Prolonged Light Exposure Induces Long-Lasting Retinal Wave Plasticity via Retrograde Melanopsin-Dopamine Signaling

Abstract

The spontaneous activity in the developing retina is necessary for the maturation of the neuronal circuitry in visual-associated brain areas. While previous studies have shown that the activation of intrinsically photosensitive retinal ganglion cells (ipRGCs) contributes to visual development, its effects and mechanisms remain largely unclear. Here, using microelectrode array recordings from both male and female mice, we demonstrated that prolong light exposure reduces the interwave interval and coupling distance of Stage 2 retinal waves, which can persist for at least 1 h after light exposure. Notably, these light-induced effects on cholinergic waves were impaired in melanopsin knock-out mice. Additionally, the light-induced retinal wave modulation is mediated by the dopaminergic pathway, primarily through D4 receptors. Using single-molecule fluorescence in situ hybridization, we identified high expression of D2-like receptors, particularly D4R, in the ganglion cell layer and ON-type starburst amacrine cells (SACs) during early postnatal development. Furthermore, we found that gap junction coupling in SACs was increased after light exposure, which can be blocked by D2-like receptor antagonists. Overall, our study reveals that the key properties of spontaneous Stage 2 retinal waves can be regulated by environmental light through ipRGCs. This regulation involves dopaminergic signaling, highlighting the critical role of ipRGC in retinal wave modulation and the convergence of experience-dependent and independent circuitry refinement processes.

Significance Statement

The formation of complex neuronal circuitry during development is regulated by both intrinsic and extrinsic factors. While the importance of experience-independent processes, such as spontaneous retinal waves, in sensory system maturation is well established, the circuitry and mechanism of internal and external factors convergence remain poorly understood. We show that light exposure modulates the properties of Stage II retinal waves through the intrinsically photosensitive retinal ganglion cell (ipRGC)–dopamine retrograde signaling pathway, revealing a critical convergence pathway between ipRGC and spontaneous retinal waves. This highlights how environmental cues interact with intrinsic developmental programs to shape the assembly of mature visual system circuitry, indicating the importance of early light exposure during development.

Introduction

The formation of complex neural networks begins with coarse wiring and is refined through subsequent axonal remodeling to establish functional circuits. Two major processes guide this development: intrinsic, experience-independent programs and sensory experience-dependent plasticity shaped by environmental input. Among intrinsic factors, axon guidance molecules direct fiber growth and connectivity (Tessier-Lavigne and Goodman, 1996; Petros et al., 2009; Kania and Klein, 2016), while spontaneous neural activity contributes broadly to circuit formation across systems, including sensory pathways (Meister et al., 1991; Garaschuk et al., 1998; Hanson and Landmesser, 2003, 2004; Tritsch et al., 2007; Watt et al., 2009).

In the developing visual system, spontaneous retinal activity—referred to as retinal waves—occurs in retinal ganglion cells (RGCs; Meister et al., 1991; Wong et al., 1995; Bansal et al., 2000). These retinal waves shape retinotopic maps (McLaughlin et al., 2003), eye-specific segregation (Muir-Robinson et al., 2002; Stellwagen and Shatz, 2002), and refining receptive fields in the dorsal lateral geniculate nucleus (dLGN) and striate cortex (Wang et al., 2009; Davis et al., 2015). Retinal waves occur in three distinct stages, each mediated by different mechanisms (Feller et al., 1996; Bansal et al., 2000; Wong et al., 2000; Syed et al., 2004). During Stage 2 retinal wave, starburst amacrine cells (SACs) release acetylcholine to drive RGC activity (Zheng et al., 2004, 2006; Ford et al., 2012). Disruption of Stage 2 retinal waves interval causes severe deficits, including stratification in the dLGN, superior colliculus, and suprachiasmatic nucleus (Grubb et al., 2003; McLaughlin et al., 2003; Negueruela et al., 2024), and disrupted eye-specific segregation (Muir-Robinson et al., 2002; Stellwagen and Shatz, 2002). In parallel, experience-dependent plasticity further sculpts neural circuits during critical developmental windows (Feldman and Knudsen, 1997; Gold and Knudsen, 1999; Miller and Knudsen, 1999; Erzurumlu and Gaspar, 2012; Cai et al., 2024). External factors, including light and temperature, have been shown to influence spontaneous retinal activity (Stafford et al., 2009; Renna et al., 2011; Chew et al., 2017). However, the mechanisms by which environmental cues interact with intrinsic programs during visual system maturation remain poorly understood.

The mammalian retina contains a distinct class of light-sensitive neurons known as intrinsically photosensitive RGCs (ipRGCs). Unlike conventional RGCs, ipRGCs express melanopsin, allowing direct light detection and transmission of luminance signals to brain regions governing nonimage-forming functions such as circadian entrainment, pupillary reflex, mood regulation, and memory (Berson et al., 2002; Hattar et al., 2002, 2003; Panda et al., 2003; LeGates et al., 2012; Fernandez et al., 2018; Lee et al., 2022; Weil et al., 2022; Huang et al., 2023; Meng et al., 2023; Lazzerini Ospri et al., 2024). ipRGCs express melanopsin as early as embryonic Day 11 (McNeill et al., 2011; Lucas and Schmidt, 2019) and influence vascular growth (Rao et al., 2013), hippocampal development (Hu et al., 2022), glucose tolerance (Clarkson-Townsend et al., 2021), and circadian function (Chew et al., 2017) during development. Recent works suggested ipRGCs also modulate retinal development, including cone lamination, maturation of the retinal clock, and retinofugal projection refinement (Tufford et al., 2018; Kinane et al., 2023; D’Souza et al., 2024). Although ipRGCs can increase Stage 2 retinal wave burst duration, how they influence retinal development retrogradely remains unclear.

Here, we show that light exposure reduces the interwave interval (IWI) and coupling distance of Stage 2 retinal waves, which are absent in melanopsin knock-out (MKO) mice. Pharmacological experiments reveal that dopamine signaling via D2-like receptors is both necessary and sufficient for light-induced modulation. Single-molecule fluorescence in situ hybridization (smFISH) further demonstrates that D2-like receptors, especially D4R, are enriched in the ganglion cell layer (GCL) and ON-type SACs during early postnatal stages. Moreover, light enhances SAC gap junction coupling through D2-like receptors. Together, these findings uncover a retrograde melanopsin–dopamine pathway that modulates Stage 2 retinal wave properties, providing a mechanism by which environmental light integrates with intrinsic activity to refine developing visual circuits.

Materials and Methods

Animals

C57BL/6J wild-type (WT) mice were purchased from the National Laboratory Animals Center. All the mice were maintained in 12:12 h light/dark cycle with ad libitum food access and aged in the range of Postnatal Day (P)3–P8. Mice were bred in house to generate the offspring for electrophysiology recording, C57BL/6 and Opn4^Cre/Cre (IMSR_JAX:035925), and for intracellular injections experiment, the ChAT-IRES-Cre (IMSR_JAX:031661) and Ai9 tdTomato reporter (IMSR_JAX: 007909). All the procedures were approved by the Institutional Animal Care and Use Committee of the National Tsing Hua University and the National Taiwan University.

Tissue preparation

In general, the animals were anesthetized by isoflurane (NDC: 66794-013, Piramal Critical Care) or hypothermia and killed by decapitation or intraperitoneal injection with 10 mg/kg ketamine and 10 mg/kg xylazine. After cardinal axis labeling by marking the dorsal site on the cornea, the eyeballs were enucleated using forceps, removed the cornea and lens, and placed in oxygenated (95% O₂ and 5% CO₂) ACSF's medium [contained (in Mm) 124 NaCl₂, 5 KCl, 1.15 KH₂PO₄, 1.15 MgSO₄, 10 glucose, 2.5 CaCl₂, and 25 NaHCO₃] for experiments. All the surgeries proceeded under the dim red light and started at zeitgeber time (ZT) 0.

Light stimulation

For electrophysiology recording, the light stimuli were delivered by blue light LED with a bandpass filter (420–490 nm; 5.11 × 10¹³ photons s⁻¹ cm⁻² at 488 nm) and focused on the retina through the optical collimator to align the light. The LED was driven by an Arduino microcontroller board (Arduino MEGA 2560 and UNO).

Electrophysiological recordings

The isolated retina was mounted on nitrocellulose membrane (9004-70-0; Merck Millipore), with the photoreceptor-side attached. Next, the retina with GCL-side down was placed on the 59 channels (electrodes) microelectrode array. Each electrode was 30 µm in diameter and spaced 200 µm apart (60MEA200/30iR-ITO-pr-T; Multichannel Systems). The platinum ring with wire was placed on the filter to flatten the retina on the MEA chip. The recording chamber was superfused with fresh oxygenated ACSF's medium (1–1.5 ml/min) and heated by the heating device (TC02, Multichannel Systems) to keep the medium at 33–35°C. The spike responses and light stimulation time were recorded simultaneously by MEA-Up-BC-System-Standard (Multichannel Systems) after 60 min of stabilization. For data analysis, the high-pass filter with a cutoff frequency of 200 Hz was applied to distinguish the spiking responses from the local field potentials in MATLAB (MathWorks).

Intracellular injection

In the intracellular injection experiment, the isolated retinas were transferred onto the nitrocellulose membrane with a 1–2 mm² hole and then mounted on the custom-made recording chamber with the GCL-side facing up under the stage of fluorescence microscope (Axioskop 2 FS Plus; Carl Zeiss). The chamber was superfused with fresh oxygenated ACSF's medium (1–1.5 ml/min) and heated by the heating device (Warner Instruments TC-324B in-line heater) to keep the medium at 33–35°C.

The micropipettes were pulled from the borosilicate capillary tube (outside diameter, 1.5 mm; inside diameter, 0.86 mm; Sutter Instruments). Electrode resistance was 8–14 MΩ. For ON SACs intracellular injection, the electrode pipettes were filled with internal solution, which contained the following (in mM): 125 K-gluconate, 8 NaCl, 4 ATP-Mg, 0.5 GTP-Li, 5 EGTA, 10 HEPES, 8% N-(2-aminoethyl)-biotinamide hydrochloride (Neurobiotin; Vector Laboratories), 2% Lucifer Yellow (Sigma-Aldrich), pH 7.3, adjusted with KOH. The internal solution was injected by the generation of oscillatory biphase current (from −0.1 to +0.7 nA at 10 Hz) for 20 min with Digidata 1440A (Molecular Devices) and then waited for neurobiotin diffusion for at least 10 min. After the solution filling, the retina samples were fixed with the fixing solution (4% PFA and 0.1% glutaraldehyde in PBS) for 2 h at room temperature, washed in PBS, and blocked with the blocking solution (8% normal donkey serum and 0.5% Triton X-100 in PBS) overnight at 4°C. Samples were then incubated with streptavidin-conjugated Alexa Fluor 635 (working dilution, 1:200; Invitrogen). Finally, the retinas were mounted on slides with the mounting medium containing DAPI (H-1200, Vector Laboratories).

Pharmacology

Dopamine (50 µM, Sigma-Aldrich), SKF 38393 (10 µM, Tocris Bioscience), quinpirole hydrochloride (20 µM, Tocris Bioscience), PD 168077 (0.5 µM, Tocris Bioscience), raclopride (8 µM, Tocris Bioscience), and L-745870 (10 µM, Tocris Bioscience) were added into the ACSF solution before the bath application, which the stock solutions were dissolved in either distilled water or DMSO. All the pharmacological experiments followed the procedure of electrophysiological experiment and superfused with drug-contained oxygenated ACSF medium, continuously bath applied for 40 min before the recordings.

RNAscope

RNAscope was performed on the samples using Multiplex Fluorescent Reagent Kit v2 Assay (Advanced Cell Diagnostics), following the manufacturer's instructions. Briefly, the retina which was collected from WT mice at age P3, P6, P9, P15, and P28 was frozen sectioned at 10 µm and mounted on glass slides. After protease pretreatment (30 min at 40°C), the samples were hybridized with RNAscope probes C1 (Mm-Drd4 #418172), C2 (Mm-Chat-C2, #408731-C2), and C3 (Mm-Drd2-C3 #406501-C3; 2 h at 40°C). The signal amplification steps followed the manufacturer's protocol with a 1:1,000 dilution of fluorophore [C1, TSA Vivid Fluorophore Kit 570 (catalog #7526); C2, TSA Vivid Fluorophore Kit 520 (catalog #7523); C3, TSA Vivid Fluorophore Kit 650 (catalog #7527)] in TSA buffer, and the samples were also counterstained with DAPI. To examine the immunostaining and RNAscope samples, we used LSM800 confocal microscope system (Carl Zeiss) with 40× and 63× objectives for imaging. All image analyses were processed in ImageJ.

Wave events identification and analysis

To identify each wave event in electrophysiological data, we followed the previously established protocols (Renna et al., 2011; Lee and Chiao, 2016). In brief, we created the peristimulus time histograms of spikes with 250 ms binning time to each channel. The bursts that were considered as the wave candidates were defined by two rules: first, the start of wave-associated bursts was identified when the firing rate was higher than two times of the averaged firing rate (10 min data) and maintained at least 500 ms, and second, the end of the burst unit was identified when the firing rate was lower than the average firing rate and maintained longer than 1 s. The time difference of burst units between neighbor channels <2 s was considered as wave-correlated bursts. After the recognition of waves, the burst duration and IWI were calculated and averaged across all channels in each retina.

For the pairwise correlation index (CI), the method had been used in previous studies (Wong et al., 1993; Blankenship et al., 2011; Kirkby and Feller, 2013) and the function was described as follows:

$$\mathrm{CI}={\displaystyle \frac{N_{AB}(\mathrm{\Delta }t)\cdot T}{N_A(T)\cdot N_B(T)\cdot 2\mathrm{\Delta }t}.}$$

N_AB(Δt) means the number of spikes pairs for which channel B within the time window ±Δt from the spikes of channel A. N_A and N_B are the total number of spikes in channel A and B between the entire recording time T. Δt is the time window, which equaled 0.1 s in our study. For the highly connected channel pairs, we defined the top 15% of the CI in each recording data as the highly connected channel pairs.

Experimental design and statistical analysis

Experiments were conducted on both male and female mice, aged from P3 to P28. All statistical analysis methods and the number of animals (n) are reported in the figure legends. All statistical analysis was performed in MATLAB (MathWorks) and SPSS. The wave-associated properties were compared using one-way ANOVA with post hoc tests or two-sample t tests. The cumulative probability was compared using the Kolmogorov–Smirnov test. The smFISH analysis was compared using Mann–Whitney U test. Two-way ANOVA with Bonferroni’s post hoc test was used for the comparison of coupling cell numbers.

Results

Early light exposure modulates the properties of Stage 2 waves in the mouse retina

To investigate the influence of early light exposure on Stage 2 retinal waves, we performed multi-channel recording of RGC activity using a microelectrode array (MEA) in P3–P8 WT mouse retinas, a period when retinal waves are driven by acetylcholine signaling during the first postnatal week (Ford and Feller, 2012). To examine the effects of light stimulation on cholinergic waves, we recorded Stage 2 retinal wave activity under six conditions: before light-ON (Dark), 10 min after light-ON (L_ON10), 30 min after light-ON (L_ON30), 20 min after light-OFF (L_OFF20), 40 min after light-OFF (L_OFF40), and 60 min after light-OFF (L_OFF60; Fig. 1A).

Figure 1.

The light exposure alters the properties of Stage 2 waves in the mouse retina. A, The experimental design paradigm for MEA recordings in WT P3–P8 mouse retina. Each time points were recorded for 10 min. B, C, The raster plots from MEA recordings in WT P5 at Dark (left) and L_ON30 (right) conditions. (Bottom) Raster plots represent the neural activity of all channels. Top, Activity maps from the rectangle time window (red box) in the raster plots. The dots represent individual electrode sites and the red dots mean the spike activities occurred. Each frame corresponds to 4 s. D–G, Summary data of (D) WAB duration (compared with Dark, L_ON10, ***p < 0.0001; L_ON30, ***p = 0.0006), (E) WAB firing rate (compared with Dark, L_ON10, **p = 0.0069), (F) IWI (compared with Dark, L_ON30, *p = 0.0481; L_OFF20, *p = 0.0141; L_OFF40, **p = 0.0020; L_OFF60, ***p = 0.0001), and (G) mean firing rate (compared with Dark, L_ON10, **p = 0.0013; L_ON30, *p = 0.0109; L_OFF40, *p = 0.0157; L_OFF60, **p = 0.0091). WT P5–P6, n = 11 from eight mice. H, The CI versus the interelectrode distance for spike pairs at Dark, L_ON30, and L_OFF20 time points in WT P5–P6. Data points represent the average of the median value from each retina. I, Cumulative probability plot of highly connected channels (top 15% of total CI) versus the interelectrode distances in three different time points (Dark vs L_ON30, ***p < 0.0001; Dark vs L_OFF20, ***p < 0.0001; L_ON30 vs L_OFF20, **p = 0.0082). J, The percentage of highly connected channels within 600 µm in WT P5–P6. Open circles correspond to the average distances among highly connected channels in each retina (Dark vs L_OFF20, *p = 0.0343). Error bars indicate mean with SEM. D–G, Repeated-measure ANOVA with Holm–Sidak post hoc test compared with the Dark condition; I, Kolmogorov–Smirnov two-sample test; J, repeated-measure ANOVA with Holm–Sidak post hoc test.

Without light stimulation, the RGC responses exhibited a periodic pattern correlated across neighboring channels, consistent with previous observations (Fig. 1B; Meister et al., 1991; Wong et al., 1995; Bansal et al., 2000). During light stimulation, few channels became continuously active, and the overall wave frequency increased (Fig. 1C). Specifically, under light stimulation, the wave-associated burst (WAB) duration was significantly longer after 10 min of light exposure compared with the dark control from P3 to P8 (the main effect of conditions in WT P5–6, F_(5,50) = 40.184; p < 0.0001; η_p² = 0.801; repeat-measure ANOVA; Fig. 1D; Fig. S1A,E), which is consistent with either single-cell or multielectrode recordings from the previous reports (Renna et al., 2011; Kirkby and Feller, 2013). However, the WAB mean firing rate decreased after 10 min of light exposure from P3 to P6 (the main effect of conditions in WT P5–6, F_(5,50) = 17.654; p < 0.0001; η_p² = 0.638; repeat-measure ANOVA; Fig. 1E; Fig. S1B). The combination of an increased burst duration and a reduced burst mean firing rate is attributable to a change in the WAB firing pattern, which exhibited a gentler decreasing slope during light exposure compared with the dark and L_OFF20 conditions (Fig. S3A). After light stimulation, the WAB duration and firing rate returned to dark baseline levels. Interestingly, IWI gradually decreased after light-ON and remained significantly lower than the dark baseline 60 min after light exposure in P5–P6 and P8 retinas (the main effect of conditions in WT P5–6, F_(5,50) = 25.234; p < 0.0001; η_p² = 0.716; repeat-measure ANOVA; Fig. 1F; Fig. S1G). Specifically, when compared with the dark control, the effect size (Cohen's d) for P5–P6 was 1.3805 during light exposure with 15.20% descend (30 min) and 2.2931 postlight exposure with 27.99% descend (60 min), suggesting a large effect of prolonged light exposure on wave frequency. Similarly, the mean firing rate significantly increased during light stimulation and remained higher than the dark baseline after light exposure (the main effect of conditions in WT P5–6, F_(5,50) = 15.761; p < 0.0001; η_p² = 0.612; repeat-measure ANOVA; Fig. 1G; Fig. S1D,H), an effect driven by a corresponding decrease in the IWI.

Additionally, we performed correlation analysis on paired channels, calculating their CIs as a function of distance (Wong et al., 1993; Blankenship et al., 2011; Kirkby and Feller, 2013) under three conditions: Dark, L_ON30, and L_OFF20. Here we observed that the CI curves between adjacent channels showed slight changes following light exposure (Fig. 1H; Fig. S2A,B). However, when focusing on the distance between electrode pairs from the top 15% of CI values, which represent the highly connected channel pairs (Kirkby and Feller, 2013), we observed a distinct shift. Specifically, these highly connected pairs were more confined to shorter interelectrode distances during and postlight exposure compared with the dark baseline (Fig. 1I; Fig. S2C,D). Furthermore, the percentage of highly connected channels within 600 µm was significantly higher in the L_OFF20 conditions (84%) compared with the dark baseline (76.28%; the main effect of conditions in WT P5–6, F_(2,20) = 5.623; p = 0.012; η_p² = 0.360; repeat-measure ANOVA; Cohen's d = 0.93 for P5–6; Fig. 1J; Fig. S2E,F). These findings demonstrate that light exposure modulates Stage 2 retinal wave properties, including IWI and coupling distance, during the first postnatal week before eye-opening.

Impaired light modulating effect on retinal waves in MKO mice

In the developing retina, ipRGCs begin to express melanopsin at Embryonic Day 11 (E11) and exhibit light responsiveness prior to the maturation of rods and cones (McNeill et al., 2011; Lucas and Schmidt, 2019). To investigate whether ipRGCs contribute to the light-induced modulation of cholinergic waves, we conducted similar MEA recordings, as indicated in Figure 1A, using MKO mice. Interestingly, in MKO mice, the WAB duration increased only at the L_OFF20 time point (the main effect of conditions in MKO P5–P6, F_(3,21) = 12.606; p < 0.0001; η_p² = 0.643; repeat-measure ANOVA), while no significant changes were observed in the IWI during light stimulation (the main effect of conditions in MKO P5–P6, F_(3,21) = 1.453; p = 0.256; η_p² = 0.172; repeat-measure ANOVA; Fig. 2A,B), and the WAB firing pattern was similar in each condition (Fig. S3B). Moreover, the reduction in IWI after light exposure was significantly attenuated in MKO mice compared with WT mice (L_ON10, t₍₁₇₎ = −1.604; p = 0.1272; L_ON30, t₍₁₇₎ = −3.194; p = 0.0053; L_OFF20, t₍₁₇₎ = −2.811; p = 0.0120; unpaired t test; Fig. 2C). Additionally, the pairwise distance-to-CI analysis showed minimal differences after light exposure in MKO mice (Fig. 2D,F). There was no significant difference between light off and baseline, although a relatively small distribution shifts in the highly connected channel pairs was observed at the L_ON30 time point (Fig. 2E). Finally, the percentage of highly connected channels within 600 µm showed no significant changes during or postlight exposure compared with the dark control (the main effect of conditions in MKO P5–P6, F_(2,14) = 0.919; p = 0.422; η_p² = 0.116; repeat-measure ANOVA; Fig. 2F). Collectively, these results suggest that melanopsin-expressing ipRGCs are essential for enhancing Stage 2 retinal wave frequency and modulating coupling properties in response to light exposure during early retinal development.

Figure 2.

The MKO mice with attenuated light-induced retinal wave modulation. A, B, (A) The WAB duration (WT, compared with Dark, L_ON10, ***p < 0.0001; L_ON30, ***p = 0.0006; MKO, compared with Dark, L_OFF20, ***p = 0.0009) and (B) IWI (compared with Dark, L_ON30, *p = 0.0481; L_OFF20; *p = 0.0141) in WT and MKO mice. WT P5–P6: n = 11 from eight mice; MKO P5–P6: n = 8 from four mice. C, The percentage of IWI reduction, normalized to the Dark condition, in WT and MKO mice (WT vs MKO, L_ON30, **p = 0.0053; L_OFF20, *p = 0.0120). D, The CI versus the interelectrode distance from MKO P5–P6 mice. E, Cumulative probability plot of highly connected channel pairs to distance in MKO P5–P6 mice (Dark vs L_ON30, *p = 0.0115). F, The percentage of highly connected channels within 600 µm for MKO P5–6 (Dark vs L_ON30, n.s. p = 0.5830; Dark vs L_OFF20, n.s. p = 0.5818). Error bars indicate mean with SEM. A, B, Repeated-measure ANOVA with Holm–Sidak post hoc test compared with Dark condition; C, unpaired t test; E: Kolmogorov–Smirnov two-sample test; F, repeated-measure ANOVA with Holm–Sidak post hoc test.

A dopaminergic pathway mediates the light-dependent Stage 2 retinal wave modulation via D2-like receptors

In adult mice, ipRGCs transmit signals to drive dopaminergic amacrine cells (DACs) through retrograde synaptic connections, thereby regulating light adaptation systems in the retina (Zhang et al., 2012; Prigge et al., 2016). To investigate whether dopaminergic signaling contributes to light-induced retinal wave modulation in early postnatal mice, we recorded neural activity in P5–P6 mouse retinas treated with dopamine in ACSF (Fig. 3A). The IWI decreased after the bath application of 0.1 µM dopamine (0.1 µM dopamine, t₍₉₎ = 4.101; p = 0.0027; paired t test; Fig. 4B), resembling the results observed at the L_ON30 and L_OFF20 time points in WT mice (the main effect of conditions, F_(3,35) = 19.653; p < 0.0001; η_p² = 0.628; one-way ANOVA; Figs. 1, 3C). Additionally, highly connected channels pairs showed distance convergent in the presence of dopamine (Fig. 3D). The percentage of highly connected channels within 600 µm also showed a trend toward increase (0.1 µM dopamine, t₍₉₎ = −2.001; p = 0.0765; paired t test; Fig. 3E). Additionally, the 0.1 µM dopamine application in MKO mice showed the similar results as the WT group (Fig. S4), indicating melanopsin is involved in dopamine release but not in receptivity to dopamine signals. Unexpectedly, the application of 50 µM dopamine caused a dramatic increase in the IWI (50 µM dopamine, t₍₆₎ = −3.038; p = 0.0229; paired t test), opposite to the effect observed with 0.1 µM dopamine and light exposure (Fig. 3B). These findings indicate that low concentration of exogenous dopamine application alters retinal wave properties similar to light exposure.

Figure 3.

Application of exogenously dopamine and dopamine receptor agonists altered the properties of retinal wave. A, The raster plots from MEA recordings in WT P5 retina before (left) and after the bath application of dopamine (DA, 0.1 µM; right). B, The IWI from bath application of 0.1 and 50 µM DA (**p = 0.0027; *p = 0.0229; 0.1 µM, n = 10 from 7 mice; 50 µM, n = 7 from six mice). C, The percentage of IWI reduction, normalized to the dark condition, in L_ON30, L_OFF20, 0.1 µM DA, and 50 µM DA condition (***p < 0.0001 for three comparison). D, E, (D) Cumulative probability plot of highly connected channel pairs to distances (***p = 0.0002) and (E) the percentage of highly connected channels within 600 µm in the control and 0.1 µM DA group. F, The IWI in control and after application of 0.1 µM DA (**p = 0.0027), D1R agonist (SKF 38393, ***p = 0.0010; n = 5 from 3 mice), D2R agonist (quinpirole; n = 8 from 4 mice), and D4R agonist (PD 168077, **p = 0.0011; n = 8 from 5 mice). G, The percentage of IWI reduction, normalized by control, in 0.1 µM DA, D1R agonist, D2R agonist, and D4R agonist (***p < 0.0001 for three comparison). Error bars indicate mean with SEM. B, E, and F, paired t test. C and G, one-way ANOVA with HSD post hoc test. D, Kolmogorov–Smirnov two-sample test.

Figure 4.

The D2-like dopamine receptor antagonists impaired the light-induced retinal wave modulation. A, B, The WAB duration (A) and IWI (B) after blocking D2-like receptor with D2R antagonist, (raclopride; compared with Dark, L_ON10, *p = 0.0135; n = 8 from 6 mice) or D4R antagonist (L-745870; compared with Dark, L_ON10, **p = 0.0013; L_ON30, **p = 0.0456; L_OFF20, *p = 0.0189; n = 9 from 5 mice). C–F, (C, E) Cumulative probability plot of highly connected channel pairs to distances (raclopride: Dark vs L_ON30, ***p = 0.0009; Dark vs L_OFF20, *p = 0.0133) and the percentage of highly connected channels within 600 µm (D, F) under the exposure of D2R and D4R antagonists. Error bars indicate mean with SEM. A, B, D, and F, repeated-measure ANOVA with Holm–Sidak post hoc test. C and E, Kolmogorov–Smirnov two-sample test.

In the rodent retina, dopamine receptors, including D1, D2, and D4, are widely distributed across various cell types (Roy and Field, 2019). While dopamine receptors in the mature retina are known to participate in visual-associated circuits (Witkovsky, 2004; Popova, 2014; Travis et al., 2018), their developmental roles remain unclear. To identify which dopamine receptor subtypes are involved in light-induced retinal wave modulation, we applied specific agonists for D1R (D1 receptor; SKF 38393, 10 µM), D2R (quinpirole hydrochloride, 20 µM), and D4R (PD 168077, 0.5 µM) during MEA recordings. The IWI was significantly reduced after D4R agonist application (PD 168077, t₍₇₎ = 5.342; p = 0.0011; paired t test), corresponding to the effects of light exposure and 0.1 µM dopamine application. Similarly, a trend of IWI reduction was observed with D2R agonist application (quinpirole hydrochloride, t₍₇₎ = 1.979; p = 0.0884; paired t test; Fig. 3F). In contrast, D1R agonist application significantly increased the IWI (SKF 38393, t₍₄₎ = −8.643; p = 0.0010; paired t test; Fig. 3F), consistent with the effects of 50 µM dopamine. The IWI reductions observed with D2R and D4R agonists were similar to the effects of 0.1 µM dopamine, while the D1R agonist produced an opposite effect (the main effect of conditions, F_(3,27) = 78.940; p < 0.0001; η_p² = 0.898; one-way ANOVA; Fig. 3G). Together, these results suggest that D2R and D4R are involved in light-induced retinal wave modulation.

To further confirm the role of D2-like receptors in light-induced retinal wave modulation, we applied D2R (raclopride, 8 µM) and D4R (L-745870, 10 µM) antagonists in ACSF before light exposure. Blocking D2R and D4R did not prevent the increase in WAB duration after light exposure (the main effect of conditions in raclopride, F_(3,21) = 15.699; p < 0.0001; η_p² = 0.692; the main effect of conditions in L-745870, F_(3,24) = 24.781; p < 0.0001; η_p² = 0.756; repeat-measure ANOVA), which remained similar to WT mice without drug application. However, L-745870 slightly reduced WAB duration at the L_OFF20 time point (Fig. 4A). Notably, D2R and D4R antagonists blocked the light-induced reduction in IWI (the main effect of conditions in raclopride, F_(3,21) = 2.718; p = 0.0705; η_p² = 0.280; the main effect of conditions in L-745870, F_(3,24) = 2.823; p = 0.0602; η_p² = 0.261; repeat-measure ANOVA; Fig. 4B). Furthermore, the D4R antagonist blocked light-induced distance convergence and the increase in the percentage of highly connected channels within 600 µm at L_ON30 and L_OFF20 time points (the main effect of conditions in L-745870, F_(2,18) = 0.814; p = 0.459; η_p² = 0.083; repeat-measure ANOVA; Fig. 4E,F). In contrast, the D2R antagonist only blocked the increase in the amount of highly connected channels within 600 µm (the main effect of conditions in raclopride, F_(2,14) = 2.984; p = 0.083; η_p² = 0.299; repeat-measure ANOVA; Fig. 4C,D). These results indicate that D2-like receptors, primarily D4R, control light-induced retinal wave modulation in early postnatal mice.

D4R is highly expressed in the GCL and ON-type SACs during the early postnatal stage

To identify the downstream targets of dopamine involved in light-induced retinal wave modulation, we performed smFISH to assess the expression of D2R and D4R mRNA in the GCL. D2R mRNA levels were higher at P3, P6, P9, and P15 compared with P28 (Fig. 5A,C; Fig. S5A,B). In addition, D4R mRNA was detected in the GCL as early as P3 and remained highly expressed through P9 (Fig. 5B,D). After P15, D4R mRNA expression dramatically decreased and remained low at P28 (Fig. 5D). Since SACs drive Stage 2 retinal waves, we next examined whether ON-type SACs express D2-like receptors during early postnatal development. Colabeling experiments for D2R and D4R mRNA with choline acetyltransferase (ChAT) mRNA, a specific marker for SACs, revealed that ON-type SACs at P6 exhibited significantly higher D4R mRNA expression compared with P15 (Fig. 5E,F). In contrast, D2R mRNA was rarely colocalized with ON-type SACs at both P6 and P15 (Fig. S5C,D). These findings demonstrate that D2-like receptors, particularly D4R, are highly expressed in the GCL, particularly in ON-type SACs during the early postnatal stage.

Figure 5.

D4R is highly expressed in the GCL and ON-type SACs during the early postnatal stage. A, B, Representative images of smFISH for D2R (A) and D4R (B) mRNA in GCL at WT P6 (top) and P15 (bottom). The right panels show higher-magnification images of the areas indicated by the dotted squares in the left panel. Scale bars, 10 µm. C, D, D2R (C) and D4R (D) mRNA abundance measured in area in WT P3, P6, P9, P15, and P28 GCL, normalized by the DAPI area. E, Representative images of smFISH for ChAT and D4R mRNA in GCL at WT P6 (top) and P15 (bottom). The dotted line in zoom images (right) represents the ON-type SAC. Scale bars, 10 µm. F, Puncta analysis for D4R mRNA in ON-type SAC at WT P6 and P15 (***p < 0.0001). WT P6, n = 16 from three mice; WT P15, n = 14 from two mice. Error bars indicate mean with SEM. F: Mann–Whitney U Test.

Light exposure promotes gap junction coupling of SACs via D2-like receptors

Previous studies have shown that blocking gap junctions can dramatically decrease the frequency of Stage 2 retinal waves (Singer et al., 2001). Interestingly, recent research demonstrated that SACs form gap junctions during retinal development, which largely disappear in the adult retina (Maruyama et al., 2023). Therefore, light exposure might influence SAC gap junction coupling to regulate retinal wave properties. To test this hypothesis, we performed tracer coupling experiments on tdTomato-expressing ON-SACs in P5–P6 mice using patch pipettes loaded with neurobiotin (Fig. 6A,B). Under dark conditions, only a few neurons exhibited streptavidin-conjugated fluorescence signals (Fig. 6A, top), indicating limited gap junction coupling to the patched SACs in GCL. Strikingly, after 30 min of light exposure, the number of tracer-coupled neurons significantly increased compared with the dark condition (Fig. 6A, bottom; Fig. 6C, main effect in conditions, F_(1,13) = 5.381; p = 0.0373; two-way ANOVA). Consistent with a previous report (Maruyama et al., 2023), there was no coupling between ON-type SACs before or after light exposure. To further investigate the spatial distribution of the coupled neurons, we normalized their positions relative to the radius of SAC dendritic fields. Under dark conditions, the coupled neurons were predominantly localized within the dendritic field radius, displaying a centralized distribution. In contrast, the spatial distribution expanded significantly following light exposure, including regions inside and outside the dendritic field (Fig. 6D).

Figure 6.

Light exposure promotes gap junction coupling of SACs via D2-like receptors. A, B, Representative images of neurobiotin (NB)-filled ON-type SACs (ChAT-Cre; tdTomato reporter) at dark (top, Dark) and light exposure (bottom, L_ON30) condition in control (A) and the bath application of D2R/D4R blocker cocktail (B). Counterstained with DAPI. Scale bars, 100 µm. C, Total number of coupling cells under the dark and L_ON30 condition from control retina or after D2R/D4R blocker application [Dark (control) vs L_ON30 (control), **p = 0.0010; L_ON30 (control) vs Dark (D2R/D4R blocker), ***p = 0.0003; L_ON30 (control) vs L_ON30 (D2R/D4R blocker), ***p < 0.0001]. D, Histogram of spatial distribution for all NB-filled cells normalized to the radius of injected SAC dendritic field (dark condition, from 5 SACs; light condition from 7 SACs). Error bars indicate mean with SEM. C, Two-way ANOVA with Bonferroni’s post hoc test.

Previous studies have established that the dopaminergic pathway regulates gap junction coupling in the adult mammalian retina, modulating the integration of environmental light signals (Bloomfield and Völgyi, 2009). To determine whether the light-induced increase in gap junction coupling in SACs during early postnatal development is mediated by D2-like receptors, widely expressed in the postnatal GCL (Fig. 5, Fig. S5), we applied a mixed D2/D4 receptor antagonist (raclopride, 8 µM; L-745870, 10 µM) during the tracer coupling experiments. As expected, the application of D2-like receptor antagonists blocked the light-induced increase in SAC gap junction coupling (Fig. 6B). Notably, in the presence of the antagonist, the number of tracer-coupled cells was similar to the dark condition and significantly lower than the light condition without antagonist application (Fig. 6C, main effects between blocks, F_(1,17) = 18.77; p = 0.0005; two-way ANOVA, and condition × block interaction, F_(1,13) = 12.32; p = 0.0038; two-way ANOVA). In summary, our findings demonstrate that light exposure facilitates gap junction coupling in ON-type SACs during early postnatal development, and this effect is dependent on D2-like receptor signaling.

Discussion

Previous studies have demonstrated that ipRGCs play a role in retinal development. For example, MKO mice exhibit impaired eye-specific segregation (Renna et al., 2011). However, the mechanisms by which ipRGCs influence the axonal refinement of regular RGCs in the brain remain unclear. In this study, we provide evidence that ipRGCs play a critical role in modulating Stage 2 retinal waves. Specifically, we showed that WAB duration increased under light exposure and returned to a prestimulation value after light exposure, consistent with previous reports (Renna et al., 2011; Kirkby and Feller, 2013). This immediate WAB duration increase is not affected by the application of the D2 and D4 receptor antagonists. Based on these results, we speculate that mechanism of light-induced WAB duration is different from long-term IWI modulation, which can last for at least 1 h after light exposure. Furthermore, light exposure significantly reduced IWI and enhanced the spatial convergence of highly connected channel pairs compared with dark conditions. Previous studies have shown that the IWI change in nicotinic acetylcholine receptor β2 subunit whole-body knock-out mice is ∼60%, despite this change being associated with a switch from cholinergic to gap junction-dependent wave mechanisms (Kirkby and Feller, 2013; Burbridge et al., 2014). Therefore, the 15–27% reduction in IWI does not appear small when compared with the effect observed in β2 KO mice. Although the light-induced IWI reduction is only 15.2–27.99% after 30 min of light exposure, it has a large effect size (Cohen's d > 0.8) and can be sustained for at least 60 min after lights off, indicating a prolonged modulatory effect. Therefore, the cumulative number of spontaneous waves after prolonged light exposure could meaningfully contribute to the modulation of retinal development. Application of dopamine or D2-like receptor agonists could mimic the light-dependent modulatory effect, while MKO and D2-like receptor antagonists impair these modulations. Additionally, we observed that ON-type SACs express D4R during early postnatal development, and light exposure increased gap junction coupling between SACs and other neurons. Together, these findings suggest that external factors, such as light exposure mediated through ipRGCs, act as key modulatory cues to influence the properties of spontaneous retinal waves.

Our study revealed that MKO mice exhibit a significant impairment in light-induced retinal wave modulation. However, we observed a slight spatial shift in highly connected channel pairs in MKO mice at the L_ON30 time point. This minor shift could result from compensatory mechanisms, such as altered retinal circuitry or opsin expression patterns in MKO mice. Alternatively, OPN5-positive RGCs, another subtype of photosensitive RGCs responsive to violet light (Tarttelin et al., 2003; Yamashita et al., 2010, 2014; Kojima et al., 2011), might contribute to this light-dependent mechanism. OPN5 RGCs have been implicated in light-dependent vascular development in the postnatal eye (Nguyen et al., 2019), and the light stimulation used in our experiments could potentially activate these cells. Nonetheless, our findings from MKO mice strongly suggest that ipRGCs play a crucial role in mediating light-dependent modulation of cholinergic retinal waves.

In the immature retina, dopamine secretion by DACs has been detected as early as P4, with tyrosine hydroxylase (TH) immunoreactivity observed in DACs during this period (Yoshida et al., 2011; Arroyo et al., 2016). Our results show that exogenous application of dopamine or D2-like receptor activation modulates IWI and the spatial distribution of cholinergic waves, closely resembling light-induced wave modulation. Interestingly, the application of 50 µM dopamine or D1R agonists significantly increased IWI, contrary to light-induced effects. These results indicate that D1-like receptor signaling maybe downstream of D2-like receptor signaling pathway or may suppress the functional contribution of D2-like receptor. However, the dopamine dose–response tests from previous study (Arroyo et al., 2016) demonstrated that during the early postnatal period, retinal dopamine concentrations predominantly engage D2-like receptor–mediated mechanisms, owing to their higher binding affinity compared with D1-like receptors (Martel and Gatti McArthur, 2020). Therefore, under normal physiological condition, the dopamine effect should mainly pass through D2-like receptor. How D1-like receptor influence retinal wave under abnormal high concentration of dopamine should be further studied under specific condition. Together, we propose that light exposure induces dopamine release at low concentrations to selectively activate D2-like receptors during retinal development. Antagonist experiments and smFISH analyses further confirm the role of D2-like receptors, particularly D4R, in mediating light-induced retinal wave modulation. While early studies on TH knock-out mice showed no differences in Stage 2 retinal wave frequency compared with WT controls (Stellwagen et al., 1999), our findings suggest that D2-like receptors are required for light-induced IWI reduction. This implies that dopamine may specifically mediate light-dependent retinal wave modulation, while intrinsic wave properties remain unaffected in the absence of light or TH knock-out mice. Taken together, we hypothesize that D2-family receptors play a critical role in light-dependent retinal wave circuits modulation during the early postnatal stage.

In recent years, ipRGCs have been reported to form retrograde synaptic connections with various types of amacrine cells and influence neurotransmitter secretion in targeted neurons, including DACs and corticotropin-releasing hormone–expressing amacrine cells (Zhang et al., 2012; Prigge et al., 2016; Pottackal et al., 2021). In addition, ipRGCs may connect with DACs during the postnatal stage with coalescing dendrites between ipRGCs and DACs at P8 (Tufford et al., 2018). Therefore, ipRGCs may directly activate DACs to regulate dopamine contents at the early developmental stage. However, due to the lack of light-induced c-Fos expression in DACs before the functional synaptic forming of rods and cones (Jain et al., 2023), combined with the observation that a low concentration of dopamine (0.1 µM; Fig. 4C) bath application can mimic light-dependent retinal wave modulation, we speculate an alternative hypothesis that ipRGC may regulate dopamine content through a DAC-independent pathway. A recent study found the existence of TH RGCs in postnatal mice, which are essential for retinal vasculature development (Liang et al., 2023). It is possible that ipRGCs may form the gap junction connection with TH RGCs to regulate dopamine content after light exposure. Alternatively, some ipRGCs may express TH and produce dopamine in the developing retina after light exposure. Therefore, further examination is required to determine how ipRGCs trigger dopamine release in the postnatal retina.

How could light-mediated dopamine signaling affect the cholinergic wave during the first postnatal week? In the mature retina, the electrical coupling is widespread around several types of retinal cells, and the coupling is mainly regulated by the dopaminergic signaling pathway (Bloomfield and Völgyi, 2009). During the developmental stage, ipRGCs form the gap junction microcircuits in the RGC layer, and the extent of gap junction coupling is modulated by the Stage 2 retinal waves (Arroyo et al., 2016). Interestingly, Singer et al. showed that the application of a gap junction blocker reduces the size and frequency of Stage 2 retinal waves (Singer et al., 2001). In our results, the gap junction connections in ON-type SACs are facilitated under light stimulation, and the IWI is reduced after light or dopamine application. These results agree with the gap junction blocker's effect on Stage 2 retinal wave. Therefore, our study suggests that light exposure can modulate retinal waves through D2-like receptor signaling by increasing the gap junction coupling in ON-type SACs.

Previous studies have shown that acute light stimulation can either initiate wave propagation or modulate eye-specific segregation during Stage 3 retinal waves but does not influence Stage 2 retinal waves (Tiriac et al., 2018). However, this finding is inconsistent with another report in MKO mice (Renna et al., 2011). In addition, it has been shown that acute light stimulation can increase the WAB duration of Stage 2 retinal waves without influencing the wave frequency (Renna et al., 2011; Kirkby and Feller, 2013). However, our data demonstrate that a long duration of light exposure (∼30 min) can increase Stage 2 retinal wave frequency. The delayed long-term light effect suggested a potential alteration of gene expression pattern, which may lead to the switching of neurotransmitters. Neuron subtype switch has been observed in recent studies. For example, long-day photoperiod exposure elevates the global neural activity in the paraventricular nucleus of the hypothalamus (PaVN), which changes the expression level of TH (Meng et al., 2018). In our data, the IWI reduction was observed since the L_ON30 time point and maintained 60 min after the light off. Due to the sustained effects of light-induced retinal wave modulation, we hypothesize that external light exposure can alter the expression level of dopamine-producing genes in DACs or TH-positive RGCs. Therefore, light exposure could gradually enhance dopamine production and produce a sustained modulatory effect on the properties of Stage 2 retinal waves. Taken together, our results illustrate a melanopsin–dopamine-associated pathway that modulates SAC gap junction conduction, thereby generating prolonged regulation of the spontaneously cholinergic wave properties (Fig. 7), which is essential for visual system development.

Figure 7.

Schematic of a melanopsin–dopamine-associated pathway in visual system development. Our findings show that environmental light can regulate the key properties of spontaneous Stage 2 retinal waves through a melanopsin–dopamine-associated pathway. Upon light exposure, ipRGCs induce dopamine release. Dopamine subsequently activates D2-like receptors, predominant through D4 receptors, on GCLs and ON-type SACs. This receptor-mediated signaling enhances gap junction coupling between SACs and other neurons, resulting in a shortened IWI and a reduced coupling distance of Stage 2 retinal waves.

Data Availability

This study includes no data deposited in external repositories.