Coculture techniques for modeling retinal development and disease, and enabling regenerative medicine

Abstract Stem cell‐derived retinal organoids offer the opportunity to cure retinal degeneration of wide‐ranging etiology either through the study of in vitro models or the generation of tissue for transplantation. However, despite much work in animals and several human pilot studies, satisfactory therapies have not been developed. Two major challenges for retinal regenerative medicine are (a) physical cell‐cell interactions, which are critical to graft function, are not formed and (b) the host environment does not provide suitable queues for development. Several strategies offer to improve the delivery, integration, maturation, and functionality of cell transplantation. These include minimally invasive delivery, biocompatible material vehicles, retinal cell sheets, and optogenetics. Optimizing several variables in animal models is practically difficult, limited by anatomical and disease pathology which is often different to humans, and faces regulatory and ethical challenges. High‐throughput methods are needed to experimentally optimize these variables. Retinal organoids will be important to the success of these models. In their current state, they do not incorporate a representative retinal pigment epithelium (RPE)‐photoreceptor interface nor vascular elements, which influence the neural retina phenotype directly and are known to be dysfunctional in common retinal diseases such as age‐related macular degeneration. Advanced coculture techniques, which emulate the RPE‐photoreceptor and RPE‐Bruch's‐choriocapillaris interactions, can incorporate disease‐specific, human retinal organoids and overcome these drawbacks. Herein, we review retinal coculture models of the neural retina, RPE, and choriocapillaris. We delineate the scientific need for such systems in the study of retinal organogenesis, disease modeling, and the optimization of regenerative cell therapies for retinal degeneration.


| Retinogenesis
The finding that retinal homeobox (Rx) + cells derived from ESCs and induced pluripotent stem cells (iPSCs) can self-aggregate in lowattachment culture to form retinal organoids with apical-basal polarity, showed that neuroectodermal cells of the developing optic cup possess an inherent self-organizing capacity. 1,2 This approach has created a new source of transplantable photoreceptor precursors and retinal sheets, which have been shown to integrate into degenerate retina forming synapses with host bipolar cells, responding to light and transmitting action potentials to the superior colliculus. [3][4][5][6][7] The process of generating retinal organoids, however, remains costly and produces variable results. 8 Furthermore, the transplantation of stem cell derived retinal tissue faces several challenges, which we will discuss.
Retinogenesis is orchestrated by a cascade of genetic masterswitches whose activation progressively narrows the developmental potential of precursors. 1,2,6,9 The developing RPE and neural retina are derived from a bi-layered invagination of the neuroectodermal optic vesicle which is surrounded by tissues of mesenchymal, neural crest, and surface ectodermal origin during development ( Figure 1A).
The space between the two layers of neuroectoderm is obliterated in humans between postconception days 37 and 42 and this event physically unites the precursors of the RPE and the photoreceptors. 10 It is after this event that the two layers differentiate into RPE and neural retina with the photoreceptor outer segments forming at the apical surface of the developing RPE and the RPE microvilli developing at the apical surface of the photoreceptor precursors. 10 Thus, functional and physical maturation occurs in the context of physical cell-cell contact. Outer segment formation occurs at approximately 112 days postconception and the RPE eventually phagocytoses photoreceptor outer segments to facilitate the recycling of visual pigments. Importantly, apposition of the developing RPE and neural retina precedes their maturation into functional retina. 10-12

| Synaptogenesis
Synaptogenesis in the developing human retina occurs sequentially from the developing fovea to the periphery, beginning with cones.
The cone-rich fovea provides high-acuity color vision. Development of cone-bipolar cell synapses begins in postconception weeks 8 to 10, before synaptic development of rods, and by week 13, all cones across the fovea express synaptic markers. However, core components of the synapse, the metabotropic glutamate receptor 6, and voltage-dependent calcium channel α1.4 are not detected until fetal week 22. 13 Rods are the predominant photoreceptor type and contact approximately 58 rod bipolar cells in the mature mouse retina. This convergence of rod afferent signals ensures high sensitivity of scotopic vision. 14 Interestingly, mature mouse bipolar cells appears to retain useful plasticity, with the ability to extend dendrites to new target rods, while reducing the number of synapses containing multiple ribbons, in order to prevent oversaturation with input signals. 15 Given the large number of cell types in the vertebrate retina, it is vital that postsynaptic target cells are correctly specified during development and remodeling. Specific pre-and postsynaptic membrane proteins linked by specific extracellular proteins that determine target type and are essential for correct synapse formation. 16

| Retinal organoids
Retinal organoids recapitulate the main events in the development of the mammalian retina both spatially and temporally and produce all seven cell types in a typical laminated fashion. 1 Naturally, they have been used to study mammalian retinal development. 17 Free-floating embryoid bodies are generated from suspensions of, or free-floating aggregates of, stem cells and then directed toward neural differentiation which is followed by eye field specification and maturation into retinal organoids. Recent research has focused on improving the efficiency of differentiation, developing disease models using patient iPSCs, testing gene therapies, staging the differentiation of retinal organoids, and single cell transcriptomics. 18 Transcriptomic studies have characterized the molecular architecture of developing retinal organoids and correlated it with developing and adult human retina, creating large repositories of temporal expression data. 19,20 Recent work has studied the substantial variability between retinal organoid lines. 8 Although the RPE-photoreceptor relationship is indispensable to photoreceptor signal transduction and therefore vision, this relationship is not reproduced in retinal organoids, which often but not always, contain a pole of RPE oriented away from neural retina ( Figure 1B) F I G U R E 1 Understanding retinogenesis is critical to developing in vitro models and transplantable tissue. A, A para-axial cross section is shown of the development of the optic vesicle and its invagination to form the bi-layered, neuroectodermal optic cup which later develops into the adult RPE and neural retina (postconception days shown). At approximately day 37, the two layers of the optic cup become apposed and the optic ventricle is obliterated. Neural crest and mesoderm condense to form the choroid and choriocapillaris. Hyaloid vessels form the blood supply to the inner retina and their vitreous branches regress in later life. B, A bright field photograph of human iPSC-derived retinal organoids at day 90 of differentiation. There is phase-bright neural retina with photoreceptors orientated outward (not shown) and a single dark pole of RPE. Scale shown. iPSC, induced pluripotent stem cell; RPE, retinal pigment epithelium tissues as occurs in vivo. Little work has been done on the coculture of retinal organoids with RPE, 21 although it has been reported that the maturation of retinal organoid photoreceptors is faster in the presence of primary RPE in contact coculture. 22 1.4 | Choriocapillaris, Bruch's membrane, RPE, and neural retina form a single functional unit in vivo Photoreceptors are supported by growth factors from the RPE, which also phagocytose their distal outer segments, which is critical in the recycling of phototransduction pigments ( Figure 2). Cone pigments are additionally recycled by Müller glia. 23 Oxygen and metabolites are delivered directly to the inner neural retina by the retinal vasculature, branches of the central retinal artery-these are directly visualized on fundoscopy. However, it is the choriocapillaris, which meets the high metabolic demand of the RPE and outer neural retina including the outer nuclear layer (ONL). 23 Distinct pathologies, which all have an endpoint of blindness, may have mechanisms which start in the choriocapillaris, RPE, or neural retina. At least 34 loci of genomic variants have been associated with age-related macular degeneration (AMD), with diverse functions including lipid metabolism, immune function, and extracellular matrix (ECM) proteins. 24 In one possible mechanism, AMD is believed to be initiated by complement dysregulation in the choriocapillaris, thereby leading to RPE loss and then photoreceptor degeneration. [25][26][27][28][29] Although the exact sequence of events is a matter of active research, it is clear that all elements of the retina are eventually affected and areas of RPE and choriocapillaris degeneration colocalize. 25,30 Similarly, in inherited chorioretinal dystrophies of the choroid and RPE, photoreceptors are sequentially lost. This occurs in choroideremia, an X-linked chorioretinal dystrophy caused by mutations in the gene Choroideremia (CHM), in which the earliest findings are loss of RPE seen on spectral domain-optical coherence tomography. This is followed by photoreceptor degeneration around areas of RPE loss and loss of choriocapillaris density. 31 Sandwiched between the RPE is a unique connective tissue sheet ( Figure 2). Bruch's membrane is a five-layered composite of RPE basement membrane, three elastin-rich layers, and the choriocapillaris basement membrane. 32 In adult life, it forms a barrier between the choriocapillaris and the RPE. The dense vessel network of the choriocapillaris is embedded within Bruch's membrane and unlike the larger choroidal vessels, develops in situ through a process called hemo-vasculogenesis. 33 One of the first pathological events in AMD is the formation of drusen and lipid deposition within Bruch's membrane, which holds prognostic significance for vision loss in AMD. 25,32,34 F I G U R E 2 Illustration of the physical relationship between the neural retina, RPE, Bruch's membrane, and choroid, which form a single functional unit. The RPE cells phagocytose photoreceptor outer segments. The dense capillary network of the choriocapillaris is the blood supply to the highly metabolically active RPE and photoreceptors. RPE, retinal pigment epithelium In this review, we will first discuss the major challenges facing cell transplant for retinal regeneration, with a focus on transplant of neural retina, although this discussion and the use of coculture models is also relevant to those interested in RPE transplant. We then will discuss techniques used in the coculture of neural retina, RPE, and choriocapillaris and how these can be used to build better in vitro models of the retina. Finally, we will discuss the current state-of-the-art in vitro models of the retina. Transplant of RPE patches has been investigated in phase I clinical trials and a phase I study using autologous iPSC-derived RPE to treat geographic atrophy associated with AMD is now in the recruitment phase, although the technique is far from being optimized in preclinical models. 35

| Transplant of dissociated photoreceptor precursors
The eye is an attractive target for regenerative cell therapies because of its relative immune privilege, accessibility, and the availability of advanced imaging technologies to noninvasively obtain highresolution images in vivo. 39,40 Initial studies using dissociated suspensions of photoreceptor precursors derived from early postnatal mice appeared to demonstrate maturation and functional integration in animal models of retinal degeneration. 41,42 However, it has become apparent that transfer of cytoplasmic material is the main mechanism of recovery of host retinal function (when no donor cells are present), regardless of whether primary cells or stem cell-derived precursors are transplanted. [43][44][45][46] Given that the majority of cells in the host ONL (the layer containing photoreceptor nuclei) expressing donor cell markers are a result of cytoplasmic exchange, proving that dissociated photoreceptor precursors mature and integrate into the host retina by forming synapses with bipolar cells has been difficult. 43,45,46 Others have reviewed this problem in detail. 47,48 Studies have shown some, albeit limited synaptic connection of photoreceptor precursors with host bipolar cells in rd1 mice where the ONL is absent and therefore cytoplasmic fusion not possible. These studies have however either not yet demonstrated visual improvement 6,46 or only modest visual improvement with survival of only a small fraction of the transplanted cells. 7 Singh et al found that only a third of eyes had surviving donor precursors at week 12 and that an average of 7.9% of donor cells survived. 7 In all cases, donor cells only partially matured without typical outer segment formation and no demonstrable interaction with the host RPE, which is critical to longterm function. 6,7,46 Recently, Garita-Hernandez et al have circumvented the problem posed by the poor outer segment development and interaction with RPE in transplanted dissociated cells by transfecting primary mouse photoreceptor precursors and human iPSC-derived photoreceptor precursors with bacterial opsins. 49 However, much work needs to be done in selecting opsins with sensitivity over a useful range of the human visual spectrum and the material exchange paradigm will continue to be problematic for human-human transplants. 49  while microaggregates could no longer be detected in the same retina. 57 Transplanted neonatal mouse microaggregates also show more maturation of outer segments than dissociated cells in the rd1 mouse. 55 Several experiments carried out in the 1990s provide data on the in vivo, long-term survival of foetal retinal sheets in humans. [51][52][53][54] The long-term survival of human iPSC retinal sheets in rat and primate models of retinal degeneration has been demonstrated (5 months and 2 years, respectively). 56 The limited integration of interneurons from transplanted planar organoids has also been demonstrated. 59 The available evidence therefore points to survival and likely functional engraftment of transplanted foetal/neonatal, ESC, and iPSC-derived retinal sheets, while the available data do not currently demonstrate long-term survival nor the exact functional abilities of integrated dissociated cells.  Figure 3B). [3][4][5]50,56,61 The graft may also lose its laminar structure and graft photoreceptors may migrate individually toward the host INL ( Figure 3C). 5 Another obstacle (which is at least partially a consequence of the above) is that From the bottom, RPE could be cultured to maturity on a suitable membrane. Retinal organoids could be hemi-sectioned to produce a flat sheet and placed photoreceptor-side down onto the RPE. Degenerate retina from the Pde6Brd1 mouse, which lacks an ONL could be placed onto a suitable carrier membrane and placed on top of the hemi-sectioned retinal organoid, thus recreating an in vitro subretinal space. For synapses to form between the Pde6Brd1 retina and the "transplanted" organoid, dendrites need to extend through the organoid inner retina (shown in gray scale). Ideally, the retinal organoid inner retinal layers, which may be redundant, could be ablated or dissected away before transplant. This experimental setup is inspired by Yanai et al. 60 INL, inner nuclear layer; ONL, outer nuclear layer; RPE, retinal pigment epithelium the formation of mature photoreceptors with complete outer segments has been highly variable in transplanted retinal sheets. The result of these three obstacles is that graft light-sensitivity has not been robust and can sometimes only be measured after a period of dark-adaptation. 3,4,56 In their study of iPSC-derived retinal sheets, Mandai et al supplied the grafted mice with intraperitoneal 9-cis retinol in the belief that the grafts would not be able to isomerize alltrans-retinol due to the lack of contact with the host RPE. 4 Optimizing the above factors is challenging owing to the difficulty In vitro models of retinal degeneration which recapitulate the neural retina-RPE interaction in disease states are needed. As we will now discuss, these models can be used to facilitate the faster optimization of regenerative cell therapies. To model the RPE more accurately in disease states, these models also require that other elements of the outer retina be present, namely Bruch's membrane and choroidal vasculature.

| COCULTURE OF NEURAL RETINA AND RPE
Over the last 30 years, several efforts have been made to coculture RPE with neural retina to study retinal development, in vitro differentiation of stem cells and to develop regenerative cell therapies.

| Models of differentiation and development
A key finding from experiments in which activation of a diphtheria toxin-expressing gene selectively ablated the RPE of developing chick embryos was that the embryonic RPE is required for the development of other ocular tissues: transgenic chicks expressing the toxin showed anophthalmia, with the presence of only the extra-ocular muscles and conjunctiva. 63 Other studies have since shown that RPE-derived trophic factors regulate mammalian neural retina development and that early RPE dysfunction during development alters the course of neural retina development, leading to neurodegeneration. 64 The transcription factor icrophthalmia-associated transcription factor (MITF) is highly expressed in RPE and other neural crest-derived tissues, and mutations in this protein cause a variety of phenotypes in mice, most notably microphthalmia and retinal degeneration. 65 Coculture models were created to study early retinal organogenesis, particularly the question of how apical-basal polarity is attained. On the other hand, the chick embryonic neural retina secretes factors which promote the maturation of the RPE. 73 These factors increase the transepithelial resistance of RPE by directing the expression and localization of claudins. 74 Interestingly, chick RPE (and human RPE) 75 can attain a high level of differentiation in isolated in vitro cultures, expressing many RPE-specific genes and showing a pigmented, cobblestone phenotype. However, full development of the RPE can only be attained with exposure to factors derived from the developing neural retina. 73 Taken together, these models point to a mutualistic relationship between developing RPE and neural retina based on promoting tissue growth through paracrine signaling which is possibly complemented by signals generated through physical interaction. gives approximately a 20-minute window before permanent damage (as measured by inability to restore a normal electroretinogram [ERG]). 88 Liberation of the neural retina from the hypoxic globe during this window is therefore important. 81 Anecdotally, avoiding killing test animals with CO 2 and ideally aiming to enucleate under deep anesthesia could improve neural retina survival ex vivo. There is also evidence that replenishing glucose immediately postenucleation could help to minimize ischemic damage. 88 Ideally, perfusion could be restored in ex vivo culture using microfluidic systems, allowing the constant replenishment of oxygen and nutrients. 89 To date, ex vivo retina strategies have used surrogate markers of retinal function such as ONL thickness 90 and immunohistochemical markers such as opsins and synaptic proteins 76,91 (quantified in terms of number of cells, area, or average intensity) or the presence of outer segment integrity. 76,86 More accurate data on functional enhancement of degenerate retinal explants could come from ERG assessments of retinal function which provide an assessment of the retinas ability to respond to light stimulation. 4  Coculture experiments have also shown that there is a bidirectional relationship between RPE and choriocapillaris. 94,96 Hamilton et al also showed that the transfer of a 4kD dextran fluorescein tracer was almost totally inhibited by a HUVEC-amnion-RPE trilayer but not a HUVEC-amnion-corneal epithelial cell trilayer, and the amnion-RPE bilayer alone was ineffective at inhibiting tracer transfer. 94 Interestingly, experiments looking at the trans-epithelial resistance in hESC-RPE monocultures found that they are less than monocultures of human foetal RPE isolated from 16-week-gestation foetuses, even after the transepithelial electrical resistance (TER) is allowed to stabilize in a medium, which encourages fetal and hESC-RPE maturation. Transcriptomic data showed that the expression of tight-junction and membrane transport genes was reduced in hESC-RPE. 97

| Disease models-AMD
The oBRB is formed by the tight junctions between RPE cells ( Figure 1). These cells are attached to their basement membrane which forms the innermost layer of Bruch's membrane, and whose outermost membrane is formed by the basement membrane of the choroidal endothelium. 32 The degree to which these three layers are interdependent is critical to accurate modeling of disease and regener-

| BIOCOMPATIBLE MATERIALS FOR COCULTURE
The development of engineered biomaterials designed to mimic the interphotoreceptor matrix (IPM), Bruch's membrane, and choriocapillaris basement membrane will allow the development of improved coculture models and regenerative treatments, with improved differentiation efficiency, reduced phenotypic drift, improved survival in long-term cultures, and improved lamination of in vitro or transplanted retina (reduced rosette formation). The use of biomaterials in in vitro models and retinal regeneration has been reviewed thoroughly. 89,104 Here, we will focus on their importance to coculture models of the retina.
It is now clear that specific physical properties of the extra-cellular environment are critical signals which direct embryonic and tissue stem cell differentiation. These include both cell-cell and cell-ECM contacts. [105][106][107][108][109][110][111] Physical queues alone can direct tissue-specific differentiation and can replace some of the factors required to maintain pluripotency in vitro. Substrate stiffness has been identified as a critical environmental queue in maintenance of pluripotency and lineage specification. 107,110,112,113 The presence of specific ECM proteins facilitates tethering to biological or synthetic substrates and thereby facilitates efficient mechano-transduction of substrate stiffness to intracellular signals. 109,111,112,114 Oligodendrocyte precursor cells are multipotent cells responsible for regeneration in the central nervous system. It has recently been shown that age-dependent loss of function in oligodendrocyte precursor cells can be recapitulated by culture on stiff hydrogels which mimic the aged connective tissue and this loss of function can be restored on soft hydrogels which mimic young brains. 110 Similarly, investigators have already shown that cultured RPE has reduced adhesion and survival on damaged or aged RPE. 115,116 Bruch's membrane is known to stiffen with age and cultured RPE show reduced phagocytic capacity on stiff vs soft scaffolds. 117,118 Resurfacing of aged Bruch's with ECM components can improve RPE survival in vitro, 119

| Biosynthetic Bruch's membrane
As RPE and choroid endothelium are known to interact synergistically, [92][93][94]96,98 development of choroid-RPE cocultures is an attractive option both for in vitro modeling and for transplantation. 122 The ideal rationally designed Bruch's membrane would have a permeability sufficient to allow metabolite diffusion between cell layers while maintaining structural integrity to physically separate choroidal endothelium and RPE. 123 It would allow cell tethering to its surface while having a stiffness which could be physically tuned to the optimum for RPE on one side and choriocapillaris on the other. It would biodegrade over a sufficient period to allow RPE and choriocapillaris to lay down basement membrane. 124 The specific complement of ECM proteins may also be important and can be recapitulated using decellularized ex vivo Bruch's 125 or decellularized choroidal ECM which can be revascularized with endothelial cells. 126 Ex vivo Bruch's membrane from individuals with outer retinal disease may serve in coculture models of these diseases. 127 Other physical properties which have been investigated are porosity, wettability, and ion diffusion capacity in a variety of synthetic and natural explants (reviewed in detail here) 128 and using a variety of surface coating to facilitate cell adhesion 122,123,129,130 ; however, these studies have not investigated stiffness or tethering properties. Unsurprisingly, several polymers and several surface coatings containing ECM proteins such as collagens and laminins were found to be suitable for culture of hESC-derived RPE, although the identification of specific conditions which optimize RPE function have been equivocal. 104,128,129 Other studies have focused on the protein composition of the RPE-secreted matrix but not on its physical properties. 125,131 Few studies have specifically examined the effect of scaffold stiffness on RPE culture and more work is required to identify the optimal tethering and stiffness properties which will facilitate cocultures models of the outer retina. 117,[131][132][133] Poly lactic-co-glycolic acid is a popular biomaterial for drug delivery and tissue engineering, and is currently being investigated as a vehicle for the subretinal delivery of iPSC-derived RPE to treat AMD. 36,124 Its advantages include the availability of fabrication methods, biocompatibility, biodegradability, and tunable mechanical properties. 134 For the transplant of retinal sheets, or perhaps sheets containing neural retina in addition to RPE and/or choroid, scaffolds which maintain cellular packing density and cellular orientation are desirable. The use of two-photon polymerization has been used to generate scaffolds with submicron features. Retinal progenitors nested within 25 μm pores oriented themselves perpendicular to the scaffold. 135

| Biosynthetic IPM
While photoreceptor soma are supported physically and physiologically by Müller cells, their outer segments are embedded in a highly specialized extra-cellular matrix, the IPM (Figure 1). The IPM is a highly complex hyaluronan matrix consisting of regularly repeating units of peanut agglutinin and wheat germ agglutinin which respectively sheath cone and rod outer segments. 136 Photoreceptor outer segments are also closely associated with hyalectans (Versican and Brevican) and Interphotoreceptor Matrix Proteoglycans (IMPG1 and IMPG2), which interact with the hyaluronan scaffold. 137 Photoreceptor outer segments and RPE do not form intercellular connections and therefore the IPM may provide an adhesive interface. 138 The IPM additionally facilitates transport of signaling molecules and metabolites between the RPE and photoreceptors. Its components undergo light-dependent conformational changes, underlying its importance in maintaining efficient phototransduction. 139 Major structural components, such as hyaluronan, are secreted by the RPE while other protein components such as interphotoreceptor retinoid binding protein are secreted by the photoreceptors. 138  feasible. Identifying the current barriers to effective stem cell-derived retinal transplant and modeling them individually will aid progress in the field of retinal regenerative therapies. As we have previously identified, these barriers include efficient synaptic integration of the transplant, interaction between photoreceptors and RPE and maturation of transplants to a fully functional form, among other issues. 144 One focus has been to "upgrade" ex vivo models and retinal organoids. By introducing microfluidics, it is possible to achieve a more physiological steady state in terms of removing metabolic by-products and introducing nutrients and oxygen. This could help to increase the longevity of retinal explants.
Retinal organoids suffer from a necrosis of centrally located cells likely due to the diffusion limit of oxygen. Recent characterization of cerebral cortical organoids has found that maturation into cortical subtypes is impaired by the activation of stress pathways. This activation is ameliorated, and subtype specification improved by transplantation of cortical organoids into the mouse cerebral cortex. Likewise, primary progenitors transplanted into cortical organoids adopt stress pathway activation. 145 This indicates it is the in vitro environment which impairs cortical organoid maturation.
F I G U R E 4 Legend on next page.
Microphysiological systems may be used to investigate local effects of cell therapy, for example, on glial activation, 146 to study the regeneration of the oBRB in the case of stem cell-derived RPE transplant, 100 to optimize regenerative therapies in disease-specific states (eg, wet AMD) 100 and to optimize synaptic integration of transplanted neural retina. 147 Su et al observed synaptogenesis between two populations of developing mouse retinal precursors in 4 μm guiding channels. 147 Compartmentalized experimental setups such as this could be used to optimize synaptogenesis between stem cell derived retina and host neurons.
Paek et al designed a polydimethylsiloxane (PDMS) culture chamber, which they filled with primary choroidal endothelium and choroidal fibroblasts laden in a fibrin/type 1 collagen hydrogel. They observed self-assembly of endothelial vascular networks which anastomosed with an input-output circuit and were thereby perfused with culture media ( Figure 4A). They cultured iPSC-derived RPE on this hydrogel and found greater basement membrane deposition, melanosome expression, and RPE-65 expression in their coculture model vs monocultures. 148 Interestingly, they were also able to create a vascularized solid tumor-on-a-chip model in which the self-assembled vascular networks anastomosed with the tumor blood supply. 148 Another strategy for in vitro vascular networks is shown in Figure 4B. 149 This raises the possibility of perfused retinal organoids in an adapted microphysiological system as has been achieved with liver organoids ( Figure   4C). 150 Furthermore, this model does not overcome the problem of necrosis at the organoid center due to the oxygen diffusion limit as the organoids are not vascularized. 152 7 | CONCLUSIONS Coculture models will advance our understanding of outer retinal physiology and disease. Furthermore, they will help optimize the delivery, integration, maturation, and functionality of stem cellderived retina for transplantation.
In our search, we have not identified a coculture system which incorporates neural retina, RPE, and choriocapillaris. As we have shown, all the techniques to achieve such a system are available. In the coculture systems we have reviewed, the neural retina is most frequently represented by retinal explants, which can be taken from animals with retinal degenerations. However, the neural retina component of coculture models may also be represented by hESC-or iPSC-retinal organoids. In this way, human retinal organogenesis and patient-specific disease models can be studied. Recapitulating embryogenesis by providing the correct developmental queues will direct stem cells toward developing in vitro models and tissues suitable for transplant. A significant barrier to successful cell F I G U R E 4 Perfusable in vitro generated vasculature may provide models of choriocapillaris. A, De novo generation of endothelial networks within an extracellular matrix-laden hydrogel. These networks anastomose with main channels (representing arteries and veins, respectively) and the endothelial network can be perfused with tissue culture medium. RPE is cultured over the vascular networks to generate a model of the RPE and choriocapillaris. Reprinted with permission from Paek et al. 148 Copyright 2019 American Chemical Society. B, Synthetic microvascular networks embedded within a hydrogel become lined with endothelium and allow remodeling of the surrounding hydrogel by parenchymal cells. In vitro perfusable tissue is thereby generated and can be linked to a host blood supply via surgical anastomoses (in this case, the rat femoral artery and vein). Adapted by permission from Springer Nature Customer Service Centre GmbH: Zhang et al. 149 Copyright 2016. C, Liver organ buds generated by recapitulation of mesenchymal condensation are perfused by the host vasculature when transplanted into the mouse cerebrum (days post-transplant shown). Adapted by permission from Springer Nature Customer Service Centre GmbH: Takebe et al. 150 Copyright 2013 Springer Nature. RPE, retinal pigment epithelium therapies may be providing immature transplants with the correct cues for integration into an adult diseased host environment. Advance in vitro coculture models will be invaluable tools in solving this problem.
In combination with retinal organoid technology, these strategies will lead to realistic, disease-specific models of ocular diseases. Using these models, we will be able to create rationally designed stem cell therapies for the retina and iteratively improve these therapies before moving to in vivo models and human trials.

AUTHOR CONTRIBUTIONS
A.E.G.: collection and/or assembly of data, data analysis and interpretation, manuscript writing; M.L., D.H.S.: conception and design, contributed to manuscript writing, final approval of manuscript.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no new data were created or analyzed in this study.