Controlling Horizontal Cell-Mediated Lateral Inhibition in Transgenic Zebrafish Retina with Chemogenetic Tools

Abstract

Horizontal cells (HCs) form reciprocal synapses with rod and cone photoreceptors, an arrangement that underlies lateral inhibition in the retina. HCs send negative and positive feedback signals to photoreceptors, but how HCs initiate these signals remains unclear. Unfortunately, because HCs have no unique neurotransmitter receptors, there are no pharmacological treatments for perturbing membrane potential specifically in HCs. Here we use transgenic zebrafish whose HCs express alien receptors, enabling cell-type-specific control by cognate alien agonists. To depolarize HCs, we used the Phe-Met-Arg-Phe-amide (FMRFamide)-gated Na⁺ channel (FaNaC) activated by the invertebrate neuropeptide FMRFamide. To hyperpolarize HCs we used a pharmacologically selective actuator module (PSAM)-glycine receptor (GlyR), an engineered Cl^- selective channel activated by a synthetic agonist. Expression of FaNaC or PSAM-GlyR was restricted to HCs with the cell-type selective promoter for connexin-55.5. We assessed HC-feedback control of photoreceptor synapses in three ways. First, we measured presynaptic exocytosis from photoreceptor terminals using the fluorescent dye FM1-43. Second, we measured the electroretinogram (ERG) b-wave, a signal generated by postsynaptic responses. Third, we used Ca²⁺ imaging in retinal ganglion cells (RGCs) expressing the Ca²⁺ indicator GCaMP6. Addition of FMRFamide significantly decreased FM1-43 destaining in darkness, whereas the addition of PSAM-GlyR significantly increased it. However, both agonists decreased the light-elicited ERG b-wave and eliminated surround inhibition of the Ca²⁺ response of RGCs. Taken together, our findings show that chemogenetic tools can selectively manipulate negative feedback from HCs, providing a platform for understanding its mechanism and helping to elucidate its functional roles in visual information processing at a succession of downstream stages.

Keywords: feedback inhibition; fluorescent protein; horizontal cell; lateral inhibition; photoreceptor; retina.

Figure 1.

Strategy for chemogenetic manipulation of HC membrane potential. A, A bicistronic construct containing the molluscan FaNaC from H. aspersa and the red fluorescent protein mCherry, both under the control of the Cx55.5 promoter, were transgenically expressed in HCs of zebrafish causing red fluorescence specific to the HC layer. mCherry exhibits intrinsic fluorescence seen in confocal imaging of a fixed retinal section (left) and in two-photon imaging of a fresh retinal flat mount (right). Puncta with saturating expression of mCherry likely represent protein aggregates within the Golgi apparatus of HCs. A diagram illustrates how binding of the agonist FMRFamide causes Na⁺ influx thereby depolarizing the membrane potential. B, A bicistronic construct containing the PSAM-GlyR and the green fluorescent protein eGFP, both under the control of the Cx55.5 promoter, were transgenically expressed in HCs of zebrafish causing green fluorescence specific to the HC layer. eGFP was immunolabeled with Alexa Fluor 488 for confocal imaging of a fixed retinal section (left) and its intrinsic fluorescence is seen in two-photon imaging of a fresh retinal flat mount (right). A diagram illustrates how binding of the designer agonist PSEM^89S causes Cl^– influx thereby clamping the membrane potential to ECl. Nuclei are stained with DAPI (blue). Scale bar: 20 μm.

Figure 2.

Monitoring changes in membrane potential elicited by FMRFamide and PSEM^89S. A, Voltage response to a 100-ms puff of PSEM^89S (500 μm) in an isolated HC recorded with the gramicidin perforated patch configuration. A puff was delivered at the time indicated by the arrow. Although the concentration of PSEM^89S was 500 μm in the pipette, the concentration of the drug is estimated to be diluted ∼50-fold at the cell (Firestein and Werblin, 1989), positioned ∼100 μm from the puffer pipette. B, Traces from three different HCs showing responses to 100-ms puffs of GABA (1 mm), chosen to highlight the variability of responses to GABA that were observed between cells. C, Response to 100-ms application of muscimol (100 μm), a GABA_A receptor agonist. Responses were uniformly hyperpolarizing in all cells tested. D, Summary data showing the resting membrane potential, and the membrane potential at the peak of the response to PSEM^89S. Large bolded symbols are the mean ± SEM for each condition (n = 6). E, Summary of the change in membrane potential evoked by puffs of GABA or muscimol. Small closed symbols are the responses of individual cells to GABA (n = 11). Open symbols are the responses to muscimol (n = 7). Open boxes are the mean ± SEM for hyperpolarizing responses, and the shaded box is the mean ± SEM for depolarizing responses. The mean amplitude of the responses evoked by muscimol was not significantly different from the amplitude of hyperpolarization evoked by GABA (p = 0.53, rank-sum test). F, Summary data for bath application of FMRFamide (30 μm). Large bolded symbols are the mean ± SEM for each condition (n = 7).

Figure 3.

Chemogenetic activation of HCs alters vesicular release of FM1-43 from cone terminals. A, Schematic representation of the mechanisms of negative (–) and positive (+) feedback of HCs onto cone photoreceptors (PR). VGCCs are voltage-gated Ca²⁺ channels, green circles are FM1-43-filled synaptic vesicles, orange dots are glutamate molecules in the synaptic cleft, AMPARs are AMPA receptors. Negative feedback is mediated by changes in the membrane potential of HCs, whereas positive feedback is mediated by increased intracellular Ca²⁺ in HCs, owing to influx of Ca²⁺ through Ca²⁺-permeant AMPA receptors. ΔCa_i is the change in intracellular Ca²⁺; ΔV_M is the change in HC membrane potential. B, Two-photon scanned images of the photoreceptor terminal layer in FaNaC zebrafish retinas. Retinas were pretreated with FM1-43 to load recycling synaptic vesicles and then treated for 20 min with the indicated receptor agonist. AMPA (20 μm) increased the rate of dye loss (destaining) and FMRFamide (10 μm) decreased destaining, but the AMPA-elicited increase was dominant when the two agonists were applied together. C, D, Time course of FM1-43 destaining in FaNaC fish. Without added agonists, cone terminals released FM1-43 at 0.87 ± 0.08% per minute (n = 10 retinas). Adding AMPA accelerated destaining rate (1.89 ± 0.13%, n = 9 retinas, p < 1 × 10−5). Adding FMRFamide (10 μm) decelerated destaining (0.54 ± 0.07% n = 9 retinas, p = 0.001). Adding AMPA and FMRFamide together resulted in a significant accelerated destaining rate from dark (p = 0.002), which was not significantly different from that with AMPA alone (1.96 ± 0.25%, n = 4, p = 0.88). Statistical differences were determined with one-way ANOVA (F(3,23) = 27.4, p = 8.9 × 10−8). E, F, Time course of FM1-43 destaining in PSAM fish. Adding PSEM^89S (30 μm) accelerated destaining (0.83 ± 0.19%, n = 9 retinas, p = 0.044). Adding AMPA accelerated destaining (2.19 ± 0.19%, n = 5 retinas, p = 0.0002). Adding PSEM^89S and AMPA together resulted in a destaining rate not significantly different from that with AMPA alone (2.05 ± 0.35%, n = 5, p = 0.82). Analysis was performed with one-way ANOVA (F(3,24) = 14.69, p = 1.2 × 10−5). For D and F open circles represent individual data points. Open boxes represent the interquartile range with whiskers. The mean is marked by x and the median by a line.

Figure 4.

Chemogenetic manipulation of HCs modulates the bipolar cell light response. Application of FMRFamide had no effect on the ERG of WT zebrafish (A), but decreased the maximal b-wave amplitude in FaNaC zebrafish (B), reaching a maximum ∼40% decrease in the b-wave (C). Application of the buffer HEPES (pH 7.35) cancelled the effects of FMRFamide (C). FMRFamide response was dose dependent having an EC₅₀ of 4.25 μm (D). Application of the agonist PSEM^89S (30 μm) also caused a decrease in the maximal b-wave amplitude, but to a lesser degree (E, F).

Figure 5.

Expression of GCaMP6f in chemogenetic zebrafish lines. A, Ex vivo retina of adult HuC-GCaMP6f zebrafish shows eGFP fluorescence in retinal flat mounts using two-photon imaging. B, Light response is easily measured in RGCs before (left) and after (right) a light flash. C, In vivo confocal imaging of eGFP in 1 dpf PTU-treated zebrafish larvae of HuC-GCaMP6f and FaNaC crossed transgenic fish. The optic nerve (ON) and the developing retina (RGC) show strong eGFP fluorescence. D, In vivo imaging of the same line of fish imaged in C showing that FaNaC-mCherry is easily visualized in HCs at 2 dpf zebrafish larvae. Some red autofluorescence is generated by scattered pigment cells on the ocular surface. RGC, retinal ganglion cell layer; IPL, inner plexiform layer; BC, bipolar cell layer; HC, horizontal cell layer.

Figure 6.

Chemogenetic manipulation of HCs alters lateral inhibition in downstream RGCs. A, Light responses were measured in HuC::GCaMP6f fish with three light stimuli, delivered at six different spot diameters. Calcium transients were recorded using fluorescence imaging. B, WT RGCs responded with maximum change in fluorescence to a 500-μm spot of light, which decreased at 1000 μm, supporting lateral inhibition. This effect was blocked by HEPES, and was unchanged by applying 10 μm FMRFamide. C, Application of FMRFamide on FaNaC retinas perturbed normal RGC center surround response, (D) as did application of PSEM^89S on PSAM- GlyR retinas. E, The LIR in various conditions shows that application of the cognate agonists to FaNaC and PSAM retina significantly disrupts lateral inhibition. Open circles represent individual data points. Open boxes represent the interquartile range with whiskers. The mean is marked by x and the median by a line.