & Creutzfeldt, 1970). With lower contrasts the efficacy of the ganglion cell in generating l.g.n. spikes improved, so that instead of about 30 0/ of the ganglion
COMPARISON OF RESPONSES IN RETINA AND L.G.N. cell impulses being effective, about 80 % trigger l.g.n. spikes. A similar pattern is present with a 0-8 deg spot, except that here the efficacy of the ganglion cell in triggering l.g.n. spikes is lower at all contrasts, consistent with more extensive activation of inhibition in the l.g.n. with the larger spot. With a 2-5 deg spot this is even more marked, so that little l.g.n. response is present even though some response is still seen in the ganglion cell. With a 10 deg spot, little response occurs for either cell. 1 2 deg 25 deg 10 deg 257 33S<. 4-2 0.52 C,,a Fig. 5. Responses of Y on-centre cell pair to flashing spots of different diameters and contrasts (the latter is indicated at the beginning of each row). Decay of response during the stimulus is more rapid in this cell type. Effect of inhibition is also apparent in this cell type with larger spot sizes. Bar indicates stimulus duration (1 s with I s between stimuli). Bin duration 10 ms. In Fig. 4, increasing stimulus size diminishes synaptic efficacy, consistent with increased activation of inhibitory interneurones with larger stimuli. This is of course not a new finding; however, several features of these inhibitory interactions deserve comment. First, attenuation of l.g.n. relative to ganglion cell firing rate occurred throughout the sustained component ofthe responses; a deeper inhibition immediately following the initial peak, as seen in intracellular records, was seldom evident. Secondly, if attenuation of ganglion cell responses is a measure of inhibitory effects, then inhibition would seem to depend in a complex way on intensity and stimulus size. For example, as contrast is increased from 3-2 to 10, the ganglion cell response increases slightly while the l.g.n. cell response diminishes substantially, an cffcct 9 v' 1 :Y369
258 B. G. CLELAND AND B. B. LEE readily explicable in terms of increased activation of an 'inhibitory pool' at higher contrasts. Fig. 5 shows a similar set of histograms for a Y on-centre cell, the cell of Fig. 2B. Flash responses are more transient as expected from this cell type. For 1-2 and 2-5 deg spots almost all ganglion-cell impulses evoke an l.g.n. spike, except at the highest contrast for the larger spot size. For a 10 deg stimulus at high contrast the attenuation of transmission by geniculate inhibition is again evident. A 33 004 deg 0-8 deg 2 5 deg 10 deg B 0-8 deg 1-2 deg 2 deg 10.,~J2~~ 0 4 deg 0-8 deg 2-4 deg 10 deg 3.2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ D 08deg 1 2deg 2deg 10 Co - I 0 Fig. 6. Responses ofx (A, B) and Y (C, D) on-centre pairs to flashing spots (A, C) and annuli (B, D). The inner diameter of the annuli is indicated; outer diameter was about 20 deg. Bar indicates stimulus duration (1 s with 1 s between stimuli). Contrast is indicated at left. The responses of the geniculate cells closely follow those of the retinal ganglion cells, though the efficacy of the afferent cell in firing the l.g.n. cell varies with stimulus conditions. Bin duration 10 ms. The lack of correlation in firing between off-centre ganglion cells and on-centre l.g.n. neurones and vice vera, would suggest that retinal afferents do not make excitatory synapses onto l.g.n. cells of the opposite centre type (Cleland et al. 1971 a). Nevertheless, it might be that impulses evoked by stimulating the surround are not directly derived from the afferent driving centre responses. We therefore stimulated cell pairs with both spots and annuli at different contrasts. The annuli had a large outside diameter (ca. 20 deg); the inside diameter was varied. Examples of X and Y on-centre cells (at just one contrast level) are shown in Fig. 6, and examples ofx and Y off-centre cells in Fig. 7. In both these Figures, A and C are responses of X and Y cells to bright spots, and B and D to bright annuli.
COMPARISON OF RESPONSES IN RETINA AND L.G.N. The histograms in Fig. 6A and C indicate the changes in synaptic efficacy with spot size shown in Figs. 4 and 5. With annuli, in B and D, off-responses are evoked in which the initial peak of the ganglion cell response is scarcely attenuated in the l.g.n. neurone, though later response components are attenuated. A similar pattern is present in the off-centre cells of Fig. 7, except that the responses occur in the opposite phase of stimulus presentation. A 04 deg 08 deg 25 deg 10deg 259 10 de2 2 deg C 08 deg 25 deg l2deg 3.2 D 0D8Odeg 1 2deg2deg W C4J Fig. 7. Responses ofx (A, B) and Y (C, D) off-centre pairs to flashing spots (A, C) and annuli (B, D). The inner diameter of the annuli is indicated; outer diameter was about 20 deg. Bar indicates stimulus duration (1 s with 1 s between stimuli). Contrast is indicated at left. L.g.n. cell responses follow ganglion cell responses. Bin duration 10 ms. It is striking in Figs. 6 and 7 how the patterns of responses of the l.g.n. cells closely follow their retinal afferents, in that peaks in l.g.n. cell firing rate are derived from peaks in ganglion cell activity, although peak size in the l.g.n. varies with stimulus conditions. Only on a few occasions is the firing rate of the ganglion cell lower than that of its l.g.n. cell, and then only following a period of inhibition. Under the different stimulus conditions, no marked changes in driving of the l.g.n. by the retinal cell were observed on the oscilloscope. In summary therefore, discharges of an l.g.n. cell appear to be derived from retinal afferents feeding the centre under all the stimulus conditions we employed. Comparison of responses to moving stimuli In the retina, amplitude of responses to moving stimuli depends in an apparently complex way on bar contrast, speed of movement, width and length. However, if flux 9-2
260 B. G. CLELAND AND B. B. LEE falling on the centre and surround and centre-surround interaction are considered, many of the effects of varying these stimulus parameters can be explained (Cleland, Harding & Tulunay-Keesey, 1983; Cleland & Harding, 1983). In particular, the differences in velocity preference between X and Y cells appear to be related to receptive field centre size. 400B 300 A 33 1i001:101100 33 60 601~~~~03 Stimulus velocity (deg/s) 1 10 100 1000 Fig. 8. Velocity-response curves of X (A) and Y (B) on-centre cell pairs. Open circles indicate ganglion cell responses, filled circles l.g.n. cell responses. Contrast is indicated for each pair of curves. Responses were obtained with a 0- x 10 deg bar and averaged for the two movement directions. For each cell the same curve template was used to draw curves through all sets of data, displaced as necessary. For the X cell pair, a constant fraction of afferent spikes generate a l.g.n. cell response irrespective of stimulus conditions. For the Y cell, some variation with contrast was found. We compared responses of members of cell pairs to moving bars, to see how far the retinal pattern is changed in the geniculate. Since an oft-quoted difference between X and Y cells is velocity tuning we concentrated on measuring velocityresponse curves at different contrasts (for eighteen pairs), but, in addition, we also changed bar length and width (for ten pairs). The velocity-response curves for nine of the fourteen on-centre pairs tested did not demonstrate radical changes in tuning between retina and geniculate. Figs. 8 and 9 illustrate this. Data in Fig. 8A and B are from X and Y on-centre cells at different contrasts, a 01 x 10 deg bar being used in each case. To avoid overlap, each pair of curves have been shifted vertically. As in Cleland &Z Harding (1983) for each cell pair an identical template was used in drawing the solid curves, the template being moved horizontally and vertically as necessary at each contrast. Within each pair of curves, only a vertical displacement was necessary. The behaviour of the ganglion cells is as expected from the earlier work, but it is striking that in the geniculate little change in velocity tuning occurs; for the X cell of Fig. 8A a constant fraction of ganglion cell spikes evoked an l.g.n. spike independent of velocity and contrast, while for the Y cell of Fig. 8B, a lower efficacy of the ganglion cell was present at the highest contrast.
COMPARISON OF RESPONSES IN RETINA AND L.G.N. A similar result was found for eight other pairs, except that for three pairs responses for higher velocities (ca. 40 deg/s and above) were attenuated, examples being seen in Fig. 9A for Y and X cell pairs (upper and lower pairs of curves). We have fitted the ascending limb of the geniculate cell curve with the ganglion cell template in each case. At higher velocities, the geniculate cells respond more weakly than expected. The Y cell shown was the most marked instance we found, a definite decrease in peak velocity occurring from retina to geniculate. B 300 A 400-100= 0 ~~~~~~40] 261 11 1 10300 40 100.\ 0\~~~33 oloo 033 *10 * 33 30-101000010000 100: Stimulus velocity (deg/s) Fig. 9. A, velocity-response curves for a Y (upper pair of curves) and an X on-centre cell pair (lower pair of curves). Curves were fitted in the same way as in Fig. 8. There is some attenuation of l.g.n. cell responsiveness at higher relative to lower speeds, especially for the Y cell pair. B, velocity-response curves for X on-centre cell pair with three different stimuli; above, 01 x 10 deg bar, middle 004x 10 deg bar and below 01 x 1-0 deg bar. Little change in velocity tuning was visible for this cell pair. In A and B, contrast is indicated at the right of each pair of curves. For retinal ganglion cells velocity tuning curves change in an orderly manner as bar contrast, width and length are altered, with the same curve template being usable to fit the data. This was also ns in the geniculate, any change relative to the retinal input being present over all the conditions we tested. For stimulus contrast, Fig. 8 can be taken as an example, while in Fig. 9B length and width ofthe bar were altered. The top pair were obtained with a 0-1 x 10 deg bar, the middle with a 04 x 10 deg bar and the lower with a 01 x 1 deg short bar. The same curve has been drawn through the data in each case, shifted along the axes to get best fit as in Fig. 8. Changes in peak amplitude and best velocity occur as expected (Cleland &iharding, 1983), both cells undergoing comparable changes. The other five on-centre pairs (three Y, two X) underwent more striking changes in velocity tuning, in the form of a sharp attenuation of responses to low velocities, with l.g.n. responses to high velocities often exceeding the ganglion-cell firing rate.
262 B. G. CLELAND AND B. B. LEE All five of these pairs belonged to a group of six on-centre pairs where the geniculate cell appeared to have more than one major excitatory input. In the case of one X cell we were fortunate in obtaining good recordings from both its major inputs. Fig. 1OA illustrates velocity-response curves at two contrast levels for the one input, and Fig. lob at one contrast level for the other. Curves at other contrasts and with a short bar showed the same steep fall in responsiveness at low speeds, as did curves with the bar moving vertically instead of horizontally. Fig. lob illustrates a similar result for a Y on-centre cell at three different contrasts; a steep decrease in responsiveness at low speeds and an increase in optimum velocity occurs. 200 A 100 * ~~~~~~~200-C ** 100:-. @33 2 3 30* X 100: 30 -, 200. 0 g 100! 200 1 0 230 0. 0 1~~~~~~~~~~~01 0 30. 00 X ~~ ~ ~~~10 100 1000* B, ''" 20 j 100: * 33~~~260 * 0 20-0 201 :3 20J~~~~~~ 1 10 100 1000 1 10 100 1000 Stimulus velocity (deg/s) Fig. 10. A, B, velocity tuning curves for an X on-centre l.g.n. cell and its two retinal afferents. Both afferents evoked a prepotential but of different size. There is attenuation of l.g.n. cell response at low but enhancement at high speeds relative to the retinal inputs. C, the same pattern occurs for a Y on-centre cell, shown here for three different contrasts. Contrast is indicated at the right for each pair of curves. Finally, Fig. 1 1 shows velocity-response curves for X and Y off-centre cells. A framework for interpretation of responses of off-centre cells to bright bars has not yet been developed, since such responses are dependent on a mixture of centre disinhibition and surround excitation and thus more difficult to analyse than centre responses of on-centre cells. For the X cell, which was probably the sole input, not too great a difference in velocity tuning can be seen. For the Y cell, another afferent probably contributed to the cell's response. Again, no great alteration in velocity tuning is apparent. The l.g.n. cell had a tendency to fire in doublets and triplets, each associated with a single afferent impulse, resulting in l.g.n. cell firing rates which sometimes exceeded those of the retinal cell. In summary, velocity tuning curves of many l.g.n. cells appear to follow those of their retinal inputs, with a constant fraction of l.g.n. impulses generating spikes. Some
COMPARISON OF RESPONSES IN RETINA AND L.G.N. 263 attenuation of responses in the l.g.n. at high or low stimulus speeds was present for other pairs. Any such transformation of velocity tuning was, to a first approximation, independent of parameters such as bar length, width and contrast. No evidence of selective effects in X or Y pathways was seen. 300 1A 100-0 0 U) w a de1 D 0 OS 300 300 B I I I I I I I I I I I I I I I I I I I I a I I I I I I I 1 10 100 1000 100 30 - I I I I I I I I I I I I I I I I I I I I I I II 1 10 100 1000 Stimulus velocity (deg/s) Fig. 11. Velocity tuning curves for X (A) and Y (B) off-centre cells, in which little change in tuning is apparent. Responses ofoff-centre cells tended to be more variable than on-centre cells with a bright bar, and there is more scatter of the data points. Contrast was 33. DISCUSSION There can now be little doubt that excitatory convergence of retinal ganglion cell afferents is minimal in the main layers of the geniculate. Indeed, for twelve of the twenty pairs we recorded, it appeared that only one ganglion cell provided the main excitatory drive to the l.g.n. cell. In this context it is worth stressing that as soon as such a ganglion cell was found, a very clear correlation appeared on the oscilloscope; neighbouring neurones, even if recorded simultaneously, did not exhibit any correlation. A critical assumption is, nevertheless, that firing of neighbouring ganglion cells is uncorrelated. Recent results of Mastronarde (1983 a) broadly confirm earlier observations of Rodieck (1967). Correlated events (that is, two cells firing within a few milliseconds of one another with a probability exceeding that expected by chance) in spike trains of neighbouring Y cells can number up to 20 % of the cells' impulses, but are usually less. Correlated events with neighbouring X cells number less than 10 % of their impulses, or are entirely absent. Further, the correlation of neighbouring ganglion cells is often spread over several milliseconds. Correlation
264 B. G. CLELAND AND B. B. LEE levels of retino-geniculate cell pairs are much higher, 100% drives being possible with tighter temporal correlation, peaks covering maximally two or three milliseconds (Cleland et al. 1971a, b). If all, or almost all, geniculate impulses derive from a single afferent, then the response of the l.g.n. cell should be always less or equal to the ganglion cell response. This was the case, even in responses evoked by annuli in the surround, except for the few impulses which might reflect some 'post-inhibitory rebound' effect. Firing rates of l.g.n. cells were also generally below that of their retinal afferents in cases where more than one ganglion cell provided input, an exception being high stimulus velocities. The responses to flashing spots and annuli demonstrate the well-known inhibitory interactions which occur in the geniculate. There exist inhibitory interneurones both in the principal layers of the geniculate and in the perigeniculate (Dubin & Cleland, 1977; Lindstrom, 1982). The former are responsible for 'feed-forward' the latter for 'feed-back' inhibition. The differential role of these pathways in modulating visual responses is obscure. Responses of l.g.n. cells in our experiments were usually very sustained and lacked the oscillations sometimes seen in the literature (e.g. Virsu, Lee & Creutzfeldt, 1977). After two days, bursting activity often became apparent in the l.g.n. and experiments were terminated. Also, presence of barbiturate in the preparation may encourage such firing patterns; it causes changes in response patterns (Schmidt & Creutzfeldt, 1968) and temporal filtering (Kaplan & Shapley, 1982). Intracellular recordings of Singer & Creutzfeldt (1970) and Singer et al. (1972) were also carried out under barbiturate anaesthesia, perhaps accounting for the pronounced 'secondary inhibition' they observed. Although 'dips' in geniculate firing occurred at the appropriate moment in some of our records, they were not marked in comparison with the prolonged inhibition of retinal input which lasted throughout the stimulus. Responses to moving stimuli In the retina, responses to moving stimuli can be related to receptive field organization of X and Y cells (Rodieck, 1965; Cleland et al. 1983; Cleland & Harding, 1983) in a relatively simple way considering the wide range of stimulus parameters. Although in our results histograms of l.g.n. cells were often modified in comparison with their retinal inputs, in more than half the cell pairs peak firing rates followed a very similar pattern as stimulus parameters were varied. The responses of the l.g.n. cell were attenuated by a constant factor vis-a-vis the ganglion cell, in a very similar way to retino-geniculate transmission in the macaque (Lee, Virsu & Creutzfeldt, 1983), permitting a similar treatment. If peak firing rates (without subtraction of maintained activity) are plotted against one another for members of a pair, a close to linear relation results, as demonstrated in Fig. 12 for the cells of Fig. 8. The slope of the relation gives the degree of attenuation, which is likely to be related to anaesthetic depth or degree of arousal (Singer & Bedworth, 1974). A further point is that the regression lines plotted through the points intersect the x axes at positive values, indicating that a threshold firing rate of the ganglion cells had to be exceeded before l.g.n. impulses were evoked. Maintained firing rates are indicated by crosses,
COMPARISON OF RESPONSES IN RETINA AND L.G.N. 265 whose position illustrates that maintained firing in the l.g.n. was substantially less than in retinal ganglion cells. This result is surprising, since some modification of responsiveness to movement, especially velocity tuning, in the l.g.n. has long been assumed (e.g. Singer & Bedworth, 1973). We would stress that inhibitory effects on l.g.n. cell responses to 200 A 100- en 0@P * @ a 0 Cfl 300 100 200 300 300 B o 0 0)~~~~~~~~~ 100 o 0 *0 0 0 100 200 300 400 500 Ganglion cell firing rate (spikes/s) Fig. 12. A plot of firing rates of retinal ganglion cell against l.g.n. cells for the cells of Fig. 8. A and B were drawn from the data for the X and Y cells respectively; all measurements in Fig. 8 are included. The crosses indicate averaged maintained activity levels. In A, a linear relation holds; a least squares regression line is drawn through the points. In B, responses at the highest contrast are indicated by open circles. They do not fall on the regression line drawn through the other points, as expected due to the greater displacement between curves for ganglion and l.g.n. cell in Fig. 8B at high contrast. flashed stimuli were clearly visible. Also, the actual histograms with moving bars clearly indicated sharpening of l.g.n. response peaks. Indeed, the X on-centre cell pair of Fig. 2A in which this sharpening was very clear, is the same one as in Figs. 8 and 12A. It is possible that moving stimuli activate l.g.n. inhibition less effectively than the flashing spots used, which generally covered a larger retinal area. How far the attenuation of l.g.n. movement responses in cell pairs was due to geniculate inhibition and how far it was due to some property of the retino-geniculate synapse is unclear.
266 B. G. CLELAND AND B. B. LEE Other cell pairs displayed some modification of velocity tuning. For those in which responses at high speeds were attenuated, one might suggest summation in inhibitory interneurones occurs with fast-moving stimuli in much the same way as summation in the surround of ganglion cells (Cleland & Harding, 1983). With slower stimuli, this temporal summation does not occur. All those cell pairs in which a shift to higher velocities was present appeared to receive more than one excitatory afferent in our small sample. If such l.g.n. cells have different properties to those with a single drive, their behaviour with other visual stimuli, such as tests for linearity of spatial summation, would be of interest. With flashed stimuli, however, we did not notice any differences. Previous studies of velocity tuning in the l.g.n. (Hess & Wolters, 1979; Frishman, Schweitzer-Tong & Goldstein, 1983) contain curves plotted in linear or log-linear coordinates, but after transformation to log-log plots many curves shown by these authors could have derived from our sample. Frishman et al. (1983) confirm Cleland & Harding (1983) in their description of velocity tuning of X and Y ganglion cells. However, they suggest that in the l.g.n. differences in velocity tuning of X and Y types are more marked than in the retina, where they mainly result from differences in field size. This is not consistent with our results, which indicate a more complex picture. Changes in velocity tuning occur in l.g.n. but irrespective of whether a cell is X or Y type. The small but reliable differences at the retinal level become blurred in the geniculate. The discrepancy could be due to comparison of cell properties at different eccentricities by Frishman et al. (1983) or by their using a bar width matched to centre size. This influences velocity tuning curves, so any systematic tendency to use different bar widths in retina and l.g.n. would distort the comparison. In any event, the direct comparison provided here indicates that a role for Y cells as 'movement detectors' cannot be based on anything specific in their velocity tuning characteristics. The cellular events underlying the relation between impulse rate in a ganglion cell afferent and its relay cell are difficult to evaluate with extracellular records. One factor will be the membrane conductance change generated by an impulse, another will be the level of inhibition. It may be necessary to consider transmission through the l.g.n. as a function of the pattern of afferent impulses per se as well as a function of inhibition in the nucleus. Nevertheless, the fact that in many cells a constant fraction of afferent impulses can evoke l.g.n. discharges under a variety of stimulus conditions indicates that visual information is transferred in a lawful manner through the nucleus. B. B. Lee gratefully acknowledges support from the Lions NSW-ACT Save Sight Foundation (Australia). REFERENCES BISHOP, P. O., BURKE, W. & DAVIS, R. (1962). The identification of single units in central visual pathways. Journal of Physiology 162, 409-431. CLELAND, B. G. (1983). Sensitivity to stationary flashing spots of the brisk classes of ganglion cells in the cat retina. Journal of Physiology 345, 15-26. CLELAND, B. G., DUBIN, M. W. & LEVICK, W. R. (1971 a). Sustained and transient neurones in the cat's retina and lateral geniculate nucleus. Journal of Physiology 217, 473-496.
COMPARISON OF RESPONSES IN RETINA AND L.G.N. CLELAND, B. G., DUBIN, M. W. & LEVICK, W. R. (1971 b). Simultaneous recording of input and output of lateral geniculate neurones. Nature 231, 191-192. CLELAND, B. G. & HARDING, T. H. (1983). Response to the velocity of moving visual stimuli of the brisk classes of ganglion cells in the cat retina. Journal of Physiology 345, 47-64. CLELAND, B. G., HARDING, T. H. & TULUNAY-KEESEY, U. (1983). Response to length of moving stimuli of the brisk classes of ganglion cells in the cat retina. Journal of Physiology 345, 27-46. DREHER, B. & SANDERSON, K. J. (1973). Receptive field analysis: responses to moving contours by single lateral geniculate neurones in the cat. Journal of Physiology 234, 95-118. DUBIN, M. W. & CLELAND, B. G. (1977). The organization of visual inputs to interneurones of the lateral geniculate nucleus of the cat. Journal of Neurophysiology 40, 410-427. FRISHMAN, L. J., SCHWEITZER-TONG, D. E. & GOLDSTEIN, E. B. (1983). Velocity tuning of cells in dorsal lateral geniculate nucleus and retina of the cat. Journal of Neurophysiology 50, 1393-1414. HESS, R. & WOLTERS, W. (1979). Responses of cells in cat's lateral geniculate nucleus and area 17 to the velocity of moving stimuli. Experimental Brain Research 24, 273-286. HUBEL, D. H. & WIESEL, T. N. (1961). Integrative action in the cat's lateral geniculate body. Journal of Phy8iology 155, 385-398. HUGHES, A. (1981). Population magnitudes and distribution of the major modal classes ofcat retinal ganglion cell as estimated from HRP filling and a systematic survey of the soma diameter spectra for classical neurones. Journal of Comparative Neurology 197, 303-339. KAPLAN, E. & SHAPLEY, R. (1982). What controls information processing in the LGN? Society for Neuroscience Abstracts 8, 405. LEE, B. B., VIRsu, V. & CREUTZFELDT, 0. D. (1977). Responses of cells in the cat lateral geniculate nucleus to moving stimuli at various levels of light and dark adaptation. Experimental Brain Research 27, 51-59. LEE, B. B., VIRsu, V. & CREUTZFELDT, 0. D. (1983). Linear signal transmission from prepotentials to cells in the macaque lateral geniculate nucleus. Experimental Brain Research 52, 50-56. LEE, B. B., VIRsu, V. & ELEPFANDT, A. (1981). The phase of responses to moving gratings in the cat retina and lateral geniculate nucleus. Journal of Neurophysiology 45, 547-577. LINDSTROM, S. (1982). Synaptic organization of inhibitory pathways to principal cells in the lateral geniculate nucleus of the cat. Brain Research 23, 447-453. MASTRONARDE, D. N. (1983a). Correlated firing of retinal ganglion cells I: Spontaneously active inputs to X- and Y-cells. Journal of Neurophysiology 49, 303-324. MASTRONARDE, D. N. (1983b). Subtypes of X relay cells in cat LGN can be distinguished solely by responses to visual stimuli. Investigative Ophthalmology and Visual Science 24, suppl., 265. RODIECK, R. W. (1965). Quantitative analysis of cat retinal ganglion cell response to visual stimuli. Vision Research 5, 563-601. RODIECK, R. W. (1967). Maintained activity of cat retinal ganglion cells. Journal of Neurophysiology 30, 1043-1071. RODIECK, R. W. & STONE, J. (1965). Response of cat retinal ganglion cells to moving visual patterns. Journal of Neurophysiology 28, 833-849. SCHMIDT, R. & CREUTZFELDT, 0. D. (1968). Changes in spontaneous and evoked activity of single neurones of the retina and lateral geniculate body of the cat induced by fractionated injection of Pentobarbitol-Na (Nembutal). Pfluigers Archiv 300, 129-147. SINGER, W. & BEDWORTH, N. (1973). Inhibitory interaction between X- and Y-units in the cat lateral geniculate nucleus. Brain Research 49, 291-307. SINGER, W. & BEDWORTH, N. (1974). Correlation between the effects of brain-stem stimulation and saccadic eye movements on transmission in the cat lateral geniculate nucleus Brain Research 72, 185-202. SINGER, W. & CREUTZFELDT, 0. D. (1970). Reciprocal lateral inhibition of on- and off-centre neurones in the lateral geniculate body of the cat. Experimental Brain Research 10, 311-330. SINGER, W., P6PPEL, E. & CREUTZFELDT, 0. D. (1972). Inhibitory interaction in the cat's lateral geniculate nucleus. Experimental Brain Research 14, 210-226. So, Y. T. & SHAPLEY, R. (1979). Spatial properties of X and Y cells in the lateral geniculate nucleus of the cat and conduction velocities of their inputs. Experimental Brain Research 36, 530-533. TROY, J. B. (1983). Spatio-temporal interaction in neurones of the cat's lateral geniculate nucleus. Journal of Physiology 344, 419-432. 267
268 B. G. CLELAND AND B. B. LEE VIRSU, V., LEE, B. B. & CREUTZFELDT, 0. D. (1977). Dark adaptation and receptive field organization of cells in the cat lateral geniculate nucleus. Experimental Brain Research 27, 35-50. WANG, C., CLELAND, B. G. & BURKE, W. (1985). Synaptic delay in the lateral geniculate nucleus of the cat. Brain Research (in the Press).