The retina has emerged as a model system for the study of γ-aminobutyric acid (GABA)¹ receptors. It possesses a diversity of GABA_A receptors (GABA_AR) (Greferath et al., 1995), at least two types of GABA_B receptor (GABA_BR) (Matthews et al., 1994; Tian and Slaughter, 1994), and a GABA_C receptor (GABA_CR) (Cutting et al., 1991). These receptors produce a variety of electrogenic effects including the ionotropic increase in chloride conductance and the metabotropic modulation of a calcium conductance and activation of a putative potassium conductance. The ability of GABA to produce this multiplicity of effects permits a precise and varied integration of synaptic information.

While a plethora of GABA receptors have been identified in retina, often they have been discretely localized to small groups of neurons. For example, in white perch retina the GABA_CR has been described in a rod-driven horizontal cell (H4) (Qian and Dowling, 1993), while in catfish retina this receptor was found in cone-driven, but not rod-driven, horizontal cells (Dong et al., 1994; Takahashi et al., 1994). The baclofen-sensitive GABA_BR has been identified in fish ganglion cells but was not seen in fish horizontal or bipolar cells (Bindokas and Ishida, 1991). In one type of goldfish bipolar cell a cis-aminocrotonic acid (CACA)-sensitive, but not the baclofen-sensitive, GABA receptor was detected (Matthews et al., 1994). This restricted expression fits a model in which GABARs are precisely localized so that each receptor subtype is associated with particular neurons. The implication is that there is a functional correlation between receptor properties, such as conductance and kinetics, and the physiology of the neuron. In this context, Pan and Lipton (1995) reported that the GABA_CR found in rat bipolar cells produced a smaller and slower current than the GABA_AR. It is tempting to conclude that there is a good match between this receptor and the relatively slow and small responses of the rod pathway.

An alternative model of GABA receptor diversity is that multiple GABARs may be localized in one cell and thereby expand the inhibitory capability of that neuron. If present, this confluence of GABA receptor subtypes could be particularly important in retinal ganglion cells, the recipients of powerful GABAergic input from amacrine cells and the final integrators of retinal information. Therefore, we examined retinal ganglion cells to determine their complement of GABA receptors. We previously reported that amphibian ganglion cells possess two types of ionotropic GABA receptor: GABA_A and GABA_C (Zhang and Slaughter, 1995). We find that these neurons also contain two types of metabotropic GABA receptor, one sensitive to baclofen and the other to CACA. Thus, ganglion cells possess every type of GABA receptor currently identified. Through these different receptors, GABA regulates chloride and potassium conductances and modulates voltage-dependent calcium conductances. This permits a graded level of inhibition by parallel deployment of different conductance mechanisms. The GABA receptors can be differentially modulated by the internal or external environment of the ganglion cell, indicating that the balance of receptor inhibition can be tuned.

Experiments were performed in aquatic tiger salamander (Ambystoma tigrinum) retinal slice preparation, which has been described in detail previously (Werblin, 1978; Wu, 1987). In brief, the animal was decapitated, double pithed, and the eye was enucleated. The cornea, lens, and vitreous were removed, exposing the retina. The retina was separated from the pigment epithelium and positioned, photoreceptor side down, on filter paper. The retina and filter paper were sectioned into 200–300-μm thick slices. These sections were placed on their sides in a perfusion chamber and positioned on the stage of an upright microscope fitted with Hoffman modulation optics. The retinal layers and individual neurons could be observed while a microelectrode was positioned for whole cell recording. Low resistance electrodes (<5 MΩ) were used to form gigaseals. Recordings were obtained from neurons in the ganglion cell layer, a somatic layer that Lukasiewicz and Werblin (1988) found to be almost exclusively populated by ganglion cells. In some cases the neurons were stained with Lucifer yellow (0.05% added to pipette solution) and retrospectively identified as ganglion cells.

A few experiments were performed on isolated ganglion cells. A similar dissection procedure and standard dissociation procedures were employed, as detailed previously (Pan and Slaughter, 1995).

Whole cell pipette solution consisted of (in mM): 106 K-gluconate, 5 NaCl, 2 MgCl₂, 5 EGTA, and 5 HEPES, buffered to pH 7.4 with KOH. To reduce calcium channel run-down, the pipette solution also contained an “ATP-regenerating cocktail” consisting of 4 mM ATP, 20 mM phosphocreatine, and 50 U/ml creatine phosphokinase. In experiments to study the CACA-sensitive GABA receptor, the ATP-regenerating system was replaced with 4 mM ATP and 0.3 mM GTP. At the recommendation of one reviewer, a few experiments were performed using nystatin perforated patches. In these cases, the same internal solution was used without the ATP-regenerating medium.

Both voltage steps and ramps were used to evoke currents. Since the barium currents that we measured were sustained and showed very little inactivation over this time course, we found that results using the step and ramp were similar. For convenience, ramp data were most often obtained. The figures in this paper display the original data and are not corrected for electrode tip potential or series resistance. The cells had high input resistances so leak subtraction was not necessary. The tip potential was measured for each pipette solution (Neher, 1992), and the correction factor is mentioned in results, particularly where reversal potential measurements are presented. The amplifier did not have series resistance compensation. Except for perforated patch recordings, access resistance was generally between 6 and 15 MΩ. This resulted in an offset in the peak barium currents. However, since we were monitoring sustained currents and examining these currents over a wide range of potentials, this offset did not significantly influence our results. In a few cases, where a large shift occurred, the data were discarded. When both tip potential and series resistance voltage offsets were subtracted, the barium current was calculated to peak near +10 mV. This information was not a topic of this paper, so these calculations were not routinely done for each cell and the raw data is shown without these corrections. Voltage clamp experiments were performed using an Axoclamp 2A amplifier in continuous voltage clamp mode and PCLAMP software on a Gateway 486 computer equipped with a Labmaster A/D board. Analysis was performed employing PCLAMP, EXCEL, and Origin software.

Control amphibian Ringer's solution consisted of (in mM): 111 NaCl, 2.5 KCl, 1.8 CaCl₂, 1 MgCl₂, 10 dextrose, and HEPES buffered to pH 7.8 and oxygenated. When calcium currents were studied in isolation, 10 mM BaCl₂ and 40 mM tetraethyl ammonium chloride (TEA·Cl) replaced equimolar NaCl in the extracellular medium. Tetrodotoxin (1 μM) was used to block sodium currents. Application of 25 μM CdCl₂ suppressed most of the inward current under these conditions, indicating that most, if not all, of the observed current arose from calcium channels. Control and drug containing solutions were applied through a gravity fed system to a manifold in connection with the perfusion chamber containing the retinal slice. Valves controlled drug application, and there was a delay of ~10 s due to exchange time of the system. Alternatively, drugs were applied through a puff pipette positioned at the edge of the retina near the cell of interest. In this case, Ringer's solution was still constantly applied to the retinal slice. When antagonists were tested, they were applied first, then after 30–60 s agonist was applied with antagonist. Generally, one cell was recorded in each retinal slice to avoid residual drug effects. Drug effects described in this paper are steady-state values, usually obtained 30–60 s after drug application.

The drug concentrations required to evoke effects in the slice preparation were often higher than anticipated, indicating that there was a diffusion or uptake barrier. For example, 100 μM GABA had only a small effect in the slice preparation, although it produced a near maximal effect when applied to isolated ganglion cells (Pan and Slaughter, 1995). This was true for glutamate as well (Shen and Slaughter, unpublished observations). Therefore, there is probably a disparity between the dose applied and the effective drug concentration at the receptor. This phenomenon has been noted in retinal slice studies from other laboratories. Also, during the course of these studies we noted a decrease in the response to baclofen and to cis-aminocrotonic acid. Whether this represented a seasonal change or other effect was unclear. A series of initial experiments were performed using 100 μM baclofen, later 50 μM baclofen was used. In results, the average effect of 100 μM baclofen is reported to be greater than the effect of 50 μM, but this is probably due to the time periods when the studies were completed. Some experiments compare the effects of second messenger antagonists with the average effect of 100 or 50 μM baclofen. These antagonist experiments were performed in periods that overlapped the experiments from which the average data were obtained, and therefore represent ganglion cell populations with appropriate control responses. In CACA experiments, 5 μM doses produced nearly 50% suppression in early experiments (winter), but only about 20% suppression in later experiments (summer). Again, this may have been due to seasonal changes or other factors. When comparing data between experiments, we only contrasted effects measured at the same time period (usually the same day or week). All statistical data are presented as means ± SEMs.

SR95531, 2-hydroxysaclofen, nifedipine, nimodipine, trifluoperazine, KN-62, forskolin, H-9, Rp-cAMP, microcystin, calmidazolium, ruthenium red, heparin, N^G-monomethyl-l-arginine, and sodium nitroprusside were purchased from Research Biochemicals, Inc. (Natick, MA). CACA was obtained from Tocris Cookson (St. Louis, MO). Cyclosporin A was obtained from Calbiochem Corp. (San Diego, CA). CGP35348 and baclofen were gifts from Ciba-Geigy (Basel, Switzerland). All other chemicals were obtained from Sigma Chemical Co. (St. Louis, MO).

Figure 1 GABA induced currents in retinal ganglion cells. This shows the current responses of a ganglion cell to voltage steps under control conditions (A), during 200-ms puffs of 500 μM GABA from a pipette near the proximal edge of the retina (B), and (more ...)

A notch is apparent in the I-V curve (Fig. 1 E) in the region where high voltage–activated (HVA) calcium currents might arise. This suggests that an additional effect of GABA is to suppress a voltage-activated calcium conductance, resulting in a reduction in an outward potassium conductance. The calcium current was blocked to test this hypothesis. When a protocol similar to that in Fig. 1 E was used in the presence of 100 μM external cadmium (this recording is from a different neuron than the one in Fig. 1, A–E), the GABA I-V curve was relatively linear and lacked the notch (Fig. 1 F). The data support the supposition that the notch was due to the GABA-induced reduction of a calcium-activated outward current. Ensuing experiments indicate that GABA does suppress calcium channel currents.

Baclofen suppressed a portion of the HVA barium current in ganglion cells (Fig. 2). The neuron was held at −70 mV, stepped to −120 mV, and then the voltage was ramped to +60 mV in 1 s. This paradigm evoked an inward barium current (Fig. 2 A1) that was suppressed approximately 30% by 100 μM baclofen (Fig. 2 A2). Adding 2-hydroxysaclofen (1 mM), an antagonist of baclofen-sensitive GABA receptors (Kerr et al., 1988), restored the inward current close to its control value (Fig. 2 A3). The lower three panels, taken from recordings of another ganglion cell, show that picrotoxin, an antagonist of both GABA_ARs and GABA_CRs, did not block the effect of baclofen on the HVA barium current. In experiments on 73 cells, 50 μM baclofen reduced the HVA current by 29 ± 1% (mean ± SEM), while in another 22 ganglion cells 100 μM baclofen reduced the barium current by 40 ± 3% (see materials and methods regarding this difference). In the presence of 100 μM picrotoxin, 100 μM baclofen reduced the barium current by 44 ± 3% (n = 6), indicating that picrotoxin did not reduce baclofen's action. This pharmacology leads to the conclusion that amphibian ganglion cells possess a baclofen-sensitive GABA_BR that can down-regulate HVA calcium currents.

Figure 2 Baclofen reduces a voltage-activated barium current. The voltage was ramped from −120 to +60 mV in 1 s. The barium current was isolated and enhanced by replacing extracellular calcium with 10 mM barium and adding 40 mM TEA (equimolar (more ...)

In previous studies (Tian and Slaughter, 1994), CGP35348 was found to block baclofen but 2-hydroxysaclofen did not. In the present experiments, both were effective baclofen antagonists. The difference is that in the prior experiments the concentration of 2-hydroxysaclofen was 100 μM while in the present experiments that concentration was raised to 1 mM.

To determine if ganglion cells possessed another GABA receptor, we applied GABA in the presence of antagonists of the GABA_AR, GABA_CR, and the baclofen-sensitive GABA_BR. Since 2-hydroxysaclofen is a weak and competitive antagonist of baclofen-sensitive GABA_BRs, we sometimes chose to saturate the baclofen-sensitive receptor instead of blocking it. When GABA was applied in the presence of SR95531, picrotoxin, and baclofen, it was still able to produce an additional suppression of the barium current (Fig. 3 A). In nine cells tested in this manner, GABA reduced the barium current by 11 ± 2%. Another GABAR has been identified in goldfish retina and shown to down-regulate HVA calcium channels in bipolar cells (Matthews et al., 1994). This has been proposed as a novel type of GABA_BR. It can be identified pharmacologically by its sensitivity to low micromolar concentrations of CACA and its insensitivity to bicuculline, SR95531, picrotoxin, baclofen, or 2-hydroxysaclofen.

Figure 3 Ganglion cells possess a CACA-sensitive GABA receptor. (A) There is a GABA effect on barium current not due to the GABAAR, the baclofen sensitive GABABR, or the GABACR. When these receptors were blocked (GABAAR and GABACR) or saturated (GABABR), GABA (more ...)

This receptor is also present in amphibian ganglion cells. A low micromolar concentration of CACA was able to reduce the HVA barium current in a reversible manner (Fig. 3 B). CACA produced a 40% decrease in the barium current in this neuron. We saw no evidence that low doses of CACA changed the conductance of the cell (monitored at negative potentials where voltage-dependent channels were not activated), although at much higher concentrations it can activate the GABA_AR as well as the GABA_CR in these cells (Zhang and Slaughter, 1995). The lower set of panels, taken from another ganglion cell, show that the effect of CACA persisted when it is applied in the presence of picrotoxin, which blocked the GABA_A&CRs, and 2-hydroxysaclofen, which blocked the baclofen-sensitive GABA_BR. As noted by Matthews et al. (1994), CACA effects were robust when the inside of the cell was dialyzed with GTP. With 0.3 mM GTP in the pipette, CACA reduced the barium current by a mean of 46 ± 5% in nine cells. When GTP was not included in the pipette solution, CACA produced a smaller reduction in the barium current of 23 ± 2% in 14 cells. This is different than the action of baclofen, which did not depend on GTP dialysis. Without GTP dialysis the recovery after CACA treatment was poor. All subsequent CACA experiments described in this report were performed using GTP dialysis. The GABA experiments in Fig. 3 did not include GTP.

Since different internal solutions were used in recording baclofen-sensitive and CACA-sensitive currents, nystatin perforated patch recordings were used to compare the actions of baclofen, CACA, and GABA while limiting the effects of cell dialysis. High concentrations were used to produce maximal responses to each agonist (50 μM CACA, 100 μM baclofen, and 200 μM GABA). The retina slice was pretreated and continuously superfused with picrotoxin (200 μM) to avoid the effects on ionotropic GABA receptors. In five cells, baclofen was found to reduce the barium current by a mean of 28% (from a control value of 1,028 ± 169 pA to 708 ± 109 pA); CACA reduced the current by 9% (to 924 ± 135 pA); GABA reduced the current by 46% (to 534 ± 82 pA). The response to each agonist was statistically significant (P < 0.05, Wilcoxin's signed-ranks test).

The ionotropic GABA_CR is sometimes referred to as CACA-sensitive because CACA can be more effective at activating the GABA_CR than the GABA_AR (Feigenspan et al., 1993; Qian and Dowling, 1993; Pan and Lipton, 1995). We previously reported that high concentrations of CACA were needed to stimulate the GABA_CR, and these concentrations also activated the GABA_AR (Zhang and Slaughter, 1995). This is in contrast to the receptor identified in Fig. 3, which is sensitive to low micromolar concentrations of CACA. To avoid confusion with reports describing CACA sensitivity of the GABA_CR or GABA_AR, we refer to this receptor as the CACA-sensitive GABA_BR (GABA_B-CACAR).

Figure 4 The GABAB-BACLOFENR suppresses a ω-conotoxin–GVIA-sensitive calcium channel. The upper left panel in A shows that baclofen reversibly reduces barium current (1 = control; 2 = baclofen; 3 = recovery). The adjacent (more ...)

The N-type channel blocker, ω-conotoxin–GVIA, had a much larger effect on both the calcium current and baclofen's action (Fig. 4, bottom). In eight cells, ω-conotoxin reduced the barium current by 57 ± 7%. In the presence of ω-conotoxin, 50 μM baclofen produced only a 14 ± 3% suppression of the remaining reduced barium current. Of the eight cells used in these experiments, seven had been previously tested with baclofen alone. Baclofen had produced a mean barium current suppression of 37 ± 2%. If baclofen acted exclusively on non–N-type calcium channels, it would be expected to suppress 86% of the remaining calcium current in the presence of ω-conotoxin. If ω-conotoxin-GVIA is acting selectively, this indicates that a large fraction of the baclofen effect is on the N-type channel.

Figure 5 The GABAB-CACAR modulates a dihydropyridine-sensitive barium current. The top trace shows recordings from one ganglion cell in which the barium current was evoked by a voltage ramp. The current was suppressed by 50 μM nimodipine or 5 μM (more ...)

Figure 7 GDP-β-S blocks the effects of CACA and baclofen. Whole cell recordings were performed while including 1 mM GDP-β-S in the pipette solution. The graphs show the peak barium current measured at fixed intervals before, during, and after (more ...)

Figure 6 Effects of CACA and baclofen on ganglion cell spike activity. (A) Ganglion cell spiking was elicited by brief (300 ms) current injections. Both baclofen and CACA decreased spike activity. (B) The reduction in spike activity produced by baclofen was (more ...)

Figure 8 Depolarizing prepulses reduce the effect of baclofen. Barium currents were elicited by stimulating pulses of +10 mV. If the conditioning potential was −70 mV, baclofen significantly reduced the barium current. But after a conditioning (more ...)

The effect of baclofen was reduced, but not eliminated, by a variety of agents that interfere with the protein kinase A (PKA) pathway. Examples of the effects of these agents and a summary of the data are shown in Fig. 9. Except for forskolin, all drugs were dialyzed into the cell by including them in the pipette solution. Therefore, the effects of baclofen during dialysis of these drugs were compared to effects of baclofen on a control population of neurons. The control data used, presented above, are that 50 μM baclofen reduced the HVA calcium current by a mean of 29% (n = 73), whereas 100 μM baclofen reduced this current by 40% (n = 22). When H-7, a protein kinase inhibitor, was dialyzed into 11 cells, the effect of 100 μM baclofen was reduced to approximately half the normal magnitude (19 ± 3%). But Rp-cAMP, a specific blocker of protein kinase A, had little effect on the response to baclofen. In the presence of Rp-cAMP, 50 μM baclofen suppressed the calcium current by 22 ± 2% (n = 24), about 75% of control. If the internal levels of cAMP were increased by dialysis then the effect of 100 μM baclofen was reduced by about 50% (19 ± 2% suppression in six cells). Forskolin was superfused onto the retina, and, therefore, the effect of forskolin on the calcium current could be measured. When ganglion cells were exposed to forskolin, it produced a 27 ± 3% suppression of the calcium current. And in the presence of forskolin, 50 μM baclofen produced a 15 ± 3% reduction in the calcium current (n = 11). If baclofen was acting on calcium channels that were not affected by forskolin, a suppression of 44% by baclofen would be anticipated. Therefore, the action of baclofen was significantly occluded by forskolin. The results suggest that activation of the PKA pathway interferes with the effects of GABA_B-BACLOFENR stimulation.

Figure 9 PKA analogues reduce the effect of baclofen. The top row (A) shows examples for each of the four agents tested while the bottom row (B) graphs the data for all cells tested. In the top row, trace 1 = PKA analogue and trace 2 = baclofen (more ...)

A few antagonists of other second messenger kinase pathways were tested and found not to alter the action of baclofen. These included a protein kinase C inhibitor, staurosporine, and a tyrosine kinase inhibitor, genistein.

The effects of PKA agonists and antagonists suggest a role for this pathway in mediating the response to baclofen, whereas the prepulse data suggest a direct G-protein action that is independent of second messengers such as PKA. This discrepancy could be explained if the PKA system autonomously regulated the baclofen-sensitive calcium channel. Prepulse experiments supported this proposition (Fig. 10). While the effect of baclofen was reversed by prepulse facilitation (response 2 in Fig. 10), the action of forskolin was unaffected by the prepulse (response 3). In addition, the effect of baclofen and forskolin together (response 4) was less than the sum of each agent individually. This indicates that baclofen and forskolin act in parallel to reduce the calcium channel activation. Forskolin acts by a mechanism that differs from that produced by GABA_B-BACLOFENR activation, yet occludes the effect of baclofen.

Figure 10 Prepulse facilitation does not influence the effect of PKA. The left side of this figure shows that when barium currents are elicited by stepping to +5 mV from a holding potential of −70 mV (1), baclofen reduces more barium current (more ...)

The action of baclofen was suppressed by calmidazolium, an inhibitor of calcium-calmodulin. In ten ganglion cells, internal dialysis with 100 μM calmidazolium reduced the effect of baclofen by an average of 72% (Fig. 11 C). Internal dialysis with 100 μM trifluoperazine, another calcium-calmodulin inhibitor, reduced baclofen's action by 57% (n = 7). A third inhibitor, W-7 (100 μM added to pipette), suppressed 43% of the baclofen effect (n = 5).

Figure 11 Calmodulin-dependent enzymes regulate the GABAB-BACLOFENR. This figure shows the effects of baclofen on barium currents in the presence of agents that inhibit calcium-calmodulin (Calmidazolium), phosphatases (Microcystin), and internal calcium release (more ...)

Calcium-calmodulin can activate protein kinases and phosphatases. We therefore examined the effect of microcystin LR, an inhibitor of protein phosphatases. An example is shown in Fig. 11, where 50 μM baclofen (Fig. 11 A, trace 2) suppressed the control barium current (trace 1) by 14% in this cell. Microcystin alone produced a very small suppression of the barium current in the same cell (Fig. 11 B, trace 1). But 50 μM baclofen in the presence of microcystin (trace 2) reduced the barium current by 29%. In eleven cells tested in this manner, microcystin enhanced the effect of 50 μM baclofen by 161% (from a mean calcium current suppression of 23 ± 3% to a mean of 37 ± 3%). Internal dialysis of cyclosporin A, a calcineurin phosphatase inhibitor, augmented baclofen's mean effect to 152% of control (n = 7). Okadaic acid, another phosphatase inhibitor, produced a similar enhancement. While blockers of calmodulin-dependent phosphatases had a dramatic effect on baclofen's action, KN-62, a very potent blocker of some calcium-calmodulin kinases, had only a slight suppressive effect (data not shown). Internal dialysis with 50–100 μM KN-62 reduced the baclofen effect by 22% (n = 5).

We did not find any evidence that the calcium-calmodulin modulation of the GABA_BR pathway resulted from calcium release from internal stores. Neither heparin, a blocker of the IP3 pathway, nor ruthenium red, a blocker of the ryanodine receptor, reduced the effect of baclofen (Fig. 11, D and E). Ruthenium red was tested by placing it in the electrode because extracellular application significantly reduced membrane barium currents. In three cells, 50 μM baclofen in the presence of ruthenium red reduced the barium current by 30 ± 5%. Heparin was also included in the patch pipette solution. After heparin dialysis, 50 μM baclofen still reduced the barium current by 33 ± 1% (n = 7). This suppression by 50 μM baclofen, in the presence of ruthenium red or heparin is very similar to the average effect of 50 μM baclofen alone (29 ± 1% suppression).

Figure 12 The GABAB-CACAR pathway is enhanced by nitric oxide. CACA's effect was enhanced when the retina was superfused with 8-bromo-cGMP or sodium nitroprusside. In each graph, the peak voltage-activated barium current was measured every 30 s and plotted. (more ...)

Blocking of calcium-calmodulin with calmidazolium, trifluoperazine, or W-7 largely suppressed GABA_B-BACLOFENR action. None of these agents alone produced an obvious alteration in the calcium current. Subtle changes would be difficult to detect because all drugs were applied by internal dialysis and therefore were compared statistically to other neurons that served as a control. Calcium-calmodulin stimulates a number of pathways, including specific phosphatases and kinases. KN-62, a potent blocker of some calcium-calmodulin kinases, did not suppress the effect of baclofen. Blockers of calcium-calmodulin dependent phosphatases, such as cyclosporin A and microcystin, enhanced the effect of baclofen. Phosphatase inhibitors did not appear to have a direct effect on the calcium current, indicating that phosphorylation enhanced the GABA_B-BACLOFENR pathway.

Thus, calmodulin-dependent phosphatases down-regulated the GABA_B-BACLOFENR. But calmodulin antagonists reduced the effect of baclofen, indicating that calmodulin stimulation is required for GABA_B-BACLOFENR activity. Therefore, another calmodulin pathway must be involved, possibly a calmodulin-dependent kinase. The ineffectiveness of KN-62 prevents clear identification of a calmodulin kinase, but the action of H-7 (coupled with the ineffectiveness of Rp-cAMP) could be interpreted as a block of a calmodulin-kinase pathway. While the precise calmodulin kinase pathway was not identified, these experiments suggest a regulatory scheme In which calmodulin modulates the activity of the GABA receptor by both up-regulation through kinase activation and down-regulation through phosphatase activation.

A tentative model of GABA_B-BACLOFENR transduction consistent with our data (Fig. 13) proposes that baclofen down-regulates N-type calcium channels through a direct G-protein pathway. The GABA_B-BACLOFENR is biphasically modulated by calmodulin: up-regulation of the receptor by calmodulin, probably through a kinase, and down-regulation through a calmodulin-dependent phosphatase. The balance between these two calmodulin-dependent enzymes would determine the phosphorylation state, and therefore the activity, of this transduction pathway. In addition, a parallel PKA pathway also down-regulates the same population of calcium channels.

Figure 13 A tentative model of the baclofen transduction pathway. Baclofen reduces the voltage-activated calcium current via a direct G-protein interaction with N-type calcium channels. Through a parallel mechanism, the cAMP pathway also reduces current through (more ...)

One ramification of the model is that the calcium-calmodulin system provides a feedback loop between the calcium channel and the GABA_B-BACLOFENR. The GABA_B-BACLOFENR pathway does not appear to require calcium from internal stores (based on the heparin and ruthenium red data). Therefore, calcium influx through the voltage-sensitive calcium channel may stimulate calmodulin, which up-regulates the GABA_B-BACLOFENR. A check on this negative feedback would be the phosphatase system, which turns off the baclofen receptor. This mechanism could help to regulate the internal calcium concentration. The model implies that the calmodulin activation of kinases and phosphatases have different calcium sensitivities or that there is higher gain in the kinase circuit.

Many regulatory mechanisms also produce differential effects on ionotropic GABARs. For example, protein kinase C reduces GABA_CR and enhances GABA_AR currents (Dong and Werblin, 1994; Feigenspan and Bormann, 1994; Wellis and Werblin, 1995). Similarly, low micromolar concentrations of zinc can reduce the GABA_CR current but enhance the GABA_AR current (Chappell et al., 1995). Zinc is believed to be synaptically released by photoreceptors (Wu et al., 1993). Overall, these modulatory mechanisms can be employed to adjust the balance of inhibitory action produced by the four GABA receptor subtypes and suggests that the effects of the ionotropic receptors are complementary, as are the metabotropic receptors.

In summary, the range of GABA receptors found on ganglion cells coupled with the ability to modulate their effects extends the processing capabilities at this final step in the retinal network. Modulation of inhibition can be a mechanism to shift the balance of weighted inputs or postsynaptic processing in the retina's output neurons.

We thank Ciba-Geigy for the generous gifts of baclofen and CGP35348.