Photostimulation improves maturation of human photoreceptors Authors: Canan Celiker, Eva Hruba, Konstantina Kyriakou, Birthe Dorgau, Francisco Molina Gambin, Kamila Weissova, Edwin van Oosten, Lucie Englmaier, Nadezda Vaskovicova, Michael A. Savage, Gerrit Hilgen, Evelyne Sernagor, Alejandro Garanto, Majlinda Lako, Tomas Barta Published: 2026 ABSTRACT The human retina contains photoreceptor cells that detect light and enable vision. The development of these cells involves a tightly regulated cascade of structural and molecular events, and their dysfunction leads to irreversible blindness in many retinal diseases. Human retinal organoids derived from stem cells have become powerful tools to model retinal development and disease, but they often remain immature and lack key features required for full function. Light is not only the sensory target of photoreceptors but also an important developmental signal in vivo. However, light has rarely been used as a deliberate stimulus during in vitro differentiation. Here we show that exposing retinal organoids to rhythmic light flickering at a specific frequency enhances photoreceptor maturation across multiple levels. This stimulation improves formation of presumptive outer segments, promotes transcriptional shifts from precursor toward more mature photoreceptor states, and is associated with enhanced responsiveness of downstream retinal neurons. These findings identify patterned light as a potent and physiologically relevant signal for driving retinal differentiation in vitro. This approach represents a non-invasive and easily scalable method for improving the quality of retinal organoids, with implications for disease modelling, drug discovery and the preparation of photoreceptors for cell-based therapies. Introduction The human retina is a complex neural tissue that captures light and initiates visual processing through the action of photoreceptor cells - rods and cones - whose development requires precise orchestration of gene expression, polarity, morphogenesis, and synaptogenesis. Retinal organoids derived from human pluripotent stem cells have emerged as powerful models to study these processes in vitro; however, they typically exhibit incomplete structural and functional maturation features1,2. These limitations are especially pronounced in long-term cultures and remain a bottleneck for both fundamental and translational applications. Additionally, they have far-reaching consequences. In disease modelling, immature photoreceptors may fail to replicate key pathological features of inherited retinal disorders, limiting the utility of organoids for studying disease progression or testing therapeutic interventions. In drug discovery, insufficient maturation hampers the ability to assess compound efficacy on functional phototransduction or synaptic signalling. Critically, in the context of photoreceptor transplantation, a promising strategy for restoring vision in degenerative retinal diseases, the ability of donor cells to survive, integrate, and contribute to visual function largely depends on the maturity of cells and synaptic machinery3,4. Therefore, strategies that promote the maturation of photoreceptors in retinal organoids are urgently needed to improve the fidelity, functionality, and clinical relevance of this model system. While advances in organoid engineering have shown that physical and sensory cues, such as mechanical strain, electrical stimulation, or circadian entrainment, can promote the maturation of other stem cell-derived tissues (e.g., cardiomyocytes, neurons, β-cells)5–10, the role of light as a developmental stimulus on the retina has been largely overlooked. This is surprising given that light is not only the physiological target of photoreceptors but also an essential instructive signal during retinal development in vivo11–15. Here, we address this fundamental and previously untested question: Can light, delivered in a controlled, rhythmic pattern, promote the maturation of human photoreceptors and strengthen synaptic connectivity? By applying 40 Hz flickering light in a 12-h light/12-h dark cycle, we demonstrate that photostimulation markedly enhances photoreceptor maturation at structural, transcriptomic, and synaptic levels while strengthening electrophysiological responses downstream of photoreceptors. In doing so, we identify light not only as a sensory end-point but as a developmental signal for guiding human retinal differentiation. Results Light stimulation promotes outer segment–like structure formation Light is a fundamental driver of retinal development12–16, and we hypothesized that controlled stimulation using light could enhance the differentiation and maturation of photoreceptor cells in human retinal organoids. Our proposed model suggests that such stimulation improves the structural and functional properties of the retina, with optimal conditions promoting the formation of outer segment–like structures (hereafter OS-like) and enhancing the expression of photoreceptor-specific markers (Fig. 1a).Fig. 1Light stimulation enhances retinal differentiation and outer segment–like structure formation.a Conceptual schematic illustrating the hypothesis. b Experimental design of light stimulation paradigms. Retinal organoids were exposed to four conditions: complete darkness, constant light, random-frequency flickering light (1–100 Hz), or 40 Hz flickering light, all delivered as white light at 15 lx in a 12 h light/12 h dark cycle for 7 days. Retina-cell figure adapted from Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). c Custom-built light stimulation device (Cell LighteR) used to deliver precisely controlled light exposure to retinal organoids within a standard incubator environment. d Representative scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images showing OS-like structures in retinal organoids cultured under different light conditions. Scale bars: SEM = 20 µm; TEM = 1 µm. OS, outer segment; IS, inner segment; CC, connecting cilium. e Quantification of OS-like length based on CONE ARRESTIN (ARR3) and ROM1 immunostaining. Data are shown as violin plots with embedded box plots and individual measurements overlaid. Each dot represents one OS-like structure. For each condition at least 3 different organoids were measured. The box represents the interquartile range, the centre line indicates the median, and whiskers extend to 1.5 × interquartile range. Red points indicate group means. Statistical significance was assessed using one-way ANOVA followed by Games–Howell post hoc multiple-comparison tests. P values shown in the figure are Holm-adjusted for multiple comparisons. f Representative immunofluorescence images of CONE ARRESTIN and ROM1 corresponding to the quantification shown in (e). Scale bar = 20 µm. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. g Western blot analysis of photoreceptor and synaptic protein expression, including CONE ARRESTIN, ROD ARRESTIN (SAG), RHODOPSIN (RHO), GNAT1, and SNAP25, across all stimulation conditions. Numbers below bands indicate mean fold change relative to the dark condition. VINCULIN serves as a loading control. Experiment was performed using three biological replicates. Uncropped blots including all replicates are shown in Fig. S5. h Principal component analysis (PCA) of bulk RNA-seq data showing segregation of transcriptomic profiles according to light stimulation paradigm. i Heatmap displaying expression of genes associated with photoreceptor outer segment (OS) and inner segment (IS) structures across different light stimulation conditions. Light stimulation enhances retinal differentiation and outer segment–like structure formation. a Conceptual schematic illustrating the hypothesis. b Experimental design of light stimulation paradigms. Retinal organoids were exposed to four conditions: complete darkness, constant light, random-frequency flickering light (1–100 Hz), or 40 Hz flickering light, all delivered as white light at 15 lx in a 12 h light/12 h dark cycle for 7 days. Retina-cell figure adapted from Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). c Custom-built light stimulation device (Cell LighteR) used to deliver precisely controlled light exposure to retinal organoids within a standard incubator environment. d Representative scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images showing OS-like structures in retinal organoids cultured under different light conditions. Scale bars: SEM = 20 µm; TEM = 1 µm. OS, outer segment; IS, inner segment; CC, connecting cilium. e Quantification of OS-like length based on CONE ARRESTIN (ARR3) and ROM1 immunostaining. Data are shown as violin plots with embedded box plots and individual measurements overlaid. Each dot represents one OS-like structure. For each condition at least 3 different organoids were measured. The box represents the interquartile range, the centre line indicates the median, and whiskers extend to 1.5 × interquartile range. Red points indicate group means. Statistical significance was assessed using one-way ANOVA followed by Games–Howell post hoc multiple-comparison tests. P values shown in the figure are Holm-adjusted for multiple comparisons. f Representative immunofluorescence images of CONE ARRESTIN and ROM1 corresponding to the quantification shown in (e). Scale bar = 20 µm. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. g Western blot analysis of photoreceptor and synaptic protein expression, including CONE ARRESTIN, ROD ARRESTIN (SAG), RHODOPSIN (RHO), GNAT1, and SNAP25, across all stimulation conditions. Numbers below bands indicate mean fold change relative to the dark condition. VINCULIN serves as a loading control. Experiment was performed using three biological replicates. Uncropped blots including all replicates are shown in Fig. S5. h Principal component analysis (PCA) of bulk RNA-seq data showing segregation of transcriptomic profiles according to light stimulation paradigm. i Heatmap displaying expression of genes associated with photoreceptor outer segment (OS) and inner segment (IS) structures across different light stimulation conditions. To test this hypothesis, we subjected human retinal organoids (Fig. S1a–d) to four different conditions: complete darkness, constant light, random-frequency flickering light (1–100 Hz), and flickering light at 40 Hz, all conditions using white light at 15 lx. We selected this gamma-band oscillation at 40 Hz because it represents a fundamental feature of neuronal network dynamics, known to facilitate sensory processing and synaptic plasticity17–19. The light intensity of 15 lx was chosen to approximate intrauterine illuminance levels (10–100 lx)20, thereby providing a physiologically relevant light environment for retinal organoids at the differentiation stage examined in this study. All conditions were applied in a 12-h light/12-h dark cycle for 7 consecutive days to mimic natural day/night rhythms, and organoids were harvested at the end of the dark phase to minimize acute light-induced effects (Fig. 1b). All experiments were performed using our custom-built Cell LighteR device21, designed to deliver precise and controlled light stimulation to organoid cultures (Fig. 1c). All experiments were conducted using organoids between day (D) 150 and D180 of differentiation (unless stated otherwise), corresponding to a stage at which post-mitotic CRX⁺ photoreceptors are present and both inner segments (IS) and OS-like structures are already detectable but remain short and immature2,22–24. Consistent with human retinal development in vivo, photoreceptor OS-like structures at this stage are expected to be short and structurally immature. In human foetal retina, rod OS begin to emerge around 24–25 gestational weeks (D168-175), but remain incomplete25. The presence of OS-like structures in organoids at this stage is supported by scanning electron microscopy (SEM) (Fig. 1d, S1c), transmission electron microscopy (TEM) (Fig. 1d, Fig. S2), immunofluorescence staining (IF) (Figs. 1f, S1d, S3a-d), and bright-field (BF) imaging (Figs. S1b, S3e). Notably, retinal organoids become transcriptionally and electrophysiologically responsive to light between D120 and D15021,26, aligning with the onset of photoreceptor maturation. To assess the impact of light stimulation on photoreceptor maturation, we used OS-like length and structure as a rapid and reliable readout, analysing their morphology with SEM and TEM (Figs. 1d, S2), while IF was used for OS-like length quantification (Figs. 1e, f, S3). Dark-cultured organoids displayed short OS-like structures, while all light-treated groups exhibited significant improvements in their length. Notably, the 40 Hz flickering condition resulted in the most pronounced OS-like elongation relative to dark controls (Figs. 1d–f, S3). Because presumptive OS at D150–180 are still immature and difficult to delineate using a single marker, we also used a multi-marker strategy to define OS-like boundaries, combining ARL13B (connecting cilium) and CRALBP (outer limiting membrane) with ROM1 or CONE ARRESTIN. This approach enables more accurate identification of OS–IS boundaries compared to using photoreceptor markers alone (Fig. S3c, d). To validate ARL13B localization, we performed additional stainings using multiple antibodies, which showed restriction to the connecting cilium (Fig. S3a). Compared to the dark condition, all light-exposed groups showed significantly elongated presumptive OS, with the most pronounced effect observed under 40 Hz flickering light. We additionally validated the observed elongation using an independent BF imaging approach (Fig. S3d–f). Measurements confirmed that the 40 Hz stimulation condition induced elongation of OS-like structures, which was directly visible in BF images (Fig. S3d). Quantification of OS-like length before and after the experimental period demonstrated a clear increase following 7 days of flickering light at 40 Hz, when compared to organoids maintained in darkness (Fig. S3e, f). OS-like structures appeared longer in bright-field imaging compared to immunofluorescence, likely reflecting differences in sample processing and fixation that can affect the morphology of these delicate structures. Because cones have naturally shorter OS than rods in the retina, we next asked whether the shorter OS-like structures in dark-cultured organoids could be explained by differences in rod–cone proportions. Quantification of ARR3⁺ cones and RHO⁺ rods using QuPath-assisted segmentation (Fig. S4a) showed that rods constituted ~76–82% of ONL nuclei and cones ~19–24%, with no differences across stimulation conditions. This finding was independently supported by single-cell–based cell-type quantification, which likewise revealed no change in the rod–cone ratio (Fig. S4b). Immunostaining for critical photoreceptor proteins further confirmed the differentiation-enhancing effect of light. Compared to dark-cultured organoids, all light conditions resulted in increased expression of key photoreceptor markers, including RHODOPSIN (RHO), GNAT1, PRPH2, and PDE6A (Fig. S3b). The expression levels, determined by Western blot, of CONE ARRESTIN, ROD ARRESTIN (gene SAG), RHODOPSIN, GNAT1, and the synaptic marker SNAP25 were all elevated in light-treated conditions compared to the dark group (Fig. 1g, S5). Among all conditions, 40 Hz flickering light led to the highest protein expression levels. To obtain a global overview of transcriptomic changes, we conducted bulk RNA sequencing. Principal component analysis (PCA) of the transcriptome clearly segregated samples by stimulation condition (Fig. 1h). The 40 Hz light group formed a distinct cluster, indicating that this condition induces the most pronounced and consistent transcriptional changes among all tested paradigms. Expression changes of genes associated with photoreceptor OS and photoreceptor IS structures, as defined by Gene Ontology (GO) annotations (https://www.ebi.ac.uk/QuickGO/), were visualized using a heatmap (Fig. 1i). The majority of these genes were significantly upregulated under light stimulation, with the strongest induction observed in the 40 Hz flickering condition. Importantly, all experiments were conducted in phenol red-free medium, as we observed that the presence of phenol red markedly interfered with the effects of light. When experiments were carried out in medium containing phenol red, no profound differences in gene expression were detected across conditions. Specifically, the expression levels of key photoreceptor genes RHODOPSIN, OPN1SW, and SAG remained uniformly low, indicating that light failed to elicit the expected transcriptional upregulation in the presence of phenol red (Fig. S6). Flickering light at 40 Hz induces cell-type specific responses in retinal organoids Among all tested conditions, 40 Hz flickering light consistently demonstrated the most pronounced effects on photoreceptors (Figs. 1, S3). Therefore, we selected the 40 Hz condition as the most effective stimulation paradigm and focused subsequent analyses on uncovering its cell-type-specific impact. We performed single-cell RNA sequencing (scRNA-seq) on human retinal organoids subjected to either dark or 40 Hz flickering light (further referred as photostimulation) in a 12-h light/12-h dark cycle for 7 days (Fig. 2a). Using the human retina atlas gene annotations27,28, we identified 13 distinct retinal cell types, including rod photoreceptors, cone photoreceptors, rod and cone precursors, horizontal cells, amacrine cells, retinal ganglion cells, (ON and OFF) bipolar cells, Müller glia, astrocytes, retinal pigment epithelium (RPE), and retinal precursors (Figs. 2b, c, S7; and Supplementary Data 1). Interestingly, photostimulation led to an increase in rod and cone photoreceptors numbers at the expense of their respective precursor populations, in line with enhanced photoreceptor maturation under flickering light conditions (Figs. 2d, 3c). A small decrease in the proportion of Müller glia was observed in the scRNA-seq dataset. However, this difference was modest and does not indicate a dramatic loss of Müller glia. Consistent with this, CRALBP immunostaining and Müller glia–specific gene expression confirmed preservation of this population across all conditions (Fig. S8). Moreover, Müller glia did not exhibit transcriptional signatures of apoptosis or oxidative stress (Fig. S9)Fig. 2Single-Cell RNA sequencing reveals specific responses to photostimulation.a Experimental setup for scRNA-seq analysis. Retinal organoids were cultured under dark or 40 Hz flickering light (Photostimulation) conditions for 7 days (12-h light/12-h dark cycle), followed by dissociation and scRNA-seq. b UMAP projection of retinal organoid cells, identifying 13 distinct retinal cell types. c Dot plot showing expression of canonical marker genes used to validate cell-type annotations across retinal populations. d Cell-type abundance analysis comparing dark vs. photostimulated organoids. e Differential gene expression analysis comparing dark vs. photostimulated organoids. f Waterfall plot displaying the top 30 differentially expressed genes in each condition. g Gene Ontology (GO) enrichment analysis for biological processes. h GO enrichment analysis for cellular components. For both panels (g, h)—Enrichment P values were calculated using a one-sided hypergeometric test/Fisher’s exact test, using all detected genes as the background universe. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method. GO terms with adjusted P < 0.05 and q < 0.2 were considered enriched.Fig. 3Photostimulation promotes maturation of rod and cone photoreceptors.a UMAP embedding of sub-clustered photoreceptors, identifying four transcriptionally distinct populations: rod photoreceptors, cone photoreceptors, rod photoreceptor precursors, and cone photoreceptor precursors. b Pseudotime trajectory analysis overlaid onto the UMAP space reveals a continuous developmental progression from precursor to photoreceptor states. c Number of cells in each photoreceptor or precursor cluster under dark and photostimulated conditions, showing enrichment of photoreceptors in the photostimulated group and accumulation of precursors in the dark condition. d Schematic summary of gene expression changes in each photoreceptor population. For each subtype, three waterfall plots depict photostimulation-induced transcriptional changes in major genes associated with outer segments (OS), inner segments (IS), and ribbon synapses (Ribb. Syn.). Bars indicate pseudobulk log10 fold changes calculated from aggregated expression values across photoreceptor populations. Each panel is independently scaled to visualize relative changes within a given functional gene set. To allow direct comparison of effect sizes across functional gene sets without independent rescaling, the same data are additionally shown as a dot plot in Fig. S13. Retina-rod-cell and retina-cone-cell figures adapted from Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) e Immunostaining for DNM1, RIMS2, IMPG1, RHODOPSIN (RHO), GNAT1, and ROD ARRESTIN (ROD AR.), CONE ARRESTIN (CONE AR.) in retinal organoid sections. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 20 μm. Graphs below show quantification of immunostainings. Each dot represents individual organoid. The box indicates the interquartile range, the centre line the median, and whiskers 1.5 × interquartile range. Red dots denote mean values. Single-Cell RNA sequencing reveals specific responses to photostimulation. a Experimental setup for scRNA-seq analysis. Retinal organoids were cultured under dark or 40 Hz flickering light (Photostimulation) conditions for 7 days (12-h light/12-h dark cycle), followed by dissociation and scRNA-seq. b UMAP projection of retinal organoid cells, identifying 13 distinct retinal cell types. c Dot plot showing expression of canonical marker genes used to validate cell-type annotations across retinal populations. d Cell-type abundance analysis comparing dark vs. photostimulated organoids. e Differential gene expression analysis comparing dark vs. photostimulated organoids. f Waterfall plot displaying the top 30 differentially expressed genes in each condition. g Gene Ontology (GO) enrichment analysis for biological processes. h GO enrichment analysis for cellular components. For both panels (g, h)—Enrichment P values were calculated using a one-sided hypergeometric test/Fisher’s exact test, using all detected genes as the background universe. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method. GO terms with adjusted P < 0.05 and q < 0.2 were considered enriched. Photostimulation promotes maturation of rod and cone photoreceptors. a UMAP embedding of sub-clustered photoreceptors, identifying four transcriptionally distinct populations: rod photoreceptors, cone photoreceptors, rod photoreceptor precursors, and cone photoreceptor precursors. b Pseudotime trajectory analysis overlaid onto the UMAP space reveals a continuous developmental progression from precursor to photoreceptor states. c Number of cells in each photoreceptor or precursor cluster under dark and photostimulated conditions, showing enrichment of photoreceptors in the photostimulated group and accumulation of precursors in the dark condition. d Schematic summary of gene expression changes in each photoreceptor population. For each subtype, three waterfall plots depict photostimulation-induced transcriptional changes in major genes associated with outer segments (OS), inner segments (IS), and ribbon synapses (Ribb. Syn.). Bars indicate pseudobulk log10 fold changes calculated from aggregated expression values across photoreceptor populations. Each panel is independently scaled to visualize relative changes within a given functional gene set. To allow direct comparison of effect sizes across functional gene sets without independent rescaling, the same data are additionally shown as a dot plot in Fig. S13. Retina-rod-cell and retina-cone-cell figures adapted from Servier Medical Art (https://smart.servier.com), licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) e Immunostaining for DNM1, RIMS2, IMPG1, RHODOPSIN (RHO), GNAT1, and ROD ARRESTIN (ROD AR.), CONE ARRESTIN (CONE AR.) in retinal organoid sections. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 20 μm. Graphs below show quantification of immunostainings. Each dot represents individual organoid. The box indicates the interquartile range, the centre line the median, and whiskers 1.5 × interquartile range. Red dots denote mean values. To determine whether photostimulation alters overall retinal cell-type composition, we performed an independent immunohistochemical cell-count analysis using established markers for major retinal cell types. Quantification of cone photoreceptors (CONE ARRESTIN), rod photoreceptors (RHODOPSIN), retinal ganglion cells (NEUN), and bipolar cells (VSX2), normalized to DAPI-positive nuclei within the corresponding retinal layers, revealed no significant differences between dark-cultured and photostimulated organoids (Fig. S10). These results indicate that photostimulation promotes maturation within existing photoreceptor populations rather than altering retinal cell fate decisions or global cellular composition. To identify genes differentially expressed between dark and photostimulated conditions, we first analysed global transcriptomic changes across all cells (Fig. 2e, f). Key photoreceptor-associated genes were upregulated in response to flickering light exposure, including PDE6G, SAG, PDE6A, PRPH2, ARR3, TTR, RHO, ESRRB, and GUCA1B. These genes are crucial for phototransduction, OS development, and photoreceptor maintenance. GO analysis of all significantly differentially expressed genes (p < 0.05) revealed strong enrichment for sensory perception of light, visual perception, detection of light stimulus, phototransduction, and photoreceptor cell maintenance, supporting the role of flickering light in promoting photoreceptor-driven signalling (Fig. 2g). Additionally, GO analysis for cellular components showed that differentially expressed genes were significantly associated with photoreceptor OS, photoreceptor cilia, photoreceptor disc membranes, and ciliary membranes, highlighting structural improvements and enhanced expression of phototransduction components following photostimulation (Fig. 2h). Interestingly, oxidative stress and apoptosis-related genes were not broadly upregulated across retinal cell types; however, a subset of these genes was increased specifically in the RPE population (Fig. S9a, b), supporting known light-responsive behaviour of RPE cells and confirming effective photostimulation29. To assess whether the effects of photostimulation persist beyond the stimulation period, retinal organoids were exposed to flickering light for 7 days and subsequently maintained for an additional 21 days in complete darkness (Fig. S11a). Quantification of OS-like length revealed that organoids exposed to flickering light retained elongated presumptive OS compared to age-matched dark controls even after this prolonged dark recovery period (Fig. S11b, c). These findings indicate that a transient (7 days) exposure to light induces stable changes rather than acute responses. In contrast, when flickering light exposure was applied to late Stage 3 retinal organoids30 (D280) for 7 days, only a minimal increase in OS-like length was observed (Fig. S12a, b). Moreover, RT-qPCR analysis revealed no robust upregulation of photoreceptor-associated genes, with SAG being the only marker showing a significant increase (Fig. S12c). These results indicate that flickering light exposure has a reduced impact on photoreceptor maturation at later differentiation stages. Photostimulation improves maturation of photoreceptors To investigate the effect of photostimulation on photoreceptors, we focused our single-cell transcriptomic analysis on rod and cone lineages. Sub-clustering of the photoreceptor population identified four transcriptionally distinct states corresponding to rod and cone photoreceptors and their respective precursor populations (Fig. 3a). Pseudotime trajectory analysis revealed a continuous developmental progression from precursor states toward more mature photoreceptor identities (Fig. 3b). Notably, cells from photostimulated organoids were enriched at later pseudotime positions, whereas dark-cultured cells preferentially occupied earlier stages along the trajectory. In line with this shift, quantitative analysis showed an increased proportion of rod and cone photoreceptors and a relative depletion of precursor populations following photostimulation (Fig. 3c). These results indicate that photostimulation biases photoreceptor populations toward more mature transcriptional states at this differentiation stage. To examine subtype-specific responses to light in more detail, we generated schematic representations of each photoreceptor category, overlaying gene expression changes induced by photostimulation (Figs. 3d, S13). Waterfall plots illustrated upregulation of major genes associated with OS, IS, and ribbon synapse development across all cell subtypes. These included functional markers such as RHO, SAG, GNAT1, ARR3, PDE6A, IMPG1, RIMS2, and DNM1, highlighting enhanced expression of genes linked to OS formation and synaptic organization in response to stimulation. To validate these transcriptomic findings at the protein level, we performed immunostaining for RHODOPSIN, GNAT1, ROD ARRESTIN, CONE ARRESTIN, DNM1, RIMS2, and IMPG1—key proteins involved in OS, synaptic vesicle recycling, ribbon synapse organization, and interphotoreceptor matrix composition, respectively. All markers showed visibly increased expression and broader distribution in photostimulated retinal organoids compared to dark controls, with strongest signal observed in layers corresponding to the outer nuclear layer and synaptic regions (Fig. 3e). These results support the conclusion that photostimulation enhances maturation of photoreceptors at both transcriptional and protein levels. To validate the maturation trajectory, we assessed the spatial expression of canonical photoreceptor markers across the UMAP embedding. CRX, a key regulator of photoreceptor fate, was broadly expressed across all photoreceptor populations (Fig. S14a). In contrast, SAG (rods) and ARR3 (cones) were restricted to differentiated cells, while PTCHD4, FOXD1, and BHLHE40 were confined to precursor clusters (Fig. S14a, b). Differential expression analysis between precursors and photoreceptors confirmed the upregulation of phototransduction, synaptic, and structural gene networks in differentiated cells (Fig. S14b). To evaluate how stimulation modulates gene expression within each photoreceptor type, we compared transcriptomic profiles of photostimulated and dark-cultured organoids. UMAP plots with overlaid condition labels showed an enrichment of photostimulated cells in photoreceptor clusters and a reciprocal enrichment of dark-cultured cells in precursor clusters (Fig. S15a, b). Expression of key functional genes, including SAG and ARR3, was elevated in photostimulated cells, whereas precursor-associated genes (PTCHD4, FOXD1, BHLHE40) were diminished (Fig. S15c–f). To further validate increased photoreceptor maturation, we performed an integrated scRNA-seq analysis comparing photoreceptors from dark-cultured organoids, photostimulated organoids, and adult human retina (a dataset published by Lukowski et al.)28 (Fig. S16). In the integrated UMAP, adult rods and cones formed clusters with minimal precursor-like features, whereas organoid-derived photoreceptors occupied intermediate transcriptional states (Fig. S16a–c). Expression of canonical markers confirmed these identities, with CRX broadly expressed, SAG and ARR3 enriched in mature photoreceptors, and BHLHE40, PTCHD4, and FOXD1 restricted to precursors (Fig. S16b). Centroid distance analysis showed that both rod and cone populations from photostimulated organoids were transcriptionally closer to adult photoreceptors than those from dark-cultured organoids (Fig. S16d). Photostimulation enhances electrophysiological responses downstream of photoreceptors Having observed enhanced photoreceptor maturation in response to photostimulation, we next assessed whether these changes translate into enhanced downstream retinal responses, using electrophysiological readouts. We first examined the expression of genes associated with synaptic transmission, with particular focus on components of the photoreceptor ribbon synapse. Comparison of dark and photostimulated conditions revealed profound upregulation of key synaptic genes, including STX3, BSN, SYP, and DLG4 (PSD95), in both rods and cones following stimulation (Fig. 4a). These genes are essential for vesicle docking, neurotransmitter release, and synaptic scaffold formation, suggesting that stimulation enhances synaptogenesis and synaptic maturity in photoreceptors. To validate transcriptional changes at the protein level, we performed immunostaining for synaptic markers in retinal organoids cultured under either dark or photostimulated conditions. STX3 and BSN showed elevated expression between the outer nuclear layer (ONL) and the inner nuclear layer (INL) (presumptive outer plexiform layer) of photostimulated organoids compared to dark controls (Fig. 4b, c). Notably, STX3 signal was also detected within the ONL and OS-like. This localization is consistent with previous reports in human retina, where STX3 is not restricted to synaptic compartments but is also present in photoreceptor inner and outer segment regions, reflecting its broader role in membrane trafficking and vesicle fusion31,32. The relatively low expression of STX3 and BSN in dark-cultured organoids likely reflects the immature state of photoreceptor synapses under these conditions, rather than defective protein trafficking. In contrast, photostimulation leads to a marked increase in their expression and more pronounced localization at the presumptive outer plexiform layer, in line with enhanced synaptic maturation.Fig. 4Photostimulation enhances synaptic marker expression and RGC responsiveness.a Dot plot showing differential expression of synaptic genes, including ribbon synapse components, in rod and cone photoreceptors under dark and photostimulated conditions. b Immunostaining for synaptic markers STX3 and BSN in retinal organoid sections. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 20 μm. c Graphs show quantification of immunostainings form (b). Each dot represents individual organoid. The box indicates the interquartile range, the centre line the median, and whiskers 1.5 × interquartile range. Red dots denote mean values. d Schematic illustration of the experimental setup for MEA recordings, including delivery of light stimulation. CorePlate™ 1 W graphics reproduced with permission from 3Brain AG, Pfäffikon, Switzerland. e Representative raster plots and corresponding peristimulus time histograms illustrating RGC spiking activity during alternating dark and light periods in dark-cultured and photostimulated organoids. f Quantification of light-evoked spiking responses of ON and OFF RGCs. Data are shown as violin plots with embedded box plots and individual measurements overlaid. The centre line indicates the median, the box represents the interquartile range, and notches indicate the approximate 95% confidence interval of the median. Whiskers extend to 1.5 × interquartile range. Red points represent the mean ± 95% bootstrap confidence interval. Statistical significance was assessed using two-sided Welch’s t-test. Photostimulation enhances synaptic marker expression and RGC responsiveness. a Dot plot showing differential expression of synaptic genes, including ribbon synapse components, in rod and cone photoreceptors under dark and photostimulated conditions. b Immunostaining for synaptic markers STX3 and BSN in retinal organoid sections. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar = 20 μm. c Graphs show quantification of immunostainings form (b). Each dot represents individual organoid. The box indicates the interquartile range, the centre line the median, and whiskers 1.5 × interquartile range. Red dots denote mean values. d Schematic illustration of the experimental setup for MEA recordings, including delivery of light stimulation. CorePlate™ 1 W graphics reproduced with permission from 3Brain AG, Pfäffikon, Switzerland. e Representative raster plots and corresponding peristimulus time histograms illustrating RGC spiking activity during alternating dark and light periods in dark-cultured and photostimulated organoids. f Quantification of light-evoked spiking responses of ON and OFF RGCs. Data are shown as violin plots with embedded box plots and individual measurements overlaid. The centre line indicates the median, the box represents the interquartile range, and notches indicate the approximate 95% confidence interval of the median. Whiskers extend to 1.5 × interquartile range. Red points represent the mean ± 95% bootstrap confidence interval. Statistical significance was assessed using two-sided Welch’s t-test. To determine whether these changes translate into functional circuit output, we next performed high-density multielectrode array (MEA-HD) recordings on retinal organoids maintained under dark or photostimulated conditions. Recordings were carried out using a combination of red (620–625 nm) and green (522–525 nm) light stimulus (Fig. 4d), as these wavelengths activate responses while minimizing contributions from intrinsically photosensitive RGCs (ipRGCs), which are primarily blue-light responsive33,34. Light-responsive RGC units were classified based on their firing behaviour during light stimulation. Units that exhibited an increase in spiking activity (≥50%) during the light period were classified as ON responses, whereas units showing a decrease in firing (≥50%) were classified as OFF responses, with activity recovering upon return to darkness (Figs. 4e, S17a). Quantitative analysis demonstrated a statistically significant, although modest, increase in light-evoked spiking frequency of ON RGCs in photostimulated organoids compared to dark controls (Fig. 4f). In contrast, although OFF cells showed a trend toward increased responsiveness, this change did not reach statistical significance, indicating a preferential enhancement of ON-pathway activity following photostimulation. To assess the presence of RGCs and evaluate a potential contribution of ipRGCs, we analysed retinal organoids using transcriptomic and immunohistochemical approaches. RGCs are known to be particularly vulnerable in vitro, often undergoing degeneration during prolonged organoid culture35. We therefore specifically assessed their presence in retinal organoids. Analysis of bulk and single-cell RNA sequencing data revealed expression of multiple established RGC markers, including RBPMS, ISL1, RBFOX3, POU6F2, and SNCG (Fig. S18a, b), supporting the presence of RGC populations in these organoids. At the protein level, immunostaining confirmed the presence of neuronal populations within the GCL, including SNCG-positive cells (Fig. S18c). While NEUN staining labels a broader neuronal population (potentially including amacrine cells), the combined transcriptional and protein-level data support the presence of RGCs in the organoids. In contrast, OPN4 expression was only minimally detected in bulk RNA sequencing and was not detected in single-cell RNA sequencing, with no identifiable OPN4-positive cells (Fig. S18a, b), indicating that ipRGCs are absent or rare in retinal organoids. Furthermore, ipRGCs are characterized by delayed and sustained responses to light36,37, whereas our stimulation paradigm (5 s light/5 s dark; Fig. 4) probes rapid, transient responses typical for photoreceptor-driven signalling. Together, these observations argue against a major contribution of ipRGC-mediated activity to the recorded responses. To further confirm that the recorded activity originated from photoreceptor-driven pathways, we applied 8-br-cGMP, a membrane-permeable analogue of cGMP (Fig. S18d). cGMP is a key second messenger in the phototransduction cascade, gating Na⁺-permeable channels in photoreceptor OS. Puffing 8-br-cGMP mimics the dark current by depolarizing photoreceptors through Na⁺ influx, thereby initiating neurotransmitter release to downstream neurons26. This approach allows for the selective activation of photoreceptor-driven circuitry independent of actual light exposure. RGCs that responded to 8-br-cGMP with a ≥ 50% decrease in spiking activity were classified as presumptive ON cells, consistent with their expected hyperpolarization upon photoreceptor activation. Conversely, RGCs that exhibited a ≥ 50% increase in firing rate were categorized as OFF cells (Fig. S17b). Organoids exposed to 8-br-cGMP exhibited ON and OFF RGC responses, with increased spiking observed in OFF cells and decreased firing in ON cells in the presence of cGMP (Fig. S18e, f). Importantly, photostimulated organoids again exhibited a selective and significant enhancement of ON-cell responsiveness, while OFF-cell activity remained unchanged relative to dark controls (Fig. S18f). These responses are consistent with enhanced photoreceptor-to-bipolar-to-RGC signalling, as supported by upregulation of synaptic markers and the presence of RGC populations in bulk, scRNA-seq data, and immunostainings. Although absolute firing rates in the 8-br-cGMP experiments were low, this is in agreement with previous reports describing weak spontaneous and evoked activity in human retinal organoids at comparable stages21,26. Baseline and evoked firing rates observed in the 8-br-cGMP experiments were lower than those recorded during light-based stimulation. This difference likely reflects methodological factors, as cGMP recordings were performed in artificial cerebrospinal fluid (aCSF), whereas light-stimulation recordings were conducted in BrainPhys™ Neuronal Medium, which supports higher neuronal excitability. In addition, variability between pluripotent stem cell lines and differences in organoid RGC maturation state used in these experiments may contribute to baseline differences in firing rates. Photostimulation is sufficient to entrain circadian gene expression in retinal organoids Our photostimulation paradigm was applied using a 12-h light/12-h dark cycle, a strategy designed not only to mimic natural diurnal rhythms but also to engage the intrinsic circadian machinery of the retina. In vivo, photoreceptors are tightly regulated by the circadian clock, which coordinates daily fluctuations in gene expression, OS turnover, and synaptic transmission38,39. We therefore hypothesized that retinal organoids exposed to rhythmic light might entrain endogenous circadian oscillators, further supporting their maturation and physiological fidelity. To investigate whether human retinal organoids possess a functional molecular clock that can respond to external cues, we employed a BMAL1-driven luciferase reporter system. Human iPSCs were transduced with a lentiviral construct encoding the BMAL1::luciferase cassette, and luciferase activity was monitored in real time using a custom-adapted version of the LuminoCell40—a platform capable of measuring luciferase activity in individual organoids in a 96-well plate format (Fig. 5a). We first examined early-stage retinal organoids (day 20) to assess whether the core circadian transcriptional machinery was active or inducible. While untreated controls showed no rhythmicity, organoids synchronized with dexamethasone (DEX) displayed robust circadian oscillations in BMAL1-driven bioluminescence (Fig. 5b), indicating that a functional molecular oscillator is already present at this early stage.Fig. 5Retinal organoids possess functional circadian machinery responsive to photic cues.a Photograph of a custom-modified LuminoCell device adapted for bioluminescence monitoring of individual retinal organoids in a 96-well format. The device continuously records luciferase activity driven by a BMAL1 promoter. b Bioluminescence traces from five day-20 retinal organoids expressing a BMAL1::luciferase reporter, either synchronized with dexamethasone (+DEX) or left unsynchronized (−DEX). Rhythmic oscillations are observed only in the synchronized group. Cosinor analysis confirms statistically significant rhythmicity in the + DEX condition. Five biological replicates were measured at each time point. Mean values across replicates were used for cosinor fitting. Coloured lines represent individual replicates, and the red curve represents a fitted 24-h cosinor model calculated from the mean response across replicates at each time point. Rhythmicity was assessed using a two-sided cosinor linear regression model including cosine and sine terms with a fixed 24-h period. No correction for multiple comparisons was applied. c qPCR analysis of PER1 expression in day-180 retinal organoids following circadian synchronization. Organoids were exposed to either phasic or antiphasic 12 h light/12 h dark cycles for 72 h, followed by monitoring under constant darkness. Dexamethasone-treated organoids served as a positive control. Both light-synchronized groups exhibited distinct oscillatory patterns of PER1 expression, indicating successful circadian entrainment by light. Each dot represents mean from three biological replicates. Error bars represent +/− SD (standard deviation of mean). Retinal organoids possess functional circadian machinery responsive to photic cues. a Photograph of a custom-modified LuminoCell device adapted for bioluminescence monitoring of individual retinal organoids in a 96-well format. The device continuously records luciferase activity driven by a BMAL1 promoter. b Bioluminescence traces from five day-20 retinal organoids expressing a BMAL1::luciferase reporter, either synchronized with dexamethasone (+DEX) or left unsynchronized (−DEX). Rhythmic oscillations are observed only in the synchronized group. Cosinor analysis confirms statistically significant rhythmicity in the + DEX condition. Five biological replicates were measured at each time point. Mean values across replicates were used for cosinor fitting. Coloured lines represent individual replicates, and the red curve represents a fitted 24-h cosinor model calculated from the mean response across replicates at each time point. Rhythmicity was assessed using a two-sided cosinor linear regression model including cosine and sine terms with a fixed 24-h period. No correction for multiple comparisons was applied. c qPCR analysis of PER1 expression in day-180 retinal organoids following circadian synchronization. Organoids were exposed to either phasic or antiphasic 12 h light/12 h dark cycles for 72 h, followed by monitoring under constant darkness. Dexamethasone-treated organoids served as a positive control. Both light-synchronized groups exhibited distinct oscillatory patterns of PER1 expression, indicating successful circadian entrainment by light. Each dot represents mean from three biological replicates. Error bars represent +/− SD (standard deviation of mean). To determine whether light could act as an entraining cue in more mature organoids, we exposed day 180 retinal organoids to a 72-h light-dark cycle, sufficient to synchronize circadian rhythms8,9, followed by 36 h of sampling under constant darkness at 4-h intervals. Using qPCR, we tracked the expression of PER1, a canonical circadian clock gene, in organoids that had been synchronized either in phase or in antiphase. In both cases, we observed clear oscillations in PER1 transcript levels, while dexamethasone-treated samples served as positive controls (Fig. 5c). These findings demonstrate that light is sufficient to entrain circadian gene expression in retinal organoids and that phase relationships are preserved across experimental conditions. Together, these results confirm that the circadian machinery in human retinal organoids is functional and light responsive. The use of a 12-h light/dark stimulation protocol, therefore, serves a dual role, both as a physiological entrainment signal and as a maturation-enhancing cue, supporting the idea that rhythmic environmental inputs can drive developmental and functional refinement in stem cell-derived neural tissues, however, the contribution of circadian entrainment to photostimulation-induced photoreceptor maturation remains to be determined. Discussion In this study, we demonstrate that temporally structured light, specifically 40 Hz flickering light, enhances maturation of human photoreceptors in retinal organoids. Our findings reveal that light is not only the physiological target of photoreceptors but also a potent environmental cue capable of guiding their differentiation. Among several stimulation paradigms tested, including constant illumination and randomly flickering light, gamma-frequency flicker (40 Hz) emerged as uniquely effective in promoting OS-like elongation, increasing expression of photoreceptor-specific genes, and producing a modest but reproducible enhancement of light-evoked responses in downstream retinal neurons. These effects were accompanied by increased synaptic protein expression and enhanced responsiveness of downstream retinal ganglion cells. The concept of using environmental stimulation to promote cell differentiation is well established in other organoid and stem cell systems. Mechanical stretch and electrical stimulation enhance sarcomere formation and alignment in cardiomyocytes5,7, electrical pulses increase neuronal excitability and accelerate the development of synaptic networks6, and circadian metabolic entrainment through time-restricted feeding supports the functional maturation of β-like cells in pancreatic islets9. These approaches apply physical or rhythmic inputs that cells would normally encounter in vivo. Light, uniquely suited to the visual system, has been largely overlooked in this context. In animal models, light exposure during defined developmental windows influences photoreceptor orientation, synaptic transmission, and the maturation of outer retinal responses, whereas dark rearing delays or attenuates these processes14,15,41–43. For example, light exposure before eye opening shapes emerging photoreceptor and bipolar cell responses and promotes maturation of cone-driven signalling in the outer retina, while prolonged darkness impairs synaptic protein expression and functional transmission at the first visual synapse42,44. Together, these studies support the view that light acts not only as a sensory input but also as a developmental cue directly impacting photoreceptor maturation and outer retinal function. Building on these observations, our study extends this concept to the human system and demonstrates that rhythmic light stimulation, specifically in the gamma frequency range, can promote advanced features of photoreceptor maturation and enhance retinal function in human retinal organoids. Our findings suggest that 40 Hz flickering light is particularly effective due to its resonance with intrinsic neural oscillations. Gamma-band oscillations (~30–80 Hz) are fundamental features of neuronal activity in the central nervous system and play critical roles in sensory processing, attention, and circuit maturation18,45. In the developing visual system, spontaneous and evoked retinal waves help pattern connectivity before photoreceptors become light-responsive46,47. Flicker at 40 Hz, which entrains gamma oscillations, has been shown to enhance synaptic plasticity, improve memory, and reduce amyloid and tau pathology in mouse models of Alzheimer’s disease19,48,49. Our data suggest that similar entrainment may also apply to immature retinal circuits, potentially activating downstream signalling or activity-dependent transcriptional programmes necessary for photoreceptor maturation. Interestingly, in a complementary study, van Oosten et al. demonstrate that non-flickering, long-term photostimulation also promotes photoreceptor maturation in retinal organoids.50 While their study focuses on constant light exposure from day 70 of organoid differentiation process, the convergence of our results demonstrates the broader principle that light itself, whether rhythmic or continuous, acts as a potent maturation cue for human photoreceptors. These findings highlight the versatility of photostimulation strategies and suggest that multiple temporal patterns of light may be harnessed to support retinal differentiation, depending on the specific maturation stage or cellular context. Importantly, our temporal analyses indicate that photostimulation acts within a defined developmental window. When applied during photoreceptor maturation (D150–180), a brief period of flickering light is sufficient to induce structural changes that persist long after stimulation has ceased. In contrast, photostimulation at late stage produces only marginal effects on OS-like and gene expression, suggesting reduced plasticity at advanced maturation stages. Single-cell transcriptomic analysis revealed that photostimulation shifts photoreceptor populations toward more advanced maturation states by increasing differentiated rod and cone photoreceptors while reducing precursor populations. Importantly, this shift does not imply that photostimulation advances organoids to a later differentiation endpoint. Rather, photostimulated organoids remain within a foetal transcriptional stage, exhibiting more advanced maturation features relative to dark controls. Consistent with this interpretation, integration of organoid scRNA-seq data with adult human retina dataset showed that rods and cones from photostimulated organoids map closer to adult photoreceptors than those from dark-cultured organoids. Furthermore, transcriptomic signatures of phototransduction, OS structure, and ribbon synapses were enriched following photostimulation, indicating comprehensive effects on photoreceptor architecture and signalling. At this stage, the present study was designed to establish the existence, selectivity, and differentiation-related impact of patterned light stimulation, rather than to resolve the underlying molecular mechanisms. In this context, although our circadian entrainment data show that retinal organoids possess a functional molecular clock, the present study does not establish a causal link between circadian transcriptional regulation and photostimulation-induced photoreceptor maturation, nor does it allow discrimination between circadian-dependent and circadian-independent mechanisms. Dissecting these cell-type–specific pathways will therefore be an important focus of future work. RGC firing rates in human retinal organoids remain substantially lower than those reported for adult retina26. Low spontaneous and evoked activity is a well-recognized feature of current human retinal organoid models and likely reflects their immature and incomplete organization. In line with the molecular and structural changes, photostimulated organoids showed increased expression of synaptic markers and modest alterations in RGC spiking responses. The two electrophysiological paradigms used in this study serve complementary purposes. The 8-br-cGMP assay acts as a pathway-specific probe to assess photoreceptor-driven synaptic transmission independently of light, whereas wavelength-restricted light stimulation provides a more physiological readout of circuit activity. Future studies incorporating pharmacological manipulation of synaptic transmission will be important to further dissect pathway specificity and definitively distinguish photoreceptor-driven responses from potential direct effects on downstream neurons. The consistent increase in ON-pathway responsiveness across both paradigms suggests that photostimulation may enhance photoreceptor-to-downstream signalling despite overall low firing rates. Similar improvements in electrophysiological responses and cell maturation have been reported in brain organoids following electrical stimulation51 and in cardiomyocytes subjected to rhythmic pacing5, suggesting that temporally structured external cues can enhance functional properties of stem cell–derived tissues in vitro. One of the most promising therapeutic strategies for retinal degenerative diseases is the transplantation of photoreceptor precursors or mature photoreceptors derived from stem cells. However, despite decades of progress, the efficiency of photoreceptor transplantation remains limited. Challenges include poor cell survival, inadequate integration into the host outer nuclear layer, and limited functional synapse formation with downstream neurons3,4,52,53. A critical bottleneck is the maturity of the donor photoreceptors: cells that lack fully developed OS, proper polarity, or synaptic machinery are unlikely to survive transplantation or restore vision effectively. In addition to pre-conditioning donor cells in vitro, a key future direction will be to explore whether in vivo photostimulation, particularly with flickering light, could further enhance integration and functionality after transplantation. Our data indicate that 40 Hz flickering light promotes photoreceptor maturation and enhanced response to light in organoid systems. It is conceivable that a similar stimulus, applied post-transplantation, could entrain host and donor retinal activity, promote synaptogenesis, and strengthen functional coupling with bipolar or ganglion cells. Such an approach would parallel therapeutic sensory entrainment strategies used in other areas of the nervous system, including auditory and cortical rehabilitation. The feasibility of delivering flickering light non-invasively and in a frequency-specific manner makes it an attractive candidate for adjunctive therapy in cell-based retinal repair. Preclinical models will be essential to determine the safety, optimal parameters, and efficacy of such interventions. Our findings suggest that pre-conditioning photoreceptor cells with structured photostimulation, particularly 40 Hz flickering light, can promote key aspects of photoreceptor maturation, including OS-like formation, expression of phototransduction machinery, and functional connectivity. This strategy may substantially enhance the readiness of stem cell-derived photoreceptors for transplantation by bringing them closer to a mature, transplantation-competent state. Another promising avenue for improving photoreceptor maturation involves the modulation of culture media composition. Previous work by West and colleagues54 demonstrated that supplementation with specific lipids enhances OS formation and disk organization in differentiating photoreceptors. Given that OS morphogenesis depends on high membrane turnover and tightly regulated lipid dynamics, integrating lipid supplementation into photostimulation protocols could further synergize structural maturation. Combining photostimulation with metabolic support through targeted lipid additives may promote both the elongation and lamellar organization of OS membranes, yielding photoreceptors with more native-like architecture. Exploring such approaches will be essential for optimizing the functional competence of retinal organoids and transplant-ready cells. It is also important to highlight the technical simplicity and physiological relevance of the photostimulation paradigm. Unlike genetic modifications or pharmacological agents, patterned light is non-invasive, easily tuneable, and directly interpretable by the photoreceptor machinery. Importantly, we found that the presence of phenol red in the medium abolished the stimulatory effects of light, a finding consistent with reports that phenol red can act as a weak oestrogen mimic or ROS scavenger55,56. While the effects of 40 Hz stimulation were robust, several limitations should be acknowledged. First, our study focused on a fixed developmental window, capturing responses to patterned light over a 7-day period. Photoreceptor maturation in humans is a long-term, multi-month process, and our experimental design therefore involves only relatively short-term light modulation. Importantly, evidence for the persistence and accumulation of light-driven maturation is provided by a complementary study from van Oosten et al.,50 in which retinal organoids exposed to daily light cycles from day 70 to day 210 exhibited progressive increases in photoreceptor gene expression, protein abundance, and OS length. Together, these findings indicate that light-induced changes are not transient responses, but can contribute to sustained maturation. In this context, our study is best viewed as defining the early cellular and molecular responses to light stimulation, including transcriptional state shifts, increased OS-like formation, and enhanced electrophysiological responses downstream of photoreceptors. It is also likely that the responsiveness of retinal progenitors and immature photoreceptors to light changes across differentiation stages, an aspect that enables further investigation for both mechanistic insight and translational relevance. Second, all experiments were performed using human pluripotent stem cell–derived retinal organoids generated from healthy donor-derived cell lines. Whether photostimulation can modulate or partially compensate for disease-associated defects in genetic models, including retinal dystrophies or ciliopathies, remains an important question for future studies. Taken together, our findings demonstrate that gamma-frequency light flicker can be used as physiologically aligned stimulus to promote photoreceptor maturation within the constraints of the intrinsic differentiation programme. This approach improves cellular maturation and is associated with enhanced downstream retinal responses in retinal organoids, offering opportunities for disease modelling, drug screening, and future translational applications such as photoreceptor transplantation. Methods Pluripotent stem cells culture and retinal organoids generation We used human induced pluripotent stem cells (hiPSCs; cell line M857–60) and human embryonic stem cells (hESCs; cell line H9) for retinal organoid generation. Each cell line was maintained, characterised, and differentiated independently in distinct laboratories in which the experiments were conducted. hiPSCs were cultured in Essential 8 medium (Gibco) on Vitronectin-coated dishes and passaged approximately every 4–5 days. hESCs were maintained in mTeSR™1 medium (STEMCELL Technologies) on Matrigel-coated plates (Corning), with routine passaging at similar intervals. Retinal organoids were derived from both cell lines using previously established protocols30,54 (Fig. S1a). Photostimulation of retinal organoids Retinal organoids between day 150 and day 180 of the differentiation process were used, unless stated otherwise. One day prior to photostimulation, organoids were transferred into a black 96-well plate with a clear, round, ultra-low attachment bottom (#4515, Corning), each well containing fresh culture medium without phenol red and a single organoid. Photostimulation was performed using the Cell LighteR device21. The LED module (Fig. 1c) was placed inside a standard tissue culture incubator, and the 96-well plate was positioned directly on top of the module. A black lid was used to minimize ambient light interference. Photostimulation began the day after seeding. The light intensity used for the stimulation was 15 lx of white light. Retinal organoid dissociation for scRNA-seq Organoids were photostimulated under a 12-h light/12-h dark cycle for one week, with the light phase consisting of 40 Hz flickering illumination. On day 7, organoids were collected during the dark phase to avoid acute light-induced effects, and all subsequent steps were performed under light-protected conditions. For each condition, 30 organoids were processed for scRNA-seq. Organoid dissociation was carried out using the Neurosphere Dissociation Kit (#130-095-943, Miltenyi Biotec) according to the manufacturer’s instructions. Briefly, organoids were washed twice with cold HBSS (#14025092, Gibco) and transferred into Eppendorf tubes containing 960 μL of preheated Buffer X (each set of 15 organoids was dissociated in a separate tube). Then, 25 μL of Enzyme P, 10 μL of Buffer Y, and 5 μL of Enzyme A were added and mixed gently. The tubes were placed on a thermal mixer with a block (Eppendorf) for 10 min at 37 °C and 800 rpm. The organoids were dissociated using a fire-polished Pasteur pipette by pipetting up and down 10 times slowly. They were then incubated again under the same conditions for 5 min in the thermal mixer and pipetted up and down 10 times with the Pasteur pipette. After adding cold HBSS containing 5% FBS to the single-cell suspension, the mixture was passed through a 40 μm cell strainer placed on a 15 mL tube. Cells were counted, and the tubes were centrifuged at 300 × g for 10 min at room temperature. The supernatant was aspirated completely, and the cells were resuspended in cold HBSS containing 5% FBS at a concentration of 1000 cells/μL and kept on ice. (Total single cell number 1,500,000/sample) ScRNA-seq; library preparation and data processing The cell concentrations were adjusted to 1000 cells/µL. These suspensions were loaded onto the Chromium Controller following the Chromium Next GEM Single Cell 3’ Gene Expression v3.1 (Dual Index) protocol, Rev E (CG000315) provided by 10x Genomics. Library preparation was conducted using the same protocol. The quality and quantity of the prepared libraries were assessed using the High Sensitivity NGS Fragment Analysis Kit (Agilent Technologies) and the QuantiFluor dsDNA System (Promega). The libraries were then sequenced using the AVITI sequencing platform (Element Biosciences), employing the AVITI 2 × 75 Sequencing Kit Cloudbreak FS High Output (150 Gb). Sequencing reads were processed using Cell Ranger software version 8.0.1 (10x Genomics), with FASTQ files aligned to the human reference genome refdata-gex-GRCh38-2020-A, also provided by 10x Genomics, to generate gene expression matrices for downstream analyses. All data analysis and visualizations were conducted using R. We performed quality control checks in R for each sample and removed any cells with <1000 reads or 500 genes or >10% mitochondrial reads. Doublets were predicted using DoubletFinder61 and filtered from the data. The Seurat R package (version 5.1.0)62 was then used to process the data. The following R packages were used to further analyse and visualize the data: SeuratExtend63, scCustomize, clusterProfiler64, monocle365, and ggstatsplot66. RNA isolation For RNA extraction, a minimum of six organoids per sample were collected and thoroughly washed with phosphate-buffered saline (PBS). Organoids were then homogenized in 300 µL of RNA Blue Reagent (#R013, Top-Bio), a Trizol analogue, using a 1 mL insulin syringe to ensure complete cell lysis. After homogenization, total RNA was extracted using the Direct-zol™ RNA Microprep Kit (#R2062, Zymo Research) according to the manufacturer’s protocol. RT-qPCR analysis For gene expression analysis, RNA was reverse transcribed into complementary DNA (cDNA) using the High-Capacity cDNA Reverse Transcription Kit (#4368813, Applied Biosystems), following the manufacturer’s instructions. Reverse transcription was carried out in a 20 µL reaction volume with incubation steps at 16 °C for 30 min, 42 °C for 30 min, and 85 °C for 5 min. The resulting cDNA was amplified using the LightCycler 480 Real-Time PCR System (Roche) and PowerUp SYBR Green Master Mix (#P552, Top-Bio) in 20 µL reactions. PCR cycling conditions included an initial denaturation at 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 minute. Primer sequences are listed in Table S1. Gene expression levels were normalized to HPRT1 mRNA as the endogenous control. All experiments included at least three independent biological replicates, each performed in technical triplicate. Individual data points are shown in each graph. Statistical significance was assessed using a two-tailed Student’s t-test, with p-values indicated in the corresponding figure panels. Comparisons with p < 0.05 were considered statistically significant. Bulk RNA sequencing For bulk RNA-seq, RNA was extracted from eight organoids for each condition in two independent biological replicates. RNA integrity was assessed using the Fragment Analyser with the RNA Kit 15 nt (Agilent Technologies). For library preparation, 500 ng of total RNA was used as input for the QuantSeq 3′ mRNA-Seq FWD with UDI 12 nt Kit (v2) (Lexogen), in combination with the UMI Second Strand Synthesis Module for QuantSeq FWD. Library quality and quantity were evaluated using the QuantiFluor dsDNA System (Promega) and the High Sensitivity NGS Fragment Analysis Kit (Agilent Technologies). The final pooled libraries were sequenced on an Illumina NextSeq platform using the High Output Kit v2.5 (75 cycles) in single-end mode. Raw single-end FASTQ reads were quality-checked using FastQC. Adaptor removal and quality trimming were performed with Trimmomatic v0.39 using the following parameters: CROP:250, LEADING:3, TRAILING:3, SLIDINGWINDOW:4:5, and MINLEN:35. Trimmed reads were aligned to the human reference genome (Ensembl GRCh38 annotation) using STAR v2.7.3a with default settings, except for --outFilterMismatchNoverLmax 0.66 and --twopassMode Basic. Post-alignment quality control—including assessment of uniquely and multi-mapped reads, rRNA contamination, mapping distribution, read coverage, strand specificity, gene biotypes, and PCR duplication—was performed using RSeQC v4.0.0, Picard toolkit v2.25.6, and Qualimap v2.2.2. Downstream analysis of the NGS datasets was conducted in RStudio using the following R packages: DESeq2, biomaRt, Rsubread, and clusterProfiler. Immunohistochemistry Organoids were washed with PBS and fixed in 4% paraformaldehyde (PFA) solution in PBS for 30 min at room temperature. After fixation, the PFA was removed, and organoids were washed three times with PBS. They were then incubated overnight in 30% sucrose (Sigma-Aldrich) solution at 4 °C for cryoprotection. Subsequently, organoids were transferred into plastic disposable base molds, the remaining sucrose solution was carefully removed, and the molds were filled with Tissue-Tek® O.C.T. Compound medium (Sakura). The molds containing organoids were placed on dry ice for 15 min to solidify the medium and stored at −20 °C or processed immediately. Frozen blocks were sectioned into 7 µm-thick slices using a Leica CM1850 Cryostat (Leica). Slides were allowed to dry at room temperature for 20 min before further processing. For immunofluorescent staining, sections were washed with PBS and blocked for 1 h at room temperature in a humidified chamber using a blocking buffer containing 0.3% Triton X-100 and 5% normal goat serum in PBS. Primary antibodies are shown in Table S2 diluted in antibody diluent (AD, PBS with 0.3% Triton X-100 and 1% bovine serum albumin) were applied overnight at 4 °C. After washing with AD, secondary antibodies (Alexa Fluor 488 or 594-conjugated Goat anti-Mouse/Rabbit IgG, Invitrogen, 1:1000) were applied for 1 h at room temperature in a humidified chamber. Nuclei were stained with DAPI (1 µg/mL in PBS) for 5 min, and sections were mounted using Fluoromount Aqueous Mounting Medium (#F4680, Sigma Aldrich). OS-like length was measured using ImageJ software. In the first approach, OS length was defined as the distance from the edge of the DAPI signal to the distal extent of CONE ARRESTIN (ARR3) or ROM1 immunoreactivity. Because OS-like structures at this differentiation stage can vary in morphology, we additionally employed a complementary multi-marker strategy. In this second approach, the distal OS-like boundary was delineated using rod- or cone-specific markers (ROM1 or ARR3) in combination with ARL13B (connecting cilium marker) and/or CRALBP (outer limiting membrane marker). Together, these approaches enabled robust and consistent quantification of OS-like length across conditions. Cell-type composition analysis was done using QuPath (v0.6.0), marker-positive cells were quantified and normalized to the number of DAPI-positive nuclei within the corresponding retinal layers (ONL, INL, GCL). Quantification was performed on multiple organoids per condition using identical segmentation and thresholding parameters. Bright-field microscopy and OS-like length quantification Retinal organoids were imaged before and after the experimental period (7 days in darkness or 40 Hz photostimulation) using a Zeiss LSM 990 microscope equipped with a 20× objective. For each organoid, multiple regions were imaged to capture representative areas of the outer retina. OS-like structures were manually measured using ImageJ. Where applicable, measurements were performed in a paired manner on the same organoids before and after treatment. Scanning electron microscopy Retinal organoids (at least three organoids derived from independent differentiation batches per condition) were initially fixed in 3% glutaraldehyde (#G5882, Sigma-Aldrich) prepared in 0.1 M cacodylate buffer (pH 7.3; #C0250, Sigma-Aldrich). Following fixation, organoids were rinsed with fresh 0.1 M cacodylate buffer to remove residual fixative. Dehydration was performed through a graded ethanol series (30%, 50%, 70%, 80%, 90%, 96%, and absolute ethanol). After dehydration, organoids were dried using a critical point dryer (CPD 030, BAL-TEC Inc.). Dried organoids were then coated with a thin layer of gold using a Bal-Tec SCD 040 sputter coater (LEICA). The gold-coated samples were imaged using a scanning electron microscope (VEGA TS 5136 XM, Tescan Orsay Holding). Transmission electron microscopy Organoids (at least three organoids derived from independent differentiation batches per condition) were fixed in 3% glutaraldehyde (#G5882, Sigma-Aldrich) in 0.1 M cacodylate buffer (pH 7.3; #C0250, Sigma-Aldrich) for 24 h at 4 °C. After fixation, samples were washed with 0.1 M cacodylate buffer and post-fixed in 1% OsO₄ (#05500, Sigma-Aldrich) in the same buffer for 1.5 h at room temperature. Following post-fixation, samples were dehydrated through a graded ethanol series (50%, 70%, 96%, and absolute ethanol) and embedded in Durcupan™ ACM resin (Sigma-Aldrich). Polymerization was carried out over 3 days with a gradual temperature increase from 60 °C to 80 °C. Resin blocks were processed according to standard protocols for transmission electron microscopy, sections were prepared on Leica EM UC6 (Leica). Ultrathin sections were examined using a Morgagni 268D transmission electron microscope (Thermo Fisher Scientific) operating at 90 kV and equipped with a Veleta CCD camera (Olympus). Western blot analysis For each sample, at least six organoids were rinsed three times with PBS and lysed in a buffer containing 50 mM Tris-HCl (pH 6.8), 10% glycerol, and 1% SDS. Lysates were sonicated to ensure complete homogenization, and protein concentrations were determined using the DC Protein Assay (Bio-Rad). Subsequently, lysates were mixed with 0.01% bromophenol blue and 1% β-mercaptoethanol, then heated at 100 °C for 5 min to denature proteins. Equal amounts of protein were loaded onto SDS-polyacrylamide gels for electrophoresis, followed by transfer to PVDF membranes (Merck Millipore) pre-activated with methanol. Membranes were blocked for 1 h at room temperature in 5% skimmed milk prepared in TBST. Primary antibodies (Table S2), diluted in either 5% skimmed milk or 5% BSA (Sigma-Aldrich) in TBST, were applied overnight at 4 °C. The following day, membranes were washed three times for 15 min with TBST and incubated for 1 h at room temperature with the appropriate secondary antibodies (Table S2). Protein bands were visualized using the ECL Plus Reagent Kit (GE Healthcare). Uncropped blots are shown in Fig. S5. Electrophysiological recordings using 8-br-cGMP puffing Electrophysiological recordings were performed as described previously21,26. The organoids (four organoids per condition) were placed in artificial cerebrospinal fluid (aCSF), containing: 118 mM NaCl, 25 mM NaHCO3, 1 mM NaH2PO4, 3 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 10 mM glucose, and 0.5 mM L-Glutamine. The aCSF was equilibrated with 95% O2 and 5% CO2. The organoids were then longitudinally opened and placed with the presumed RGC layer facing down onto a 4096 channel multielectrode array (MEA). A translucent polyester membrane filter (Sterlitech Corporation) was used to flatten the organoids. The organoids were allowed to settle for a minimum of 1 h. Recordings were performed on the BioCam4096 MEA platform with BioChips 4096S+ (3Brain). The platform contained 4096 square microelectrodes in a 64 × 64 array configuration. Spike extraction was done using a quantile-based event detection method67. An automated spike sorting method for dense, large-scale recordings was used to sort single-unit spikes68. Statistical analysis and firing rate analysis were conducted using RStudio (R v4.4.2) and ggstatsplot library66. cGMP (8-Bromoguanosine 3′,5′-cyclic monophosphate, # B1381, Sigma-Aldrich) was puffed into the recording chamber at a final concentration of 100 μM. Activity was recorded continuously for 10 min, starting with the spontaneous activity recording (baseline, 5 min) followed by the cGMP puff recording (5 min). Electrophysiological recordings using light stimulation One day before the recordings, retinal organoids (two organoids per condition) were transferred into BrainPhys™ Neuronal Medium (STEMCELL Technologies), which supports higher neuronal excitability compared to aCSF used for 8-br-cGMP recordings. Prior to measurement, organoids were longitudinally opened and positioned with the presumptive RGC layer facing downward onto a 4096-channel multielectrode array (MEA; BioCam4096 platform with BioChips 4096S + , 3Brain), and organoids were allowed to settle for at least 1 h before recording. Recordings were conducted using the BioCam MEA platform. During the experiment, organoids on the MEA were stimulated with red (620–625 nm) and green (522–525 nm) light (intensity: 150 lx) to evoke RGCs responses. Blue light was avoided to prevent activation of intrinsically photosensitive RGCs (ipRGCs). Spike sorting was done using the same approach and settings, as documented for 8-br-cGMP experiments. Statistical analyses and firing rate quantifications were conducted in RStudio RStudio (R v4.4.2) and ggstatsplot library66. Experiments were performed in three independent runs; in each run, two organoids per condition were measured. Circadian rhythms measurement For BMAIL1 expression monitoring using luciferase reporters, hiPSCs were transduced using lentiviral particles containing the pABpuro-BluF vector (pABpuro-BluF was a kind gift from Steven Brown [plasmid #46824; Addgene; http://n2t.net/addgene:46824; RRID:Addgene_46824])69. Lentiviral particles were generated, as described in refs. 58,70,71. Upon transduction, hiPSCs BMAL1::Luc were cultured in the presence of Puromycin (0.5 μg/ml). For dexamethasone synchronization, the culture medium was replaced with pre-warmed medium containing 100 nM dexamethasone (#D4902, Merck). Organoids were incubated at 37 °C for 1 h to allow synchronization. Following incubation, the dexamethasone-containing medium was removed, and the cells were washed three times with warm PBS to eliminate any residual compound. The PBS was then replaced with fresh, pre-warmed culture medium. For luciferase activity measurement, we used a modified version of LuminoCell40. This version is capable to measure luciferase activity in individual wells of 96-well plate (Fig. 5a). Individual organoids were seeded into wells of a black 96-well plate with a clear, round, ultra-low attachment bottom (#4515, Corning) in the presence of 100 μM D-Luciferin sodium salt (#L6882; Merck). The LuminoCell was placed into a tissue incubator, the 96 well plate was positioned on the luminometer and covered by a lid to protect the cells from potential incoming light. Luciferase activity was measured using a built-in serial monitor in the Arduino IDE software on a laptop computer Lenovo SL500 with an Ubuntu operating system (20.04 version). The measured values were saved into a CSV file and processed in R studio. Ethics approval and consent to participate All methods were performed in accordance with relevant guidelines and regulations. The human pluripotent stem cell lines used in this study were obtained and used under appropriate approvals (MZP/2018/750/2676, MSMT-19838/2022-7). No new human or animal samples were collected specifically for this study, and separate ethics approval or consent to participate was not required. This study does not include any identifiable images or data from human participants. Therefore, consent for publication is not applicable. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary Transparent Peer Review file Source data Source Data Source Data