The Journal of Neurophysiology Vol. 77 No. 2 February 1997, pp. 675-689
Copyright ©1997 by the American Physiological Society

Pharmacology of Directionally Selective Ganglion Cells in the Rabbit Retina

Christopher A. Kittila and Stephen C. Massey

Department of Ophthalmology and Visual Science, University of Texas Medical School, Houston, Texas 77030

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Kittila, Christopher A. and Stephen C. Massey. Pharmacology of directionally selective ganglion cells in the rabbit retina. J. Neurophysiol. 77: 675-689, 1997. In this report we describe extracellular recordings made from ON and ON-OFF directionally selective (DS) ganglion cells in the rabbit retina during perfusion with agonists and antagonists to acetylcholine (ACh), glutamate, and -aminobutyric acid (GABA). Nicotinic ACh agonists strongly excited DS ganglion cell in a dose-dependent manner. Dose-response curves showed a wide range of potencies, with (±)-exo-2-(6-chloro-3pyridinyl)-7-azabicyclo[2.2.1] heptane dihydrochloride (epibatidine)  nicotine > 1,1-dimethyl-4-phenylpiperazinium iodide = carbachol. In addition, the mixed cholinergic agonist carbachol produced a small excitation, mediated by muscarinic receptors, that could be blocked by atropine. The specific nicotinic antagonists hexamethonium bromide (100 µM), dihydro--erythroidine (50 µM), mecamylamine (50 µM), and tubocurarine (50 µM) blocked the responses to nicotinic agonists. In addition, nicotinic antagonists reduced the light-driven input to DS ganglion cells by ~50%. However, attenuated responses were still DS. We deduce that cholinergic input is not required for directional selectivity. These experiments reveal the importance of bipolar cell input mediated by glutamate. N-methyl-D-aspartic acid (NMDA) excited DS ganglion cells, but NMDA antagonists did not abolish directional selectivity. However, a combined cholinergic and NMDA blockade reduced the responses of DS ganglion cells by >90%. This indicates that most of the noncholinergic excitatory input appears to be mediated by NMDA receptors, with a small residual made upb y    - a m i n o - 3 - h y d r o x y - 5 - m e t h y l - 4 - i s o x a z o l e p r o p i o n i c   a c i d(AMPA)/kainate (KA) receptors. Responses to AMPA and KA were highly variable and often evoked a mixture of excitation and inhibition due to the release of ACh and GABA. Under cholinergic blockade AMPA/KA elicited a strong GABA-mediated inhibition in DS ganglion cells. AMPA/KA antagonists, such as 2,3-dihydroxy-6-nitro-7-sulfamoylbenzo(F)quinoxaline dione and GYKI-53655, promoted null responses and abolished directional selectivity due to the blockade of GABA release. We conclude that GABA release, mediated by non-NMDA glutamate receptors, is an essential part of the mechanism of directional selectivity. The source of the GABA is unknown, but may arise from starburst amacrine cells.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

As described by Barlow and Levick (1965), directionally selective (DS) ganglion cells respond vigorously to a bar of light moved across the retina in the preferred direction and little, or not at all, to the same stimulus moved in the opposite or null direction. Although the actual mechanism is not yet known, theoretical models of directional selectivity (Ariel and Daw 1982b; Barlow and Levick 1965; Borg-Graham and Grzywacz 1992; Hildreth and Koch 1987; Koch et al. 1982; Torre and Poggio 1978; Vaney et al. 1989) suggest that DS responses are due to the nonlinear interaction(s) of excitatory and inhibitory neurotransmitter inputs to DS ganglion cells. Of the neurotransmitter candidates described in these models, acetylcholine (ACh) and -aminobutyric acid (GABA) have received extensive, if not almost exclusive, consideration for mediating the excitatory and inhibitory inputs to DS ganglion cells. Indeed, there is considerable evidence that null inhibition of DS ganglion cells is mediated by a GABA_A mechanism, because it is clear that the application of GABA antagonists reversibly abolishes directional selectivity (Ariel and Daw 1982b; Caldwell et al. 1978; Kittila and Massey 1995a; Wyatt and Daw 1976). However, as we detail in this report, the case for ACh is not as clearly defined.

The morphologies of ON and ON-OFF DS ganglion cells in the rabbit retina have been identified by intracellular recording and staining (Amthor et al. 1984, 1989; Oyster et al. 1993). Further work with electron microscopy has indicated an intimate association of DS ganglion cell dendrites and the synaptic terminals of starburst amacrine cells (Famiglietti 1992). Cholinergic amacrine cells have been identified as starburst amacrine cells because of their distinctive dendritic morphology (Famiglietti 1983; Vaney 1984) and are unique in that they are the sole source of ACh in the rabbit retina (Baughman and Bader 1977; Famiglietti 1983; Masland and Mills 1979; Tauchi and Masland 1984). More recently, starburst amacrine cells have been shown to contain GABA (Brecha et al. 1988; Vaney and Young 1988). The exquisite sensitivity of DS ganglion cells to the application of ACh, especially nicotinic analogues, has long supported a functional role for starburst amacrine cells in the mechanism of directional selectivity (Ariel and Daw 1982a,b; Masland and Ames 1976). However, several reports have shown that the blockade of cholinergic input does not block directional selectivity (Ariel and Daw 1982b; Cohen and Miller 1995; Kittila and Massey 1995a). Additionally, muscarinic cholinergic receptors are abundant in the inner retina (Neal 1983; Neal and Dawson 1985; Puro 1985; Sugiyama et al. 1977) and responses mediated by muscarinic receptors have been reported in inner retina of cat (Schmidt et al. 1987). Given the intimate anatomic arrangement of starburst amacrine cell processes with the dendrites of DS ganglion cells and the strong sensitivity of DS ganglion cells to cholinergic drugs, it is a reasonable assumption that ACh is critical to the DS mechanism. In this paper we seek to specifically test the necessity of cholinergic input to the mechanism of directional selectivity under controlled perfusion of cholinergic agonists and antagonists.

From Famiglietti (1992) it is clear that starburst amacrine cells represent only a portion of the synaptic input to ON-OFF DS cells; a second major input is derived from cone bipolar cells. Cone bipolar cells carry visual information from the photoreceptors to ganglion and amacrine cells. Synaptic connections between cone bipolar cells and DS ganglion cells provide an additional pathway for direct input to DS ganglion cells. This direct input would probably be mediated by glutamate, because there is strong evidence that cone bipolar cells use glutamate as a neurotransmitter (Massey 1990; Massey and Maguire 1995; Massey and Miller 1988).

Ionotropic glutamate receptors are most simply divided into N-methyl-D-aspartic acid (NMDA) and -amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/kainate (KA) (non-NMDA) receptors. There is ample evidence for a spectrum of different AMPA/KA receptors (Hamassaki-Britto et al. 1993; Hughes et al. 1992; Seeburg 1993, 1996); however, they cannot be reliably separated with current non-NMDA glutamate antagonists. Cohen and Miller (1995) reported that most of the glutamate input to DS cells is mediated by NMDA receptors. In the present paper we confirm these results and expand this work to reveal interactions between glutamate, ACh, and GABA at the amacrine and ganglion cell level. Preliminary reports of these findings have been described in abstract form (Kittila and Massey 1995b, 1996).

	METHODS

Abstract Introduction Methods Results Discussion References

Preparation and surgery

Extracellular recordings were made from the everted, perfused rabbit eyecup preparation with the use of methods previously described (Kittila and Massey 1995a; Miller et al. 1986). Briefly, pigmented rabbits (3-5 kg) were deeply anesthetized with urethan (loading dose 1.5 g/kg ip), and, before surgery, the orbit was irrigated with 2% lidocaine. After enucleation, the eye was hemisected and everted over a Teflon pedestal.

Perfusion

Eyecups were continuously perfused with bicarbonate-based buffer solution according to Ames and Nesbett (1981) and included all additives except serum. Perfusion solutions were freshly prepared and maintained at pH 7.4 by bubbling with 95% O₂-5% CO₂. The perfusate was warmed through an in-line heater and maintained between 35 and 37°C at the efflux. Drugs were added to the perfusate through a 12-port manifold in the main perfusion line. Drug solutions were formulated freshly from 100-fold stock solutions. Unless otherwise noted, all drugs were obtained from Sigma Chemical, with the exceptions of 2,3-dihydroxy-6-nitro-7-sulfamoylbenzo(F)quinoxaline dione (NBQX), (±)-3-(2-carboxypiperazin-4-yl phosphonic acid (CPP), and epibatidine (Tocris Cookson, St. Louis, MO); dihydro--erythroidine (Research Biochemical International); and GYKI-53655 (kindly donated by Eli Lilly).

Electrophysiology

Extracellular recordings were made with the use of tungsten-in-glass microelectrodes (Levick 1972). Recordings were band-pass filtered from 1 to 3 kHz and action potentials were passed through a window discriminator to an analog ratemeter, which proved convenient to display large amounts of data. All recordings were stored on video tape (VETTER 420) for off-line analysis. Peristimulus time histograms were constructed via computer with the use of a standard binwidth of 50 ms and sum-averaged over five preferred-null scan repetitions. Dose-response data were fit to the equation V/V_max = [agonist]ⁿ/(EDⁿ₅₀ + [agonist]ⁿ). The dose-response plots, ED₅₀ values, and Hill coefficients were determined from a least-squares nonlinear curve fit of dose-response data (SCIENTIST, Micro Math).

Stimulation

Experiments were carried out under dim light conditions with the use of photopic (white light) stimuli (full intensity: 3.14 × 10¹³  cm²s¹ at 560 nm). ON-OFF DS ganglion cells were physiologically identified by their characteristic and stereotypical responses to a moving bar, small flashed spot, and full-field (diffuse) stimulation. Stimulation was provided by a computer monitor imaged on the retinal surface. Moving stimulation consisted of a single bar of light that traversed the receptive field in any of the eight major compass directions. Bar width was held constant and subtended 3.0° of visual angle (~500 µm on the retina). The bar length was long enough to extend across the entire eyecup. ON-OFF DS ganglion cells were optimally stimulated with bar speeds >5.8°/s (1,000 µm/s) and ON DS ganglion cells with bar speeds <1.7°/s (300 µm/s) (Oyster 1968).

For each DS ganglion cell, the preferred-null axis was empirically determined as that plane that showed the greatest asymmetry in number of action potentials to back-and-forth movement of the bar. Movements in the preferred direction always produced the greatest number of spikes and movements in the null direction produced few, if any, spikes. Scans in all other planes showed intermediate levels of activity.

	RESULTS

Abstract Introduction Methods Results Discussion References

Cholinergic agonists

The exquisite sensitivity of DS ganglion cells to ACh and nicotine suggests that the mechanism of directional selectivity is mediated by a nicotinic ACh receptor (Ariel and Daw 1982a,b; Masland and Ames 1976). We evaluated the potency of three specific nicotinic ACh agonists, nicotine, 1,1-dimethyl-4-phenylpiperazinium iodide (DMPP), and epibatidine, and one mixed nicotinic-muscarinic ACh agonist, carbachol. Dose-response curves for these agonists are displayed in Fig. 1A. Qualitatively, application of ACh agonists strongly excited DS ganglion cells, generating dose-dependent increases in firing rate up to a drug-specific maximum dose. Above the maximum dosage, DS ganglion cells quickly entered a depolarization block and were inhibited from firing. ED₅₀s were determined by least-squares fit and are displayed in Table 1. These results show a wide range of potencies, with epibatidine  nicotine > DMPP = carbachol. Equivalent doses were derived from these curves for subsequent pharmacology. The maximum dose was defined as the concentration required to produce a maximum firing rate before reduction due to depolarization block.

View larger version (24K): [in this window] [in a new window]   FIG. 1. A: dose-response curves for cholinergic agonists from an ON-OFF directionally selective (DS) ganglion cell. ED50, determined by computer analysis, was as follows: epibatidine, 5.36 nM; nicotine, 1.38 µM; 1,1-dimethyl-4-phenylpiperazinium iodide (DMPP), 16.3 µM; carbachol, 15.4 MM. B: response of an ON-OFF DS ganglion cell to the nicotinic agonist epibatidine (20 nM). Typical of nicotinic agonists, epibatidine produced a large increase in spontaneous depolarization. ACh, acetylcholine.

One of the most interesting compounds we evaluated was the novel nicotinic agonist epibatidine (Badio and Daly 1994; Qian et al. 1993; Spande et al. 1992). Epibatidine, a natural extract from the South American frog Epipedobates tricolor, proved to be an extremely potent nicotinic agonist for DS ganglion cells. Acting at a maximum dose of 20 nM, epibatidine was found to be 2.5 log units more potent than nicotine and 3.5 log units more potent than DMPP and carbachol (see Fig. 1B).

Throughout these experiments we had the opportunity to examine the effects of cholinergic agonists on several other classes of retinal ganglion cells. In general, ON DS and ON sluggish ganglion cells showed strong sensitivities to nicotinic agonists, with dose-response characteristics nearly identical to those of ON-OFF DS cells. In contrast, brisk-classed ON and OFF wide-field ganglion cells were generally much less responsive. Even under elevated doses of nicotinic agonist (>10 µM nicotine), wide-field cells often showed only mild increases in spontaneous depolarization and were not easily driven to depolarization block. In general, our findings are in agreement with results from previous studies (Ariel and Daw 1982a,b; Masland and Ames 1976). Clearly, it appears that DS ganglion cells are among the most cholinergic sensitive cells in the retina.

Effect of cholinergic blockade on directional selectivity

With theoretical models of directional selectivity emphasizing interactions of excitation and inhibition, the sensitivity of DS ganglion cells to cholinergic agonists, and the costratification of DS ganglion cells and starburst amacrine cells, a role for ACh seems almost perfectly outlined. However, the initial report by Ariel and Daw (1982b), which stated that the systemic infusion of the nicotinic antagonist mecamylamine did not abolish directional selectivity, raises fundamental questions about the necessity of excitatory cholinergic input to the DS mechanism. Recording directly in a perfused eyecup arrangement minimizes many of the uncertainties involved with systemic drug delivery, and a wider variety of drugs can be used. We tested the effects of four well-documented nicotinic cholinergic antagonists: mecamylamine (50 µM), dihydro--erythroidine (50 µM), hexamethonium bromide (100 µM), and tubocurarine chloride (50 µM). All of these antagonists blocked the excitatory effects of the nicotinic agonists described earlier, including epibatidine, the most potent nicotinic agonist to date. Our goal was to test the necessity of cholinergic input to the mechanism of directional selectivity by blocking cholinergic receptors with established cholinergic antagonists.

Figure 2 shows peristimulus time histograms from a representative cholinergic blockade experiment in which the noncompetitive nicotinic antagonist mecamylamine was used. The first record shows the control response from an ON-OFF DS ganglion cell to a bar of light moved through the receptive field along the preferred-null axis. The preferred-direction scan produced two groups of spikes as the leading (ON) edge and trailing (OFF) edges of the bar traversed the receptive field; null-direction movement produced no response. In the second record a positive control of the specific nicotinic agonist DMPP (50 µM) was applied. As expected, DMPP strongly excited the DS ganglion cell, indicating a sensitivity of the cell to nicotinic ACh input. Next, nicotinic receptors were blocked by direct application of 50 µM mecamylamine. In the third record, mecamylamine attenuated the preferred-direction responses by about half, but had little or no effect on directional selectivity. In other words, the blockade of nicotinic cholinergic receptors did not abolish DS responses. To confirm the existence of a complete blockade of nicotinic input to the DS ganglion cell, a second control application of 50 µM DMPP was added to the perfusate. As shown in the fourth record, mecamylamine completely blocked all excitatory effects of DMPP, yet the cell remained DS. After the washout of mecamylamine, the fifth record shows a restoration of nicotinic sensitivity as the DS cell once again responded to a control application of 50 µM DMPP with a large increase in spontaneous firing. In the sixth record there was a return of the control responses after the washout of all drugs.

View larger version (23K): [in this window] [in a new window]   FIG. 2. Nicotinic antagonists do not block directional selectivity. Peristimulus time histograms from an ON-OFF DS ganglion cell. The nicotinic agonist DMPP (50 µM) strongly excited DS ganglion cells (2ndrecord). The noncompetitive nicotinic antagonist mecamylamine (50 µM) attenuated preferred-direction responses but did not abolish directional selectivity (3rd record). The complete blockade of nicotinic input is verified by lack of excitation by a control application of DMPP under mecamylamine (4th record). Return of nicotinic sensitivity follows washout of mecamylamine (5th record). Return of control response follows washout of DMPP (6th record).

This self-contained experiment establishes several important points. First, ON-OFF DS ganglion cells are sensitive to the application of the nicotinic agonist DMPP. Second, the nicotinic ACh input can be reversibly blocked by the application of 50 µM mecamylamine. Third, ON-OFF DS ganglion cells remained DS during a confirmed blockade of all nicotinic ACh input. Fourth, light-driven responses under nicotinic blockade were generally attenuated by ~50%. Our results were confirmed in similar experiments with the use of a variety of competitive nicotinic antagonists including 100 µM hexamethonium bromide (n = 12), 50 µM dihydro--erythroidine (n = 4), and 50 µM tubocurarine chloride (n = 3), all with similar results. In similar experiments with 29 ON-OFF DS cells and five ON DS cells, where cholinergic blockade included both nicotinic and muscarinic receptors, we never observed any loss of directional selectivity in response to cholinergic blockade. In general, our results confirm and extend the findings of Ariel and Daw (1982b) and strongly indicate that cholinergic input is not required to produce DS responses.

Muscarinic responses of DS ganglion cells

Cholinergic receptors can be broadly subdivided into nicotinic and muscarinic varieties. Of the two, nicotinic receptors are best understood and have dominated the literature on directional selectivity. However, at an early stage of these experiments it became apparent that the mixed agonist carbachol, a structural analogue of ACh, produced an excitation of DS ganglion cells that could not be blocked by specific nicotinic antagonists. We wished to rule out a muscarinic mechanism for DS, so we conducted experiments with the muscarinic antagonist atropine.

The peristimulus time histograms of Fig. 3 compare the carbachol-induced excitation of an ON-OFF DS ganglion cell under a selective nicotinic blockade by 100 µM hexamethonium bromide (left) versus a combined nicotinic-muscarinic blockade by 100 µM hexamethonium bromide and 2 µM atropine (right). The top record shows the control response from an ON-OFF DS ganglion cell. In this case a small response to null-direction movement was present. Next, cholinergic receptors were blocked with selective antagonists, and the completeness of the blockade was verified by application of the mixed agonist carbachol. Under a selective nicotinic blockade (Fig. 3, bottom left), the application of 50 µM carbachol increased peak preferred-direction responses from 21 spikes per bin in the control to 35 spikes per bin. In addition, there was an obvious potentiation of the small null response compared with the control. Under the same conditions of nicotinic block, but now supplemented with 2 µM atropine (Fig. 3, bottom right), carbachol had no potentiating effect on either preferred- or null-direction responses. Importantly, the responses remained DS during the combined nicotinic and muscarinic blockade. In fact, under these conditions, peak responses were reduced by 24% from control to a level of 16 spikes per bin.

View larger version (13K): [in this window] [in a new window]   FIG. 3. Muscarinic ACh responses of ON-OFF DS ganglion cells. The nicotinic antagonist hexamethonium bromide (100 µM) unmasked a carbachol-induced potentiation of the light-driven responses of DS ganglion cells. This 2-fold potentiation was blocked by the addition of 2 µM atropine, a muscarinic antagonist.

Because carbachol acts at both muscarinic and nicotinic ACh receptors, the potentiation of the light-driven response under hexamethonium alone (Fig. 3, bottom left) reflects an incomplete blockade of cholinergic receptors. Subsequently we were led to control for any possible muscarinic interactions by including 2 µM atropine in all experiments where we describe a complete cholinergic block. These experiments indicate that cholinergic input, whether nicotinic or muscarinic, is not required to generate DS responses.

NMDA input to DS ganglion cells

The substantial residual responses of DS ganglion cells under cholinergic blockade suggest the existence of an additional independent, excitatory input to the DS mechanism. The persistence of DS responses during a complete cholinergic block is compelling evidence that ACh does not provide all of the excitatory input to DS ganglion cells. As previously discussed, the presence of cone bipolar cell inputs to DS ganglion cells suggests that this pathway should also be sensitive to glutamate. Indeed, ON-OFF DS ganglion cells have previously been shown to be sensitive to NMDA (Cohen and Miller 1995; Kittila and Massey 1995b; Massey and Miller 1990); however, the relative independence of NMDA to excite DS cells in the absence of cholinergic input has not been demonstrated. The ratemeter trace of Fig. 4 shows a contiguous record of responses from an ON-OFF DS ganglion cell to scans in the preferred-null axis. The first two scans, detailed in an expanded time scale in the bottom left, show a strong preferred-direction response and no response to movement in the null direction. (Because of the compressed nature of this recording and the lack of a null response, ratemeter peaks will necessarily highlight preferred-direction responses of the cell.) Starting from the left edge, five relatively uniform control responses are interrupted by a control application of 50 µM DMPP. As before, DMPP produced a large increase in the spontaneous firing of the ON-OFF DS ganglion cell, indicating a strong nicotinic input to the cell. After a return of the control response, a 30-s pulse of 200 µM NMDA was applied. Similar to the response to the nicotinic agonist, the cell responded with a large increase in firing rate, indicating a sensitivity to NMDA. After a return of the normal response pattern, cholinergic receptors were blocked by perfusion with 100 µM hexamethonium bromide plus 2 µM atropine. As before, we establish the existence of a complete cholinergic block through a control application of 50 µM DMPP, which produced no change in firing rate. Once again, aside from the noticeable attenuation of the preferred-direction responses, cholinergic blockade did not abolish directionally selectivity. Next, still during cholinergic block, a control pulse of 200 µM NMDA was applied and the ON-OFF DS ganglion cell responded with large increases in spontaneous firing. This result confirms that hexamethonium selectively blocks cholinergic responses, but other drugs added to the perfusate can still produce excitation. More importantly, we have established that ON-OFF DS ganglion cells can be independently excited by NMDA in the absence of cholinergic input. After the response to NMDA, cholinergic antagonists were washed out, and a postdrug control of DMPP showed a return of normal cholinergic sensitivity. The results of this experiment demonstrate the existence of an independent glutamate input to DS ganglion cells, an input mediated by NMDA receptors. As suggested by anatomic evidence, this direct input could be provided by cone bipolar cells, which are known to release glutamate. Because we obtained similar results in three of three ON DS ganglion cells, direct glutamate input appears to be common feature among retinal ganglion cells that produce DS responses.

View larger version (30K): [in this window] [in a new window]   FIG. 4. Ratemeter recording from an ON-OFF DS ganglion cell continuously stimulated with a moving bar of light in the preferred-null axis. The 1st 2 scans are shown in expanded time scale in the bottom left. Predrug controls show large increases in spontaneous activity to the application of the nicotinic agonist DMPP (50 µM), and, later, the glutamate agonist N-methyl-D-aspartic acid (NMDA) (200 µM). Under complete cholinergic blockade with hexamethonium bromide (100 µM) and atropine (2 µM), preferred responses are reduced but directional selectivity is maintained. Under hexamethonium bromide and atropine, the response to DMPP is blocked but the response to NMDA is not affected.

Effect of NMDA and ACh antagonists

The sensitivity of ON-OFF DS ganglion cells to NMDA under cholinergic blockade suggests glutamate receptor involvement in the synaptic mechanism. Further support is generated from experiments with specific NMDA antagonists such as aminophosphonoheptanoic acid (Ap-7; 100 µM) and CPP (5 µM). Although previous work with aminophosphonoheptanoic acid has shown that the blockade of NMDA receptors does not abolish directional selectivity (Fig. 12 in Massey and Miller 1990), the reported 34% response attenuation by NMDA antagonists so complemented the residual responses measured under cholinergic blockade that we were led to examine the effect of concurrent NMDA and ACh blockade.

View larger version (10K): [in this window] [in a new window]   FIG. 12. Proposed schematic models of the pharmacological inputs to DS ganglion cells in the rabbit retina. A: DS ganglion cells (DSG) receive excitatory (+) cholinergic (ACh) and inhibitory () GABAergic (GABA) input from starburst amacrine cells (SA) and direct excitatory glutamate input (Glu) from cone bipolar cells (CB). Glutamate receptors on the starburst cell show strong AMPA/KA sensitivity and a reduced sensitivity to NMDA. Glutamate receptors on the DS ganglion cell show strong NMDA sensitivity and a reduced sensitivity to AMPA/KA. Additional receptors on the DS ganglion cell include nicotinic ACh and GABAA receptors. B: similar to A, except GABAergic input to DS ganglion cells is provided by a separate, yet unknown, amacrine cell (?A). Both A and B suggest presynaptic GABA release to DS ganglion cells is mediated by non-NMDA glutamate receptors.

In Fig. 5A, ratemeter peaks highlight the preferred-direction responses from an ON-OFF DS ganglion cell. Cholinergic receptors were blocked by the application of 100 µM hexamethonium bromide and 2 µM atropine. As before, the cholinergic blockade attenuated the preferred-direction responses but did not effect the directional selectivity of the cell. Next, 5 µM CPP was added to block NMDA receptors. In independent experiments, 5 µM CPP or 100 µM aminophosphonoheptanoic acid were shown to selectively abolish responses to a saturating dose of NMDA (data not shown; see also Massey and Miller 1990). The addition of CPP during cholinergic block produced an additional reduction of the preferred-direction response, but did not abolish directional selectivity. On washout, control responses were obtained once again.

View larger version (35K): [in this window] [in a new window]   FIG. 5. Effect of ACh and NMDA blockade. A: ratemeter record from an ON-OFF DS cell under continuous preferred-null stimulation. The 1st 3 responses are shown in expanded time scale (bottom left). B: same data in A displayed as peristimulus time histograms (left). Sequential application of the cholinergic antagonists hexamethonium bromide (HEX, 100 µM) and atropine (2 µM) and the NMDA antagonist (±)-3-(2-carboxypiperazin-4-yl phosphonic acid (CPP, 5 µM) reduced preferred-direction responses but did not abolish directional selectivity. As a control, the order of antagonist application was reversed (right). This figure shows that nicotinic block and NMDA block are additive, but even in the presence of both antagonists small DS responses remain.

To better quantify these results in terms of spike counts, peristimulus time histograms were formed from this same data (Fig. 5B, left). As shown in these four histograms, perfusion of cholinergic antagonists reduced the preferred-direction response of the DS ganglion cell by almost 70% (70 spikes vs. 235 control spikes). The addition of NMDA antagonists produced an additional 26% reduction, leaving <4% (9 vs. 235 spikes) of the original response. As a control, the experiment was repeated in reverse order, blocking NMDA receptors first, then ACh (Fig. 5B, right). In this case NMDA antagonists abolished 21% of the response (162 spikes vs. 204 control spikes) and application of both ACh and NMDA antagonists abolished nearly 94% (14 spikes vs. 204 control spikes) of the response. Yet, the remaining response was still DS. We obtained similar results from six other ON-OFF DS ganglion cells, and the residual responses following ACh and NMDA blockade are summarized in Fig. 6. For this group of cells, NMDA antagonists produced a 31 ± 7.29% (mean ± SE) reduction in the light-driven spike count. Likewise, cholinergic antagonists attenuated the light-driven responses by 54 ± 6.67%. Applied together, antagonists of ACh and NMDA reduced the responses of ON-OFF DS ganglion cells by nearly 92 ± 2.43% . Presumably, the 8% residual response undersimultaneous ACh and NMDA blockade represents the non-NMDA (AMPA/KA) input to the ON-OFF DS ganglion cell.

View larger version (34K): [in this window] [in a new window]   FIG. 6. Summary of results from 6 ON-OFF DS ganglion cells showing the percentage of residual response (compared with control spike count values) produced by selective blockade of ACh and NMDA inputs. On average, NMDA antagonists blocked 31 ± 7.29% of the light-driven responses and cholinergic antagonists blocked 54 ± 6.67%. Simultaneous application of ACh and NMDA antagonists blocked 92 ± 2.43% of the light-driven responses of these cells. Cholinergic input was selectively blocked by perfusion of 100 µM hexamethonium bromide and 2 µM atropine. NMDA input was blocked with 10 µM CPP.

From a statistical analysis of the data in Fig. 6, it is clear that ACh and NMDA antagonists showed large variances in the degree of response attenuation. Although such variations can often be dismissed as abnormal drug delivery, this does not appear to be the case, because this variance was significantly reduced when antagonists to both ACh and NMDA were applied. This finding suggests that cholinergic and glutamatergic inputs are not rigidly fixed in their contributions to the final response. Indeed, it was often noted that in cases where one antagonist produced a below-average response reduction, the other antagonist had an above-average effect. This suggests that ACh and NMDA act together, each transmitter supplementing the other to produce an appropriate level of excitation to feed the DS mechanism. During a cholinergic block, it appears that most of the excitatory input is mediated by NMDA receptors. Similar results were obtained independently by Cohen and Miller (1995).

Non-NMDA input to DS ganglion cells

It would be convenient to describe the excitatory inputs to DS ganglion cells as a complementary mixture of NMDA and ACh, but the results in Figs. 5 and 6 reveal that the blockade of both of these inputs rarely abolished all of the light-driven activity (only 2 of 11 cells showed a complete blockade). In almost all cases a small percentage of the response remained, usually <5%, and this response remained DS. ON-OFF DS ganglion cells, like all retinal ganglion cells in rabbit, are known to be excited by the glutamate agonist KA (Massey and Miller 1988). Because this excitation could be elicited in the presence of a synaptic blockade with the use of Co²⁺ (Massey, unpublished result), the presence of KA receptors on the ganglion cell membrane is suggested. As seems to be the case throughout the nervous system, NMDA receptors are rarely expressed alone. Patterns of NMDA receptor expression generally show a parallel distribution with that of non-NMDA (AMPA) receptors (Watkins et al. 1990).

To evaluate the role of AMPA/KA receptors in directional selectivity, we sought to determine dose-response relationships for AMPA and KA. Although both AMPA and KA are excitatory glutamate analogues, lone application of either of these drugs generally produced unpredictable results in DS ganglion cells. Compared with NMDA, which always produced dose-dependent increases in spontaneous firing, KA rarely induced clear and repeatable excitation. The responses of DS ganglion cells to KA generally showed weak excitation that was immediately followed by a reduction in light-driven activity. Because KA has been shown to cause massive release of ACh in the rabbit retina (Linn et al. 1991), we surmised that unusual dose-response relationships observed with AMPA and KA were being affected by the excitatory action of ACh. To control for released ACh and to isolate KA responses, our tactic was to apply KA under a cholinergic blockade. Although we fully expected to see a more controlled excitatory response to KA (such as with NMDA in Fig. 4), the opposite occurred. Instead of an excitatory response to KA, there was complete inhibition of the light-driven response. Figure 7 is a ratemeter display from an ON-OFF DS ganglion cell stimulated with a moving bar under cholinergic blockade with 50 µM hexamethonium bromide and 2 µM atropine. After the addition of 10 µM KA the light-driven response of the cell was completely abolished. This result may be clearly distinguished from a depolarizing block, which is always preceded by excitation and increased firing rate. In this example, there is no preceding excitation (see arrows). Furthermore, the dose of KA employed (2 µM) is sufficient to drive the cells maximally (Massey and Miller 1988). On washout of the drugs the response of the cell returned to control levels. It is clear that under cholinergic blockade the application of KA (or AMPA) produced a complete inhibition of the ON-OFF ganglion cell. Similar responses were obtained in two of two ON DS ganglion cells. The responses of DS ganglion cells to AMPA and KA resembled a composite of excitation and inhibition acting simultaneously. These results suggest that AMPA and KA, in addition to their direct excitatory effects, act to promote the release of an inhibitory neurotransmitter as well as releasing ACh.

View larger version (52K): [in this window] [in a new window]   FIG. 7. Response inhibition by kainate (KA) under cholinergic blockade. Ratemeter record from an ON-OFF DS ganglion under continuous stimulation by a moving bar of light. Ratemeter peaks highlight preferred-direction responses. Under the blockade of cholinergic receptors by 50 µM hexamethonium bromide and 2 µM atropine, application of 10 µM KA produced a complete inhibition of the light-driven response of the cell. After washout, control responses returned. Arrows: fall in ratemeter record with no sign of previous excitation that would indicate a depolarizing block. This figure shows that KA causes the release of an inhibitory intermediate.

Of the inhibitory neurotransmitters known to exist in the inner retina, GABA has been shown to be directly involved in the mechanism of directional selectivity (Ariel and Daw 1982b; Kittila and Massey 1995a; Wyatt and Daw 1976). If the unusual responses of DS ganglion cells to the application of non-NMDA agonists were due to a presynaptic release of GABA, then a blockade of postsynaptic GABA receptors should reduce or eliminate the inhibition. The ratemeter record of Fig. 8 shows the result of an experiment from an ON-OFF DS ganglion cell in which KA was applied both alone and under GABA blockade by the specific GABA_A antagonist SR95531 (Kittila and Massey 1995a). In Fig. 8, the application of 10 µM KA produced a weak excitation in the ON-OFF DS ganglion cell. In this case, the KA-induced excitation was approximately half of the peak (predrug) light-driven levels. In addition, the excitatory effects of KA were short lived and produced a period of reduced light response during washout. Next, GABA_A receptors were blocked by the application of 2 µM SR95531. The blockade of GABA_A receptors was confirmed by the loss of directional selectivity (i.e., the emergence of null-direction responses) and is apparent in the ratemeter record as a period of spike doubling. Under the conditions of GABA blockade, 10 µM KA was applied a second time. With GABA_A receptors blocked, the excitatory effects of KA were greatly potentiated, producing a sustained excitation nearly twofold larger than that of control. Clearly, under GABA_A blockade the inhibition previously observed with the perfusion of KA is eliminated. Because KA is known to cause ACh release in the rabbit retina (Linn et al. 1991), it is clear that at least a portion of the excitation would be mediated by ACh. However, in similar experiments (n = 5) where we included cholinergic antagonists in addition to SR95531, the results were qualitatively similar. Additionally, the effect is reversible, as shown following the washout of SR95531, where a postdrug control of KA produced a renewed attenuation of the DS ganglion cell response. Clearly, under GABA blockade KA acts in manner that is much more consistent with the actions of an excitatory amino acid transmitter. Therefore we suggest that the unusual inhibitory effect of KA may be explained by the release of GABA, which, in turn, inhibits DS ganglion cells. Similar results were obtained with other glutamate analogues, such as AMPA.

View larger version (35K): [in this window] [in a new window]   FIG. 8. KA responses are suppressed by -aminobutyric acid (GABA) release. Ratemeter recording from an ON-OFF DS ganglion cell. KA (10 µM), an excitatory amino acid, produced weak excitation of DS ganglion cells when perfused alone. After excitation by KA, a short period of reduced light-driven activity can be observed. Under GABAA blockade by SR-95531 (2 µM), directional selectivity is lost (revealed as an apparent spike doubling in the ratemeter record), and the response to KA has more than doubled. The slight delay in the KA excitation is due to a 30-s delay in perfusion. Clearly, GABAA receptors remain blocked during the KA excitation, as evidenced by the continued spike doubling following washout of KA. This figure suggests that the inhibitory effects of KA are mediated by GABA.

Non-NMDA antagonists

Further evidence of non-NMDA receptor involvement in GABA release was derived from experiments with specific antagonists to AMPA/KA receptors. Recently, it has been shown that the application of NBQX abolishes directional selectivity in ganglion cells of the rabbit retina (Cohen and Miller 1995), and we have confirmed these findings in seven of seven ON-OFF DS and two of two ON DS ganglion cells. Figure 9A shows a raw recording from an ON-OFF DS ganglion cell. The control response produced 25 spikes to movement in the preferred direction and no response to movement in the opposite direction. After the application of 10 µM NBQX, the preferred and null responses became equal, each producing 21 spikes. Interestingly, aside from the obvious loss of directional selectivity, NBQX reduced the number of preferred-direction spikes from 25 to 21. (A similar reduction can be seen in Fig. 1 of Cohen and Miller 1995, where preferred-direction spike counts are reduced from 34 to 20 under NBQX.) From these data it appears that NBQX produced two major effects on the responses of ON-OFF DS ganglion cells. First, NBQX induced a complete loss of directional selectivity. Second, preferred-direction responses were attenuated. The loss of directional selectivity observed with NBQX is similar to results obtained with the selective AMPA antagonist GYKI-53655 (Paternain et al. 1995). Like NBQX, 10 µM GYKI-53655 has been found to block the release of ACh in the rabbit retina (unpublished results) and also blocks directional selectivity (see Fig. 9B). We have found GYKI-53655 to act as a relatively selective AMPA antagonist with a potency similar to that of NBQX. As we have previous demonstrated throughout these results, the blockade of ACh attenuated preferred-direction responses but never blocked directional selectivity. Thus the loss of directional selectivity cannot be attributed to a reduction of cholinergic input to the DS ganglion cell. The fact that GABA antagonists are known to abolish directional selectivity, that KA responses appear to be greatly potentiated in the presence of GABA antagonists, and that the blockade of non-NMDA glutamate input produces a complete loss of directional selectivity suggests an involvement of non-NMDA receptors in the presynaptic release of GABA. Under this scenario, blockade of AMPA/KA receptors by NBQX and GYKI-53655 blocks the release of GABA (abolishing directional selectivity) and also blocks the release of ACh (revealed as an attenuation of preferred-direction responses).

View larger version (15K): [in this window] [in a new window]   FIG. 9. -Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)/KA antagonists abolish directional selectivity. A: raw recording from an ON-OFF DS ganglion cell. Similar to the results of Cohen and Miller (1995), application of 10 µM 2,3-dihydroxy-6-nitro-7-sulfamoylbenzo(F)quinoxaline dione (NBQX) abolished directional selectivity in both ON and ON-OFF DS ganglion cells. NBQX promoted null-direction responses to equal those of the preferred direction, and slightly attenuated preferred-direction responses (compared with control). After washout of NBQX, directional selectivity was regained. B: peristimulus time histograms from an ON-OFF DS ganglion cell showing that the AMPA/KA antagonist GYKI-53655 (10 µM) abolishes directional selectivity similarly to NBQX (10 µM).

The idea that NBQX blocks the release of ACh and GABA is supported by further experiments combining NMDA and non-NMDA antagonists. In Fig. 10, the first record shows control responses from an ON-OFF DS ganglion cell. In the second record NMDA is selectively blocked by the addition of 10 µM CPP. As before, blockade of NMDA receptors reduced the responses of DS ganglion cells, but did not block directional selectivity. Under these conditions, the remaining excitatory input is mediated mostly by nicotinic ACh receptors (cf. Fig. 5B, right). The further addition of NBQX (3rd record) abolished all light-driven responses of the DS ganglion cell. This indicates the block of ACh release onto the DS ganglion cell. In this example the null responses, because of the NBQX block of GABA release, do not emerge because all excitatory input to the ganglion cell has been blocked.

View larger version (15K): [in this window] [in a new window]   FIG. 10. Effect of NMDA and non-NMDA blockade. Peristimulus time histograms from an ON-OFF DS ganglion cell. Blockade of NMDA receptors by CPP (10 µM) reduced preferred-direction responses, but did not abolish directional selectivity (2nd record). Coapplication of CPP and the AMPA/KA antagonist NBQX (10 µM) abolished all light-driven responses (3rd record). After washout of CPP and NBQX, control responses returned (4th record). This figure shows that the combined block of NMDA and non-NMDA receptors abolishes light-driven responses of DS ganglion cells. Conversely, the responses remaining after NBQX alone are mediated entirely by NMDA receptors.

If the loss of directional selectivity and the reduction of preferred-direction responses of DS ganglion cells seen with non-NMDA antagonists are due to a blockade of GABA and ACh release, then findings with NBQX should be mimicked by simultaneous blockade of GABA and ACh. We demonstrate this effect in Fig. 11. The first histogram shows the control response from an ON-OFF DS ganglion cell, with two bursts of spikes to movement in the preferred direction and almost no response to movement in the opposite direction. Next we blocked cholinergic input with 100 µM hexamethonium bromide and 2 µM atropine. As in preceding examples, cholinergic blockade reduced preferred-direction responses, but did not alter directional selectivity. In the third histogram 10 µM NBQX is added along with cholinergic antagonists. The addition of NBQX produced a complete loss of directional selectivity. It is our interpretation that NBQX acts to block an additional component of the DS mechanism that is responsible for null-direction inhibition. After a return to control (4th histogram), the responses under NBQX, hexamethonium bromide, and atropine are mimicked by the perfusion of hexamethonium bromide, atropine, and 2 µM SR95531. Qualitatively, blockade of both ACh and GABA_A receptors clearly mimics the responses obtained with non-NMDA receptor antagonists and supports our claim that AMPA/KA receptors control the release of ACh and GABA to DS ganglion cells.

View larger version (28K): [in this window] [in a new window]   FIG. 11. NBQX blocks release of ACh and GABA. Peristimulus time histograms from an ON-OFF DS ganglion cell. As before, blockade of cholinergic input by 100 µM hexamethonium bromide and 2 µM atropine reduced preferred-direction responses, but did not abolish directional selectivity (2nd record). The addition of 10 µM NBQX to the cholinergic antagonists evoked a complete loss of directional selectivity (3rd record). After a brief return to control (4th histogram), a qualitatively similar response to that in the 3rd record is obtained by perfusion of 100 µM hexamethonium bromide, 2 µM atropine, and 2 µM SR95531, a GABAA antagonist (5th record). After washout there is a return of control responses (6th record). These results suggest the loss of directional selectivity with NBQX is due to a blockade of GABA release and not to a loss of cholinergic input.

	DISCUSSION

Abstract Introduction Methods Results Discussion References

Summary

DS retinal ganglion cells in rabbit show a strong sensitivity to ACh, glutamate, and GABA. DS ganglion cells were excited by nicotinic cholinergic agonists and by an atropine-blocked, muscarinic component of carbachol. In addition, DS ganglion cells were excited by the glutamate agonist NMDA, with these receptors appearing to account for most of the glutamate input to DS ganglion cells. NMDA and cholinergic antagonists reduced the responses of DS ganglion cells, but even together did not abolish directional selectivity. The non-NMDA glutamate agonists AMPA and KA generally produced a mixed response of excitation and inhibition in DS ganglion cells that we attributed to the release of ACh and GABA, as well as some direct, but minor, AMPA/KA activation. The non-NMDA antagonists NBQX and GYKI-53655 reduced the light-driven responses and abolished directional selectivity due to the block of ACh release and GABA release, respectively. These results underlie the importance of GABA release to the mechanism of directional selectivity.

ACh agonists

In our survey of nicotinic agonists we examined the dose-response characteristics of three well-established drugs (DMPP, nicotine, and carbachol) and one novel nicotinic agonist (epibatidine). Of these four compounds, epibatidine proved to be an extremely effective nicotinic agonist with a potency nearly 3 log units over that of nicotine itself. Although epibatidine has been reported to possess certain analgesic properties (Qian et al. 1993; Spande et al. 1992), the responses obtained from DS ganglion cells at the extracellular level appear indistinguishable from those obtained with nicotine itself. Given the extreme potency of this drug and its susceptibility to blockade by specific nicotinic antagonists, we find epibatidine to be a potentially useful nicotinic agonist. We note that many ganglion cell types respond to nicotinic cholinergic drugs (Ariel and Daw 1982a,b), and ACh has also been implicated in the generation of spontaneous retinal waves in ganglion cells of the developing retina (Feller et al. 1996). Clearly, there is much to learn about cholinergic input to ganglion cells and epibatidine is sufficiently potent to be used for intraocular application.

Although we do not dispute the dominant role of nicotinic receptors in DS pharmacology, the excitation produced by the mixed agonist carbachol in the presence of nicotinic antagonists suggest muscarinic receptors may have some impact on the responses of DS ganglion cells. Somewhat disturbing, however, is the fact that the application of atropine alone had no measurable effect on the extracellularly recorded responses of either ON or ON-OFF DS cells. Additionally, the application of general muscarinic agonists such as 50 µM methyfurtrehonium and 50 µM oxotremorine produced no obvious changes in spike output of DS ganglion cells. Given the extensive variety of muscarinic receptors known to exist in the inner retina, numerous mechanisms (many of which are indirect) could be proposed to explain this effect. Indeed, the pharmacology of muscarinic receptors in the retina, including their role in visual function, is not well understood. In any event, our results suggest that muscarinic receptors have at least a minor role in the modifying responses of DS ganglion cells and that this input can be blocked by atropine.

ACh antagonists

The interpretation of our results with nicotinic antagonists is unequivocal. We used four different antagonists with similar results and have confirmed the original result from Ariel and Daw (1982b) that directional selectivity is not blocked by nicotinic antagonists. Furthermore, in Figs. 2 and 4 we have demonstrated, by the application of nicotinic agonists, that a nicotinic cholinergic block was in effect during the time that DS responses were obtained. The fact that the blockade of nicotinic receptors does not abolish directional selectivity eliminates the necessity of a nicotinic cholinergic synapse (whether direct or indirect) for the generation of a DS response.

NMDA input to DS ganglion

Although most of the present models invoke nicotinic cholinergic input as the sole source of excitatory input to the DS ganglion cell, our results suggest that glutamate (mostly via an NMDA receptor) carries a mutual responsibility. Both glutamate and ACh are present in presynaptic cells, both show light-driven release, and both have postsynaptic actions that excite DS ganglion cells. Perhaps the most startling finding regarding the glutamate input to DS ganglion cells is the preponderance of NMDA input over that of AMPA/KA. In subsequent experiments (data not shown) in which we attempted to quantify the NMDA and non-NMDA components of the glutamate input by sequentially blocking ACh and GABA receptors, then NMDA and AMPA/KA receptors (and vice versa), the results suggested that nearly 90% of the glutamate input to DS ganglion cells was NMDA driven. To our knowledge, such dominance of NMDA input over that of AMPA/KA is unprecedented in the retina. Previous work from several groups suggests that NMDA receptors carry a minor portion, often ~30%, of glutamate input to ganglion cells (Cohen and Miller 1994; Coleman and Miller 1988, 1990; Mittman et al. 1990; Taylor et al. 1995).

The finding that antagonists to both ACh and NMDA, even in combination, do not abolish directional selectivity is especially important in understanding the role of excitatory input to the DS mechanism. The implication is that GABA-mediated inhibition is primarily responsible for the phenomenon of directional selectivity, as first proposed by Wyatt and Daw (1976). This is not to say that excitation plays no role in the DS mechanism, because it is clear that DS ganglion cells do require some form of excitatory input. Our interpretation favors feed-forward, light-driven excitation evoked by ACh and glutamate (via the NMDA receptor) that is inhibited under null-direction motion by GABA. The reason for two pharmacologically distinct forms of excitation to ON-OFF DS ganglion cells is not clear at this point, but may provide some additional form of receptive field tuning. Such an argument is suggested by the study of Grzywacs and Amthor (1993), in which preferred-direction facilitation first described by Barlow and Levick (1965) appeared to be modulated by an unnamed excitatory agent. Additionally, Smith et al. (1996) have suggested that directional selectivity in the turtle retina may be mediated by several mechanisms other than just asymmetric GABA inhibition. Certainly more work is indicated to identify these potential mechanisms and how they might interact to produce directional responses, especially with regard to any neurotransmitter signatures that might be involved.

Non-NMDA-receptor-mediated GABA and ACh release

In contrast to the excitatory effects of NMDA, perfusion with AMPA and KA evoked mixed responses. When applied alone, AMPA or KA produced a weak excitation followed by a prolonged depression. It is well known that AMPA and KA cause massive ACh release (Linn et al. 1991), so these experiments were repeated during a nicotinic block. Perfusion with hexamethonium (both with and without atropine) revealed an AMPA/KA-mediated inhibition, which displayed none of the characteristics of a depolarizing block. Further experiments revealed that this strong inhibition of DS ganglion cells was mediated by GABA. NMDA application always produced excitation and never produced mixed responses in DS ganglion cells. The implication is that GABA release is controlled by AMPA/KA receptors and not by NMDA receptors.

Our interpretation of the AMPA/KA effects is reinforced by the results with AMPA/KA antagonists such as NBQX. As previously reported by Cohen and Miller (1995), NBQX reduced the firing rate of DS ganglion cells and blocked DS by promoting null responses. The remaining responses in the presence of NBQX are mediated by NMDA receptors (Fig. 10). In fact, the effect of NBQX is mimicked by the combination of hexamethonium bromide, atropine and SR95531 (Fig. 11). Therefore the reduction in firing rate is due to the block of ACh release, and the emergence of null responses is due to the block of GABA release. In turn, this indicates that the amacrine cells that release ACh and GABA are selectively excited by AMPA/KA and not NMDA receptors.

There is ample precedent for amacrine cells that respond to AMPA/KA but not NMDA. Dixon and Copenhagen (1992) demonstrated that sustained amacrine cells responded selectively to non-NMDA glutamate agonists, but many other amacrine cells were excited by both NMDA and non-NMDA agonists. Recordings from the rat retina also showed that AII amacrine cells were selectively excited by non-NMDA agonists such as AMPA and KA (Boos et al. 1993). More interestingly for the system at hand, Linn et al. (1991) showed that KA and AMPA caused massive ACh release and that the light-evoked release of ACh was abolished by AMPA/KA antagonists such as 6,7-dinitroquinoxaline-2,3-dione. NMDA was ineffective in releasing ACh in the presence of physiological levels of Mg²⁺. In support of these results, Zhou and Fain (1995) showed that starburst amacrine cells in rabbit had very weak responses to NMDA, but AMPA/KA were potent agonists.

The pharmacological profile of ACh release and starburst amacrine response is the same as the profile of GABA release reported in this paper. Therefore these results may be consistent with the release of GABA from the same source, i.e., the starburst amacrine cell. However, it is clear that other amacrine cells are selectively stimulated by non-NMDA agonists. It should be noted that O'Malley et al. (1992) showed that ACh and GABA release have different properties, although the interpretation of GABA release is complicated by the effects of glial uptake.

Model of bipolar cell input to DS ganglion cells and starburst amacrine cells

Our results suggest that presynaptic GABA release to DS ganglion cells is controlled by an AMPA/KA receptor on a GABAergic interneuron, most likely an amacrine cell. Schematic diagrams of these possible arrangements are shown in Fig. 12. In both of these models, excitatory input to the DS ganglion cell is provided by ACh (from starburst amacrine cells) and an NMDA-dominated glutamate input (from cone bipolar cells) on the DS ganglion cell membrane. One arrangement (Fig. 12A) uses the starburst amacrine cell as the GABAergic interneuron to the DS circuit. A second arrangement (Fig. 12B) substitutes an independent GABAergic amacrine cell and is similar to a scheme described by Ariel and Daw (1982b, Fig. 14, scheme 2).

Of the two possible arrangements, Fig. 12A, which utilizes the starburst amacrine cell, is particularly attractive for several reasons. 1) Starburst amacrine cells are known to contain GABA and GABA precursors (Brecha et al. 1988; Vaney and Young 1988). 2) Starburst amacrine cells costratify and make synaptic connections with the dendrites of DS ganglion cells (Famiglietti 1991, 1992). 3) Patch-clamp studies have shown that starburst amacrine cells contain AMPA/KA receptors (Zhou and Fain 1995). 4) Under normal physiological conditions, application of KA, but not NMDA, evokes release of the colocalized neurotransmitter ACh (Linn et al. 1991). Although we cannot offer definitive proof that the GABA output from the starburst amacrine cell is responsible for the null inhibition of directional selectivity, this cell is well situated to meet the needs of the system without additional neural hardware. However, the evidence favoring GABA release from the starburst amacrine cells is circumstantial. The importance of GABA seems clear, but the source of GABA is unknown. We are unable to rule out contributions from other GABA amacrine cells, although these experiments do indicate that GABA release is controlled by AMPA/KA receptors. In due course, this information may indicate which GABA amacrine cells are involved, but definitive evidence will require further experimentation. The emphasis will be on local interactions, because Yang and Masland (1992, 1994) have shown that the receptive field of a DS ganglion cell closely matches the dendritic field. This implies that cellular, as opposed to local dendritic, interactions from starburst or other amacrine cells that extend well beyond the dendritic field of the DS ganglion cell are minimal. In addition, recent recordings have shown that starburst amacrine cells themselves do not produce DS responses (Peters and Masland 1996; Taylor and Wässle 1994). Thus the mechanism of retinal directional selectivity remains a neural computation integrated by the DS ganglion cell through synaptic connections in the inner plexiform layer.

	ACKNOWLEDGEMENTS

  The authors thank Eli Lilly and Company for the generous donation of GYKI-53655.

  Support for this research was provided by National Eye Institute Grants F32-EY-06566 to C. A. Kittila and EY-06515 to S. C. Massey, and a vision core grant. Additional support was provided by an unrestricted grant from Research to Prevent Blindness to the Department of Ophthalmology and Visual Science.

	FOOTNOTES

  Address for reprint requests: C. A. Kittila, Dept. of Ophthalmology and Visual Science, University of Texas Medical School, 6431 Fannin St., Suite 7.024, Houston, TX 77030.

  Received 19 August 1996; accepted in final form 17 October 1996.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• AMES, A., NESBETT, F. B. In vitro retina as an experimental model of the central nervous system. J. Neurochem. 37: 867-877, 1981.[Medline]
• AMTHOR, F. R., OYSTER, C. W., TAKAHASHI, E. S. Morphology of on-off directional-selective ganglion cells in the rabbit retina. Brain Res. 298: 187-190, 1984.[Medline]
• AMTHOR, F. R., TAKAHASHI, E. S., OYSTER, C. W. Morphologies of rabbit retinal ganglion cells with complex receptive fields. J. Comp. Neurol. 280: 97-121, 1989.[Medline]
• ARIEL, M., DAW, N. W. Effects of cholinergic drugs on receptive field properties of rabbit retinal ganglion cells. J. Physiol. Lond. 324: 135-160, 1982a.[Abstract/Free Full Text]
• ARIEL, M., DAW, N. W. Pharmacological analysis of directionally sensitive rabbit retinal ganglion cells. J. Physiol. Lond. 324: 161-185, 1982b.[Abstract/Free Full Text]
• BADIO, B., DALY, J. W. Epibatidine, a potent analgesic and nicotinic agonist. Mol. Pharmacol. 45: 563-569, 1994.[Abstract]
• BARLOW, H. B., LEVICK, W. R. The mechanism of directionally selective units in rabbit's retina. J. Physiol. Lond. 178: 477-504, 1965.[Free Full Text]
• BAUGHMAN, R. W., BADER, C. R. Biochemical characterization and cellular localization of the cholinergic system in chicken retina. Brain Res. 138: 469-485, 1977.[Medline]
• BOOS, R., SCHNEIDER, H., WÄSSLE, H. Voltage- and transmitter-gated currents of AII-amacrine cells in a slice preparation of the rat retina. J. Neurosci. 13: 2874-2888, 1993.[Abstract]
• BORG-GRAHAM, L. J., GRZYWACZ, N. A model of the direction selectivity circuit in retina: transformations by neurons singly and in concert. In: Single Neuron Computations, edited by T. McKenna, J. Davis, and S. F. Zorneter . New York: Academic, 1992, p. 347-375
• BRECHA, N., JOHNSON, D., PEICHL, L., WÄSSLE, H. Cholinergic amacrine cells of the rabbit retina contain glutamate decarboxylase and -aminobutyrate immunoreactivity. Proc. Natl. Acad. Sci. USA 85: 6187-6191, 1988.[Abstract/Free Full Text]
• CALDWELL, J. H., DAW, N. W., WYATT, H. J. Effects of picrotoxin and strychnine on rabbit retinal ganglion cells: lateral interactions for cells with more complex receptive fields. J. Physiol. Lond. 276: 277-298, 1978.[Abstract/Free Full Text]
• COHEN, E. D., MILLER, R. F. The role of NMDA and non-NMDA excitatory amino acid receptors in the functional organization of primate retinal ganglion cells. Visual Neurosci. 11: 317-332, 1994.[Medline]
• COHEN, E. D., MILLER, R. F. Quinoxlaines block the mechanism of directional selectivity in ganglion cells of the rabbit retina. Proc. Natl. Acad. Sci. USA 92: 1127-1131, 1995.[Abstract/Free Full Text]
• COLEMAN, P. A., MILLER, R. F. Do N-methyl-D-aspartate receptors mediate synaptic responses in the mudpuppy retina? J. Neurosci. 8: 4728-4733, 1988.[Abstract]
• DIXON, D. B., COPENHAGEN, D. R. Two types of glutamate receptors differentially excite amacrine cells in the tiger salamander retina. J. Physiol. Lond. 449: 589-606, 1992.[Abstract/Free Full Text]
• FAMIGLIETTI, E. V. "Starburst" amacrine cells and cholinergic neurons: mirror-symmetric ON and OFF amacrine cells of rabbit retina. Brain Res. 261: 138-144, 1983.[Medline]
• FAMIGLIETTI, E. V. Synaptic organization of starburst amacrine cells in rabbit retina: analysis of serial thin sections by electron microscopy and graphic reconstruction. J. Comp. Neurol. 309: 40-70, 1991.[Medline]
• FAMIGLIETTI, E. V. Dendritic co-stratification of ON and ON-OFF directionally selective ganglion cells with starburst amacrine cells in rabbit retina. J. Comp. Neurol. 324: 322-335, 1992.[Medline]
• FELLER, M. B., WELLIS, D. P., STELLWAGEN, D., WERBLIN, F. S., SHATZ, C. J. Requirement for cholinergic synaptic transmission in the propagation of spontaneous retinal waves. Science Wash. DC 272: 1182-1187, 1996.[Abstract]
• GRZYWACZ, N. M., AMTHOR, F. R. Facilitation in ON-OFF directionally selective ganglion cells of the rabbit retina. J. Neurophysiol. 69: 2188-2199, 1993.[Abstract/Free Full Text]
• HAMASSAKI-BRITTO, D. E., HERMANS-BORGMEYER, I. H., HEINEMANN, S., HUGHES, T. E. Expression of glutamate receptor genes in the mammalian retina: the localization of Glu1 through Glu7 mRNAs. J. Neurosci. 13: 1888-1898, 1993.[Abstract]
• HILDRETH, E. C. AND KOCH, C. The analysis of visual motion: from computational theory to neuronal mechanisms. In: Annual Review of Neuroscience, edited by W. M. Cowan, E. M. Shooter, C. F. Stevens, and R. F. Thompson. Palo Alto: Annual Reviews, 1987, vol. 10, p. 477-533.
• HUGHES, T. E., HERMANS-BORGMEYER, I., HEINEMANN, S. Differential expression of glutamate receptor genes (GluR1-5) in the rat retina. Visual Neurosci. 8: 49-55, 1992.[Medline]
• KITTILA, C. A., MASSEY, S. C. The effect of ON pathway blockade on directional selectivity in rabbit retina. J. Neurophysiol. 73: 703-712, 1995a.[Abstract/Free Full Text]
• KITTILA, C. A., MASSEY, S. C. Directionally selective ganglion cells show sensitivity to glutamate under cholinergic blockade (Abstract). Invest. Ophthalmol. Visual Sci. 36, Suppl.: 865, 1995b.
• KITTILA, C. A., MASSEY, S. C. Pharmacological inputs to directionally selective retinal ganglion cells in the rabbit retina (Abstract). Invest. Ophthalmol. Visual Sci. 37, Suppl.: 1056, 1996.
• KOCH, C., POGGIO, T., TORRE, V. Retinal ganglion cells: a functional interpretation of dendritic morphologies. Philos. Trans. R. Soc. Lond. B Biol. Sci. 298: 222-264, 1982.
• LEVICK, W. R. Another tungsten microelectrode. Med. Biol. Eng. 10: 510-515, 1972.[Medline]
• LINN, D. M., BLAZYNSKI, C., REDBURN, D. A., MASSEY, S. C. Acetylcholine release from the rabbit retina mediated by kainate receptors. J. Neurosci. 11: 111-122, 1991.[Abstract]
• MASLAND, R. H., AMES, A. Responses to acetylcholine of ganglion cells in an isolated mammalian retina. J. Neurophysiol. 39: 1220-1235, 1976.[Abstract/Free Full Text]
• MASLAND, R. H., MILLS, J. W. Autoradiographic identification of acetylcholine in the rabbit retina. J. Cell Biol. 83: 159-178, 1979.[Abstract/Free Full Text]
• MASSEY, S. C. Cell types using glutamate as a neurotransmitter in the vertebrate retina. Prog. Ret. Res. 9: 399-425, 1990.
• MASSEY, S. C., MILLER, R. F. Glutamate receptors of ganglion cells in the rabbit retina: evidence for glutamate as a bipolar cell transmitter. J. Physiol. Lond. 405: 635-655, 1988.[Abstract/Free Full Text]
• MASSEY, S. C., MILLER, R. F. N-methyl-D-aspartate receptors of ganglion cells in rabbit retina. J. Neurophysiol. 63: 16-29, 1990.[Abstract/Free Full Text]
• MASSEY, S. C., MAGUIRE, G. The role of glutamate in retinal circuitry. In: Excitatory Amino Acids and Synaptic Transmission, edited by H. Wheal, and A. Thomson , 2nd ed.. New York: Academic, 1995, p. 201-221
• MILLER, R. F., ZALUTSKY, R. A., MASSEY, S. C. A perfused rabbit retina preparation suitable for pharmacological studies. J. Neurosci. Methods 16: 309-322, 1986.[Medline]
• MITTMAN, S., TAYLOR, W. R., COPENHAGEN, D. R. Concomitant activation of two types of glutamate receptor mediates excitation of salamander retinal ganglion cells. J. Physiol. Lond. 428: 175-197, 1990.[Abstract/Free Full Text]
• NEAL, M. J. Cholinergic mechanisms in the vertebrate retina. Prog. Ret. Res. 191-212, 1983.
• NEAL, M. J., DAWSON, C. Muscarinic cholinergic receptors in rabbit retina. J. Pharm. Pharmacol. 37: 60-61, 1985.[Medline]
• O'MALLEY, D. M., SANDELL, J. H., MASLAND, R. H. Co-release of acetylcholine and GABA by the starburst amacrine cells. J. Neurosci. 12: 1394-1408, 1992.[Abstract]
• OYSTER, C. W. The analysis of image motion by the rabbit retina. J. Physiol. Lond. 199: 613-635, 1968.[Abstract/Free Full Text]
• OYSTER, C. W., AMTHOR, F. R., TAKAHASHI, E. S. Dendritic architecture of on-off direction-selective ganglion cells in the rabbit retina. Vision Res. 33: 579-608, 1993.[Medline]
• PATERNAIN, A. V., MORALES, M., LERMA, J. Selective antagonism of AMPA receptors unmasks kainate receptor-mediated responses in hippocampal neurons. Neuron 14: 185-189, 1995.[Medline]
• PETERS, B. N., MASLAND, R. H. Responses to light of starburst amacrine cells. J. Neurophysiol. 75: 469-480, 1996.[Abstract/Free Full Text]
• PURO, D. G. Cholinergic systems. In: Retinal Transmitters and Modulators: Models for the Brain, edited by and W. W. Morgan . Boca Raton, FL: CRC, 1985, p. 63-91
• QIAN, C. G., LI, T. C., SHEN, T. Y., LIBERTINEGARAHAN, L., ECKMAN, J., BIFTU, T., IP, S. Epibatidine is a nicotinic analgesic. Eur. J. Pharmacol. 250: 13-14, 1993.
• SCHMIDT, M., HUMPHREY, M. F., WÄSSLE, H. Action and localization of acetylcholine in the cat retina. J. Neurophysiol. 58: 997-1015, 1987.[Abstract/Free Full Text]
• SEEBURG, P. H. The TiPS/TINS Lecture: the molecular biology of mammalian glutamate receptor channels. Trends Pharmacol. Sci. 14: 297-303, 1993.[Medline]
• SEEBURG, P. H. The role of RNA editing in controlling glutamate receptor channel properties. J. Neurochem. 66: 1-5, 1996.[Medline]
• SMITH, R. D., GRZYWACZ, N. M., BORG-GRAHAM, L. J. Is the input to a GABAergic synapse the sole asymmetry in turtle's retinal directional selectivity? Visual Neurosci. 13: 423-439, 1996.[Medline]
• SPANDE, T. F., GARRAFFO, H. M., EDWARDS, M. W., YEH, H. J. C., PANNELL, L., DALY, J. W. Epibatidine: a novel (chloropyridyl)azabicycloheptane with potent analgesic activity from an Ecuadoran poison frog. J. Am. Chem. Soc. 114: 3475-3478, 1992.
• SUGIYAMA, H., DANIELS, M. P., NIRENBERG, M. Muscarinic acetylcholine receptors of the developing retina. Proc. Natl. Acad. Sci. USA 74: 5524-5528, 1977.[Abstract/Free Full Text]
• TAUCHI, M., MASLAND, R. H. The shape and arrangement of the cholinergic neurons in the rabbit retina. Proc. R. Soc. Lond. B Biol. Sci. 220: 501-508, 1984.[Medline]
• TAYLOR, W. R., CHEN, E., COPENHAGEN, D. R. Characterization of spontaneous synaptic currents in salamander retinal ganglion cells. J. Physiol. Lond. 486: 207-221, 1995.[Abstract/Free Full Text]
• TAYLOR, W. R. AND WÄSSLE, H. Receptive field properties of cholinergic amacrine cells in the rabbit retina (Abstract). Invest. Ophthalmol. Visual Sci. 36, Suppl. 864: 3960, 1994.
• TORRE, V., POGGIO, T. A synaptic mechanism possibly underlying directional selectivity to motion. Proc. R. Soc. Lond. B Biol. Sci. 202: 409-416, 1978.[Abstract/Free Full Text]
• VANEY, D. I. "Coronate" amacrine cells in the rabbit retina have the "starburst" dendritic morphology. Proc. R. Soc. Lond. B Biol. Sci. 220: 501-508, 1984.
• VANEY, D. I., COLLIN, S. P., YOUNG, H. M. Dendritic relationships between cholinergic amacrine cells and direction-selective retinal ganglion cells. In: Neurobiology of the Inner Retina, edited by R. Weiler, and N. N. Osborne . Berlin: Springer-Verlag, 1989, vol. H31, p. 157-168
• VANEY, D. I., YOUNG, H. M. GABA-like immunoreactivity in cholinergic amacrine cells of the rabbit retina. Brain Res. 438: 369-373, 1988.[Medline]
• WATKINS, J. C., KROGSGAARD-LARSEN, P., HONORÉ, T. Structure-activity relationships in the development of excitatory amino acid receptor agonists and competitive antagonists. Trends Pharmacol. Sci. 11: 25-33, 1990.[Medline]
• WYATT, H. J., DAW, N. W. Specific effect of neurotransmitter antagonists on ganglion cells in rabbit retina. Science Wash. DC 191: 204-205, 1976.[Abstract/Free Full Text]
• YANG, G., MASLAND, R. H. Direct visualization of the dendritic and receptive fields of directionally selective retinal ganglion cells. Science Wash. DC 258: 1949-1952, 1992.[Abstract/Free Full Text]
• YANG, G., MASLAND, R. H. Receptive fields and dendritic structure of directionally selective retinal ganglion cells. J. Neurosci. 14: 5267-5280, 1994.[Abstract]
• ZHOU, Z. J., FAIN, G. L. Neurotransmitter receptors of starburst amacrine cells in rabbit retinal slices. J. Neurosci. 15: 5334-5345, 1995.[Abstract]

0022-3077/97 $5.00 Copyright ©1997 The American Physiological Society