Glycine transporter 1 modulates GABA release from amacrine cells by controlling occupancy of coagonist binding site of NMDA receptors

Abstract

The occupancy of coagonist binding sites of NMDA receptors (NMDARs) by glycine or d-serine has been thought to mediate NMDAR-dependent excitatory signaling, as simultaneous binding of glutamate and a coagonist is obligatory for NMDAR activation. Amacrine cells (ACs) mediating GABAergic feedback inhibition of mixed bipolar cells (Mbs) in the goldfish retina have been shown to express NMDARs. Here we studied whether NMDAR-mediated GABAergic inhibitory currents (I_GABA) recorded from the axon terminals of Mbs are influenced by experimental manipulations altering retinal glycine and d-serine levels. Feedback I_GABA in Mb axon terminals was triggered by focal NMDA application or by synaptically released glutamate from depolarized Mb terminals. In both cases, blocking the coagonist binding sites of NMDARs eliminated the NMDAR-dependent I_GABA, demonstrating that coagonist binding is critical in mediating NMDAR activity-triggered GABA release. Glycine transporter 1 (GLYT1) inhibition increased I_GABA, indicating that coagonist binding sites of NMDARs on ACs providing GABAergic feedback inhibition to Mbs were not saturated. Focal glycine application, in the presence of the ionotropic glycine receptor blocker strychnine, triggered a GLYT1-dependent current in ACs, suggesting that GLYT1 expressed by putative glycinergic ACs controls the saturation level of NMDARs' coagonist sites. External d-serine also increased NMDAR activation-triggered I_GABA in Mbs, further substantiating that the coagonist sites were unsaturated. Together, our findings demonstrate that coagonist modulation of glutamatergic input to GABAergic ACs via NMDARs is strongly reflected in the AC neuronal output (i.e., transmitter release) and thus is critical in GABAergic signal transfer function in the inner retina.

activation of nmda receptors (NMDARs) requires simultaneous occupancy of their glutamate and coagonist binding sites (Kleckner and Dingledine 1988) located on the NR1 subunits (Cull-Candy and Leszkiewicz 2004; Lynch et al. 1994). In the central nervous system both glycine (Johnson and Ascher 1987) and d-serine (Mothet et al. 2000) can bind to the coagonist binding site of the NMDAR, which appears to be nonsaturated in some brain regions and saturated in others (Mothet et al. 2000; Wilcox et al. 1996).

In the retina, investigations addressing the nature of endogenous coagonist and saturation level of its binding sites on the NMDARs have focused on ganglion cells (GCs) (Daniels and Baldridge 2010; Gottesmann and Miller 1992; Gustafson et al. 2007; Kalbaugh et al. 2009; Lukasievicz and Roeder 1995; Reed et al. 2009; Stevens et al. 2003; Sullivan et al. 2011; Sullivan and Miller 2012; Yazejian and Fain 1992), as NMDARs are known to contribute to light-evoked synaptic currents (Diamond and Copenhagen 1993; Lukasiewicz and McReynolds 1985; Massey and Miller 1990; Mittman et al. 1990). It was proposed that ambient d-serine, produced by Müller cells (Stevens et al. 2003), sets a baseline for synaptic NMDAR activation (Kalbaugh et al. 2009), whereas glycine, released in the inner retina by glycinergic amacrine cells (ACs), might provide a more dynamic and rapid fine-tuning (Kalbaugh et al. 2009). Retinal glycine levels are controlled by the highly abundant glycine transporter 1 (GLYT1) (Pow and Hendrickson 1999), and, indeed, coagonist binding sites of NMDARs expressed by GCs of GLYT1-deficient mice appeared to be saturated (Reed et al. 2009). More recent studies in the retinas of transgenic mice lacking the endogenous d-serine-synthesizing enzyme serine racemase (SR) demonstrated a complete absence of light-evoked NMDAR-mediated component of GC responses (Sullivan et al. 2011). This finding suggests that d-serine is absolutely critical for NMDAR function in GCs and that in the absence of d-serine ambient glycine may not substitute for d-serine to provide any light-evoked NMDAR activation in GCs.

Retinal ACs are also known to express NMDARs in a variety of species (Dixon and Copenhagen 1992; Hartveit and Veruki 1997; Matsui et al. 1998, 2001; Mittman et al. 1990); NMDARs have been shown to contribute to the light-evoked synaptic responses of ACs particularly in nonmammalian retinas (Matsui et al. 1998; Mittman et al. 1990). In the mammalian retina NMDARs expressed by ACs have been shown to play a direct role in releasing glycine from ACs (Chavez and Diamond 2008).

The objective of the present study was to elucidate whether the modulation of NMDAR signaling via their coagonist binding sites can influence the transmitter release from GABAergic ACs. We recorded NMDAR activation-evoked GABAergic inhibitory postsynaptic currents (IPSCs) from the axon terminals of Mb-type bipolar cells (Mbs) in acute slices of the goldfish retina. IPSCs were evoked by focal NMDA puffs or by glutamate released synaptically in response to depolarization of a single Mb axon terminal. Our results suggest that coagonist binding sites of NMDARs on ACs providing GABAergic feedback inhibition to bipolar cells (BCs) are not saturated and GLYT1, presumably localized on glycinergic ACs, can control GABA release from GABAergic ACs by influencing the occupancy of the glycine-binding subunits of their NMDARs. In keeping with this notion, exogenous d-serine also potentiated both synaptic glutamate-triggered (reciprocal) and NMDA-evoked GABAergic IPSCs in Mbs. However, in our hands NMDAR-mediated reciprocal IPSCs recorded from Mbs in slice preparation were not eliminated or reduced by acute manipulations known to reduce endogenous d-serine levels in the retina. These results indicate that the role of d-serine is less critical in occupying the coagonist binding sites of NMDARs mediating reciprocal inhibition of Mbs than what has been shown in various species at NMDARs expressed by retinal GCs (Kalbaugh et al. 2009; Stevens et al. 2010a; Sullivan et al. 2011).

Our results add a new line to the growing body of evidence showing that regulating the occupancy of coagonist binding sites of NMDARs can influence the neuronal response to glutamatergic input. Specifically, on the basis of our results we propose that regulating the saturation level of these binding sites on GABAergic ACs provides a route for regulation of inhibitory signaling in the retinal circuitry.

MATERIALS AND METHODS

Retinal slice preparation.

Animals were handled according to protocols reviewed and approved by the Colorado State University Institutional Animal Care and Use Committee, and all procedures met US Public Health Service guidelines. Light-adapted retinal slices were prepared from the retina of goldfish (Carassius auratus) of either sex under standard room light, as described previously (Palmer et al. 2003b). Slices (200 μm thick) were superfused at ∼3 ml/min with a Ringer solution containing (in mM) 100 NaCl, 2.5 KCl, 1.0 MgCl₂, 2.5 CaCl₂, 25 NaHCO₃, 0–0.2 ascorbic acid, and 12 glucose (osmolarity 260 ± 2 mosM and pH 7.45, set with NaOH). The Ringer solution was gassed continuously with 95% O₂-5% CO₂. NMDA (500 μM, in Ringer solution) or glycine (1 mM) was puff applied from small-tipped (∼1-μm diameter) glass pipettes positioned over the inner plexiform layer (IPL) of the retina at ∼20–30 μm away from the Mb terminal with Picospritzer III (Parker, Cleveland, OH). Short (100 ms) puffs were aimed toward the BC terminals or ACs, parallel to and in the direction of the flow of perfusion. Other drugs were bath applied in the perfusing medium for 3–5 min before their effect was tested. (S)-(+)-α-amino-4-carboxy-2-methylbenzeneacetic acid (LY 367385), 6-imino-3-(4-methoxyphenyl)-1(6H)-pyridazinebutanoic acid hydrobromide (SR95531), (1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid (TPMPA), N-[(3R)-3-([1,1′-biphenyl]-4-yloxy)-3-(4-fluorophenyl)propyl]-N-methylglycine hydrochloride (ALX 5407), 3-((R)-2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid [(R)-CPP], d-(−)-2-amino-5-phosphonopentanoic acid (d-AP5), and 5,7-dichlorokynurenic acid (DCKA) were obtained from Tocris (Bristol, UK). Alexa Fluor 488 was purchased from Life Technologies (Grand Island, NY). All other chemicals and salts were obtained from Sigma (St. Louis, MO).

Electrophysiology.

Whole cell current recordings from voltage-clamped giant terminals of Mb BCs and ACs were performed with the EPC-10 USB patch-clamp amplifier (HEKA) in conjunction with Patchmaster software (version 2.30) at room temperature during daytime to avoid circadian changes in transmitter release from BCs (Hull et al. 2006). One set of experiments was performed on intact Mb BCs (see Fig. 2Aii); the rest of the Mb recordings were made from Mb terminals whose axon was severed during the slicing procedure (see Fig. 1B). Axotomized Mb terminals in the IPL were identified on the basis of their electrophysiological signature as described previously (Palmer et al. 2003b). ACs in the inner nuclear layer (INL) were identified by means of fluorescent dye injection (0.1% Lucifer yellow or 25 μM Alexa Fluor 488) via the recording pipette. Dye-filled cells were visualized by using a mercury light source (X-cite 120Q, LumenDynamics) and appropriate filter cubes (Zeiss). The INL of the goldfish retina is formed by somas of horizontal cells, BCs, ACs, Müller cells, and occasional large, round displaced GCs (Marc et al. 1990). During these experiments, cells labeled by fluorescent dye injection via the recording pipette were considered to be putative ACs if they 1) had a soma <10 μm wide (to exclude displaced GCs); 2) were located close to the border of IPL (to exclude horizontal cells); and 3) had no apparent process toward the outer plexiform layer (to exclude Müller cells and BCs). Voltage-clamp series resistance (R_s) errors were not electronically compensated, and liquid junction potential was not corrected. Membrane potential of recorded cells was held at −60 mV unless otherwise noted. Recordings from Mb terminals with uncompensated R_s > 30 MΩ or leak > 50 (−pA) at a holding potential of −60 mV were excluded from further evaluation. Recordings were obtained with 6- to 12-MΩ patch pipettes pulled from 1.5-mm-diameter thick-walled borosilicate glass (World Precision Instruments, Sarasota, FL) on a horizontal puller (model P-97, Sutter, Novato, CA), coated with dental wax (Cavex, West Chester, PA) to reduce pipette capacitance, and filled with solution containing (in mM) 95 Cs-gluconate, 25 HEPES, 10 TEA-Cl, 3 Mg-ATP, 0.5 Na-GTP, and 2 EGTA, adjusted to pH 7.2 with CsOH. For experiments addressing the amplitude of reciprocal feedback IPSCs in BCs, in order to remove exocytosed proton-mediated inhibition of presynaptic voltage-gated calcium current (I_Ca) (DeVries 2001; Palmer et al. 2003a) methylamine HCl (10 mM) was included to buffer vesicular pH (Cousin and Nicholls 1997; Vigh et al. 2005). In these experiments equimolar Cs-gluconate was removed to keep osmolarity constant.

Fig. 2.

Glycine transporter 1 (GLYT1) blocker increased NMDA receptor (NMDAR)-mediated GABA release from ACs. Ai: NMDA puff-evoked GABAergic feedback to axotomized Mb terminals (control) was increased by GLYT1 blocker ALX 5407 (1 μM) at −60 mV holding potential. Aii: NMDA puff-evoked GABAergic feedback to axon terminals of intact Mbs (control) was increased by GLYT1 blocker ALX 5407 (1 μM) at −60 mV holding potential. Bi: epifluorescence of a Lucifer yellow-filled putative AC, located at the innermost portion of INL of the goldfish retina. Scale bar, 10 μm. Bii: NMDA puff-evoked inward current (control) in ACs was increased by GLYT1 blocker ALX 5407 (1 μM) at −60 mV holding potential. C: summary diagram of normalized increase of NMDA-evoked inhibitory currents (I_GABA) recorded from Mb terminals with severed (Mb_T) and intact (Mb_I) axons as well as NMDAR-mediated excitatory currents from ACs. Data are presented as means ± SE for numbers of cells in parentheses.

Fig. 1.

Focal application of NMDA onto mixed bipolar cell (Mb) axon terminals evoked GABAergic feedback from amacrine cells (ACs). Ai: experimental setting for recording NMDA-evoked currents from a voltage-clamped axon terminal of Mb, with severed axon. BC, bipolar cell. Aii: montage of IR-DIC and epifluorescence of Alexa Fluor 488-filled, axotomized Mb terminal in the innermost layer of the inner plexiform layer (IPL) of the goldfish retina slice during recording. Note the termination of the severed axon (arrow) at the border of IPL. INL, inner nuclear layer; GCL, ganglion cell layer. Bi: focal NMDA puff (500 μM, 100 ms)-evoked inward current (control) in Mb terminals held at −80 mV. This inward current was increased by prefusion with Ringer solution containing no Mg²⁺ (0 Mg), and it was eliminated upon addition of SR95531 (25 μM) and TPMPA (100 μM) even in the absence of extracellular Mg²⁺ (+SR+TP/−80 mV). Bii: same cell as in Bi; consecutive traces are divided into 2 panels for better visibility. “0 Mg” trace is same as in Bi. In the presence of SR95531 and TPMPA NMDA puff did not evoke any current in axotomized Mb terminals at −30 mV holding potential (+SR+TP/−30 mV). In this representative cell incomplete recovery of the NMDA puff-evoked inward current at −80 mV was seen after 15 min of wash with control Ringer solution (1 mM Mg²⁺). C: summary diagram of normalized NMDA-evoked inward currents recorded from Mb terminals in 0 Mg and in the presence of SR95531 and TPMPA at −80 and −30 mV holding potentials. Data are presented as means ± SE for numbers of cells in parentheses.

For real-time measurements of membrane capacitance of axotomized Mb terminals a Sine+DC technique was used, in which a 1-kHz sinusoidal voltage command (20 mV peak to peak) was added to the holding potential of −60 mV (see Fig. 4Di, stimulus marker of holding potential, HP) and the resulting current was analyzed at two orthogonal phase angles by the lock-in amplifier (Gillis 2000).

Fig. 4.

Inhibition of GLYT1 increased NMDAR-mediated GABAergic reciprocal feedback to Mbs. Ai: depolarization of axotomized Mb terminals from −60 to −20 mV (holding potential, HP) for 100 ms evoked GABAergic reciprocal feedback to axotomized Mb terminals (I_GABA) overlying inward voltage-gated calcium current (I_Ca). I_GABA was insensitive to strychnine (STR, 5 μM) but was absent in picrotoxin (PTX, 100 μM). Note that PTX trace was obtained in separate experiments, averaged (n = 29), and scaled to the I_Ca peak (materials and methods). Aii: net reciprocal I_GABA in control and in STR obtained from the experiment shown in Ai by subtracting the PTX trace from the corresponding traces (see materials and methods). B: representative experiment showing that net reciprocal I_GABA was not increased by addition of external glycine (Gly, 100 μM) in the presence of STR. C: addition of GLYT1 blocker ALX 5407 (1 μM) along with glycine (100 μM) in the presence of STR increased the reciprocal I_GABA, particularly its later phase (arrow). Di: GLYT2 inhibitor ALX 1393 did not affect reciprocal I_GABA and had no effect on the depolarization (from −60 to −20 mV, for 100 ms)-evoked presynaptic glutamate release either, as evidenced by the membrane capacitance (C_m) measurements. Dii: same cell as in Di; consecutive traces are divided into 2 panels for better visibility. GLYT1 inhibitor ALX 5407 increased the reciprocal I_GABA, particularly its later components (arrow). Note that ΔC_m remained the same across treatments. E: summary diagram of changes in normalized reciprocal I_GABA across the experiments presented in Ai–Dii. Data are presented as means ± SE for numbers of cells in parentheses.

Data analysis.

Off-line data analysis was performed with IGOR Pro software (version 5.03; WaveMetrics, Lake Oswego, OR). Quantification of NMDA puff-evoked IPSCs was performed by integrating the current trace to calculate charge transfer (Q). Reciprocal feedback IPSCs recorded from Mb terminals triggered by 100-ms depolarization of the Mb terminal from −60 to −20 mV were quantified according to Li et al. (2007). In short, to isolate feedback IPSCs from I_Ca, we first recorded the pure I_Ca in 100 μM picrotoxin (PTX, a blocker of both GABA_A and GABA_C receptors in goldfish; Vigh et al. 2005) in response to a 100-ms depolarization. The traces (n = 29) were first normalized to the peak I_Ca amplitude and then were averaged. This pure I_Ca “template” was scaled to the peak I_Ca of each individual current response with superimposed reciprocal IPSCs (see Fig. 4A). The template trace then was subtracted from the one with IPSCs. The resulted difference trace was taken as the isolated GABAergic reciprocal feedback response, and we used its area (i.e., the total chloride charge transfer Q_Cl) (Vigh et al. 2005) to quantify the amount of GABAergic feedback (Li at al. 2007). These experiments were performed in the presence of LY 367385 (100 μM) to avoid reciprocal feedback potentiation due to AC mGluR1 metabotropic glutamate receptor activation (Vigh et al. 2005). Recordings showing >15% rundown of I_Ca peak amplitude or in the membrane depolarization-evoked increase in membrane capacitance (ΔC_m) across treatments were excluded from analysis. ΔC_m, when measured, was calculated as the difference between the baseline average C_m value measured for 100 ms before the depolarizing step and the average C_m value measured during 100 ms after the step, starting 200–300 ms after repolarization to allow time for all evoked conductances to have decayed (Vigh and von Gersdorff 2005). Statistics were calculated with SigmaPlot (version 11; Systat Software). Paired two-tailed Student's t-tests were used to compare data sets obtained in a self-controlled manner. For comparing means of two independent data sets with known, calculated variance, a z-test was used. Data are reported as means ± SE.

RESULTS

Focal NMDA application evokes GABAergic IPSCs in Mb terminals.

Whole cell current recordings were made from axotomized Mb terminals voltage clamped at −80 mV (Fig. 1Ai). These large bulbous structures are located in the ON sublamina of the IPL, close to the GC layer in the goldfish retinal slice preparation (Fig. 1Aii). NMDA puff (500 μM, 100 ms) evoked large inward current (n = 15, Fig. 1Bi) in our standard external solution containing 1 mM Mg²⁺ (materials and methods). Perfusing the preparation with Ringer solution containing no Mg²⁺ significantly increased the NMDA-evoked inward current (Fig. 1Bi) to 335.8 ± 57.3% of control (n = 6, P < 0.007, Student's t-tests; Fig. 1C). The NMDA-evoked inward current was eliminated by simultaneous application of GABA_A receptor blocker SR95531 (25 μM) and GABA_C receptor blocker TPMPA (100–150 μM) at −80 mV holding potential (Fig. 1Bi), regardless of the presence or absence of Mg²⁺ in the Ringer solution (n = 5 with 1 mM Mg²⁺, n = 5 without Mg²⁺ summed in Fig. 1C; 5.2 ± 1.5% of control). These results are consistent with our earlier reports showing that Mb terminals receive GABA_A and GABA_C receptor-mediated reciprocal and lateral feedback inhibition from ACs and NMDAR activation on ACs presynaptic to Mb terminals contributes to both reciprocal and lateral feedback mechanisms (Vigh and von Gersdorff 2005; Vigh et al. 2011). To exclude the possibility that exogenous NMDA acts directly on NMDARs expressed by, and located on, the Mb terminals, the membrane potential of axotomized terminals was held at −30 mV to facilitate the removal of voltage-dependent Mg²⁺ block of putative NMDARs (Lukasiewicz and Roeder 1995). Importantly, under these conditions NMDA puff did not evoke membrane current in Mg²⁺-free Ringer solution in the presence of SR95531 and TPMPA (3.8 ± 2.7% of control, n = 5; Fig. 1, Bii and C). This negative result is consistent with the observation that expression of NMDARs is limited to ACs and GCs in the goldfish inner retina as suggested by immunohistochemistry and in situ hybridization (Vandenbranden et al. 2000a) and thus functional NMDARs are not expressed at Mb terminals.

Together, these results suggested that short, focal NMDA application around Mb terminals triggered GABA release from AC processes that acted on GABA_A and GABA_C receptors to mediate an inward current (I_GABA) at the Mb terminals held at −80 mV. Experiments reported from this point on were performed in our standard external solution containing 1 mM Mg²⁺, and voltage-clamped retinal neurons were subjected to −60 mV holding potential.

Glycine transporter block increases NMDAR-mediated GABA release from ACs.

Next we tested whether GLYT1 controls the occupancy of NMDARs' coagonist binding site in GABAergic ACs as shown for GCs in the mouse retina (Reed et al. 2009) and, importantly, whether GLYT1 block could influence the NMDAR activation-triggered GABA release from ACs. To study the GLYT1 activity mediating GABAergic feedback to Mbs, we recorded NMDA-evoked (GABAergic) IPSCs from Mb terminals. We found that perfusion of the recording chamber with the GLYT1 blocker ALX 5407 (1 μM) significantly increased NMDA puff (500 μM, 100 ms)-evoked I_GABA at Mb axon terminals regardless of the integrity of the axon: in axotomized terminals (Fig. 2Ai) this increase was 51.0 ± 7.2% (n = 5, P < 0.03, Student's t-tests; Fig. 2C) compared with the control vs. 32.0 ± 7.2% (n = 5, P < 0.02, Student's t-tests; Fig. 2C) in intact cells (Fig. 2Aii). We found no significant difference between the effects ALX 5407 exerted on NMDA puff-evoked I_GABA recorded from intact cells and axotomized terminals (P < 0.39, z-test), further strengthening the notion that putative outer retinal NMDARs were not activated by the NMDA puffs in our experiments. This is in concert with the observation that NMDAR distribution is restricted to the inner retina in goldfish (Vandenbranden et al. 2000a).

A plausible explanation for the above results is that GLYT1 inhibition might have increased NMDAR responses in ACs, as shown for retinal GCs (Reed et al. 2009), which in turn led to increased GABA release. To test this notion directly, putative ACs were targeted for recording of NMDA puff-evoked currents in the INL. Besides using their retinal location and morphological characteristics (see materials and methods), cells responding to NMDA puff with inward current at the holding potential of −60 mV in the presence of PTX (100 μM) were considered ACs. At this concentration, PTX not only blocks GABA_A and GABA_C receptors in the goldfish retina (Vigh et al. 2005) it was also shown to strongly inhibit (ionotropic) glycine receptor-mediated currents (Li and Slaughter 2007). Figure 2Bi shows the image of a putative AC, which responded to a short (100 ms), focal application of NMDA (500 μM) with an inward current (I_NMDA) (Fig. 2Bii). Blocking GLYT1 with ALX 5407 (1 μM) enhanced I_NMDA in ACs (Fig. 2Bii), and this increase was found to be significant (190.8 ± 33.2%, n = 5, P < 0.01, Student's t-tests; Fig. 2C). Note that the AC identification method used in these experiments is sufficient to separate ACs from the rest of the cell types whose somas are located in the INL (i.e., horizontal cells, Müller cells, BCs) with high certainty but cannot differentiate between the two major AC classes in the vertebrate retina, GABAergic and glycinergic ACs (MacNeil and Masland 1998; Menger et al. 1998; Wässle 2004). Although in the goldfish retina the ratio of GABAergic and glycinergic ACs is 70% to 30%, respectively (Marc 1989; Marc et al. 1990), which favors random recording from GABAergic ACs over glycinergic ACs, we cannot claim that these recordings were made solely from GABAergic ACs.

Nevertheless, these results together support the notion that blocking GLYT1 might elevate the extracellular glycine concentration in the retinal slice high enough to increase glycine binding to the coagonist binding sites at NMDARs, in turn increasing NMDAR signaling in ACs and NMDAR activation-evoked GABA release onto Mb terminals.

To test this notion directly, we studied the effect of GLYT1 blocker ALX 5407 (1 μM) on NMDAR signaling in ACs in external solution containing strychnine and PTX, in the presence of DCKA (25 μM), a competitive antagonist at the coagonist site (Kleckner and Dingledine 1988). DCKA eliminated the NMDA puff (500 μM, 100 ms)-evoked inward current in putative ACs (n = 3; Fig. 3Ai). Addition of GLYT1 blocker ALX 5407 (1 μM) in the presence of DCKA did not increase I_NMDA (92.2 ± 11.9%, n = 3, P < 0.85, Student's paired t-tests), which recovered after 15 min of wash with normal Ringer solution in one of the three cases (Fig. 3Aii). Similar experiments were performed in the absence of PTX, recording NMDA-evoked I_GABA from Mb terminals. In light of our previous results showing no difference in the effect of ALX 5407 on NMDA-evoked I_GABA recorded from axotomized terminals and intact Mbs (Fig. 2A, i and ii, and C), for these experiments we pooled data obtained from either. DCKA (25 μM) eliminated the NMDA-evoked I_GABA (Fig. 3Bi). Application of ALX 5407 (1 μM) in the presence of DCKA did reveal a tiny I_GABA in two of the six experiments (Fig. 3Bii). Although ALX 5407 increased NMDA-evoked I_GABA to 150.6 ± 60.7% of that recorded in the presence of DCKA, this increase remained statistically nonsignificant across our experiments (n = 6, P < 0.79, Student's paired t-tests). Partial recovery of NMDA-evoked I_GABA was achieved after 15- to 20-min wash in two of five experiments (Fig. 3Bii).

Fig. 3.

GLYT1 blocker effect on NMDAR-dependent AC signaling is mediated via the coagonist binding sites of NMDARs. Ai: NMDA puff-evoked inward current (I_NMDA, control) in ACs was eliminated by the NMDAR coagonist site blocker 5,7-dichlorokynurenic acid (DCKA, 25 μM) at −60 mV holding potential. Aii: same cell as in Ai; consecutive traces are divided into 2 panels for better visibility. In the presence of DCKA, GLYT1 blocker ALX 5407 (1 μM) could not recover I_NMDA, which was recovered after 20 min of wash in normal Ringer solution. Bi: DCKA (25 μM) eliminated NMDA-evoked GABAergic feedback (I_GABA) to axotomized Mb terminals (control). Bii: same cell as in Bi; consecutive traces are divided into 2 panels for better visibility. In the presence of DCKA, GLYT1 blocker ALX 5407 (1 μM) revealed a tiny I_GABA, which was partially recovered after 20 min of wash in normal Ringer solution.

These data suggest that 1) agonist binding to the coagonist binding sites at NMDARs expressed by ACs is absolutely critical for NMDA-evoked GABAergic feedback to Mb terminals and 2) once the coagonist binding sites are blocked by DCKA, GLYT1 blocker ALX 5407 is largely ineffective. Furthermore, these results indicate that although ALX 5407 appears to increase NMDA-evoked I_GABA by elevating retinal glycine and, in turn, NMDAR signaling, even the elevated glycine is not sufficient to outcompete 25 μM DCKA at the coagonist sites.

Block of GLYT1 increases NMDAR-mediated GABAergic reciprocal feedback to Mbs.

Next we investigated whether blocking glycine transporters can influence NMDAR function under conditions in which NMDARs, expressed by GABAergic ACs, are activated by endogenous glutamate, synaptically released from Mbs. We depolarized a single axotomized Mb terminal from −60 to −20 mV for 100 ms in the presence of the Cs-based internal solution. The depolarization of presynaptic Mb terminal releases glutamate onto AMPA receptors (AMPARs) and NMDARs on the ACs, triggering GABAergic reciprocal feedback (Fig. 4A) mediated by GABA_A and GABA_C receptors on Mb terminals with a ∼1.5-ms delay (Vigh et al. 2005; Vigh and von Gersdorff 2005). Accordingly, the outward reciprocal inhibitory current (I_GABA) overlying the inward voltage-gated calcium current (I_Ca), which was absent in the presence of PTX (100 μM) (Fig. 4Ai), was insensitive to the ionotropic glycine receptor blocker strychnine (5 μM; Fig. 4Aii, 98.0 ± 1.0% of control, n = 5, P < 0.17, Student's t-tests; Fig. 4E, see materials and methods for quantification). This finding not only confirms that (ionotropic) glycinergic signaling does not contribute directly to the reciprocal inhibition of the Mb terminals but also shows that GABAergic ACs providing reciprocal inhibition to Mb terminals were not disinhibited by strychnine. In other words, under our experimental conditions, those GABAergic ACs that form reciprocal inhibitory synapses with the Mb terminals do not seem to receive active glycinergic inhibition. Note that GABA_A receptor blocker SR95531 can actually increase feedback to Mbs (Li et al. 2007), indicating that GABAergic inputs to ACs via GABA_A receptors (Watanabe et al. 2000) inhibit GABA release onto Mb terminals via serial inhibitory synapses (Marc and Liu 2000), similar to other vertebrate retinas (Eggers and Lukasiewicz 2010; Zhang et al. 1997).

Although this finding suggested that (ionotropic) glycinergic signaling does not contribute to the reciprocal inhibitory signaling at Mb terminals, we performed the following experiments in the constant presence of strychnine (5 μM) to avoid potential interference with ionotropic glycine effects exerted elsewhere in the retinal circuitry. Under these conditions, application of glycine (100 μM) elevated the reciprocal feedback in two of five cells (to ∼142% of the control in average), whereas it was slightly reduced in the other three experiments (to ∼82% of the control in average), making the overall effect a nonsignificant change (106.8 ± 15.3%, n = 5, P < 0.7, Student's t-tests; Fig. 4, B and E). We interpreted these bimodal results as follows: GLYTs might prevent the saturation of the NMDARs' coagonist binding site (Reed et al. 2009) by exogenous glycine (n = 3) applied at a dose that should be sufficient to increase NMDAR signaling (Supplisson and Bergman 1997; Wilcox et al. 1996) in retinal slice preparation, provided that neuronal processes expressing GLYTs in the proximity of recorded Mb terminals are preserved and functional. However, if these processes are destroyed during slicing, the GLYT function might be impacted such that the exogenous glycine reaches the NMDARs, increasing the occupancy of the coagonist binding sites and, in turn, increasing NMDAR activation-dependent GABA release onto Mb terminals (n = 2). To test this notion further, in the next set of experiments exogenous glycine (100 μM) was applied together with the GLYT1 blocker ALX 5407 (1 μM) in the presence of strychnine. The GABAergic reciprocal feedback increased (Fig. 4C) every time (n = 8), although the enhancement varied between 1% and 71%, I_GABA averaging 130.0 ± 10.0% of the control (P < 0.01, Student's t-tests; Fig. 4E). Our interpretation of these results is that blocking GLYT1 can significantly increase the GABAergic feedback, presumably by increasing glycine binding to the NMDARs' coagonist binding site.

Nonetheless, it was not clear whether or not exogenous glycine was critical in this process, or whether GLYT1 block by itself could elevate ambient retinal glycine level. In particular, GLYT1 block alone may increase endogenous glycine in the synaptic cleft around the Mb terminal, resulting in increased NMDAR-mediated reciprocal feedback. Therefore, we tested whether GLYT1 or GLYT2 activity can influence the reciprocal inhibition of Mbs without exogenous glycine.

Application of the GLYT2-specific blocker ALX 1393 (1 μM), in the presence of strychnine (5 μM), did not increase the reciprocal feedback (Fig. 4Di) to the Mb terminals but instead slightly reduced it, although this reduction was not significant (93.8 ± 2.3% of control, n = 3, P < 0.09, Student's t-test; Fig. 4E) and might have been related to rundown. However, application of GLYT1 blocker ALX 5407 (1 μM) increased I_GABA (Fig. 4Dii) to 136.6 ± 8.1% of control (n = 6, P < 0.005, Student's t-test; Fig. 4E). Importantly, GLYT1 block increased the reciprocal GABAergic feedback at the same Mb terminals where GLYT2 block was ineffective (Fig. 4, Di vs. Dii). Furthermore, in these experiments ALX 5407 did not increase the depolarization-evoked I_Ca peak amplitude in Mb terminals (95.0 ± 2.7% of control, n = 6, P < 0.07, Student's t-test; Fig. 4Dii) or the amount of consequently released glutamate (ΔC_m: 104.4 ± 11.7% of control, n = 6, P < 0.59, Student's t-test; Fig. 4Dii), both of which could potentially contribute to the increase in reciprocal feedback (Vigh and von Gersdorff 2005).

Collectively, these data supported the notion that the site of action through which GLYT1 blocker ALX 5407 modulated the reciprocal inhibitory feedback was downstream from the Mb terminal. Note that GLYT1 block seemed to increase particularly the late portions of the reciprocal I_GABA (Fig. 4, C and Dii). This is consistent with the notion that at Mb terminals the initial transient component of the reciprocal feedback is triggered by high-affinity, relatively desensitizing AMPARs, whereas the late, more sustained parts are evoked primarily by low-affinity, slowly desensitizing NMDAR activation (Vigh and von Gersdorff 2005).

To test this notion directly, we applied GLYT1 blocker ALX 5407 (1 μM) in the presence of strychnine and a selective NMDAR antagonist, (R)-CPP (20 μM). A representative experiment is shown in Fig. 5, Ai and Aii. (R)-CPP reduced the late portions of the reciprocal I_GABA and did not change the transient initial component (Fig. 5Aii), in agreement with our earlier reports where d-AP5, another selective NMDAR blocker, was used (Vigh and von Gersdorff 2005; see also Fig. 7E). In the presence of (R)-CPP, GLYT1 blocker ALX 5407 did not increase the reciprocal GABAergic feedback [Fig. 5, Ai and Aii; 100.1 ± 10.1% of (R)-CPP alone, n = 3, P < 0.50, Student's t-tests]. The coagonist binding site antagonist of NMDARs DCKA (25–50 μM) also reduced the late components of reciprocal I_GABA in the presence of strychnine (Fig. 5, Bi and Bii). Importantly, we found that in the presence of DCKA addition of ALX 5407 did not increase the reciprocal GABAergic feedback to Mb terminals (Fig. 5, Bi and Bii; 98.8 ± 1.3% of DCKA alone, n = 5, P < 0.56, Student's t-tests).

Fig. 5.

GLYT1 blocker effect on GABAergic reciprocal inhibition of Mbs is mediated via the coagonist binding sites of NMDARs expressed by ACs. Ai: blocking NMDARs with (R)-CPP (20 μM) markedly reduced the reciprocal I_GABA in Mb terminals without altering presynaptic glutamate release (ΔC_m). In the presence of (R)-CPP (20 μM), GLYT1 blocker ALX 5407 (1 μM) did not alter I_GABA. Aii: net reciprocal I_GABA obtained from the traces shown in Ai. Note that NMDAR block with (R)-CPP markedly inhibited the late portions of the reciprocal I_GABA while leaving the transient initial portion intact. Bi: blocking the coagonist binding sites of NMDARs with DCKA (25 μM) markedly reduced the reciprocal I_GABA in Mb terminals without altering presynaptic glutamate release (ΔC_m). In the presence of DCKA, GLYT1 blocker ALX 5407 (1 μM) did not alter I_GABA. Bii: net reciprocal I_GABA obtained from the traces shown in Bi. Note the similarities between the effects of NMDAR block with (R)-CPP (Aii) and DCKA: 1) both reduced primarily the late portions of the reciprocal I_GABA, and 2) GLYT1 inhibitor ALX 5407 was ineffective in the presence of both (R)-CPP and DCKA.

Fig. 7.

Exogenous d-serine increased NMDAR-mediated GABA release from ACs onto Mb terminals. A: focal NMDA puff (500 μM, 100 ms) evoked inward current (control) in Mb terminals held at −60 mV. This inward current was increased by perfusion with Ringer solution containing d-serine (100 μM). In this representative cell, addition of PTX (100 μM) to the solution eliminated NMDA puff-evoked inward current in the presence of d-serine. B: application of serine racemase inhibitor phenazine ethosulfate (10 μM) together with d-amino acid oxidase (100 μg/ml) via the Ringer solution (Phen+DAO) slightly reduced NMDA-evoked I_GABA over 10 min of application compared with the control. Ci: in axotomized Mb terminals depolarization-evoked reciprocal I_GABA was increased when d-serine (100 μM) was added to the Ringer solution, particularly the later components (arrow). Note that ΔC_m was not increased by d-serine. Cii: net reciprocal I_GABA obtained from the traces shown in Ci. D: application of phenazine ethosulfate (10 μM) together with d-amino acid oxidase (100 μg/ml) via the Ringer solution (Phen+DAO) did not reduce reciprocal I_GABA over 10 min of application compared with the control. E: after 2 h of incubation of retinal slices with phenazine ethosulfate (10 μM) and d-amino acid oxidase (100 μg/ml), the presynaptic depolarization-evoked reciprocal I_GABA contained a large, d-(−)-2-amino-5-phosphonopentanoic acid (d-AP5)-sensitive (NMDAR mediated) component. Note that ΔC_m remained unchanged by the d-AP5 treatment.

ALX 5407 inhibits GLYT-associated membrane currents in amacrine cells.

The above result indicated that GLYT1 blocker ALX 5407 acted on GLYT1-expressing cells to elevate retinal glycine levels, which in turn enhanced NMDAR function on GABAergic ACs and NMDAR-activation triggered GABA release onto Mb terminals. GLYT-mediated glycine uptake is known to be coupled to the sodium gradient across the membrane (Betz et al. 2006) and associated with inward current at −60 mV holding potential (Zhang et al. 2008). GLYT1 is expressed primarily by glycinergic ACs in the mammalian and avian (Pow and Hendrickson 1999) retina, and it plays an essential role in maintaining their glycine content (Pow 1998). In amphibians, GLYT1 is expressed by Müller cells (Jiang et al. 2007) and cones (Zhang et al. 2008), whereas GLYT2 is expressed by ACs (Jiang et al. 2007). In the goldfish retina, Aa-type ACs and I2 interplexiform cells exhibit high-affinity glycine uptake (Marc et al. 1978; Marc and Lam 1981). However, the molecular homology between GLYTs expressed by these cells in the goldfish retina and other vertebrate GLYT types is not known.

To test directly whether ALX 5407 acted on GLYTs expressed by ACs, we targeted putative glycinergic Aa-type ACs in the INL (see materials and methods) to record membrane currents associated with GLYT activation. GLYT “currents” were sought to be evoked at −60 mV holding potential by focal glycine puff (1 mM, 100 ms) in the presence of the ionotropic glycine receptor blocker strychnine (5–10 μM), similar to the protocol used for recording GLYT1-mediated current from bullfrog cones (Zhang et al. 2008). Figure 6 summarizes the results of successful experiments, that is, when glycine puff did evoke an inward current (I_Gly) in the presence of strychnine. These GLYT “currents” appeared insensitive to GLYT2 inhibitor ALX 1393 (1 μM; Fig. 6A) application. However, detailed statistical analysis of the data revealed a small but significant reduction of I_Gly by ALX 1393 (to 96.6 ± 0.5% of the control, P < 0.03, n = 5, Student's t-test; Fig. 6C). The GLYT1-specific blocker ALX 5407 caused a much more prominent inhibition of I_Gly (Fig. 6B), reducing it to 52.8 ± 10.0% of the control (P < 0.03, n = 5, Student's t-test; Fig. 6C). Nonetheless, the inhibition caused by ALX 5407 was never complete. These results are in line with our experiments demonstrating a significant increase of the reciprocal inhibition at Mb terminals by ALX 5407 but not by ALX 1393. Collectively, they suggest that ALX 5407 acted on Aa-type ACs arborizing primarily in the IPL (Marc and Lam 1981) to potentiate the NMDAR signaling mediating the reciprocal feedback to Mb terminals.

Fig. 6.

ALX 5407 inhibited GLYT-associated membrane currents in ACs. A: in the presence of strychnine (5 μM), focal glycine puff (1 mM, 100 ms) triggered slow inward current (I_Gly) at −60 mV holding potential in putative ACs (control), which was rather insensitive to GLYT2 inhibitor ALX 1393 (1 μM). B: focal glycine puff (1 mM, 100 ms)-triggered slow inward current (I_Gly) in putative ACs was inhibited by GLYT1 inhibitor ALX 5407 (1 μM). C: summary diagram of normalized effects of ALX 1393 and ALX 5407 on glycine-evoked membrane currents (I_Gly) recorded from putative ACs. Data are presented as means ± SE for numbers of cells in parentheses.

Exogenous d-serine increases NMDAR-mediated GABA release from ACs.

To test the effect of exogenous d-serine on the NMDAR-mediated GABA release from ACs, NMDA puff (100 μM, 100 ms)-evoked GABAergic IPSCs were recorded from Mb terminals in the absence (control) and in the presence of d-serine (100 μM). We found that addition of d-serine to the external solution significantly increased the NMDA-evoked GABAergic IPSCs (Fig. 7A), to 145.0 ± 14.3% of control (P < 0.04, n = 5, Student's t-test). Next, we tested whether endogenous d-serine contributes to NMDA-evoked GABAergic IPSCs in Mb terminals. In the retina, d-serine is produced in Müller cells by SR (Kalbaugh et al. 2009; Stevens et al. 2003, 2010a) and broken down by d-amino acid oxidase (DAO) (Gustafson et al. 2007; Kalbaugh et al. 2009; Molla et al. 2006). Application of the SR inhibitor phenazine ethosulfate (10 μM) together with DAO (100 μg/ml) caused a small but significant reduction (Fig. 7B; 89.4 ± 2.7% of control, n = 5, P < 0.02, Student's t-test) of the NMDA-evoked GABAergic IPSCs in Mb terminals over 5–10 min of treatment. However, we could not achieve recovery by washing the slices with normal Ringer solution; therefore these results are indistinguishable from rundown.

The GABAergic reciprocal feedback IPSCs, evoked by depolarizing the Mb terminals from −60 to −20 mV for 100 ms, were also enhanced when d-serine was added to the Ringer solution (Fig. 7, Ci and Cii). The enhancement to 139.3 ± 17.7% (n = 5) of the control was statistically significant (P < 0.01, Student's t-test). Importantly, much like the effect of ALX 5407 on the reciprocal feedback inhibition (Fig. 4Dii), the exogenous d-serine-mediated increase of the reciprocal feedback (Fig. 7Ci) was not associated with increase in either I_Ca amplitude or (presynaptic) glutamate release as evidenced by membrane capacitance measurements (I_ca peak: 99.7 ± 1.1% of control, P < 1; ΔC_m: 97.4 ± 6.2% of control, P < 0.36, n = 5, Student's t-test), indicating that d-serine presumably acted on NMDARs postsynaptic to Mb terminals. It is noteworthy that d-serine exerted its effect primarily on the late components of reciprocal GABAergic IPSCs (Fig. 7, Ci and Cii), consistent with increasing NMDAR function on ACs participating in the reciprocal inhibition of Mbs (Vigh and von Gersdorff 2005). We tested whether attempts to acutely reduce retinal d-serine content could also reduce the GABAergic reciprocal feedback. Surprisingly, in our hands up to 10-min application of the SR inhibitor phenazine ethosulfate (10 μM) together with DAO (100 μg/ml) did not alter the reciprocal feedback inhibition of Mb terminals (Fig. 7D) (n = 3). To the contrary, application of phenazine ethosulfate has been found to reduce retinal d-serine content to 50% after 10-min exposure even without DAO and also inhibited light-evoked NMDAR signaling in GCs (Stevens et al. 2010a). In light of these findings, we attributed our negative results (Fig. 7D) to the short exposure of retina to phenazine ethosulfate and DAO. However, longer acute treatment during recording was associated with significant (>15%) rundown of I_Ca and/or ΔC_m. It should be noted that these recordings were made directly from presynaptic axon terminals that had survived axotomy. In this preparation the negative consequences of “washout” in whole cell voltage-clamp experiments during long recordings are intensified. Therefore, in the next set of experiments we incubated the retinas for 2 h in oxygenated Ringer solution containing phenazine ethosulfate (10 μM) and DAO (100 μg/ml). The preparation was exposed to phenazine ethosulfate and DAO throughout the experiment, including sectioning and recording, totaling at least 3 h of exposure for the cells reported here. Under these conditions, reciprocal I_GABA recorded from Mb terminals (Fig. 7E) was drastically reduced by the NMDAR specific blocker d-AP5, on average to 32.2 ± 12.0% of its control value (n = 4), leaving the transient AMPA-type receptor-mediated I_GABA (50 μM; Fig. 7E), similar to the effect of d-AP5 reported previously (Vigh and von Gersdorff 2005) and that of (R)-CPP shown in Fig. 5Ai. This finding clearly demonstrates the presence of a large NMDAR-mediated GABAergic feedback component even after long incubation with agents shown to reduce retinal d-serine levels in previous studies (Kalbaugh et al. 2009; Stevens et al. 2010a).

Together, these data showed that (exogenous) d-serine could increase NMDAR signaling and consequent transmitter release from GABAergic ACs mediating inhibitory feedback to Mb terminals. Nonetheless, in our hands, acute attempts to reduce endogenous d-serine content in the retina failed to eliminate NMDAR-mediated reciprocal inhibition to Mbs, although the observed small reduction in the NMDA-triggered I_GABA indicated that endogenous d-serine could contribute to NMDAR signaling that mediates lateral feedback to Mbs.

DISCUSSION

The role of NMDARs' coagonist binding site in mediating excitatory postsynaptic potentials (EPSPs) and GC signaling has been extensively studied in various structures of the brain (Mothet et al. 2000; Wilcox et al. 1996) as well as in amphibian (Gustafson et al. 2007; Lukasiewicz and Roeder 1995; Stevens et al. 2003, 2010a, 2010b) and mammalian (Daniels and Baldridge 2010; Kalbaugh et al. 2009; Reed et al. 2009; Sullivan et al. 2011; Sullivan and Miller 2012) retina preparations. The present study presents new information about the functional consequences of occupancy of NMDARs' coagonist binding site for inhibitory signaling in the retina and NMDAR-mediated GABA release from ACs. We studied the GABAergic inhibitory feedback from ACs to Mb axon terminals in the goldfish retina (Kaneko and Tachibana 1987), where the majority of GABAergic IPSCs have been shown to be triggered by activation of NMDARs (Fig. 5, Ai and Aii, Fig. 7E; Vigh and von Gersdorff 2005; Vigh et al. 2011).

The main finding of this study is that in the goldfish retinal slice preparation NMDAR activation-evoked GABA release to Mb terminals can be enhanced by indirect elevation of retinal glycine concentration through blocking GLYT1 or by (exogenous) d-serine, regardless of whether the NMDARs were activated by NMDA puff or synaptic glutamate released from Mb terminals. Together, these results indicate that the coagonist binding sites of NMDARs that contribute to the release of GABA onto Mb terminals are unsaturated. It is important to consider whether or not the nonsaturated state of the coagonist binding sites on ACs was a preparation-introduced artifact, as it is reasonable to believe that in slice preparations the washout of endogenous glycine and/or d-serine could result in an amount potentially lower than in the intact retina. If we assume that a synaptic site is sufficiently exposed for cotransmitters to be washed out, by the same token we might expect that the same synaptic sites would be more accessible for exogenous cotransmitters. We found that glycine alone could only increase reciprocal I_GABA recorded from Mb terminals in two of five cases. However, glycine more reliably increased NMDAR signaling at the Mb terminals' reciprocal synapses when GLYT1s were blocked (Fig. 4, C, Dii, and E). Importantly, the increase of reciprocal inhibition after GLYT1 block was rather similar in the presence and absence of exogenous glycine (130.0 ± 10.0% vs. 136.6 ± 8.1%, respectively; Fig. 4E), as well as in those two cases when glycine alone enhanced reciprocal feedback (143.7 ± 6.5%). Together, these results suggest that in our retinal slices 1) at AC→Mb synapses, GLYT1s control the saturation level of the NMDAR's coagonist site saturation, even in the presence of exogenous glycine, and 2) once GLYT1s are blocked, the saturation level of the coagonist binding sites cannot be increased further by exogenous glycine. This is in perfect agreement with the data showing that although global retinal glycine levels are high enough (Reed et al. 2010) to saturate the NMDARs (Johnson and Ascher 1987), high-affinity glycine uptake by GLYT1s can maintain a local environment, with low glycine concentration that keeps the coagonist sites unsaturated (Reed et al. 2009; Supplisson and Bergman 1997). Note that we found remarkably similar enhancement of I_GABA recorded from Mb terminals (∼30–40%) by exogenous d-serine, both comparable to that found for NMDAR-mediated light responses recorded from ON GCs in the mouse retina (Sullivan and Miller 2012). These observations are also in line with results obtained in isolated GlyT1^−/+ mouse retinas, where exogenous d-serine produced less increase in NMDA-evoked GC responses compared with wild-type retinas (Reed et al. 2009), indicating that reduced GLYT1 function can increase retinal glycine levels, which, in turn, can increase saturation of the coagonist binding site. These data suggest in extension that coagonist sites of NMDARs at some synapses in the retina are nonsaturated. In summary, although we cannot exclude the possibility that washout plays some role in maintaining a low level of occupancy at coagonist binding sites of NMDARs in slice preparations, we propose that the nonsaturated state is not a preparation artifact. It is important to note that our findings are different from what was reported for NMDARs expressed by ON GCs in the rat retina, where synaptically driven coagonist release saturated the coagonist sites (Kalbaugh et al. 2009). However, those experiments were also performed in retinal slices, supporting our conclusion that the slice preparation did not introduce substantial artifact.

The structure of the goldfish retina, and particularly the synaptic connections made by Mb axon terminals, has been studied in remarkably great detail (Witkovsky and Dowling 1969; Marc and Liu 2000; Studholme and Yazulla 1988). Mb terminals in the IPL were found to be surrounded almost entirely by pre- and postsynaptic GABAergic profiles (Marc and Liu 2000; Studholme and Yazulla 1988), in numbers estimated at 300–350 per Mb terminal (Witkovsky and Dowling 1969; Marc and Liu 2000). [³H]glycine-accumulating (glycinergic) AC profiles were not found to contact Mb terminals directly (Marc and Lam 1981), which remained consistent with the lack of glycine immunoreactivity in those AC terminals making synaptic contacts with Mb terminals (Studholme and Yazulla 1988). Additionally, the infrequent vesicle-free, GABA- electron lucent neural profiles adjacent to Mb ribbons were considered to be GC dendrites (Marc and Liu 2000). These utrastructural data are consistent with the results presented in this report showing the lack of a glycinergic component in the reciprocal feedback (Fig. 4, Ai and Aii) and with the conclusion of our earlier works characterizing the reciprocal and lateral feedback to Mb terminals (Vigh and von Gersdorff 2005; Vigh et al. 2011).

Glycinergic ACs have been shown to provide glycine to activate NMDARs on GC in the salamander retina (Lukasiewicz and Roeder 1995). Furthermore, the coagonist sites of NMDARs expressed by ON GCs appeared to be activated by synaptically released glycine (Kalbaugh et al. 2009). However, it is not known whether the glycine acting as coagonist at NMDARs is released directly into the synaptic cleft of excitatory (glutamatergic) synapses or reaches it by spillover from nearby glycinergic synapses. Our data show that reciprocal GABA release from ACs onto Mb terminals can be increased by a GLYT1 blocker (Fig. 4Dii). As glycinergic AC processes were not found adjacent to Mb terminals in electron microscopic studies (Marc and Liu 2000; Studholme and Yazulla 1988), we propose that glycine is not released directly into the synaptic cleft surrounding Mb terminals but reaches the coagonist binding sites of NMDARs expressed by GABAergic ACs forming reciprocal synapses with Mb terminals by spillover from nearby glycinergic inhibitory synapses, as is the case in the spinal cord (Ahmadi et al. 2003). Although glycine does not inhibit Mb terminals (Fig. 4, B and E) directly via transient ionotropic glycine receptors known to be expressed by BC terminals as seen in mouse retina (Wässle et al. 2009), our study proves that glycine can contribute to the inhibition of glutamate release from Mbs in a more powerful manner by increasing a more sustained, NMDAR-mediated GABA release from AC processes onto slowly desensitizing GABA_C receptors expressed on Mb terminals (Vigh and von Gersdorff 2005).

Extensive literature suggests that endogenous d-serine is critical for mediating NMDAR function at glutamatergic synapses in the CNS (Mothet et al. 2000; Panatier et al. 2006) including the (ON) BC to GC synapses in the retina (Kalbaugh et al. 2009; Sullivan et al. 2011; Sullivan and Miller 2010). This is strengthened by a recent report convincingly showing that AMPAR-dependent, rapid d-serine release from Müller cells provides complex regulation of synaptic NMDAR function in retinal GCs (Sullivan and Miller 2012). The proposed model implies that in the mammalian retina Müller cell processes are very close to both the presynaptic BC terminal and the postsynaptic site GC dendrite. This model is similar to the spatial arrangement found in the supraoptic nucleus, where glutamatergic, NMDAR-expressing synapses are surrounded by glial processes that can release d-serine to modulate NMDARs (Panatier et al. 2006). However, Müller cell processes were not found to make direct contacts with Mb terminals (Marc and Liu 2000), and in concert with this, glutamate clearance from the synaptic cleft is thought to be mediated primarily by glutamate transporters expressed by the presynaptic Mb terminals (Palmer et al. 2003b; Vandenbranden et al. 2000b). Therefore, a contribution of (Mb) activity-triggered rapid d-serine release from Müller cells to the reciprocal feedback is unlikely, and this is consistent with our results showing that acute attempts for d-serine degradation did not alter synaptically evoked responses (Fig. 7, D and E). Nonetheless, the presence of substantial NMDAR-dependent reciprocal I_GABA component after combined DAO-phenazine ethosulfate treatment (Fig. 7E) does not mean that endogenous d-serine does not play any role at Mb→AC synapses in mediating NMDAR signaling, but it suggests that glycine might substitute for d-serine. Namely, it has been proposed that the occupancy of the coagonist site might depend on the availability of the coagonist: if d-serine is not present (because its generation is blocked and/or it is broken down by DAO) the binding sites might be more accessible for glycine (Stevens et al. 2010b). Furthermore, transporter systems controlling retinal glycine and d-serine levels might be differently altered in slice preparations. In any case, our results suggest a less critical role for endogenous d-serine at the Mb→AC synapses that drive the reciprocal feedback inhibition of Mbs than what has been reported at BC→GC feedforward synapses in the SR knockout mouse retina, where the light-evoked NMDAR component was eliminated (Sullivan et al. 2011). Along the same lines, it is possible that light-stimulated endogenous d-serine levels in the synaptic cleft surrounding the Mb terminals could be higher in the intact retina than in the retinal slice, and so is the contribution of endogenous (background or spillover) d-serine to the NMDAR-mediated GABA release onto Mb terminals. Our results also show that endogenous d-serine might contribute to NMDA puff-evoked GABAergic feedback to Mbs. Focal NMDA application around the Mb terminals most likely triggers transmitter release from AC processes mediating both lateral and reciprocal feedback inhibition of the Mbs; thus these results suggest that endogenous d-serine might be more important in mediating NMDAR signaling in ACs providing lateral inhibition to the Mb terminal. This is in concert with observations made in both fish and rat retinas (Vigh et al. 2011 and Chavez et al. 2010, respectively) indicating that lateral and reciprocal inhibitory feedback circuits do not overlap or, in other words, they are mediated by distinct ACs. In extension, a potential difference in d-serine contribution to NMDAR signaling that mediates lateral vs. reciprocal feedback inhibition to Mb terminals suggests that NMDAR function might be regulated through the coagonist binding site in a synapse-specific manner.

Our data strongly support the view that NMDAR activity in retinal ACs can be modulated via regulation of the occupancy of their coagonist binding sites. Furthermore, our experiments prove that this modulation of input via NMDARs is strongly reflected in the neuronal output (i.e., transmitter release) of GABAergic ACs. Therefore, we propose that the saturation level of coagonist binding sites on NMDARs expressed by ACs is critical in determining the gain of the GABAergic signal transfer function in the inner retina. Besides the feedback inhibition to BC terminals exploited in our study, GABAergic ACs provide feedforward inhibition to GCs and other ACs (reviewed in Masland 1988; Wässle 2004). It is not known whether or not feedforward GABAergic inhibition depends on NMDARs and whether it can be modulated via their coagonist binding sites in particular. Nonetheless, the main excitatory input to GCs is provided by BCs. Therefore, it is likely that modulation of NMDAR-mediated feedback inhibition affects glutamate release from (ON) BCs via GABA_C receptors (Sagdullaev et al. 2011; Vigh and von Gersdorff 2005) and, in turn, excitation of postsynaptic GCs contributing to the retinal output.

GRANTS

This work was supported by a grant from National Eye Institute R01 EY-019051 (J. Vigh).

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

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