Different types of retinal inhibition have distinct neurotransmitter release properties

Abstract

Neurotransmitter release varies between neurons due to differences in presynaptic mechanisms such as Ca²⁺ sensitivity and timing. Retinal rod bipolar cells respond to brief dim illumination with prolonged glutamate release that is tuned by the differential release of GABA and glycine from amacrine cells in the inner retina. To test if differences among types of GABA and glycine release are due to inherent amacrine cell release properties, we directly activated amacrine cell neurotransmitter release by electrical stimulation. We found that the timing of electrically evoked inhibitory currents was inherently slow and that the timecourse of inhibition from slowest to fastest was GABA_C receptors > glycine receptors > GABA_A receptors. Deconvolution analysis showed that the distinct timing was due to differences in prolonged GABA and glycine release from amacrine cells. The timecourses of slow glycine release and GABA release onto GABA_C receptors were reduced by Ca²⁺ buffering with EGTA-AM and BAPTA-AM, but faster GABA release on GABA_A receptors was not, suggesting that release onto GABA_A receptors is tightly coupled to Ca²⁺. The differential timing of GABA release was detected from spiking amacrine cells and not nonspiking A17 amacrine cells that form a reciprocal synapse with rod bipolar cells. Our results indicate that release from amacrine cells is inherently asynchronous and that the source of nonreciprocal rod bipolar cell inhibition differs between GABA receptors. The slow, differential timecourse of inhibition may be a mechanism to match the prolonged rod bipolar cell glutamate release and provide a way to temporally tune information across retinal pathways.

neural systems rely on a proper balance between the timing of excitation and inhibition to function optimally (Buzsaki et al. 2007; Mann and Mody 2010; Scholl and Wehr 2008). In the retina, the excitatory pathway consists of light detection by photoreceptors that signal to bipolar cells and ganglion cells using glutamate (Masland 2001). The output of the excitatory pathway is modulated by inhibitory input from amacrine cells onto bipolar cells and ganglion cells in the inner retina (Masland 2001). GABAergic and glycinergic amacrine cells shorten the timecourse of bipolar cell glutamate release, which is essential to retinal temporal tuning (Dong and Werblin 1998). Bipolar cells use graded membrane depolarization to evoke sustained glutamate release (Heidelberger and Matthews 1992); therefore, the timing of amacrine cell-mediated inhibition must be prolonged to contribute to shaping bipolar cell output.

Amacrine cells provide inhibition to bipolar cell axon terminals by releasing GABA or glycine onto GABA_A, GABA_C, or glycine receptors (R) (Eggers and Lukasiewicz 2006a and 2006b; Euler and Masland 2000). The timing of light-evoked inhibitory postsynaptic currents (L-IPSCs) in bipolar cells is prolonged relative to an initial light stimulus (Eggers and Lukasiewicz 2006b). The timecourse of inhibition has been shown to vary due to the kinetics of each receptor type (Chang and Weiss 1999; Eggers and Lukasiewicz 2006b; Frech and Backus 2004), and due to prolonged amacrine cell neurotransmitter release (Eggers and Lukasiewicz 2006b; Gleason et al. 1993). However, it is unclear whether prolonged release from amacrine cells is a result of the time it takes to process the light signal upstream of amacrine cells at the photoreceptor-bipolar cell or bipolar cell-amacrine cell synapses or if it is due to an inherent mechanism of amacrine cell function.

Unlike the graded membrane depolarization used by bipolar cells, many amacrine cell types rely on action potentials to mediate transmitter release (Bloomfield 1992; Eggers et al. 2013; Shields and Lukasiewicz 2003; Taylor 1999). Spike-dependent neurotransmitter release is a fast synchronous event that generally involves the rapid increase of intracellular Ca²⁺ near release sites and subsequent neurotransmitter release to occur within 10–20 ms (Kaeser and Regehr 2013). However, previous studies have established that release from amacrine cells can occur asynchronously over several hundred milliseconds after a single stimulus (Borges et al. 1995; Eggers et al. 2013; Eggers and Lukasiewicz 2006b; Gleason et al. 1994). In contrast to the fast local rises in intracellular Ca²⁺ that mediate synchronous release, slow asynchronous release relies on global increases in intracellular Ca²⁺ far from the release site (Chung and Raingo 2013; Goda and Stevens 1994; Kaeser and Regehr 2013; Sakaba and Neher 2001; Scheuss et al. 2007). Amacrine cells may need to utilize asynchronous release in response to an action potential as a mechanism to match the timing of sustained bipolar cell glutamate release.

We have previously shown that asynchronous release after a single stimulus is a primary mechanism used by a population of spiking amacrine cells to release GABA onto rod bipolar cell GABA_CRs (Eggers et al. 2013). However, although rod bipolar cell L-IPSCs, mediated by GABA_CRs, GABA_ARs, and glycineRs are all prolonged relative to a brief light stimulus, the release kinetics of GABA and glycine underlying these light responses vary (Eggers and Lukasiewicz 2006b), and it is not known whether all amacrine cells use asynchronous release. Additionally, rod bipolar cells receive two types of inhibition: lateral inhibition from GABAergic and glycinergic amacrine cells that often have action potentials and reciprocal inhibition where rod bipolar cells activate a GABAergic amacrine cell called the A17 (Chavez et al. 2010; Chavez et al. 2006; Eggers and Lukasiewicz 2010; Grimes et al. 2009) that is nonspiking and uses Ca²⁺ entry through Ca²⁺-permeable α-amino-3-hydroxy-5-methyl-1-4-isoxazolepropionic acid (AMPA) receptors to activate GABA release back onto the rod bipolar cell (Chavez et al. 2006). Here, we ask if asynchronous release is the release mechanism inherent to all amacrine cells giving input to rod bipolar cell axon terminals by using an electrical stimulus to stimulate isolated input from many amacrine cell types and by electrically activating reciprocal feedback inhibition. In the present study, we show that the distinct timing of rod bipolar cell GABA_CR, glycineR, and GABA_AR inhibitory currents is mediated by the distinct timing of release from amacrine cells independent of photoreceptor or bipolar cell signaling. We present evidence that asynchronous release due to prolonged Ca²⁺ signaling is a primary mechanism of release from amacrine cells that synapse onto GABA_CRs and glycineRs on rod bipolar cells.

MATERIALS AND METHODS

Ethical approval.

Animal protocols were approved by the University of Arizona Institutional Animal Care and Use Committee and conformed to the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Retinal slice preparation.

As previously described (Eggers and Lukasiewicz, 2006), C57BL/6J male mice (Jackson Laboratories, Bar Harbor, ME) 35–60 days of age were killed using carbon dioxide. The eyes were enucleated, the cornea and lens were removed, and the eyecup was incubated in cold extracellular solution (see Solutions and drugs) with 800 U/ml of hyaluronidase for 20 min. The eyecup was washed with cold extracellular solution, and the retina was removed. The retina was trimmed into an approximate rectangle and mounted onto 0.45-μm nitrocellulose filter paper (Millipore, Billerica, MA). The filter paper containing the retina was transferred to a hand chopper. The filter paper was sliced into 250-μm thick slices, rotated 90°, and mounted onto glass cover slips using vacuum grease.

Solutions and drugs.

Extracellular solution used as a control bath for dissection and whole cell recordings was bubbled with a mixture of 95% O₂-5% CO₂ and contained (in mM): 125 NaCl, 2.5 KCl, 1 MgCl₂, 1.25 NaH₂PO₄, 20 glucose, 26 NaHCO₃, and 2 CaCl₂. The intracellular solution in the recording pipette used for monitoring electrically stimulated inhibitory rod bipolar cell currents contained (in mM): 120 CsOH, 120 gluconic acid, 1 MgCl₂, 10 HEPES, 10 EGTA, 10 TEA-Cl, 10 phosphocreatine-Na₂, 4 MgATP, 0.5 Na-GTP, and 50 μM Alexa Fluor 488 (Invitrogen, Carlsbad, CA) and was adjusted to pH 7.2 with CsOH. The intracellular solution used for monitoring electrically stimulated amacrine cell depolarization contained (in mM): 120 KOH, 120 gluconic acid, 1 MgCl₂, 10 HEPES, 10 EGTA, 10 TEA-Cl, 10 phosphocreatine-Na₂, 4 MgATP, 0.5 Na-GTP, and 50 μM Alexa Fluor 488 (Invitrogen) and was adjusted to pH 7.2 with CsOH. The intracellular solution used for monitoring reciprocal feedback inhibitory currents contained (in mM): 120 CsCl, 1 MgCl₂, 10 HEPES, 0.1 EGTA, 10 TEA-Cl, 10 phosphocreatine-Na₂, 4 MgATP, 0.5 Na-GTP, and 50 μM Alexa Fluor 488 and was adjusted to pH 7.2 with CsOH. With these concentrations the driving force for Cl⁻ was calculated as 60 mV in all solutions.

Antagonists were used to isolate receptor input. Strychnine (500 nM-1 μM) was used to block glycineRs, 20 μM SR-95531 was used to block GABA_ARs, and 50 μM 1,2,5,6-(tetrahydropyridin-4-yl)methylphosphinic acid hydrate (TPMPA) was used to block GABA_CRs. Tetrodotoxin (TTX, 500 Nm; Alomone Labs, Jerusalem, Israel) was used to block voltage-gated Na⁺ channels. EGTA-AM (50 μM; Invitrogen) and BAPTA-AM (50 μM; Invitrogen) were used to increase intracellular Ca²⁺ buffering. EGTA-AM and BAPTA-AM were applied to the bath for 5 min before the experiment and were present during the recordings for a total of 10 min. All antagonists were applied to the slice by a gravity-driven superfusion system (Cell Microcontrols, Norfolk, VA) at a rate of ∼1 ml/min. Chemicals were purchased from Sigma-Aldrich (St. Louis, MO), unless otherwise indicated.

Whole cell recordings.

All recordings were performed in light-adapted conditions. Retinal slices on glass cover slips were placed in a custom chamber and heated to 32° by temperature-controlled thin-stage and inline heaters (Cell Microcontrols). Whole cell voltage clamp recordings from rod bipolar cells in retinal slices were made as previously described (Eggers and Lukasiewicz 2006b). Electrically evoked inhibitory postsynaptic currents (eIPSCs) were recorded from rod bipolar cells clamped at 0 mV, the reversal potential for currents mediated by nonselective cation channels. Electrodes were pulled from borosilicate glass (World Precision Instruments, Sarasota, FL) using a P97 Flaming/Brown puller (Sutter Instruments, Novato, CA) and had resistances of 5–7 MΩ. Liquid junction potentials of 20 mV were corrected for before recording.

For electrically evoked rod bipolar cell current responses, rod bipolar cells with intact axon terminals were identified morphologically by exciting the dye in the intracellular solution using a fluorescence microscope. A 1-ms, 4- to 20-μA stimulus was delivered through a stimulating pipette placed near rod bipolar cell axon terminals by an S48 stimulator (Grass, Warwick, RI) with attached PSIU6 photoelectric isolation unit (Grass). The stimulus was delivered every 60 s. Electrically evoked amacrine cell voltage responses were recorded in current clamp mode in response to the same electrical stimulus from a stimulating electrode located near a fluorescently identified amacrine cell process in the distal inner plexiform layer, where rod bipolar cell terminals are located. Electrically evoked responses were filtered at 5 kHz using a four-pole low-pass Bessel filter on an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA). The response was digitized at 10 kHz using a Digidata 1440A data acquisition system (Molecular Devices) and Clampex software (Molecular Devices). Reciprocal IPSCs were elicited every 45–60 s by a 10-ms step depolarization from a holding potential of −60 to −20 mV. Confirmation of rod bipolar cell (Ghosh et al. 2004) or amacrine cell morphology was done at the end of each recording using an Intensilight fluorescence lamp and Digitalsight camera controlled by Elements software (Nikon Instruments, Tokyo, Japan).

Data analysis and statistics.

Traces of evoked and reciprocal IPSCs and evoked excitatory postsynaptic potentials for each condition were averaged using Clampfit (Molecular Devices). The peak amplitude, charge transfer (Q), time to peak and decay to 37% of the peak (D37) were determined. The timecourse and amount of transmitter release mediating evoked and reciprocal IPSCs were calculated with custom written Matlab software. Release functions were calculated by convolution analysis (Diamond and Jahr 1995) using the relationship:

$$eIPSC\left(t\right)=release\left(t\right)\otimes \hspace{0.17em}\mathrm{s}IPSC\left(t\right),$$ (1)

such that

$$release\left(t\right)=F^{-1}\frac{F\left[\text{eIPSC}\left(t\right)\right]}{\text{F}\left[\text{sIPSC}\left(t\right)\right]},$$ (2)

where sIPSC(t) is the average spontaneous GABA_CR, GABA_AR, or glycineR-mediated IPSC, and F and F⁻¹ represent the Fourier transform and inverse Fourier transform of the function, respectively. Briefly, software data processing occurred as follows: data were down sampled from 10 to 1 kHz and smoothed using a moving average filter (20 points). The software calculated vesicle release from current-time curves by performing deconvolution of the average rod bipolar cell GABA_AR-, GABA_CR-, or glycineR-mediated current evoked by a single vesicle release [sIPSC(t)] (Eggers and Lukasiewicz 2006b). The deconvolution was performed by dividing the fast-Fourier transform (FFT) of the single vesicle release into the FFT of the electrically evoked response [eIPSC(t)] and again smoothed similar to the initial current trace. The resulting release traces were further analyzed in Clampfit to determine the number of vesicles released and the D37. The area under the curve was used to determine the amount of vesicle release. Release occurred in early and late phases. The early phase of release was determined to be the first 100 ms after the initial stimulus because it included the peak for release onto all receptor inputs. The late phase was determined to be the remaining release that occurred from 100 ms until the response returned to baseline.

Paired Student's t-tests were used to compare values between conditions for the same cell. For each cell, a normalized data value of the percent of total GABA_AR-, GABA_CR-, or glycineR-mediated eIPSC that remained after the drug treatment was calculated. Nonpaired Student's t-tests were used to compare these values across two groups of cells. For comparisons of three or more groups of cells, these values were compared across drug conditions with ANOVA tests, using the Student-Newman-Keuls (SNK) method for pairwise comparisons. Differences were considered significant when P ≤ 0.05. All data are reported as means ± SE.

RESULTS

Release from GABAergic and glycinergic amacrine cells occurs with prolonged but distinct timing.

Light-evoked GABA and glycine release have been shown to occur with prolonged, but distinct, timing and to mediate the sustained timecourse of GABAergic and glycinergic L-IPSCs (Eggers and Lukasiewicz 2006b). Previous work has shown that some amacrine cells use asynchronous release as a primary mechanism to release GABA onto GABA_CRs (Eggers et al. 2013). It is not known whether GABAergic amacrine cells also use asynchronous release to release GABA onto GABA_ARs or if glycinergic amacrine cells use asynchronous release. To determine whether prolonged GABA and glycine release is an inherent characteristic of amacrine cells, we isolated the amacrine cell inputs to rod bipolar cells using an electrical stimulus at the amacrine cell-rod bipolar cell synapse in the inner plexiform layer in the presence of antagonists to isolate GABA_AR-, GABA_CR-, or glycineR-mediated inputs (see materials and methods). We previously determined that this electrical stimulus activates isolated amacrine cell input to rod bipolar cells because the response was not reduced by blocking glutamate receptors with 6-cyano-7-nitroquinoxaline-2,3-dione (Eggers et al. 2013). This electrical stimulus evoked a brief depolarization (D37 = 1 ± 0.1 ms, n = 6) in recorded amacrine cells (see materials and methods), suggesting that any prolonged signals are due to inherent release properties of the amacrine cell and not prolonged depolarization.

eIPSCs mediated by GABA_ARs and glycineRs had a prolonged response that lasted much longer than the 1-ms stimulus (Fig. 1A), similar to eIPSCs of GABA_CRs (Eggers et al. 2013). However, there were significant timing differences between the three inputs. GABA_CR eIPSCs had the slowest decay time (D37) followed by glycineR eIPSCs and GABA_AR eIPSCs (P < 0.001 ANOVA, SNK post hoc P < 0.05; Table 1 and Fig. 1B). The D37 times reported here follow the same hierarchy from the slowest time to fastest time as L-IPSCs measured in a previous study: GABA_CRs > glycineRs > GABA_ARs (Eggers and Lukasiewicz 2006b). To determine if differences in the eIPSCs were due to variability in stimulus electrode placement, we measured eIPSCs of GABAR and glycineR inputs from the same rod bipolar cell while sequentially applying GABA_CR, GABA_AR, and glycineR antagonists. We found that the D37 times followed the same hierarchy as isolated receptor inputs compiled from multiple rod bipolar cells with GABA_CR eIPSCs having the slowest D37 and GABA_AR eIPSCs having the fastest D37 (GABA_CR: 2,365 ± 208 ms > glycineRs: 1,307 ± 235 ms > GABA_ARs: 258 ± 97 ms; n = 3, P < 0.001 ANOVA, SNK post hoc P < 0.01). This suggests that, although the electrical stimulus evokes long-lasting IPSCs, it does not affect the relative timing of GABA_CR, GABA_AR, and glycineR inputs.

Fig. 1.

Electrically evoked (e) inhibitory postsynaptic currents (IPSCs) of rod bipolar cells are prolonged by slow GABA and glycine release. A: eIPSCs mediated by GABA_C receptors (R), glycineRs, and GABA_ARs measured from rod bipolar cells were prolonged relative to the stimulus (1 ms, gray bar, stimulus artifact removed) similar to light-evoked IPSCs. eIPSCs shown were normalized to the peak of the response to emphasize the timecourse differences. B: GABA_CR eIPSCs had an average time to peak and decay to 37% of the peak (D37) significantly longer than the average D37 for glycineRs [1-way ANOVA P < 0.05, Student-Newman-Keuls (SNK) post hoc P < 0.05] and GABA_ARs (P < 0.01). C: GABA and glycine release was estimated from eIPSCs with the deconvolution of an average GABA_CR, glycineR, or GABA_AR spontaneous IPSC (sIPSC) from rod bipolar cells. Traces are shown normalized to the peak. D: the timecourse of release onto GABA_CRs and glycineRs was similar and significantly longer than release onto GABA_ARs (SNK post hoc P < 0.05). *P < 0.05 and **P < 0.01.

Table 1.

Timecourse of eIPSCs and estimated transmitter release

eIPSC Type	eIPSC D37, ms	Release D37, ms	n
GABACR	1,410 ± 116***	907 ± 108*	50
GlycineR	717 ± 164*†	878 ± 233*	13
GABAAR	73.4 ± 33.5†	288 ± 154†	15

Data are average values of decay to 37% of peak (D37) of electrically evoked inhibitory postsynaptic currents (eIPSCs) and estimated vesicle release.

n, No. of cells; R, receptor.

^*P < 0.05 and

^***P < 0.001 for statistical comparisons with GABA_AR and

^†P < 0.05 for comparisons with GABA_CR.

The decays of spontaneous IPSCs mediated by GABA_ARs (2.0 ± 0.5 ms), GABA_CRs (34.1 ± 2.1 ms), and glycineRs (3.6 ± 0.5 ms) (Eggers and Lukasiewicz 2006b) that measure individual receptor kinetics are also different and share a similar relationship to the eIPSC and L-IPSC kinetics. However, the sIPSC decays are much briefer than the eIPSC D37s, in contrast to other synapses where the decay times of spontaneous and electrically evoked currents are similar (Lu and Trussell 2000). This suggests that, similar to GABA release onto GABA_CRs, prolonged GABA and glycine release likely contributes to the prolonged timecourse of GABA_AR and glycineR eIPSCs.

To test this, we used deconvolution analysis (Diamond and Jahr 1995; Eggers et al. 2013; Eggers and Lukasiewicz 2006b) to estimate the timecourse of GABA and glycine release that underlies the eIPSCs. This analysis assumes that vesicles released are linearly summed by postsynaptic receptors, which may not be the case. This analysis then represents the minimum release required to create the eIPSCs measured. We found that the timecourse of GABA and glycine release was prolonged relative to the stimulus (Fig. 1C). We expected the timing of amacrine cell transmitter release to follow the same hierarchy as that observed for the eIPSC D37. Surprisingly, the D37 of GABA release onto GABA_CRs and glycine release were similar (P = 0.9; Table 1 and Fig. 1D), despite the difference in GABA_CR vs. glycineR eIPSCs. This is likely because the slow kinetics of the GABA_CRs prolong the timecourse of the eIPSC in response to the same release timecourse (Eggers et al. 2013; Eggers and Lukasiewicz 2006a). The D37 of GABA release onto GABA_ARs was significantly faster than that of GABA_CRs and glycineRs (P < 0.05 ANOVA, SNK post hoc P < 0.05 Table 1 and Fig. 1D) so the hierarchy of the timecourse of transmitter release was GABA_CRs = glycineRs > GABA_ARs. This suggests that GABA release onto GABA_ARs may originate from a different population of amacrine cells than GABA release onto GABA_CRs. This also suggests that GABAergic amacrine cells that release onto GABA_CRs and glycinergic amacrine cells may share similar mechanisms that determine the timecourse of release. Although the timecourse of release onto GABA_ARs was faster than that onto GABA_CRs or glycineRs, all were consistent with the timing of asynchronous release that occurs over several hundred milliseconds (Chung and Raingo 2013; Goda and Stevens 1994; Kaeser and Regehr 2013; Sakaba and Neher 2001; Scheuss et al. 2007).

Slow Ca²⁺ buffering does not decrease GABA release onto GABA_ARs.

These data suggest that transmitter release from amacrine cells that mediates light-evoked (Eggers and Lukasiewicz 2006a) and electrically evoked IPSCs (Fig. 1) may be asynchronous release. Similar to synchronous release, asynchronous release is regulated by changes in intracellular Ca²⁺ concentrations. However, asynchronous release often occurs in response to tetanic stimulation that leads to a global accumulation of intracellular Ca²⁺ (Chung and Raingo 2013; Kaeser and Regehr 2013; Sakaba and Neher 2001; Scheuss et al. 2007). Asynchronous release is susceptible to Ca²⁺ buffering by the slow Ca²⁺ chelator EGTA because of its dependence on global changes in intracellular Ca²⁺. EGTA primarily affects asynchronous release while sparing fast synchronous release that occurs rapidly and in response to an increase in Ca²⁺ near Ca²⁺ channels and release sites (Augustine et al. 2003; Chung and Raingo 2013; Hefft and Jonas 2005; Kaeser and Regehr 2013; Neher 1998; Rozov et al. 2001; Sakaba and Neher 2001).

The timecourse of amacrine cell GABA release onto GABA_ARs (Fig. 1) is slow, lasting for hundreds of milliseconds, which is characteristic of asynchronous release. To test if amacrine cell release onto GABA_ARs occurs asynchronously, we used EGTA-AM, the membrane permeant analog of the chelator, to limit asynchronous release. We added low concentrations (50 μM) of EGTA-AM to the bath, and 10 mM EGTA was in the recording pipette so that EGTA affected only the presynaptic amacrine cell terminal. If prolonged GABA release onto GABA_ARs is occurring asynchronously, we would expect EGTA-AM to reduce the GABA_AR-mediated input by reducing GABA release and shortening the decay time.

As shown in Fig. 2A, EGTA-AM did not reduce GABA_AR-mediated eIPSCs. There was no significant change in the Q, D37, or peak (Table 2 and Fig. 2, C–E), although on average the Q and D37 of GABA_AR eIPSCs were increased compared with control. Although the underlying GABA release onto GABA_ARs occurred with prolonged timing (125 ± 33 ms) that is characteristic of asynchronous release (Fig. 2F), EGTA-AM did not reduce the timing, peak, or the amount (Table 3 and Fig. 2, H–J) of release. In comparison, EGTA-AM reduced the Q, peak amplitude, and D37 (Table 2 and Fig. 2, C–E) of GABA_CR eIPSCs (Fig. 2B) by shortening the timecourse and limiting the amount of GABA release (Table 3 and Fig. 2, G–J) (Eggers et al. 2013).

Fig. 2.

GABA release onto GABA_ARs is not reduced by Ca²⁺ buffers. Representative traces of GABA_AR (A) and GABA_CR (B) eIPSCs before and after treatment with 50 μM EGTA-AM. EGTA-AM (EG) and BAPTA-AM (BA) reduced the charge transfer (Q; C), shortened the D37 (D), and reduced the peak (E) of GABA_CR, but not GABA_AR eIPSCs. Representative traces of the timecourse of GABA release estimated through deconvolution onto GABA_AR (F) and GABA_CRs (G) are shown. The amount (H), timing (I), and peak (J) of GABA release onto GABA_ARs after treatment with EGTA-AM or BAPTA-AM were similar to control values. In contrast, both Ca²⁺ buffers significantly reduced the timing and amount of GABA release onto GABA_CRs, and BAPTA-AM reduced the peak. All values are normalized to the control GABA_AR or GABA_CR eIPSC (C–E) or GABA release (H–J) for each cell. Gray bar = 1-ms electrical stimulus. *P < 0.05 and **P < 0.01.

Table 2.

Average values for eIPSCs recorded in the presence of EGTA-AM or BAPTA-AM (normalized to %respective eIPSC control values)

eIPSC Type	Q (norm)	D37 (norm)	Peak (norm)	n
GABACR 50 EGTA-AM	33.6 ± 6.7***	57.3 ± 5.8***	58.2 ± 8.1**	7
GABACR BAPTA-AM	31.6 ± 5.2***	50.5 ± 4.0***	53.5 ± 8.3***	11
GlycineR 50 EGTA-AM	33.3 ± 4.8***	47.4 ± 9.1**	68.2 ± 5.7**	5
GlycineR BAPTA-AM	47.1 ± 12.6**	42.2 ± 10.7**	80 ± 11.8 (P = 0.2)	6
GABAAR 50 EGTA-AM	235 ± 66.4 (P = 0.09)	211.2 ± 62.1 (P = 0.13)	85.1 ± 6.2 (P = 0.06)	6
GABAAR BAPTA-AM	173 ± 68.6 (P = 0.34)	209.1 ± 92.3 (P = 0.3)	84.8 ± 14.5 (P = 0.4)	5

Data are average values of charge transfer (Q), D37, and peak amplitude of eIPSCs in the presence of the drug shown in the left column, normalized (norm) to the control value for each cell.

n, No. of cells; 50, 50 μM.

GABA_CR data taken from Eggers et al. 2013 are shown for comparison.

^**P < 0.01 and

^***P < 0.001 compared with respective controls.

Table 3.

Average values for release from deconvolution of eIPSCs recorded in the presence of EGTA-AM or BAPTA-AM (normalized to %respective control values)

	Q (no. of vesicles, norm)	D37 (norm)	Peak (norm)	n
GABACR 50 EGTA-AM	56.6 ± 14.7*	50.4 ± 11.5**	65.4 ± 15.0	7
GABACR BAPTA-AM	37.0 ± 5.0***	42.8 ± 9.7***	62.1 ± 12.5*	11
GlycineR 50 EGTA-AM	47.1 ± 5.6***	58 ± 11.6*	77.2 ± 15.7 (P = 0.2)	6
GlycineR BAPTA-AM	33.8 ± 14.2**	37.5 ± 14.1**	54.7 ± 16.4*	5
GABAAR 50 EGTA-AM	121 ± 24 (P = 0.4)	144 ± 30 (P = 0.2)	140 ± 34 (P = 0.3)	6
GABAAR BAPTA-AM	162 ± 82 (P = 0.6)	158.2 ± 37.4 (P = 0.2)	224 ± 75 (P = 0.2)	5

Data are average values of the no. of vesicles (Q), D37, and peak amplitude of release in the presence of the drug shown in the left column, normalized (norm) to the control value for each cell.

n, No. of cells.

GABA_CR data taken from Eggers et al. 2013 are shown for comparison.

^*P ≤ 0.05,

^**P ≤ 0.01, and

^***P < 0.001 compared with respective controls.

If GABA is released from the same population of amacrine cells, then GABA release from GABAergic amacrine cells should occur by a similar mechanism. However, our data indicate that GABA release onto rod bipolar cell GABA_ARs may occur by a mechanism other than the asynchronous release, dependent on prolonged Ca²⁺ signals, that is observed for GABA_CRs (Eggers et al. 2013). Due to the faster timecourse of the GABA release that underlies the GABA_AR-mediated eIPSCs, it is possible that synchronous release of GABA onto GABA_ARs accounts for the majority of the release. Because EGTA-AM is a Ca²⁺ chelator with slow kinetics (Augustine et al. 1991; Tymianski et al. 1994) it cannot act quickly enough to limit synchronous release. Therefore, we tested if GABA release onto GABA_ARs could be reduced by BAPTA-AM, a membrane-permeant analog Ca²⁺ chelator with fast kinetics (Tymianski et al. 1994). BAPTA-AM did not reduce the Q, D37, or peak (Table 2 and Fig. 2, C–E) of GABA_AR-mediated eIPSCs, although on average the value was increased but not significantly different. The amount, timing, and peak (Table 3 and Fig. 2, H–J) of GABA release onto GABA_ARs were also not reduced in the presence of BAPTA-AM, in contrast with the effects on release onto GABA_CRs. There was no significant difference between the effect of EGTA-AM or BAPTA-AM on release onto GABA_ARs (current Q P = 0.5, D37 P = 1, peak P = 1; release Q P = 0.7, D37 P = 0.8, peak P = 0.4). GABA release onto GABA_ARs is prolonged, and the timing is characteristic of asynchronous release, but it is not susceptible to Ca²⁺ buffering by the fast Ca²⁺ chelator BAPTA-AM or the slow Ca²⁺ chelator EGTA-AM, suggesting that release is tightly coupled to Ca²⁺ (Augustine et al. 2003). As we have isolated the amacrine cell-rod bipolar cell synapse using a direct electrical stimulus, the differences in the effects of Ca²⁺ buffering on GABA release onto GABA_ARs and GABA_CRs is not a result of upstream signaling from photoreceptors or bipolar cells. Thus, amacrine cell GABA release onto GABA_ARs and GABA_CRs occurs with distinct timing that may reflect GABA release from two different populations of amacrine cells.

Slow Ca²⁺ buffering reduces glycine release from amacrine cells.

Our results indicate that the timecourse of evoked GABA release onto GABA_CRs and glycine release is similar (Fig. 1D). Release of glycine from amacrine cells in response to glutamate puffs in the inner plexiform layer was shown to be enhanced by Ca²⁺-induced Ca²⁺ release (Chavez and Diamond 2008) that mediates asynchronous GABA release (Eggers et al. 2013), suggesting that glycine release may be mediated by slow Ca²⁺ signaling as well. We tested whether glycine release from amacrine cells could be reduced by slow Ca²⁺ buffering by recording glycineR eIPSCs in the presence of EGTA-AM. We found that EGTA-AM reduced the eIPSC Q, D37, and peak amplitude (Table 2 and Fig. 3, A–D). Deconvolution analysis showed that EGTA-AM also reduced the timing and amount of vesicle release (Table 3 and Fig. 3, E–H). These results show that glycine release from amacrine cells onto rod bipolar cell terminals occurs asynchronously.

Fig. 3.

Glycine release from amacrine cells is reduced by Ca²⁺ buffering. Representative traces of glycineR eIPSCs (A) and timecourse of glycine release (E) recorded before and after treatment with EGTA-AM are shown. Increased Ca²⁺ buffering with EGTA-AM or BAPTA-AM reduced Q (B), shortened D37 (C) of glycineR-mediated eIPSCs, and EGTA-AM reduced the peak (D). EGTA-AM and BAPTA-AM also reduced the amount (F) and D37 (G) of vesicle release onto glycineRs, and BAPTA-AM reduced the peak (H). All values are normalized to the glycineR eIPSCs (B–D) or the release onto glycineRs (F–H) for each cell. Gray bar = 1-ms electrical stimulus. *P < 0.05 and **P < 0.01.

We tested whether any additional glycine release from amacrine cells could be reduced by fast Ca²⁺ buffering with BAPTA-AM. BAPTA-AM reduced the glycineR eIPSC Q, D37, and peak by similar amounts as EGTA-AM with no further decrease (EGTA-AM vs. BAPTA-AM Q P = 0.3; D37 P = 0.7; peak = 0.4; Fig. 3, B–D). Consistent with these results, deconvolution analysis showed that BAPTA-AM also reduced the total, peak, and timing of vesicle release (Fig. 3, F–H). Interestingly, BAPTA-AM, but not EGTA-AM, reduced the peak value (Fig. 3H) of transmitter release. These results indicate that much of amacrine cell glycine release onto rod bipolar cell terminals is asynchronous with some synchronous release that can be reduced by fast Ca²⁺ buffering. This is similar to the results of GABA release onto GABA_CRs. It is likely that amacrine cell glycine release and GABA release onto GABA_CRs share similar mechanisms, and, although GABA release onto GABA_ARs still occurs more slowly than you would expect for purely synchronous release, it may require higher intracellular Ca²⁺, or those neurons have higher Ca²⁺-buffering capacities that our addition of EGTA-AM or BAPTA-AM did not disturb.

GABA release onto GABA_ARs occurs in two phases.

Deconvolution analysis suggested that GABA release onto GABA_ARs occurred in two phases, an early fast phase represented by a transient spike of release occurring immediately after the stimulus followed by a later slow sustained phase of vesicle release (Fig. 2F). A fast and slow phase of GABA release onto GABA_ARs was found in previous studies of depolarization-induced transmission between cultured chick retinal amacrine cells (Borges et al. 1995). We determined the number of vesicles that were released in each phase relative to total vesicle release. We defined the fast phase as release that occurred 100 ms after the stimulus because this timing included the peak of release onto each receptor input. The slow phase was defined as the amount of vesicle release remaining between 100 ms and the return of the response to baseline (see materials and methods). We found that on average 55 ± 12 vesicles (∼33% of all vesicles) were released onto GABA_ARs during the fast phase (Fig. 4A). We compared the number of vesicles released onto GABA_ARs during both phases with that released onto GABA_CRs and glycineRs during the early (first 100 ms) and late (100 ms, baseline return) time periods. During the early time period of release, amacrine cells released significantly fewer vesicles onto GABA_CRs (7 ± 1%, ANOVA P < 0.001, SNK post hoc P < 0.001) and glycineRs (9 ± 2%, P < 0.001) compared with release onto GABA_ARs (Fig. 4A). Most of the vesicle release onto GABA_CRs and glycineRs occurred during the later time period (Fig. 4B). Although most of the GABA release onto GABA_ARs occurred during the fast phase, a large proportion of the release also occurred during the slow phase. At conventional synapses, much of synchronous release occurs within 10–20 ms. In our experiments, only 0.19% of total vesicle release onto GABA_ARs, 0.07% of glycine release, and 3.3% of release onto GABA_CRs (Eggers et al. 2013) occurred within 10 ms of the stimulus. These results support the hypothesis that amacrine cells predominantly use asynchronous release to release onto glycineRs and GABA_CRs, and may rely on a distinct mode of release onto GABA_ARs.

Fig. 4.

GABA release onto GABA_ARs occurs in the fast and slow phases of release. The amount of vesicle release that occurred during early (A) and late (B) time periods normalized to the total amount of vesicle release is shown for release onto GABA_CRs, glycineRs, and GABA_ARs. An example trace of vesicle release onto GABA_ARs with the fast and slow phases separated (gray dotted line) is shown in the inset. Vesicle release onto GABA_CRs and glycineRs occurred primarily during the late time period (B, GABA_C = 93 ± 1%, n = 18; glycine = 91 ± 2%, n = 11) compared with release onto GABA_ARs (66 ± 8%, n = 11, 1-way ANOVA P < 0.01, SNK post hoc P < 0.01); 33 ± 8% of vesicle release onto GABA_ARs occurred during the early time period (A). **P < 0.01.

Electrical stimuli preferentially activate spiking GABAergic amacrine cells.

Bipolar cells receive inhibitory input from slowly depolarizing (nonspiking) and spiking (Bloomfield 1992; Hartveit 1999; Ivanova et al. 2006; Taylor 1999) amacrine cells. Some of the inhibitory input arises from a reciprocal rod bipolar cell-A17 amacrine cell circuit, where nonspiking A17 amacrine cells require direct activation by rod bipolar cell glutamate release (Chavez et al. 2006; Grimes et al. 2009; Hartveit 1999; Menger and Wässle 2000; Nelson and Kolb 1985; Singer and Diamond 2003). The remaining rod bipolar cell inhibition is mediated by nonreciprocal inhibition mediated by amacrine cells activated by cone bipolar cells (Chavez and Diamond 2008; Cui et al. 2003; Ivanova et al. 2006). It is likely that an electrical stimulus will activate multiple amacrine cell inputs. Previous studies have shown that nonreciprocal inhibition is due in part to spiking amacrine cells (Chavez and Diamond 2008; Chavez et al. 2010; Cui et al. 2003; Lukasiewicz and Shields 1998) and that using an electrical stimulus in the outer plexiform layer to activate bipolar cell inhibition predominantly recruits spiking amacrine cells (Shields and Lukasiewicz 2003). It is possible that asynchronous release may be used primarily by spiking amacrine cells, since it could be a way to match timing with the graded activation of bipolar cells.

We previously showed that the asynchronous release onto GABA_CRs that underlies GABA_CR eIPSCs was due to the activation of spiking amacrine cells, since GABA_CR eIPSCs were blocked by 500 nM TTX (Eggers et al. 2013). Here, we have extended these studies to test if eIPSCs mediated by GABA_ARs and glycineRs were also primarily due to input from spiking amacrine cells. We found that blocking action potentials with 500 nM TTX eliminated GABA_AR eIPSCs similar to the significant decrease seen with TTX on GABA_CR eIPSCs (Fig. 5, A and B) (Eggers et al. 2013). These results suggest that our electrical stimulus is primarily activating nonreciprocal connections between populations of spiking GABAergic amacrine cells and rod bipolar cells. However, blocking action potentials did not decrease glycineR eIPSCs (Fig. 5C). On average, TTX actually increased the glycineR eIPSC Q, potentially due to actions on inputs to glycinergic amacrine cells (Chavez and Diamond 2008), although the difference was not significant (P = 0.2). There was also no significant decrease in the peak (120 ± 40% of control, P = 0.5) or D37 (150 ± 40% of control, P = 0.2). This suggests that nonspiking glycinergic amacrine cells are activated by our electrical stimulus and that both spiking and nonspiking amacrine cells demonstrate asynchronous release.

Fig. 5.

Tetrodotoxin (TTX)-sensitive GABAergic amacrine cells are activated in response to an electrical stimulus. Representative traces of rod bipolar cell eIPSCs recorded in the presence of TTX are shown for GABA_AR (A), GABA_CR (B), and glycineRs (C). D: Q of GABA_AR (Student's t-test, P < 0.01, n = 4)- and GABA_CR (P < 0.01, n = 5)-mediated eIPSCs are attenuated by TTX. TTX did not significantly change glycineRs (P = 0.2, n = 5) eIPSCs. All values are normalized to the control GABA_AR (A), GABA_CR (B), or glycineR (C) current for each cell. GABA_CR eIPSCs modified from Eggers et al., 2013. Gray bar = 1-ms electrical stimulus. **P < 0.01.

Timing of GABA release onto GABA_ARs and GABA_CRs at the rod bipolar cell-A17 reciprocal synapse.

Our data suggest that the distinct timecourse of GABA release onto GABA_ARs and GABA_CRs comes from different amacrine cells. To test this idea, we examined release at the reciprocal rod bipolar cell circuit, where both GABAR types are activated by one type of presynaptic amacrine cell called the A17 (Chavez et al. 2010; Chavez et al. 2006; Eggers and Lukasiewicz 2010; Grimes et al. 2009) that relies on Ca²⁺ entry through Ca²⁺-permeable AMPA receptors to activate GABA release (Chavez et al. 2006). This reciprocal feedback inhibition will not be activated by our electrical stimulus as we show in Fig. 5 that TTX blocks electrically stimulated GABAergic inhibition to rod bipolar cells and previous reports showed that the A17 amacrine cell is nonspiking (Chavez et al. 2006; Grimes et al. 2009; Hartveit 1999; Menger and Wässle 2000; Nelson and Kolb 1985; Singer and Diamond 2003) and that the A17 feedback inhibition is not affected by TTX (Chavez et al. 2006). Therefore, we asked if the distinct timing of GABA release also occurs at this reciprocal synapse where both receptor types are activated by the same presynaptic amacrine cell.

We used an electrode containing a high-Cl⁻ intracellular solution (see materials and methods) to deliver a brief (10 ms) depolarizing step from −60 to −20 mV to a rod bipolar cell to activate feedback inhibitory currents (fIPSCs). This caused an initial transient inward Ca²⁺ current (Hartveit 1999) followed by a more sustained inward inhibitory current that lasted several hundred milliseconds after the initial 10-ms step (Fig. 6A). The timing of fIPSCs was prolonged and had an average D37 of 132 ± 28 ms (n = 5), which is considerably faster than that detected in response to an electrical (Fig. 1 and Eggers et al. 2013) or light (Eggers and Lukasiewicz 2006b) stimulus. When TPMPA was added to the bath to block GABA_CR input, the remaining GABA_AR fIPSC (Fig. 6A) was more transient, and the D37 on average was 32 ± 13 ms (n = 5, Fig. 6C). GABA_CR input (Fig. 6A) was determined by subtracting the GABA_AR fIPSC from the total response. The D37 of the GABA_CR fIPSC was 191 ± 58 ms (n = 5) and significantly longer than that of the GABA_AR fIPSC (P < 0.05, Fig. 6C). Deconvolution analysis (Fig. 6B) indicated that GABA release underlying the fIPSCs occurred much faster than that induced by a light stimulus (Eggers and Lukasiewicz 2006b) or the summed inputs activated by an electrical stimulus (Fig. 1). This is likely due to the rapid activation of Ca²⁺-permeable AMPA receptors on the A17 amacrine cells by depolarization of the rod bipolar cell (Chavez et al. 2006). Although the timing of the GABAR currents was different, the timing of the underlying GABA release was not significantly different. On average release onto GABA_CRs was slightly faster than release onto GABA_ARs (P = 0.07, Fig. 6D), in contrast with our results with an electrical stimulus (Fig. 2) where release onto GABA_CRs is considerably slower. This shows that the slower timecourse of GABA_CR fIPSCs was due to slow GABA_CR kinetics, and not slow GABA release. These results suggest that, at the rod bipolar cell-A17 synapse, GABAR kinetics predominantly shape GABAergic input to rod bipolar cells and that GABA release onto GABA_ARs and GABA_CRs occurs by a similar mechanism in contrast to GABA release from amacrine cells that provide nonreciprocal input to rod bipolar cells.

Fig. 6.

The distinct timing of GABA release onto GABA_ARs and GABA_CRs can be measured at the rod bipolar cell-A17 amacrine cell synapse. A: representative traces of reciprocal feedback (f) IPSCs from A17 amacrine cells onto rod bipolar cells are shown. GABAergic fIPSCs of rod bipolar cells were evoked by a 10-ms step from −60 to −20 mV (top) and recorded in the absence (dark gray) and presence of 1,2,5,6-(tetrahydropyridin-4-yl)methylphosphinic acid hydrate (TPMPA) to isolate GABA_AR-mediated IPSCs (light gray trace). The GABA_CR (black trace)-mediated input was determined after subtracting the GABA_AR IPSC from the total response. B: estimated release onto GABA_ARs and GABA_CRs from the example in A. C: GABA_AR (n = 5) fIPSCs D37 was briefer than the total fIPSC D37 (n = 5) and isolated GABA_CR fIPSCs (n = 5, P < 0.05 for both). The GABA_CR fIPSC was not different from control (P = 0.2). D: the timing of GABA release onto GABA_ARs and GABA_CRs was not significantly different. *P < 0.05.

DISCUSSION

Our results show that the slow release of glycine and GABA onto GABA_ARs is an inherent characteristic of amacrine cells, similar to GABA release onto GABA_CRs (Eggers et al. 2013). Glycine and GABA release onto GABA_CRs occurs asynchronously and can be reduced by both EGTA-AM and BAPTA-AM. Although the timing of GABA release onto GABA_ARs was also inherently slow, it occurred more rapidly than glycine release and GABA release onto GABA_CRs and was not reduced by slow or fast Ca²⁺ buffering. The differential timing of GABA release was not observed at the reciprocal rod bipolar cell-A17 amacrine cell synapse, indicating that different release mechanisms are employed in reciprocal and nonreciprocal rod bipolar cell-amacrine cell synapses and that GABA release at nonreciprocal synapses likely comes from different GABAergic amacrine cells. The differences in the timecourse of release onto GABAR and glycineR inputs may provide another level of control on retinal inhibition and a way to tune information across retinal pathways.

Comparisons with release at other synapses.

At some conventional synapses, fast synchronous release that occurs within tens of milliseconds is followed by prolonged asynchronous release that can last for several hundred milliseconds (Atluri and Regehr 1998; Cummings et al. 1996; Goda and Stevens 1994; Jiang et al. 2012; Lu and Trussell 2000; Otsu et al. 2004; Sakaba 2006), often in response to repetitive stimuli (Best and Regehr 2009; Cummings et al. 1996; Hefft and Jonas 2005; Jiang et al. 2012; Lu and Trussell 2000; Otsu et al. 2004; Sakaba 2006). Hippocampal neurons utilize asynchronous release at excitatory (Cummings et al. 1996; Otsu et al. 2004) and inhibitory (Hefft and Jonas 2005; Jensen et al. 2000) synapses to maintain synaptic transmission during high-frequency stimulation after the depression of synchronous release (Otsu et al. 2004). We have shown here and in previous studies (Eggers et al. 2013; Eggers and Lukasiewicz 2006b) that a single brief stimulus can evoke prolonged GABAR and glycineR IPSCs. These prolonged responses are mediated by prolonged GABA and glycine release, and are not artifacts of our electrical stimulus because stimulated amacrine cells show a brief depolarization and the D37 of the eIPSCs followed the same hierarchy (GABA_CRs > glycineRs > GABA_ARs; Fig. 1) as the D37 of L-IPSCs (Eggers and Lukasiewicz 2006b) even when each receptor input was measured from the same rod bipolar cell. Asynchronous release has also been observed to occur after step depolarization at the calyx of Held (Sakaba 2006) and after a high-intensity stimulus in cerebellar granule cells (Atluri and Regehr 1998), but neither rely on asynchronous release as a primary mode of transmitter release. Amacrine cells belong to a unique subset of neurons in that they inherently release transmitter with prolonged timing after a single stimulus. The use of asynchronous release as a primary mode of transmitter release has so far been observed in only one other neuronal system where asynchronous release accounts for >90% of all release from neurons that project from the deep cerebellar nuclei to the inferior olive in response to low-frequency stimulation (Best and Regehr 2009).

Mechanisms of asynchronous release from glycinergic amacrine cells.

Our results suggest that glycine release from amacrine cells is similar to GABA release onto GABA_CRs (Eggers et al. 2013). Amacrine cell glycine release occurs asynchronously because it is attenuated by the slow Ca²⁺ buffer EGTA-AM (Augustine et al. 2003). Asynchronous release of GABA has been observed at other synapses (Best and Regehr 2009; Hefft and Jonas 2005; Jiang et al. 2012; Lu and Trussell 2000), but asynchronous release of glycine has not been widely reported. A recent study in neurons of the avian auditory brain stem that release both GABA and glycine showed that asynchronous glycine release could only be evoked by high-frequency stimulation after GABA was depleted (Fischl et al. 2014). The authors suggest that this is due to the recruitment of glycine to vesicles in the absence of adequate amounts of GABA. In the retina, colocalization and corelease of GABA and glycine has not been observed, and both transmitters are released from different populations of amacrine cells (Masland 2012); therefore, it is unlikely that GABA release influences glycine release from the same amacrine cell.

In contrast to GABA release, electrically evoked glycine release from amacrine cells was not blocked by TTX, suggesting that the glycinergic amacrine cells activated by our electrical stimulus do not rely on the spikes to trigger release. This is likely because the span of glycinergic amacrine cell processes is compact, and they do not extend across the retina as widely as GABAergic amacrine cells (Masland 2001). Thus, they would not require spikes to support transmission of a signal across long distances as may be the case for some GABAergic amacrine cells (Eggers et al. 2013; Shields and Lukasiewicz 2003). Glycine release from amacrine cells is Ca²⁺ dependent (Bieda and MacIver 2004; Chavez and Diamond 2008; Habermann et al. 2003) and relies in part on activation of slowly inactivating L-type Ca²⁺ channels (Bieda and MacIver 2004; Chavez and Diamond 2008; Habermann et al. 2003) and Ca²⁺-induced Ca²⁺ release (Chavez and Diamond 2008) that would cause prolonged Ca²⁺ signals. We have previously shown that the activation of L-type Ca²⁺ channels and Ca²⁺-induced Ca²⁺ release underlies asynchronous release from GABAergic amacrine cells onto rod bipolar cell GABA_CRs (Eggers et al. 2013), suggesting the mechanisms for slow release may be similar between glycinergic and GABAergic amacrine cells.

Differential GABA release from spiking GABAergic amacrine cells.

An unexpected result from this present study shows that GABA release from amacrine cells onto GABA_ARs occurs differently than release onto GABA_CRs even though they both come from GABAergic amacrine cells. Differential release of the same transmitter from different types of neurons onto the same postsynaptic neuron has been reported in other neuronal systems (Best and Regehr 2009; Hefft and Jonas 2005). Our data showing different Ca²⁺ buffering sensitivity of release onto GABA_ARs and GABA_CRs suggest that the differences in the timing of GABA release from spiking amacrine cells may reflect differences in inputs from multiple types of GABAergic amacrine cells onto the same rod bipolar cell terminals. There are ∼30 types of amacrine cells found in the retina that receive excitatory input from rod and cone pathways, and some subset of these could provide inhibitory input to rod bipolar cells (Masland 2001). Also, GABA_ARs and GABA_CRs cluster at different synapses on rod bipolar cell terminals (Koulen et al. 1998), so their inputs could be independent. It is possible that the electrically evoked GABAergic input on to each receptor arises from nonreciprocal amacrine cells that have distinct release mechanisms. A unique feature of rod bipolar cells is that, in addition to receiving nonreciprocal amacrine cell inhibition, they also receive reciprocal inhibition from a nonspiking GABAergic amacrine cell, the A17 amacrine cell, which accounts for ∼50% of GABAergic inhibitory input (Chavez et al. 2010; Chavez et al. 2006; Hartveit 1999; Nelson and Kolb 1985). Although our focal electrical stimulus strongly activated nonreciprocal amacrine cell inputs to rod bipolar cells (Eggers et al. 2013), it likely did not induce GABA release from A17 amacrine cells that do not generate spikes (Chavez et al. 2006; Menger and Wässle 2000). A17 amacrine cells rely predominantly on the opening of Ca²⁺-permeable AMPA receptors that are directly activated specifically by rod bipolar cell glutamate release to trigger GABA release (Chavez et al. 2006). We found that the timing of depolarization-induced GABA release from A17 amacrine cells onto GABA_ARs and GABA_CRs was not different (Fig. 6). Therefore, our results indicate that rod bipolar cells receive nonreciprocal inhibition from different amacrine cells with distinct release mechanisms that differs from the release that mediates reciprocal inhibition.

Differences in spiking GABAergic amacrine cells release could be due to differences in Ca²⁺ signaling such as the expression of voltage-dependent Ca²⁺ channels, the extent of Ca²⁺-induced Ca²⁺ release, vesicle availability, the Ca²⁺ sensors found on the vesicles, and the amount of Ca²⁺ buffers that together influence how Ca²⁺ triggers release (Kaeser and Regehr 2013). In addition to slow L-type Ca²⁺ channels, amacrine cells also express N-type and P/Q-type Ca²⁺ channels (Bieda and MacIver 2004; Chavez and Diamond 2008; Chavez et al. 2010; Vigh and Lasater 2004), which would not contribute to a prolonged Ca²⁺ signal (Eggers et al. 2013; Randall and Tsien 1995). Our results suggest that GABA release occurs more rapidly from spiking amacrine cells that release GABA onto GABA_ARs because neither the slow Ca²⁺ buffer EGTA-AM nor the fast Ca²⁺ buffer BAPTA-AM affected GABA release onto GABA_ARs. Another potential explanation for the differential effects of the Ca²⁺ buffers on amacrine cell GABA release is that it is due to differing sensitivities of amacrine cells to the AM ester analogs of EGTA and BAPTA that may cause varied accumulation of the buffers in amacrine cells. However, we think that this possibility is unlikely because a higher concentration of EGTA-AM (100 μM) did not affect GABA release onto GABA_ARs (unpublished observations). We think it is more likely that, in spiking GABAergic amacrine cells, release onto GABA_ARs is more tightly coupled to Ca²⁺ (Hefft and Jonas 2005) possibly due to the close proximity of vesicles to sites of Ca²⁺ entry where Ca²⁺ is present in concentrations as high as 100 μM (Augustine et al. 2003; Kaeser and Regehr 2013) that could saturate BAPTA-AM, thereby reducing its buffering capacity (Cummings et al. 1996; Tymianski et al. 1994). Neurotransmitter release at the parvalbumin interneuron-granule cell synapse has also been shown to be only weakly sensitive to Ca²⁺ buffering with BAPTA-AM (Hefft and Jonas 2005). Alternatively, GABAergic amacrine cells that release onto GABA_ARs could have a higher Ca²⁺-buffering capacity that prevents a global buildup of Ca²⁺, as has been seen in other retinal neurons (Mehta et al. 2014). In addition to various sources of Ca²⁺, some studies have suggested that different synaptic proteins are involved in mediating asynchronous release (Kaeser and Regehr 2013), including synaptotagmin 7 (Schonn et al. 2008; Wen et al. 2010), Doc2 (Yao et al. 2011), and the SNARE (soluble N-ethylmaleimide-sensitive factor attachment receptor) protein vAMP4 (vesicle-associated membrane protein 4) (Raingo et al.). These proteins could be expressed in distinct amacrine cells at different levels and may contribute to differential release. Additionally, although we saw no differences in the timing of amacrine cell depolarization, the possibility remains that different amacrine cells have distinct voltage responses to the electrical stimulus that would affect the timing of release. However, this hypothesis is difficult to test because the identities of all the amacrine cells that inhibit rod bipolar cell output are unknown.

Functional implications.

Light stimuli are processed by the rod pathway when rod photoreceptors are active in dim light and the cone pathways when cone photoreceptors are active at bright light intensities (Wang and Kefalov 2009). Amacrine cell-mediated inhibition limits the extent of bipolar cell output in the cone and rod pathways by releasing GABA or glycine onto different proportions of GABARs and glycineRs with distinct kinetics (Eggers and Lukasiewicz 2006b), and our results reported here and previously (Eggers et al. 2013; Eggers and Lukasiewicz 2006b) suggest that differential asynchronous release also contributes. Previous studies have shown that light-evoked rod bipolar cell glutamate release is prolonged with increasing light intensities (Berntson and Taylor 2000; Euler and Masland 2000). Given that light-evoked GABA release onto rod bipolar cell terminals is also increased and occurs on a slower timecourse at increasing light intensities (Eggers and Lukasiewicz 2006b), it is possible that differences in prolonged GABA and glycine release are a way to maintain long-lasting inhibition at higher light intensities to match prolonged rod bipolar cell glutamate release. At other inhibitory synapses, asynchronous release has been shown to increase after the depression of synchronous release to maintain tonic inhibitory input (Fischl et al. 2014; Hefft and Jonas 2005; Jensen et al. 2000; Otsu et al. 2004). Asynchronous release may also be more adaptable to changes in stimuli (light intensity) than receptor expression and kinetics because release can be adjusted rapidly by modulating vesicle availability and Ca²⁺ signaling (Kaeser and Regehr 2013).

Release onto GABA_ARs and glycineRs provides the initial inhibitory input that shapes the peak of rod bipolar cell IPSCs (Eggers and Lukasiewicz 2006b) while prolonged release onto glycineRs and GABA_CRs controls the slow timecourse of rod bipolar cell IPSCs, and maintains inhibition to match sustained rod bipolar cell glutamate release (Eggers and Lukasiewicz 2006b). Because the kinetics of light-evoked IPSCs have been shown to affect the kinetics of rod bipolar cell output (Eggers and Lukasiewicz 2006a), increased inhibition would limit the output of the rod pathway as the light intensity increases to cone levels. Our results suggest that the differential timing of slow transmitter release is used in amacrine cells presumably to match the sustained rod bipolar cell excitatory output. Assuming that differential amacrine cell release occurs in the “on” and “off” pathways, the difference in timing of release combined with known differences in receptor kinetics (Eggers et al. 2007) would make the kinetics of inhibition different between each pathway allowing for rapid adaptation to changes in light intensities.

GRANTS

This work was supported by the National Institutes of Health (EY-018131 to E. D. Eggers, Graduate Training in Systems and Integrative Physiology Grant 5T32-GM-008400 appointment to R. E. Mazade, and Cardiovascular Training Grant 2T32-HL-7249-36A1 appointment to J. M. Moore-Dotson) and Science Foundation Arizona (R. E. Mazade).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

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