J Physiol Volume 508, Number 1, 223-236, April 1, 1998
The Journal of Physiology (1998), 508.1, pp. 223-236
© Copyright 1998 The Physiological Society
Spatio-temporal receptive fields in carp retinal horizontal cells
Osamu Umino and Tomomi Ushio
Department of Information Science, Toho University, Funabashi-shi, Chiba 274, Japan
Received 23 July 1997; accepted after revision 25 November 1997.
 |
ABSTRACT |
- The dynamics of the receptive fields of retinal horizontal cells were examined by applying a spatio-temporal modulated light signal to the retina.
- The spatio-temporal receptive fields of both cone- and rod-driven horizontal cells, estimated through cross-correlation between the modulated light signal and the cells' responses, showed their receptive fields (the space-dependent component) to be reduced in size with time.
- In cone-driven horizontal cells, the reduction in receptive field size was initially small but then rapidly became prominent with time. The time to peak of the time-dependent component of spatio-temporal receptive fields did not depend on the distance from the centre.
- Application of a small amount of Co2+, an agent blocking the cone-driven horizontal cells' feedback action on cones, or GABA, resulted in a reversal of the time-dependent shrinkage of receptive fields to time-dependent expansion.
- In rod-driven horizontal cells, the receptive field shrinkage was slow. The time to peak of the time-dependent component decreased with the distance from the centre.
- Image processing experiments examining the response pattern in the horizontal cell layer (neural image) to a moving square of light showed smudging of the neural image when the time-dependent receptive field expansion was present, while there was essentially no smudging under conditions of receptive field shrinkage.
|
INTRODUCTION |
Natural images impinging on the retina vary with time as well as space. Light signals are converted to electrical signals by photoreceptors, then synaptically transmitted to horizontal cells. Horizontal cells mediate lateral inhibition in the outer retina and thereby contribute to the early processing of spatio-temporal information (Werblin & Dowling, 1969; Kaneko, 1970).
The receptive fields of horizontal cells have conventionally been examined by using specific stimuli such as a spot or slit of light, and the current evidence indicates that the highly symmetrical organization of these cells' receptive fields far exceeds the dendritic arborization of the cell processes (Naka & Rushton, 1967; Lamb, 1976). When light is delivered to the receptive field centre, the light-evoked response at photoreceptors is directly transmitted to the horizontal cell located at the centre, while for peripheral illumination the peripheral response is transmitted to the central horizontal cell via various elements including membrane conductance, coupling conductance, and interactions between photoreceptors and horizontal cells. Thus, because the dynamics of responses recorded at the centre depends on the position of a light stimulus, the resultant receptive fields are expected to depend on time. In fact, measurement of the receptive fields of horizontal cells at different intervals after the onset of a single slit stimulus showed both time-dependent expansion and time-dependent shrinkage of receptive fields (Byzov & Shura-Bura, 1983; Kamermans et al. 1996). Numerical simulations also showed how the elements can affect the dynamics of horizontal cells in response to a single slit of light (Kamermans et al. 1989; Ohshima et al. 1995; Kamermans, Haak, Habraken & Spekreijse, 1996).
The dynamics of receptive fields can be most adequately examined by using a light signal modulated in both space and time (Yasui, Davis & Naka, 1979; Powers & Arnett, 1981; Hida & Naka, 1982). Furthermore, the modulation of illuminance around a mean more nearly approximates the visual environment an animal encounters in nature (Naka, Itoh & Chappell, 1987). Naka and colleagues used a random, travelling light grate to estimate the horizontal cell receptive fields through cross-correlation between the light input and the response output (Chappell, Naka & Sakuranaga, 1985). Their estimated receptive fields provided useful information related to spatio-temporal receptive fields, but gave only an approximation of the stationary receptive fields (Yasui et al. 1979). Thus, in contrast to studies using a single slit or a spot of light, our understanding of the dynamics of receptive fields in horizontal cells using a modulated light is limited.
In this study, we examined the dynamics of receptive fields in both cone-driven and rod-driven horizontal cells by using a light signal modulated simultaneously in both space and time. The spatio-temporal receptive fields of horizontal cells were calculated by cross-correlating the light with the responses of the cells. The spatio-temporal receptive field is an indicator of the way in which the intracellular potential depends on how far in relative retinal distance and how long ago the stimulus has occurred. Our findings indicate that: (1) in both rod- and cone-driven horizontal cells, the spatio-temporal receptive fields are gradually reduced in size with time; (2) in cone-driven horizontal cells the negative feedback action of horizontal cells on cones causes time-dependent shrinkage in their receptive fields, whereas the receptive field shrinkage of rod-driven horizontal cells may reflect response spread involving the rod layer; and (3) the smudging of the neural image which appears in the activity pattern in both the cone- and rod-driven horizontal cell layers for a moving light signal was eliminated by time-dependent shrinkage of receptive fields, thereby 'sharpening' the neural image for further processing in the proximal layers of the retina. Some of these results have appeared in abstract form (Ushio & Umino, 1996).
|
METHODS |
Horizontal cells from the carp retina (Cyprinus carpio) were prepared and measured as described previously (Umino, Maehara, Hidaka, Kita & Hashimoto, 1994; Umino, 1997). Carp were anaesthetized in a solution containing 3-aminobenzoic acid ethyl ester (0·01 %, Sigma) and pithed, and the eyes were excised and the retinae detached under dim red light. Isolated retinae were mounted receptor-side up on Millipore filters with a square hole (
5 mm each side) in the centre, then placed in a plastic chamber; in the receptive field experiments, light stimulated the retina through the square hole (see below). Recordings were made with 4 M potassium acetate-filled glass microelectrodes with resistances in the range of 30-60 M
. In most experiments recordings were conducted from the somata of cone-driven luminosity-type and rod-driven horizontal cells, but in some experiments the electrode penetrated the axon terminals of cone-driven cells. For cone-driven horizontal cell experiments, white diffuse light flashes (1 × 10-8 W cm-2) of 0·5 s duration were delivered every 5 s for a minimum of 10 min to the retinae so that they were moderately light adapted before recording these cells. Cells were identified on the basis of their responses to light stimuli, recording depth, and staining with Lucifer Yellow.
To examine the spatio-temporal receptive fields of horizontal cells, images were displayed onto the retina with a computer-controlled cathode-ray tube monitor (Umino et al. 1994; Umino, 1997). Because horizontal cells show a symmetrical receptive field, their receptive fields can be evaluated using a slit light (Lamb, 1976). In our experiments, the dark/light spatio-temporal random stripe images, in which one image was composed from forty-eight slits (50 µm × 1 cm for one slit on the retina, see Fig. 1), were displayed on the monitor. In this case, the light signal was a function of both time t and space x. The intensity in each slit was modulated by the maximal-length linear binary sequence (m sequence, m = 11), pseudorandom binary noise. The contrast ratio of a bright slit intensity to dim one-in-one stripe image was 100. The change of each stripe image was synchronized with the monitor frame cycle, so that new stripes were presented every 1/60 s without flicker. The m sequence is repetitive, and thus shows a discrete power spectrum. The cyclic period was 35 s in our experiments, yielding a harmonic separation of the spectrum as small as 3 × 10-2 Hz. Thus, the power spectrum of the m sequence used in the present experiments can be regarded as being continuous. The bandwidth of the spectrum was approximately 19·2 Hz (0·32 × 60 Hz) (Roberts, Eng & Davis, 1966), i.e. higher than the frequency bandwidth of cone photoreceptors (Toyoda, 1974). Under the present experimental conditions the response of cone-driven horizontal cells to the modulated light has a flat power spectrum with a cut-off frequency of 4-7 Hz. When the same grate of modulated light was delivered, only a very slow membrane potential fluctuation with an amplitude of less than 1 mV peak to peak was evoked in rod-driven horizontal cells because these cells respond very slowly to temporal change in the light stimulus. Due to the high sensitivity of rod-driven horizontal cells to a light stimulus, the intensity of the modulated light used in the rod-driven horizontal cell experiments was usually 1/1000th of that used in cone-driven horizontal cell experiments. To evoke a larger response fluctuation in rod-driven horizontal cells we lowered the frequency of the change in the modulated grid image from 60 to 20 Hz. The power spectrum of the 20 Hz modulated light was nearly flat up to 6·4 Hz, which fully covered the frequency range (0·5-2 Hz) of rods (Toyoda & Coles, 1975): the cut-off frequency of the power spectrum of rod-driven horizontal cell responses to the modulated light was 0·3-1 Hz. The auto-correlation function of the light signal exhibited a property essentially the same as that of the Dirac delta function, being nearly zero unless its parameters were as follows: time displacement < 16 ms for cone-driven horizontal cells and < 48 ms for rod-driven horizontal cells, and the absolute value of position displacements < 50 µm for both cells. Thus, the stripe stimulus was for all practical purposes a spatio-temporal white-noise signal for retinal neurons. The spatio-temporal receptive field, h1(
, x'), was calculated by cross-correlating the stripes, i(t x), with the response of the cell, r(t), as follows:
h1(, x') = (1/P) r(t) i(t-, x - x') dt dx, (1)
where x' is the position displacement on the retina and is the time displacement but, in this study, x' and are generally referred to as position (on the retina) and time, respectively. P is a constant corresponding to the power spectral density of I, and h1 can be called the first-order spatio-temporal Wiener kernel (Yasui et al. 1979). Note that r is an AC (time-varying fluctuation) response and that i is the space- and time-varying modulation in light intensity. Both the light and the response were digitized off-line at 500 Hz. Cross-correlation was computed for the stationary part of the response after the initial transient hyperpolarization (see Fig. 1), and the data length provided for the analysis was typically 35 s for cone-driven horizontal cells, and 280 s for rod-driven horizontal cells.
Cone-driven horizontal cells receive synaptic input from cones, and in return send their responses to cones (Stell, Lightfoot, Wheeler & Leeper, 1975). The feedback pathway from cone-driven horizontal cells to cones is mediated by GABA (Schwartz, 1982). It has been reported that the GABAergic feedback pathway is blocked by the GABA antagonists bicuculline and picrotoxin (Murakami, Shimoda, Nakatani, Miyachi & Watanabe, 1982), indicating that the GABAA receptor may be involved in the feedback pathway. In our carp experiments, however, neither agent eliminated the depolarizing response to red light in carp biphasic chromaticity (RG)-type horizontal cells (Umino, Watanabe & Hashimoto, 1989), which is known to be produced by the feedback action from luminosity-type horizontal cells (Stell, Lightfoot, Wheeler & Leeper, 1975; Murakami et al. 1982). Furthermore, questions as to the involvement of the GABAA receptor in the feedback pathway were also recently raised by Verweij, Kamermans, Aker & Spekreijse (1996). To block the feedback pathway, the following two agents were used in the present study; a small amount of Co2+ (50 µM) or GABA (0·5 mM). A small amount of Co2+ can block the feedback pathway by suppressing the GABA-induced current at the cone postsynaptic region (Kaneko & Tachibana, 1986). The blocking actions of both agents were experimentally confirmed by disappearance of the depolarizing response to red light in RG-type horizontal cells in solutions containing one of the two agents (Umino et al. 1989).
|
RESULTS |
Cone-driven horizontal cells
When the grate of modulated light, composed of forty-eight dark/light alternating slits, was applied to the retina, the horizontal cells rapidly hyperpolarized, then gradually depolarized and reached a steady state; the membrane potential of horizontal cells showed a sharp fluctuation with time in response to the light-grate stimuli (Fig. 1). The spatio-temporal receptive fields, estimated by cross-correlating the light signal with the response of the cell, are plotted on the plane composed of the space (x', position displacements in millimetres on the retina) and the time displacement (, in milliseconds), and are viewed from three points in Fig. 2.
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 1. Cone-driven luminosity-type horizontal cell response to light modulated randomly in space and time
When the modulated light was delivered to the retina (the downward arrow), the horizontal cells showed a rapid hyperpolarization followed by a slow declining phase after which they reached a stationary state. Horizontal cells showed a sharp membrane potential fluctuation in response to the modulated light. A portion of the membrane potential fluctuation was recorded at a higher sweep speed (upper trace). Lower stripes indicate the initial part of the modulated light; each image, composed of forty-eight narrow slits (50 µm × 1 mm for one slit on the retina) the intensity of which, modulated in m-sequence fashion, was sequentially delivered as indicated by the numbers 1, 2, 3, etc. at intervals of 1/60 s. The light intensities were 0·35 × 10-4 µW cm-2 for dark slits and 0·2 × 10-2 µW cm-2 for light slits.
|
The spatio-temporal receptive field describes the response elicited by a slit of light presented at position x' and = 0, thereby showing how the horizontal cells gathered signals in space and time. Viewed from the space axis (Fig. 2B), which is a space-dependent component and is the traditional view of receptive fields (Lamb, 1976), the spatio-temporal receptive field decays almost exponentially with distance from either side of the centre (0 mm) and approaches zero at 0·5 mm. When viewed from the time axis (Fig. 2C), which is a time-dependent component and corresponds to the responses of the cell to a slit of light, the spatio-temporal receptive field begins to increase its amplitude at 20 ms, peaks at 70 ms, and returns to nearly the original dark level within a time span of 130 ms at all position displacements. As will be shown later, the time to peak of the time-dependent component of the spatio-temporal receptive fields was nearly constant for all position displacements. The late rebound of depolarization (downward in this plot), which would be due to the feedback action from horizontal cells to cones (Lam, Lasater, & Naka, 1978; Ohshima, Yagi & Funahashi, 1995), was seen in the spatio-temporal receptive field at the centre at 170-270 ms.
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 2. Estimated spatio-temporal receptive fields of cone-driven luminosity-type horizontal cells
The spatio-temporal receptive field, h1, calculated by cross-correlating the response with the stripes (see Methods), was plotted on the plane composed of the space axis (x', position displacements in mm on the retina) and the time displacement axis ( in ms), and then viewed from three different angles (A, B, C). Plots in B correspond to the traditional receptive fields, while plots in C correspond to the cell's responses to a slit flash. Note that horizontal cells showed a hyperpolarizing response to light, but the spatio-temporal receptive field was plotted with the negative values upward. Small fluctuations on the plane were caused mainly by the electrical noise included in the response recorded. Although no absolute value is shown, the ordinate for the spatio-temporal receptive field is in units of V × unit area W-1 s-1.
|
The time course of the receptive fields, the space-dependent component in spatio-temporal receptive fields, was examined in detail. In Fig. 3A, the relative receptive fields (one side) at various times were superimposed on a log scale, while in Fig. 3B, the top view of relative receptive fields was plotted on a contour map. As shown, the receptive fields were gradually reduced in size with time displacement. The shrinkage was less evident near the receptive field centre (0 mm), but became prominent as the distance from the centre increased; in the peripheral region the slope of the receptive field in which the response fell to e-1 was 140 µm at 70 ms and, surprisingly, only 30 µm at 120 ms (Fig. 3A), which was much smaller than the smallest space length constant value of 100 µm for horizontal cells reported to date (Lamb, 1976; Umino, 1997). Similar shrinkage of spatio-temporal receptive fields was seen in seventeen cone-driven horizontal cells. The decrease in the space length constant observed in seven spatio-temporal experiments showing a time to peak of approximately 70 ms is shown in Fig. 4. Since the receptive field profiles changed significantly with time, the receptive fields were also evaluated according to band width. From 50 to 100 ms, the 50 % receptive field, which shows the width at 50 % of its peak receptive field amplitude, decreased from (in mm) 0·29 ± 0·07 to 0·22 ± 0·07, and the 20 % receptive field from 0·64 ± 0·18 to 0·43 ± 0·1 (n = 7).
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 3. Shrinkage of spatio-temporal receptive field with time displacement
Cone-driven luminosity-type horizontal cells. A, receptive fields, the space-dependent component of the spatio-temporal receptive field, at various time displacements were normalized with their maximum value at the centre (0 mm) and were superimposed logarithmically; note that one side of each receptive field was plotted. Numbers near the plots indicate the time displacement in ms; the spatio-temporal receptive field has its peak at 70 ms (see Fig. 2C). Time displacement (ms): ×, 50;  , 70 (peak);  , 90;  , 100;  , 120. A straight line was drawn to fit the peripheral part of the three receptive fields, and their slopes correspond to a length constant of 144 µm at 70 ms (peak), 59 µm at 100 ms, and 30 µm at 120 ms. B, a counter map showing the time courses of relative receptive fields. As in A, individual receptive fields at different times were normalized with their peak value at the centre (0 mm) and viewed from the top. In other words, this illustration corresponds to a top view of Fig. 2A, but with each receptive field normalized using the receptive field centre values. Numbers near the traces indicate the value (in %) of receptive fields relative to their maximum amplitudes at the centre (0 mm). The dashed line marked 100 % indicates the receptive field centre. Data from Figs 1 and 2.
|
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 4. Dynamics of the sizes of receptive fields
Cone-driven luminosity-type horizontal cells. Receptive field size was evaluated from the slope of the space-dependent components of spatio temporal receptive fields. The slope values were measured by fitting an exponential function to the data at the periphery of the space-dependent components (see Fig. 3A) and were plotted as a function of the time displacement. n = 11.
|
Similar spatio-temporal receptive field experiments were conducted in solutions containing either a small amount of Co2+ (50 µM) or GABA (0·5 mM). In these solutions, the membrane potential fluctuated in response to the gate stimulus (not shown), a result quite similar to that for the control shown in Fig. 1. However, the cross-correlation between responses and the grate stimulus revealed an obvious difference in their spatio-temporal receptive fields. The results of the Co2+ experiments are plotted in Fig. 5. As shown, in contrast to the time-dependent shrinkage of the receptive fields in normal solution, those in Co2+-containing solution gradually expanded with time. The expansion of receptive fields became prominent as the distance from the centre increased. From 50 to 100 ms, the 50 % receptive field decreased (in mm) from 0·35 ±0·06 to 0·37 ± 0·15, and the 20 % receptive field from 0·62 ± 0·22 to 0·84 ± 0·19 (n = 8). In contrast to a minimal change in the time to peak of the time-dependent component of the spatio-temporal receptive fields in normal solution, those in Co2+-containing solution gradually increased as the position displacements from the centre increased (Fig. 6). The late rebound of depolarization seen in normal solution (Fig. 2) was not produced in Co2+-containing solution. Application of 0·5 mM GABA also caused a gradual expansion of spatio-temporal receptive fields. Furthermore, the time to peak of the time-dependent component of spatio-temporal receptive fields gradually increased as the distance from the centre increased (n = 4). Thus, the effect of GABA on the spatio-temporal receptive fields was essentially the same with that of Co2+.
Next, spatio-temporal receptive field experiments were conducted on the axon terminals of horizontal cells, which have been shown to form a receptive field separate from that of the somata (Teranishi, 1983) and to have no synaptic feedback action on cones. Interestingly, the receptive fields of horizontal cell axon terminals did not shrink, but rather expanded slightly (n = 5). The time to peak of the time-dependent component of spatio-temporal receptive fields increased with the distance from the centre (85 ± 3 ms at the centre and 92 ± 5 ms at the periphery, 0·7 mm from the centre).
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 5. Expansion of spatio-temporal receptive fields with time displacement
Cone-driven luminosity-type horizontal cells. Experiments were conducted in a Co2+ (50 µM)-containing solution, which blocks the feedback signal transmission from horizontal cells to cones (Umino et al. 1989). A, receptive fields at various times were normalized with their maximum value at the centre (0 mm) and were superimposed logarithmically. Time displacement (ms) is indicated by the numbers near the symbols. The peak of the spatio-temporal receptive field was at 80 ms. The slopes of the straight lines correspond to a length constant of 93 µm at 20 ms, 144 µm at 80 ms (peak), and 223 µm at 140 ms. B, a counter plot showing the dynamics of relative receptive fields. Plots as in Fig. 3B.
|
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 6. The time to peak of the time-dependent component of the spatio-temporal receptive field as a function of the position displacements
The mean of the time to peak was plotted. Cone-driven luminosity-type horizontal cells. A, in normal solution. n = 7. B, in a Co2+-containing solution. n = 8. The standard deviation at each position was within 15 ms.
|
Rod-driven horizontal cells
The response of rod-driven horizontal cells to the grate of modulated light is illustrated in Fig. 7. In the experiments, the change of the modulated grid image was lowered to 20 Hz (see Methods). As shown, the membrane potential of rod-driven horizontal cells fluctuated very slowly in response to the modulated light. The estimated spatio- temporal receptive fields of rod-driven horizontal cells is shown in Fig. 8. In comparison with the spatio-temporal receptive fields of cone-driven horizontal cells, the spatio-temporal receptive fields decay faster with the distance from the centre and thus have narrower receptive fields (Fig. 8B). Furthermore, the dynamics of the spatio-temporal receptive fields were very slow and had a peak at approximately 285 ms (Fig. 8C). As to these dynamics, after the peak the spatio-temporal receptive fields returned to the original dark level very slowly (Fig. 8C), and the late rebound of depolarization, which appeared in the spatio-temporal receptive fields of cone-driven horizontal cells (Fig. 2), was not seen. The receptive field size of spatio-temporal receptive fields of rod-driven horizontal cells also decreased with time (Fig. 9). In comparison with the peripheral-prominent decay of receptive fields in cone-driven horizontal cells (Fig. 3), rod-driven horizontal cells showed a decay in all locations in the receptive field (Fig. 9). The spatio-temporal receptive fields of rod-driven horizontal cells sometimes showed an initial increase in receptive field size (arrows in Fig. 9B), which was not observed in the spatio-temporal receptive fields of cone-driven horizontal cells.
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 7. Rod-driven horizontal cell response to light modulated randomly in space and time
The modulated light was delivered to the retina at the point indicated by the downward arrow. Lower stripes indicate the initial part of the modulated light; each image, composed of 48 narrow slits (50 µm × 1 mm for one slit on the retina) the intensities of which were modulated in m- sequence fashion, was sequentially delivered as indicated by the numbers 1, 2, 3, etc. at intervals of 1/20 s. The light intensities were 0·35 × 10-7 µW cm-2 for dark slits and 0·2 × 10-5 µW cm-2 for light slits.
|
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 8. Estimated spatio-temporal receptive fields of rod-driven horizontal cells
The spatio-temporal receptive field was plotted on the plane composed of the space axis (x', position displacements in millimetres on the retina) and the time displacement axis ( in ms), and was viewed from three different angles (A, B, C). For further explanation, see Fig. 2.
|
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 9. Shrinkage of spatio-temporal receptive field with time displacement. Rod-driven horizontal cells
A, receptive fields, the space-dependent component of the spatio-temporal receptive fields, at various time displacements were normalized with their maximum value at the centre (0 mm) and were superimposed logarithmically; note that one-side of each receptive field was plotted. Numbers near the plots indicate the time displacement in ms; the spatio-temporal receptive field has its peak at 290 ms (see Fig. 8). Numbers near the plots indicate the time displacement in ms. Data from Fig. 8. B , a counter map showing the time courses of relative receptive fields. As in A, individual receptive fields at different times were normalized with their peak value at the centre (0 mm) and viewed from the top. Numbers near the traces indicate the value (in %) of receptive fields relative to their maximum amplitudes at the centre (0 mm). The dashed line marked 100 % indicates the receptive field centre. For further explanation, see Fig. 3.
|
The initial rising phase of the time-dependent component of the spatio-temporal receptive fields in the periphery was relatively steep as compared with that of the component at the centre (Fig. 10A), and as a result the time to peak decreased with the distance from the centre (Fig. 10B). The slow shrinkage of receptive fields and the shortening of the time to peak in the time-dependent component of spatio-temporal receptive fields were consistently observed in five cells. From 300 to 500 ms the 50 % receptive field decreased from 0·15 ± 0·3 to 0·12 ± 0·3 mm. The mean time to peak was 298 ± 34 ms, and the mean shortening of the time to peak at the periphery 0·2 mm away from the centre was 8 %.
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 10. Changes in the time-dependent components of spatio-temporal receptive fields. Rod-driven horizontal cells
Data from Fig. 8. A, superimposed components at the centre (continuous line) and periphery, 0·15 mm from the centre (dotted line). B, the time to peak of the time-dependent component plotted as a function of the distance from the centre.
|
Image processing experiments
To examine the functional role of the observed shrinkage of receptive fields, image processing experiments were conducted. In this case, the input was a moving square pattern (Fig. 11A) while the output was a pattern of activity (a neural image) at the horizontal cell layer (Fig. 11Bb and Cb) estimated by the convolution of the input pattern (Fig. 11A) with the theoretical spatio-temporal receptive field (Fig. 11Ba for the time-dependent shrinkage of the spatio-temporal receptive field and Fig. 11Ca for the time-dependent expansion of the spatio-temporal receptive field). The time to peak of the time-dependent component of the theoretical spatio-temporal receptive fields did not depend on the distance from the centre, such that the theoretical spatio-temporal receptive fields mimicked the spatio-temporal receptive fields of cone-driven horizontal cells. When the square pattern was presented (at 0 ms in Fig. 11A), no response was seen in the neural image of the horizontal cell layer because of the delayed spatio-temporal receptive field response, as shown in Fig. 11Bb and Cb, where the dark region represents horizontal cell membrane potential in the dark. However, 100 ms later, the input square pattern is shifted slightly to the right and the horizontal cell responses are manifested as a white area in the filtered image; the pattern of horizontal cell activity is blurred because the horizontal cells have a large receptive field. Virtually no difference was seen at this time between the filtered images with and without feedback. When the input square was moved further (200, 300, and 400 ms), similar blurred images of horizontal cell activity were seen in the filtered image as the spatio-temporal receptive fields shrank with time (Fig. 11Bb), while as the spatio-temporal receptive fields expanded with time the smudging (after image) of the neural image of horizontal cell activity became pronounced (Fig. 11Cb).
|
 |
View larger version
[in this window]
[in a new window] |
|
|
Figure 11. Image processing experiments examining the activity pattern of the horizontal cell layer in response to a moving square
A, input pattern of one white moving square. One image was 100 × 100 pixels and each image was connected. A similar connecting of images was conducted for the neural images, Bb and Cb. B, control. Ba, theoretical spatio-temporal receptive field reduced in size with time. The theoretical spatio-temporal receptive field, F(t, x, y), is represented by:
F(t, x, y) = (t/tpeak)n exp(-t n/tpeak) exp(-(x2 + y2)/ 2),(2)
tpeak = 70, for t < tpeak,
tpeak = 70 - 0·1(exp(0·5 (x2 + y2)0·5) - 1), for t >= tpeak,
where, n is 2 and is 5 pixels. The size of F is 20 × 20 pixels. tpeak is the time (ms) to the peak of F. Bb, neural image of the horizontal cell layer. The dark area represents the membrane potential of horizontal cells in darkness, while the white regions are the horizontal cell activities of hyperpolarizing responses to the moving square in A. Neural images, H(t, x, y), were calculated by convolution of the input signal (A), M(t, x, y), with the theoretical spatio-temporal receptive field (Ba), F (t, x, y), as follows:
H(t, x, y) = M(t - , x - x', y-y') F (, x', y') d dx'dy'. (3)
Neural activity (white area) was not seen at 0 ms because of the delay in the horizontal cell response to the input square. Activity patterns appeared in the neural image from 100 ms. The neural activity of white regions was blurred because of the large receptive fields of the horizontal cells. C, spatio-temporal receptive field expanded with time. Ca, theoretical spatio-temporal receptive field, which can be computed from eqn (2) except for the following parameter:tpeak = 70 + 0·1(exp(0·5(x2 + y2)0·5) - 1), for t >= tpeak. (4)Cb, neural image of the horizontal cell layer. Neural activity was nearly identical to that for the control at 100 ms (Bb), while the smudging became evident at 200, 300 and 400 ms.
|
|
DISCUSSION |
In this study, we showed that the feedback action from horizontal cells onto cones shrinks the horizontal cell receptive fields with time displacement. Since horizontal cells possess voltage-dependent membrane conductance (Byzov, Trifonov, Yu, Chailahian & Golubtzov, 1977), it could be argued that they effect the shrinkage of the receptive fields of horizontal cells. In our spatio-temporal receptive field experiments, the horizontal cell membrane potential typically fluctuated within a small range of ±3 mV. Furthermore, time-dependent shrinkage of horizontal cell receptive fields also occurred when the membrane potential fluctuated within a smaller range of ±1 mV. Within such a small membrane potential range, the horizontal cell response to a light stimulus is known to be nearly linear (Chappell, Naka & Sakuranaga, 1985). In fact, the mean square error between the predicted response from h1 and the real response (Chappell et al. 1985) was less than 10 % in our experiments. Furthermore, in spatio-temporal receptive fields the attenuation was rather small in the central region but quite prominent in the peripheral region (Fig. 3), a profile very different from those described by Lamb (1976) and Umino (1997), in studies in which changes occurred at all positions in the receptive fields, and the receptive field change was explained by the voltage dependence of membrane conductance. Thus, the voltage dependency of membrane conductance does not appear to be involved, at least as a major factor, in the time-dependent shrinkage of horizontal cell receptive fields.
Byzov & Shura-Bura (1983) and Kamermans Haak, Habraken & Spekreijse (1996) examined the dynamics of horizontal cell receptive fields by measuring the receptive field size at different intervals after the onset of a single slit flash. They showed that the receptive field size initially increases sharply, then decreases with time. In our experiments, however, the initial expansion of receptive fields was not observed (Fig. 3). In an effort to explain this difference, we will first discuss possible mechanisms for the initial increase in the size of the receptive field. The simplest explanation for this initial increase is based on the low-pass filter property of the horizontal cell layer. When a light is delivered to the region outside the receptive field centre, the light-evoked response there is transmitted to the central horizontal cell in the horizontal cell layer. An isolated horizontal cell is represented as parallel resistor-capacitance circuits, and thus has a low-pass filter property (Johnston & Lam, 1981). As was shown by numerical simulations, after transmission through such low-pass horizontal cells, the response slows (Ohshima et al. 1995) such that the receptive field is initially narrow but then gradually expands (Kamermans et al. 1996). This explanation also predicts the delay in the response peak with increasing distance from the centre. Electrophysiological experiments using an intact retina, however, showed that the relatively rising phase of the peripheral response was almost the same as that of the central response and no difference was seen in the time to peak between the central and the peripheral response (Fig. 6A; Ohshima et al. 1995). Rather, Byzov & Shura-Bura (1983) observed a decrease in the time to peak of the response with increasing distance from the centre in turtle horizontal cells. Thus, the low-pass filter property is unlikely to be a major mechanism producing the initial receptive field expansion. Because receptive field expansion was seen in a solution containing either a small amount of Co2+ (Fig. 5) or GABA and because the time to peak of the time-dependent component of the spatio-temporal receptive fields increased with the distance from the centre in these solutions (Fig. 6), the low-pass filter property of horizontal cells might be masked in the intact retina by the feedback action from horizontal cells to cones. An alternative explanation for the initial expansion of the receptive fields is that this expansion is caused by increased membrane resistance. Experimentally, a transient increase in membrane potential was demonstrated by injecting a hyperpolarizing current into the cell while recording the membrane potential change from the neighbouring horizontal cells (Byzov & Shura-Bura, 1983). The membrane resistance can also explain why the receptive field expansion was not seen in the initial rising phase in the present experiment. That is, in contrast to the relatively large response, up to 10 mV, having a plateau phase in the study by Byzov & Shura-Bura (1983), the response in our experiment was small, typically 3 mV peak to peak. Thus, in our experiments the transient increase in membrane resistance would be small. In the spatio-temporal experiments, the membrane potential fluctuation includes responses to increments and decrements in the modulated light around the mean. The transient increase in membrane resistance was not seen with depolarization of the membrane potential, i.e. a decremental response (Byzov & Shura-Bura, 1983). Therefore, this property of membrane resistance can also contribute to minimizing the change in receptive field size in our spatio-temporal experiments. Another difference is that values of space length constants in their slit flash experiments were in the 0·3-2·5 mm range, while the minimum value of the slope of the space-dependent component in our spatio-temporal receptive field study was as low as 30 µm. The receptive field size of gold fish horizontal cells reported by Kamermans et al. (1996) was ranged from 0·5 to 2 mm, values larger than those of the typical receptive fields of teleost horizontal cell somata and close to the receptive field size of horizontal cell axon terminals (Teranishi, 1983; Umino, 1997). As mentioned above, the receptive fields of horizontal cell axon terminals expanded with time.
In solution containing a small amount of Co2+ or GABA, the shrinkage of receptive fields was blocked. A small amount of Co2+ was shown to block the feedback pathway by suppressing the GABA-induced current at the cone postsynaptic region (Kaneko & Tachibana, 1986; also see Umino et al. 1989) while externally applied GABA would occupy the GABA receptors on the cone terminals preventing endogenous GABA released from horizontal cells from binding their receptors (Murakami et al. 1982). Thus, both experiments strongly suggest the involvement of the GABAergic feedback pathway in receptive field shrinkage. This suggestion is consistent with electrophysiological studies (Verweij et al. 1996) and numerical simulation studies of responses to a single slit flash (Kamermans et al. 1989; Ohshima et al. 1995; Kamermans et al. 1996; Ushio & Umino, 1996). The spatio-temporal receptive fields in normal solution had the following two characteristics. First, receptive fields initially decreased in size very slowly but then shrank markedly with time (Figs 3 and 4). Thus, the shrinkage occurs with delay. Second, the decay in the response distribution was less prominent in the region near the receptive field centre, but was prominent at the periphery. The size of the region where the decay was small was approximately 0·1 mm, a value roughly corresponding to the morphological size of one horizontal cell. These two characteristics can be explained by the feedback pathway as follows. When a flash is delivered, nearly the same response is evoked in the central and peripheral region of the cell's receptive field. In the centre, the light-evoked response of horizontal cells feedbacks onto cones with a delay. The feedback action accelerates the recovery depolarizing phase after the peak (Baylor, Fuortes & O'Bryan, 1971), but the change in the recovery phase would be virtually the same in the region of one central horizontal cell, such that the receptive fields measured there would be nearly the same. The peripheral response is transmitted to the centre with a very small delay via electrical coupling between horizontal cells, but the horizontal cells also negatively feed their responses back to the cones with a significant delay (Baylor et al. 1971; Umino & Hashimoto, 1991) such that the horizontal cell response amplitudes are attenuated with a delay. The attenuation becomes prominent for the peripheral response far from the centre because the number of negative feedback signals increases, resulting in a corresponding decrease in receptive field size with the distance from the centre as well as with time.
In rod-driven horizontal cells, a small receptive field expansion was initially seen in the spatio-temporal receptive fields (arrows in Fig. 9B). A similar receptive field expansion was also reported in slit-displacement experiments (Kamermans et al. 1996), though the expansion was much more prominent in the slit displacement than in the spatio-temporal receptive field experiments. The time to peak in the time-dependent component of the spatio-temporal receptive fields decreased with the distance from the receptive field centre. Thus, as discussed above, a low-pass filter property of rod-driven horizontal cells does not appear to be a major source of receptive field expansion (see below). The alternative explanation is that, as in the cone-driven horizontal cells mentioned above, the transient increase in membrane resistance enlarges the receptive field in a time-dependent manner.
The major observation on the spatio-temporal receptive fields of rod-driven horizontal cells was a gradual shrinkage of receptive fields with time. Rod-driven horizontal cells receive input from rod photoreceptors. In turtle rods, Detwiler, Hodgkin & McNaughton (1978) used slit-displacement experiments to demonstrate that the response in unilluminated rods became faster as the distance from the source increased, and that rods accelerate the initial hyperpolarizing phase to the peak, as well as the recovery depolarizing phase after the peak, such that the peak occurs more quickly in the periphery. These reported properties of rods are qualitatively very similar to the observations in the spatio-temporal receptive fields of rod-driven horizontal cells. Thus, the response spread in the rod-network may account for the gradual receptive field shrinkage in rod-driven horizontal cells. Quantitatively, however, the space-dependent shortening of the time to peak of responses in the rod-driven horizontal cell layer is less prominent than that in the rod layer. From the centre to the periphery (0·1 mm from the centre), the time to peak of the horizontal cells decreased from 285 to 265 ms (Fig. 10) but in rods from 1·2 to 0·65 s (Detwiler et al. 1978). The receptive field shrinkage is also less prominent in horizontal cells. Thus, an additional mechanism, such as a low-pass filter and/or increased membrane resistance, would be required to blunt the shortening of the time to peak and the receptive field shrinkage. No evidence has been reported for a feedback action of rod-horizontal cells on rods.
The spatio-temporal receptive fields of cone-driven horizontal cells shrank dramatically with time, while in rod-driven horizontal cells the shrinkage of receptive fields was slow and slight. Cone-driven horizontal cells have large receptive fields due to electrical coupling (Kaneko, 1971). As shown in the experiments using either Co2+ (Fig. 5) or GABA (also, see Lam et al. 1978), without the feedback action, the response passing through the horizontal cell layer slows down. Thus, although electrical coupling is an ingenious mechanism by which horizontal cells attain a large receptive field, the low-pass filter property of the horizontal cell membrane expands the receptive fields with time. In rod-driven horizontal cells, the shrinkage was slow and less prominent than that of cone-driven horizontal cells. Furthermore, the spatio-temporal receptive field response lasts longer than that of cone-driven horizontal cells (Figs 2 and 8). In comparison with cone-driven horizontal cells, the coupling between rod-driven horizontal cells is minimal (Teranishi, Negishi & Kato, 1984) and, as discussed above, the receptive field shrinks with time at the rod-layer. Thus, the requirement to reduce receptive fields in size may be lower in rod-driven horizontal cells than in cone-driven horizontal cells.
The observed time-dependent shrinkage of receptive fields indicates that neurons exhibiting such behaviour are initially affected by light signals over a large area on the retina, whereas the effects become more localized with time. Detwiler et al. (1978) proposed that such a time-dependent shrinkage of receptive fields in the rod network may have evolutionary advantages when objects are viewed under conditions of very dim light. In the present study, the time-dependent shrinkage of receptive fields in cone-driven horizontal cells was seen under photopic conditions. Therefore, such advantages are not applied to these cells. Although neural images in the present image processing experiments were developed for the network of cone-driven horizontal cells, the possibility of eliminating neural-image smudging can be extended to the study of all neurons, processing time-dependent shrinkage of receptive fields, including turtle rods and both carp rod-driven and cone-driven horizontal cells.
Based on the present results, the feedback action from cone-driven horizontal cells to cones eliminates smudging of the neural image which appears in the horizontal cell layer in response to the moving pattern, by shrinking the receptive field size with time. Coupling conductance also modifies the horizontal cell receptive field size (Lamb, 1976). That is, the increased coupling conductance expands the steady-state receptive fields while the increased membrane conductance shrinks the steady-state receptive fields. Numerical simulations showed that when coupling conductance is increased, the response of horizontal cells to light becomes faster and the response after the peak returns to the original level is even more rapid (Ohshima et al. 1995), a phenomenon attributable to the coupling conductance-induced decrease in the input impedance (Poznanski & Umino, 1998). Thus, in contrast to the feedback action which eliminate smudging by decreasing the receptive field size, the coupling conductance can decrease the smudging with accompanying expansion of the steady-state receptive fields. Coupling conductance is high in the dark while being low in the light-adapted retina (Shigematsu & Yamada, 1988; Baldridge & Ball, 1991; Umino, Lee & Dowling, 1991; Kurz-Isler, Voigt & Wolburg, 1992). Low coupling conductance in the light shrinks the steady-state receptive fields allowing the details of patterns to be detected. In this case, smudging would appear in the neural image, but because the feedback pathway is active in the light (Kirsch, Wagner & Djamgoz, 1990) such smudging is possibly reduced by the feedback. Based on numerical simulations it was also shown that the decreased coupling conductance between horizontal cells accelerates the recovery phase of the response to the basal level in cones (Ohshima et al. 1995). The signal in the outer retina is transmitted to the inner retina via bipolar cells. The central receptive field of bipolar cells is provided by photoreceptors while their surrounding receptive field response is from horizontal cells. Based on the above discussion, in the photopic state smudging in central receptive fields is reduced by decreasing the coupling conductance between cone-driven horizontal cells while smudging in the surrounding receptive fields is eliminated by the feedback action from cone-driven horizontal cells to cones. In the scotopic state, the receptive field sizes of both rods and rod-driven horizontal cells are reduced with time such that smudging in both the centre and the surround of the receptive fields is reduced in bipolar cells.
|
REFERENCES |
Baldridge, W. H. & Ball, A. K. (1991). Background illumination reduces horizontal cells receptive field size in both normal and 6-hydroxydopamine-lesioned goldfish retinas. Visual Neuroscience 7, 441- 450.
[Medline]
Baylor, D. A., Fuortes, M. G. F. & O'Bryan, P. M. (1971). Receptive fields of single cones in the retina of the turtle. The Journal of Physiology 214, 265-294.
[Medline]
Byzov, A. L. & Shura-Bura, T. M. (1983). Spread of potentials along the network of horizontal cells in the retina of the turtle. Vision Research 23, 389-397.
[Medline]
Byzov, A. L., Trifonov, Yu, A., Chailahian, L. M. & Golubtzov, K. W. (1977). Amplification of graded potentials in horizontal cells of the retina. Vision Research 17, 265-273.
[Medline]
Chappell, R. L., Naka, K.-I. & Sakuranaga, M. (1985). Dynamics of turtle horizontal cell response. Journal of General Physiology 86, 423-453.
[Abstract]
Detwiler, P. B., Hodgkin, A. L. & McNaughton, P. A. (1978). A surprising property of electrical spread in the network of rods in the turtle's retina. Nature 274, 562-565.
[Medline]
Hida, E. & Naka, K.-I. (1982). Spatio-temporal receptive fields as revealed by spatio-temporal random noise. Zeitschrift fur Naturforschung 37C, 1048-1049.
Johnston, D. & Lam, D. M.-K. (1981). Regenerative and passive membrane properties of isolated horizontal cells from a teleost retina. Nature 292, 451-454.
[Medline]
Kamermans, M., Haak, J., Habraken, J. B. A. & Spekreijse, H. (1996). The size of the horizontal cell receptive fields adapts to the stimulus in the light adapted goldfish retina. Vision Research 36, 4105-4119.
[Medline]
Kamermans, M., Van Dijk, B. W. & Spekreijse, H. (1989). Lateral feedback from monophasic horizontal cells to cones in carp retina. II. A quantitative model. Journal of General Physiology 93, 695-714.
[Abstract]
Kaneko, A. (1970). Physiological and morphological identification of horizontal, bipolar, and amacrine cells in the goldfish retina. The Journal of Physiology 207, 623-633.
[Medline]
Kaneko, A. (1971). Electrical connexions between horizontal cells in the dogfish retina. The Journal of Physiology 213, 95-105.
[Medline]
Kaneko, A. & Tachibana, M. (1986). Blocking effects of cobalt and related ion on the 
-aminobutyric acid-induced current in turtle retinal cones. The Journal of Physiology 373, 463-479.
[Abstract]
Kirsch, M., Wagner, H. J. & Djamgoz, M. B. A. (1991). Dopamine and plasticity of horizontal cell function in the teleost retina: regulation of a spectral mechanism through D1-receptors. Vision Research 31, 401-412.
[Medline]
Kurz-Isler, G., Voigt, T. & Wolburg, H. (1992). Modulation of connexon densities in gap junctions of horizontal cell perikarya and axon terminals in fish retina: effects of light/dark cycles, interruption of the optic nerve and application of dopamine. Cell Tissue Research 268, 267-275.
[Medline]
Lam, D. M. K., Lasater, E. M. & Naka, K.-I. (1978). -Aminobutyric acid: A neurotransmitter candidate for cone horizontal cells of the catfish retina. Proceedings of the National Academy of Sciences of the USA 75, 6310-6313.
[Medline]
Lamb, T. D. (1976). Spatial properties of horizontal cell responses in the turtle retina. The Journal of Physiology 263, 257-286.
[Medline]
Murakami, M., Shimoda, Y., Nakatani, K., Miyachi, E.-I. & Watanabe, S.-I. (1982). GABA-mediated negative feedback from horizontal cells to cones in carp retina. Japanese The Journal of Physiology 32, 911-926.
[Medline]
Naka, K.-I., Itoh, M.-A. & Chappell, R. L. (1987). Dynamics of turtle cones. Journal of General Physiology 89, 321-337.
[Abstract]
Naka, K.-I. & Rushton, W. A. H. (1967). The generation and spread of S-potentials in fish (Cyprinidae). The Journal of Physiology 192, 437-461.
[Medline]
Ohshima, S., Yagi, T. & Funahashi, Y. (1995). Computational studies on the interaction between red cone and H1 horizontal cell. Vision Research 35, 149-160.
[Medline]
Powers, R. L. & Arnett, D. W. (1981). Spatio-temporal cross-correlation analysis of catfish retinal neurons. Biological Cybernetics 44, 179-196.
Poznanski, R. & Umino, O. (1998). The effects of dendro-dendritic gap junctions on cable parameters of neurons. In Modeling in the Neurosciences - From Ionic Channels to Neural Networks, ed. Poznanski, R. Harwood Academic Publishers (in the Press)
Roberts, P. D., Eng, B. & Davis, B. Sc. (1966). Statistical properties of smoothed maximal-length linear binary sequences. Proceedings of the IEE 113, 190-196.
Schwartz, E. A. (1982). Calcium-independent release of GABA from isolated horizontal cells of the toad retina. The Journal of Physiology 323, 211-227.
[Abstract]
Shigematsu, Y. & Yamada, M. (1988). Effects of dopamine on spatial properties of horizontal cell responses in the carp retina. Neuroscience Research 8, S69-80.
Stell, W. K., Lightfoot, D. O., Wheeler, T. G. & Leeper, H. F. (1975). Goldfish retina: Functional polarization of cone horizontal cell dendrites and synapses. Science 190, 989-990.
[Medline]
Teranishi, T. (1983). Lateral spread of light-induced response at the cell body and axon terminal levels of external horizontal cells in the carp retina. Japanese The Journal of Physiology 33, 417-428.
[Medline]
Teranishi, T., Negishi, K. & Kato, S. (1984). Regulatory effect of dopamine on spatial properties of horizontal cells in carp retina. Journal of Neuroscience 4, 1271-1280.
[Abstract]
Toyoda, J. (1974). Frequency characteristics of retinal neurons in the carp. Journal of General Physiology 63, 214-234.
[Medline]
Toyoda, J.-I. & Coles, J. A. (1975). Rod response to sinusoidally flickering light. Vision Research 15, 981-983.
[Medline]
Umino, O. (1997). A light stimulator composed of a cathode-ray tube monitor and its use in intracellular recordings from retinal neurons. Journal of Tokyo Women's Medical College 67, 322-331.
Umino, O. & Hashimoto, Y. (1991). The role of the GABA-ergic feedback pathway from horizontal cells to cones in responsiveness to light stimulus. Neuroscience Research 15, S67-76.
Umino, O., Lee, Y. & Dowling, J. E. (1991). Effect of light stimuli on the release of dopamine from interplexiform cells in the white perch retina. Visual Neuroscience 7, 451-458.
[Medline]
Umino, O., Maehara, M., Hidaka, S., Kita, S. & Hashimoto, Y. (1994). The network properties of bipolar-bipolar cell coupling in the retina of teleost fishes. Visual Neuroscience 11, 533-548.
[Medline]
Umino, O., Watanabe, K. & Hashimoto, Y. (1989). Neural mechanisms of chromatic adaptation in L-type cone horizontal cells of the carp retina. Japanese The Journal of Physiology 39, 725-742.
[Medline]
Ushio, T. & Umino, O. (1996). The involvement of the horizontal cell's feedback action on cones in the receptive field properties of horizontal cells. Investigative Ophthalmology and Visual Science 37, 1054.
Verweij, J., Kamermans, M., Aker, E. C. & Spekreijse, H. (1996). Modulation of horizontal cell receptive fields in the light adapted goldfish retina. Vision Research 36, 3913-3923.
[Medline]
Werblin, F. S. & Dowling, J. E. (1969). Organization of the retina of the mudpuppy, Necturus maculosus. II. Intracellular recording. Journal of Neurophysiology 32, 339-355.
[Medline]
Yasui, S., Davis, W & Naka, K.-I. (1979). Spatio-temporal receptive field measurement of retinal neurons by random pattern stimulation and cross correlation. IEEE Transactions on Biomedical Engineering BME 26, 263-272.
[Medline]
Acknowledgements
We are indebted to Dr Y. Hashimoto, members of her laboratory, Dr T. Ohtsuka and Dr R. Poznanski for helpful discussions. The technical assistance of T. Yokoshima, M. Ono, O. Shishido, S.Miyaji and G. Utui was of great value. Dr B. Barford corrected the English text. Support was provided by grants 04680037 and 07680439 from the Japanese Ministry of Education.
Corresponding author
O. Umino: Department of Information Science, Toho University, 2-2-1 Miyama, Funabashi-shi, Chiba 274, Japan.
Email: umino{at}is.sci.toho-u.ac.jp
Top
Introduction
