Retinal Bipolar Cell Input Mechanisms in Giant Danio. II. Patch-Clamp Analysis of on Bipolar Cells

Kwoon Y. Wong, Ethan D. Cohen, John E. Dowling


Glutamate receptors on giant danio retinal on bipolar cells were studied with whole cell patch clamping using a slice preparation. Cone-driven on bipolars (Cbs) and mixed-input on bipolars (Mbs) were identified morphologically. Most Cbs responded to the excitatory amino acid transporter (EAAT) substrate d-aspartate but not to the group III metabotropic glutamate receptor (mGluR) agonist l-(+)-2-amino-4-phosphonobutyric acid (l-AP4) or the AMPA/kainate receptor agonist kainate, suggesting EAATs are the primary glutamate receptors on Cbs. The EAAT inhibitor dl-threo-β-benzyloxyasparate (TBOA) blocked all light-evoked responses of Cbs, suggesting these responses are mediated exclusively by EAATs. Conversely, all Mbs responded to d-aspartate and l-AP4 but not to kainate, indicating they have both EAATs and group III mGluRs (presumably mGluR6). The light responses of Mbs involve both receptors because they could be blocked by TBOA plus (RS)-α-cyclopropyl-4-phosphonophenylglycine (CPPG, a group III mGluR antagonist) but not by either alone. Under dark-adapted conditions, the responses of Mbs to green (rod-selective) stimuli were reduced by CPPG but enhanced by TBOA. In contrast, both antagonists reduced the responses to red (cone-selective) stimuli, although TBOA was more effective. Furthermore, under photopic conditions, TBOA failed to eliminate light-evoked responses of Mbs. Thus on Mbs, rod inputs are mediated predominantly by mGluR6, whereas cone inputs are mediated mainly by EAATs but also by mGluR6 to some extent. Finally, we explored the interactions between EAATs and mGluR6 in Mbs. Responses to d-aspartate were reduced by l-AP4 and vice versa. Therefore mGluR6 and EAATs suppress each other, and this might underlie mutual suppression between rod and cone signals in Mbs.


In the companion paper, we used the electroretinogram (ERG) to study the glutamate receptor mechanisms that underlie the rod and cone contributions to the b-wave in the giant danio retina. Assuming the b-wave is an indicator of on bipolar cell light responses, we confirmed the presence of 2 distinct input mechanisms generating the responses of these cells (Grant and Dowling 1996; Saito et al. 1979). One of these mechanisms, mGluR6 (a group III metabotropic glutamate receptor), was found to account for nearly all rod-driven b-wave activity, consistent with the study by Nawy and Copenhagen (1987). In addition, mGluR6 was found to make a small but significant contribution to the cone-driven b-wave response. However, the majority of the cone b-wave was shown to be mediated by excitatory amino acid transporters (EAATs), based on the observation that the EAAT blocker dl-threo-β-benzyloxyasparate (TBOA) eliminated most of the photopic b-wave.

Although the ERG is a useful tool for analyzing the summed response of many on bipolar cells in a retina, it gives no information about the response of individual bipolar cells. In particular, the ERG study described in the companion paper cannot answer 3 important questions. 1) Because there are exclusively cone-driven and mixed-input on bipolars in teleosts (Sherry and Yazulla 1993), does the mGluR6 component of the cone-driven response reside in just one or both of these bipolar cell subtypes? 2) Because on bipolar cells with both glutamate receptor types exist (Connaughton and Nelson 2000; Grant and Dowling 1996), do the mechanisms interact, and what might be the significance of such an interaction? 3) Because Nawy and Copenhagen (1990) found that intracellular cesium reduced the non-mGluR6 mechanism in mixed on bipolar cells, suggesting that a potassium conductance was involved in the reception of cone signals in fish, is there a third type of glutamate receptor on mixed on bipolars?

To answer these questions, we performed whole cell patch-clamp recording from on bipolar cells in slices using the giant danio. As mentioned above, there are 2 major subtypes of on bipolar cells in teleosts, and they can be distinguished from each other morphologically. Cone-driven on bipolar cells (abbreviated as Cb’s–cone-driven bipolar cells with axon terminals in sublamina b) receive inputs exclusively from cone photoreceptors and have smaller somata and axon terminals (Fig. 1, top left). Mixed on bipolars (Mbs) receive inputs from both rods and cones and have considerably larger somata and axon terminals (Fig. 3, top left). The axon terminals of both cell types are in sublamina b of the inner plexiform layer (IPL), although those of the Mbs are slightly closer to the ganglion cell layer (Sherry and Yazulla 1993). The cells reported in the present study were identified based on these morphological criteria.

FIG. 1.

Cone-driven on bipolars (Cbs) and their responses to glutamate receptor agonists. Top left: morphology of a Cb revealed by Lucifer yellow staining. Its axon terminal in sublamina b is indicated by the arrow. Right: voltage-ramp responses of a Cb to 3 mM d-aspartate, 100 μM kainate, and 300 μM l-(+)-2-amino-4-phosphonobutyric acid (l-AP4). Top inset: raw data used to determine this cell’s response to d-aspartate. Command voltage is shown in the top trace. Holding potential was −39 mV, and a voltage ramp (from −129 to 71 mV) was applied first before and then during the d-aspartate puff illustrated in the bottom trace. Difference between the 2 ramp responses corresponds to the d-aspartate response. Bottom left: responses of another Cb to the same agonists recorded under current clamp. Number next to the response trace on the left indicates the resting potential. Duration of the puffs (indicated by the steps at the bottom) was 1 s. ONL, outer nuclear layer; OPL, outer plexiform layer; INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer.



All experiments were performed on adult giant danios (Segrest Farms, Gibsonton, FL; EkkWill Waterlife Resources, Gibsonton, FL; Central Mass Aquatics, Worcester, MA) between 1.5 and 3 in. in length. Animals were maintained according to Harvard University and National Institutes of Health Guidelines.

Whole cell patch recording

All procedures were approved by the Institutional Animal Care and Use Committee at Harvard University. Animals were dark-adapted for at least 1 h before each experiment. The entire slicing procedure was performed in dim red light. After anesthesia in 0.016% (g/mL) tricaine, animals were decapitated and their eyes removed. After the removal of the cornea and the lens, the retinas were isolated from the pigment epithelium, sclera, and choroid in a chilled solution that consisted of 25% (mL/mL) L-15 culture medium (Invitrogen, Carlsbad, CA) and 75% Ringer solution (see Chemicals and solutions) with picrotoxin, strychnine, and d-(−)-2-amino-5-phosphonopentanoic acid (d-AP5) omitted. Each retina was cut in half and placed photoreceptor side up on a piece of filter paper (type 0.45-μm HA; Millipore, Billerica, MA), and vacuum was applied across the filter for about 1 min to remove some of the vitreous humor and to attach the retina onto the filter. During the suction, the L-15/Ringer solution was continuously added to the retina to prevent it from drying out. The retina–filter complex was held in place on a glass slide and cut into 200-μm slices using a chopper with Feather blades (Ted Pella, Redding, CA). The slices were incubated in the same solution at 8°C for 10 min to 8 h before recording. The slices were not oxygenated during the incubation.

Whole cell patch clamp was used to record from bipolar cells in the slices. Experiments were done under dim red light unless stated otherwise. Slices were rotated 90° and held in that position by slits in 2 rows of silicone cement during recording, and were viewed through a 60× water-immersion objective lens. A microscope with a mercury arc lamp attached was used. The slices were superfused throughout with the Ringer solution or drug solutions at room temperature. The solutions were not oxygenated.

Electrodes were pulled from borosilicate glass (1.5 mm OD, 0.86 mm ID) on a Flaming/Brown P-87 puller (Sutter Instruments, Novato, CA), with resistance in the bath of 5–10 MΩ, and connected to a Dagan 8900 amplifier (Dagan, Minneapolis, MN). After clamping in the whole cell configuration, recording typically started in about 1 min. Data were acquired using PCLAMP software (Axon Instruments, Union City, CA). Signals were low-pass filtered at 1 kHz for most experiments and at 10 kHz when the capacitative current was used to estimate the series resistance, and the sampling rates were 2.02 and 20.4 kHz, respectively. Most voltage-clamp recordings were partially compensated for series resistance. After the compensation, the remaining series resistances were 74.3 ± 14.3 MΩ (SE; n = 16) for cone-driven on bipolar cells and 35.9 ± 1.3 MΩ (n = 31) for mixed on bipolars. These were likely to be overestimates because the amplitudes of the capacitative currents were probably attenuated by the 10-kHz filter. All reversal potential values were adjusted for series resistance with Ohm’s law, Verror = IholdRseries, where Verror is the error in the reversal potential measurement, Ihold is the holding current at the reversal potential, and Rseries is the series resistance. The liquid junction potentials were measured in the manner described in Neher (1992), and they were 9.033 ± 0.076 mV (n = 6) and 13.43 ± 0.28 mV (n = 6) for cesium- and potassium-based internal solutions (see Chemicals and solutions), respectively. For both current- and voltage-clamp experiments, membrane potentials were adjusted for the liquid junction potentials. The reference electrode was an Ag/AgCl wire placed in the bath.

Unless stated otherwise, the white stimulus was a light-emitting diode (LED, 25 μW cm−2; HLWW-L51A; Roithner Lasertechnik, Vienna, Austria), positioned about 3 cm from the slices. For the experiments that studied the responses of mixed on bipolar cells to red and green stimuli at various intensities, the light source was a 150-W halogen light bulb. Stimulus flash duration was controlled by an electromechanical shutter, and full-field flashes were presented through the camera port and focused onto the cell by the objective lens. Different intensities and wavelengths were obtained by putting neutral-density and interference filters into the light path. The wavelengths used and their unattenuated intensities were 490 nm (0.64 μW cm−2), 520 nm (5.6 μW cm−2), 640 nm (15 μW cm−2), and 670 nm (6.5 μW cm−2). Light response data were either single responses or averages of 2–4 responses, and their amplitudes were measured from the prestimulation baseline to the peak of the responses during the light flash. At the end of each experiment, the morphology of the cell was visualized with Lucifer yellow staining. When visible, the glycine-positive terminal-free band in the IPL was used as the border between sublaminae a and b (Connaughton and Nelson 2000). In cases where this band could not be seen clearly, the midpoint of the IPL was used as the border (Sherry and Yazulla 1993). All error estimates, including error bars in the figures, are SEs. Origin software (Microcal, Northampton, MA) was used to calculate P values, with the significance level set at 0.05.

Chemicals and solutions

The Ringer solution contained (in mM) 120 NaCl, 2 KCl, 1 MgCl2, 2–3 CaCl2, 4 HEPES, 4 d-glucose, 0.1–0.2 picrotoxin, 0.002–0.005 strychnine, and 0.03 d-AP5, and was set to pH 7.65–7.75 with NaOH. For experiments where synaptic release was blocked, CaCl2 in the Ringer solution was substituted with 3 mM CoCl2. In experiments where potassium conductances were blocked, the control solution had the same ingredients as the above-mentioned Ringer solution (with CoCl2 instead of CaCl2), except that 43 mM NaCl was substituted with equimolar choline-Cl. The solution used to block potassium conductances contained (in mM) 30 tetraethylammonium (TEA)-Cl, 10 BaCl2, 77 NaCl, 1 MgCl2, 3 CoCl2, 4 HEPES, 4 d-glucose, 0.1–0.2 picrotoxin, 0.002–0.005 strychnine, and 0.03 d-AP5, and was set to pH 7.65–7.75 with NaOH. The intracellular solution contained (in mM) 104 K-gluconate or Cs-methanesulfonate, 12 KCl or CsCl, 0.1 CaCl2, 1 EGTA, 2 Na2-ATP, 4 HEPES, 1 Na-GTP, and 0.1–1 Na-cGMP, and its pH was set to 7.4 with KOH or CsOH. Approximately 0.05% Lucifer yellow was added to the intracellular solution to visualize the morphology of the cells. In all voltage-clamp experiments where reversal potentials were determined, the cesium-based internal solution was used. In the current-clamp experiments where only agonist-induced (but not light-evoked) responses were studied, both internal solutions were used, but for consistency between the voltage-clamp and current-clamp experiments, all the raw data shown in the figures were obtained using the cesium-based solution. The potassium-based internal solution was used in most current-clamp experiments where light-evoked responses were recorded, and all the data shown in the figures were obtained using this solution unless stated otherwise.

l-AP4 [l-(+)-2-amino-4-phosphonobutyric acid], d-AP5 [d-(−)-2-amino-5-phosphonopentanoic acid], CNQX (6-cyano-7-nitroquinoxaline-2,3-dione), CPPG [(RS)-α-cyclopropyl-4-phosphonophenylglycine], l-glutamate, NBQX (2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo-[f]quinoxaline-7-sulfonamide), and TBOA (dl-threo-β-benzyloxyaspartate) were purchased from Tocris (Ellisville, MO). TTX (tetrodotoxin) was purchased from CalBiochem (San Diego, CA) and Sigma (St. Louis, MO). Kainate was purchased from Ocean Produce International (Shelburne, Canada). All other chemicals were purchased from Sigma. Picrotoxin powder was dissolved directly into all recording solutions just before each experiment. For other antagonists and agonists, stock solutions were made by dissolving the drugs in distilled water, with the addition of NaOH or HCl to adjust the pH, and were kept frozen at −20°C. Most stock solutions had osmolarities between 200 and 300 mOsm to minimize the change in the osmolarity of the final solution. Stock solutions were added to the Ringer without substitution, and the pH of all final solutions was adjusted to pH 7.65–7.75. Chemicals were either bath-applied by a gravity-fed manifold system, or locally puffed using a multibarrel puffer pipette with a combined tip diameter of about 10–15 μm placed about 10–20 μm from the dendrites of the bipolar cell being recorded. Puffing was controlled by a Pneumatic PicoPump (PV 820; World Precision Instruments, Sarasota, FL). Drugs were puffed at intervals of 20–90 s, depending on the speed of recovery from the previous puff.


Morphological identification of on bipolar cells

The slice preparation allowed us to select visually the 2 cell types for recording. The Mbs, which are the largest bipolar cells in teleosts (about 10 μm soma diameter), could be reliably identified, and postexperiment staining with intracellular Lucifer yellow confirmed their identity. All the other bipolar cells, including the Cbs, are about 5 to 6 μm in soma diameter and thus could not be distinguished from one other until stained with Lucifer yellow at the end of each experiment. We found that these smaller cells could have one or more terminals in sublamina b, one or more terminals in sublamina a, or multiple terminals in both sublaminae. The first of these 3 groups corresponds to the Cbs.

Cone-driven on bipolar cells


The 3 known types of fast-acting glutamate receptors on teleost bipolar cells are mGluR6, EAATs, and AMPA/kainate receptors. To investigate which of these receptors are on the Cbs, 3 selective agonists [l-AP4 (150–300 μM), d-aspartate (1.5–3 mM), and kainate (100 μM)] were puffed onto the dendritic region of each cell with a multibarrel puffer pipette. In all puffing experiments, calcium in the Ringer was replaced by 3 mM cobalt to block calcium-dependent synaptic release. The picrotoxin (100–200 μM) and strychnine (2–5 μM) included in the Ringer should further suppress any remaining GABAergic and glycinergic synaptic transmission. Because d-aspartate is also an agonist for the N-methyl-d-aspartate (NMDA) receptor (Foster and Fagg 1987), the NMDA receptor antagonist d-AP5 (30 μM) was also included. Magnesium (1 mM) was also added to the Ringer to suppress any remaining NMDA receptor response generated by d-aspartate. Although functional NMDA receptors have not been found on on bipolar cells in any fish or other cold-blooded species, anatomically or physiologically (Grant and Dowling 1995; Kawai 1999; Thoreson and Witkovsky 1999; Vandenbranden et al. 2000; Yazulla and Studholme 2001), these precautions were undertaken to ensure that responses to d-aspartate resulted from activation of the EAATs.

Although these agents minimize polysynaptic interactions, the possibility remained that d-aspartate could induce responses from the cell being recorded by inhibiting EAATs on neighboring cells. Because cobalt blocks only calcium-dependent transmitter release, glutamate might still be released from the photoreceptors through any calcium-independent mechanisms present. Glutamate would then be continuously removed from the synaptic cleft by EAATs on surrounding cells. Thus when d-aspartate was puffed onto the synapse, it could compete with glutamate for these EAATs, resulting in a transient increase in the level of glutamate in the cleft, which might then activate glutamate receptors on the cell being analyzed. To test this possibility, d-aspartate was first puffed alone, and then copuffed with the EAAT inhibitor TBOA (1–1.5 mM) onto 7 Cbs and 2 Mbs. In all cases, TBOA blocked most or all of the d-aspartate responses (not shown). If d-aspartate had been acting through inhibition of EAATs on surrounding cells, the coapplication of TBOA would, if anything, enhance the responses to d-aspartate. This control experiment confirmed that d-aspartate acted directly onto the bipolar cells being recorded.

Fifty-five Cbs were analyzed in these puffing experiments. They had a mean input resistance of 1.82 ± 0.22 GΩ in the Ringer’s containing cobalt (n = 16). Of these, 52 responded to d-aspartate but not to l-AP4 or kainate, and 3 responded to d-aspartate and kainate, but not to l-AP4. Two of those 3 cells were tested for light-evoked responses, and they depolarized to a white light stimulus (not shown). Because cells that responded to both kainate and d-aspartate were encountered at such a low frequency, they are probably rare exceptions rather than the norm. In fact, their responses to kainate could even have been the result of artifact. Thus the remainder of this study will focus on those cells that responded only to d-aspartate.

To measure the reversal potential of the d-aspartate response, a voltage ramp was presented first in the absence of, and then in the presence of, d-aspartate (applied from the puffer pipette). The difference between the responses recorded under these 2 conditions was taken as the response induced by d-aspartate. Only cells recorded with the cesium-based internal solution were used for these measurements. d-Aspartate elicited a conductance increase that reversed at −48.4 ± 1.5 mV (n = 25; Fig. 1, right), somewhat more positive than the calculated Nernst potential for chloride (−60 mV) in these experiments. This reversal potential was more negative than the resting potential of these cells, which was −36.5 ± 3.8 mV with intracellular potassium (n = 8), and −29.8 ± 1.3 mV with intracellular cesium (n = 15). (The slightly more positive resting potential measured in the presence of internal cesium probably reflects the blockade of potassium-conducting leak channels.) Therefore in current clamp, d-aspartate induced a hyperpolarization from this resting potential (n = 21; Fig. 1, bottom left), consistent with EAATs role in the generation of sign-inverting responses to the endogenous transmitter glutamate. Besides these 52 cells that had intact axon terminals, 28 additional cells were encountered that had similar sized somata but had lost their axon terminals during the slicing process, and that responded only to d-aspartate but not to the other 2 agonists (not shown). This indicates that the d-aspartate response was not dependent on EAATs on the axonal terminals. Results from these agonist experiments are summarized in Table 1.

View this table:

Summary of the responses of Cbs and Mbs to glutamate receptor agonists

The above experiments showed that a vast majority of Cbs have EAATs but not mGluR6 or AMPA/kainate receptors. However, it is important to rule out the presence of any other novel glutamate receptor types besides these 3 known types. To confirm that EAATs are indeed the only glutamate receptor type on Cbs, l-glutamate (1 mM) was puffed onto the cells to activate all glutamate receptors (except, of course, NMDA receptors because d-AP5 and Mg2+ were included in the bath). In voltage clamp, l-glutamate induced a conductance increase that reversed at −41.5 ± 1.6 mV (n = 3). For all 3 cells, when the EAAT antagonist TBOA (500 μM) was copuffed with l-glutamate, the responses to l-glutamate were blocked (Fig. 2, top). This finding provides additional evidence that EAATs are the only glutamate receptors on Cbs.

FIG. 2.

Excitatory amino acid transporters (EAATs) constitute the only glutamate receptor that mediates the light responses of Cbs. Top: 500 μM dl-threo-β-benzyloxyaspartate (TBOA) blocked all of the l-glutamate (1 mM) responses of this Cb. Bottom: 100 μM TBOA blocked all light responses of another Cb. Stimulus was a 500-ms full-field white flash, indicated as a step under each response trace. Cesium-based internal solution was used for this cell.


To study light-evoked responses, cobalt was replaced with calcium to permit synaptic transmission from photoreceptors to second-order neurons. Responses to full-field white flashes were recorded under current clamp. In darkness, the resting potentials of Cbs were −53.5 ± 1.5 mV with intracellular potassium (n = 21) and −36.0 ± 2.0 mV with intracellular cesium (n = 2). All Cbs tested gave a depolarization at light on, and they returned to the resting potential after a brief overshoot at light off (n = 28; Fig. 2, bottom left). This light-evoked response was blocked by bath-applied TBOA (100 μM) (n = 8; Fig. 2, bottom right), and the resting potentials were raised to −40.8 ± 4.2 mV and −28.0 ± 1.0 mV with intracellular potassium (n = 5) and cesium (n = 2), respectively. These values are comparable to the resting potentials measured when calcium-dependent release of glutamate from photoreceptors was blocked by cobalt (see above), consistent with the blockade of most if not all glutamatergic inputs by TBOA. Taken together, these results confirm that EAATs are the only type of glutamate receptor on Cb’s, and that they mediate these cells’ light-evoked responses.

Mixed on bipolar cells


d-Aspartate, l-AP4, and kainate were puffed onto Mbs in the presence of cobalt, picrotoxin, strychnine, and d-AP5. In these blockers, the Mb cells had a mean input resistance of 330 ± 31 MΩ (n = 16). All Mbs tested (n = 74) responded to both d-aspartate and l-AP4, but not to kainate (Table 1). Reversal potentials were measured by applying a voltage ramp to the cells as described above for Cbs. d-Aspartate activated a conductance increase that reversed at −50.2 ± 1.4 mV, consistent with the opening of a chloride-conducting channel, whereas l-AP4 generated a conductance decrease that reversed at 21.7 ± 4.8 mV (n = 27; Fig. 3, right). This value for the Erev of the l-AP4 response is considerably higher than that measured by most other investigators, the values of which are much closer to zero and are consistent with the involvement of a nonselective cation channel in the mGluR6 pathway (e.g., Connaughton and Nelson 2000; Euler et al. 1996; McGillem and Dacheux 2001). One possible explanation is that Mbs in the giant danio have 2 mGluR6-activated mechanisms, one that leads to the closure of nonselective cation channels and another that opens channels to ions with relatively negative reversal potentials. The latter mechanism has been reported in salamander on bipolar cells (Hirano and MacLeish 1991). Because the Mbs in giant danio responded to l-AP4 with a net conductance decrease, the closure of nonselective cation channels probably dominates the mGluR6-mediated response.

FIG. 3.

Mixed-input on bipolars (Mbs) and their responses to glutamate receptor agonists. Top left: morphology of an Mb revealed by Lucifer yellow staining. Right: responses of an Mb to 3 glutamate receptor agonists recorded under voltage clamp. Bottom left: responses of another Mb to the same agonists recorded under current clamp. Duration of the puffs was 1 s.

Mbs had resting potentials of −32.1 ± 4.1 mV (n = 6) or −28.4 ± 1.2 mV (n = 51) when potassium- or cesium-based internal solutions were used, respectively. Because the reversal potential of the d-aspartate response was more negative than the resting potential, and the response to l-AP4 involved the closure of a channel with a reversal potential more positive than the resting potential (see above), both agonists hyperpolarized Mbs from rest in current-clamp mode (n = 94 for l-AP4 and 92 for d-aspartate; Fig. 3, bottom left). Palmer et al. (2003) reported the presence of EAATs on axon terminals of Mbs in goldfish, raising the possibility that the d-aspartate–induced response was generated by these EAATs. Two lines of evidence argue against this possibility. First, the response to d-aspartate was larger when it was puffed onto the outer plexiform layer than when it was puffed onto the IPL (n = 3; not shown), suggesting the EAATs activated by d-aspartate were located mainly in the dendritic region. Furthermore, 7 large, Mb-like cells whose axon terminals had been cut off were encountered, and all of them responded to both d-aspartate and l-AP4 puffs, indicating the axonal EAATs were not necessary for the d-aspartate response.

The presence of these 2 receptors on Mbs was confirmed by the effect of TBOA on l-glutamate puffs. Under voltage clamp, a conductance increase was induced in all 6 cells tested when l-glutamate (1 mM) was puffed alone. However, when copuffed with TBOA (500 μM), the response of these 6 cells to l-glutamate became a conductance decrease that reversed at −1.1 ± 3.2 mV (Fig. 4, left). On the other hand, when l-AP4 was constantly present in the bath to saturate mGluR6, the response to 1.5 mM l-glutamate was totally blocked by 500 μM TBOA (n = 5; Fig. 4, right). This experiment confirmed that EAATs and mGluR6 account for all glutamate receptor types on Mbs. As in the case for Cbs, d-AP5 and Mg2+ were included in the Ringer’s and thus any NMDA receptors present would not have been activated by the l-glutamate puffs. Nevertheless, because functional NMDA receptors have not been reported on any on bipolars in all fish species studied, it is extremely unlikely that these receptors were on any of the Cbs and Mbs tested.

FIG. 4.

EAATs and mGluR6 receptors are both present on Mbs. Left: when puffed alone, 1 mM l-glutamate activated both EAATs and mGluR6 receptors on this Mb. When copuffed with 500 μM TBOA, the EAAT response was blocked, leaving only the mGluR6 response to l-glutamate. Right: another Mb was recorded in a Ringer’s containing 200 μM l-AP4, which constantly activated mGluR6. Therefore l-glutamate (1.5 mM) could activate only EAATs, and TBOA blocked all of the l-glutamate responses.


To analyze light-evoked responses in Mbs, a calcium-based Ringer was used. The resting potentials of Mbs were −52.85 ± 0.90 mV with internal potassium (n = 56) and −42.8 ± 1.1 mV with internal cesium (n = 32). Full-field white stimuli elicited a depolarization in current clamp for all 68 cells tested (Fig. 5 top left). In contrast to its effect on light responses of Cbs, 100 μM TBOA did not completely block the light responses of Mbs. For all 18 Mbs tested, a substantial portion of their light responses was resistant to TBOA. Some cells’ responses were reduced, whereas responses of others were actually enhanced. The average reduction was 3 ± 11% (n = 18). Additionally, the more or less square waveform in the control Ringer now rose more gradually at light on, and returned to the resting potential more slowly at light off (Fig. 5, top left). The slower rising phase during light stimulation may be attributable to the slower removal of glutamate from the synaptic cleft by EAATs on neighboring cells, which would be operating less efficiently in TBOA, whereas the slower falling phase at light off may in part reflect the slower kinetics of mGluR6. Adding CPPG (1 mM), a group III mGluR antagonist, on top of TBOA completely blocked the remaining light responses (Fig. 5, top left) (n = 5). In the reverse experiment, CPPG applied alone reduced the amplitude of the depolarization by 36.6 ± 7.0% (n = 11), without significantly changing the waveform of the response. Applying TBOA in addition to CPPG blocked all remaining responses (Fig. 5, bottom left) (n = 7). It was concluded that both EAATs and mGluR6 directly mediate all light-evoked responses of Mbs, and that the concentrations of TBOA and CPPG used (100 μM and 1 mM, respectively) were adequate to completely inhibit these receptors. The percentage reductions by TBOA alone (3%) and by CPPG alone (37%) do not add up to 100%, and the reason for that will be presented in the discussion.

FIG. 5.

Both EAATs and mGluR6 contribute to the light responses of Mbs. Left: examples of raw data from 2 cells. All stimuli were 500-ms full-field white flashes. Top right: summary of results from 18 cells tested for the effect of 100 μM TBOA. Bottom right: summary of results from 11 cells tested for the effect of 1 mM (RS)-α-cyclopropyl-4-phosphonophenylglycine (CPPG). In both summary diagrams, each cell’s response amplitude in the control was normalized to 1.

When all light responses were blocked by CPPG plus TBOA, the resting potential for cells recorded with internal potassium was elevated to −46.3 ± 3.3 mV (n = 8), whereas that for cells with internal cesium was raised to −36.3 ± 5.7 mV (n = 3). These values are more negative than the resting potentials measured in the cobalt-based Ringer (P values of 0.0183 and 0.1383, respectively). One plausible explanation is that when calcium is present in the Ringer, it suppresses the cation channel associated with the mGluR6 pathway, reducing the influx of cations and thus leads to a more hyperpolarized state (Nawy 2000).


Grant and Dowling (1996), by combining their own findings with earlier reports by Saito et al. (1979) and Nawy and Copenhagen (1987), postulated that rod inputs onto on bipolar cells are mediated by mGluR6, and that cone inputs are mediated mainly by EAATs, although mGluR6 could possibly be involved as well. The light response experiment described above demonstrated that Cbs indeed receive cone signals by EAATs. Here we will discuss the rod-driven mechanism, which was studied by recording from Mbs. Because Mbs receive inputs from cones as well, the responses of these cells to cone signals were also studied, to test whether they use EAATs or a combination of EAATs and mGluR6.

To achieve this aim, we compared the effects of CPPG or TBOA on the responses of Mbs to red (670 or 640 nm) and green (520 or 490 nm) stimuli. The red stimulus preferentially stimulates the cones, whereas the green stimulus is detected mainly by the rods (Palacios et al. 1996; Robinson et al. 1993; Schwanzara 1967). A series of increasing intensities of red and green flashes was presented to each cell, first in the control Ringer’s, and then in the presence of CPPG (1 mM) or TBOA (100 μM). Although some cells gave unambiguous responses to the dimmest flashes tested (−4 and −3 log I), the small amplitudes of most responses elicited by these dim stimuli made it difficult to distinguish them from the high level of noise in most records. Therefore only data obtained with the 3 brightest intensities (−2 to −0 log I) will be considered here. At these high intensities, both rods and cones probably respond to both wavelengths, although the rods still respond more strongly to green than to red, whereas the cones prefer red. Responses to the green flashes were reduced by blocking mGluR6 with CPPG (by 14 ± 18% at −2 log, 29 ± 13% at −1 log, and 51.6 ± 7.8% at −0 log, or 30.8 ± 8.3% on average; n = 10), but were enhanced significantly by blocking EAATs with TBOA (by 89 ± 22% on average; n = 6) (Fig. 6, top). In contrast, responses elicited by the red flashes were reduced by both antagonists, although the reduction by TBOA (47.9 ± 9.1% on average; n = 5) was somewhat larger than that by CPPG (27.5 ± 5.0% on average; n = 10) (P = 0.038; Fig. 6, bottom).

FIG. 6.

Responses of Mbs to green (rod-selective) and red (cone-selective) stimuli were differentially affected by CPPG and TBOA. Top: responses to the green stimulus were reduced by 1 mM CPPG (left; n = 10), but enhanced by 100 μM TBOA (right; n = 6). Bottom: responses to the red stimulus were reduced by both CPPG (left; n = 10) and TBOA (right; n = 5). In all 4 plots, the biggest response from each cell obtained in the control was normalized to 1.

These results strongly suggest that the rod-driven light response of Mbs is mediated mainly if not exclusively by mGluR6, because TBOA not only did not reduce the rod-driven responses (which would have occurred if EAATs mediated most of the rod inputs) but enhanced them. A possible explanation for such amplification will be presented in the discussion. On the other hand, inputs from cones appear to be mediated by both EAATs and mGluR6, although the former appear more important. However, because part of the responses to the brighter red flashes was probably driven by the rods, the reduction by CPPG could simply be a result of its action on the rod inputs. To test whether mGluR6 has a contribution to the cone-driven responses, the following experiment was performed on a total of 7 Mbs. These cells were first perfused with 100 μM TBOA for 5–10 min in the dark-adapted state to block the EAATs as completely as possible. The condenser light source of the microscope (110–560 μW cm−2) was then turned on to saturate the rod photoreceptors. Responses of each cell to white flashes were then monitored for 5–10 min. All 7 cells continued to give robust responses during that time period (Fig. 7). Control experiments showed that the red appearance of isolated retinas, which reflects the presence of unbleached rhodopsin, completely disappeared within about 5 min when these retinas were exposed to a 54.5 μW cm−2 white background (not shown), suggesting that rhodopsin in the rod photoreceptors was mostly bleached shortly after the condenser light was turned on. Even if rhodopsin was not completely bleached, the rods were most likely saturated because 0.07 μW cm−2 had been found to be sufficient to eliminate the contribution of rod signals to ganglion cells in goldfish (Mackintosh et al. 1987). Therefore the light-evoked responses from all 7 cells were most likely exclusively cone-driven. Because 100 μM TBOA had been shown to be sufficient to completely block EAATs on Mbs (see Fig. 5, bottom left), the responses obtained in this photopic experiment were presumably mediated by mGluR6. Indeed, when CPPG was added to 3 of these 7 cells, their light responses were completely blocked (the other 4 cells were not tested with CPPG; Fig. 7). This experiment confirmed that Mbs receive cone signals not only by EAATs but also by mGluR6.

FIG. 7.

In the presence of 100 μM TBOA, Mbs could respond to 500-ms white flashes (about 104 μW cm−2, presented through the camera port) under photopic conditions. Subsequent application of 1 mM CPPG rapidly and completely blocked the light response. In this case, the adapting background light had an intensity of 110 μW cm−2. Times shown at the top indicate how long the background light had been on.


Whereas the above experiments demonstrated that mGluR6 and EAATs are probably the only glutamate receptors on Mbs, contradictory results from 2 previous studies deserved reexamination. By recording from on bipolars in goldfish, Nawy and Copenhagen (1990) showed that when potassium-filled intracellular electrodes were used, glutamate had no effect on the membrane conductance of these cells even though it hyperpolarized them. However, when cesium-filled electrodes were used, l-glutamate induced a conductance decrease with the same reversal potential as the responses to l-AP4. These authors thus proposed that the non-mGluR6 component in teleost on bipolar cells is associated with the opening of a cesium-sensitive potassium channel. By contrast, in a more recent study, Grant and Dowling (1995) reported that the non-mGluR6, EAAT-like l-glutamate response could be induced whether the internal solution contained potassium or cesium.

Therefore the study by Nawy and Copenhagen might implicate a third type of glutamate receptor on teleost on bipolar cells, in addition to mGluR6 and EAATs. We tested this possibility by performing the following experiment on 8 Mbs. l-AP4 (200 μM), CNQX (60 μM), and d-AP5 (30 μM) were added to the cobalt-based Ringer and puffed solutions to block mGluR6 (by saturating it), AMPA/kainate receptors, and NMDA receptors. With these receptors blocked, the only remaining functional glutamate receptors would be the EAATs, plus the potassium-conducting glutamate receptor in question. A potassium-based internal solution was used. l-Glutamate (1.5 mM) was first puffed in the normal Ringer solution, which contained 2 mM KCl as well as cobalt, picrotoxin, and strychnine, and 43 mM NaCl replaced with equimolar choline chloride. The perfusion was then switched to a solution with potassium omitted. Furthermore, the choline chloride in the Ringer was replaced with 30 mM TEA-Cl and 10 mM BaCl2, and l-glutamate was puffed again. TEA and Ba2+ would presumably block most potassium channels (Hille 2001). In a control experiment performed in the cobalt-based Ringer under current clamp, similar doses of TEA and barium were shown to bring the resting potentials of Mbs to much more positive levels (by >10 mV; n = 4), indicating that these concentrations were effective in blocking potassium channels on Mbs (not shown). For all 8 cells tested, blocking potassium channels in this manner had only slight effects on the l-glutamate–induced response, and the averaged response amplitude measured at −40 mV (a physiologically relevant membrane potential) was virtually identical to that obtained under the control condition (Fig. 8). A possible explanation for the suppression of the non-mGluR6 l-glutamate response by internal cesium observed by Nawy and Copenhagen will be considered in the discussion.

FIG. 8.

Response of Mbs to l-glutamate does not involve a potassium channel. In this experiment, a potassium-based internal solution was used. Left: this cell gave similar voltage-ramp responses to 1.5 mM l-glutamate in the control and in a solution with 30 mM tetraethylammonium (TEA) and 10 mM barium. Right: summary of the results from 8 cells. Each cell’s response amplitude in the control was normalized to 1.

Interactions between EAATs and mGluR6 in Mbs


Because EAATs and mGluR6 are both present on Mbs, an important question is how they interact. To address this, l-AP4 (150–300 μM) and d-aspartate (1.5–3 mM) were used to selectively activate mGluR6 and EAATs, respectively. As in the puffing experiments described above, the Ringer contained cobalt, picrotoxin, strychnine, and d-AP5 to ensure direct actions of l-AP4 and d-aspartate onto the Mbs. To study the effect of activating mGluR6 on EAAT responses, d-aspartate was puffed onto Mbs first in the control Ringer, and then with l-AP4 in the bath. When this was performed in current-clamp mode, bath-applied l-AP4 hyperpolarized the cells and reduced their responses to d-aspartate by 43.7 ± 9.7% (n = 7; Fig. 9). A simple explanation for this reduction is that the l-AP4–induced hyperpolarization brought the resting potential closer to the Erev for EAATs, reducing the driving force on chloride to enter the cell. To test this hypothesis, the experiment was repeated in voltage-clamp mode for all 7 cells. Under voltage clamp, l-AP4 was prevented from changing the membrane potential, and bath-applied l-AP4 had negligible effects on the d-aspartate responses. The slope of the voltage-ramp response, which corresponds to the conductance increase induced by d-aspartate, was reduced by only 2.3 ± 4.1% in the presence of l-AP4 (Fig. 10). Therefore a change in the membrane potential by l-AP4 was probably responsible for the reduction of the d-aspartate responses in current clamp.

FIG. 9.

In current-clamp mode, the responses of Mbs to 1.5–3 mM d-aspartate were reduced by 150–300 μM l-AP4. Top: example of raw data obtained from one cell. Duration of the puffs was 1.5 s. Bottom: summary of the results from 7 cells. Each cell’s response amplitude in the control was normalized to 1.

FIG. 10.

In voltage-clamp mode, l-AP4 had little effect on the responses of Mbs to d-aspartate. Left: example of raw data from one cell. Right: summary of the results from the same 7 cells whose responses were measured under current clamp in the experiment described in Fig. 9. Each cell’s d-aspartate–induced conductance change in the control was normalized to 1.


The converse experiment was performed to test whether EAATs suppress mGluR6 in current-clamp mode. Bath-applied d-aspartate hyperpolarized Mbs and reduced their responses to l-AP4 by 71.6 ± 3.3% (n = 14), suggesting mutual suppression occurs between the 2 types of glutamate receptors (Fig. 11). However, this reduction cannot be explained by the change in resting potential caused by d-aspartate because the reversal potential for the l-AP4 response was ≥0, and hyperpolarizations would actually enhance the response. A more likely explanation is that the cells became leakier when d-aspartate opened the EAAT channels. To test whether this explanation fully accounts for the reduction of l-AP4 responses, the experiment was repeated in voltage-clamp mode for 10 of the 14 cells. If the d-aspartate–induced membrane conductance change was the only cause of the reduced l-AP4 response under current clamp, and l-AP4 could still close the same number of cation channels when EAATs were activated, then d-aspartate should have little effect on the l-AP4 response under voltage clamp because the same number of channels closed by l-AP4 would be reflected by the same conductance change. However, this was not observed. Instead, the slope of the l-AP4 response to the voltage ramp was reduced by 64.7 ± 5.8% (Fig. 12), meaning that when EAATs were activated by d-aspartate, a smaller number of channels were closed by l-AP4. It is possible that this effect of d-aspartate acts in combination with the leakiness resulting from the opening of EAATs to lead to the final reduction in the l-AP4 response amplitude measured in current-clamp mode.

FIG. 11.

In current-clamp mode, the responses of Mbs to l-AP4 were reduced by d-aspartate. Top: example of raw data obtained from one cell. Duration of the puffs was 3 s. Bottom: summary of the results from 14 cells. Each cell’s response amplitude in the control was normalized to 1.

FIG. 12.

In voltage-clamp mode, d-aspartate reduced the responses of Mbs to l-AP4. Left: bath-applied d-aspartate reduced the slope of the voltage-ramp response to l-AP4. Right: summary of the results from 10 cells. Each cell’s l-AP4–induced conductance change in the control was normalized to 1.


Rods signal onto mGluR6, whereas cones signal onto both EAATs and mGluR6

The ERG study reported in the companion paper provided evidence that in the giant danio retina, the rod-driven b-wave is mediated primarily by mGluR6, whereas the cone-driven b-wave is mediated by a combination of EAATs and mGluR6. The patch-clamp analysis herein confirmed those findings, and yielded additional information about the glutamatergic input mechanisms on the 2 subtypes of on bipolar cells. The Cbs, which receive input exclusively from the cone photoreceptors (e.g., Scholes 1975; Stell 1967; Van Haesendonck and Missotten 1984), were found to possess only EAATs and not mGluR6 or AMPA/kainate receptors (Fig. 1). EAATs were also shown to be solely responsible for the generation of light-evoked responses in these cells (Fig. 2), suggesting that the mGluR6 component of the photopic b-wave reported in the companion paper does not reside in Cbs.

Because Mbs are the only bipolar cell type possessing mGluR6 (based on their responses to l-AP4; see Table 1 and Fig. 3), and because they receive input from cones as well as rods (Saito et al. 1983, 1985; Stell 1967), they are an obvious candidate to be the source of the mGluR6 component of the cone-driven b-wave. Indeed, when EAATs were blocked with TBOA, Mbs were still able to respond to light flashes under photopic conditions, arguing that Mbs can respond to cones by mGluR6 (Fig. 7). However, mGluR6 is only one of the cone-responding mechanisms on Mbs. Besides mGluR6, Mbs also possess EAATs, which appear to mediate a majority of the cone input onto Mbs, based on the differential effects of TBOA and CPPG on the responses of Mbs to red stimuli (Fig. 6, bottom). Although the comparatively high variability of the data from that experiment has made such a conclusion somewhat equivocal (notice the size of the error bars in Fig. 6, bottom), the cone-induced conductance changes measured by Saito et al. (1979) lend further support for this conclusion. They showed that in the carp retina, light signals from cones caused a decrease in the conductance of the on bipolar cells, implying that glutamate released from these photoreceptors results in a conductance increase. Because activation of EAATs by glutamate leads to a conductance increase, and activating mGluR6 receptors results in a conductance decrease (Fig. 3, right), a majority of the response must be generated by EAATs to account for the fact that the net change in conductance is an increase.

Besides participating in the cone-evoked light responses, mGluR6 receptors on Mbs also mediate rod-driven responses. In fact, both the patch-clamp results in this paper and the ERG data in the companion paper suggest that rod inputs are mediated predominantly by mGluR6 (Fig. 6, top). These results confirm previous findings for goldfish on bipolar cells (Nawy and Copenhagen 1987).

EAATs and mGluR6 are the only photoreceptor-responding glutamate receptors on teleost on bipolar cells

Both the ERG study in the companion paper and the present patch-clamp experiments provide strong evidence that mGluR6 and EAATs are the only types of glutamate receptors that transmit signals from photoreceptors to on bipolar cells in giant danio. As noted in the introduction, Nawy and Copenhagen (1990) reported evidence for a glutamate-responsive potassium channel on goldfish on bipolar cells. Grant and Dowling (1995) refuted that conclusion, showing that even when internal cesium was used, and mGluR6 receptors blocked, l-glutamate responses could still be obtained from on bipolar cells in white perch and bass. They argued that the non-mGluR6 receptor was a chloride-permeant channel, most likely an EAAT. However, with the limited variety of EAAT antagonists available at that time, Grant and Dowling were able to suppress the non-mGluR6 glutamate responses by <50%, and the existence of a third glutamate receptor on the on bipolars remained a possibility.

Using the potent, selective, and nontransportable EAAT blocker TBOA, we were able to completely block all l-glutamate and light-evoked responses in Cbs, and all non-mGluR6 responses in Mbs, providing strong evidence for the EAAT hypothesis proposed by Grant and Dowling. In addition, by applying TEA and barium extracellularly, we provide evidence against the presence of a glutamate-responding, potassium-permeable channel on on bipolars (Fig. 8). A possible explanation for the results reported by Nawy and Copenhagen (1990) is that potassium is one of the ions involved in the transport of glutamate through EAATs. The transport of each glutamate molecule into the cell is coupled to the cotransport of 2 or 3 sodium ions and one proton, and the countertransport of a potassium ion. The chloride current that enables glutamate to hyperpolarize on bipolar cells is uncoupled from the transport process (Seal and Amara 1999). When intracellular potassium is replaced with cesium, the transport process is slowed, and the net current generated is reduced by one third or more (Barbour et al. 1991; Bergles et al. 2002). Nawy and Copenhagen found that l-glutamate could not induce a conductance change in the presence of internal potassium, indicating that the conductance increase triggered by the activation of EAATs canceled out the conductance decrease associated with the activation of mGluR6. However, when internal cesium was used, the efficacy of the EAATs was reduced, and the mGluR6 response was relatively enhanced, resulting in a net decrease in cell conductance.

In a recent anatomical study, Henry et al. (2003) reported the presence of cyclic nucleotide-gated (CNG) channels on some Cbs in the goldfish retina. Because CNG channels were, for a number of years, a candidate for the cation channel modulated by the mGluR6 pathway in on bipolar cells (Nawy and Jahr 1991; Shiells and Falk 1990), the authors suggested that their finding could be evidence for a role of mGluR6 in the generation of light-evoked responses in Cbs. However, none of the giant danio Cbs we examined responded to l-AP4, arguing that the CNG channels on these cells probably serve other functions.

Although the present study has shown that mGluR6 and EAATs are the only glutamate receptors on the on bipolars, it does not address how many types of EAATs are present because d-aspartate and TBOA act on all 5 types of EAATs (Bridges et al. 1999; Shigeri et al. 2001; Shimamoto et al. 2000). Types 1, 2, 3, and 5 have been found in the retina, but type 5 is the only one with a conductance carried predominantly by chloride; thus it is the prime candidate for the EAATs on teleost on bipolar cells (Thoreson and Witkovsky 1999). The currents through the other 3 types are carried mostly by cations, leading to more positive reversal potentials (Eliasof et al. 1998a,b; Wadiche et al. 1995). In giant danio, d-aspartate responses were found to reverse at around −50 mV for Mbs and −48 mV for Cbs, which are significantly more positive than the calculated ECl of −60 mV, raising the possibility that besides EAAT5, one or more of the other 3 EAAT types exist on these bipolar cells. This is consistent with anatomical studies of several other species, which showed the presence of EAAT2 on some on bipolar cells (e.g., rat: Euler and Wässle 1995; rat and monkey: Rauen and Kanner 1994; salamander: Eliasof et al. 1998a,b). However, this does not necessarily mean that these other EAATs directly contribute to light responses.

Mutual inhibition between EAATs and mGluR6 in Mbs may subserve rod–cone suppression

In scotopic and mesopic states, both rods and cones release glutamate onto Mbs, resulting in the simultaneous activation of mGluR6 and EAATs. Even under photopic conditions, when rods are saturated and thus release little if any transmitter, glutamate from cones can still activate both mGluR6 and EAATs. Thus we studied the interactions between these receptors by selectively activating them with l-AP4 and d-aspartate, respectively. Under current clamp, d-aspartate triggered a larger voltage change when applied alone than when applied in the presence of l-AP4, suggesting that mGluR6 inhibits the EAATs (Fig. 9). By repeating the experiment in voltage-clamp mode (Fig. 10), we showed that l-AP4 induces such inhibition probably by bringing the membrane potential closer to the Erev for EAATs. Because rods signal mainly onto mGluR6, and cones signal mainly onto EAATs, this kind of interaction may have relevance to rod–cone suppression, a phenomenon that has been observed in many species (e.g., Dong et al. 1988; Goldberg et al. 1983; Hess et al. 1992; Krizaj et al. 1994; Levine et al. 1987; MacLeod 1972; Margrain and Thomson 1997; Sandberg et al. 1981; Whitten and Brown 1973). Cone-driven light responses are stronger in photopic states when the rods are inactive and, conversely, rod-driven light responses are stronger under scotopic conditions when the cones are not responding to light. Such a phenomenon allows the visual system to select rod signals over cone signals under scotopic states, and cone signals over rod signals under brighter lighting conditions. The suppression of EAATs by mGluR6 may give rise to a similar type of suppression in teleost Mbs. When dark-adapted, rods release glutamate onto mGluR6 receptors, reducing the amplitude of the EAAT-mediated cone responses. On light adaptation, mGluR6 receptors are less activated, and glutamate from cones triggers a bigger voltage change through the EAATs.

The converse experiment showed that EAATs could also inhibit mGluR6 because l-AP4 elicited larger responses when puffed alone than in the presence of d-aspartate (Fig. 11). Multiple mechanisms may contribute to this inhibition, including an increase in membrane conductance and a reduction in the number of cation channels closed by l-AP4 (Fig. 12). Such observations predict that glutamate released from cones in scotopic conditions reduces the efficacy of rod signals. Supporting this notion, the responses of giant danio Mbs to green stimuli were substantially enhanced when EAATs were blocked by TBOA, implying that before TBOA was applied, the rod input had been under suppression by EAATs (Fig. 6, top). This enhancement is unlikely attributable to TBOAs inhibiting EAATs on neighboring cells, which would have caused glutamate to be removed from the synaptic cleft during the light stimulus less efficiently and led to smaller light responses. However, it seems counterproductive to suppress the rod inputs under scotopic conditions where it is desirable to amplify small signals induced by dim stimuli. Therefore the significance of this type of suppression is unclear.

The mutual inhibition between EAATs and mGluR6 can explain why the reductions of Mb responses to white light by TBOA alone and by CPPG alone added up to much less than 100% (Fig. 5, right). When the EAATs were blocked by TBOA, the light responses generated by mGluR6 became larger than those generated by mGluR6 in the control because inhibition exerted by the EAATs was removed. Conversely, when mGluR6 was blocked by CPPG, the EAAT-mediated components of the light responses were enhanced as a result of the removal of inhibition from mGluR6.

Why are two glutamate receptors needed?

An obvious question is why are there 2 different glutamate receptors on the on bipolar cells? One possible reason, as the previous section suggested, is to allow the rod and cone channels to partially suppress each other in the mixed bipolar cells. However, there are other plausible reasons. For example, EAATs directly induce voltage changes, whereas mGluR6 opens/closes cation channels by a pathway that involves multiple steps. Consequently, the former has faster kinetics than the latter. Shiells and Falk (1994) reported that after glutamate was applied, there is a delay of ≥10 ms before a response is initiated in isolated dogfish on bipolar cells that use mGluR6. In contrast, Auger and Attwell (2000) demonstrated that the chloride current through postsynaptic EAAT4 on Purkinje cells in the cerebellum reached peak amplitude within about 7 ms after the transporter was activated by glutamate. Thus the use of EAATs may enable the cone channel to have faster kinetics.

Furthermore, these glutamate receptors have different dose–response functions. Eliasof et al. (1998b) showed that the dose–response curve for EAAT5 expressed on oocytes is spread over 4 log units. In contrast, Shiells and Falk (1994) reported that the dose–response curve for mGluR6 on dogfish on bipolar cells saturated about 2 log units above the threshold concentration. Therefore another advantage of using EAATs is to allow the cone channel to distinguish among a wider range of stimulus intensities.

In addition, the second-messenger cascade activated by mGluR6 may greatly amplify the weak, single-photon signals in rods, a process critical for night vision. This is analogous to the amplification of weak light responses by the phototransduction pathway in photoreceptors. Ashmore and Falk (1980) reported that rod signals could be amplified by more than 2 orders of magnitude in the on bipolars, and part of this amplification probably arises from the mGluR6 cascade.

Finally, the G protein and second messengers activated by mGluR6 can conceivably trigger modulatory pathways that are responsible for slower but long-lasting physiological changes. The mGluRs are involved in various kinds of modulatory interactions in the retina (reviewed in Thoreson and Witkovsky 1999), and thus mGluR6 may have other functions besides the rapid detection of light signals from photoreceptors.

on bipolar cell mechanisms in nonteleost species

At present, it is commonly believed that teleosts are the only species where EAATs are directly involved in the generation of the on bipolar cell response, and that in all other species mGluR6 is the sole mechanism mediating this response (Thoreson and Witkovsky 1999). In salamander, where the input mechanisms of on bipolar cells have been studied the most thoroughly, mGluR6 is believed to be the only glutamate receptor on these cells because l-glutamate and l-AP4 induce the same conductance changes (Nawy and Jahr 1991), and l-AP4 totally blocks the light-induced depolarization of on bipolar cells (Hensley et al. 1993; Nawy and Jahr 1991; Wu and Maple 1998). l-AP4 likewise completely suppresses the light-evoked depolarization of on bipolars in mudpuppy (Slaughter and Miller 1981). However, the reversal potential of the on bipolar cell light response is somewhat more negative when the internal solution contains low chloride than when a high-chloride internal solution is used (Thoreson and Miller 1993). Similarly, some mixed-input on bipolar cells in the salamander respond to l-glutamate puffed onto the dendrites with reversal potentials about 20 mV more negative than the Erev of the l-glutamate responses of cone-dominant on bipolar cells (Wu and Maple 1998). These 2 observations have raised the possibility that EAATs make a contribution to the responses of on bipolar cells in amphibians.

Mammalian on bipolar cells have been studied less extensively. De la Villa et al. (1995) showed that on bipolar cells isolated from the cat responded to l-glutamate with a conductance decrease that reversed at around 0 mV, but a high-chloride internal solution was used in that study, making it difficult to separate the mGluR6-mediated response from any EAAT-mediated response present. In another study where a low-chloride internal solution was used, the light responses of rod bipolar cells in rat were found to reverse at close to 0 mV, indicating the presence of only mGluR6, but on cone bipolars were not analyzed in detail (Euler and Masland 2000). On the other hand, both rod-driven and on cone bipolar cells in the mouse were studied by Berntson and Taylor (2000), using low-chloride internal solutions, and both cell types responded to light with reversal potentials at about 0 mV, suggesting EAATs are not involved. However, in the rabbit, 30 μM l-AP4 failed to completely suppress the depolarizations of on bipolars to light, and it remains to be tested whether EAATs are responsible for the residual responses (Cohen and Miller 1999).

Besides single-cell studies, on bipolar cell mechanisms have also been investigated using the ERG. l-AP4 has been shown to completely eliminate the b-wave in nearly all species tested, such as cat (Gargini et al. 1999), monkey (Kondo and Sieving 2001; Sieving et al. 1994), mudpuppy (Stockton and Slaughter 1989), rabbit (Lei and Perlman 1999), and rat (Green and Kapousta-Bruneau 1999). The caveat in these studies is that inhibitory synapses were not blocked before the addition of l-AP4, and much of the removal of the b-wave could have been the result of polysynaptic inhibitory effects of l-AP4 (see the companion paper). However, in the skate, all of the b-wave could be eliminated when l-AP4 was applied in the presence of the GABA receptor antagonist picrotoxin, showing that mGluR6 may indeed be the on bipolar cell mechanism in this species (Chappell and Rosenstein 1996); however, skate retinas contain only rods (Dowling and Ripps 1970). In frog, on the other hand, a sizeable b-wave seems to be resistant to 100–200 μM l-AP4 even when the agonist is applied alone (Arnarsson and Eysteinsson 2000; Popova et al. 1995; Szikra and Witkovsky 2001). Furthermore, some b-wave is present in the ERG of mice deficient for mGluR6 (Masu et al. 1995) or for Gαο, the G protein believed to be necessary for the mGluR6-mediated intracellular cascade (Dhingra et al. 2000, 2002).

In summary, it appears likely that EAATs may well be involved in the generation of on bipolar cell responses in species other than fish. Of interest will be to test this possibility in the retinas of various species and also to determine whether EAATs play such a fundamental role in mediating synaptic transmission elsewhere in the CNS.


This work was supported in part by National Eye Institute Grants EY-00824 and EY-00811.


We thank Drs. Alan Adolph, Clint Makino, Markus Meister, Venki Murthy, and Sam Wu for helpful comments on the paper.

Present address of E. D. Cohen: Center for Devices and Radiological Health, FDA, HFZ130, 12725 Twinbrook Parkway, Rockville, MD 20852.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract