Age-related macular degeneration (AMD) is a progressive, degenerative disease of the macula, a small area at the center of the retina [1]. It is in adults older than 50 years in industrialized countries and its incidence has been rising as the life expectancy of the global population has increased [2]. There are two types of AMD: non-exudative (‘dry’) AMD and exudative (‘wet’) AMD. Non-exudative AMD and exudative AMD account for 90% and 10% patients diagnosed with AMD, respectively [3]. The clinical hallmark of non-exudative AMD is drusen, yellow deposits made up of lipoproteins, found in between the retinal pigment epithelium (RPE) and Bruch’s membrane (BrM) [4]. This non-exudative AMD is divided into early, intermediate, and late stages based on the size and the degree of drusen deposition. The early stage is recognized by having multiple small (< 63 µm) or more than one intermediate drusen (63 µm ~ 125 µm); the intermediate stage is characterized by many intermediate or one large drusen (> 125 µm); the late stage is characterized by geographic atrophy (GA) involves degeneration of the RPE, retina and the choriocapillaris with well-demarcated borders, resembling the map of a ‘continent’. Wet AMD is characterized by vascular growth from the choroid penetrating Bruch’s membrane, referred to as macular neovascularization (MNV), which accounts for 10% of total AMD patients [5]. The newly formed blood vessels leak fluid, causing visual impairment and blindness, which is mediated by drusen-mediated activation of vascular endothelial growth factor (VEGF) expression in the lesion [6]. Thus, anti-angiogenic therapies using anti-VEGF agents in the past 15 years have significantly reduced the blindness rates for those wet AMD patients with MNV [7]. Even though anti-VEGF is effective in exudative AMD, there is no treatment available to prevent or treat non-exudative (or ‘dry’) AMD initiated by drusen formation and leading to MA. Thus, the drusen formation and MA remain an area of significant unmet medical need [8].

We acknowledge that animal models have been widely used to study ocular biology and disease such as AMD [9]. Despite the contributions of the animal models, it is often difficult to isolate the relative contributions of complex biological factors and biophysical cues in the animal models, including inflammatory cytokines, reactive oxidative species (ROS), interstitial pressure, etc., thus these in vivo models are often not as helpful in identifying detailed mechanisms underlying multifactorial AMD pathogenesis [10]. By contrast, traditional cell cultures in two-dimensional (2D) dishes or transwell permit such identification, as they are a highly controllable model system, but they do not recapitulate the three-dimensional (3D) in vivo organization of these ocular tissues [11,12,13]. Therefore, there has been a clear, unmet need for 3D culture that reconstitutes human ocular structure to permit controlled experiments investigating drusen formation, deposition, and macular atrophy in AMD [9]. The purpose of the current review is to provide an update on the in vitro bioengineered methods to study AMD.

Main textBasic biology of age-related macular degeneration (AMD)

The retina is one of the highest oxygen consuming tissues in the body, making it one of the most metabolically active tissues (Fig. 1A). It therefore requires an efficient mechanism to maintain homeostasis as well as meet metabolic needs of the retinal neurons [14]. There are three major cellular types including vascular cells, glial cells, and neurons that comprise the retina. Similar to the central nervous system, these cells are intimately interacting to each other in the retina to form a functional unit called the retinal neurovascular unit (NVU) whose function is to deliver the metabolic fuels required for vision in a timely manner, and remove the metabolic wastes generated during the metabolism of vision [15]. The retinal NVU is composed of the photoreceptors, Bruch's membrane (BrM), the retinal pigment epithelium (RPE), choriocapillaris (CC), and deep capillary plexus (Fig. 1B) [16]. The photoreceptors are the group of cells that are converting light into biological signals that are sent to the brain. Because positively charged ions consistently enter unstimulated photoreceptors to keep dark the photoceptor disks continuously, retina is one of the most metabolically active parts in the human body [17]. The choriocapillaris (CC) is a fenestrated layer of capillaries that provides nutrients and removes waste products from the photoreceptors and the RPE. Bruch's membrane (BrM) is an extracellular matrix that lies between the CC and the RPE, which functionally serves as a biological barrier for transport of the nutrients and metabolites [18]. It consists of an inner collagenous layer (adjacent to the RPE), a middle elastic layer, and an outer collagenous layer (adjacent to the CC). The retinal pigment epithelium (RPE) is a monolayer of polarized hexagonal cells whose apices have multiple villous processes that are intimately related to the photoreceptor outer segment. By phagocytosing the outer segment discs that are shed daily, the RPE plays a critical role in retinoid recycling during phototransduction [19]. Finally, the lateral intercellular junctional complexes that join contiguous RPE cells form the outer blood retinal barrier [20, 21].

Fig. 1

Eye structure and AMD pathology. A Physiology structure of an entire eye. B Enlarged description of macular and Bruch’s membrane structure. C Pathology of dry AMD, including drusen deposition (yellow) and RPE elevation. D Pathology of 3 types of wet AMD. Type 1 macular neovascularization (MNV) is an ingrowth of vessels without outer retinal disruption. Type 2 MNV is the proliferation of new vessels with outer retinal disruption. Type 3 MNV is a downgrowth of vessels from the retinal vascular plexus toward the outer retina [22]

The AMD is a complex, chronic, multifactorial disease characterized by an aging macula in individuals with genetic mutations and environmental risk factors whose pathophysiology is poorly understood [23]. It appears that genetic mutations, particularly in the complement genes, play an important role in establishing the disease.A large genome-wide association study identified 34 loci that clustered in three main pathways in AMD progression (in ascending order of p-value): the complement system, extracellular matrix remodeling, and lipoprotein metabolism [24]. Risk factors of AMD include aging, smoking, higher body mass index, cardiovascular disease, high fat diet, etc., however, aging is the most consistent risk factor among them [25]. Aging causes choroidal thinning which is most likely due to a decreased choroidal vascular volume, secondary to the resultant decrease in choroidal blood perfusion. The severity of the choroidal perfusion deficits, particularly in the central macula, has been correlated with AMD severity and progression [26,27,28]. These changes in choroidal hemodynamics appear to play a crucial role in the progression of intermediate AMD to late AMD [29]. Meanwhile, aging also influences the transport function of BrM, by thickening and cross-linking the barrier. As the process goes on, the lipoproteins secreted from functional RPE cells would not be able to transport through the BrM and choroidal blood endothelium, causing the impaired egress [30]. Hence, the BrM accumulates with lipids, especially lipoproteins that contain apolipoprotein B and E, cholesterol and 7-ketocholesterol. As a result, the aging macula experiences the deposition of lipid rich extracellular deposits known as drusen and subretinal drusenoid deposits (SDD), causing dry AMD (Fig. 1C) [31]. The lipoprotein deposited between the RPE and BrM (the subRPE space) is called Drusen, and the deposition between the RPE and the photoreceptors (the subretinal space) is called SDD [18, 30, 32]. These deposits trigger the complement cascade, recruit the macrophage through the breaching BrM into sub RPE space, and activate the microglia in subretinal space, which further leads to local inflammatory responses. The complement system, a part of the innate immune system is unspecific, thus can lead to macular neovascularization (wet AMD) and macular atrophy (Fig. 1D). Patients with genetic polymorphisms of the complement genes are unable to turn off the complement cascade in the disease progression [33, 34].

Preclinical animal models and tissue-engineered approaches for in Vivo transplantation

In the past, there was a strong reliance on animal models to further our understanding of the mechanisms of AMD and to test new therapies. A major limitation in translating observations gained from animal models to human disease is the differences between the anatomy of these animals and humans [35]. For instance, the rodents, such as mice and rats, has a similar cone density in entire retina, hence don’t have anatomical macula [36]. Non-human primates are the only animals that has macular structure that is similar to human, showed spontaneously develop drusenoid lesions. However, non-human primates haven’t showed advanced AMD hallmarks, such as MA or MNV [37, 38]. Meanwhile, these are costly to maintain, difficult to manipulate genetically, and manifest a slow time course of disease progression compared with rodent models [39]. Despite the absence of macula, rodent models have been able to recreate several of the histological features of dry AMD including drusen-like extracellular deposits between RPE and BrM, choroid neovascularization, and subretinal microglia accumulation [40]. Notably, the microglia accumulation was not found in human eyes with AMD, suggesting another distinction between human and rodent model [41]. The laser-induced macular neovascularization (MNV) in both rodents and non-human primates have served as important models to study drug treatments in exudative AMD [42,43,44]. The optimal rodent model differs depending on which stage or type of AMD phenotype one wants to model. Although several animal models of AMD have been developed, none of them has been able to demonstrate the full human AMD spectrum [40]. Therefore, a lack of adequate animal models has hampered progress in our understanding of the mechanisms that lead to macular atrophy.

Prior to the introduction of pharmacological therapy with anti-VEGF drugs, surgical approaches were tried in eyes with exudative AMD. One example is macular translocation, which is a surgical procedure that moves the macular neuroretina to the area of RPE without neovascularization [45]. Although most eyes that underwent the macular translocation did not recover vision, some eyes experienced functional improvement once the macular photoreceptors were coupled to a bed of healthy RPE cells. This study, together with the fact that the RPE can heal after surgical removal of neovascular membranes, has served as a proof of concept that reconstitution of the macular function was possible by transplanting healthy RPE cell bed [46]. Transplantation of sheets of homologous fetal RPE, suspensions of autologous iris pigment epithelial cells or RPE cells have been tried but with little functional success [47, 48]. In contrast, an autologous free RPE-choroid graft was able to improve the visual acuity in eyes with wet AMD. However, serious intra-operative and post-operative complications, including retinal detachment, proliferative vitreoretinopathy, and recurrence of the MNV, were reported as side effects of these surgical procedures [49].

Ever since the observation in 1987 that transplanted rabbit RPE cells into the subretinal space are able to survive and phagocytose the photoreceptor outer segments [50], several groups have banked on cell therapy as a promising treatment alternative for AMD. RPE cell suspension was positioned adjacent to neural retina. Before injecting RPE cells, a buffer solution was slowly injected through retina into subretinal space to form bleb detachment. With continued RPE cell injection, cell suspension enters subretinal space [51] (Fig. 2A). While retinal allografts in pigs and primates are associated with a significant immune response, an autologous source of cells is preferred to avoid immunosuppressive drugs that are expensive, toxic, and life-long. Cells derived from the patient for whom they are intended would be the best possible immunologic match.

Fig. 2

Tissue-Engineered Approaches for In vivo Transplantation. A A schematic illustration of transplantation of RPE cell suspension into neural rerina. (a) Micropipette containing cell suspension is positioned adjacent to neural retina. (b) Small amount of buffered salt solution (BSS) at tip of the pipette is slowly injected through retina into subretinal space. As bleb detachment forms, patches of host RPE cells lift off with neural retina, creating areas of bare Bruch's membrane. (c) With continued injection, cell suspension enters subretinal space. (d) Reattachment of retina occurs within 24 to 48 h of the RPE cell transplantation. (e) SEM image of two attached RPE cells onto the bare BrM surface one hour after injection [50]. B Biodegradable 3D gelatin, chondroitin sulfate, and hyaluronic acid (GCH) scaffold for retina cell differentiation and transplant. (a) Scanning electron micrograph shows three faces of a block of scaffold. (b) A higher magnification of one face. Light micrograph (c) and scanning electron micrograph (d) show embryoid bodies on the scaffold one-day post-seeding. (e) One-week post-seeding, cultures were stained with DAPI (blue) to reveal cell nuclei. Three-dimensional reconstruction from confocal micrographs demonstrated that cells migrated the thickness of the scaffold. (f) After three weeks, cells homogenously populated most of the scaffold, but acellular areas were present [52]. C Transplantation of organoid-derived retina-like sheets with rd1 host retina. Schematic diagrams show three typical patterns of integration with rd1 host retina of the transplanted grafts. (a) Pattern 1: laminar interception. Graft INL was present between host INL and graft ONL. (b) Pattern 2: direct contact. The graft ONL was adjacent to the host INL. (c) Pattern 3: cell integration. The graft ONL structure was disorganized, similar to what was observed for cell transplantation. (d) A typical image of pattern 1. RHODOPSIN + photoreceptors from DD16 Nrl-GFP miPSC-derived retinal sheets migrate toward the host retina (white arrowhead). H, host; G, graft. (e) A typical image of pattern 2. DD16 Nrl-GFP miPSC-derived retinal sheets show structured ONL directly contacting host INL. H, host; G, graft. (f) A typical image of pattern 3. DD18 Rx-GFP mESC-derived retinal grafts show disorganized patterns similar to those observed for cell transplantation. H, host; G, graft. Scale bars, (B) a, 1000 μm; b&c, 500 μm; d, 100 μm; e&f, scale in μm. (C) d-f, 50 μm (D–F); I, 20 μm. Figures were adapted with permission from [51,52,53]

The main autologous cell sources for retinal implants include embryonic or fetal derived stem cells, adult tissue derived stem cells and induced pluripotent stem cells (iPSCs) [35]. The plasticity and unlimited capacity for self-renewal of embryonic stem cells need to be balanced against the potential safety issues including tumorigenesis, potential immune rejection, possible ethical issues, and the risk of differentiating into unwanted cell types. Adult tissues derived those stem cells may serve as a source of autologous multipotent cells but are not easy to come by and may contain the genetic cause of the disease. iPSCs are developed by harvesting adult somatic cells from the patient’s own tissues [54]. These somatic cells are reprogrammed to a pluripotent state where they can be expanded and differentiated into different retinal cell types including cones, rods, and retinal ganglion cells [55]. Since iPSCs are derived from the same patient there is a very low risk of rejection and ethical issues. Nevertheless, they can still harbor the diseased genes and develop tumors. CRISPR/Cas9 genome editing can be used to correct the mutations in the iPSCs before differentiating them into retinal cells [56].

Apart from direct injection of cell suspension, scientists have also tried bioscaffolds and organoid-derived retina-like cell sheets. Singh et al. (2018) developed a biodegradable porous scaffold composed of gelatin, chondroitin sulfate and hyaluronic acids (GCH) and seed human embryonic stem cells (hESCs) onto the GCH scaffold (Fig. 2B) [52]. They differentiated hESCs into retinal progenitor cells (RPCs) and implanted the differentiated cell sheets (including a variety of retina cell types such as photoreceptors, ganglion cells, and RPE cells, etc.) into the subretinal space of an AMD mouse model. This bioscaffold caused minimal immune reaction after implantation and survived for at least 12 weeks. Although they observed some cells migrated from the scaffold into the inner layer of retina, whether the migrated cells integrated functionally with the host retina is yet to be determined. Assawachananont et al. (2014) developed a 3D retinal tissue from embryonic stem cells and induced pluripotent stem cells (ESCs/iPSCs) [53]. They observed different integration and synaptic patterns after transplantation, and in the cell integration group, the 3D retinal tissue developed structured outer nuclear layer with complete inner and out segments (Fig. 2C).Although they proved the feasibility of retina-like cell sheet transplantation, they did not find the deterministic factors between different integration patterns. Despite the advances in direct cell transplantation with different cell sources and delivery methods (bioscaffolds), these approaches have not been able to significantly improve vision, long term cell survival, nor functionally integrate into the host retina [35], which requires future investigations for better translation into human trials.

Tissue-engineered approaches for in vitro modeling of AMD

Several in vitro models have been developed to study AMD mechanisms and identify potential therapeutics. They are divided into two major groups: two-dimensional (2D) models and three-dimensional (3D) models. 2D models can be used for researchers to study not only the single cell behaviors, but also cell–cell interactions using conditioned media and transwells [57, 58]. 2D models can mimic some aspects of the RPE and Bruch’s membrane, because these tissues exhibit a layered structure in vivo, which can be modeled in 2D culture system like transwells. Meanwhile, 2D cell culture models have also shown the drusen formation [59, 60]. However, 2D models could not contain photoreceptors (PR) and choriocapillaris structures, which are not easy to be recapitulated in 2D planar. The 3D models can take advantage of bio-scaffolds, organoids, and eye explants to coculture different cell types in AMD, including RPE, PR, endothelial cells, etc. [61, 62]. Moreover, with the help of microfluidic devices, fluid (aqueous humor) perfusion and mechanical force (blinking motion) can be introduced into 3D models, which may lead to a better biomimetic model for the AMD [63, 64].

Regarding cell sources, both cell lines and primary cells are available for modeling retina [65]. Cell lines are particularly suited for rapid and efficient screening of drug candidates for their toxicity and biocompatibility. Different cell lines that have been used for retinal research include the ARPE-19, RPE-1, D407 (RPE cell type), human WERI-RB1, Y79 retinoblastoma derived cell lines, 661 W (photoreceptor cell type), RGC-5 (ganglion cell type), and MIO-M1(glial Müller cell type) among others [66,67,68]. Although cell lines are widely accessible and easily cultured, up to 20% of currently used cell lines are contaminated or erroneously categorized, thus, validation of its origins and characteristic properties are of paramount importance prior to initiating work with them [69]. Meanwhile, the RPE cell lines were usually isolated from whole eyecups, which means all macular specific characteristics could be lost in those cell lines. Unlike cell lines, primary cells need to be harvested from humans or animals making them less available than cell lines [70]. However, they may better emulate the in vivo scenario than immortalized cell lines. Primary cell cultures used in retinal research include primary RPE cells, primary photoreceptors cells, primary retinal ganglion cells, primary microglia, and primary Müller cells [71].

2D tissue-engineered models for AMD study

The simplest 2D model would be culturing retinal cells on plastic dishes. For example, Johnson et al. (2011) developed an AMD model in 2D cell culture. They seeded primary human RPE cells onto a laminin-coated porous supports and cultured with “Miller medium” to trigger AMD pathogenesis. They have observed various pathological aspects of AMD, including drusen-like deposition, activation of the complement system, and deposition of terminal complement complexes in their model. This model highly recapitulates the AMD characteristics [59].

In 2D cultures, cell–cell interactions in AMD were studied with the help of the conditioned media or transwell. Incubating one cell type with the conditioned medium derived from another cell type enabled to study interactions between two cell types in 2D condition. For example, Nebel et al. (2017) incubated ARPE-19 cells, an RPE cell line, with human microglial cell conditioned media [11]. When microglia cells were activated by lipopolysaccharide (LPS) or Leu-Leu-O-Me (LLOMe), the resulting microglial cell conditioned media induced ARPE-19 cells be disorganized and accumulate lipid deposits, which indicates the role of microglia in RPE inflammation during AMD progression. However, their results have not been verified in other primary RPE cell lines or in other in vivo models. In another study, Leclaire et al. (2019) incubated microglia cells with lipofuscin (LP), a waste material that accumulated in human RPE in AMD [57]. They found that microglia phagocytosed LP and increased the expression of proinflammatory cytokines and VEGF, leading to inflammatory reaction and angiogenesis in AMD, respectively.

Using transwells, researchers could achieve coculture of two cell types and enable their direct communication. Meanwhile, RPE cells can form a polarized and functional monolayer on transwell [