Disinhibitory recruitment of NMDA receptor pathways in retina

Abstract

Glutamate release at bipolar to ganglion cell synapses activates NMDA and AMPA/kainic acid (KA) ionotropic glutamate receptors. Their relative strength determines the output signals of the retina. We found that this balance is tightly regulated by presynaptic inhibition that preferentially suppresses NMDA receptor (NMDAR) activation. In transient ON-OFF neurons, block of GABA and glycine feedback enhanced total NMDAR charge by 35-fold in the ON response and 9-fold in the OFF compared with a 1.7-fold enhancement of AMPA/KA receptors. Blocking only glycine receptors enhanced the NMDAR excitatory postsynaptic current 10-fold in the ON and 2-fold in the OFF pathway. Blocking GABA_A or GABA_C receptors (GABA_CRs or GABA_ARs) produced small changes in total NMDAR charge. When both GABA_ARs and GABA_CRs were blocked, the total NMDAR charge increased ninefold in the ON and fivefold in the OFF pathway. This exposed a strong GABA_CR feedback to bipolar cells that was suppressed by serial amacrine cell synapses mediated by GABA_ARs. The results indicate that NMDAR currents are large but latent, held in check by dual GABA and glycine presynaptic inhibition. One example of this controlled NMDAR activation is the cross talk between ON and OFF pathways. Blocking the ON pathway increased NMDAR relative strength in the OFF pathway. Stimulus prolongation similarly increased the NMDAR relative strength in the OFF response. This NMDAR enhancement was produced by a diminution in GABA and glycine feedback. Thus the retinal network recruits NMDAR pathways through presynaptic disinhibition.

ionotropic glutamate receptors are of two types, NMDA receptors (NMDARs) and non-NMDARs [AMPA/kainic acid receptors (AMPA/KARs)]. Retinal ganglion cells (RGCs), the output neurons of the retina, receive excitatory glutamatergic input from bipolar cell terminals and express both AMPA/KARs and NMDARs (Bloomfield and Dowling 1985; Gottesman and Miller 1992; Lukasiewicz and McReynolds 1985; Slaughter and Miller 1983). NMDARs are characterized by higher glutamate affinity and slower kinetics (Bekkers and Stevens 1989; Chen and Diamond 2002; Jahr and Stevens 1987; McBain and Dingledine 1992; Mittman et al. 1990; Patneau and Mayer 1990). Hence NMDARs and AMPA/KARs are thought to be complementary, extending the response range of the RGCs to read the output of bipolar cells (Buldyrev et al. 2012; Manookin et al. 2010). NMDARs and AMPA/KARs can combine to provide a linear correlation between excitatory synaptic input and spiking behavior over a range of stimulus strengths (Diamond and Copenhagen 1995). It has been suggested that the relative activations of NMDARs and AMPA/KARs also induce variations in response kinetics and contrast sensitivities in specific cell types (Sagdullaev et al. 2006).

Studies determining the relative strength of NMDAR to AMPA/KAR activation during ganglion cell synaptic stimulation have been conflicting. Some indicate that blocking NMDARs has a small effect on the light-evoked synaptic activity, even though functional NMDARs are expressed (Cohen and Miller 1994; Coleman and Miller 1988, 1989; Massey and Miller 1988, 1990; Slaughter and Miller 1983). Other studies indicate that NMDARs produce a significant component of the RGC synaptic responses to light, electrical, and pharmacological stimuli (Buldyrev et al. 2012; Chen and Diamond 2002; Diamond and Copenhagen 1993, 1995; Kalbaugh et al. 2009; Manookin et al. 2010; Matsui et al. 1998; Mittman et al. 1990; Sagdullaev et al. 2006, 2011; Taylor et al. 1995). It is possible that these reports are from different cell types with dissimilar expression of NMDARs accounting for the variable activation of NMDARs (Manookin et al. 2010). However, it is also likely that the differences are an outcome of strong regulation of NMDAR activation (Matsui et al. 2001; Sagdullaev et al. 2006). This work aimed to address that possibility and to determine the degree and mechanisms involved. The results indicate that synaptic input to ON-OFF ganglion cells can transition from AMPA/KA to NMDA receptor dominance under the control of presynaptic bipolar cell inhibition. We found that the cross talk between ON and OFF pathways provides one example where this transition occurs in the normal operation of the retina.

METHODS

Tissue preparation.

Larval tiger salamanders (Ambystoma tigrinum) were obtained from Charles Sullivan (Nashville, TN) and Kons Scientific (Germantown, WI) and were kept in tanks maintained at 4°C on a 12:12-h light-dark cycle. The animals were decapitated, and the eyes were enucleated. All procedures were performed in accordance with the US Animal Welfare Act and the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the University Animal Care Committee at the State University of New York. The eyeballs were hemisected under infrared light, and the posterior eye cup was placed in oxygenated Ringer solution. The retina was detached from the pigment epithelium and flat mounted on a glass coverslip (Bellco Glass, Vineland, NJ) coated with poly-l-lysine (Sigma-Aldrich, St. Louis, MO) with ganglion cells facing up. For slices, the retina was flat mounted ganglion side up on a 0.22-μm-pore membrane filter (Millipore, Bedford, MA) and sliced at 150–250 μm with a tissue slicer (Stoelting, Wood Dale, IL). Slices were rotated 90° and mounted on coverslips with vacuum grease (Dow Corning, Midland, MI). All electrophysiological experiments were done under infrared light. Coverslips with either a whole mounted retina or a retinal slice were transferred to the recording chamber attached to an upright Zeiss Axioskop2 FS fluorescent microscope, equipped with a ×40 Achroplan water immersion objective. An infrared-sensitive CCD camera (Hamamatsu) was used to capture the image of the preparation.

The tissue was constantly superfused with oxygenated Ringer solution containing (in mM) 111 NaCl, 2.5 KCl, 1.8 CaCl₂, 1 MgCl₂, 5 HEPES, and 10 dextrose buffered to pH 7.8 with NaOH. A gravity-fed perfusion system was used to maintain a flow rate of ∼1.5 ml/min for control Ringer solution.

Electrophysiology.

Recordings were made from neurons in the ganglion cell layer (GCL) of both wholemounts and slices at room temperature. In wholemount retina, the glial end feet were removed with an 8- to 10-MΩ electrode filled with Ringer solution to expose the soma of ganglion cells. First, the exposed neurons were sampled for extracellular spike activity by a loose seal (25–50 MΩ) with an 8- and 10-MΩ electrode filled with Ringer solution. On the basis of the extracellular spike recordings, ON-OFF transient cells were identified and then patched for whole cell recordings with a 5- to 7-MΩ electrode containing (in mM) 100 potassium gluconate, 5 NaCl, 1 MgCl₂, 5 HEPES, and 5 EGTA buffered to pH 7.4 with KOH.

Data were acquired with a Multiclamp 700B Amplifier (Molecular Devices, Sunnyvale, CA). Analog signals were low-pass filtered at 2 kHz and sampled at 10 kHz with the Digidata 1322A analog-to-digital board (Molecular Devices). Clampex 10.1 software (Molecular Devices) was used to control the voltage command outputs, acquire data, and trigger stimuli. The currents shown are raw data and were not corrected for electrode junction potential and access resistance. Both the series resistance and membrane capacitance were constantly monitored by a −20-mV square pulse (50-ms duration) before every light stimulus. Cells in which neither parameter changed during the entire course of the experiment were considered for further analysis. Drug solutions were delivered through a pressure-fed Octaflow 2 perfusion system (ALA Scientific Instruments, Farmingdale, NY). Picrotoxin, strychnine, meclofenamic acid (MFA), and 18α-glycyrrhetinic acid (αGA) were purchased from Sigma-Aldrich; d-2-amino-5 phosphonovaleric acid (d-AP5), l-(+)-2-amino-4-phosphonobutyric acid (l-AP4), 6-imino-3-(4-methoxyphenyl)-1(6H)pyridazinebutanoic acid hydrobromide (SR-95531), (1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA), and 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f] quinoxaline-7-sulfonamide (NBQX) were obtained from Tocris Bioscience (Minneapolis, MN).

Light stimulation.

Photoreceptors were stimulated by a 200-μm spot from a red light-emitting diode (LED, λ_max = 640 nm) projected through the objective lens. The irradiance between 620 and 660 nm of the LED was ∼1.2 μW/cm², measured by a RPS900-R wideband spectroradiometer (International Light, Peabody, MA). This light intensity has been used to stimulate cones (Song and Slaughter 2010). A 1-s-duration light stimulus was presented every 25 s.

Electrical stimulation.

Bipolar cells in retinal slices were directly stimulated by short pulses (1 ms) of current delivered through an electrode filled with Ringer solution that was placed into the outer plexiform layer directly above the patched ganglion cell (Awatramani and Slaughter 2001). The pulses were generated with a constant-current stimulator (Grass S48 with stimulus isolation unit PSIU6, Grass Instruments, West Warwick, RI).

Data analysis.

Traces were imported into IGOR Pro 6.22 (WaveMetrics) for making figures and further analysis. Unless otherwise mentioned, the total charge transferred by the excitatory postsynaptic current (EPSC) after light onset and offset was used as a measure of the ON and OFF light responses, respectively. The pooled data were imported to Microsoft Excel 2007 to make graphs and for statistical tests. Pooled data are expressed as means ± SE. Student's t-test was used to compare values under different conditions and was paired unless otherwise mentioned. Differences were considered significant when P ≤ 0.05.

RESULTS

Presynaptic inhibition regulates activation of NMDARs in ON-OFF transient cells.

ON-OFF transient cells in the GCL were initially identified based on their light-evoked spike activity with a loose-patch recording. They were characterized by a short transient burst of spikes at both the onset and offset of light (Fig. 1A, inset). Once identified, whole cell patch electrodes were used to record light-evoked transient ON-OFF EPSCs (L-EPSCs), representing glutamate output from bipolar cells (Fig. 1A). The L-EPSCs were recorded at −70 mV, close to the calculated Cl⁻ reversal potential (E_Cl = −71 mV). At −70 mV, NMDARs are subject to magnesium block by the control Ringer solution (1 mM Mg²⁺). However, removal of Mg²⁺ from the Ringer solution (nominally Mg²⁺ free) did not produce a statistically significant change in the L-EPSC (ON: 96 ± 13%, OFF: 107 ± 10%) (Fig. 1, A and B). A competitive NMDAR antagonist, d-AP5 (50 μM), in Mg²⁺-free Ringer solution slightly reduced the L-EPSCs (ON: 92 ± 9% remaining, OFF: 83 ± 6%); only the change at OFF was statistically significant (Fig. 1, A and B). Overall, removing magnesium did not produce a significant enhancement and NMDAR block produced a relatively small suppression in synaptically driven light responses under our control experimental conditions. This is consistent with previous studies indicating that light-evoked glutamatergic output from bipolar cells activates mainly AMPA/KARs (Coleman and Miller 1988, 1989; Slaughter and Miller 1983). However, when bipolar cell dendrites were electrically stimulated in slices, it activated both AMPA/KARs and NMDARs. Mg²⁺-free Ringer solution increased the electrically evoked EPSC (E-EPSC) compared with control (184 ± 20%) (Fig. 1, C and D). d-AP5 significantly reduced the enhanced E-EPSC (25 ± 10% remaining) (Fig. 1, C and D). Hence NMDARs can be synaptically activated with strong stimulation, but their activation by light stimulation is small under our control conditions. Furthermore, NMDARs are subject to Mg²⁺ block, but a Mg²⁺-free Ringer solution had little effect on the network under our control conditions.

Fig. 1.

Comparison of the NMDA receptor (NMDAR) current in light-evoked excitatory postsynaptic currents (L-EPSCs) and electrically evoked EPSCs (E-EPSCs). A shows L-EPSCs for an ON-OFF transient cell in control, Mg²⁺-free Ringer solution, and Mg²⁺-free Ringer solution with 50 μM d-2-amino-5 phosphonovaleric acid (d-AP5). The 1-s light stimulus is represented by the bar at top. Inset below the current traces shows the raster plot for spike activity evoked by a series of three 1-s light stimuli. B shows the quantification of the effect of Mg²⁺ free Ringer solution alone or with d-AP5 compared with control (n = 13). Mg²⁺-free Ringer solution had a small effect on the L-EPSCs (ON: P = 0.77, OFF: P = 0.47). d-AP5 also had a small effect on L-EPSCs, statistically significant only in the OFF pathway (ON: P = 0.44, OFF: *P < 0.05). C shows the E-EPSCs of a cell in the ganglion cell layer of a slice preparation in control and in Mg²⁺-free Ringer solution alone and with d-AP5. The stimulus was a 1-ms pulse applied to the outer plexiform layer. D shows the quantification of the effect of Mg²⁺-free Ringer solution and d-AP5 compared with control (n = 4). Mg²⁺-free Ringer solution greatly enhanced the E-EPSCs (*P < 0.05), and d-AP5 blocked a large portion of the E-EPSC (^#P ≤ 0.005).

The possibility that weak activation of NMDARs was due to low concentrations of the coagonist, either glycine or d-serine, was tested (Stevens et al. 2003). However, under control conditions the addition of 100 μM d-serine did not alter the light responses in ON-OFF cells (data not shown).

When inhibition was blocked there was a marked increase in the L-EPSC total charge and a large increase in light-evoked synaptic NMDAR responses in ON-OFF cells, similar to E-EPSCs shown in Fig. 1. In retinas treated with 100 μM picrotoxin [PTX, which blocks GABA_A receptors (GABA_ARs) and GABA_C receptors (GABA_CRs) in amphibian retina] and 10 μM strychnine [STR, which blocks glycine receptors (GlyRs)] in Mg²⁺-free Ringer solution, the L-EPSCs increased (ON: 407 ± 72%, OFF: 262 ± 34%) (Fig. 2, A and B). d-AP5 reduced the enhanced L-EPSCs (ON: 32 ± 5% remaining, OFF: 39 ± 5%) (Fig. 2, A and B). Overall, PTX and STR increased the ON EPSC by 407% and increased the percentage of the NMDAR component from 8% to 68% (Fig. 2C). Therefore, the NMDAR EPSC charge increased ∼35-fold. In the OFF pathway, PTX + STR increased the L-EPSC by 262% and the NMDAR fraction increased from 17% to 61%; the NMDAR charge increased by over ninefold (Fig. 2C). The AMPA/KAR EPSC in PTX and STR was not augmented as much (Fig. 2, A and B). To evaluate this, d-AP5 was used to isolate the non-NMDAR component of the L-EPSCs. PTX + STR increased the non-NMDAR EPSC total charge by 1.7-fold in both ON and OFF pathways (ON: 1.72 ± 0.39, OFF: 1.72 ± 0.29) (Fig. 2, D and E). In summary, PTX and STR increased the non-NMDAR ON and OFF responses by <2-fold while it increased the NMDAR responses by 35-fold in the ON and 9-fold in the OFF EPSC. Hence presynaptic inhibition disproportionately suppressed the NMDAR component of L-EPSCs. In four cells, the combination of 50 μM d-AP5 and 10 μM NBQX, a competitive AMPA/KAR antagonist, blocked the PTX + STR-enhanced L-EPSCs almost completely (ON: 5 ± 3% remaining, OFF: 8 ± 4%) (Fig. 2F), indicating that almost all the excitatory input to ganglion cells could be eliminated by combining AMPA/KAR and NMDAR antagonists.

Fig. 2.

L-EPSCs had a large NMDAR component when presynaptic inhibition was blocked. A shows the ON-OFF L-EPSCs of a cell in control and 100 μM picrotoxin (PTX) + 10 μM strychnine (STR) in Mg²⁺-free Ringer solution alone or with 50 μM d-AP5. B shows the quantification of the effects of PTX+STR and d-AP5+PTX+STR in Mg²⁺-free Ringer solution compared with control (n = 6). PTX+STR greatly enhanced the L-EPSCs (ON: *P < 0.01, OFF: **P < 0.005), and d-AP5 blocked a large portion of the enhancement (ON: ^#P < 0.001, OFF: ^#P < 0.001). C shows the comparisons of the NMDAR component in PTX+STR vs. control in the ON and OFF pathways (^#P < 0.001, unpaired t-test). D shows the ON-OFF L-EPSCs of a cell in d-AP5 alone and then with the addition of PTX+STR. E shows the change in the total control L-EPSC produced by d-AP5 alone and by d-AP5+PTX+STR in both ON and OFF pathways (n = 6; ON: P = 0.12, OFF: *P ≤ 0.05). F shows the ON-OFF L-EPSCs of a cell in control, then in PTX+STR in Mg²⁺-free Ringer solution, and then after addition of 50 μM d-AP5 + 10 μM 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f] quinoxaline-7- sulfonamide (NBQX) to the PTX+STR Mg²⁺-free Ringer solution.

Glycine inhibition plays a major role in regulation of NMDAR activation.

On the basis of these findings, experiments were performed to determine whether NMDAR activation was preferentially regulated by GABA_ARs, GABA_CRs, or GlyRs. The role of glycine inhibition in regulating the activation of NMDARs was tested with 10 μM STR. In six cells, STR in Mg²⁺-free Ringer solution increased the L-EPSCs in both ON and OFF pathways (ON: 155 ± 18%, OFF: 134 ± 14%) (Fig. 3, A and B). d-AP5 reduced the STR-enhanced L-EPSCs (ON: 49 ± 8% remaining, OFF: 72 ± 9%). Thus glycine inhibition produced a 10-fold increase in the NMDAR charge in the ON pathway and a 2-fold increase in the OFF pathway. This is qualitatively similar to the effects of PTX + STR but about a quarter of the magnitude.

Fig. 3.

Effects of GABA and glycine antagonists on the NMDAR current. A shows the L-EPSCs of a cell in Mg²⁺-free Ringer solution, then with addition of 10 μM STR, and then with 50 μM d-AP5 and STR in Mg²⁺-free Ringer solution. B shows the quantification of the effects of STR alone and d-AP5 + STR compared with Mg²⁺-free Ringer solution (n = 6). STR enhanced the L-EPSCs (ON: *P < 0.05, OFF: P = 0.06), and d-AP5 blocked a significant portion of the enhanced L-EPSC (ON: ^#P < 0.005, OFF: ^#P < 0.005). C shows the L-EPSCs of a cell in Mg²⁺-free Ringer solution, then + SR-95531 (SR), and then + SR and d-AP5. D shows the quantification of the effects of SR alone or of d-AP5 + SR compared with Mg²⁺-free Ringer solution (n = 7). The effects on both the total and peak L-EPSCs are shown. SR produced a statistically significant increase in total synaptic charge only in the OFF L-EPSC (ON: P = 0.75, OFF: *P < 0.05). d-AP5 significantly reduced only the OFF L-EPSC (ON: P = 0.1, OFF: *P < 0.05). SR increased the peaks of the L-EPSC (ON: ⁺P ≤ 0.01, OFF: ⁺P ≤ 0.01), while d-AP5 significantly reduced only the enhanced OFF peak (ON: P = 0.075, OFF: *P < 0.05). E shows the L-EPSCs of a cell in Mg²⁺-free Ringer solution, then + 100 μM (1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA), and then + TPMPA and d-AP5. F shows the quantification of the effects of TPMPA, and d-AP5 in TPMPA, compared with Mg²⁺-free Ringer solution (n = 7, 8). TPMPA enhanced the L-EPSCs (ON: **P < 0.001, OFF: *P < 0.05), but d-AP5 did not reduce the enhanced L-EPSCs significantly (ON: P = 0.13, OFF: P = 0.41). G and H show the comparison of NMDAR components in STR, SR, and TPMPA compared with control or PTX+STR in the ON and OFF pathways, respectively. In the ON pathway, only STR had a NMDAR component significantly higher than control (STR: ^#P < 0.005, SR: P = 0.67, TPMPA = 0.6, unpaired t-tests). In the OFF pathway, none of the 3 antagonists had a significantly higher NMDAR component than control (STR: P = 0.34, SR: P = 0.37, TPMPA = 0.36, unpaired t-tests).

In another three ON-OFF cells STR reduced the L-EPSCs (ON: 29 ± 4%, P < 0.005; OFF: 47 ± 4%, P < 0.005) and d-AP5 had a small and statistically insignificant effect on the STR L-EPSC (ON: 131 ± 14%, P = 0.167; OFF: 102 ± 26%, P = 0.95). This STR-induced reduction of L-EPSCs is probably the result of serial inhibition (Zhang et al. 1997), as described below.

The role of GABA_A inhibition was tested with SR-95531 (SR, gabazine), a selective GABA_AR antagonist. SR (10 μM) in Mg²⁺-free Ringer solution increased the peak of the L-EPSC (ON: 224 ± 36%, OFF: 168 ± 20%) but also abbreviated the response (Fig. 3, C and D, inset). Only the total synaptic charge in the OFF response was enhanced by SR (ON: 106 ± 18%; OFF: 192 ± 34%) (Fig. 3D). In the ON pathway, SR enhanced the initial peak but suppressed the prolonged component, resulting in little change in the total synaptic charge at light onset. The reduction in the prolonged component (portion of the EPSC after the peak) probably results from unmasking other inhibitory circuits when GABA_ARs are blocked. In the ON pathway, d-AP5 did not reduce the L-EPSC significantly (ON peak: 83 ± 8% remaining; ON area: 87 ± 9%) (Fig. 3, C and D). In the OFF pathway, where SR increased both the peak and total charge of the light response, d-AP5 reduced both (OFF peak: 73 ± 9% remaining; OFF area: 72 ± 9%) (Fig. 3, C and D, inset). The NMDAR component in the OFF pathway increased from 17% in control to 28% when GABA_ARs were blocked, and there was a threefold increase in the total NMDAR charge.

Glycine and GABA_A pathways had similar effects in suppressing NMDAR circuits in the OFF pathway. GABA_ARs had little effect in the total charge in the ON pathway, but interpretation was difficult because SR made the ON response more transient. When blocking an inhibitory pathway results in less excitation, it is likely that a serial inhibitory network is involved. Since GABA_CRs are particularly important in feedback inhibition to ON bipolar cells, the antagonist TPMPA was used to explore the role of GABA_C inhibition in regulating the activation of NMDARs. TPMPA (100 μM) in Mg²⁺-free Ringer solution increased the L-EPSCs in both the ON and OFF pathways (ON: 136 ± 5% remaining, OFF: 125 ± 10%) (Fig. 3, E and F). d-AP5 had a small effect on the TPMPA-enhanced L-EPSCs in both the ON and OFF pathways (ON: 86 ± 8% remaining, OFF: 93 ± 8%) (Fig. 3, E and F). This would suggest that GABA_CR pathways play a small role in the activation of NMDARs in both ON and OFF pathways. Since TPMPA alone did not reduce the L-EPSC in any of the cells, it seems likely that GABA_C inhibition does not regulate other inhibitory circuits (serial inhibition). This is consistent with studies indicating that GABA_CRs are localized primarily to bipolar cell terminals (Lukasiewicz and Werblin 1994; Sagdullaev et al. 2006).

In summary, the NMDAR component in the ON pathway increased from 8% in control to 51% in STR, to 13% in SR, and to 14% in TPMPA (Fig. 3G). Only STR induced an ON NMDAR component comparable to PTX + STR. In the OFF pathway, the NMDAR component changed from 17% in control to 28% in STR, 27% in SR, and 7% in TPMPA (Fig. 3H). The OFF NMDAR component in the presence of PTX + STR (61%) was significantly higher.

GABA_A inhibition regulates activation of GABA_C inhibition.

Unlike the data presented above, previous reports suggest that GABA_CRs regulate the activation of NMDARs (Matsui et al. 2001; Sagdullaev et al. 2006, 2011). This could be a species difference, as the other studies were performed in mouse retina. Another possible explanation is that our recordings were from retinal wholemounts while most of the data in these reports were from retinal slices (but see Sagdullaev et al. 2011). Such a discrepancy has been reported before in the effect of d-serine as a coagonist for NMDAR activation. The coagonist site seems to be saturated in retinal slices but not in wholemounts (Kalbaugh et al. 2009; Stevens et al. 2010). When we recorded from ON-OFF transient cells in the retinal slice preparation, 100 μM TPMPA in Mg²⁺-free Ringer solution increased the L-EPSCs in both ON and OFF pathways (ON: 148 ± 13%, OFF: 145 ± 10%) (Fig. 4, A and B). d-AP5 (50 μM) reduced the TPMPA-enhanced L-EPSCs in both the ON and OFF pathways (ON: 51 ± 12% remaining, OFF: 39 ± 11%) (Fig. 4, A–C). This result implies that GABA_AR or GlyR circuits suppress activation of GABA_CR synapses in the wholemount retina preparation. These inhibitory circuits may be truncated in retinal slices, thereby disinhibiting GABA_CR pathways. Studies have shown that GABA_AR inhibition regulates activation of GABA_CRs (Buldyrev and Taylor 2013; Eggers and Lukasiewicz 2006; Roska et al. 1998; Vigh et al. 2011; Zhang et al. 1997). This leads to the prediction that blocking GABA_CRs along with GABA_ARs would increase the NMDAR component of the L-EPSCs in wholemount retina.

Fig. 4.

GABA_C receptors (GABA_CRs) suppressed a significant NMDAR component in L-EPSCs in the retinal slice. In the retinal wholemount, GABA_CRs suppressed a large NMDAR component only when GABA_A receptor (GABA_AR) inhibition was also blocked. A shows the L-EPSCs of a cell in a retinal slice in Mg²⁺-free Ringer solution, + 100 μM TPMPA and then 50 μM d-AP5 in TPMPA. B shows the quantification of the effects of TPMPA and of d-AP5+TPMPA compared with Mg²⁺-free Ringer solution (n = 6). TPMPA enhanced the L-EPSCs (ON: *P ≤ 0.01, OFF: *P < 0.01), and d-AP5 reduced the enhanced L-EPSCs significantly (ON: *P ≤ 0.01, OFF: ^#P < 0.005). C shows the comparison of the NMDAR component of the TPMPA-enhanced L-EPSCs in retinal slices and wholemounts. This component was significantly higher in retinal slices compared with wholemounts (ON: *P < 0.05, OFF: ^#P < 0.005, unpaired t-tests). D shows the L-EPSCs of a cell in the retinal wholemount in Mg²⁺-free Ringer solution, SR+TPMPA, and then d-AP5 in SR+TPMPA. E shows the quantification in the wholemount retina of the effects of SR+TPMPA vs. d-AP5+SR+TPMPA compared with Mg²⁺-free Ringer solution (n = 5). SR+TPMPA increased the L-EPSCs (ON: *P < 0.05, OFF: *P < 0.05), and d-AP5 reduced the enhanced L-EPSCs significantly (ON: *P < 0.05, OFF: ^#P < 0.005). F shows the comparison of the NMDAR component of L-EPSC in the wholemount retina in SR+TPMPA compared with control. SR+TPMPA had a higher NMDAR component than control (ON: P = 0.09, OFF: *P < 0.01, unpaired t-tests).

TPMPA (100 μM) with SR (10 μM) in the wholemount retina increased the L-EPSCs in transient ON-OFF cells (ON: 187 ± 26%, OFF: 197 ± 33%) (Fig. 4, D and E). d-AP5 (50 μM) reduced the SR + TPMPA-enhanced L-EPSCs (ON: 63 ± 13% remaining; OFF: 53 ± 7%) (Fig. 4, D and E). Thus SR + TPMPA produced a significant NMDAR component in the ON pathway (37%), which was not observed with either drug alone. Furthermore, in the OFF pathway, SR + TPMPA produced an NMDAR component significantly higher than control (Fig. 4F). Thus the combined effect of SR + TPMPA was more profound than summation of their individual effects. This supports the premise of serial inhibition and that GABA_AR inhibition suppresses GABA_CR feedback to bipolar cells.

The combination of SR + TPMPA increased the NMDAR EPSC in the ON response ninefold and in the OFF response fivefold. Compared with glycine feedback, combined GABA feedback produces a similar suppression of the ON NMDAR EPSC but twice the suppression in the OFF pathway. The results confirm that GABA_CR inhibition can regulate activation of NMDARs.

Stimulus duration differentially activates NMDARs.

The pharmacological manipulations may reveal NMDAR pathways that are utilized by the retina, or they may be a physiologically irrelevant epiphenomenon. We were able to differentially regulate NMDAR activation in ON-OFF cells by simply lengthening the light stimulus duration, thus providing evidence that NMDAR regulation by presynaptic inhibition occurs under physiological conditions.

The duration of the light stimulus was varied from 1 to 2 and 3 s under control conditions. In the OFF response, the 1 s L-EPSC was significantly smaller than the 2 s, which was similar to the 3 s L-EPSC (Table 1, Fig. 5, A and B, inset). The larger response after 2 s probably results from photoreceptor adaptation. In two cells, the stimuli were increased to 4 and 5 s, but this did not further increase the L-EPSC (data not shown). As expected, the ON L-EPSCs were similar in all three stimuli (Table 1, Fig. 5, A and B, inset). The ratio of excitation (OFF L-EPSC/ON L-EPSC) increased from 0.62 in 1-s to 0.99 in 2-s and 1.04 in 3-s stimuli (Table 1, Fig. 5B). Thus the total charge of the OFF EPSC and the OFF/ON excitation ratio increased as the stimulus duration increased from 1 to 2 s, and there was little change with further increase in stimulus duration.

Table 1.

ON and OFF L-EPSCs for 1-, 2-, and 3-s light stimuli

	1-s Light		2-s Light		3-s Light
	ON	OFF	ON	OFF	ON	OFF
Control	100%	100%	96 ± 3%	162 ± 12%	90 ± 2%	159 ± 15%
OFF-to-ON ratio	0.62 ± 0.08%		0.99 ± 0.11%		1.04 ± 0.11%
Control subset	100%	100%	92 ± 4%	191 ± 22%	90 ± 3%	202 ± 30%
d-AP5 on subset	90 ± 9%	81 ± 6%	84 ± 5%	64 ± 5%	91 ± 10%	73 ± 4%
d-AP5 on l-AP4 treated subset	—	51 ± 5%	—	58 ± 6%	—	—

Comparison of the ON and OFF light-evoked excitatory postsynaptic currents (L-EPSCs) recorded for 1-, 2-, and 3-s light stimuli in control, 50 μM d-2-amino-5 phosphonovaleric acid (d-AP5), and 20 μM l-(+)-2-amino-4-phosphonobutyric acid (l-AP4) + d-AP5. Data rows 1 and 2: The OFF response to 2-s or 3-s stimulation is larger than to 1-s stimulation. Data rows 3–5: This tabulates a subpopulation of the cells in the first row in which the OFF response to a 2-s light stimulus is at least 50% larger than the OFF response to a 1-s light stimulus. d-AP5 suppressed the 1-s, 2-s, and 3-s OFF responses compared with the control subset (P ≤ 0.01). When the ON response was blocked by l-AP4, d-AP5 suppressed the 1-s and 2-s OFF response (P < 0.001). All responses are normalized to the control ON and OFF EPSCs for a 1-s light stimulus.

Fig. 5.

NMDAR component of the OFF L-EPSC increased with stimulus duration. A shows the L-EPSCs of a transient ON-OFF cell to a series of light stimuli of 1-, 2-, and 3-s duration in Mg²⁺-free Ringer solution and then with 50 μM d-AP5 added. B shows the ratio of OFF to ON L-EPSC for these stimuli (n = 17, 18). The ratio is significantly higher for 2- and 3-s stimuli compared with the 1-s stimulus (2 s: *P ≤ 0.01, 3 s: *P < 0.01). Inset shows the 2 and 3 s L-EPSCs compared with the 1 s ON and OFF L-EPSCs. The 2- and 3-s stimuli have a significantly higher OFF L-EPSC compared with the 1-s stimulus (Table 1; ^#P < 0.001). However, the ON L-EPSC did not change significantly (Table 1). C shows the comparison of the NMDAR component in the OFF L-EPSC for the 3 stimuli (n = 7). The NMDAR components in the 2- and 3-s stimuli were significantly higher than in the 1-s stimulus (Table 1; 2 s: **P ≤ 0.01, 3 s: *P < 0.05).

To study the NMDAR EPSC in these responses, we selected neurons in which the 2 s OFF L-EPSC was at least 50% larger than the 1 s OFF L-EPSC. In these cells, the 2 s and 3 s OFF L-EPSCs were ∼90% higher than the 1 s OFF L-EPSCs (Table 1). d-AP5 (50 μM) reduced the OFF L-EPSCs for all three stimulus durations (Table 1, Fig. 5A). The NMDAR component of the 1 s OFF L-EPSC was 19% and increased to 36% with a 2-s stimulus and to 27% with a 3-s stimulus (Fig. 5C). In the two cells in which the stimulus was increased to 4 and 5 s the NMDAR component was ∼35%. Compared with the 1 s OFF L-EPSC, the 2 s OFF L-EPSC had a 3.6-fold increase in NMDAR EPSC; the 3 s OFF L-EPSC had a 2.7-fold increase. Therefore, prolonging the stimulus from 1 s to 2 s increased the OFF EPSC, the total NMDAR charge, and the NMDAR fraction of the total EPSC.

The ON pathway regulates NMDAR activation of the OFF pathway.

The ON pathway was blocked by l-AP4 to test whether NMDAR recruitment was due to interaction between ON and OFF pathways (Slaughter and Miller 1981). d-AP5 was used to determine the amount of the EPSC that was due to NMDAR activation (Fig. 6A). l-AP4 (20 μM) in Mg²⁺-free Ringer solution blocked the ON L-EPSC of transient cells completely and increased the 1 s OFF L-EPSC compared with control (223 ± 40%) (Fig. 6, A and B). d-AP5 (50 μM) reduced the l-AP4-enhanced OFF EPSC to 51 ± 5% (Fig. 6, A and B). Thus when the ON response was blocked the NMDAR component in the OFF response was significantly enhanced. Compared with control, where the ON response was present, there was a 2.6-fold increase in the OFF NMDAR component (from 19% to 49%) of the EPSC and a 5.8-fold increase in total NMDAR EPSC charge. The AMPA/KAR EPSC did not show a significant increase in l-AP4 (1.1-fold increase compared with control, P = 0.31, unpaired t-test). Thus there are endogenous circuits between the ON and OFF pathways that selectively control the level of NMDAR activation. When this cross talk is blocked, then the OFF response increases because of recruitment of NMDAR activation.

Fig. 6.

Blocking cross talk increased the NMDAR component of the OFF L-EPSC. A shows a cell's L-EPSCs to a 1-s light stimulus in control, then with 20 μM l-(+)-2-amino-4-phosphonobutyric acid (l-AP4) in Mg²⁺-free Ringer solution, then with the addition of 50 μM d-AP5 for 1-s stimuli. B shows the mean total OFF L-EPSC produced by a 1-s light stimulus, normalized to the OFF response in control (n = 8). l-AP4 in Mg²⁺-free Ringer solution blocked the ON L-EPSC completely and increased the OFF L-EPSC significantly (*P < 0.05). d-AP5 reduced a large fraction of the enhanced OFF L-EPSC (^#P < 0.005). C shows the comparison of the NMDAR component of the OFF L-EPSC in control and l-AP4 for 1- and 2-s stimuli. The NMDAR component of the OFF response in the presence of l-AP4 was not significantly different between 1- and 2-s stimuli (P = 0.29). The NMDAR component of the 1 s OFF L-EPSC was significantly greater in l-AP4 (^#P < 0.005, unpaired t-test). However, the NMDAR component of the 2 s OFF L-EPSC was not significantly different (P = 0.49, unpaired t-test). D shows L-EPSCs of a cell in PTX+STR in Mg²⁺-free Ringer solution containing 50 μM 18α-glycyrrhetinic acid (18-αGA) and then with the addition of l-AP4 for 1-s stimuli. E shows the effect of l-AP4 and d-AP5 on the relative mean total OFF L-EPSC in Mg²⁺-free Ringer solution containing either 18-αGA or 100 μM meclofenamic acid (MFA). l-AP4 increased the OFF EPSC (*P < 0.005; n = 8), while d-AP5 blocked a large component of the enhanced EPSC (^#P < 0.001; n = 5). F shows the normalized total OFF L-EPSC in PTX+STR (in Mg²⁺-free Ringer solution containing 18-αGA/MFA) and the effect of l-AP4. l-AP4 did not increase the OFF L-EPSC in the presence of PTX+STR (P = 0.42; n = 7).

Blocking the ON pathway with l-AP4 augmented the NMDAR OFF component with a 1-s stimulus (Fig. 6). However, when a 2-s stimulus was used, the NMDAR component in control was 36% of the total EPSC, and this was not significantly increased when cross talk was blocked with l-AP4 (42%) (Table 1, Fig. 6C). These results correlate with the experiments prolonging the light stimulus (Fig. 5), where the OFF NMDAR EPSC was augmented with a 2-s stimulus under control conditions. Thus the OFF NMDAR component was relatively small during a 1-s stimulus and l-AP4 enhanced this component, but the OFF NMDAR-component was larger during a 2-s stimulus and l-AP4 produced little enhancement. This suggests that suppression of NMDAR pathways was due to short-term cross talk from the ON pathway and this cross talk diminished with a 2-s stimulus.

To evaluate involvement of GABA and glycine inhibition in ON-OFF cross talk, the effects of l-AP4 on the OFF L-EPSC were tested after pretreatment of the tissue with PTX + STR. In our experiments, we found that both gap junctions and inhibition influenced the effects of l-AP4. For the purpose of this study, we performed experiments in the presence of gap junction blockers in order to focus on the impact of cross inhibition. To block gap junctions, the tissue was pretreated for 10 min with either 50 μM αGA or 100 μM MFA. The effects of both gap junction blockers were similar, and hence data were pooled for statistical analysis. l-AP4 increased the 1 s OFF L-EPSC by 212 ± 23% in the presence of αGA/MFA (Fig. 6E). d-AP5 reduced the enhanced L-EPSC to 41 ± 5%. Hence the effects of l-AP4 and d-AP5 were similar in control (Fig. 6B) and in αGA/MFA (Fig. 6E), indicating that gap junction blockers did not alter the drug responses. The application of PTX and STR (in the presence of αGA/MFA) enhanced both the ON and OFF responses, but now the effect of l-AP4 was occluded (94 ± 7%) (Fig. 6, D and F). Thus ionotropic GABAergic and glycinergic cross talk from the ON pathway regulates the NMDAR component in the OFF pathway. This illustrates an endogenous inhibitory circuit that regulates activation of NMDA receptor pathways. If gap junctions were active but inhibition was blocked, l-AP4 also increased the 1 s OFF response. This second pathway through gap junctions was not explored.

DISCUSSION

This study demonstrates the prominent role that presynaptic inhibition plays in the control of NMDAR activation in retina. More importantly, it illustrates that this regulation is dynamic and employed in normal visual function. In summary, 1) NMDAR activation is minimal when presynaptic inhibition is active, 2) glycinergic inhibition strongly suppresses NMDAR activation during the ON L-EPSC in ganglion cells, 3) concatenated synapses between GABA_AR and GABA_CR pathways reduce NMDAR current in both the ON and OFF L-EPSCs, and 4) short-term cross talk from ON to OFF pathways inhibits NMDAR activation in the OFF EPSC.

Background.

The role of NMDARs in retina has been paradoxical. In early studies comparing NMDA and KA effects on third-order neurons in rabbit and salamander retina, both agonists produced strong depolarizations, yet blocking NMDARs had little effect on light-evoked excitatory postsynaptic potentials (EPSPs) (Bloomfield and Dowling 1985; Massey and Miller 1990; Slaughter and Miller 1983). Subsequent studies in amphibian retina demonstrated synaptic NMDAR responses, but these experiments were performed in the presence of STR and PTX (Diamond and Copenhagen 1993; Mittman et al. 1990). Still later it was shown that blocking inhibition allowed for spillover of glutamate, leading to activation of perisynaptic NMDARs in rodent retina (Sagdullaev et al. 2006; Zhang and Diamond 2006). The conclusion was that NMDARs were activated under conditions of excessive bipolar cell transmitter release, and the implication was that this might not occur in the normal functioning of the retina.

A few specific populations of ganglion cells receive large synaptic NMDAR EPSCs, particularly OFF cells. For example, OFF alpha ganglion cells in the guinea pig retina have large NMDAR conductances, but ON alpha cells do not (Manookin et al. 2010). Similarly, OFF brisk-sustained ganglion cells, but not ON brisk-sustained ganglion cells, are driven largely by NMDAR activation in rabbit retina (Buldyrev et al. 2012; Buldyrev and Taylor 2013). In mouse retina, OFF cells have a larger NMDAR component compared with ON cells (Yang et al. 2011). However, the ON alpha cells in mouse have significant NMDAR conductances (Manookin et al. 2010).

The objective of our study was to evaluate the influence of GABAergic and glycinergic pathways in the control of light-driven NMDAR current in amphibian ganglion cells. It was somewhat surprising that feedback inhibition of bipolar cells completely reversed the relative dominance of AMPA and NMDA receptor currents at the ganglion cell. The implication is that a reduction in local feedback inhibition would be sufficient to augment NMDARs at particular synapses and this could change the EPSC dynamics. We could not determine whether the small NMDAR current observed under our control experimental stimulus conditions represented synapses that were disinhibited, rather than unregulated, constitutive NMDAR circuits.

Cross talk regulates NMDARs in OFF pathway.

One circuit in which NMDAR activation is strongly regulated is the cross talk between ON and OFF pathways. The OFF bipolar cell output to third-order neurons is enhanced when the light stimulus is prolonged or when the ON bipolar light response is blocked. The enhanced OFF response was almost entirely due to recruitment of NMDARs. The enhanced OFF bipolar cell output may lead to spillover, but that response is part of the natural physiological repertoire of the retina.

Cross talk between ON and OFF pathways is a prominent feature of visual processing and takes a number of forms. In crossover inhibition, excitation of the ON pathway inhibits the OFF pathway and vice versa. Crossover inhibition has been shown to be due to glycinergic amacrine cells in rabbit retina (Hsueh et al. 2008; Molnar and Werblin 2007). This type of cross talk enhances the OFF response because it now represents both excitation and disinhibition. It functions to improve linearity (Molnar et al. 2009). l-AP4 blocks the crossover inhibitory pathway by blocking the ON pathway (Hsueh et al. 2008; Zaghloul et al. 2003). However, we find that l-AP4 enhances the OFF response, suggesting a different mechanism of cross talk. This has been reported previously in the mudpuppy retina (Arkin and Miller 1987, 1988). The effect of l-AP4 on presynaptic mGluRs in bipolar terminals can also be ruled out, as it would reduce the L-EPSC, opposite to our results (Awatramani and Slaughter 2001). Although we have determined that this cross talk is through ionotropic inhibitory circuits, the mechanism by which it regulates the OFF L-EPSC is yet to be deciphered.

A similar cross talk may be present in the OFF brisk-sustained ganglion cells in rabbit retina (Buldyrev et al. 2012). The ganglion cells receive bipolar input that activates both AMPARs and NMDARs and is regulated by glycinergic input from the ON pathway, both presynaptically to bipolar cell terminals and postsynaptically to the ganglion cells. However, there is no evidence that glycine inhibition alters the balance of NMDA and AMPA/KA activation in the pathway.

Cross talk can also be excitatory. In amphibian retinal ON-OFF ganglion cells there is a small population in which the ON response is suppressed by GABA/glycine receptor antagonists and a larger group in which these antagonists block the OFF response (Pang et al. 2007). It is postulated that this indirect inhibitory input to one set of bipolars, from the opposing set, is mediated by inhibitory amacrine cells. We have not encountered examples of this type of cross talk.

NMDAR activation is disproportionately regulated by presynaptic inhibition in both ON and OFF synapses.

We confirmed previous studies finding that NMDARs are expressed in third-order neurons (Bloomfield and Dowling 1985; Kalbaugh et al. 2009; Mittman et al. 1990; Zhang and Diamond 2006, 2009) but their activation is minimal when presynaptic inhibitory circuits are intact (Coleman and Miller 1988, 1989; Slaughter and Miller 1983). Several studies found synaptic NMDARs only in the OFF pathway (Sagdullaev et al. 2006; Zhang and Diamond 2009). This might explain why the NMDAR component, although small, was significant only in the OFF L-EPSC under our control conditions. Amacrine cells also possess NMDARs and AMPA/KARs (Dixon and Copenhagen 1992). This could mean that NMDAR antagonists would reduce amacrine cell excitation and consequently reduce feedback inhibition to bipolar cells. The reduced inhibitory feedback could potentially enhance NMDAR activation on ganglion cells, but this would be blocked in our experiments because of the NMDAR antagonist.

When inhibitory circuits are blocked, then NMDARs contribute significantly to light responses in ON-OFF transient cells (Diamond and Copenhagen 1993, 1995; Mittman et al. 1990; Taylor et al. 1995). However, the extent of this transition had not been characterized previously. Hence our observations bridge the varied literature in the long-standing debate on whether NMDARs contribute to light responses of RGCs. The sheer magnitude of the total increase in ON pathway synaptic (4×) and NMDAR EPSCs (35×) regulated by presynaptic inhibition is noteworthy. When comparing this to the 1.7× increase in AMPA/KAR EPSCs, it is clear that NMDAR activation is disproportionately regulated by presynaptic inhibition. The supposition is that AMPA/KARs are located at the synapse and saturated by glutamate release. Therefore, increases in AMPA/KAR currents represent recruitment of additional synapses. On the other hand, NMDARs may be largely perisynaptic and can be stimulated by the spillover that results from more glutamate release at the active sites (Chen and Diamond 2002; Sagdullaev et al. 2006) (Fig. 7). Most of the studies on the relative strength of activation of NMDARs and AMPA/KARs in the past have blocked inhibitory circuits in order to isolate excitation (Chen and Diamond 2002; Kalbaugh et al. 2009; Matsui et al. 1998; Sagdullaev et al. 2006). Our observations suggest that these protocols not only simplify the retinal network but fundamentally alter the properties of excitatory signaling.

Fig. 7.

Proposed model of inhibitory feedback to bipolar cells that regulates their glutamate release and activation of ganglion cell excitatory receptors. Retinal ganglion cells (RGCs) express AMPA/kainic acid receptors (AMPA/KARs) and NMDARs in the synaptic and perisynaptic space, respectively. The activation of NMDARs is determined by enhanced glutamate release from bipolar terminals, regulated by inhibition from glycinergic and GABAergic amacrine cells. The GABA_AR pathway acts to both inhibit bipolar cells and inhibit GABA_CR pathways. Under the control conditions depicted in the figure, AMPA/KARs, but few NMDARs, are activated. GABA_CR feedback is low because of GABA_AR inhibition. Block of glycinergic inhibition or both GABA_AR and GABA_CR inhibition augments glutamate release, and consequently perisynaptic NMDARs become activated.

GABA and glycine inhibition.

Glycinergic amacrine cells in mammalian retina are generally narrow field (Menger et al. 1998; Pourcho and Goebel 1987) and form ∼50% of the amacrine cell population (Koontz et al. 1993; MacNeil and Masland 1998; Wassle et al. 1986). In salamander retina there are multistratified wide-field and narrow-field glycinergic amacrine cells (Yang et al. 1991), and glycinergic input is found both at the synaptic axon terminal and more distally at bipolar cell dendrites (Maple and Wu 1998). The axon terminal input is from amacrine cells and the dendritic input from interplexiform cells (Shen and Jiang 2007). We found that GlyRs exerted more control in directly regulating NMDAR activation than either GABA_CRs or GABA_ARs. This may be the result of amphibian bipolar cells receiving dual glycinergic input. However, glycinergic input from OFF-responding interplexiform cells was found to inhibit ON bipolar cells in the dark (Maple and Wu 1998). Thus ON signals would get opposing glycinergic signals at light onset: disinhibition from OFF interplexiform cells and inhibition from ON and ON-OFF amacrine cells.

Signals from GABA receptors were only found at the axon terminal of amphibian bipolar cells (Maple and Wu 1996). Mouse rod bipolar terminals express GABA_ARs, GABA_CRs, and GlyRs, although GABA_CR currents dominate (Eggers and Lukasiewicz 2006) and GABA_C inhibitory circuits regulate NMDAR activation in mouse (Matsui et al. 2001; Sagdullaev et al. 2006, 2011). Our findings in intact salamander retina are that NMDAR pathway regulation by GABA_CRs only becomes prominent when GABA_AR pathways are blocked (Fig. 7).

Serial inhibition indirectly regulates NMDAR activation through GABA_CRs.

Serial inhibition, in which amacrine cells inhibit other amacrine cells, makes it difficult to fully evaluate the importance of the individual transmitter systems (Eggers and Lukasiewicz 2006; Eggers et al. 2007; Roska et al. 1998; Vigh et al. 2011; Zhang et al. 1997). This is most evident in our experiments for the GABAergic pathways in the intact retina, where selective suppression of either GABA_ARs or GABA_CRs had only a small effect on synaptic excitation, but when both receptors were blocked then the NMDAR total charge in the ON EPSC increased ninefold. This suggests interdependence between the two receptors. Blocking only GABA_ARs made the ON L-EPSCs transient, indicating an increase in sustained inhibitory circuits, most likely GABA_CR circuits. We did not observe reciprocal effects on GABA_AR pathways when blocking GABA_CRs. Hence it is likely that GABA_ARs provide inhibition not only to bipolar terminals but also to amacrine cells activating GABA_CRs (Fig. 7). A similar circuit has been described recently in the surround inhibition of rabbit RGCs (Buldyrev and Taylor 2013). Sometimes even blocking GlyRs decreased the L-EPSCs, indicating that GABAergic amacrine cells can be regulated by GlyR activation. This is evident in the ON NMDAR responses, where blocking GlyRs increased the response 10-fold, blocking GABA receptors produced a 9-fold increase, but blocking all ionotropic GABA receptors and GlyRs increased the ON NMDAR total charge by 35-fold. The reason that the individual receptor blockers do not add up to the total block is presumably the effects of serial inhibition. Therefore, it is likely that we are underestimating the magnitude of each transmitter system in regulation of NMDAR pathways.

GRANTS

This work was supported by National Eye Institute Grant EY-05725.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

Author contributions: S.S. and M.M.S. conception and design of research; S.S. performed experiments; S.S. and M.M.S. analyzed data; S.S. and M.M.S. interpreted results of experiments; S.S. and M.M.S. prepared figures; S.S. and M.M.S. drafted manuscript; S.S. and M.M.S. edited and revised manuscript; S.S. and M.M.S. approved final version of manuscript.