Contributions of glutamate transporters and Ca²⁺-activated Cl⁻ currents to feedback from horizontal cells to cone photoreceptors

Abstract

Lateral inhibitory feedback from horizontal cells (HCs) to cones establishes center-surround receptive fields and color opponency in the retina. When HCs hyperpolarize to light, inhibitory feedback to cones increases activation of cone Ca²⁺ currents (I_Ca) that can in turn activate additional currents. We recorded simultaneously from cones and HCs to analyze cone currents activated by HC feedback in salamander retina. Depolarization-activated inward tail currents in cones were inhibited by CaCCinh-A01 that inhibits both Ano1 and Ano2 Ca²⁺-activated Cl⁻ currents (I_Cl(Ca)). An Ano1-selective inhibitor Ani9 was less effective suggesting that Ano2 is the predominant I_Cl(Ca) subtype in cones. CaCCinh-A01 inhibited feedback currents more strongly when intracellular Ca²⁺ in cones was buffered with 0.05 mM EGTA compared to stronger buffering with 5 mM EGTA. By contrast, blocking glutamate transporter anion currents (I_Cl(Glu)) with TBOA had stronger inhibitory effects on cone feedback currents when Ca²⁺ buffering was strong. Inward feedback currents ran down at rates intermediate between rundown of glutamate release and I_Cl(Ca), consistent with contributions to feedback from both I_Cl(Ca) and I_Cl(Glu). These results suggest that Cl⁻ channels coupled to glutamate transporters help to speed inward feedback currents initiated by local changes in intracellular [Ca²⁺] close to synaptic ribbons of cones whereas Ano2 Ca²⁺-activated Cl⁻ channels contribute to slower components of feedback regulated by spatially extensive changes in intracellular [Ca²⁺].

1. Introduction

One of the essential early circuits in vision involves negative feedback from second-order horizontal cells (HCs) to cone photoreceptor cells in the retina. Cones hyperpolarize to light and negative feedback from HCs depolarizes cones to restore their synaptic output (Baylor et al., 1971). HCs have a large receptive field, collecting inputs from many cones, and so negative feedback from HCs allows comparisons between local changes detected by an individual cone with the average response from many surrounding cones. Negative feedback from HCs to cones plays a key role in detecting changes in both luminance and chromatic contrast (Thoreson and Mangel, 2012; Thoreson and Dacey, 2019).

Central to this negative feedback, hyperpolarization of HCs by light causes both an increase in peak amplitude and a negative activation shift in the voltage-dependence of cone calcium currents (I_Ca) (Verweij et al., 1996). The combined effects of this negative activation shift and increase in peak amplitude increases I_Ca over the normal physiological range of cone membrane potentials (ca. −35 to −60 mV). The mechanisms by which negative feedback from HCs alter cone I_Ca are not fully understood, but involve extracellular pH changes (Cadetti and Thoreson, 2006; Grove et al., 2019; Thoreson and Mangel, 2012; Vroman et al., 2014; Wang et al., 2014; Warren et al., 2016a).

When HCs hyperpolarize to light, the increased Ca²⁺ influx produced by feedback exerts a depolarizing influence on cones and increases release of glutamate-filled synaptic vesicles. Ca²⁺ entry into the cone also activates Ca²⁺-dependent Cl⁻ channels in cones. The feedback-induced depolarization in cones exhibits Cl-dependence and can be inhibited by the Ca²⁺-activated Cl⁻ channel blocker, niflumic acid (Barnes and Deschenes, 1992; Kraaij et al., 2000; Packer et al., 2010; Thoreson and Burkhardt, 1991; Verweij et al., 2003). Cl⁻ channel activity can influence feedback by effects of Cl⁻ flux on cone membrane potential, changes in ephaptic currents (Endeman et al., 2012) and by direct effects on I_Ca (Thoreson et al., 1997). In addition to effects on post-synaptic glutamate receptors, glutamate released from cones can also have a presynaptic influence through the activity of glutamate transporters in cones. Following release from a cone, glutamate molecules are retrieved by presynaptic EAAT2 transporters in cone terminals. EAAT glutamate transporters possess an uncoupled anion channel that opens during the transport of glutamate (Eliasof and Werblin, 1993; Picaud et al., 1995a; Rauen and Kanner, 1994; Vandenbranden et al., 1996). Cones possess two EAAT2 splice variants (Eliasof et al., 1998b; Rauen et al., 2004; Reye et al., 2002; Rowan et al., 2010; Schneider et al., 2014) and although some EAAT2 isoforms exhibit little anion current, the retina-specific variant of EAAT2 shows prominent anion currents (Schneider et al., 2014). Release of glutamate can activate EAAT anion currents in both the cone that released the glutamate (Picaud et al., 1995b) and neighboring cones (Arriza et al., 1997; Szmajda and Devries, 2011). Vroman and Kamermans (2015) showed that activation of glutamate-associated Cl⁻ currents in neighboring cones by diffusion of glutamate helps to restore their membrane potential back into the range where I_Ca is active. This study used surround illumination to stimulate HC feedback in goldfish cones and so was not able to assess the role of locally released glutamate on feedback currents. We investigated the contributions of both I_Cl(Ca) and glutamate transporter anion currents (I_Cl(Glu)) to feedback-induced currents in cones. Using simultaneous whole-cell patch clamp recordings from both cones and HCs, we were able to study the effects of changes in HC membrane potential on cone membrane currents. Our results confirm and extend earlier studies showing that changes in I_Ca, I_Cl(Ca), and I_Cl(Glu) all contribute to the negative feedback currents in cones evoked by changes in HC membrane potential. Contributions of I_Cl(Glu) had a net effect of speeding feedback currents and were more prominent when Ca²⁺ was buffered to restrict its spread. In contrast, I_Cl(Ca) showed a greater contribution to slower components of feedback and played a more prominent role when Ca²⁺ was allowed to spread further from Ca²⁺ channels. The combined effects of these currents provide depolarizing influences that can adjust feedback to compensate for changes in the spatial and temporal patterns of illumination.

2. Methods

2.2. Animal care and use

Aquatic tiger salamanders (Ambystoma tigrinum, 18 to 25 cm; Waddell’s Waterdogs, Fayetteville, TX, USA) were maintained on a 12-h light/dark cycle and sacrificed after ≥1 h of dark adaptation. Salamanders were anaesthetized by immersion in 0.25 g/L Tricaine-S (tricaine methanesulfonate, Western Chemical, WA, USA) for >15 min, decapitated with heavy shears, and then pithed. Animal care and use procedures were approved by the University of Nebraska Medical Center Institutional Animal Care and Use Committee.

2.3. Retinal slices

A detailed description of retinal slice preparation and whole cell recording techniques has been provided previously (Van Hook and Thoreson, 2013). Briefly, after enucleation, the front of the eye was removed and the resulting eyecup cut into 2-4 pieces. A piece of retina was placed vitreal side down on a piece of nitrocellulose membrane (5 × 10 mm; type AAWP, 0.8 μm pores; EMD Millipore). The filter paper was submerged in cold amphibian saline and the sclera peeled away, leaving the retina attached to the membrane. The retina was cut into 125-μm slices using a razor blade tissue slicer (Stoelting Co, Wood Dale, IL, USA). Slices were rotated to view the retinal layers and anchored in the recording chamber by embedding ends of the nitrocellulose membrane in vacuum grease. The recording chamber was mounted on an upright fixed-stage microscope (Olympus BH2-WI, Tokyo, Japan) and slices were superfused at ~1 ml min⁻¹ with bicarbonate-buffered amphibian saline solution (in mM): 101 NaCl, 22 NaHCO₃, 2.5 KCl, 2.0 CaCl₂, 0.5 MgCl₂, 11 glucose (pH 7.4). Solutions were bubbled continuously with 95% O₂/5% CO₂.

2.4. Patch-clamp electrophysiology

Patch pipettes were fabricated from borosilicate glass pipettes (1.2 mm OD, 0.9 mm ID, with an internal filament; World Precision Instruments, Sarasota, FL, USA) using a PC-10 or PP-830 vertical puller (Narishige, Tokyo, Japan) to achieve tip diameters of ~1 μm and resistances of 10-15 MΩ.

We used a number of pipette solutions that differed in anion levels and Ca²⁺ buffering. In many experiments, we used a Cs-Gluconate/Cs-Glutamate-based pipette solution for cones that contained (in mM): 50 Cs-gluconate, 40 Cs-glutamate, 10 TEA-Cl, 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 10 ATP-Mg, 0.5 GTP-Na, 10 HEPES, and 5 EGTA. The liquid junction potential (LJP) for this solution calculated from the calculator incorporated into Clampex (Axon Instruments) was −12.7 mV. The inclusion of glutamate in the cone pipette solution enhances excitatory post-synaptic currents (EPSCs) in HCs and slows rundown during paired whole-cell recording (Bartoletti and Thoreson, 2011). We also tested a CsGluconate pipette solution that omitted glutamate and consisted of (in mM): 90 Cs-gluconate, 10 TEA-Cl, 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 10 ATP-Mg, 0.5 GTP-Na, 10 HEPES, and 5 EGTA. The LJP for this solution was −12.2 mV. We did not see any significant difference in effects of TBOA on feedback in recordings with and without added glutamate so we combined these data in the graphs shown in Figure 6. When combining these data, we applied an LJP of −12.5 mV. We also tested cone pipette solutions in which the primary anion was chloride and contained (in mM): 90 CsCl, 10 TEA-Cl, 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 9.4 ATP-Mg, 0.5 GTP-Na, 10 HEPES, and 5 EGTA. For CsCl-based pipette solutions, the LJP was −3 mV. In some experiments, we lowered EGTA to 0.05 mM. In experiments where we tested TBOA effects on glutamate transporter currents in cones, we enhanced anion currents by using a pipette solution containing (in mM): 90 KSCN, 10 TEA-Cl, 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 9.4 ATP-Mg, 0.5 GTP-Na, 10 HEPES, and 5 EGTA (LJP = −4.5 mV). For gramicidin-perforated patch recordings in cones, the pipette solution contained (in mM): 90 KGluconate, 10 TEA-Cl, 3.5 NaCl, 1 CaCl₂, 1 MgCl₂, 9.4 ATP-Mg, 0.5 GTP-Na, 10 HEPES, and 5 EGTA. To this solution, we added 5 μg/ml gramicidin from a stock solution prepared in 95% ethanol (5 mg/ml). Gramicidin is only permeable to monovalent cations and differences in monovalent cation levels between the pipette and bath solutions predict an LJP across the perforated patch of <1 mV. Given the small value and uncertainties about calculating the LJP under perforated patch conditions, we did not correct for the LJP in perforated patch, current clamp experiments. However, we did correct for predicted LJPs in ruptured patch, voltage clamp recordings

Figure 6. Glutamate contributes to feedback currents in cones.

(A) Feedback currents in a cone voltage-clamped at −30 mV and recorded with a pipette solution containing CsCl and 5 mM EGTA. Application of TBOA (300 μM; red trace) inhibited the feedback current. There was partial recovery after 3 min of washout (gray trace). Inward feedback currents increased with time constants of 35.7 ms and 31.3 ms in control and TBOA, respectively. (B) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=6), TBOA (filled red triangles), and wash (open gray triangles) when using a cone pipette solution with 5 mM EGTA and E_Cl = 0 mV. Feedback currents evoked when cones were held at −33 mV were significantly reduced by TBOA (paired t-test, P = 0.0013). (C) Amplitude of inward feedback currents in cones plotted against cone holding potential when using pipette solutions containing CsGluconate or a combination of CsGluconate and 40 mM CsGlutamate (E_Cl = −46 mV) along with 5 mM EGTA (N=11 cell pairs). Control: black squares. TBOA: filled red triangles. Wash: open gray triangles. Feedback currents evoked when cones were held at −43 mV were significantly reduced by TBOA (paired t-test, P = 0.005) and increased at −23 mV (P=0.0064), +7 mV (P=0.0033), and +17 mV (P<0.0001). (D) An example of feedback currents in a cone voltage-clamped at −33 mV recorded with a pipette solution containing CsCl and 0.05 mM EGTA before (black trace), during (red), and after (gray) bath application of TBOA (300 μM). (E) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=6), TBOA (filled red triangles), and wash (open gray triangles). Pipette solution: 0.05 mM EGTA with E_Cl = 0 mV. Feedback currents evoked when cones were held at −33 mV were significantly reduced by TBOA (paired t-test, P = 0.0041). (F) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=5), TBOA (filled red triangles), and wash (open gray triangles). Pipette solution: 0.05 mM EGTA with E_Cl = −46 mV. Feedback currents evoked when cones were held at −43 mV were significantly reduced by TBOA (paired t-test, P = 0.0077).

Most of our experiments involved paired whole-cell recordings from cones and HCs using a Multiclamp 700B amplifier, pClamp 10.7 software, and Digidata 1440A digitizer (Molecular Devices, Sunnyvale, CA, USA). Cones were identified by their morphology. HCs were identified by their morphology, position in the slice, and electrophysiological characteristics (Van Hook and Thoreson, 2013). In cones, the membrane capacitance averaged 77.6 ± 30.0 (mean ± SD) pF, membrane resistance averaged 447.6 ± 258.8 MΩ, and series resistance averaged 38.2 ± 16.1 MΩ (N=145). In horizontal cells, the membrane capacitance averaged 42.7 ± 18.0 pF, membrane resistance averaged 353.8 ± 368.8 MΩ, and series resistance averaged 45.7 ± 21.5 MΩ (N=129). Voltages were not corrected for series resistance.

2.5. Reagents

DL-TBOA, CaCCinh-A01, Ani9, and 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) were from Tocris (Bristol, UK). Unless otherwise noted, other reagents were from Sigma-Aldrich Chemicals (St. Louis, MO, USA). Test solutions were bath applied.

2.6. Statistics

Data were analyzed using Clampfit (Molecular Devices) and Prism 7 (GraphPad). Statistical significance was evaluated by regression analysis, Student’s t-tests, and ANOVA with p < 0.05. Except where otherwise noted, values are reported as mean ± SEM.

3. Results

3.1. Depolarization of cones by feedback from single HCs

Inhibitory feedback from HCs to cones was originally identified by the ability of surround illumination to depolarize cones during steady central illumination (Baylor et al., 1971). The amplitude of surround-evoked depolarization varies among experimental preparations and it is thought that this variability is largely due to differences in E_Cl. We examined effects of HC feedback on cone membrane potential in simultaneous, paired recordings from individual cones and HCs in salamander retina. We voltage-clamped the HC at −43 mV, similar to its resting membrane potential in darkness, and strongly activated inhibitory feedback by applying a 60 mV hyperpolarizing step. Figure 1 shows the impact of this change in HC membrane potential on the membrane voltage measured simultaneously in a cone. The voltage protocol for the HC is shown at the top of Figure 1A. To retain endogenous Ca²⁺ buffers and endogenous Cl⁻ levels in the cone, we used gramicidin perforated patch recordings to record voltage responses from cones. We adjusted the holding current to vary the cone membrane potential from −60 to +10 mV (Figs. 1A-C). When the cone was at a relatively positive potential near −9 mV, stepping the HC from −43 to −103 mV evoked a small depolarization (Fig. 1A). When the cone membrane potential was near −29 mV, depolarizing the HC from −73 to −43 mV caused the cone to hyperpolarize by about 1 mV and then subsequently hyperpolarizing the HC from −43 to −103 mV produced a 1.8 mV feedback depolarization (Fig. 1B). This feedback-evoked depolarization virtually disappeared when the cone was held at −59 mV (Fig. 1C). The graph in Figure 1D plots the feedback-evoked depolarization (evoked by the step from −43 to −103 mV applied to a simultaneously voltage-clamped HC) measured in 5 cones at various membrane potentials. Figure 1E shows the same data after normalizing responses from each cone to the largest response in that cone. The red dots show the distribution of membrane potentials from the cone shown in Figures 1A-C. When fit with Gaussian functions, the mean values for the two plots were −25.4 (Fig. 1D) and −26.1 mV (Fig. 1E). The largest feedback-evoked depolarization in each cone was attained at an average cone membrane potential of −28.2 ± 0.48 mV (N=5). Thus, the largest feedback depolarization is attained at potentials close to the midpoint activation voltage for I_Ca (e.g., see Fig. 3). When measured at cone membrane potentials closest to −40 mV in these 5 cones, feedback-evoked depolarization averaged 0.5 ± 0.2 mV. These data show that with endogenous Ca²⁺ buffering and intracellular Cl⁻ levels, feedback from a single HC can evoke depolarizing responses in a cone. In experiments described below, we present evidence that chloride currents, I_Cl(Ca) and I_Cl(glu) contribute more strongly at more negative potentials whereas I_Ca contributes more strongly to responses at more positive potentials.

Figure 1. Depolarization of cones by feedback from single HCs during paired recordings.

(A) Gramicidin-perforated patch recordings that retain endogenous Cl⁻ levels were used to record voltage changes in cones while simultaneously voltage-clamping a post-synaptic HC. The voltage protocol used to stimulate the HC is shown at the top (HC E_m). The resulting changes in cone membrane potential are shown below. The steady holding current applied to the cone was adjusted to yield average membrane potentials of −9 (A), −29 (B) and −59 mV (C). A larger feedback response was evoked when the cone was held near −29 mV than at the other two potentials. (D) The depolarizing voltage change produced in cones by hyperpolarizing the HC from −43 to −103 mV are plotted against the cone membrane potential (N = 5 pairs, 46 data points). The filled red circles show the data from the example cone/HC showed in (A-C). The data were fit with a Gaussian curve (mean = −25.4 mV, SD = 10.7 mV). (E) The same data plotted after normalizing responses to the largest response in each cell and fit with a Gaussian distribution (mean = −26.1 mV, SD = 12.8 mV).

Figure 3. Comparison of feedback changes in I_Ca to step-evoked feedback currents in cones.

(A) Leak-subtracted cone I_Ca plotted against the cone holding potential measured using CsGluc/Glut pipette solution while a post-synaptic HC was voltage-clamped at either −43 (red trace) or −103 mV (black trace). I_Ca was evoked by a ramp voltage protocol applied to the cone (0.5 mV/ms; −90 to +60 mV). (B) Difference current obtained by subtracting cone I_Ca measured when the HC holding potential was −103 mV from that measured when the HC was held at −43 mV. (C) Average difference current from 9 cone/HC pairs plotted as mean ± 95% confidence interval (C.I., gray trace, right axis). The same graph also shows the peak amplitude of inward feedback currents in cones evoked by HC steps from −43 to −103 mV plotted against cone holding potential (left axis). Black squares (dashed line) show results when using CsGluconate-(N=11 pairs) and red triangles (solid line) shows results obtained using CsCl-based pipette solutions (N=7). 5 mM EGTA was the Ca²⁺ buffer in both. Each data points show average ± 95% C.I..

3.2. Chloride currents contribute to inward feedback currents in cones.

To analyze currents responsible for the voltage changes in Figure 1, we measured feedback currents in voltage-clamped cones while applying a hyperpolarizing voltage step to a simultaneously voltage-clamped HC (Warren et al., 2016b). As in Figure 1, we held the HC at −43 mV and then hyperpolarized it by 60 mV. We measured effects of this change in HC membrane potential on cone membrane currents while holding a cone at number of different membrane potentials. The top of Figure 2A shows the HC voltage protocol (top) and membrane current (bottom). The top of Figure 2B shows the cone voltage protocol (top) and resulting membrane current (bottom). When the HC and cone were synaptically coupled , depolarizing the cone to −30 mV or above evoked a transient inward EPSC in the HC (Fig. 2A, arrow). Hyperpolarizing the HC from −43 to −103 mV evoked a 10 pA inward feedback current when the cone was voltage-clamped at −33 mV (Fig. 2B, bottom). The inward feedback current evoked in this cone is shown on an expanded time scale in Figure 2C. Consistent with the voltage changes seen in Figure 1, the largest feedback currents were typically seen in cones held near −30 mV; feedback current amplitude generally declined at more negative or positive potentials.

Figure 2. Testing for inward feedback currents in cone/HC pairs.

(A) The voltage step waveform applied to the voltage-clamped HC is shown at the top (HC V_m), and the resulting HC membrane currents accompanying the five different test steps applied to cones are overlaid below (HC I_m). The presence of excitatory postsynaptic currents (EPSCs) evoked in the HC when the cone was depolarized to at least −33 mV (arrow), shows that the HC and cone were synaptically connected. (B) Voltage step protocol applied to a simultaneously voltage-clamped cone (Cone E_m) is shown at the top and the resulting cone membrane currents shown below (Cone I_m). The maximal inward feedback current evoked by hyperpolarizing the HC from −43 to −103 mV was observed when the cone was held at −33 mV (blue rectangular). The cone pipette solution was a CsCl-based solution with 5 mM EGTA. (C) Magnified view of the inward feedback current evoked when the cone was voltage-clamped at a nominal potential of −33 mV. The feedback current was fit with a single exponential curve (τ = 79 ms). (D) Charging curve in the HC evoked by a step from −73 to −43 mV. Input resistance of this HC was 942 MΩ and the charging curve was well fit by a single exponential function (τ = 1.0 ms) showing that this HC had a compact electrotonic structure amenable to voltage clamp. (E) Feedback time constants (τ) measured in cones held at −33 mV plotted as a function of HC input resistance (R_in). Slope of the linear regression (R²=0.0026) was not significantly different from zero (F-test, P=0.81).

The feedback currents in cones evoked by hyperpolarizing steps applied to HCs show relatively slow kinetics (Warren et al., 2016b). For example, the feedback current shown in Figure 2C was fit by a single exponential function with a time constant of 79 ms (red line). The slowest feedback currents were seen when cones were held near −30 mV. Slow kinetics were not a result of poor voltage clamp of the HC membrane potential. Predictions of an anatomically realistic model show that a voltage change applied to the HC soma should alter the voltage at the tip of an HC dendrite in <0.1 ms (Warren et al., 2016b). The HC recording shown in Figure 2 provides further evidence of good voltage clamp. As in other cell pairs, the HC in this example had a high input resistance of 942 MΩ suggesting it was largely uncoupled from neighboring HCs. Consistent with this, the charging curve could be fit by a single exponential function (τ = 1 ms) demonstrating that this HC had a compact electrotonic structure amenable to voltage clamp (Fig. 2D). If slow feedback time constants were a result of HC coupling, then they should become progressively slower with more strongly coupled HCs that have a lower input resistance. However, as shown in Figure 2E, there was no relationship between cone feedback time constants and HC input resistance.

Another way to assess the potential impact of HC coupling on feedback currents in cones is to examine cell pairs that do not communicate directly with one another. We analyzed 12 cone/HC pairs in which stimulation of the cone failed to evoke an EPSC in the adjacent HC (out of more than 200 paired recordings used for this study), suggesting that the cone does synapse directly onto the HC. 11 of these 12 pairs showed no detectable feedback. One cone showed a small, slow feedback current (3 pA, τ=195 ms) during a single test step to −33 mV in the cone. It is possible that this might reflect feedback effects transmitted via coupled HCs, but it is also possible that these two cells were synaptically connected but feedforward release was compromised. If current spread to neighboring HCs was a major contributor to cone feedback currents then we should have seen effects in the other 11 cell pairs. Taken together, these various results indicate that while other factors may occasionally contribute, the time course of cone feedback currents in these experiments is largely determined by synaptic interactions between the voltage-clamped HC and its presynaptic cone.

Previous studies using paired recordings from cones and HCs have shown that hyperpolarizing a synaptically-coupled HC from −43 to −103 mV introduces a feedback signal to the cone that shifts cone I_Ca activation towards more negative voltages and increases the peak amplitude of I_Ca (Cadetti and Thoreson, 2006; Verweij et al., 1996; Warren et al., 2016a). The trace in Figure 3A shows examples of I_Ca measured in a cone while the HC was voltage-clamped at −43 and −103 mV. Figure 3B shows the difference current created by subtracting I_Ca recorded when the HC was held at −103 (black trace, Fig. 3A) from I_Ca recorded when the HC was held at−43 mV (red trace). As shown by the difference current, the leftward activation shift enhanced cone I_Ca above −40 mV with the maximal change in I_Ca produced by HC feedback attained slightly above −30 mV. Figure 3C shows the average difference current from nine cones evoked by the same protocol. On the graph in Figure 3C, we also plotted the peak amplitude of inward feedback currents evoked by hyperpolarizing HC steps measured at various cone holding potentials using the protocol shown in Figure 2. Feedback currents obtained when using the cone pipette solution containing mostly CsGluconate (E_Cl = −46 mV) and moderately strong Ca²⁺ buffering (5 mM EGTA) are plotted as the filled squares in Figure 3C. These data were aligned on the graph so that the mean amplitude of step-evoked feedback currents matched the amplitude of difference currents at −23 mV. The bars for these data sets show 95% confidence intervals. Large inward feedback currents were evoked in cones held at −43 mV where difference currents were quite small and confidence intervals of the two data sets do not overlap. Using a CsCl-based pipette solution (E_Cl = 0 mV; 5 mM EGTA), inward feedback currents were further accentuated (red triangles, Fig. 3C). Together, these data confirm other studies showing a significant contribution to feedback from Cl⁻ currents (Barnes and Deschenes, 1992; Kraaij et al., 2000; Packer et al., 2010; Thoreson and Burkhardt, 1991; Verweij et al.,2003).

3.3. Ano2 calcium-activated chloride currents contribute to feedback currents in cones.

To identify the chloride currents contributing to inward feedback currents in cones, we turned to pharmacology. I_Cl(Ca) has been suggested to contribute to feedback (Barnes and Deschenes, 1992; Kraaij et al., 2000; Packer et al., 2010; Thoreson and Burkhardt, 1991; Verweij et al., 1996; Verweij et al., 2003). In addition to the Cl⁻ dependence described above, one key piece of evidence is the ability of niflumic acid to block this current. However, niflumic acid can also inhibit I_Ca directly (Barnes and Deschenes, 1992; Thoreson et al., 2003). There is evidence for two types of Ca²⁺-activated Cl⁻ channels in cones: anoctamin-1 (Ano1/ TMEM16a) and anoctamin-2 (Ano2/ TMEM16b) (Dauner et al., 2013; Jeon et al., 2013; Mercer et al., 2011b; Stohr et al., 2009). We therefore tested two Ca²⁺-activated Cl⁻ channel antagonists: Ani9 that is selective for Ano1 over Ano2 (Seo et al., 2016) and CaCCinh-A01 that shows similar inhibitory potency at both types (Cherkashin et al., 2016). Activation of I_Cl(Ca) by strong depolarizing steps results in a large inward tail current at the end of the step (Mercer et al., 2011b). We measured the tail current amplitude 150 ms after the end of the test pulse relative to baseline currents when the cone was held at −73 mV. Application of steps to −13 mV in cones using a CsCl-based cone pipette solution (E_Cl = 0 mV) with 0.05 mM EGTA as the Ca²⁺ buffer (similar to the weak endogenous buffering in photoreceptors) (Van Hook and Thoreson, 2015) evoked large inward tail currents (Fig. 4A). As illustrated by this example, bath application of CaCCinh-A01 (50 μM) inhibited cone tail currents. Baseline currents aligned in Fig. 4A. The holding current in this example decreased from −129 to −108 pA. Overall, currents required to voltage clamp cones at −73 mV declined from −127 ± 24.8 pA to −96 ± 17.8 pA (N=5) after bath application of CaCCinh-A01 (50 μM, P=0.036, paired t-test).

Figure 4. Ano2 calcium-activated chloride currents in cones.

(A) Cone membrane currents (Cone I_m) evoked by a strong depolarizing test step (−70 to −10 mV, 3.2 s) show a large inward I_Cl(Ca) tail current at offset of the step (arrow). With a pipette solution where E_Cl = 0 mV and Ca²⁺ was buffered with 0.05 mM EGTA, the Ca²⁺-activated Cl⁻ channel inhibitor CaCCinh-A01 (50 μM), strongly inhibited this tail current. (B) Fractional inhibition of tail currents by 1, 10, and 50 μM CaCCinh-A01 (filled circles) and 10 μM Ani9 (filled triangle) using the same pipette solution. Open circles show fractional inhibition by 50 μM CaCCinh-A01 when using 5 mM EGTA in the cone pipette solution. (C) Application of TBOA (300 μM) reversibly inhibited tail currents in a cone recorded using a pipette solution where E_Cl = 0 mV and Ca²⁺ was buffered with 5 mM EGTA, showing that glutamate transporter currents significantly contribute to tail currents under these conditions.

CaCCinh-A01 showed a dose-dependent inhibition of tail currents with 75 ± 5.4% (N=5) inhibition at 50 μM, 62 ± 5.1% inhibition at 10 μM (N=8), and weaker inhibition at 1 μM (Fig. 4B). Bath application of 10 μM Ani9 inhibited tail currents by 22 ± 8.6% (N=7), less than the same concentration of CaCCinh-A01 (Fig. 4B). Higher concentrations of Ani9 did not go into solution. After subtracting the LJP of 3 mV, cone currents blocked by 50 μM CaCCinh-A01 reversed at −3.1 ± 1.37 mV (N=6), consistent with a predicted value for E_Cl of 0 mV. Tail currents showed little recovery after washout of these two compounds and so we tested them only once per preparation, preparing new retinal slices after each experiment with either CaCCinh-A01 or Ani9. Unlike niflumic acid, CaCCinh-A01 (50 μM) did not inhibit I_Ca in cones. CaCCinh-A01 (50 μM) caused a statistically insignificant increase in the peak amplitude of I_Ca (11.7 ± 12.1%; N=7, P=0.371, paired t-test). The greater efficacy of CaCCinh-A01 over Ani9 is consistent with other evidence that the predominant Ca²⁺-activated Cl⁻ channel subtype in photoreceptors is Ano2 (Dauner et al., 2013; Stohr et al., 2009).

Tail currents in cones were much smaller when Ca²⁺ buffering was increased to 5 mM EGTA, averaging only 66.4 ± 12.6 pA (N=18) with 5 mM EGTA compared to 304.2 ± 18.5 pA (N=40) when using a CsCl pipette solution with 0.05 mM EGTA. Inhibition of tail currents by 50 μM CaCCinh-A01 was also reduced to 28 ± 7.3% when using 5 mM EGTA (N=4, Fig. 4B). Much of the inward tail current observed with the CsCl/5 mM EGTA pipette solution (E_Cl=0 mV) appears to be due to glutamate transporter anion currents since it could be blocked by a glutamate transport inhibitor, TBOA (300 μM; control: 54.2 ± 17.7 pA; TBOA: 2.6 ± 2.6 pA; N=9; P=0.023, paired t-test). This is consistent with evidence that glutamate release sites are located within Ca²⁺ nanodomains close to Ca²⁺ channels (Mercer et al., 2011b) and that glutamate can diffuse rapidly to nearby transporters (Vroman and Kamermans, 2015). Tail currents were also reduced by application of TBOA when using 0.05 mM EGTA as the Ca²⁺ buffer (control: 294 ± 33.2 pA; TBOA: 164 ± 32.2 pA; N=6; P=0.0004, paired t-test)

We next tested effects of CaCCinh-A01 (50 μM) on feedback currents in cones evoked by hyperpolarizing steps applied to simultaneously voltage-clamped HCs. Figure 5A shows currents recorded from a cone using CsCl and 0.05 mM EGTA. The combination of weak Ca²⁺ buffering and a value of 0 mV for E_Cl increased the amplitude of inward feedback currents. As we discuss later, this pipette solution also yielded slow feedback currents. The enhancement accompanying weak Ca²⁺ buffering involves a substantial contribution from Ca²⁺-activated Cl⁻ channels. Feedback currents measured with this pipette solution when the cone was held at −33 mV were significantly inhibited by CaCCinh-A01 (50 μM), declining by 70% from −29.4 ± 11.0 to −8.8 ± 1.51 pA (N=5). Consistent with relatively weak effects of Ani9 on I_Cl(Ca) tail currents, inward feedback currents diminished only slightly after applying this compound (Fig. 5C). As shown later, feedback currents diminished by ~30% in the first 7 min when using this pipette solution that lacked supplemental glutamate. Thus, some of the decline seen with CaCCinh-A01 and much of the decline with Ani9 may be due to rundown. Feedback currents evoked in cones at −33 mV when using CsCl/0.05 mM EGTA did not always attain a plateau during the HC test step, so effects of CaCCinh-A01 on feedback current amplitude measured at −33 mV may underestimate the total contribution from Ca²⁺-activated Cl⁻ channels at this potential. In addition to promoting I_Cl(Ca), weak Ca²⁺ buffering may facilitate glutamate release at more negative potentials in cones. This may account for the relatively large feedback currents seen at −53 mV that were not blocked by CaCCinh-A01 (Fig. 5B). We explore contributions from I_Cl(Glu) further in experiments described below.

Figure 5. Ano2 calcium-activated chloride currents contribute to feedback currents in cones.

(A) Feedback currents in a cone voltage-clamped at −30 mV and recorded with a pipette solution containing CsCl and 0.05 mM EGTA. Application of CaCCinh-A01 (50 μM; red trace) inhibited the feedback current. There was no recovery after 5 min of washout (gray trace). Control feedback current (black trace) increased with a time constant of 286 ms. The time constant declined to 133 ms after application of CaCCinh-A01. (B) Amplitude of inward feedback currents in cones plotted against cone holding potential when using a CsCl-based pipette solution with 0.05 mM EGTA (N=6 cell pairs). Control: black squares. CaCCinh-A01: filled red triangles. Wash: open gray triangles. (C) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=6), Ani9 (10 μM; filled red triangles), and wash (open gray triangles). Pipette solution contained CsCl and 0.05 mM EGTA. (D) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=5), CaCCinh-A01 (10 μM; filled red triangles), and wash (open gray triangles). Pipette solution contained CsCl and 5 mM EGTA.

We also tested a CsCl cone pipette solution (E_Cl ~ 0 mV) but with 5 mM EGTA as the Ca²⁺ buffer (Fig. 5D). Using this pipette solution, CaCCinh-A01 did not inhibit feedback currents, consistent with results of Fig. 4B showing it also had little effect on tail currents with this pipette solution. These data suggest that the feedback currents observed with weak intracellular Ca²⁺ buffering are more strongly shaped by I_Cl(Ca). Our results confirm and extend earlier studies suggesting a role for Ano2 I_Cl(Ca) in feedback (Barnes and Deschenes, 1992; Endeman et al., 2012; Kraaij et al., 2000; Packer et al., 2010; Thoreson and Burkhardt, 1991; Verweij et al., 2003).

3.4. Glutamate transporter chloride currents contribute to feedback currents in cones.

In addition to stimulating I_Cl(Ca), Ca²⁺ influx into cones stimulates the release of glutamate into the synaptic cleft. After release, glutamate molecules are recycled by excitatory amino acid transporters (EAAT) on the plasma membrane of glial Muller cells and photoreceptors. Activation of EAATs in cones by glutamate is accompanied by the opening of transporter-associated anion channels (Eliasof and Werblin, 1993). We tested a role for EAAT anion currents in feedback by using a glutamate transporter inhibitor, DL-TBOA (300 μM). TBOA can inhibit EAAT1-5, but is particularly effective in blocking EAAT2 (Shimamoto et al., 1998), the subtype found in cones (Eliasof et al., 1998b; Rowan et al., 2010; Schneider et al., 2014). We confirmed the ability to detect presynaptic glutamate transporter anion currents in cones by showing that inward tail currents in cones following depolarizing steps (50 ms, −74.5 to −14.5 mV) were reduced from 58.6 ± 12.9 pA to 4.4 ± 4.4 pA (N=5; not shown) after bath application of TBOA (300 μM). For this experiment, we enhanced anion currents in cones by using SCN⁻ as predominant anion in the patch pipette solution. TBOA did not cause a significant change in I_Ca measured in cones using a CsGluconate (5 mM EGTA) pipette solution and tested with a ramp voltage protocol (0.5 mV/ms; control: 139 ± 16.7 pA; TBOA: 125.8 ± 18.5 pA, P=0.1, paired t-test, N=5).

We measured effects of TBOA on feedback currents in cones using four different pipette solutions. We used pipette solutions with two values for E_Cl (−46 or 0 mV) to examine the Cl-dependence of feedback currents. We also tested two levels of intracellular Ca²⁺ buffering (0.05 or 5 mM EGTA). Strong buffering restricts Ca²⁺ diffusion whereas weak buffering promotes greater spread, allowing us to examine the impact of near and far Ca²⁺-sensitive sites on the cone response.

Figure 6A shows an example of feedback currents recorded using a pipette solution containing 5 mM EGTA and E_Cl = 0 mV. Hyperpolarizing a synaptically-coupled HC from −43 to −103 mV evoked an inward current in the cone (E_m = −33 mV) that was reversibly suppressed by TBOA (300 μM). Inward feedback currents increased with time constants of 35.7 ms and 31.3 ms in control and TBOA, respectively. Figure 6B plots the amplitude of feedback currents as a function of cone holding potential using this same pipette solution (with 5 mM EGTA and E_Cl = 0 mV). Consistent with contributions from I_Cl(Glu) to feedback, blocking glutamate transporters with TBOA significantly reduced inward feedback currents at −33 and −53 mV (Fig. 6B) (N=5 pairs; −53 mV: P=0.0472, paired t-test; −33 mV: P=0.0011). We observed partial recovery after washout.

Figure 6C plots the amplitude of feedback currents vs. membrane potential from cones recorded using CsGluconate-based pipette solutions where E_Cl = −46 mV and the EGTA concentration was 5 mM (Fig. 6C). To slow rundown of synaptic glutamate release during whole cell recording, we sometimes used a pipette solution in which 40 mM gluconate was replaced with glutamate (Bartoletti et al., 2010). Application of TBOA produced similar effects when using CsGluconate-based pipette solutions with (N=7) or without glutamate (N=4 pairs) so we combined data using both pipette solutions. When E_Cl was −46 mV, inhibiting glutamate transporter anion currents (I_Cl(Glu)) by bath application of TBOA (300 μM) significantly reduced feedback currents at a cone holding potential of −43 mV (N= 11 pairs; P=0.0003, ratio paired t-test) but increased inward feedback currents at more positive potentials (Fig. 6C). This suggests that TBOA inhibited a current that reverses from inward to outward around ~−30 mV. This is above the predicted value for E_Cl, but counter-transport of glutamate anions can shift the reversal potential for transporter currents towards more positive values (Wadiche et al., 1995). With supplemental glutamate in the pipette solution, feedback currents recovered after washout.

Figure 6D illustrates an experiment in which we lowered Ca²⁺ buffering to 0.05 mM EGTA using the CsCl-based pipette solution. The feedback current increased slowly with a time constant of 264 ms in control conditions. After treatment with TBOA, the feedback current was smaller and slower (τ ~ 1400 ms). As mentioned when considering I_Cl(Ca), the slow feedback currents observed with this pipette solution often failed to attain a plateau during HC steps applied when the cone was held at −33 mV (Fig. 6D) so measurements at this potential under-estimated the peak amplitude attained by the feedback current. Nevertheless, feedback currents measured at a cone holding potential of −33 mV were significantly inhibited by TBOA (paired t-test, P=0.004) in cones with 0.05 mM EGTA and E_Cl = 0 mV (Fig. 6E). Without supplemental glutamate in the pipette (e.g., when E_Cl = 0), we typically observed only partial recovery after washout. Figure 6F plots the feedback current amplitude evoked when the cone pipette solution contained 0.05 mM EGTA and where E_Cl = −46 mV. This experimental condition most closely approximates endogenous values for E_Cl and Ca²⁺ buffering (Thoreson and Bryson, 2004; Van Hook and Thoreson, 2014). With this pipette solution containing supplemental glutamate, TBOA reversibly inhibited inward feedback currents at −43 mV (P=0.0077, paired t-test).

Using paired recordings, we can examine effects of feedback onto cones independent of feedforward effects of cones onto HCs. Nevertheless, blocking glutamate re-uptake with TBOA will elevate glutamate levels in the outer retina and this can depolarize other HCs that may also contact the voltage-clamped cone. This could in turn impact the overall strength of inhibitory feedback onto the voltage-clamped cone and lead to a reduction in I_Ca amplitude. To avoid such effects, we also tested TBOA after inhibiting HC glutamate receptors with CNQX. To ensure large feedback currents in these experiments, we used CsCl pipette solutions where E_Cl = 0 mV, but as before we also tested weak (0.05 mM EGTA) and strong (5 mM EGTA) Ca²⁺ buffering. By itself, blocking glutamate receptors on HCs with CNQX had no significant effect on the amplitude of cone I_Ca amplitude or cone feedback currents (CsCl, 5 mM EGTA; Figs. 7A-B). In vehicle control experiments (0.03% DMSO) with this same pipette solution, we also did not see a significant change in the amplitude of feedback currents (N=3). Although feedback currents did not decline when using this pipette solution, EPSCs in HCs declined by 60% during the period of drug application and another 38% during recovery from washout (final amplitude 22% of control), indicating a decline in feedforward glutamate release.

Figure 7. Glutamate transporter chloride currents contribute to feedback currents in cones.

(A) I_Ca amplitude in 5 cones in control conditions, after application of CNQX (100 μM), after combined application of CNQX and TBOA (300 μM), and following washout of both drugs. The small differences in amplitude among groups were not statistically significant by paired t-tests. I_Ca was measured with a ramp voltage protocol using a pipette solution containing CsCl and 5 mM EGTA. (B) Feedback current amplitude plotted against cone membrane potential comparing control (black squares; N=6), CNQX (100 μM; filled red triangles), and wash (open gray triangles). Pipette solution: 5 mM EGTA with E_Cl = 0 mV. (C) Feedback currents in a cone voltage-clamped at −33 mV and recorded with a pipette solution containing CsCl and 5 mM EGTA in the presence of CNQX (100 μM, black trace), CNQX (100 μM) + TBOA (300 μM, red), and after washout (gray). Time constants in CNQX and CNQX plus TBOA were 76 and 155 ms, respectively. (D) Feedback current amplitude plotted against cone membrane potential comparing CNQX (black squares; N=5), CNQX + TBOA (filled red triangles), and wash (open gray triangles). Pipette solution: 5 mM EGTA with E_Cl = 0 mV. Feedback currents evoked when cones were held at −33 mV were significantly reduced by TBOA (paired t-test, P = 0.0098). (E) Feedback currents in a cone voltage-clamped at −53 mV and recorded with a pipette solution containing CsCl and 0.05 mM EGTA in the presence of CNQX (100 μM, black trace), CNQX (100 μM) + TBOA (300 μM, red), and after washout (gray). (F) Feedback current amplitude plotted against cone membrane potential comparing CNQX (black squares; N=6), CNQX + TBOA (100 μM; filled red triangles), and wash (open gray triangles). Pipette solution: 0.05 mM EGTA with E_Cl = 0 mV. Feedback currents evoked when cones were held at −53 mV were significantly reduced by TBOA (paired t-test, P = 0.0067).

When using 5 mM EGTA as the Ca²⁺ buffer with E_Cl at 0 mV, application of TBOA in the presence of CNQX inhibited feedback when the cone was held at both −33 and −53 mV (Figs. 7C-D). TBOA-sensitive feedback currents did not recover after washout of both CNQX and TBOA, perhaps due to rundown of presynaptic glutamate release during these protracted experiments. This is similar to effects of TBOA in the absence of CNQX (Fig. 6B) and contrasts with effects of CaCCinh-A01 that did not inhibit feedback currents when using this same pipette solution (Fig. 5D). This suggests that when the spread of Ca²⁺ within cone terminals is limited to regions close to Ca²⁺ channels, glutamate transporter anion currents are an important component of feedback currents.

Using a pipette solution containing CsCl and 0.05 mM EGTA, application of TBOA in the presence of CNQX caused a significant reduction in feedback currents only when cones were held at −53 mV (Figs. 7E-F). Figure 7E shows an example in which hyperpolarizing the HC evoked a large, inward feedback current in the cone held at −53 mV (in the presence of CNQX). This current was substantially inhibited by TBOA and then recovered after washout. While TBOA reversibly inhibited feedback currents when the cone was held at −53 mV, it had little impact on feedback currents evoked when the cone was held at −33 mV. Application of CNQX hyperpolarizes HCs, and while the change in I_Ca amplitude did not attain statistical significance, the changes in feedback strength caused by hyperpolarization of surrounding HCs with CNQX caused a significant negative activation shift in cone I_Ca (−7.2 mV ± 1.61 mV, N=15 pairs, P=0.042, paired t-test). This negative activation shift in I_Ca may explain the smaller impact of TBOA on feedback at −33 mV and enhanced impact at −53 mV in the presence of CNQX.

3.5. Differences in rundown also suggest that I_Cl(Glu) and I_Cl(ca) both contribute to feedback

In another approach to analyzing the contribution of different currents to feedback, we compared the rundown of different current components during paired recordings. To generate large feedback currents suitable for monitoring rundown, we used the cone pipette solution containing CsCl and 0.05 mM EGTA. We measured EPSCs in HCs, inward feedback currents in cones, and depolarization-evoked cone tail currents using the voltage protocols illustrated in Fig. 2 except that we only tested a single cone membrane potential by stepping the cone from −73 to −33 mV. We waited three minutes and then tested feedback-induced changes in voltage dependence and amplitude of cone I_Ca. Similar to Fig. 3, we measured changes in I_Ca by applying ramp voltage protocols (−103 to +47 mV, 0.5 mV/ms) to the cone while holding the HC alternately at −103 or −43 mV. We waited one minute between each of the two ramps and switched the order between each pair of HC test steps. We then determined the changes in peak amplitude and midpoint activation voltage for I_Ca in the cone produced by these changes in HC holding potential.

As shown in Fig. 8A, feedback-induced changes in voltage-dependence and amplitude of cone I_Ca showed little or no rundown, remaining relatively stable for more than 20 min (Fig. 8A). By contrast, glutamate release assessed from HC EPSCs evoked by steps to −33 mV in cones ran down with a time constant of 6.1 min and was abolished after 25 min (Fig. 8B, solid squares). Thus, while the presence of a synapse between the cone and HC appears to be necessary, feedback-induced changes in cone I_Ca persist after the loss of presynaptic glutamate release (Cadetti and Thoreson, 2006).

Figure 8. Kinetics of rundown also suggest that I_Cl(Glu) and I_Cl(Ca) both contribute to feedback currents in cones.

Feedback currents, HC EPSCs, and tail currents were measured in cones using the protocol illustrated in Figure 2. Feedback-evoked changes in cone I_Ca were measured with voltage ramps (0.5 mV/ms, −103 to +47 mV) applied to the cone while holding the HC alternately at −103 or −43 mV (1 min interval between ramps). We alternated the sequence of HC holding potentials between each pair of ramps. Measurements of inward feedback currents were alternated with measurements of I_Ca every 3 min for 25 min. (A) Change in amplitude (open circles, dashed line) and shift in voltage dependence (filled triangles, solid line) of I_Ca caused by hyperpolarizing a post-synaptic HC from −43 to −103 mV. (B) Amplitude of HC EPSCs evoked by stepping a cone to −33 mV (filled squares, dashed line), inward feedback currents evoked in cones held at −33 mV by stepping the HC from −43 to −103 mV (filled circles, solid line, I_feedback), and inward tail currents evoked by returning the cone membrane potential from −33 to −73 mV (open triangle, solid line, I_tail). All data plotted as a function of time. Cone pipette solution contained CsCl with 0.05 mm EGTA.

Cone tail currents that are largely due to I_Cl(Ca) (Fig. 4) showed less rundown than HC EPSCs. Tail currents evoked by a 2 s step to −33 mV declined by only 23% in 25 min (Fig. 8B, open triangles).

While feedback-induced changes in the amplitude and voltage-dependence of I_Ca did not show rundown, inward feedback currents in cones diminished during recording with a time constant of 12.5 min. Feedback currents were evoked by applying hyperpolarizing steps (−43 to −103 mV) to the HC while a presynaptic cone was voltage-clamped at −33 mV. Feedback currents declined at a rate intermediate between the decline in glutamate release and the decline in I_Cl(Ca), falling by 64% in 25 min. This provides further support to the hypothesis that I_Ca, glutamate transporter anion currents, and I_Cl(Ca) all contribute to feedback currents.

3.6. Feedback kinetics

We examined the contribution of these two Cl⁻ conductances to feedback kinetics. To obtain large currents for reliable measurement, we focused on CsCl-based pipette solutions. We determined the single exponential function that provided the best fit to feedback currents evoked at a cone holding potential of −33 mV. As illustrated by the examples in Figure 6, feedback currents showed significantly faster kinetics when the cone pipette solution was buffered with 5 mM EGTA (E_Cl = 0 mV; τ = 59.6 ± 6.5 ms, N=34 cone-HC pairs) than when using 0.05 mM EGTA (E_Cl = 0 mV; τ = 208.6 ± 19.8 ms, N=42; P <0.0001, unpaired t-test). With the CsCl, 5 mM EGTA pipette solution, inhibiting glutamate transporter anion currents with TBOA while blocking glutamate receptors with CNQX slowed the dominant feedback time constant (τ) from 38.5 ± 16.2 to 111.9 ± 22.1 ms (N=5; P=0.0067, paired t-test; Fig. 9A). However, when using the 0.05 mM EGTA/CsCl pipette solution, TBOA applied in the presence of CNQX did not significantly alter feedback time constants (Fig. 9B). This is consistent with the stronger effect on feedback current amplitude in cones seen with TBOA when using the 5 mM EGTA pipette solution (Fig. 7D) compared to 0.05 mM EGTA (Fig. 7F). Conversely, the Ca²⁺-activated Cl⁻ channel inhibitor CaCCinh-A01 (50 μM) did not significantly alter τ when using the 5 mM EGTA pipette solution, but shortened the time constant from 307 ± 75.2 ms (N=6) to 104.2 ± 16.9 ms (P=0.04, paired t-test) when using 0.05 mM EGTA. This is consistent with the stronger effects of CaCCinh-A01 on feedback current amplitude observed in cones held at −33 mV when using 0.05 μM EGTA (Fig. 5B) than when using 5 mM EGTA (Fig. 5D) in the pipette solution. Taken together, these data suggest that the more rapid kinetics of HC feedback seen when using 5 mM EGTA to restrict the spread of Ca²⁺ in cone terminals are shaped by contributions from I_Cl(Glu) whereas the slower kinetics seen with 0.05 mM EGTA reflect increased contributions from I_Cl(Ca).

Figure 9. I_Cl(Glu) and I_Cl(Ca) contribute to different kinetic components of feedback.

(A) CaCCinh-A01 (50 μM) did not significantly alter the dominant feedback time constant (τ) measured at a holding potential of −33 mV compared to the prior control when using the 5 mM EGTA pipette solution with E_Cl = 0 mV (N=5). Control: 61.4 ± 29.2 ms; CaCCinh-A01: 56.4 ± 19.9 ms, N=5. Inhibiting I_Cl(Glu) with TBOA (300 μM) while blocking glutamate receptors with CNQX (100 μM) slowed τ from 38.5 ± 16.2 to 111.9 ± 22.1 ms (N=5; P=0.0067, paired t-test). (B) When using the CsCl, 0.05 mM EGTA pipette solution, CaCCinh-A01 (50 μM) shortened the time constant from 307 ± 75.2 ms to 104.2 ± 16.9 ms (N=6, P=0.04, paired t-test). With this same pipette solution, TBOA applied in the presence of CNQX did not significantly alter the feedback time constant. CNQX: 193.1 ± 55.7 ms; CNQX+TBOA: 225.7 ± 34.8 ms, N=12.

4. Discussion

When HCs hyperpolarize to light, lateral inhibition from HCs to cones increases the amplitude of I_Ca and lowers its activation threshold (Verweij et al., 1996). Our results confirm earlier studies suggesting that inward feedback currents initiated by these changes in cone I_Ca can be amplified by I_Cl(Ca) (Barnes and Deschenes, 1992; Endeman et al., 2012; Kraaij et al., 2000; Thoreson and Burkhardt, 1991; Verweij et al., 2003) but also show a role for Cl⁻ channels coupled to glutamate transporters, consistent with Vroman and Kamermans (2015). Furthermore, we found that I_Cl(Ca) contributes preferentially when Ca²⁺ is allowed to spread throughout the terminal whereas I_Cl(Glu) is more sensitive to local Ca²⁺ changes near the ribbon.

In salamander photoreceptors, antibodies to anoctamin 1 (Ano1 or TMEM16A) have been shown to label rod and cone terminals (Mercer et al., 2011b). In mammalian retina, there is evidence for both Ano1 and Ano2 (TMEM16B) (Jeon et al., 2013; Stohr et al., 2009). In rat retina, Ano2 is expressed in rods but not cones (Dauner et al., 2013). The inhibitor, Ani9, is selective for Ano1 over Ano2 (Seo et al., 2016) whereas CaCCinh-A01 shows similar potency at both types (Cherkashin et al., 2016). The finding that CaCCinh-A01 inhibited I_Cl(Ca) in cones stronger than Ani9 thus suggests that Ano2 is the major subtype responsible for I_Cl(Ca) currents in cones.

It had been previously shown that activation of I_Cl(Glu) by glutamate released from neighboring cones contributes to feedback currents evoked by illumination of the receptive field surround (Vroman and Kamermans, 2015). Our experiments using simultaneous paired whole cell recordings from HCs and cones allowed us to study the role of glutamate at synapses between individual HCs and cones. We found that using TBOA to block I_Cl(Glu) significantly reduced cone feedback currents evoked by changes in HC membrane potential. Some of the reduction in feedback currents produced by TBOA may involve a reduction in cone I_Ca caused by enhanced feedback inhibition that would accompany the increase in synaptic cleft glutamate levels and the attendant depolarization of nearby HCs. However, TBOA also inhibited feedback after blocking glutamate receptors on HCs with CNQX to eliminate this interaction. It is unlikely that effects of TBOA on feedback reflect actions on glutamate transporters in horizontal cells which, unlike cat and rat retina (Fyk-Kolodziej et al., 2004; Rauen et al., 1996; Reye et al., 2002), are not thought to be present in salamander retina (Eliasof et al., 1998a). Thus, our results show that in addition to activation by spillover from neighboring cones, feedback-induced changes in cone I_Ca can alter glutamate release that in turn shapes feedback currents by local activation of I_Cl(Glu) in the same cone.

It has been hypothesized that currents flowing through post-synaptic glutamate receptors on HC dendrites might contribute to feedback by generating extracellular, ephaptic voltage changes (Byzov and Shura-Bura, 1986; Kamermans et al., 2001). When evoked by surround illumination, glutamate antagonists block HC to cone feedback by antagonizing feedforward glutamate release from cones onto HCs (Thoreson and Burkhardt, 1990; Verweij et al., 1996; Vroman and Kamermans, 2015). By using paired recordings from cones and HCs, we were able to test whether glutamate receptors participate directly in feedback at the synapse from HCs back onto cones. Control experiments showed that glutamate release is likely to have declined by 60% during the period of CNQX application which would diminish any impact of glutamate. Nevertheless, our results showed that while CNQX slightly increased the overall amplitude of I_Ca as expected from its ability to hyperpolarize other HCs that contact the same cone, it did not significantly alter feedback effects between the voltage-clamped HC and its paired cone.

Endogenous values of E_Cl are close to the resting membrane potential of cones in darkness (Thoreson and Bryson, 2004). When cones are hyperpolarized by light, E_Cl would thus be more positive than the membrane potential and so activating Cl⁻ currents would depolarize the cone membrane. Using a cone pipette solution where E_Cl = −46 mV, close to endogenous values, we saw inward feedback currents below −30 mV. With this pipette solution, the reduction in feedback currents produced by TBOA suggested that I_Cl(Glu) reversed around −30 mV, consistent with evidence that the net reversal potential for transporter currents is typically 15-20 mV more positive than E_Cl due to the inward counter-transport of glutamate anions (Wadiche and Kavanaugh, 1998).

Our results show that at least three different currents contribute to the depolarizing cone response evoked by negative feedback from post-synaptic HCs. When a HC hyperpolarizes to light, feedback to a cone increases the amplitude of I_Ca within the physiological voltage range. This is a consequence of a lower threshold for activation and increase in the peak amplitude of I_Ca. While the mechanisms for these changes in I_Ca are not entirely understood, it is generally agreed that they involve extracellular pH changes in the synaptic cleft (Chapot et al., 2017; Kramer and Davenport, 2015). The increase in Ca²⁺ influx into a cone during feedback stimulates glutamate release. The membrane voltage changes initiated by changes in I_Ca caused by HC feedback are further shaped by both I_Cl(Ca) and I_Cl(Glu). Below, we discuss how these two currents may contribute under different conditions.

Glutamate release from cones occurs entirely at ribbons (Snellman et al., 2011; Van Hook and Thoreson, 2015) and cone Ca²⁺ channels are clustered beneath ribbons (tom Dieck et al., 2005). Release of vesicles is regulated by highly localized Ca²⁺ nanodomains beneath these channels (Bartoletti et al., 2011; Mercer et al., 2011a). Ca²⁺-activated Cl⁻ channels are not localized to ribbons, but distributed more diffusely throughout cone terminals (Mercer et al., 2011b). Consistent with this, contributions of I_Cl(Glu) to feedback currents were more prominent when Ca²⁺ was buffered with 5 mM EGTA whereas I_Cl(Ca) played a bigger role when buffering was lowered to 0.05 mM EGTA, allowing Ca²⁺ to spread further throughout the terminal. Brief changes in HC membrane potential will lead to brief changes in the strength of feedback. Brief feedback-induced changes in cone I_Ca will in turn tend to have a bigger impact on Ca²⁺ levels close to ribbons and thus a bigger impact on I_Cl(Glu) than I_Cl(Ca). Larger, more protracted changes in cone I_Ca caused by larger, more protracted changes in HC membrane potential will have a bigger impact on Ca²⁺ levels far from the ribbon. Endogenous buffering is weak, equivalent to 0.05-0.1 mM EGTA, favoring the spread of Ca²⁺ and enhancing contributions from I_Cl(Ca) (Van Hook and Thoreson, 2014).

Blocking I_Cl(Glu) when Ca²⁺ changes were restricted near ribbons with strong Ca²⁺ buffering led to a slowing of feedback kinetics suggesting that the rapid activation of I_Cl(Glu) speeds feedback-induced depolarization of cones. However, sustained elevation of intracellular Ca²⁺ will deplete the releasable pool of glutamate-filled vesicles along the base of the ribbon (Jackman et al., 2009), causing I_Cl(Glu) to decline. Sustained illumination of the surround also leads to a decline in I_Cl(Glu) (Vroman and Kamermans, 2015). However, as I_Cl(Glu),is declining, sustained elevation of Ca²⁺ will begin to saturate endogenous buffers, allowing Ca²⁺ ions to spread further into the terminal and stimulate Ca²⁺-activated Cl⁻ channels. I_Cl(Ca) is activated more slowly than I_Cl(Glu), contributing to slower components of the feedback depolarization. The slow activation of I_Cl(Ca) would help to sustain inward feedback currents even after glutamate release and I_Cl(Glu) have diminished following depletion of the readily releasable pool of vesicles.

Feedback arising from different HCs can independently modulate activity at different ribbons in the same cone terminal (Grassmeyer and Thoreson, 2017). Thus, the strength of HC feedback on cone I_Ca and glutamate release can vary among different ribbons in the same cone if those ribbons contact different HCs receiving different stimuli. Unlike Ca²⁺ changes close to individual ribbons that can be influenced by feedback from individual HCs, regions of the synaptic terminal between ribbons will be impacted by intracellular Ca²⁺ derived from multiple nearby ribbons. And while I_Ca can be regulated at individual ribbons, the glutamate transporters on the cone can be impacted by extracellular glutamate released by multiple nearby ribbons. Thus, both I_Cl(ca) and I_Cl(Glu) may both be influenced by feedback-induced changes at multiple cone ribbons. Diffusion of glutamate released from adjacent cones can also act through I_Cl(Glu) to adjust the cone membrane potential in response to changes in luminance (Vroman and Kamermans, 2015). The combination of different components with different kinetics allows cones to adjust their output in response to changes in the temporal and spatial properties of illumination.

Highlights:

• Feedback from horizontal cells to cones is essential for color and contrast vision.

• Horizontal cell feedback acts on multiple cone currents.

• Calcium currents and glutamate transporter currents show fast kinetics.

• Calcium-activated chloride currents show slower kinetics.

• Feedback adjusts cone responses to temporal and spatial changes in luminance.

Abbreviations:

HC

Horizontal cell

I_Cl(Ca)

Ca²⁺-activated Cl⁻ current

I_Cl(Glu)

Glutamate transporter anion current