Abstract
Synapses in the inner plexiform layer of the retina undergo short-term plasticity that may mediate different forms of adaptation to regularities in light stimuli. Using patch-clamp recordings from axotomized goldfish Mb bipolar cell (BC) terminals with paired-pulse light stimulation, we isolated and quantified the short-term plasticity of GABAergic lateral IPSCs (L-IPSCs). Bright light stimulation evoked ON and OFF L-IPSCs in axotomized BCs, which had distinct onset latencies (∼50–80 and ∼70–150 ms, respectively) that depended on background light adaptation. We observed plasticity in both the synaptic strength and latency of the L-IPSCs. With paired light stimulation, latencies of ON L-IPSCs increased at paired-pulse intervals (PPIs) of 50 and 300 ms, whereas OFF L-IPSC latencies decreased at the 300 ms PPI. ON L-IPSCs showed paired-pulse depression at intervals <1 s, whereas OFF L-IPSCs showed depression at intervals ≤1 s and amplitude facilitation at longer intervals (1–2 s). This biphasic form of L-IPSC plasticity may underlie adaptation and sensitization to surround temporal contrast over multiple timescales. Block of retinal signaling at GABAARs and AMPARs differentially affected ON and OFF L-IPSCs, confirming that these two types of feedback inhibition are mediated by distinct and convergent retinal pathways with different mechanisms of plasticity. We propose that these plastic changes in the strength and timing of L-IPSCs help to dynamically shape the time course of glutamate release from ON-type BC terminals. Short-term plasticity of L-IPSCs may thus influence the strength, timing, and spatial extent of amacrine and ganglion cell inhibitory surrounds.
Introduction
Inhibitory networks in the inner plexiform layer (IPL) of the retina, consisting of amacrine cell (AC) inputs to bipolar cell (BC) presynaptic terminals, serial synapses between AC dendrites, and direct inputs to ganglion cell (GC) dendrites, perform computations that shape GC output to the brain. For example, these inputs mediate spatial integration and refine GC center-surround receptive fields (Cook and McReynolds, 1998; Jacobs and Werblin, 1998; Ichinose and Lukasiewicz, 2005) and GC orientation selectivity (Venkataramani and Taylor, 2010). The goldfish Mb-type BC, a counterpart of the mammalian rod BC (RBC) (Euler and Masland, 2000; Joselevitch and Kamermans, 2009; Snellman et al., 2009), depolarizes in response to light stimulation (Wong et al., 2005) and has a large (∼10 μm) synaptic terminal that stratifies in sublamina b (ON layer) of the IPL (Witkovsky and Dowling, 1969). It is thus possible to patch clamp a single Mb BC terminal in retinal slices (Palmer et al., 2003), which makes it an excellent model for the study of inhibitory processing in the IPL.
Each Mb terminal makes ∼50 synapses with GC dendrites and ∼300 synapses with AC boutons, of which ∼50% are reciprocal and ∼50% are lateral (Marc and Liu, 2000). Reciprocal synapses use AMPA and NMDA receptors on the AC and GABAAR and GABACR synapses on the Mb (Vigh and von Gersdorff, 2005), whereas lateral synapses exhibit unidirectional release of GABA onto the Mb (Marc and Liu, 2000). Several studies provide evidence that lateral and reciprocal synapses are spatially and functionally distinct (Marc and Liu, 2000; Zhang and Wu, 2009; Vigh et al., 2011). Immunohistochemical (Koulen et al., 1998) and electrophysiological (Palmer, 2006) evidence suggests that GABAA and GABAC receptors are present at separate synapses. We have shown recently that light-evoked lateral IPSCs (L-IPSCs) at the Mb terminal consist of GABAA and GABAC currents, which arise from pathway-specific ON or crossover OFF inputs (Vigh et al., 2011).
Short-term depression (STD) of reciprocal inhibition at AC → BC synapses is thought to modulate BC responses during high-frequency light stimulation (Li et al., 2007) and has been shown to prevent depression of excitatory ON cone BC → GC synapses in the mouse (Sagdullaev et al., 2011). However, the plasticity of L-IPSCs at the Mb terminal has not been explored previously. Here, we recorded directly from axotomized Mb terminals, which allowed us to quantify L-IPSCs in the absence of reciprocal feedback, and stimulated with pairs of full-field light flashes to characterize L-IPSC plasticity. We found that short-term plasticity (STP) differed for ON and OFF responses and was altered differentially by blockage of GABAA and AMPA receptors. Such dynamic regulation of lateral inhibition at BC terminals likely mediates rapid surround modulation and adaptation of GC and AC responses. We propose that STP of L-IPSCs acts to adjust the strength and spatial extent of postsynaptic AC and GC inhibitory surrounds by regulating both Mb BC glutamate release during feedforward subthreshold depolarization (Joselevitch and Kamermans, 2007) and the ability of Mb terminals to initiate regenerative Ca2+ spikes (Zenisek and Matthews, 1998; Protti et al., 2000; Arai et al., 2010; Baden et al., 2011).
Materials and Methods
Retinal slice preparation and electrophysiology.
Slices were prepared from pieces of goldfish (Carassius auratus; 8–14 cm) retina, as described previously (Palmer et al., 2003). For experiments in which paired recordings of ACs and Mb BC terminals were performed, the goldfish retina was gently removed from the eyecup and embedded in low-gelling-temperature agar (3% w/v in slice solution; agarose type VII-A, A-0701; Sigma-Aldrich), as described by Kim et al. (2012). For all experiments, goldfish, of either sex, were dark adapted for 1–2 h before dissection, and slicing solution contained the following (in mm): 119.0 NaCl, 2.5 KCl, 3.2 MgCl2, 0.3 CaCl2, 12.0 glucose, 12.0 HEPES, and 0.2 ascorbic acid. The pH was set to 7.45 with NaOH, and osmolarity was set to 260–265 mOsm. Dissection was performed under either dark or dim red light conditions, and recording was performed under either mesopic (1.01 × 103 to 5.03 × 103 photons μm−2 s−1) or scotopic (5.03 × 101 photons μm−2 s−1) background conditions (Krizaj, 2000). Background light levels were measured with an ILT-1700 photometer and SE033 detector from International Light Technologies. Transverse slices (250 μm thick) were cut with a Narishige ST-20 vertical slicer and transferred to a Sylgard (Dow Corning) recording chamber, in which they were secured in parallel lanes of petroleum jelly. For paired recordings, a diced agar block containing the retina was cut into 200-μm-thick slices using a Vibratome slicer (VT1000; Leica). The chamber and slices were then moved to an upright microscope (Axioskop; Carl Zeiss), in which the slices were constantly perfused with external recording solution bubbled with 95% O2 and 5% CO2 mixed gas at 4–5 ml/min and viewed with infrared differential interference contrast (IR-DIC) optics through a 40× water-immersion objective (Carl Zeiss) coupled with 2.0 pre-magnification and an IR CCD camera (XC-75; Sony). The output of the CCD camera was sent to a Hamamatsu Camera Controller C2741-62 and then to a 13-inch Sony black and white monitor for viewing. Axotomized BC terminals (i.e., with severed axons) were identified in the IPL based on the following: (1) Mb-shaped (bulbous) terminal morphology (Fig. 1A), (2) single-exponential membrane time constant (data not shown), and (3) the presence of an L-type Ca2+ current, reciprocal feedback, and ΔCm jump (Fig. 1B; Palmer et al., 2003) associated with exocytosis (von Gersdorff and Matthews, 1999).
Axotomized BC terminals in retinal slices were voltage clamped at −60 or −70 mV (uncorrected for liquid junction potential) in the whole-cell mode using a HEKA Elektronik EPC-9 patch-clamp amplifier in conjunction with Pulse software running the xChart extension (Pulse version 8.53). The Sine+DC technique (Gillis, 2000) was used for real-time measurements of membrane capacitance. Briefly, a 30 mV peak-to-peak, 1–2 kHz sine wave was superimposed on the holding potential of the cells (−60 mV) and used by online analysis software to calculate time-resolved membrane capacitance. Standard external recording solutions contained the following (in mm): 100.0 NaCl, 2.5 KCl, 2.5 CaCl2, 1.0 MgCl2, 25.0 NaHCO3, 0.2 ascorbic acid, and 12.0 glucose, pH 7.45 (osmolarity, 260–265 mOsm). Patch pipettes (6–12 MΩ) were pulled, either using a vertical (PP-830; Narishige) or horizontal (P-97; Sutter Instruments) puller, from thick-walled, 1.5-mm-outer-diameter borosilicate capillary glass from World Precision Instruments (1B150F-4) and coated with dental wax (Cavex) to reduce pipette capacitance. Internal pipette solutions contained the following solution (in mm): 60.0 Cs gluconate, 40.0 CsCl, 10.0 TEA-Cl, 25.0 HEPES, 3.0 Mg-ATP, 0.5 Na-GTP, and 2.0 EGTA. Internal solutions for paired recordings contained 28.0 HEPES and 1.0 Na-GTP but were otherwise the same. Some internal solutions contained 3.0 mm ascorbic acid and/or 3.0 mm reduced glutathione. All internals were set to pH 7.2 with CsOH, and osmolarity was adjusted to 250 mOsm with Cs gluconate and/or millipure H2O. NBQX, TTX, TPMPA [(1,2,5,6-tetrahydropyridin-4-yl)methylphosphinic acid], and SR-95531 [6-Imino-3-(4-methoxyphenyl)-1(6H)-pyridazinebutanoic acid hydrobromide] (SR) were obtained from Tocris Bioscience. All other chemicals were obtained from Sigma.
Recordings were performed at 20–22°C during the daytime (morning/afternoon) to avoid circadian variation in glutamate release from BCs (Hull et al., 2006b). Voltage-clamp series resistance (Rs) was not compensated, and liquid junction potential was not corrected. Cells with Rs > 30 MΩ or resting leak >50 pA at a holding potential of −60 mV were excluded from analysis. Average values for Rs and leak current at a holding potential of −60 mV were 22.0 ± 1.2 MΩ and −34.2 ± 2.6 pA, respectively (mean ± SE; n = 75).
Light stimulation.
Slices were stimulated with a white LED connected via soldered wire and BNC cable to a digital-to-analog output of the HEKA EPC-9 amplifier, as described previously (Vigh et al., 2011). The LED was positioned at an ∼30° angle above and behind the recording chamber, at a distance of 3 cm. Full-field light flashes for all experiments were delivered by application of voltage steps from 0 to between 3 and 5 V, which evoked a photon flux at the slice of between 5.69 × 104 and 7.32 × 105 photons μm−2 s−1 (unless otherwise noted), which is above the threshold for cone photoreceptor activation (∼102 photons μm−2 s−1; Busskamp et al., 2010). The timing and amplitude of voltage steps was controlled from within the Pulse Software (HEKA) running the EPC-9 amplifier. Calibration of light flash timing (onset and offset) was performed with a photodiode connected to the EPC-9 amplifier through an ITHACO 4302 dual 24 dB octave filter. Light flash onset and offset had rise and decay time constants of 0.1 and 6.3 ms, respectively. Onset and offset times did not vary as a function of flash duration between 100 and 1000 ms or with the presentation of paired flashes with intervals ranging from 300 to 1900 ms. Calibration of light intensity was performed with an ILT-1700 photometer and SE033 detector from International Light Technologies. Factory calibration determined the photopic illuminance response sensitivity of the detector to be 2.60 × 10−8 A ft2 lm−1 or 2.415 × 10−9 A lux−1, assuming 3215 K color temperature.
Analysis of amplitude, charge, and onset latency of L-IPSCs.
Light responses at each paired-pulse interval (PPI) were repeated at least five times per cell and averaged, unless otherwise stated. An interval of 20 or 30 s was presented between stimulations to allow for recovery from depression in neighboring Mb presynaptic terminals (von Gersdorff and Matthews, 1997). Under these conditions, light responses were generally stable for up to 30 min. Nonetheless, stimulations with PPIs ranging from 50 to 2300 ms were interleaved to avoid systematic errors attributable to rundown. For PPIs shorter than 300 ms, alternating single and double flashes were presented so that single responses could be subtracted from double responses. This allowed for isolation of the first OFF response and second ON response so that amplitude, charge transfer, and onset latencies could be reliably measured at short intervals (i.e., 50 ms). L-IPSC amplitude was calculated by subtracting a baseline ON and OFF current (first 20 ms after onset and offset of light flash, respectively) from the peak current amplitude during the 400 ms light flash and the 300 ms after the offset of the light flash for ON and OFF responses, respectively. Charge transfer was calculated by integrating current traces during these ON (400 ms) and OFF (300 ms) response windows after baseline subtraction.
L-IPSC onset latencies were determined using a custom IGOR procedure that detected the time at which the L-IPSC crossed a threshold current set at baseline minus 2 SDs, at which the baseline and SD were calculated from the 20 ms period at the beginning of either the ON or OFF response windows. The threshold crossing was determined by moving backward in time from the peak of the L-IPSC, which was detected as the global minima of the current recording during the response window. Latencies calculated with a similar technique in which the threshold was set at 10% of the difference between baseline and peak current were not significantly different from those calculated with the 2 SD criteria. Manual “by eye” detection of latencies also yielded comparable results but showed statistically significant underestimation of paired Δ onset latency compared with both the 2 SD criteria for ON L-IPSCs at the 300 ms PPI (p < 0.01, unpaired Student's t test, two tailed) and the 10% baseline to peak criteria at the 50 ms (p < 0.05) and 300 ms (p < 0.05) PPI. None of the three techniques showed significant differences for OFF L-IPSC latencies at either the 50 or 300 ms PPIs (p > 0.05). We report latencies using the 2 SD criteria here because the results of this technique showed the fewest significant differences with the results of other techniques.
Immunohistochemistry and confocal imaging.
During paired patch-clamp recording, the Mb cell and AC were filled via the patch pipette with Alexa Fluor 555 and 488 hydrazide, respectively (100–200 μm; Invitrogen). Immediately after the recording, retinal slices were transferred into 4% (w/v) paraformaldehyde in 1× PBS (catalog #70013; Invitrogen) for 30 min. Slices were mounted onto Superfrost slides (Thermo Fisher Scientific) in aqueous mounting medium with GEL/MOUNT anti-fade solution (Biomeda). Alexa Fluor-containing Mb BCs and ACs were viewed with laser lines at 488 and 555 nm using a 40× water-immersion objective on a confocal laser-scanning microscope (LSM 710; Carl Zeiss). Stacked confocal images were reconstructed with Imaris software (Bitplane Scientific Software).
DIC/epifluorescence overlays.
During some recordings of single axotomized Mb terminals, the internal solution contained 150–200 μm Alexa Fluor 555 hydrazide. DIC montage images were recorded in iMovie (Apple) by connecting the Hamamatsu Camera Controller to a MacBook Pro (Apple) via a Dazzle Hollywood DV-Bridge frame grabber. Epifluorescence imaging was performed after recording by illumination of the preparation with an AttoArc HBO 100 W mercury lamp (Carl Zeiss). A filter cube matching the emission range of Alexa Fluor 555 was placed between the objective and the CCD camera. Fluorescence images were also acquired with iMovie and were later aligned with DIC montages using NIH ImageJ and Photoshop (Adobe Systems).
Statistical analyses.
Statistics were performed on averaged traces, unless otherwise noted, by using Prism (version 4; GraphPad Software). Two-tailed paired or unpaired Student's t tests were used to compare datasets when appropriate, and one-sample t tests with a hypothetical mean of 0 or 1 were used to test paired-pulse plasticity of onset latency or size, respectively, at each PPI. Data were reported as mean ± SEM, and values of p < 0.05 were considered statistically significant.
Results
Isolation of light-evoked GABAergic ON and OFF L-IPSCs at the Mb BC terminal
To examine the synaptic properties of light-evoked lateral inhibition, we performed direct whole-cell voltage-clamp recordings of axotomized goldfish Mb BC terminals (Fig. 1A,B). By recording with intracellular solution containing 40 mm CsCl and 10 mm TEA-Cl (ECl = −20.1 mV) and voltage clamping the axotomized terminal at −60 or −70 mV, we were able to evoke large inward L-IPSCs by stimulating the retinal slice with full-field, 400 ms duration, white light flashes (between 5.69 × 104 and 7.32 × 105 photons μm−2 s−1, unless otherwise noted; see Materials and Methods) under scotopic background conditions (5.03 × 101 photons μm−2 s−1; Fig. 1C). We refer to these events as L-IPSCs because they are expected to be inhibitory (or shunting) under physiological conditions. Previous studies have estimated ECl at the BC terminal to be between −42 and −60 mV (Billups and Attwell, 2002; Duebel et al., 2006), with a resting membrane potential (Em) of between −40 and −53 mV (Protti et al., 2000; Wong et al., 2005; Baden et al., 2011).
The control light response (Fig. 1C, black trace) consisted of distinct ON (∼50 ms latency) and OFF L-IPSCs (∼100 ms latency), which could both be blocked completely by bath application of 25 μm SR and 150 μm TPMPA (Fig. 1C, red trace). Similar results were obtained with 12.5 μm SR (data not shown), a concentration less likely to produce significant nonspecific block of glycine receptors (GlyRs; Wang and Slaughter, 2005). This block was reversible during washout (Fig. 1C, blue trace). This result confirmed our previous findings that both ON and OFF L-IPSCs at the Mb terminal consist entirely of GABAA and GABAC receptor-mediated responses and are not mediated by GlyRs (i.e., not blocked by 1 μm strychnine; Vigh et al., 2011).
In Figure 1D, we show a schematic of the patterns of connectivity that are likely responsible for propagation of ON and crossover OFF lateral inhibition from the dendrites of neighboring or distant ON and OFF BCs to the Mb terminal. ON lateral inhibition is driven by mixed rod and cone inputs (Vigh et al., 2011) and arrives via a direct trisynaptic pathway that passes through neighboring ON BCs and ON ACs. However, OFF lateral inhibition is primarily driven by cone inputs (Vigh et al., 2011) and travels via either a direct trisynaptic pathway or an indirect, disinhibitory multisynaptic pathway (Fig. 1D). The direct pathway likely consists of an input from an OFF BC to an OFF AC that forms synaptic contacts onto the Mb axon shaft or an input from an OFF BC to a bistratified or diffuse AC that makes synaptic contacts directly onto the Mb terminal. The indirect OFF pathway could arise from an OFF BC via contacts onto an OFF AC that then contacts either an ON diffuse or bistratified AC that makes direct synaptic contacts onto the Mb terminal. These indirect pathways, or nested loops of serial AC → AC GABAA synapses (Fig. 1D, dashed lines; Marc and Liu, 2000; Hsueh et al., 2008), may be either homotypic (i.e., ON → ON or OFF → OFF) or heterotypic (i.e., ON → OFF or OFF → ON) and have been shown to regulate the strength of BC inhibitory surrounds (Eggers and Lukasiewicz, 2010).
ON and OFF L-IPSCs exhibit different temporal profiles of STP
To determine the likely effects of dynamic surround stimulation on the size and timing of lateral feedback inhibition at the Mb terminal, we tested STP of L-IPSCs over the PPI range of 50–2300 ms. Here, PPI refers to the interval between the offset of the first light flash and the onset of the second light flash. Short PPIs (e.g., 300 ms; Fig. 2Ai) tended to produce STD of ON and OFF L-IPSC amplitude (∼20%) and charge (∼15%), whereas long PPIs (e.g., 1900 ms; Fig. 2Bi) tended to produce STD of ON L-IPSC amplitude and charge (10–15%) and short-term facilitation (STF) of OFF L-IPSC amplitude (∼25%). Subtraction of the first light response (both ON and OFF components; windowed from t = 0 ms at light onset to t = 700 ms at termination of OFF response; ON response ends at t = 400 ms, OFF response begins at t = 400 ms) from the second light response of the pair yielded difference traces that showed clear STD of ON and OFF L-IPSCs at a PPI of 300 ms (Fig. 2Aii; same cell as Ai), and a combination of ON STD and OFF STF at a PPI of 1900 ms (Fig. 2Bii; same cell as Bi). Note that, for the 300 ms interval, ON STD in the difference trace is seen as a positive deflection relative to a baseline with an initial negative offset attributable to incomplete decay of the first OFF L-IPSC (Fig. 2Aii). For this reason, STP at the 50 ms PPI was determined with a protocol of alternating single and double light flashes. This allowed for the isolation of first and second ON and OFF L-IPSCs by subtraction of the single response from the double response (see Materials and Methods).
ON and OFF L-IPSCs showed different temporal profiles of amplitude (Fig. 3A) and charge transfer STP (Fig. 3B). STP of both amplitude and charge transfer were quantified as paired-pulse ratios (PPRs), and depression or facilitation were defined as PPRs that were <1 or >1, respectively (Fig. 3, gray asterisks). In addition, we performed direct statistical comparisons between ON and OFF PPRs at each PPI (Fig. 3, black asterisks). ON L-IPSC amplitude depression was highly statistically significant at PPIs of 50 ms (PPR = 0.57 ± 0.05; n = 14; p < 0.001, one-sample t test) and 300 ms (0.78 ± 0.06; n = 34; p < 0.001) recovered at 1100 ms (p > 0.05) and showed small but significant depression at 1900 ms (0.87 ± 0.04; n = 39; p < 0.01) that recovered by 2300 ms. In contrast, OFF amplitude depression was sometimes strong at a PPI of 50 ms, although high variability at this short PPI prevented the depression from reaching statistical significance (PPR = 0.94 ± 0.16; n = 6; p > 0.05). OFF depression was present at 300 ms (0.82 ± 0.07; n = 32; p < 0.05) but was replaced by facilitation at 1100 ms (1.21 ± 0.10; n = 32; p < 0.05) that continued at 1900 ms (1.26 ± 0.09; n = 37; p < 0.01) before recovery at 2300 ms. ON L-IPSCs were significantly depressed relative to OFF L-IPSCs at PPIs of 50 ms (ΔPPR = −39.4%; p < 0.01) and 1900 ms (Δ PPR = −30.6%; p < 0.001, unpaired Student's t test, two tailed). These differences in the STP profiles of ON and OFF L-IPSC amplitude likely allow them to act as independent temporal filters on the early and late phases, respectively, of light responses at Mb terminals. When taken together, they suggest a mechanism for regulation of the balance between surround temporal contrast adaptation (i.e., OFF facilitation) and sensitization (i.e., ON and OFF depression) at Mb → GC and Mb → AC inputs. This balance is likely adaptive during rapid transitions between visual environments containing high and low contrast (Kastner and Baccus, 2011).
Next, we analyzed the total integrated charge transfer of baseline-subtracted ON and OFF L-IPSCs (Fig. 3B). ON L-IPSC charge transfer exhibited strong depression at a PPI of 50 ms (PPR = 0.39 ± 0.08; n = 14; p < 0.001, one-sample t test). This depression recovered at 300 ms, and small but significant depression reemerged at PPIs of 1100 ms (0.78 ± 0.08; n = 34; p < 0.05) and 1900 ms (0.79 ± 0.06; n = 39; p < 0.01). ON depression was not significant at 2300 ms. OFF charge transfer STP (Fig. 3B) followed a time course of recovery from depression that was similar to that of the ON response, with strong depression at 50 ms (PPR = 0.47 ± 0.20; n = 6; p < 0.05), recovery at 300 ms, moderate depression at 1100 ms (0.78 ± 0.06; n = 32; p < 0.001), and recovery at 1900 and 2300 ms PPIs. Similarly to the result for amplitude PPR, ON L-IPSCs were significantly depressed relative to OFF L-IPSCs at the 1900 ms PPI (ΔPPR = −24.9%; p < 0.05, unpaired Student's t test, two tailed).
ON and OFF L-IPSCs exhibit differential onset latency STP
Interestingly, ON and OFF L-IPSCs also exhibited distinct STP of onset latency (Fig. 4). Such changes in the timing of presynaptic lateral inhibition may strongly determine the degree to which L-IPSCs inhibit or shunt light-evoked EPSPs (Oltedal et al., 2009) or light-evoked Ca2+ action potentials (Protti et al., 2000) at the Mb BC terminal, whether ACs act to either inhibit or disinhibit GC spiking (Manu and Baccus, 2011), and whether L-IPSCs primarily shape BC representations of either temporal contrast or mean luminance (Oesch and Diamond, 2011). To quantify the STP of onset latency, we first identified ON and OFF L-IPSC onset latencies as the crossing point of a threshold defined as baseline current minus 2 SDs of the baseline, working backward from the peak of the L-IPSC toward the beginning of the response window (see Materials and Methods; Fig. 4Ai,Aii, vertical dotted lines).
We found that ON and OFF L-IPSCs exhibited different forms of onset latency plasticity. ON L-IPSCs showed a strong paired-pulse delay (i.e., increase) in onset latency (Fig. 4Ai,Bi, blue) at PPIs of 50 ms (Δ onset latency = +75.0 ± 28.3 ms; n = 9; p < 0.05, one-sample t test, green asterisks) and 300 ms (+34.0 ± 11.3 ms; n = 34; p < 0.01). These effects were large relative to the mean onset latency of the first L-IPSCs of each pair (50 ms PPI, +96.0 ± 38.6%; 300 ms PPI, +60.2 ± 18.3%; Fig. 4Bii). ON onset latency STP was not significant at PPIs of 1100 and 1900 ms (data not shown). This delay of ON L-IPSCs at short PPIs may be caused by depletion of the readily releasable vesicle pool at ON BC ribbon synapses (von Gersdorff and Matthews, 1997), which would result in a strong reduction of the initial, fast component of glutamate release onto AC dendrites.
OFF L-IPSCs, which may be mediated by a class of ACs with release properties that are different from those of the ACs that mediate ON L-IPSCs (Vigh et al., 2011), showed paired-pulse advances (i.e., decreases) in onset latency at short PPIs. OFF onset latency (300 ms PPI example; Fig. 4Aii) showed a significant advance at the 300 ms PPI (Δ onset latency = −29.8 ± 10.1 ms; n = 31; p < 0.01, one-sample t test; Fig. 4Bi, red). There was no significant onset latency STP of OFF L-IPSCs at the 50 ms PPI (+5.4 ± 3.2 ms; n = 5; p > 0.05) or at the 1100 and 1900 ms PPIs (data not shown). When normalized to first L-IPSC latencies, OFF Δ onset latency was relatively modest (50 ms PPI, +5.0 ± 3.8%; 300 ms PPI, −13.9 ± 5.8%; Fig. 4Bii). Importantly, OFF Δ onset latency was significantly different from ON Δ onset latency at both the 50 ms (ON–OFF Δ onset latency = +63.8 ± 15.2 ms; p < 0.0001, unpaired Student's t test, two tailed; Fig. 4Bi, purple asterisks) and 300 ms PPIs (ON–OFF Δ onset latency = +69.5 ± 28.5 ms; p < 0.05). There was no significant difference in Δ onset latency between scotopic (Fig. 4Bi, open circles) or mesopic (Fig. 4Bi, filled circles) background conditions at the 300 ms PPI for ON L-IPSCs (scotopic, +44.6 ± 27.1 ms, n = 11; mesopic, +28.9 ± 10.9 ms, n = 23; p = 0.52, unpaired Student's t test, two tailed), although there was a nonsignificant trend toward larger Δ onset latencies for OFF L-IPSCs under mesopic background (scotopic, +4.6 ± 27.1 ms, n = 5; mesopic, −36.5 ± 11.0 ms, n = 26; p = 0.14).
ON and OFF L-IPSC size and timing differentially adapt to background light
Feedforward light responses in the intact Mb terminal consist of either analog EPSPs with an onset latency of ∼35–50 ms under mesopic background conditions (Wong et al., 2005) or a combination of EPSPs and Ca2+ action potentials with onset latencies ranging from ∼50–100 ms (Joselevitch and Kamermans, 2007) or ∼75–90 ms (Protti et al., 2000; Baden et al., 2011), respectively, under scotopic background conditions. Therefore, the functional interpretation of ON and OFF L-IPSC size and latency STP requires examination of the degree to which changes in the absolute size and latency of single responses is influenced by background light. Furthermore, the degree to which ON and OFF L-IPSCs overlap or act independently at the Mb terminal during exposure to rapidly changing naturalistic visual stimuli will depend on the difference between their onset latencies under different background illumination. To test the sensitivity of L-IPSC amplitude and onset latency to light adaptation state, we varied background light intensity from scotopic (5.03 × 101 photons μm−2 s−1) to mesopic (5.03 × 103 photons μm−2 s−1) levels (Fig. 5).
Mixed rod/cone-mediated ON L-IPSCs (Vigh et al., 2011) were expected to show delayed (i.e., increased) onset latencies when background light was shifted from mesopic to scotopic levels (Fig. 5A) as a result of an increased relative contribution of slow rod inputs via mGluR6 (Wong et al., 2005). In addition, ON L-IPSCs were expected to increase in amplitude under scotopic background conditions as a result of increased recruitment of rods (Wong et al., 2005; Vigh et al., 2011) and decreases in both L-type Ca2+ channel inactivation and steady-state depletion of the readily releasable vesicle pool (Jarsky et al., 2011; Oesch and Diamond, 2011) in neighboring Mb terminals. Cone-dominant OFF L-IPSCs were also expected to increase in amplitude under scotopic background because of enhancement of voltage-gated Na+ channel conductances in cone BCs after a putative reduction of dopaminergic signaling (Zenisek et al., 2001; Ichinose and Lukasiewicz, 2007).
Consistent with these expectations, the amplitudes of ON (scotopic background, 45.4 ± 6.5 pA, n = 20; mesopic background, 29.6 ± 3.7 pA, n = 20; p < 0.05, unpaired Student's t test, two tailed; Fig. 5B, top left) and OFF (scotopic background, 34.9 ± 5.5 pA, n = 18; mesopic background, 20.5 ± 2.6 pA, n = 18; p < 0.05; Fig. 5B, top right) L-IPSCs under scotopic background were larger than under mesopic background. There was no significant dependence of GABACR-mediated standing leak current (Hull et al., 2006a) on background light (scotopic, −28.3 ± 3.0 pA, n = 20; mesopic, −27.8 ± 3.1 pA, n = 20; p = 0.90, unpaired Student's t test, two tailed). ON L-IPSC latency showed the expected delay (i.e., increase) after the transition to scotopic background (scotopic, 75.6 ± 7.4 ms, n = 20; mesopic, 52.4 ± 2.6 ms, n = 20; p < 0.01; Fig. 5B, bottom left). However, OFF L-IPSC latency surprisingly showed a significant advance (i.e., decrease; scotopic background, 68.7 ± 7.5 ms, n = 18; mesopic background, 146.7 ± 12.3 ms, n = 15; p < 0.0001; Fig. 5B, bottom right). This decreased latency of OFF L-IPSCs, which are driven primarily by the light responses of cone photoreceptors (Vigh et al., 2011), may have been attributable to acceleration of glutamate release from cone BC terminals caused by enhanced Na+ channel amplification of cone BC responses under scotopic background (Ichinose et al., 2005; Ichinose and Lukasiewicz, 2007).
These experiments do not exclude the possibility that background light may have caused these changes in OFF L-IPSC latencies via light-dependent changes in voltage-gated Na+ channel amplification of OFF AC EPSPs (Vigh et al., 2011). Interestingly, the advance of OFF L-IPSCs under scotopic background resulted in the elimination of the significant differences in ON versus OFF onset latency that we observed under mesopic background (mesopic, OFF–ON latency, +94.4 ± 12.0 ms, p < 0.0001, unpaired Student's t test, two tailed; scotopic, OFF–ON latency, −6.8 ± 10.8 ms, p > 0.05). Thus, ON and OFF L-IPSCs evoked in response to naturalistic stimuli containing rapid fluctuations in surround luminance would be expected to significantly overlap with each other under scotopic background and to remain temporally distinct under mesopic (or photopic) background light levels.
Contribution of feedback inhibition and serial inhibitory circuits to STP of ON and OFF L-IPSCs
To explore the mechanisms that underlie differential STP of ON and OFF L-IPSCs, we used a pharmacological approach to dissect the roles of circuits that mediate horizontal cell (HC) → photoreceptor (PR) feedback inhibition in the outer retina (Fahrenfort et al., 2005, 2009) and AC→AC serial inhibition in the inner retina (i.e., “nested loops,” Fig. 1D; Zhang et al., 1997; Watanabe et al., 2000; Hsueh et al., 2008; Eggers and Lukasiewicz, 2010). Block of these circuits would be expected to increase the dynamic range of ON and OFF BC voltage responses to light stimulation and to disinhibit ACs that mediate the direct pathway for L-IPSCs (Fig. 1D), respectively. Thus, we hypothesized that functional removal of these outer and inner retinal subcircuits would result in the enhancement of vesicle depletion at AC (Li et al., 2007) and BC (von Gersdorff and Matthews, 1997) presynaptic terminals at different locations in the trisynaptic circuits that mediate ON and OFF L-IPSCs and thus differentially shape light-evoked depression and facilitation.
First, we examined the effect of AMPAR blockade on STP of ON L-IPSC charge transfer by bath applying 10 μm NBQX (Fig. 6A, blue). We showed previously that ON L-IPSC propagation at ON BC → AC synapses can occur in the absence of AMPAR signaling and that block of AMPARs both eliminates OFF L-IPSCs and enhances ON L-IPSC charge transfer at the Mb terminal, likely via block of HC → PR feedback inhibition (Vigh et al., 2011). Previous studies have demonstrated that this feedback is activated by PR release of glutamate onto AMPA receptors on HCs in the dark, followed by HC depolarization and either pH-mediated or ephaptic inhibition of Ca2+ channels on PR terminals (Fahrenfort et al., 2005, 2009). Our examination of STP in the presence of NBQX (Fig. 6C, left, blue) revealed that ON L-IPSCs exhibited strong depression of charge transfer at the 300 ms PPI (PPR, 0.19 ± 0.04; n = 6; p < 0.0001, one-sample t test; green asterisks) that recovered at PPIs of 1100 and 1900 ms. Depression in the presence of NBQX at the 300 ms PPI was significantly enhanced relative to control (p < 0.0001, unpaired Student's t test with Welch's correction; same control data as Fig. 3B, purple asterisks).
Next, we tested the effects of blocking GABAARs on ON and OFF STP over a range of PPIs from 50 to 1900 ms. Bath application of 25 μm SR, which blocks GABAARs at reciprocal AC → Mb (Vigh and von Gersdorff, 2005) and serial AC → AC (Dowling and Werblin, 1969; Zhang et al., 1997; Watanabe et al., 2000; Eggers and Lukasiewicz, 2010) synapses, caused a clear disinhibition of both ON and OFF L-IPSCs (Fig. 6B, red; Vigh et al., 2011). In the presence of SR, ON depression was not significantly altered relative to control PPRs (Fig. 6C, middle; same control data as Fig. 3B), except that depression at the 1900 ms PPI was not significant with SR (PPR, 0.92 ± 0.07; n = 26; p > 0.05, one sample t test). However, STP of OFF L-IPSC charge transfer (Fig. 6C, right) and peak amplitude (data not shown) were transformed in the presence of SR. Depression of OFF charge transfer was significantly reduced relative to control at the 300 ms PPI (control PPR, 0.87 ± 0.07, n = 32; SR PPR, 1.20 ± 0.12, n = 20; p < 0.01, unpaired Student's t test, two tailed), and facilitation of OFF peak amplitude was significantly reduced at PPIs of 1100 ms (control PPR, 1.21 ± 0.10, n = 32; SR PPR, 0.80 ± 0.10, n = 18; p < 0.01) and 1900 ms (control PPR, 1.26 ± 0.09, n = 37; SR PPR, 0.84 ± 0.06, n = 18; p < 0.01).
Any potential contribution of GABAAR desensitization to total depression was eliminated by bath application of SR, because GABACRs do not exhibit desensitization (Matthews et al., 1994; Lukasiewicz et al., 2004). Therefore, these results are consistent with the idea that GABAAR desensitization at lateral AC → Mb synapses may contribute to OFF depression, but not ON depression, at short PPIs. In addition, SR block of OFF peak amplitude facilitation at long intervals suggested that GABAA-mediated signaling at serial AC → AC synapses may act primarily to regulate STP at AC terminals that mediate OFF, but not ON, L-IPSCs. However, ON depression, which remained intact after GABAAR blockade, may depend primarily on a distinct mechanism, such as vesicle depletion at AC → Mb GABACR synapses.
L-IPSCs triggered by direct depolarization of ACs presynaptic to the Mb terminal display large latencies and asynchronous release
Consistent with the TTX sensitivity of glutamate puff-evoked lateral inhibition onto RBCs in the rat IPL (Chávez et al., 2010), our previous work showed that TTX attenuated OFF L-IPSCs and had mixed effects on ON L-IPSCs (Vigh et al., 2011). This suggested that two separate populations of ACs, only one of which expresses voltage-gated Na+ channels (OFF), might mediate ON and OFF L-IPSCs at the Mb terminal. Furthermore, differential effects of TTX (0.2–5 μm in the presence of 25 μm SR) on the onset latencies of ON L-IPSCs (SR latency, 68.3 ± 3.4 ms, n = 29; TTX + SR latency, 74.9 ± 4.1 ms, n = 23; p > 0.05, unpaired Student's t test, two tailed) and OFF L-IPSCs (SR latency, 158.8 ± 11.1 ms, n = 22; TTX + SR latency, 205.6 ± 17.9 ms, n = 10; p < 0.05) in the present study (data not shown) supported the idea that voltage-gated Na+ channels selectively amplify and accelerate light responses in the ACs that mediate OFF L-IPSCs. Such a role for voltage-gated Na+ channels in AC light responses has been demonstrated previously for inhibitory inputs to GCs mediated by AII ACs in mouse retina (Tian et al., 2010).
To discover the identity of the distinct AC classes involved in ON and OFF L-IPSCs, we performed paired recordings of intact Mb terminals and nearby ACs (Fig. 7A,B). In a small percentage of ACs, a somatic step depolarization from −70 to 0 mV evoked L-IPSCs in the Mb terminal (only 3 of 39 pairs; connectivity, ∼7.7%; Fig. 7A,B). Of these three laterally connected ACs, none of which showed AC EPSCs after Mb depolarization (data not shown), two were diffuse or bistratified, and one (Fig. 7A) appeared to stratify primarily in the ON sublamina of the IPL. In the remaining 36 of 39 paired recordings, EPSCs were evoked in ACs after step depolarization of the Mb terminal from −70 to 0 mV, but stimulation in the reverse direction (from AC → Mb) failed to evoke L-IPSCs. Although it is possible that ACs that exhibited unidirectional lateral synaptic connectivity in either direction (Mb → AC or AC → Mb) belong to separate functional classes, we were not able to discern clear morphological differences between these two groups based on reconstructions from confocal images. Because all ACs selected were within 50 μm of the soma of the recorded Mb, it is possible that we failed to characterize certain classes of wide-field ACs that may make the majority of the lateral inhibitory synapses onto the Mb terminal.
The AC → Mb paired recordings revealed an unexpectedly slow L-IPSC onset latency of 106.1 ± 76.0 ms (n = 3; Fig. 7A,B). Such long latencies, which are much larger than the time course of a typical synaptic delay (∼1 ms), may have been attributable to incomplete voltage clamp of distal sections of AC dendritic arbors (Koizumi et al., 2005). This would produce slow depolarization and eventually activate voltage-gated Na+ channels, triggering an escaping dendritic action potential (Koizumi et al., 2005) that would result in the relatively sharp, consistent IPSC onsets that we observed in the Mb terminal (Fig. 7B, IPSC latency). A similar delayed dendritic Na+ action potential in OFF ACs, subsequent to spatiotemporal summation of small EPSCs arriving from multiple BC presynaptic terminals, could explain the long onset latencies that we observed for light-evoked OFF IPSCs, especially under mesopic background conditions (Figs. 4, 5). However, the rare occurrence and extreme technical difficulty of obtaining synaptically connected pairs of AC → Mb recordings prevented us from further testing this mechanism with bath or focal application of TTX.
Interestingly, Mb L-IPSCs in paired recordings also showed strong asynchronous release after a somatic repolarizing step to −70 mV in the recorded AC (Fig. 7B). Response duration in Mb terminals stimulated with a 200 ms depolarizing step in ACs was 383.9 ± 95.8 ms (n = 3). A possible mechanism for this asynchronous GABA release from ACs (183.9 ± 95.8 ms in duration), which could act to both truncate Mb glutamate release and impose a relative refractory period for light-evoked depolarization, is the accumulation of cytoplasmic Ca2+ subsequent to activation of voltage-dependent Ca2+ channels (Gleason et al., 1994). Such residual Ca2+ accumulation in the AC boutons that mediate OFF L-IPSCs would provide a possible mechanism for light-evoked OFF facilitation and latency advance (Figs. 3, 4).
Discussion
We have shown that the light-evoked, trisynaptic ON and OFF L-IPSCs at Mb BC terminals exhibit distinct temporal profiles of STP. ON L-IPSCs (Fig. 8, left, blue) exhibited amplitude depression (STD) of ∼45% at 50 ms that recovered to ∼10% at 1900 ms, and onset latency delay of ∼95% at 50 ms that fully recovered at 1900 ms. OFF L-IPSCs (Fig. 8, right, red) displayed amplitude STD of ∼20% at 300 ms, amplitude facilitation (STF) of ∼25% at 1900 ms, and latency advance of ∼15% at 300 ms that recovered at 1900 ms. The recovery of ON and OFF STD (τ ∼500 ms) was rapid compared with STD of Mb exocytosis (τ ∼8 s; von Gersdorff and Matthews, 1997) and reciprocal inhibition (τ ∼12 s; Li et al., 2007) and therefore likely acts as a temporal high-pass filter of Mb glutamate release onto AC and GC dendrites (Sagdullaev et al., 2011). To clarify the functional roles of L-IPSC STP, we showed that changing background light from mesopic to scotopic increased the size and latency of ON L-IPSCs but increased the size and decreased the latency of OFF L-IPSCs. Next, we used pharmacology to identify HC → photoreceptor feedback and AC → AC serial inhibitory circuits that shape ON and OFF STP, respectively (Fig. 8, middle). Finally, we showed that L-IPSCs evoked by depolarizing voltage steps at AC somata had long (∼100 ms) latencies, consistent with the large delays of light-evoked OFF L-IPSCs, and exhibited sustained and asynchronous release (∼200 ms), consistent with the idea that Ca2+ accumulation in AC boutons may account for OFF facilitation and latency advance (Fig. 7). Together, these findings describe a novel scheme for STP of convergent, temporally independent lateral feedback inhibitory inputs at the BC presynaptic terminal.
Effects of L-IPSC plasticity on Mb terminal light responses likely depend on background light conditions
Under scotopic background, in which the Mb resting potential (Em) has been measured at between −40 and 45 mV (Protti et al., 2000; Baden et al., 2011), light responses at intact Mb terminals consist of either “analog” EPSPs with onset latencies of ∼50 to 100 ms (Joselevitch and Kamermans, 2007) or “digital” Ca2+ action potentials with latencies of 75–90 ms (Protti et al., 2000; Baden et al., 2011). We expect, under these conditions, that ON L-IPSCs with latencies of 50–100 ms and a conductance of 1 nS (Fig. 5) at a membrane potential of −40 mV could transiently hyperpolarize the Mb terminal by up to 10 mV, assuming an ECl of −60 mV and an input resistance of ∼500 MΩ (Protti et al., 2000), or reduce the EPSP peak by up to 70% (Oltedal et al., 2009). ON L-IPSC depression of 10–45% (Fig. 8, left), then, would be expected to decrease the size of subsequent IPSPs to between 5 and 9 mV and to reduce attenuation of the EPSP peak to between 40 and 60%. In addition, ON latency delay of 95% during high-frequency surround stimulation (Fig. 8, left) would eliminate block of the EPSP peak (Oltedal et al., 2009) but allow significant inhibition of the sustained component, which can last from ∼200 to 500 ms (Joselevitch and Kamermans, 2007). We expect that OFF L-IPSCs, with latencies of ∼70 ms and amplitudes of ∼35 pA (Fig. 5), would have effects similar to ON L-IPSCs under these conditions during naturalistic transient changes in surround luminance but would exert increased influence over the EPSP peak during repeated surround stimulation at intermediate and long PPIs as a result of ∼15% latency advance and ∼25% amplitude facilitation (Fig. 8, right).
The above estimates of the effects of L-IPSC plasticity on Mb light responses are based on a model of the rat RBC (Oltedal et al., 2007, 2009). Although this model uses a slightly higher axial resistance in the axon, higher specific membrane resistance, and smaller total cell capacitance than the two compartment model of goldfish Mb BCs (Mennerick et al., 1997), these differences result in only 39% shorter length constants and 25% faster time constants for the RBC compared with the Mb BC. Because the cutoff frequency of the RBC model is already quite high (∼500 Hz) relative to the overall speed of retinal processing, it is unlikely that these small differences would have a significant effect on our estimates over the paired-pulse frequency range that we examined (∼0.5–20 Hz). In addition, although the Mb BC would be expected to display both slightly faster EPSP propagation from its soma to its axon terminal and slower EPSP and IPSP decay time constants than the rat RBC, the effective cutoff frequency for the output of either class of BC is likely rate limited not by passive cable properties but by the slow opening kinetics of L-type Ca2+ channels (Freeman et al., 2011).
Under scotopic background, the Mb terminal can generate Ca2+ spikes when Em is more depolarized than −48 mV (Baden et al., 2011). Thus, hyperpolarization of 5–10 mV provided by ON or OFF L-IPSCs might be sufficient to prevent spike initiation at the Mb terminal. In addition, shunting inhibition by L-IPSCs likely exerts significant control over Ca2+ spike initiation at the Mb terminal (Hull et al., 2006a). Therefore, depression of ON and OFF L-IPSCs may be permissive for digital signaling at the Mb terminal. Furthermore, latency delay of ON L-IPSCs at short PPIs may decrease their ability to prevent spikes by shifting the time of their arrival to after threshold crossing. However, under circumstances in which surround stimulation is dominated by low-frequency decreases in luminance (i.e., 1900 ms PPI), facilitation of OFF L-IPSCs may result in reduced probability of Mb terminal Ca2+ spikes.
The effects of L-IPSC STP on Mb EPSPs under mesopic background would be expected to differ from those described above for scotopic background for two reasons. First, the latencies of light-evoked EPSPs at the Mb terminal likely advance under mesopic conditions (∼35–50 ms; Wong et al., 2005), thus decreasing the effect of ON L-IPSCs on the EPSP peak. This decreased inhibition will be exacerbated during high-frequency stimulation, when ON L-IPSCs are even further delayed (Fig. 8, left). Second, the arrival delay between ON and OFF L-IPSCs in response to rapid, naturalistic fluctuations in surround luminance will increase after the shift from scotopic to mesopic conditions, from near coincidence to ∼100 ms (Fig. 5B). Thus, we would expect ON and OFF L-IPSCs under these conditions to exert independent frequency-dependent effects (i.e., via the strength of STP), acting to truncate the transient and sustained components of the analog Mb light response, respectively. This would allow ON L-IPSCs to shape the Mb representation of temporal contrast and OFF L-IPSCs to shape its representation of mean luminance (Oesch and Diamond, 2011). Thus, although ON and OFF L-IPSCs might act in concert to exert gain control and gate spike initiation at the Mb terminal under scotopic background, they may perform separate computational roles under cone dominant conditions.
ON and OFF L-IPSC STP are shaped by distinct mechanisms
Because block of GABAARs had little effect on ON L-IPSC depression (Fig. 6), it is unlikely that ON depression was attributable to GABAAR desensitization or the influences of plasticity at upstream GABAA synapses. In addition, enhancement of ON depression at the 300 ms PPI (Fig. 6) after block of AMPARs with NBQX (Fig. 8, middle) suggested that surround feedback inhibition from HCs to photoreceptors (Fahrenfort et al., 2005, 2009) may significantly compress the dynamic range of ON BC light responses and thereby minimize vesicle depletion at ON BC and ON AC synapses. Thus, minimization of HC membrane potential fluctuations via block of AMPARs on HC dendrites would allow ON BCs to be more hyperpolarized in the dark and to experience greater depolarization during light stimulation of photoreceptors. This might both accelerate recovery of the readily releasable pool of glutamatergic vesicles at the Mb terminal in the dark and enhance depression attributable to increased vesicle depletion during the light stimulus (Jarsky et al., 2011). Therefore, it seems likely that ON depression and latency delay under control conditions are driven by a combination of vesicle depletion at neighboring BC terminals (von Gersdorff and Matthews, 1997) and presynaptic ACs (Li et al., 2007) and that these mechanisms are enhanced by removal of feedback inhibition in the outer retina.
STP of OFF L-IPSCs was regulated by mechanisms distinct from those involved in ON STP. OFF depression at short PPIs was diminished or absent after blockade of GABAARs (Fig. 6). This result suggested that GABAAR desensitization at the Mb terminal might contribute to OFF depression. This is consistent with the previous finding that Mb GABAARs can recover fully from desensitization within 1–2 s (Li et al., 2007). OFF amplitude facilitation at long PPIs (Fig. 8, right), which was attenuated by SR block of GABAARs at AC → AC synapses, may be attributable to progressive disinhibition via serial AC pathways (i.e., depression within nested loops; see Fig. 8, middle). In addition, the sustained and asynchronous L-IPSCs evoked in our paired recordings (Fig. 7), along with the latency advances displayed by paired OFF L-IPSCs (Fig. 8, right), are both consistent with the idea that OFF facilitation may be driven, in part, by intracellular residual Ca2+ accumulation and increased release probability at OFF AC presynaptic boutons (Gleason et al., 1994).
Potential physiological roles of L-IPSC STP at BC terminals
Filtering of transient or sustained glutamate release by STP of ON and OFF L-IPSCs at the Mb terminal may enable postsynaptic GCs (and ACs) to rapidly modify the strength and extent of their inhibitory surrounds after changes in surround temporal contrast. Such dynamic spatial coupling may play a role in the rapid coding of spatial information as relative GC spike onset latencies (±30 ms) after saccadic gaze relocation, wherein local spatial contrast determines the ratio between convergent OFF (early) and ON (late) BC inputs onto individual GC dendrites (Gollisch and Meister, 2008). STP in the size and timing of ON (early) and OFF (late) L-IPSCs driven by distal BC inputs would introduce frequency-dependent correlations between spike latencies across a population of GCs, thus enabling the extraction of spatial information by downstream decoders (Usrey et al., 2000; Chase and Young, 2007).
Footnotes
- Received February 5, 2012.
- Revision received June 9, 2012.
- Accepted June 15, 2012.
This research was funded by National Institutes of Health/National Eye Institute Grants EY014043 (H.v.G.) and EY EY019051 (J.V.) and was partially supported by Korea Research Foundation Grant KRF-2008-357-E00032 (M.H.K.). We thank Drs. Ilya Buldyrev and W. Rowland Taylor for helpful discussions and comments on this manuscript.
- Correspondence should be addressed to Henrique von Gersdorff, The Vollum Institute, Oregon Health and Science University, 3181 SW Sam Jackson Park Road, Portland, OR 97239. vongersd{at}ohsu.edu
- Copyright © 2012 the authors 0270-6474/12/3211688-12$15.00/0