Reserpine helps protect light-sensing eye cells in retinal disease by fixing protein balance and cell structure problems

We designed an unbiased HTS assay to identify drug candidates that might improve photoreceptor development and/or survival. For a primary screen, we chose mouse retinal organoids because of a much shorter photoreceptor differentiation time and efficient generation of rudimentary outer segments using our modified HIPRO protocol (Chen et al., 2016a). We decided to use a karyotypically normal iPSC line derived from the Nrl-GFP containing rd16 mice (Figure 1—figure supplement 1A), which expresses green fluorescent protein (GFP) in all rods and carries an in-frame deletion in the myosin tail homology domain of CEP290 (Figure 1—figure supplement 2A). The rd16 mice are a model of photoreceptor degeneration observed in LCA10 patients (Chang et al., 2006). The iPSC lines derived from rd16 mice displayed similar morphology, proliferation rate, and stem cell properties as those derived from the wild-type (WT) (Figure 1—figure supplement 2B and C). WT and rd16 retinal organoids also showed comparable morphology at the early stages of differentiation (Figure 1A). However, at later stages, rod photoreceptors in rd16 organoids displayed aberrant morphology, with malformed or missing ciliary axoneme, connecting cilium and ciliary rootlets (Figure 1—figure supplement 2D). We then performed flow cytometry analyses to evaluate the differentiation of GFP+ rod photoreceptors in the rd16 organoid cultures, with the goal to identify quantifiable phenotypes for HTS. The rd16 organoids demonstrated as much as 50% lower organoid viability and almost 60% fewer GFP+ rod photoreceptors at day (D) 32 (Figure 1B), the end stage of differentiation in mouse retinal organoids.

The phenotypes in GFP+ rd16 rod photoreceptors permitted the development of an HTS screening platform to identify bioactive small-molecule candidates for augmenting rod cell differentiation and/or survival (Figure 1—figure supplement 2E). To avoid the impact of organoid variability on drug effects, we dissociated rd16 organoids into single cells at D25 and performed the treatment using two-dimensional cultures from D26 to D28. These cultures recapitulated the phenotypes detected in non-dissociated organoids, and rd16 cells displayed reduced viability and fewer GFP+ rod cells at D28 compared to the WT (Figure 1C).

Our drug discovery pipeline is illustrated in Figure 1D. In the primary screens, rd16 retinal organoid-derived cells were treated with over 6000 small molecules; of these, 114 compounds revealed higher GFP signal intensity, indicating a positive effect on rod photoreceptor survival. We then eliminated the compounds that showed toxicity or autofluorescence even in the absence of GFP expression (false positives). Fourteen small molecule compounds were then selected for further screening to identify a lead compound.

We then treated rd16 retinal organoids with the 14 selected small molecules in a secondary assay, which started at D22 and lasted for three days to maximize the drug effect without causing toxicity (Figure 2A). WT, untreated, and treated rd16 organoids were harvested at D29 to allow a sufficient period for the restoration of photoreceptors. The lead compound was selected based on the efficacy of the small molecule drugs to improve photoreceptor development and/or survival. Immunostaining of rhodopsin and S-opsin, markers of rod and cone photoreceptors respectively, was performed to quantify the drug effect (Figure 2B). Rhodopsin is highly expressed and polarized to the apical side of rod photoreceptors in the neural retina of WT organoids. In contrast, rd16 photoreceptors exhibited lower expression of rhodopsin, and the polarity was considerably diminished. Two of the compounds, B05 and to some extent B03, were able to enhance rhodopsin expression, with B05 improving the polarity of expression in the neural retina as well. We note that S-cone photoreceptors are difficult to be maintained in mouse organoid culture (DiStefano et al., 2018); nonetheless, compounds B01 and B05 were able to augment the expression and polarity of S-opsin. Three of the 5 compounds (B01, B03, B05) that showed no toxicity revealed a dramatic increase in rod and/or cone photoreceptors in D29 rd16 organoids (Figure 2C); of these, B05 was the most effective molecule and chosen as the lead compound. Interestingly, the other four molecules (B01-B04) were found to be derivatives of B05, further validating our results. B05 was identified as reserpine, a small-molecule drug that has been approved by the FDA for the treatment of hypertension and schizophrenia (Armitage et al., 1956; Kalliomäki and Kasanen, 1969), thereby holding a potential for drug repurposing (Figure 2D).

To further examine the efficacy of the lead compound reserpine, we utilized iPSC-derived retinal organoids from a previously-reported LCA10 family (Shimada et al., 2017), which comprised of a heterozygous unaffected mother (referred to as control henceforth) and two affected compound heterozygous children (LCA-1 and LCA-2; Figure 3—figure supplement 1A and B). Immunoblot analysis revealed a profound reduction of full-length CEP290 in patient organoids compared to controls (Figure 3—figure supplement 1C). Defects in rod development and outer segment biogenesis were evident in patient organoids from D120 onwards, as revealed by the near loss of connecting cilium marker ARL13B and absence or mislocalization of rod-specific protein rhodopsin (Figure 3—figure supplement 1D). In concordance with the observation of retained central cones in LCA10 patients, cone photoreceptors were only slightly compromised even at a late stage of differentiation in patient organoids.

Given that photoreceptor, outer segment biogenesis begins around D120 in organoids (Kaya et al., 2019) and aberrant phenotypes in patient organoids were detectable at this stage (Figure 3—figure supplement 1D), reserpine was added to patient organoids at D107 at a concentration of 10 µM, 20 µM, or 30 µM based on the EC₅₀ (half maximal effective concentration) in the primary screens (16.7 µM) (Figure 3A). We could detect improved connecting cilium immunostaining and rod development as early as D125 with the addition of 30 µM reserpine (Figure 3—figure supplement 2A). Even using the lowest tested dose (10 µM), we observed more polarized rhodopsin and connecting cilium at the apical side of photoreceptors in treated patient organoids, demonstrating a favorable effect of the drug (Figure 3—figure supplement 2A). Although variations in drug effects on organoids derived from the two patients could be observed, reserpine treatment significantly increased the fluorescence intensity of rhodopsin in retinal organoids derived from two patients (Figure 3B, C), suggesting an improvement of rod photoreceptor development. To avoid clonal variation, we also tested reserpine on another clone of each patient. Although statistically non-significant, these retinal organoids showed an obvious positive trend of rhodopsin staining upon reserpine treatment (Figure 3—figure supplement 2B, C). Such differences could be attributed to variability in response to dose and/or treatment windows of retinal organoids derived from various iPSC lines. We then performed transmission electron microscopy to uncover additional structural details of the primary cilium in photoreceptors of patient organoids after treatment. Reserpine increased the percentage of mother centrioles harboring ciliary vesicles in LCA-1 patient organoids (Figure 3D and Figure 3—figure supplement 3A), which were reportedly missing in a substantial fraction of LCA photoreceptors (Shimada et al., 2017). LCA-2 organoids demonstrated almost 50% more elongated ciliary axonemes after reserpine treatment though ciliary vesicles were hardly noticed. Quantification of the length of the ciliary axoneme indicated a significant increase in both patient organoids after treatment (Figure 3E). Notably, well-organized disc-like structures could be distinguished in treated samples (Figure 3—figure supplement 3B), suggesting a positive impact of reserpine treatment on photoreceptor primary cilium development.

We note that reserpine reportedly interferes with the sympathetic nervous system by inhibiting the transport of neurotransmitters into presynaptic vesicles (Bernstein et al., 2014); yet, even at 30 µM, no adverse effect was detected on the development of cone photoreceptors, ribbon synapses, presynaptic vesicles, bipolar cells, and Müller glia in treated patient organoids (Figure 3—figure supplement 4).

To elucidate the mechanism of action of reserpine and to gain insights into photoreceptor cell death in LCA10, we performed RNA-seq analyses of control and patient retinal organoids. The organoids were harvested right after the treatment with 30 µM reserpine at D150 (Figure 4—figure supplement 1A). Consistent with the moderate response of LCA-2 patient organoid to reserpine (Figure 3—figure supplement 2C), we did not identify sufficient significantly differentially expressed (DE) genes between untreated and treated groups of LCA-2 for downstream analyses (data not shown) and thereby focused only on LCA-1. Principal component analysis (PCA) revealed substantial alterations in patient organoid transcriptome compared to control and after reserpine treatment (Figure 4A). The largest principal component PC1 accounted for up to 36.2% of the total variation and is likely due to the drug treatment. PC2 explained another 21% of the total variation and seemed to be mainly contributed by the disease status. Notably, the reserpine-treated group was closer to the control compared to the untreated group in PC2. A total of 355 genes were significantly differentially expressed between untreated and treated organoids at thresholds of 5% FDR and twofold change (Figure 4—source data 1). Consistent with the PCA plot, heatmap analysis indicated that DE genes in treated patient organoids showed a relative transcriptomic shift from the untreated group toward the control ones (Figure 4B). However, compared to the controls, most of the downregulated genes (e.g. those involved in proteostasis) in untreated LCA-1 organoids had an even higher expression upon drug treatment. Similarly, the expression of genes upregulated in untreated samples was lowered even more than the controls by reserpine treatment. Curiously, treatment of LCA-1 organoids with 10 µM or 20 µM reserpine did not yield sufficient DE genes for further analyses (data not shown).

We first analyzed the expression level of genes specific to different retinal cell types and those involved in phototransduction and photoreceptor outer segment structure/function. Reserpine treatment modulated the expression of marker genes for inner retina cell types, including Müller glia (Figure 4—figure supplement 1B); however, surprisingly, several rod and cone photoreceptor genes (e.g. OPN1SW and GRK7) showed lower expression (Figure 4C, Figure 4—figure supplement 1C, Figure 4—source data 1), with CEP290-associated cilia genes (Rachel et al., 2012) exhibiting varying changes in expression (Figure 4—figure supplement 1D).

As reserpine did not seem to directly regulate the expression of key components involved in disease pathology, we mapped the overexpressed genes to KEGG pathways and created an enrichment network. Using a cutoff of three genes with minimum overlap and 5% impact, the annotation network was plotted to visualize leading ontology terms (Figure 4D). Clusters of metabolism, proteostasis, and immune pathways, together with cell survival-related processes, were apparent in the network. To better understand functional connectivity among DE genes, we performed a Random-walktrap analysis on their protein-protein interactions network to identify co-functioning over- and under- expressing genes and identified three prominent modules (Figure 4—figure supplement 1E). Functional module 1 was comprised of extracellular matrix (ECM) and ECM-receptor interaction, and advanced glycation end products (AGE)-Receptor for AGE (RAGE) signaling pathway genes. Müller glia-specific genes MMPL14, TIMP1, and VIM were included in this module. Functional module 2 consisted of inflammation- and proteasome-related genes, which are consistent with the reported role of reserpine as an autophagy modulator (Lee et al., 2015). Notably, functional module 3 contained critical cell survival factors such as p53 signaling and cellular senescence-associated genes. We further investigated the trend of the p53 network to characterize its role in response to drug treatment. Expression of TP53, the gene coding p53, was found to increase and match the level of control organoids (Figure 4—figure supplement 1F). In addition, a widespread modulation of each component of the p53 signaling network was evident upon reserpine treatment (Figure 4—figure supplement 1G). Downstream targets of p53, including metabolic modulator TSC and mTOR complex genes, responded to reserpine and returned to the level of control organoids (Figure 4—figure supplement 1H). Two components of the mTORC1 signaling pathway, RHEB, and LAMTOR1, were over-expressed after the drug treatment (Figure 4—source data 1). As mTORC1 is a key regulator of cellular metabolism, we observed significantly elevated expression of SLC2A1 (GLUT) which is the primary glucose transporter in neurons (Figure 4—figure supplement 1I).

Given the reported actions of reserpine in neuronal cells (Lee et al., 2015), consistent with activation of mTORC1 activation (Kim and Guan, 2015) and ‘proteasome’ subunit-encoding genes (Figure 4D), we performed a focused gene set enrichment analysis to test the impact on the proteostasis network (PN). A PN geneset was manually curated from KEGG and Reactome using keywords reviewed from previous publications (Jayaraj et al., 2020; Klaips et al., 2018). We detected a significant positive net enrichment of PN in patient organoids after reserpine treatment (NES = 1.37, adj. p-value = 1.6e−11; Figure 4E), as measured by log₂ fold change in expression of all PN genes (Figure 4F). We also identified global overexpression of proteasomal subunits, some of which were notably induced with fold change >2 (Figure 4G). In concordance, expression of the key regulators of proteostasis and members of the p53 network, p62 (SQSTM1) and NRF2 (NFE2L2), were significantly augmented in reserpine-treated patient organoids (Figure 4H, I; Figure 4—source data 1).

To experimentally examine the role of PN in the survival of LCA10 photoreceptors, we supplemented organoid cultures with various autophagy inhibitors (MRT68921, Lys05, chloroquine, hydroxychloroquine, ROC-325) that target different steps of the autophagy pathway using the same timeline as reserpine treatment (Figure 5—figure supplement 1A, B). MRT68921 and Lys05 inhibit the initiation of autophagy (Galluzzi et al., 2017) and were highly toxic even at one-fourth of the reported EC₅₀. Hydroxychloroquine and a more specific and efficient small molecule autophagy inhibitor ROC-325 enhanced the number of RHO+ cells in patient organoids with more polarized localization of rhodopsin to the apical side (Figure 5—figure supplement 1C). Quantification of the fluorescence intensity of rhodopsin immunostaining demonstrated a consistent trend and hydroxychloroquine- and ROC-325-treated organoids had a significantly higher expression of rhodopsin (Figure 5—figure supplement 1D). Although statistically non-significant, chloroquine-treated organoids also showed a positive trend in rhodopsin expression (Figure 5—figure supplement 1D), providing further evidence in support of the autophagy pathway as a target for designing therapies.

To further determine the role of autophagy, we evaluated the level of cargo receptor sequestosome 1 (SQTM1) or p62, which recognizes cellular components and helps in the formation of autophagosomes for proteolysis (Mizushima, 2007; Figure 5A). We noted that LCA-1 patient organoids had a significantly lower level of p62 compared to the control, and reserpine treatment significantly elevated p62 levels by more than twofold (Figure 5B), consistent with previous studies (Lee et al., 2015; Bresciani et al., 2018). Significantly higher levels of LC3-II in patient organoids also indicated accumulation of autophagosomes (defects in autophagy), and reserpine treatment revealed a trend of decreasing LC3-II levels (though not statistically significant) in patient organoids (Figure 5B). The comparable ratio of LC3-II/LC3-I indicated a similar rate of autophagosome formation between control and patient organoids, suggesting that increased LC3-II levels in patient organoids could be due to overexpression of autophagy pathway components. Consistent with the reported function of reserpine as an autophagy inhibitor, the treated patient organoids had a significantly lower LC3-II/LC3-I ratio, although we did not detect a significant difference between the control and patient organoids (Figure 5B).

We then applied Bafilomycin A1 (BafA1), an inhibitor of autophagosome-lysosome fusion, to patient organoid cultures using the same timeline as reserpine treatment (Figure 5—figure supplement 2A). A short 6-hr treatment with BafA1 did not alter the autophagy pathway in the control, as shown by comparable levels of p62 and LC3-II as well as LC3-II/LC3-I ratio (Figure 5—figure supplement 2B); however, patient organoids demonstrated a significant (up to 70%) increase of LC3-II and LC3-II/LC3-I ratio, suggesting an elevated autophagic flux in patient organoids, consistent with the rescue by reserpine treatment.

The ubiquitin-proteasome system (UPS) and autophagy are the two key pathways in proteostasis and are reported to interact through p62 (Kumar et al., 2022; Liu et al., 2016; Figure 5C). We, therefore, investigated the response of UPS in patient organoids upon reserpine treatment. Untreated patient organoids revealed a high level of 20S proteasome, which could be barely detected in the control (Figure 5D). Nonetheless, both the control and untreated patient organoids showed comparable total catalytic activity and reserpine significantly elevated the proteasome activity (Figure 5E). These results suggest that an increase in 20S expression is likely a compensatory mechanism for compromised proteasome activity in patient organoids and that reserpine facilitates the clearance of accumulated cellular components and/or autophagosomes (Figure 5B).

We were intrigued by the reported links of autophagy to primary cilium biogenesis (Pampliega et al., 2013; Tang et al., 2013; Yamamoto and Mizushima, 2021) and, therefore, looked at the expression of key regulators involved in ciliary transport, cilium assembly, and disassembly (Figure 5F). We identified higher expression of OFD1 (oral-facial-digital syndrome 1), which when eliminated by autophagy is shown to promote ciliogenesis (Tang et al., 2013), in patient organoids compared to the control even though autophagy activity was higher in the latter (Figure 5F). This apparent ambiguity could be due to variations in autophagic adapter machinery for cargo identification. Nonetheless, reserpine treatment reduced OFD1 levels in patient organoids and should facilitate the initiation of cilia biogenesis. Another key regulator, histone deacetylase 6 (HDAC6), which deacetylates microtubules and destabilizes the primary cilium for disassembly (Simões-Pires et al., 2013; Pugacheva et al., 2007), was significantly elevated in patient organoids compared to the control. Dramatic reduction of HDAC6 by reserpine (Figure 5F) would also have a favorable impact on cilia biogenesis. Consistent with this hypothesis, the addition of a selective HDAC6 inhibitor Tubastatin A to patient organoids improved rod photoreceptor development as shown by a higher number of RHO + cells (Figure 5—figure supplement 2C). Tubastatin A treatment enhanced the polarity of not only rhodopsin but also S-opsin, likely due to increased stability of intracellular microtubules and improved intracellular trafficking. Quantification of the fluorescence intensity of rhodopsin indicated a significant increase of rhodopsin expression in tubastatin A-treated organoids (Figure 5—figure supplement 2D), suggesting a favorable effect on rod photoreceptor development probably through the inhibition of HDAC6.

To validate the effect of reserpine, we assessed the autophagy machineries and relevant ciliary markers in a less responsive LCA-2 clone F upon reserpine treatment. Consistent with LCA-1, reserpine treatment significantly increased the p62 level but did not alter LC3-II/LC3-I ratio in LCA-2 organoids (Figure 5—figure supplement 3A, B). Although statistically non-significant, the level of autophagosome showed a decreasing trend in organoids treated with reserpine (Figure 5—figure supplement 3A, B). As retinal organoids contain abundant retinal cell types besides photoreceptors, we performed immunostaining of p62 on control, untreated, and treated LCA1 and LCA2 organoids. As shown in Figure 5—figure supplement 3C, p62 staining could barely be observed in both patient organoids compared to the control. Reserpine treatment dramatically increased p62 staining in both photoreceptors and other retinal cell types. We also quantified the level of OFD1 and HDAC6, the two ciliary regulators that mediated the favorable effect of reserpine on outer segment biogenesis (Figure 5—figure supplement 3D). We observed a significant decrease of HDAC6 in LCA2 organoids upon reserpine treatment. Although statistically non-significant, a lower expression of OFD1 was shown in reserpine-treated LCA2 organoids. Therefore, the action mechanisms of reserpine are independent of the responsiveness of patient organoids.

We then performed intravitreal injection of 40 µM reserpine into rd16 mouse eyes at postnatal day (P)4, and the retinas were harvested at P21 (Figure 6A). Reserpine was able to maintain photoreceptor survival, with a significantly thicker outer nuclear layer (ONL) in both the central and peripheral retina (Figure 6B). As a small molecule drug, reserpine is highly diffusive, and we observed the effect of the drug in the contralateral eyes of the treated mice (Figure 6—figure supplement 1). Indeed, bilateral therapeutic effects have been reported following unilateral injection in clinical treatments (Michalska-Małecka et al., 2016; Rouvas et al., 2009; Calvo et al., 2016; Zlotcavitch et al., 2015; Rotsos et al., 2014). Reserpine-treated rd16 mice showed a trend of increase in scotopic a-wave and a significant increase in scotopic b-wave (Figure 6C and D), suggesting functional recovery of rod photoreceptors upon reserpine treatment. No statistically significant difference was found in photopic a-wave or b-wave (Figure 6D). Notably, no systemic toxicity was observed in the treated mice. Injection of reserpine into WT mice did not lead to structural or functional alterations (Figure 6—figure supplement 2A and B), suggesting reserpine to be a safe treatment option.

Consistent with the effects observed on human retinal organoids, reserpine treatment increased p62 levels (Figure 7A and B), dramatically reduced the 20S proteasome (Figure 7C), and significantly enhanced the proteasome activity (Figure 7D). Further analyses confirmed an improvement in the structure of photoreceptor outer segments of the treated rd16 retina. Reserpine partially restored outer segment axonemes that were largely missing in untreated mouse retina (Figure 7E). In addition, substantial ciliary rootlets were conspicuous in the inner segment (shown by GFP) and extended into the ONL. Phototransduction proteins including rhodopsin and Pde6β were transported to the outer segment (Figure 6—figure supplement 3A). In addition, the treatment with 40 µM reserpine did not alter the morphology of other retinal cell types or structures including bipolar neurons, amacrine cells, and Müller glia (Figure 6—figure supplement 3B and C). Notably, the retinal stress marker Gfap was significantly increased in untreated rd16 retina as compared to the WT and was significantly reduced upon reserpine treatment (Figure 6—figure supplement 3D and E), suggesting a favorable effect of reserpine to retinal homeostasis.

B6.Cg-Cep290rd16/Boc mice (Strain #: 012283; RRID: IMSR_JAX:012283↗) were obtained from the Jackson Laboratory and crossed to Nrlp-eGFP mice (Akimoto et al., 2006) to generate Nrlp-EGFP; Cep290^rd16/rd16 mice (referred to as rd16). The absence of the rd8 mutation in the colony was assessed by PCR. All animal procedures were approved by the Animal Care and Use committee of the National Eye Institutes (Animal study protocol NEI-650) and adhered to ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Mice were housed in an atmosphere controlled-environment (temperature: 22°C ± 2°C, humidity: 30–70%), under a 12 hr dark/12 hr light cycle, and supplied with food and water ad libitum. Food, water, and nesting material were changed weekly.

The mouse Nrl-GFP WT and rd16 iPSC lines were obtained by infection of the E14.5 Nrlp-eGFP and rd16 mouse embryonic fibroblasts with Dox-inducible lentiviral vectors carrying Pou5f1, Sox2, Klf4, and Myc genes individually, as previously described (Chen et al., 2016a; Mookherjee et al., 2018). To obtain human iPSC lines, fibroblasts of the familial control, two clones of LCA10 patient 1 (annotated as LCA-1 and LCA-1-clone C), and two clones of LCA10 patient 2 (annotated as LCA-2 and LCA-2-clone D) were isolated from skin biopsies and reprogrammed using integration-free Sendai virus by the iPSC Core Facility at National Heart, Lung, and Blood Institute of the National Institutes of Health using the established protocol (Shimada et al., 2017; Beers et al., 2015). All mouse and human iPSC lines were karyotypically normal except the rd16 line 2, which had one missing sex chromosome and thus, was not used for subsequent high-throughput screening and secondary validation assays (Figure 1—figure supplement 1). All cultures were tested for mycoplasma contamination by real-time PCR periodically.

The WT and rd16 iPSC lines were maintained as previously described (Chen et al., 2020). Briefly, the iPSC lines were maintained on feeder cells (Millipore) in a maintenance medium constituted by Knockout DMEM (ThermoFisher Scientific), 1 x MEM non-essential amino acids (NEAA) (Sigma), 1 x GlutaMAX (ThermoFisher Scientific), 1 x Penicillin-Streptomycin (PS) (ThermoFisher Scientific), 2000 U/ml leukemia inhibitory factor (LIF) (Millipore), and 15% ES cell-qualified fetal bovine serum (FBS) (ThermoFisher Scientific) at 37 °C, 5% CO₂. Full media change was performed every day, with 55 µM β-Mercaptoethanol (2-ME) (ThermoFisher Scientific) freshly added. Cells were passaged using TrypLE Express (ThermoFisher Scientific) every two days.

Human iPSC lines were maintained in Essential 8 (E8) (ThermoFisher Scientific) on growth factor-reduced (GFR) Matrigel (Corning)-coated plates, with media fully changed daily. Cells were maintained at 37 °C, 5% O₂, and 5% CO₂ and passaged at 60–80% confluency using the EDTA-based dissociation method (Kelley et al., 2020).

The modified HIPRO protocol was used to differentiate mouse iPSCs into retinal organoids (Chen et al., 2016a; Chen et al., 2020). At differentiation day (D)0, iPSCs were plated in low adhesion U-shaped 96-well plate (Wako) at a density of 3000–5000 cells per well in 100 µl retinal differentiation medium consisting of GMEM (ThermoFisher Scientific), 1 x NEAA, 1 x sodium pyruvate (Sigma) and 1.5%(v/v) knockout serum replacement (KSR) (ThermoFisher Scientific). At D1, 240 µl GFR-Matrigel (>9.5 mg/ml) was diluted in 1.8 ml retinal differentiation medium and 20 µl diluent was aliquoted to each well of the 96-well plate. Retinal organoids from one 96-well plate were transferred to a 100 mm poly(2-hydroxyethyl methacrylate) (Sigma)-coated petri dish with 12 ml DMEM/F12 with GlutaMAX, 1x N2 supplement, and 1 x PS. The media were replaced by DMEM/F12 with GlutaMAX, 1 x PS, 1x N2 supplement, 1 mM taurine (Sigma), 500 nM 9-cis retinal (Sigma), and 100 ng/ml insulin-like growth factor 1 (IGF1) (ThermoFisher Scientific) at D10, and half-media change was performed every other day, with 55 µM 2-ME freshly added to the media. From D26 onward, 1 x NEAA, 1 x B27 supplement without Vitamin A (ThermoFisher Scientific), and 2%(v/v) FBS (ThermoFisher Scientific) were added to the culture. Half-media exchanges were performed every two days, with 55 µM 2-ME freshly added to the media. The cultures were incubated in 5% O₂ from D0 to D10 and in 20% O₂ from D10 onwards.

Human retinal organoid differentiation was performed as previously described (Regent et al., 2020). Briefly, small clumps dissociated from iPSCs in one well of a six-well plate were resuspended in E8 medium supplemented with 10 µM Y-27632 (Tocris) and transferred into one 100 mm polyHEMA-coated petri dish for embryoid body (EB) formation. Media were supplied with neural induction media (NIM) (DMEM/F-12 (1:1)) (ThermoFisher Scientific), 1x N2 supplement, 1 x NEAA, 2 µg/ml heparin (Sigma) at D1 and D2 at a ratio of 3:1 and 1:1, respectively, and fully switched to NIM at D3. D7 EBs from one 100 mm petri dish were plated onto one GFR Matrigel-coated 60 mm dish and cultured in NIM, with media changed every 2–3 days. In the application requiring a large-scale production of retinal organoids, nicotinamide was added to the culture to reach a final concentration of 5 mM from D0 to D8 (Regent et al., 2022). NIM was replaced by a 3:1 retinal induction medium (RIM) consisting of DMEM/F-12 supplemented with 1 x B27 without Vitamin A (ThermoFisher Scientific), 1% antibiotic-antimycotic solution (ThermoFisher Scientific), 1% GlutaMAX and 1 X NEAA at D16 and media change was performed every day until D28, on which the adherent cells were scraped off into small clumps (<5 mm²) and split into two polyHEMA-coated 100 mm Petri dishes. The floating clumps were cultured in RIM supplemented with IGF1 (ThermoFisher Scientific) and 1 mM taurine (Sigma), with 55 µM 2-ME freshly added. From D38 onward, 10% fetal bovine serum was added to RIM supplemented with 20 ng/ml IGF-1, and 1 mM taurine. 1 µM 9-cis retinal was freshly supplemented to the cultures during media change from D63 to D91. From D91 till the end of differentiation, the concentration of 9-cis retinal was reduced to 0.5 µM and B27 without Vitamin A was replaced by N2. Once scraped, media were half changed every 2–3 days, with IGF1, taurine and 9-cis retinal freshly added to the media under dim light environment.

Mouse retinal organoids were transferred into a 15 ml centrifuge tube using wide-bored transfer pipets and washed one time with 10 ml 1 x PBS (ThermoFisher Scientific). Prewarmed 1 ml 0.25% trypsin-EDTA was then added, and the tube was incubated at 37 °C for 10 min, with pipetting up and down for 10 times using P1000 pipetman at 5- and 10 min incubation. 10 ml retinal maturation media (DMEM/F12 with GlutaMAX, 1 x PS, 1x N2 supplement, 1 mM taurine, 500 nM 9-cis retinal, 1 x NEAA, 1 x B27 supplement without Vitamin A, 2%(v/v) FBS) was added to the tube. The tube was inverted several times before centrifugation at 200 g for 5 min. After removal of the supernatant, the cell pellet was resuspended in 1 ml retinal maturation media and filtered through a 40 µm cell strainer (BD Bioscience) before proceeding to subsequent analyses.

We used three different small molecule libraries for our screening efforts. The library of 1280 pharmacologically active compounds (LOPAC 1280) consists of a collection of small molecules with characterized biological activities commonly used to test and validate screening assays. The library of U.S. Food and Drug Administration (FDA)–2800 approved drugs was set up internally. The compounds were first dissolved in 100% DMSO to generate a stock concentration with a final concentration of 10 mM and subsequently diluted into seven concentrations and dispensed into 1536-well plates. The NCATS Mechanism Interrogation Plate (MIPE) 5.0 library contains a collection of 1912 compounds (approved for clinical trials).

To optimize the assay for HTS, D25 WT, and rd16 mouse retinal organoids were dissociated into single-cell suspension and plated in 5 µl retinal maturation medium II at a concentration of 3000, 4000, and 5000 cells per well in a 1536-well plate. After culturing in a 37 °C humidified incubator overnight, 5000 cells/well yielded an optimal plating density for both WT and rd16 cultures and were selected for subsequent assays.

In the screening experiments for drug candidates, dissociated D25 WT-positive control and rd16 mouse retinal organoids were filtered through a 35 µm cell strainer and dispensed into a 1536-well plate at a density of 5000 cells/well in 5 µl freshly prepared retinal maturation medium II using the Multidrop Combi Dispenser (Thermo Scientific). Cells were incubated in a 37 °C humidified incubator overnight for cell recovery and attachment to the plates. Compounds in the 1536-well drug source plates were added to the 1536-well assay plates at a volume of 23 μl/well using an NX-TR pintool station (WAKO Scientific Solutions). Cells were treated with compounds for 48 hr, followed by the addition of 0.5 µl quencher, and GFP and DAPI signal intensities were quantified using the acumen Cellista (TTP, Labtech). The positive hits with a significant increase of GFP compared to the DMSO-treated cells were retested using dissociated retinal organoids from iPSCs without a GFP marker to remove autofluorescent false positives. The remaining drug candidates were further tested using a full 11-concentration setting (1:3 serial dilutions starting at 10 mM) in triplicate plates to prioritize the hits based on efficacy.

The primary and confirmatory screening data were analyzed using software developed internally at the NIH Chemical Genomics Center (NCGC) (Wang et al., 2010). The plate data was processed as follows: We first ran quality control on the plate data by visual inspection, masking wells showing erroneous signals, e.g., localized groupings of wells exhibiting enhanced or inhibited signals. We then performed intra-plate normalization as follows:[1]Efficacy=100×GFPRD16+C-GFP(RD16+DMSO)GFP(WT+DMSO)-GFP(RD16+DMSO)[2]Toxicity=-100×DAPI(RD16+C)-DAPI(RD16+DMSO)DAPI(RD16+DMSO)[3]Fluorescence=100×GFP(Parental+C)-GFP(Parental+DMSO)GFPWT+DMSO-GFP(Parental+DMSO)

Where:

When the data is normalized in this fashion, efficacy, toxicity, and fluorescence range from 0 to 100%. With this normalization completed, we compute dose-response curves as follows: ⁴ A ( ) = + C i A 0 ( − ) A ∞ A 0 [ + ( ] 1 10 ( ( ) − ( ) ) L o g A c L o g 50 c i ) n

Where C_i = the i’th concentration, A(C_i) = activity (efficacy, toxicity, or fluorescence) at concentration i, A₀ = activity at zero concentration, A_∞ = activity at infinite concentration, and EC₅₀ = the concentration giving a response half-way between the fitted top (100%) and bottom (0) of the curve.

Compounds are desired that have high efficacy (EC_50efficacy = small, A_{∞, efficacy =} high), low toxicity, (EC_50toxicity = large), and low fluorescence (EC_{50fluorescence} = large).[5]EC50toxicity/EC50efficacy>>1[6]EC50flurorescence/EC50efficacy>>1[7]A∞,efficacy=high

In addition, the dose-response curves for each measured quantity must be well behaved, i.e., (i) curves exhibit values near zero at low concentration, (ii) values increase with increasing concentration, (ii) curves show a well-defined inflection point at the EC₅₀ concentration, and (iv) curves show a well-defined plateau at high concentration (Wang et al., 2010).

At least three organoids in each batch were homogenized in 100 μl of RIPA buffer (Sigma) supplemented with 1 x protease inhibitor (Roche) and 1 x phosphatase inhibitor (Roche). The lysate was agitated at 4 °C for half an hour, before centrifugation at 12,000 g for 10 min at 4 °C. The supernatant was either stored at –80 °C until use or quantified by Pierce bicinchoninic acid (BCA) protein assay (ThermoFisher Scientific). Approximately 20 μg protein was diluted 4:1 in reducing 4 x Laemmli buffer (Biorad) and boiled for 10 min. The samples were separated at 150–180 V for 1 hr on 4–15% precast polyacrylamide gel (Biorad) and transferred to polyvinylidene fluoride (PVDF) membranes using a TransBlot Turbo Transfer System (Biorad). After blocking in 5% milk or 5% bovine serum albumin (BSA) for 1 hr at room temperature, the blots were incubated in antibody cocktails (Key Resources Table) overnight in 1% milk or BSA in 1 X TBS-T at 4 °C overnight with gentle agitation. Membranes were subsequently washed in 1 X TBS-T for three times, 10 min each, and incubated in 1 X TBS-T with horseradish peroxidase-conjugated secondary antibodies (1:5000) for 1 hr at room temperature, followed by another three 10 min wash. Before imaging, the membranes were exposed to SuperSignal West Pico enhanced chemiluminescence (ECL) solution (ThermoFisher Scientific) for 5 min, and chemiluminescence was captured using a Bio-Rad ChemiDoc touch (Bio-Rad).

The chymotrypsin-like protease activity associated with the proteasome complex in organoids or mouse retina was quantified using the Proteasome 20 S Activity Assay Kit (Sigma) following the manufacturer’s protocol. In short, at least three retinal organoids or one mouse retina were homogenized in 55 µl PBS mixed with 55 µl reconstituted Proteasome Assay Loading Solution and incubated on ice for 30 min. After centrifuging at 10,000 g for 10 min at 4 °C, 10 µl supernatant was taken for BCA assay to determine the protein amount and the remaining 100 µl was transferred to one well of black/clear 96-well plate for incubation at 37 °C. Protease activity of individual samples was measured by the fluorescence intensity (λex = 480–500 nm/λem = 520–530 nm) normalized to the protein amount.

Retinal organoids were processed for transmission electron microscopy (TEM) analysis as previously described (Shimada et al., 2017). Briefly, the organoids were fixed in 4% formaldehyde and 2% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4 (Tousimis) for 2 hr at room temperature, followed by three washes in cacodylate buffer before further fixation in osmium tetroxide (1% v/v in 0.1 M cacodylate buffer; Electron Microscopy Sciences) for 1 hr at room temperature. The organoids were then washed in the same buffer for three times, followed by one wash in acetate buffer (0.1 M, pH 4.2), and en-bloc staining in uranyl acetate (0.5% w/v; Electron Microscopy Sciences) in acetate buffer for 1 hr at room temperature. The samples were dehydrated in a series of ethanol solutions (35%, 50%, 75%, 95%, and 100%) and then by propylene oxide. The samples were subsequently infiltrated in a mixture of propylene oxide and epoxy resin (1:1, v/v) overnight, embedded in a flat mold with pure epoxy resin, and cured at 55 °C for 48 hr. 70–80 nm sections were made with an ultramicrotome (UC 7) and diamond knife (Diatome), attached on a 200-mesh copper grid, and counter-stained in the aqueous solution of uranyl acetate (0.5% w/v) and then lead citrate solutions. The thin sections were stabilized by carbon evaporation before the EM examination. The digital images were taken using a digital camera equipped with an electron microscope (H7650) (AMT).

After dissociating retinal organoids, the cells were resuspended in DPBS (ThermoFisher Scientific) containing 1 mM EDTA (Millipore) and filtered through a 40 μm cell strainer. 4',6-diamidino-2-phenylindole (DAPI) (ThermoFisher Scientific) was added to the samples before being analyzed by FACSAriaII (BD Bioscience). Cell viability was evaluated by integrity (gated by DAPI), size (gated by forward scatter, FSC), and granularity (gated by side scatter, SSC). WT retinal organoids without a GFP marker on the same differentiation day were used to set the gating for GFP + cells.

Total RNA was purified from homogenized retinal organoids using TriPure isolation reagent (Roche). Quality of isolated RNAs was assessed using Bioanalyzer RNA 6000 nano assays (Agilent) and high-quality total RNA (RNA integrity number ≥7.5) was used for the construction of the mRNA sequencing library. 100 ng total RNA was used to construct the strand-specific libraries using TruSeq RNA Sample Prep Kit-v2 (Illumina).

RNA sequencing was performed as described (Kim et al., 2016). Paired-end 125 bp reads were generated using Illumina sequencing. Reads were quality checked and mapped to the reference transcriptome using Kallisto. Alignments were imported into R using tximport for downstream analyses. The edgeR and limma pipeline were employed for differential expression analysis, while pathway annotation was performed using ClueGO (Bindea et al., 2009). Protein interaction data were obtained from STRING (v11.5) with evidence cut off of 700, via the R package STRINGdb (Szklarczyk et al., 2021). Network analysis was performed using the igraph package [https://igraph.org/↗] and genes were mapped to pathways using gProfileR (Raudvere et al., 2019). Proteostasis network genes were manually collected by pooling KEGG and Reactome pathways with keywords from Jayaraj et al., 2020 and (Klaips et al., 2018), and the proteosome map was adapted from KEGG (hsa03050). Gene set enrichment analysis was done using the R package fgsea. All other plotting and analyses, unless otherwise mentioned, were performed using tidyverse and base R packages.

Mouse and human organoids were fixed in 4% PFA (Electron Microscopy Sciences) for 1 hr at room temperature, washed one time, and cryoprotected in 15% sucrose for at least 2 hr at room temperature, followed by 30% sucrose at 4 °C overnight. The next day, organoids were embedded in Shandon M-1 Embedding Matrix (ThermoFisher Scientific). The blocks were sectioned at 10 μm thickness and incubated at room temperature for at least 1 hr, before immunostaining or storage at −80 °C. After incubating in blocking solution (5% donkey serum in PBS) for 1 hr at room temperature, the sections were supplied with diluted primary antibodies (Key Resources Table) at 4 °C overnight. After three 10 min washes in PBS, Species-specific secondary antibodies conjugated with Alexa Fluor 488, 568, or 637, together with DAPI, were diluted in blocking solution (1:500; ThermoFisher Scientific) and incubated with the sections for 1 hr at room temperature. After washing in PBS for three times, 10 min each, the sections were mounted for imaging.

For mouse retinal sections, rd16 mice of both sexes were injected at postnatal (P) day 4 and recovered at P21 after euthanasia using a CO₂ atmosphere. Eyes were enucleated before being pierced in the center of the cornea using a 26-gauge needle. Eyecups were then incubated for 15 min in 4% paraformaldehyde (PFA). The cornea and lens were then dissected, and the eyecups were placed in 4% PFA for 15 min at room temperature (RT) before proceeding to cryoprotection in 20 and 30% sucrose-PBS at 4 °C for 1 hr and overnight, respectively. Eyecups were then quickly frozen in Shandon M-1 Embedding Matrix (Thermo Fisher Scientific) and cut at 12 μm. Retinal sections were washed twice in PBS and then blocked in PBS containing 5% Donkey serum and 0.3% Triton X-100 (PBST) for 1 hr at room temperature. Slides were incubated overnight at 4 °C with the primary antibody diluted in PBST at an appropriate concentration (see Key Resources Table). Sections were then washed three times with PBS and incubated with a secondary antibody and 1 μg/ml 4,6-diamidino-2-phenylindole (DAPI) for 1 hr at room temperature. After three washes in PBS, the sections were mounted in Fluoromount-G (SouthernBiotech).

Bright field images were taken using an EVOS XL Core Cell Imaging System (ThermoFisher). Fluorescence images were acquired with LSM-880 confocal microscope (Zeiss) with Zen software. FIJI and Photoshop CC 2019 software was used for image export, analyzing, and processing. Rhodopsin and S-opsin fluorescent intensity of organoid sections were quantified with FIJI using the maximum intensity projections of z-stack images. Multi-channel RGB (red, green, blue) images were separated into 8-bit grayscale images, and regions of interest were identified by applying the “Moment” threshold algorithm of FIJI. The same threshold algorithm was used for all images. Area, raw, and integrated fluorescence intensity in each image were then quantified with Fiji and plotted using RStudio version 1.1.463.

Animal experiments were conducted in the animal facility at the National Eye Institute, the National Institutes of Health. The facility approved the animal care and experimental procedures used in this study (NEI-ASP650).

For evaluating the effect of reserpine, we performed two sets of experiments and examined mouse retinal function by electroretinography (ERG) and structure by histology and/or immunohistochemistry. In the first set, the mice were injected with reserpine at P4 and harvested at P21. In the other set, the mice were given two injections of the drug at P7 and P14, and the ERG was recorded at P24 before harvesting the retina for structural analysis. For intravitreal injections at P4, pups of either sex of Nrl-GFP rd16 mice were anesthetized on ice and their eyelids were opened using a 30-gauge needle to gently expose the eye. Ketoprofen (0.5 µl of 1 mg/ml in PBS) was administered for analgesia. DMSO or reserpine (0.5 µl) was blindly injected using a glass micropipette (World Precision Instruments) produced by a Flaming-Brown micropipette puller (Sutter Instruments) and connected to an Eppendorf Femtojet air compressor (Eppendorf). For intravitreal drug injections at P7 and P14, mouse pups were anesthetized by intraperitoneal injection of ketamine (50 mg/kg body weight) and xylazine (5 mg/kg body weight). DMSO or reserpine (0.5–1 µl) was intravitreally injected using a Hamilton syringe with a 34-guage needle. Mice with eye/lens damage following the injection were excluded from further analyses.

ERG responses were recorded as previously published (Hargrove-Grimes et al., 2020). Briefly, after an overnight dark adaptation, mice were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). Eyes were dilated by topical administration of tropicamide (1% wt/vol, Alcon) and phenylephrine (2.5% wt/vol, Alcon), and the body temperature was maintained at 37 °C by a heating pad. After placing the reference electrode in the mouth and applying hypromellose ophthalmic demulcent solution (2.5% wt/vol, Gonak; Akorn) to each eye, gold wire loop electrodes were placed on the center of each cornea and ERG responses were recorded with an Espion E2 Visual Electrophysiology System (Diagnosys). Scotopic responses were recorded at increasing light intensities from 0.0001 to 10 cd·s/m² with inter-stimulus intervals ranging from 5 s to 60 s depending on stimulus intensity. After a light adaptation for 2 min, photopic responses were recorded at increasing light intensities from 0.3 to 100 cd·s/m² under a background light saturating rod function. The a-wave amplitudes were measured from the baseline to the negative trough and the b-wave amplitudes were measured from the a-wave trough to the wave peak. Mice with eye/lens damage following the injection were excluded from further analyses.

Based on the initial data, the sample size of animals used in each group was determined by an unpaired t-test to be less than six mice in each group (http://www.biomath.info/↗). For experiments involvng mouse or human retinal organoids, at least four independent experiments, each of which had at least three organoids, were performed unless specified. For animal studies, at least two retinas from one mouse of at least four different litters were evaluated. All data were expressed as mean ± standard deviation (SD) unless specified. An unpaired t-test was used to compare the mean between the two groups. For the comparison of three or more groups, one-way AVOVA was performed. Results with a <i>p-value <0.05 were considered statistically significant.

All materials including the pluripotent stem cell lines and animals are available upon request. A Material Transfer Agreement complying with the guidelines of the National Institutes of Health has to be established before shipment.