Toward Retinal Organoids in High-Throughput

Abstract

Human retinal organoids recapitulate the cellular diversity, arrangement, gene expression, and functional aspects of the human retina. Protocols to generate human retinal organoids from pluripotent stem cells are typically labor intensive, include many manual handling steps, and the organoids need to be maintained for several months until they mature. To generate large numbers of human retinal organoids for therapy development and screening purposes, scaling up retinal organoid production, maintenance, and analysis is of utmost importance. In this review, we discuss strategies to increase the number of high-quality retinal organoids while reducing manual handling steps. We further review different approaches to analyze thousands of retinal organoids with currently available technologies and point to challenges that still await to be overcome both in culture and analysis of retinal organoids.

Organoids are three-dimensional cellular ensembles of multiple organ-specific cell types. The cell types are grouped and spatially organized similarly to an organ and recapitulate specific functions of that organ (Lancaster and Knoblich 2014b; Clevers 2016; Pașca et al. 2022). Organoids can be generated in vitro from different types of stem cells, such as organ-restricted adult stem cells or pluripotent stem cells (PSCs). Adult organ stem cells isolated from biopsies have the capacity to generate organoids resembling the source organ (e.g., intestinal organoids) (Clevers 2016; Kim et al. 2020). PSCs with the potential to differentiate into most cell types of the body can be subdivided into embryonic stem cells, derived from the inner cell mass of preimplantation embryos, and induced pluripotent stem cells (iPSCs), reprogrammed from adult cells such as fibroblasts or blood mononuclear cells (Liu et al. 2020; Yamanaka 2020). All these stem cells are presently being used to produce organoids that mimic different organs of the whole body (Clevers 2016; Kim et al. 2020). The organoids vary greatly in size and complexity according to the diverse nature of the tissues they model. As an example, intestinal organoids that are already used for high-throughput applications such as drug screenings (Lukonin et al. 2020) are rather small (on average 100 µm), develop relatively quickly (differentiated cells are present 5–7 d after plating), and typically contain less than 10 different cell types (Boonekamp et al. 2020; Pleguezuelos-Manzano et al. 2020). On the other hand, organoids of the central nervous system, such as cerebral and retinal organoids, are quite large (more than 1 mm), develop slowly (during 2–9 mo), and can include more than 40 different cell types arranged in a complex manner (Lancaster et al. 2013; Lancaster and Knoblich 2014a; Cowan et al. 2020; He et al. 2022).

Organoids as model systems built from human cells are a promising new tool for studying human development and disease in vitro. Source cells with a disease mutation can be used to generate patient-specific disease models, either by reprogramming patient cells to iPSCs or by introducing a precise patient mutation into control iPSCs by genome engineering. Patient-specific disease models allow not only the study of mechanisms of human disease in an unprecedented manner but also the testing of therapies such as gene therapy, small-molecule drugs, or oligonucleotides (Afanasyeva et al. 2021; Zhang et al. 2021a).

While studying organ development and disease phenotypes may require just a few dozen organoids, screening for novel drugs, for example, necessitates the generation of thousands of organoids in a high-throughput manner. However, this is easier said than done. The complex differentiation protocols to generate organoids typically include many manual handling steps. The maximum differentiation state may be reached only after weeks or months of culture, and the random 3D organization of organoids with high variability in size and shape is a problem for automated analysis. Hence, novel approaches to the culture and analysis of organoids in a high-throughput manner must be found. This review addresses this need in the context of human retinal organoids.

HUMAN RETINAL ORGANOIDS AS A MODEL SYSTEM

Retinal Organoids Mimic the Development of Human Retina

Retinal organoids were first generated by the group of Sasai from mouse embryonic stem cells (Eiraku et al. 2011) and shortly thereafter from human PSCs (Meyer et al. 2011; Nakano et al. 2012). Mouse retinal organoids develop in just a few weeks while human retinal organoids reach comparable stages of development only after several months (Llonch et al. 2018), reflecting the difference in the developmental rates of mice and humans. Several studies have shown that human retinal organoids mature at a rate comparable to the developing human retina, in terms of gene-expression profiles, cell-type generation, and morphological properties (Cowan et al. 2020; Lu et al. 2020; Sridhar et al. 2020; Kim et al. 2023).

Comparison of Retinal Organoids to Human Retina, Validating Them as a Model System

The human retina is a highly complex organ of more than 100 cell types with different morphologies, functions, and gene-expression profiles (Masland 2012; Zeng and Sanes 2017). Cell bodies in the human retina are arranged in three nuclear layers: the outer nuclear, inner nuclear, and ganglion cell layer. Photoreceptor cell bodies are located in the outer nuclear layer, horizontal, bipolar, amacrine, and Müller cells in the inner nuclear layer, and amacrine and ganglion cells in the ganglion cell layer. Synaptic connections between cells of adjacent nuclear layers are established in the outer- and inner-plexiform layers, respectively (Masland 2012). Finally, photoreceptors of the neural retina are in close contact with retinal pigment epithelial cells, which are important for photoreceptor function (Strauss 2005; Sparrrow et al. 2010).

Validation of retinal organoids as a model of the human retina depends on several criteria, namely, the presence of retinal cell types, their correct arrangement, responses to light, and synaptic transmission (Lancaster and Knoblich 2014b). The major retinal cell classes (photoreceptors, horizontal-, bipolar-, amacrine-, ganglion- and Müller cells) have been identified immunohistochemically in organoids generated with several different culture protocols (Zhong et al. 2014; Capowski et al. 2018). Photoreceptors were typically found on the outside of the organoids, while other cell types were located more internally. Only recently were we able to generate five-layered retinal organoids with three nuclear and two synaptic layers and to confirm the correct localization of organoid cell classes within the appropriate layers by comparisons to the adult retina by immunostaining (Cowan et al. 2020). Moreover, organoid photoreceptors are light sensitive and able to transmit light responses via synapses to second- or third-order cells (Zhong et al. 2014; Cowan et al. 2020; Saha et al. 2022). Single-cell genomics confirmed the presence of all major retinal cell classes in retinal organoids and a high similarity of organoid cell types to those of the human retina (Kim et al. 2023).

By comparing organoids to healthy, light-responsive human retina (fovea and periphery), we furthermore validated the expression of many disease-associated genes in the same cell types as in human adult retina (Cowan et al. 2020). Thus, existing retinal organoids can be accepted as models for the human retina that recapitulate aspects of cell-type diversity, arrangement, gene expression, and function.

Retinal organoids, however, can still be improved as they currently lack some cell types and anatomical features of the human retina. Organoid photoreceptors are currently missing contact with retinal pigment epithelium, which is often present in isolated patches on retinal organoids. Microfluidic chips have been developed to establish contact of retinal pigment epithelium with photoreceptor outer segments (Achberger et al. 2019). Additionally, efforts are ongoing to integrate missing microglia and blood vessels into retinal organoids, respectively (Gao et al. 2022; Chichagova et al. 2023; Li et al. 2023). Further, retinal organoids lack an optic nerve because ganglion cells are lost in mature retinal organoids. Therefore, they are typically not connected to brain structures targeted by retinal ganglion cell axons (Zhong et al. 2014; Cowan et al. 2020). In assemblies of retinal and brain organoids, axonal projections toward brain structures have been described (Fligor et al. 2021; Fernando et al. 2022). Finally, retinal organoids resemble more the retinal periphery and, although cone-rich organoids and organoids with some anatomical characteristics of the fovea-parafovea region have been described, organoids containing a fovea or organoids of fovea identity still await publication (Kim et al. 2019; Cowan et al. 2020; Völkner et al. 2022). Organoid (co-)culture modifications toward a more complete retina model are promising, but need to become more robust, reproducible, and automation friendly for high-throughput applications.

RETINAL ORGANOIDS FOR THERAPY DEVELOPMENT

The treatment of retinal diseases requires a deep understanding of the mechanisms and genes involved. Many genes are associated with inherited retinal degeneration, where mutations in a single gene can lead to deterioration or even complete loss of vision, and different mutations in the same gene can result in different clinical phenotypes (RetNet, the Retinal Information Network). Certain genes are associated with specific retinal diseases and some of these genes are important for the function of specific cell types (Berger et al. 2010).

Only recently has single-cell RNA sequencing of the healthy human retina allowed comprehensive identification of the cell types that express known disease-associated genes. Expression of these genes is often restricted to one or a few cell types (Cowan et al. 2020), and, thus, in many cases, the treatment of retinal diseases will be cell-type specific.

There are currently four main approaches to the therapy of retinal diseases: gene therapy, optogenetic vision restoration, cell therapy, and small molecule treatment. First, gene therapy aims to introduce a functional copy of the mutated gene into the impacted cells using viral vectors or to repair the mutation with genome-engineering tools. To prevent side-effects on cell types that do not express the disease-relevant gene, entry of the virus or expression of the virus cargo should be cell-type specific. Organoids produced in large numbers will help identify viral vector variants with affinity to specific cell types, as well as to screen for cell-type-specific expression of the virus cargo mediated by promoter elements (Gonzalez-Cordero et al. 2018; Jüttner et al. 2019; Tornabene et al. 2019; Garita-Hernandez et al. 2020; Völkner et al. 2021; McClements et al. 2022; Tso et al. 2023). Furthermore, organoids could be used to optimize new gene-editing tools such as base editing, prime editing (da Costa et al. 2021), small oligos (Dulla et al. 2018), or homology-directed repair to repair disease mutations (Gallego et al. 2020; Pasquini et al. 2020). Gene therapy needs to be adapted to individual mutations and more than 250 genes are associated with inherited retinal diseases (RetNet). Several gene-therapy approaches are already in clinical trials (clinicaltrials.gov) and Luxturna to treat RPE65-based retinal degeneration has been approved (Russell et al. 2017).

Second, artificial stimulation of the retina by optogenetic tools (Boyden et al. 2005) can be first tested and optimized in retinal organoids (Garita-Hernandez et al. 2018). Although it has been shown that retinal organoids contain functional synapses transmitting signals from photoreceptors to second- and third-order cells (Cowan et al. 2020), it is still unclear to what extent retinal circuits are present in organoids (Fathi et al. 2021).

Third, cell therapy aims to replace degenerated cells in patients and the cells for transplantation could be purified from retinal organoids. However, the functional integration of photoreceptors, for example, into existing retinal circuits is a challenge (Llonch et al. 2018).

Fourth, small-molecule treatments aim to slow or prevent retinal degeneration. The small-molecule drugs are usually applied systemically or via intravitreal injection and they can easily target the entire retina (Maneu et al. 2022). However, these drugs are not usually cell-type specific and may also interfere with healthy cells. This can be modeled by first applying the compound to the organoid culture medium. To identify small-molecule drugs with the desired effect, typically thousands of chemical compounds are tested in parallel in several replicates and this would require the generation of many thousand retinal organoids. Due to the present technical limitations, high-throughput, small-molecule screens have not yet been performed on human retinal organoids. However, initial attempts at large-scale screens have been made using simpler organoid models: Lukonin et al. have published results of a screen of 2789 compounds that used approximately 450,000 intestinal organoids (Lukonin et al. 2020). Tumor biopsies can also be cultured under organoid growth conditions and have been used for drug screenings. These models can facilitate personalized medicine by testing the response of an individual's specific tumor to various treatments, allowing for the selection of the optimal treatment option (Aboulkheyr Es et al. 2018).

Finally, the toxicology of treatments developed for medical conditions not primarily affecting the eye needs to be studied early in drug development. Several compounds such as thioridazine or tamoxifen can cause severe damage to the retina under high dosage (Corradetti et al. 2019). Retinal organoids can be used to screen for negative effects of potential drugs on the retina (Fig. 1; Dorgau et al. 2022).

Figure 1.

Main therapeutic approaches for retinal diseases that can be tested on retinal organoids. Organoids can also be a source of cells for transplantation in cell therapy. In the center of the figure is an illustration of a retinal organoid with a black bud of retinal pigment epithelium (RPE) attached to it. (Figure created with BioRender.com.)

SCALING UP RETINAL ORGANOID NUMBERS

Organoids can be generated from patient cells or from cells in which precise disease mutations have been introduced by genome engineering. Such retinal organoids have been used already in studies to elucidate disease phenotypes and the mechanisms of disease, as well as to test targeted therapies (Zhang et al. 2021a) and a small number of compounds (Dorgau et al. 2022). Relatively few organoids in the range of several dozen are sufficient for these types of studies. Studies like high-throughput compound screens with the aim to produce and analyze organoids in the range of thousands have not yet been conducted due to the limitations of both organoid culture and analysis that need to be overcome.

TOWARD HIGH-THROUGHPUT ORGANOID CULTURE

Although various protocols for organoid generation exist, they are mostly based on common principles (Zhang et al. 2021b) and can be divided into several steps. First, iPSCs are subjected to neural induction to initiate differentiation of a neural identity. Second, eye-field development is induced and, third, retina tissue is harvested and cultured floating in 3D for further development and differentiation during several months. Each of these steps is a bottleneck because they are each typically associated with time-consuming manual procedures that limit the rate at which organoids can be handled. Furthermore, the efficacies of each of these steps vary and the success rate of each procedure needs to be maximized to generate organoids in large numbers. In short, high-throughput culture of organoids requires the optimizing of organoid variability and numbers and the reduction of labor-intensive manual handling.

Given that the differentiation time-course of retinal organoids resembles that of the human retina in utero, optimal retinal organoid culture should closely mimic human embryonic development. During embryonic development, the neural tube gives rise at its anterior end to the forebrain, midbrain, and hindbrain. The forebrain gives rise to the optic vesicles that further develop into retina. Several publications have reported that PSC lines differ in their capacity to form retinal organoids (Capowski et al. 2018; Wang et al. 2018; Mellough et al. 2019; Rashidi et al. 2022). Indeed, a common by-product of retinal organoid differentiation are brain organoids without retina (Meyer et al. 2011; Fernando et al. 2022). Similarly to retinal organoids, different PSC lines when subjected to differentiation into cerebral organoids form different parts of the forebrain, such as dorsal forebrain, ventral forebrain, and retina, which suggests that the ratio of tissues generated is cell-line dependent (Kanton et al. 2019).

Hence, choosing an appropriate cell line is the first step toward high-throughput generation of retinal organoids. Some protocols add growth factors or small molecules to improve differentiation into retina, but here again the responses to both can be cell-line dependent and each line needs to be optimized (Chichagova et al. 2020; Regent et al. 2022). Most protocols perform neural induction on embryoid bodies (EBs), which are small clumps of PSCs aggregated in 3D. The size of the EBs can influence the efficiency of retinal organoid generation and some cell lines only form retinal organoids efficiently from EBs of a particular size (Mellough et al. 2019; Cowan et al. 2020). EB size can be regulated by seeding individualized PSCs at defined numbers into the round bottom cavity of a cell-culture dish (Choi et al. 2010; Cowan et al. 2020). Traditionally, this is performed in low-attachment U-bottom or V-bottom 96-well plates (Nakano et al. 2012) with one EB per well. However, these plates are expensive, feeding requires large amounts of medium, and several thousand PSCs per well need to be seeded for efficient EB formation. More high-throughput-friendly EBs can be formed in hydrogel microwell arrays (Cowan et al. 2020; Decembrini et al. 2020; Rashidi et al. 2022). The molds to prepare agarose microwell arrays are commercially available, reusable, and take up to 256 EBs that can easily be cultured in a standard multiwell plate. For two PSC lines that generate five-layered human retina, we have shown that EBs generated from only a few hundred cells lead to higher yields of retina tissue than seeding more than a thousand cells per microwell. Using agarose microwells, up to 4000 organoids can be generated from a single well of a six-well plate of iPSCs (Cowan et al. 2020), thus allowing high-throughput organoid production. Some protocols perform neural induction on PSCs in 2D without EB formation (Singh et al. 2015), but this requires many wells of evenly seeded PSCs to produce thousands of retinal organoids.

For differentiation into the eye field, EBs are either maintained in 3D in the presence of growth factors and extracellular matrix components (Nakano et al. 2012; Rashidi et al. 2022) or they are plated on Matrigel-coated dishes (Zhong et al. 2014). Most protocols require isolation of retina tissue from floating 3D aggregates or from Matrigel plates by manual microdissection. This step is not only time consuming and very challenging technically, but it also leads to discarding of retina that cannot be identified morphologically or handled within a reasonable amount of time. However, detaching the entire content of the Matrigel plate by scraping allows complete harvesting of all retina tissue from each plate (Cowan et al. 2020; Regent et al. 2020). Avoiding manual microdissection is the second critical step toward high-throughput production of retinal organoids.

Even though AMASS (agarose microwell-assisted seeding and scraping) allows the production of thousands of retinal organoids (Cowan et al. 2020), many further manual steps need to be eliminated for high-throughput organoid culture. One of these is the changing of organoid culture media several times per week during up to 38 wk of development, depending on the time point chosen for analysis. Given free-floating cultures in low-attachment 10-cm dishes, automatizing media exchange would require highly advanced and potentially custom-built robotic equipment that is currently not available in most academic laboratories. It has been suggested to culture organoids in bioreactors; although this seems to be beneficial for early organoid development, it is damaging to older organoids that contain fragile structures such as photoreceptor outer segments (Ovando-Roche et al. 2018).

A further challenge to producing thousands of organoids is the inefficiency of organoid differentiation itself, where a percentage of the 3D aggregates will not contain retina. Hence, these organoids need to be excluded from experiments or the number of replicates needs to be increased.

For many procedures, such as drug testing, imaging-based readouts, or transduction with gene therapy vectors, retinal organoids must be singularized in wells of multiwell plates. This involves labor-intensive manual intervention to transfer organoids to an assay plate that may also be subject to bias when selecting organoids for inclusion in an experiment that must meet certain quality criteria, such as the presence of retina. Automatizing this step to allow experiments on thousands of retinal organoids, for example in a drug screen, would require a custom-built machine that uses artificial intelligence to assess organoid quality and to sort them into assay plates without reducing organoid viability or disrupting fragile outer segments. Such an algorithm has been successfully developed to assess the quality of mouse retinal organoids (Kegeles et al. 2020).

Organoid Morphology

Optimal organoids for accurate high-throughput screening should be consistent in size, regular in shape, and include only retina. However, retinal organoids obtained by scraping typically contain several buds of retina as well as further buds of nonretinal identity that do not negatively impact retina development. The retina within the buds may be regularly organized in layers but the size, number, and shape of the buds differ greatly from organoid to organoid. Furthermore, retinal organoids are not perfectly round and on their long axis can reach diameters from one to three millimeters. In a high-throughput experiment such as a drug screen, the inconsistent sizes of the organoids may influence the outcome. Because of their relatively large size, retinal organoids do not fit into 1536-well plates and only barely into 384-well plates, where media consumption and thus viability may differ according to size and so may potentially influence experiment outcome.

Culture Time

Human retinal organoids in culture reach stages in which their cell types are comparable to adult retinal cells at around 30 wk of organoid development (Cowan et al. 2020). Although it has been reported that human retinal organoids can be cryopreserved (Nakano et al. 2012; Reichman et al. 2017), it remains to be shown whether cryopreserved organoids retain a layered structure with at least a defined photoreceptor layer and inner nuclear layer with viable cells. Thus, at present, cultures cannot be paused and experiments need to be strictly planned many months in advance. This long culture time brings the risk of losing organoids from events such as contamination, equipment failure, and supply-chain interruptions, or due to human error. The long culture time also places pressure on laboratory space that may limit the number of organoids that can be cultured. Thus, organoid analysis time points need careful selection to balance the required degree of organoid maturation against the shortest possible culture time. Ways to increase the rate of organoid development without severely impacting organoid quality have so far proven elusive. One possible solution for some experiments is to use organoids from primates, which develop faster in accordance with the shorter gestation times of the animals (Fig. 2; Jacobo Lopez et al. 2022).

Figure 2.

Challenges of retinal organoid culture and measures to increase throughput. Each step of the cultivation process requires careful optimization. (iPSC) Induced pluripotent stem cells, (EB) embryoid body, (CNS) central nervous system, (RPE) retinal pigment epithelium.

High-Throughput Organoid Analysis

Although organoids are open to analysis by a wide range of routine assays, depending on the scientific question, the analysis of thousands of retinal organoids remains a challenge.

Imaging

The most widely used readouts for retinal organoids are imaging based. Although retinal organoids are 3D model systems, phenotype analysis is often performed by immunohistochemistry on thin sections of fixed tissue. This can produce data on positioning and the abundance of cell types and antigens, but each stained section shows the state at only one time point. Since organoid size and cell-type composition are not constant, reliable comparison of sections may be difficult, especially when the differences in phenotype are small.

Preparing organoid sections for imaging is a labor-intensive, low-throughput manual task. However, cultured murine retinal organoids in arrays of hydrogel milliwells allowed imaging of individual organoids at different time points. Furthermore, milliwell arrays containing organoids can be embedded for thin sectioning and arrayed immunohistochemical analyses (Decembrini et al. 2020). This method has yet to be tested on human retinal organoids.

Using tissue-clearing methods, retinal organoids can be stained and imaged without sectioning and the associated sectioning problems, but this may complicate data analysis and image acquisition (Reichman et al. 2017; Cora et al. 2019). However, tissue clearing and 3D analysis are becoming more automation friendly and could be adapted for retinal organoids (Ueda et al. 2020).

Live Imaging

A more scalable approach is to use live imaging. This allows analysis of an organoid over several time points and, thus, comparison of each organoid before the onset of a phenotype or the administration of a treatment. The data analysis then is more robust. Several imaging techniques have been used to characterize retinal organoids. Methods like phase contrast, differential interference contrast, or standard brightfield imaging of organoids can assess overall viability and the quality of the tissue (Browne et al. 2017). This is suitable for high-throughput screening experiments but only a limited amount of information can be drawn from these images.

The analysis of organoids in which fluorophores are expressed cell-type specifically retrieves additional information at the cell-type level. This can be carried out with engineered stem cell lines or after delivery of a transgene to more developed organoids.

Engineering stem cell lines to carry reporter genes and their differentiation into retinal organoids has been successful and may be a useful tool to study individual cell types in a high-throughput manner (Vergara et al. 2017; Phillips et al. 2018; Lam et al. 2020; Jones et al. 2022; Nazlamova et al. 2022). A common approach is to engineer a fluorophore gene into the locus of a cell-type-specific gene and so to couple the fluorophore expression pattern to that of the cell-type-specific gene. This leads to uniform labeling of a specific cell type. However, this process is very time consuming since proper validation of the line can only be achieved after the full development of the organoid. Tagging cell-type-specific genes with a fluorophore may also lead to undesired effects on gene expression and the development of that cell type. This would be particularly troublesome when patient-derived organoids are used. Furthermore, gene editing of PSCs can cause off-target mutations and hence influence differentiation efficiency, making it more difficult to generate organoids for high-throughput applications. A simpler approach is to deliver the fluorophore transgene into more developed organoids.

Although transfection can be used for organoids (Lancaster et al. 2013; Fujii et al. 2015), the use of a viral vector as a delivery vehicle is more common. In efforts to use retinal organoids as preclinical models for gene therapy, several groups have shown that transduction of organoids using adeno-associated viral vectors is a feasible way to label a multitude of cell types. These vectors in combination with cell-type-specific promoter elements can label specific cell types in mature retinal organoids (Gonzalez-Cordero et al. 2018; Völkner et al. 2021; McClements et al. 2022; Tso et al. 2023). One disadvantage of viral delivery is that transduction efficiency may vary between organoids and be a further source of variability in high-throughput applications.

Although promising, live imaging of organoids for high-throughput applications also comes with certain problems. Retinal organoids are inherently opaque, and light from the imagers used for screening may fail to completely penetrate; this will lead to an asymmetric readout. This might be acceptable if the organoids stay in the same orientation during an experiment. However, minor turbulence in the medium (e.g., during media change) can result in movement of the organoid that exposes a different part of the tissue and makes comparisons between time points unreliable. Anchoring of retinal organoids and maintaining a sufficient supply of nutrients need to be addressed.

Molecular Assays

Common 2D cell-culture assays can be adapted for retinal organoids. Western blots, qPCR, flow cytometry, as well as recent “omic” technologies such as transcriptomics, epigenomics, or proteomics have been used to study organoids (Afanasyeva et al. 2021). These assays can be performed on entire organoids or on organoids dissociated into single cells. Modified versions of whole-organoid RNA sequencing can be explored as scalable readouts. The difficulty of these assays is to retain cell-type-specific information when applied to high-throughput applications. Analysis of single or sorted cells retains information about cell types but loses information about cell positioning. Dissociation of organoids to single cells must avoid reducing cell viability, and the dissociation of large quantities of organoids simultaneously is very challenging.

Molecular assays typically used for high-throughput screening, such as luminescence, fluorescence, or colorimetric assays, have yet to be explored in the context of retinal organoids.

An alternative to high-throughput screens that test just one condition per retinal organoid are pooled screening assays that combine several perturbations inside a single organoid. For example, CRISPR screening can modulate the expression of different genes in individual cells within a single organoid and the effects can be read out by single-cell RNA sequencing (Ungricht et al. 2022). Relatively few organoids are required for pooled screens but they yield information about changes in gene expression in a large number of cells.

Functional Assays

The capacity to recapitulate some of the functions of an organ is a key feature of organoid biology. Functional assays, such as patch clamp recordings (Meyer et al. 2011; Zhong et al. 2014; Deng et al. 2018; Li et al. 2021), multielectrode array recordings (Buskin et al. 2018; Hallam et al. 2018; Chichagova et al. 2023), and calcium imaging (Cowan et al. 2020), have been applied to retinal organoids. This has revealed that organoid photoreceptors are intrinsically light sensitive and transmit visual input information through synapses to inner nuclear cells, similarly to the human retina.

Functional assays like patch-clamping are not feasible as a high-throughput readout due to the extensive manual work required to measure just a few cells. Scaling up multielectrode array-based assays is becoming more feasible and has been applied to cerebral organoids (Durens et al. 2020). This may be adaptable for retinal organoids; however, multielectrode arrays would not allow comparison of cell types but could be used to study responses of ganglion cells as the only spiking cells in the retina. Current drawbacks of MEA recordings to study retinal ganglion cells in retinal organoids are the loss of retinal ganglion cells in organoids aged more than 20 wk (Buskin et al. 2018; Fligor et al. 2021) and the internal location of the retinal ganglion cells.

Calcium imaging may be the most sensitive functional assay for screening applications since it is based on live imaging and the expression of the calcium sensor can be targeted to certain cell types. Combining calcium imaging with visual stimulation in larger screens would require specialized equipment and sophisticated data analysis.

While many different assays are available for studying retinal organoids, very few of them are suitable for high-throughput applications. The challenge is to choose a readout that minimizes manual handling time per organoid and that, ideally, can be automated without loss of information at a cell-type level. A simple and clear readout is the essential basis of any high-throughput screening (Fig. 3).

Figure 3.

Methods for retinal organoid analysis and their suitability for high-throughput applications. Some imaging-based, molecular, or functional assays can be more easily adapted for analyzing high numbers of retinal organoids (right side of figure). (Figure created with BioRender.com.)

DISCUSSION

Retinal organoids are a new and powerful tool to study retinal biology, development, and disease. Their use as preclinical models to test new therapeutic approaches, especially in high-throughput, could revolutionize drug and therapy research of retinal diseases.

For reliable use in screening, several features of retinal organoid cultures need to be optimized and they must become more automation friendly. The production of thousands of retinal organoids is mostly theoretical at present but is essential for high-throughput screening. In addition, most differentiation protocols rely on expensive supplements, and more cost-efficient alternatives are necessary for routine high-throughput organoid screens. Another significant challenge is to sort out genuine organoids from other cell aggregates that might form and so to ensure reliable organoid quality throughout a screen.

Finding suitable phenotypes for high-throughput screening is also a challenge, as most reported disease-modeling phenotypes are monitored by transcriptome changes or immunohistochemistry (Chirco et al. 2021; Kruczek et al. 2021; Chahine Karam et al. 2022; Völkner et al. 2022). These phenotypes need to be adapted in some way and/or other more screening-friendly phenotypes sought to test and screen therapies on diseased organoids. Improvements in both organoid throughput in culture and the analysis of retinal organoids are important not only for high-throughput screening of organoids but also to facilitate the generation of high-quality retinal organoids for research that exploits this exciting new model.