Retinal ganglion cells in a dish: Current strategies and recommended best practices for effective in vitro modeling of development and disease

Introduction

Retinal ganglion cells (RGCs) are projection neurons of the central nervous system (CNS) that play a primary role in the transmission of visual information between the retina and the brain. RGCs receive visual input from photoreceptors through bipolar cells or amacrine cells, and connect with downstream targets in the brain via long axonal projections (Crair and Mason, 2016; Erskine and Herrera, 2014). The function of RGCs is also dependent upon their highly compartmentalized structure, in which RGC dendrites are found within the inner plexiform layer (IPL) in the retina, the RGC soma integrates signals within the ganglion cell layer (GCL), and finally RGC axons project toward the brain, initially within the nerve fiber layer along the inner retinal surface before entering the optic nerve and exiting the eye (Conforti et al., 2007; Whitmore et al., 2005; Yu et al., 2013). The proximal part of the RGC axon remains unmyelinated until the optic nerve head (ONH) through the lamina cribrosa, and then becomes myelinated and eventually forms synaptic contacts with downstream neurons primary in the lateral geniculate nucleus (LGN), superior colliculus (SC) and pretectal areas of the brain (Parmhans et al., 2021; Syc-Mazurek and Libby, 2019). Notably, glial cells including astrocytes and microglia interact with RGCs along each RGC compartment in different ways. For example, astrocytes in the nerve fiber layer (NFL) within the inner retina secret neurotrophic factors that are responsible for RGC development, and microglia regulate synaptic pruning as well as phagocytosis of dead RGCs within the IPL, GCL and NFL(Anderson and Vetter, 2019; Au and Ma, 2022; Vecino et al., 2016). As the sole projection neurons that covey visual signals from the eye to the brain, RGCs are long (~50mm), postmitotic projection neurons that are particularly susceptible to acute injury or chronic neurodegeneration characteristic of glaucoma or other optic neuropathies, leading to irreversible progressive loss of vision(Yu et al., 2013). Historically, research in rodent models has successfully investigated many mechanisms of RGC injury in a variety of disease states(Agostinone et al., 2018; Belforte et al., 2021; Duan et al., 2015; Tran et al., 2019), and several studies have analyzed cellular mechanisms underlying RGC neurodegeneration and identified factors that can promote RGC regeneration in experimental rodent models(Agostinone et al., 2018; Belforte et al., 2021; Jacobi et al., 2022; Lindborg et al., 2021). While these studies have provided a wealth of knowledge pertaining to the mechanisms associated with RGC damage, more recent studies have revealed numerous differences between RGCs from rodent and primate sources (Peng et al., 2019), as well as significant variances in glia across rodents and primates (Galatro et al., 2017a; Galatro et al., 2017b; Hodge et al., 2019; Oberheim et al., 2009; Patir et al., 2019; Smith and Dragunow, 2014; Zhang et al., 2016), suggesting that perhaps at least some of these mechanisms may not necessarily be conserved in patients. Thus, a critical need exists to develop a complimentary model with which we can compare findings in experimental rodent models and also confirm phenotypes in a more human-relevant cellular system.

Human pluripotent stem cell (hPSC)-derived retinal organoids have been developed as a tool for understanding human retinal development as well as modeling retinal diseases (Capowski et al., 2019; Fligor et al., 2018; Meyer et al., 2011; Nakano et al., 2012; Wahlin et al., 2017). hPSCs can be further subdivided as either human embryonic stem cells (hESCs) or human induced-pluripotent stem cells (hiPSCs) but collectively, hPSCs have the ability to give rise to all cell types of human tissues. As hiPSCs can be generated through the reprogramming of adult somatic cells such as fibroblasts or peripheral blood mononuclear cells(Takahashi et al., 2007; Yu et al., 2007), they can be used as a novel in vitro model of genetic contributions to disease when derived from patient samples harboring these genetic variants.

Historically, it has been exceptionally difficult to obtain human RGCs for the study of neurodegenerative features and as such, hPSCs have become a valuable resource to produce retinal cells for studying human retinogenesis, drug screening, disease modeling, or transplantation. We and others have previously demonstrated the generation of three-dimensional retinal organoids from hPSCs in a stepwise manner, mimicking many temporal and spatial characteristics of human retinogenesis, consisting of all of the major cell types of the human retina(Capowski et al., 2019; Eldred et al., 2018; Fligor et al., 2018; Meyer et al., 2009; Nakano et al., 2012; Ohlemacher et al., 2015; Wahlin et al., 2017; Zhong et al., 2014). Alternatively, it may often be preferable to study cell autonomous features of retinal diseases, in which case it becomes more desirable to isolate specific cell types such as RGCs into two-dimensional cultures, with specific culture medium formulations that promote RGC survival, neurite outgrowth, and functional maturation(Gomes et al., 2022; Ohlemacher et al., 2016; Teotia et al., 2019; VanderWall et al., 2019). We and others have previously demonstrated the use of hPSC-RGCs for studying RGC development and modeling glaucoma-associated neurodegeneration in this manner(Gomes et al., 2022; Teotia et al., 2020; Teotia et al., 2017b; VanderWall et al., 2020; VanderWall et al., 2019). Moreover, as improvements are continually made to hPSC-RGC differentiation protocols, it has become increasingly possible to use these cells for drug screening as well as for transplantation with the goal of RGC replacement(Patel et al., 2020; Zhang et al., 2021). However, as recent studies have demonstrated differential gene signatures between hPSC-RGCs compared to RGCs found within the fetal or adult retina(Sridhar et al., 2020), there are likely some challenges to further adapt hPSC-RGCs to create the most physiologically relevant model for human RGC neurodegeneration. Among the possibilities that may account for the differences between human RGCs and hPSC-RGCs is the environment in which they grow and mature. hPSC-RGCs typically lack the compartmentalized and polarized structure of RGCs in vivo, including the interactions with other cell types. More so, while RGCs interact with many types of neurosupportive cells in vivo including astrocytes, microglia, and vascular cells, these cells are not typically found within retinal organoids or other hPSC-RGC cultures. Without the support of these surrounding cells, these cultures are likely limited in their ability to provide the factors necessary to support the maturation of hPSC-RGCs in vitro. To create the most physiologically relevant hPSC-RGCs and to be able to mimic the neurodegenerative conditions that affect RGCs within the patient eye, we here discuss current RGC differentiation systems as well as current applications for modeling RGC diseases with hPSC-RGCs. Further, we provide our opinions on how to reduce variability across experimental groups by characterizing the features required for the definitive identification and characterization of hPSC-RGCs, as well as the necessary approaches for the advancement of hPSC-RGCs toward more translational applications.

Established Approaches for the Directed Differentiation of Human RGCs

Approaches have been developed by multiple groups to successfully generate hPSC-RGCs appropriate for a variety of applications including in vitro studies of retinogenesis, disease modeling, and cell replacement related to glaucoma or other optic neuropathies (Risner et al., 2021; Sluch et al., 2017; Teotia et al., 2020; Teotia et al., 2017b; VanderWall et al., 2020; VanderWall et al., 2019). These protocols to derive hPSC-RGCs can be divided into three main categories: (1) 2D differentiation, in which the entire differentiation process from hPSCs to RGCs is conducted in adherent cultures for the duration of the differentiation process; (2) 3D differentiation, in which at least part of the differentiation process (especially the latter stages) occur as suspension cultures to better simulate the spatial features of the developing neural environment, particularly including the formation of retinal organoids; and (3) direct differentiation, in which cells are not guided in a stepwise manner to become RGCs but rather, either undifferentiated hPSCs or unrelated somatic cells such as fibroblasts are genetically reprogrammed with RGC-associated transcription factors to drive their fate towards RGCs. While each of these three approaches have been used to effectively produce RGCs, they also have their own strengths and weaknesses that need to be considered when choosing an appropriate protocol for the derivation of RGCs for downstream applications (Figure 1).

Figure 1.

Schematic summary of current methods to derive RGCs in vitro.

The generation of hPSC-RGCs through the use of 2D differentiation protocols are often conducted by directing the differentiation of hPSCs toward an RGC fate by treatment with small molecule inhibitors and activators of key signaling pathways. Many of these protocols primarily follow dual SMAD inhibition methods to induce the initial differentiation of neural progenitors(Chambers et al., 2009), with subsequent modifications for RGC enrichment(Croteau et al., 2022; Risner et al., 2021; Sluch et al., 2017). 2D differentiation approaches provide advantages including a typically faster differentiation timeline to yield hPSC-RGCs. Further, as adherent cultures, these systems are typically less complicated to adopt, and their adherent nature lends itself to the large-scale production and screening of hPSC-RGCs. However, 2D differentiation protocols do not necessarily recapitulate the 3D retinal microenvironment and because they are unable to mimic the spatial structure of the developing retina in the dish, these approaches may not always be the most physiologically relevant, especially when cellular interactions are an important consideration for experimental questions. More so, in 2D cultures of differentiating RGCs, it becomes challenging to definitively identify and characterize hPSC-RGCs. While within the retina, RGCs can be identified based upon their location and morphological features, these features are not reliably preserved in a culture dish. Further, the expression of characteristic molecular markers including BRN3 or ISL1 can also be used to identify RGCs, as these markers are specific to RGCs uniquely within retinal tissue (Badea and Nathans, 2011; Wu et al., 2015). However, these common RGC transcription factors are often expressed elsewhere in the body, decreasing their utility when starting from a pluripotent source. For example, previous studies have shown the expression of RGC markers such as BRN3 outside of the retina in auditory and somatosensory neurons (Badea et al., 2012), while ISL1 is a well-known marker for motor neurons(Liang et al., 2011). Thus, given the ability of hPSCs to generate all cell types of the body, the mere expression of a single transcription factor is not sufficient to definitively identify a cell as an RGC, and further steps must be taken to definitively prove an RGC fate.

3D differentiation approaches, particularly the use of retinal organoids, provide a promising solution to identify hPSC-RGCs using traditional markers such as BRN3 and ISL1. As retinal organoids are often described as “mini-retinas” that spatially and temporally follow the major stages of retinogenesis in a dish, these approaches can help to eliminate the possibility of non-retinal lineages in differentiating cultures at early stages(Capowski et al., 2019; Eldred et al., 2018; Fligor et al., 2018; Fligor et al., 2021; VanderWall et al., 2020). While the differentiation of retinal organoids provides a more definitive source for the study of RGCs, the RGC layer within these retinal organoids is often highly disorganized and previous studies have shown the loss of RGCs in long-term cultures of retinal organoids (Capowski et al., 2019; Fligor et al., 2020; Kobayashi et al., 2018; Rabesandratana et al., 2020). Thus, many studies focused upon RGCs have obtained organoids and then manually isolated RGCs following enzymatic dissociation, allowing for the prolonged growth and maturation of hPSC-RGCs in 2D cultures for extended periods of time (Sluch et al., 2017; VanderWall et al., 2020; VanderWall et al., 2019). hPSC-RGCs derived from 3D differentiation protocols offer significant advantages as they more closely mimic the major stages of human retinogenesis, providing a better in vitro developmental model. Furthermore, given their isolation from retinal organoids, the identification of hPSC-RGCs based upon the expression of molecular markers becomes more accurate. However, 3D culture systems typically require more technical skills to establish and maintain, and the percentage yield of hPSC-RGCs within retinal organoids is not always consistent across different differentiation experiments.

To further enhance the purity of RGCs derived from either 2D or 3D differentiation methods, previous studies have applied CRISPR/Cas9 genome editing approaches to establish an RGC reporter line that can also be used for the purification of RGCs from a heterogenous population, based on either expression of a fluorescent reporter or the expression of cell surface antigens including Thy1.2(Sluch et al., 2017). To date, these reporters have been broadly established in multiple cell lines and proven to be strongly applicable for studies related to RGC development and degeneration in diseases related to glaucoma, as well as studies focused on neuron-glia interactions as well as efforts for hPSC-RGC transplants for cell replacement(Croteau et al., 2022; Gomes et al., 2022; Patel et al., 2020; Risner et al., 2021; VanderWall et al., 2020; VanderWall et al., 2019; Zhang et al., 2021).

Although both 2D and 3D differentiation approaches have reliably generated large quantities of RGCs, they both require relatively longer periods of time to reach an RGC fate. As a result, some other studies have focused upon the genetic reprogramming of RGCs from other cell types, including either skin fibroblasts or undifferentiated hPSCs, through the forced expression of critical transcription factors(Liu et al., 2013; Pang et al., 2011). As hPSC-RGCs retain a relatively young, immature phenotype due to the process of iPSC reprogramming erasing epigenetic marks of aging(Huh et al., 2016; Lo Sardo et al., 2017; Mertens et al., 2018; Takahashi et al., 2007), stem cell-derived RGCs may be limited in their ability to properly serve as a tool for the study of adult onset RGC neurodegenerative diseases such as glaucoma. To further address age-related phenotypes, previous studies have directly converted fibroblasts into a variety of types of neurons through the delivery of defined neuron-associated transcription factors, bypassing the embryonic-like stage (Miskinyte et al., 2017; Pfisterer et al., 2011; Qin et al., 2020; Sepehrimanesh et al., 2021; Vierbuchen et al., 2010; Yang et al., 2020a). This method has been highly effective for modeling age-related phenotypes associated with multiple neurodegenerative diseases including Alzheimer’s diseases, tauopathies, and Huntington disease(Capano et al., 2022; Mertens et al., 2021; Victor et al., 2014), and recent studies have begun to focus on the ability to directly reprogram RGCs from fibroblasts following similar methods, albeit at relatively low efficiencies (Wang et al., 2020; Xiao et al., 2020). The direct reprogramming of RGCs provides a distinct advantage of often being much faster than traditional directed differentiation protocols, and may provide better opportunities for the study of age-related phenotypes. Nevertheless, direct conversion of cells into RGCs likely yields a heterogenous population of RGC-like cells and non-RGCs, depending on the efficiency of transduction as well as the process of reprogramming itself. Previous studies focused on the direct conversion of fibroblasts into neurons have also shown the partial reprogramming of some cells that no longer possess all of the features of the starting cell type, but also lack some of the features of the desired differentiated cell type. Given the fact that RGCs, like other neurons, are postmitotic, a significant obstacle exists in the purification of fully reprogrammed RGCs apart from other cell types, as well as the sufficient generation of appropriate numbers of fully reprogrammed RGCs for desired analyses. Further, a close characterization of directly reprogrammed RGCs will also be needed when following these approaches to ensure that bona fide RGCs are produced, due to the overlapping transcriptional signatures with other neural cell types (Xiao et al., 2020).

Essential characteristics for hPSC-RGCs

While many approaches may exist to derive hPSC-RGCs, an important next decision involves the level of maturation necessary to achieve the goals of the planned study, and whether or not the predicted phenotypes can be effectively modeled in vitro. hPSC-RGCs are intrinsically a developmental model, owing to their derivation from a pluripotent cell source. In this context, they provide substantial opportunities for the study of human retinogenesis, as well as the specification of RGCs and perhaps their numerous subtypes, as each of these areas of study do not necessarily require advanced stages of maturation. However, for the study of neurodegenerative features associated with optic neuropathies, including glaucoma, these are conditions that are more typically associated with advanced age (Mancino et al., 2018; Mirzaei et al., 2017). Thus, it is likely that hPSC-RGC models will need to possess numerous characteristics of advanced stages of maturation to best serve as an in vitro disease model. To that end, we propose 5 essential characteristics of hPSC-RGCs that should be taken into consideration when properly designing experimental approaches. Depending on the requirements of the experimental plan, differentiated hPSC-RGCs should (1) be derived from a retinal lineage, (2) express appropriate RGC-associated markers, (3) establish a complex neuronal morphology with extensive and elaborate neurite outgrowth, (4) become polarized during maturation to exhibit identifiable somatodendritic and axonal compartments, and (5) exhibit proper excitable properties by electrophysiological analyses (Figure 2).

Figure 2.

Proposed criteria outlining the characteristics for hPSC-RGCs essential for proper in vitro modeling.

First, and perhaps the easiest of the criteria to meet, is the definitive identification of differentiated cells as RGCs. As mentioned above, while it is easy in the retina to simply identify RGCs based upon their location and morphology, this is not always possible in a differentiating culture of hPSCs, particularly given their pluripotent nature in which these cells can become any cell type of the body. Thus, it is often necessary to identify RGCs based upon the expression of appropriate molecular markers. While RGCs have unique transcription factors for their identification within the retina, they share the expression of similar transcription factors within the central nervous system and thus, the identification of hPSC-RGCs based upon the expression of transcription factors individually may not be sufficient to ensure the definitive identification of hPSC-RGCs. To address this, presumptive RGCs should ideally be derived along a documented retinal lineage, such as through an earlier population of retinal progenitors or by the generation of retinal organoids, to help exclude the possibility that hPSCs differentiated into non-retinal cells(Fligor et al., 2018; Ohlemacher et al., 2016; Teotia et al., 2017a; VanderWall et al., 2020).

Additionally, to further ensure that differentiated cells are indeed RGCs, a panel of molecular markers should be used to characterize the hPSC-RGCs (Figure 3). While some of these markers may be found in other lineages, they often are not found among the same non-retinal cells. For example, BRN3 is an RGC marker that can be found in auditory or somatosensory neurons(Badea et al., 2012), ISL1 is an RGC marker that can be expressed in many other types of tissues, including motor neurons of the central nervous system(Liang et al., 2011), while RBPMS is yet another RGC marker that is found in several mesodermal cell types(Bartsch et al., 2022; Nakagaki-Silva et al., 2019). However, the combinatorial expression of these markers in an individual cell would help to ensure proper RGC specification. For instance, while neither the expression of the transcription factors BRN3 nor PAX6 individually are sufficient to identify an RGC from an hPSC source, co-expression of PAX6 and BRN3 have been suggested as a proposed set of combinatorial transcription factors for the identification of hPSC-RGCs (Figure 3B–E) (Xiao et al., 2020). Additionally, in the identification of these protein markers, one should especially take into consideration the proper localization of these proteins (e.g. transcription factor expression within the nucleus, not cytosol).

Figure 3. Schematic of RGC compartmentalization and current status of immunocytochemical detection.

(A) Properly mature RGCs exhibit compartmentalized structure with the expression of RGC-associated protein markers in appropriate locations. (B-I) Immunostaining characterizes the expression of RGC-associated markers. The commonly used BRN3:tdTomato:Thy1.2 reporter is highlighted. PAX6 and ISLET1 are RGC transcription factors localized within the nucleus. MAP2 is characteristic of somatodendritic structures and labels neurites of hPSC-RGCs. Scale bar: 50 μm. IPL, inner plexiform layer; GCL, ganglion cell layer; AIS, axon initial segment.

Among cell types of the retina, one of the most distinguishing characteristics of RGCs are their morphological features. RGCs typically have extensive dendritic arborizations extending in a stratified manner into the inner plexiform layer, while also extending long axons out of the eye to connect with the brain (Liu and Sanes, 2017; Troilo et al., 1996). While these precise features are not perfectly recapitulated in a culture dish, many studies have been able to demonstrate extensive neurite outgrowth that would further aid in the identification of hPSC-RGCs, as other retinal cell types do not develop equally elaborate and lengthy morphological features. These elaborate neurite outgrowths can be used not only as an identifying feature of RGCs compared to other retinal cell types, but previous studies have also suggested that the complexity of neurite outgrowth can serve as a parameter for overall RGC health(Agostinone et al., 2018). In hPSC models, recent studies have also identified robust neurite outgrowth in healthy RGCs, while disease models of glaucoma-associated neurodegeneration have demonstrated decreased neurite complexity compared to controls(Gomes et al., 2022; Teotia et al., 2019; VanderWall et al., 2020). Taken together, morphological features of hPSC-RGCs can not only serve as an important parameter for the identification of RGCs, but also as an essential readout of RGC health.

For some applications for disease modeling, particularly for a disease like glaucoma, the compartmentalization of RGCs may also be essential(Donato et al., 2019; Syc-Mazurek and Libby, 2019). It is well-established that the primary site of injury is along the initial part of the axon, and degeneration of RGC compartments occurs by different mechanisms in the axonal and somatodendritic regions(Syc-Mazurek and Libby, 2019). During the growth and maturation of RGCs, multiple neurites initially extend from the cell body, and eventually one of these neurites extends to become an axon while the rest become dendrites. To determine the extent of RGC maturation, this compartmentalization can be analyzed by staining for features such as MAP2 expression as a dendrite marker while markers including tau or neurofilament can serve as appropriate axonal markers. In immature RGCs, these proteins are not yet properly segregated and overlapping patterns of expression will exist, while more mature RGCs will exhibit more restricted and compartmentalized expression of these markers. The expression of synaptic proteins by RGCs can also serve as an indicator of RGC maturation, with more mature RGCs expressing higher levels of these proteins, although it has been suggested that hPSC-RGCs derived to date often lack the overall proper punctal localization of these synaptic proteins on either dendrites or axon terminals(Croteau et al., 2022; VanderWall et al., 2019; Xiao et al., 2020).

Finally, to fully recapitulate the properties of RGCs in an hPSC model, the functional properties of these cells should also be taken into consideration. As the projection neurons of the visual system that connect the eye with the brain, RGCs are excitatory neurons that convey visual information by transducing action potentials down their long axons(Chen and Chiao, 2014; You et al., 2013). Previous studies have demonstrated the ability of hPSC-RGCs to exhibit at least some appropriate excitable properties, including the firing of action potentials (both spontaneously and in response to a depolarizing current) as well as the conduction of both inward and outward ionic currents(Croteau et al., 2022; Gomes et al., 2022; Sluch et al., 2017; VanderWall et al., 2019). These excitable properties can also serve as an indicator of the overall health of the hPSC-RGCs when properly controlled, as previous studies have demonstrated that RGCs with a glaucoma-associated mutation were more easily excitable than parallel isogenic control RGCs(Gomes et al., 2022; VanderWall et al., 2020). While whole cell patch clamp provides the greatest cellular resolution, it is also technically challenging and low throughput, making it difficult for all labs to implement. As such, other approaches may also be suitable as a parameter for hPSC-RGC activity, including both calcium imaging or multielectrode array (MEA) recordings. In particular, MEA approaches can not only provide a measure of neuronal activity, but they can also represent network connectivity and allow for the analysis of the same cellular population over time(Fujii et al., 2016). However, as MEA approaches analyze a relatively large population of cells at a time, it is essential to maintain a high purity of hPSC-RGCs to eliminate variability in mixed populations.

Modeling RGC neurodegeneration with hPSCs and the necessity for proper controls

Glaucoma is the second leading cause of blindness worldwide with a current incidence of approximately 70 million individuals, and the loss of vision associated with the disease due to damage to RGCs. To date, the mechanisms underlying neurodegeneration in glaucoma are not completely understood, and limited options exist to protect or regenerate patient RGCs. Glaucoma itself is often considered a group of diseases with a variety of risk factors, such as increased age as well as the elevation of intraocular pressure (IOP) (Artero-Castro et al., 2020). However, as these conditions are difficult to properly mimic in an hPSC-RGC model, glaucoma-associated monogenic risk factors have become a powerful approach to study features of RGC neurodegeneration in a dish, especially with many of the genetic risk factors adequate to cause glaucoma under a normal range of IOP. Several groups have established patient-specific hPSC-RGCs with underlying genetic mutations, including Optineurin (OPTN, E50K), SIX6(H141N), and TBK1(duplication) mutations, all of which have revealed disease-related phenotypes(Ohlemacher et al., 2016; Teotia et al., 2017b; Tucker et al., 2014). In addition to glaucoma, hPSC-RGCs have also been used for the study of inherited optic neuropathies, including Leber’s hereditary optic neuropathy (LHON) (Danese et al., 2022; Nie et al., 2022; Wong et al., 2017; Yang et al., 2020b) and autosomal dominant optic atrophy (DOA)(Sladen et al., 2022). The use of patient-specific hPSCs with relevant gene variants provides an important tool for investigating molecular or cellular mechanisms underlying RGC neurodegeneration. Further, compared with animal models of glaucoma, hPSC-RGCs are an important complimentary model that can help to simplify the cellular complexity found within numerous cell types present in retinal tissue, and can allow for the study of cell autonomous features of disease states within RGCs themselves.

While hPSC-RGCs provide opportunities for the investigation of disease-associated phenotypes in a human cellular model, the variability present between individuals (and by extension, different lines of hPSCs) may confound analyses to identify differences between disease and control cell lines. In some cases, this variability may mask disease-related phenotypes while in other situations, phenotypes due to variability between individuals may be mistaken for disease-associated phenotypes. To overcome these issues, the use of gene editing strategies such as CRISPR/Cas9 is an essential part of any current hPSC disease modeling study, as these approaches can be applied to precisely insert or delete nucleotides in the genome, with the insertion of a disease related risk allele in the genome of otherwise healthy cell lines, or the correction of the disease-related allele when cells are derived from patient-specific hPSCs. In this capacity, CRISPR/Cas9 genome editing represents a powerful approach for disease modeling that allows for the specific analysis of the role of certain gene variants in any cell type, without confounding variables due to other differences in genomic backgrounds across cell lines (Musunuru, 2013). In hPSC-RGCs, CRISPR/Cas9 genome editing has been used successfully to model aspects of RGC neurodegeneration associated with glaucoma through the insertion of an OPTN(E50K) mutation in a wild-type hPSC line as well as the correction of this variant in a patient-specific hPSC line with the OPTN(E50K) mutation(VanderWall et al., 2020). Similar approaches have also been recently adopted for the study of dominant optic atrophy (DOA) through the gene correction of a patient-specific hPSC line with the OPA1(R445H) mutation (Sladen et al., 2021). Collectively, these studies have demonstrated that genome editing can minimize the effects of variability between hPSC lines and allows for the study any gene variants of interest without misinterpretation of other genomic differences between cell lines. Taken together, future studies of hPSC-RGCs for disease modeling will need to explore similar gene editing strategies to ensure that experiments are conducted with proper isogenic controls.

Future directions for advancing hPSC-RGC technologies

In the past several years, the use of hPSC-RGCs has provided numerous opportunities for the investigation of RGC development and disease. As recent studies have demonstrated a low degree of RGC conservation between rodents and primates (including humans) (Peng et al., 2019), many studies have become further reliant upon hPSC-RGCs as a compensatory human cellular model. However, despite the human cellular origins of hPSC-RGCs, current culture systems often do not allow for the complete in vitro maturation of hPSC-RGCs. This failure to fully mature is likely not due to an intrinsic inability to reach full maturation, but is likely due more to the environment in which these cells are found in a dish. As media formulations have improved over the past several years, studies have demonstrated increased physiological maturation of hPSC-RGCs(Ohlemacher et al., 2016; VanderWall et al., 2020). Additionally, the lack of other surrounding cell types, particularly neighboring supportive glia, likely also prevents the full maturation of hPSC-RGCs. Indeed, previous studies have demonstrated that the co-culture of hPSC-RGCs with astrocytes can improve and expedite the functional maturation of hPSC-RGCs(VanderWall et al., 2019). However, while these factors may present some limitations for current systems, the recognition of these limitations also provides new opportunities for the study and advancement of hPSC-RGC cell culture models that will further support hPSC-RGC maturation.

Among the existing issues for the use of hPSC-RGCs is the lack of proper compartmentalization of these cells, with defined somatodendritic and axonal compartments. During the growth of hPSC-RGCs, robust neurite outgrowth has been observed, yet the presence of clear, defined somatodendritic and axonal compartments has not yet been reported, indicating a degree of morphological immaturity. To overcome this barrier, recent studies have begun to induce proper polarization of hPSC-RGCs through the use of microfluidic platforms that can potentially promote better RGC compartmentalization(Fligor et al., 2021; Teotia et al., 2019). The use of microfluidic devices recruits axonal growth from hPSC-RGCs through microgrooves and further enables hPSC-RGCs polarization, with important implications for the study of axonal regeneration(Teotia et al., 2019). However, while this is an important step toward hPSC-RGC compartmentalization, whether hPSC-RGC neurites specify into distinct axonal and somatodendritic compartments in microfluidic devices has yet to be determined.

Another important consideration for hPSC-RGCs is the fact that in the dish, they do not necessarily interact with other neurons which may induce cell death due to the lack of connectivity and survival factors. In vivo, RGCs receive input from bipolar and amacrine cells, and send information to downstream targets largely within the lateral geniculate nucleus and superior colliculus, where RGC axons connect and receive retrograde neurotrophic factors that can support RGC survival and maturation(Crair and Mason, 2016; Erskine and Herrera, 2014). Previous studies have described increased RGC death in 3D retinal organoids in long term cultures(Capowski et al., 2019), yet this RGC loss can be largely attenuated through the formation of assembloids between retinal and brain organoids, creating a 3D environment that may help to better mimic RGC compartmentalization and communication with downstream targets(Fligor et al., 2021). Assembloid models will also provide a novel way to assess the role of the microenvironment in hPSC-RGC maturation, as well as how neighboring cells can adversely affect RGCs in disease states.

While certain aspects of RGC neurodegeneration may occur in a cell autonomous fashion, glial cells located in the nerve fiber layer within retina as well as distal optic nerve are known to interact with RGCs and contribute to RGC maturation as well as induce certain aspects of disease states. For example, microglia regulate RGC development including through the pruning of extraneous synapses, and they phagocytose RGCs that die off(Anderson et al., 2019). Under disease states, astrocytes and microglia are known to either become neuroprotective or neurotoxic to RGCs(Bordone et al., 2017; Gomes et al., 2022; Guttenplan et al., 2020; Zhao et al., 2021), suggesting that RGC survival is also regulated by non-cell autonomous responses. Therefore, when modeling RGC diseases, it is also important to consider the contribution of glial cells. Indeed, we have previously demonstrated that diseased astrocytes can promote the degeneration of otherwise healthy RGCs, while healthy astrocytes can rescue disease RGCs, highlighting the fact that astrocytes can greatly modulate RGC neurodegenerative features(Gomes et al., 2022). To further advance the maturation of hPSC-RGCs and create more physiologically relevant disease models, the use of hPSC-derived glial cells in combination with hPSC-RGCs will be an important avenue for investigation.

To further our abilities to study molecular aspects of RGC neurodegeneration, the use of CRSIPR/Cas9 gene editing will likely be instrumental. Indeed, previous studies have developed an efficient method to purify hPSC-RGCs from otherwise mixed populations, with the use of a combined tdTomato and mouse Thy1.2 antigen(Sluch et al., 2017). This hPSC reporter line has already been used in many studies for hPSC-RGCs(Croteau et al., 2022; Gomes et al., 2022; Patel et al., 2020; Risner et al., 2021; VanderWall et al., 2020; VanderWall et al., 2019; Zhang et al., 2021). Further, related studies have also used CRISPR/Cas9 gene editing to create disease models with paired isogenic controls, which limits variability between cell lines. In future studies, it is apparent that CRISPR/Cas9 genome editing technologies will become more commonplace, whether it is used to create new disease models, or perhaps tag subcellular structures or organelles with fluorescent proteins for subsequent analyses in disease states. Of interest, it may also be possible to establish unique reporter cell lines for the identification and analysis of RGC subtypes(Langer et al., 2018), as previous studies have suggested differential susceptibility of RGC subtypes to disease states(Daniel et al., 2018).

Concluding Remarks

Significant advances have already been made in the translational application of hPSC-RGCs for studies of retinal development, in vitro disease modeling, and cell replacement. However, in the future pursuit of RGC studies, it will be of paramount importance to carefully assess hPSC-RGC integrity, ensuring that these cells meet the highest of standards needed to ascertain unique disease-related phenotypes.