Interaction Between Mesenchymal Stem Cells and Retinal Degenerative Microenvironment

Retinal degenerative diseases (RDDs) are a group of diseases contributing to irreversible vision loss with yet limited therapies. Stem cell-based therapy is a promising novel therapeutic approach in RDD treatment. Mesenchymal stromal/stem cells (MSCs) have emerged as a leading cell source due to their neurotrophic and immunomodulatory capabilities, limited ethical concerns, and low risk of tumor formation. Several pre-clinical studies have shown that MSCs have the potential to delay retinal degeneration, and recent clinical trials have demonstrated promising safety profiles for the application of MSCs in retinal disease. However, some of the clinical-stage MSC therapies have been unable to meet primary efficacy end points, and severe side effects were reported in some retinal “stem cell” clinics. In this review, we provide an update of the interaction between MSCs and the RDD microenvironment and discuss how to balance the therapeutic potential and safety concerns of MSCs' ocular application.

While there are currently no effective treatments for RDDs as the human retina has no regenerative ability, new approaches-such as gene therapy, neuroprotection, anti-VEGF (vascular endothelium growth factors), and stem cell therapy-have the potential to delay vision loss or even to restore vision . Gene therapy can cure certain types of RDDs caused by single recessive gene defects, if applied in the early stages of the disease process while the photoreceptors are still alive (Scholl et al., 2016). Neuroprotective reagents are believed to be the best method to treat progressive photoreceptor cell degradation; however, they are unable to regenerate previously lost retinal neurons (Oswald and Baranov, 2018). Anti-VEGF therapy is the standard of care for neovascular AMD; it is also used to treat DR and many other retinal neovascular conditions. However, it has no effect on dry AMD and non-proliferative DR (Ferrara and Adamis, 2016).
In contrast to these treatments, cell-based therapy presents an opportunity for preserving or restoring vision in advanced stages of retinal degeneration and can be utilized even when photoreceptor cell degeneration has already occurred (Scholl et al., 2016). There are numerous pre-clinical and clinical trials regarding the application of several types of cells-including neural stem cells, mesenchymal stromal/stem cells, embryonic stem cells, and induced pluripotent stem cells-in RDDs (Li et al., 2016b;Ding et al., 2017;Enzmann et al., 2017;Mclelland et al., 2018;Jin et al., 2019;Shen, 2020). Stem cells have the capacity for self-renewal and are able to differentiate into multiple cell types (Ludwig et al., 2019). Degenerated retinal cells can benefit from stem cell transplantation, which can act as vehicles for drug delivery, immune modulators, or sources for direct tissue regeneration (Aharony et al., 2017).
Mesenchymal stem or stromal cells (MSCs) are multipotent cells isolated from a variety of tissues, such as adult bone marrow (bone marrow-derived MSC, BM-MSC), adipose tissue (adiposemesenchymal derived stem cells, ASC), and dental pulp (dental pulp stem cells, DPSC) (Mead et al., 2013. They can also be isolated from neonatal tissues and fluids, such as umbilical cord (umbilical cord-derived MSCs, UC-MSC), Wharton's jelly MSCs (WJ-MSCs), amniotic membrane, amniotic fluid, and placenta (Ding et al., 2011). Although MSCs are derived from different tissues, they share some common features. In 2006, the International Society for Cellular Therapy established the minimal criteria for designating a cell as MSC (Dominici et al., 2006). MSCs are able to adhere to plastic in standard culture conditions and have the potential to differentiate into adipocytes, chondroblasts, and osteoblasts. In general, they do not express hematopoietic and endothelial cell markers, such as CD34 CD45, CD11b, CD11c, CD14, CD19, CD79α, CD86, and HLA class II molecules, but they do phenotypically express a distinct set of cell surface markers for CD105, CD90, CD44, CD9, and CD73 (Dominici et al., 2006;Marino et al., 2019;Naji et al., 2019). The low immunogenicity (low levels of HLA class I and absence of HLA class II expression) makes MSCs a good candidate for cell transplantation (Ryan et al., 2005).
In addition to their wide distribution and ease to harvest, MSCs are also known to possess minimal susceptibility to malignant transformation and are capable of avoiding immune rejection (Ding et al., 2017). MSCs have therapeutic efficacy to promote regeneration of multiple tissues and cells, including central nervous system (CNS) neurons (Nakano et al., 2016). Thus, allogeneic or autologous cell transplantation of MSCs shows promises for potential therapeutic applications in RDDs. Indeed, several pre-clinical trials of MSCs in the treatment of rodent RDDs (such as streptozotocin or STZ-induced diabetic rodent models, rodent retinal degeneration models, and rodent glaucoma and retinal ischemia models) generated encouraging results (Lund et al., 2007;Guan et al., 2013;Tzameret et al., 2014;Ezquer et al., 2016;Mead et al., 2016;Roth et al., 2016). More than 40% of clinical trials of stem cell therapy for retinal diseases are also using bone marrow or umbilical cord-derived stem cells (Park et al., 2017;Shen, 2020).
However, there are several reports that RDD patients lost their vision after receiving intraocular injection of autologous bone marrow/adipose tissue-derived stem cells (Leung et al., 2016;Kuriyan et al., 2017;Rong et al., 2018;Khine et al., 2020). These devastating outcomes are primarily due to rhegmatogenous retinal detachments with severe proliferative vitreoretinopathy (PVR). The subretinal delivery of human umbilical tissue-derived cells (palucorcel) into eyes with geographic atrophy of AMD patients was also associated with a high rate of retinal perforation and detachment (Ho et al., 2017). From a scientific perspective, a probable cause is the trans-differentiation of injected MSCs into myofibroblast-like cells, which can induce fibrosis or PVR. Another possible mechanism is the toxicity to the retina or optic nerve caused by the injected material, which may have included enzymes used in the preparation process (such as trypsin). Periocular injection of autologous bone marrow stem cells to treat RP has also been reported to induce central retinal artery occlusion and vision loss (Boudreault et al., 2016). These severe complications call for a better understanding of the interaction between the donor MSCs and the degenerative host retinal environment. It is also imperative to have stringent quality control of MSC preparation and careful surgical procedures for ocular or intraocular delivery of MSCs (Apatoff et al., 2018;Singh et al., 2020). The aim of this brief review is to provide an update on the interaction between MSCs and the RDD microenvironment and to discuss methods by which to balance the therapeutic potential and safety concerns of MSCs ocular application.

THE INTERACTION BETWEEN MSCs AND THE RETINAL ENVIRONMENT
The communication between MSCs and host cells is mainly mediated by paracrine activity, tunneling nanotubes (TNTs), and extracellular vesicles (such as exosomes and microvesicles) that contain reparative molecules (Spees et al., 2016;Holan et al., 2019). The secretory activity of MSCs is strongly influenced by the retinal environment (Holan et al., 2019). Collectively, these mechanisms confer MSCs both trophic and immunomodulatory properties, which are essential for their application for RDDs (Figure 1).

MSCs Promote Retinal Cell Survival or RPE Phagocytic Function
MSCs can protect retinal cells by secreting growth factors, which can be classified into two categories (Murray et al., 2014). In the first category are factors promoting cell proliferation, including transforming growth factor-alpha (TGF-α), TGF-β, hepatocyte growth factor (HGF), epidermal growth factor (EGF), and fibroblast growth factor-4 (FGF-4); the second category consists of factors enhancing angiogenesis, including VEGF, interleukin-8 (IL-8), and insulin-like growth factor-1 (IGF-1) (Murphy et al., 2013;Kocan et al., 2017). Under stress conditions, such as hypoxia and growth factor deprivation, MSCs can increase the secretion of these factors (Anderson et al., 2016). The trophic effects of MSCs on retinal ganglion cells (RGCs), photoreceptors, RPE cells, and multiple cell types in diabetic retinopathy have been observed in several pre-clinical studies ( Figure 1A).

A-Retinal Ganglion Cells
Intravitreal injection of BM-MSCs promoted the survival of RGCs for at least four weeks by secretion of neurotrophic factors in an ischemia and reperfusion rat model (Li et al., 2009) and increased the overall RGC axon survival and reduced the RGC axon loss rate in a laser-induced ocular hypertensive glaucoma rat model (Johnson et al., 2010). However, intravenously injected MSCs did not migrate to the injured eyes and had no effects on optic nerve damages (Johnson et al., 2010). Intravitreal transplants of DPSCs promoted significant RGC survival and axon regeneration in a surgically induced optic nerve crush injury rat model, where the effects were mediated by neurotrophins and abolished after TrK receptor blockade (Mead et al., 2013).
These results were also verified by a coculture study of the porcine retinal cells and human MSCs, separated in a Transwell system, which suggested that MSCs produce brain-derived neurotrophic factor (BDNF) and ciliary neurotrophic factor (CNTF) (Labrador-Velandia et al., 2019).
Interestingly, overexpression of neurotrophin in MSCs can enhance these protective effects. Intravitreal transplanted MSCs engineered to express BDNF and green fluorescent protein (GFP) survived in rat eyes with chronic hypertension and protected retinal and optic nerve function and structure (Harper et al., 2011). Similarly, intravitreally injected MSCs engineered to overexpress nerve growth factor (NGF) survived and promoted RGC survival and regeneration in a NGF-dependent manner (Levkovitch-Verbin et al., 2010).
There is evidence that exosomes mediate the effect of MSCs on damaged RGCs (Mathew et al., 2019). Exosomes are membraneenclosed extracellular vesicles that contain mRNA, microRNA, and proteins including neurotrophins. MSC-derived exosomes can rapidly reach RGCs to supply them with neurotrophins and suppress their cell death (Mead and Tomarev, 2017). TNT-mediated mitochondria transfer from MSCs to RGCs is another mechanism to explain the protective effects. TNTs are communicating intercellular transport networks formed between many cells (Rustom et al., 2004). The TNTs are tubular structures, with diameters varying between 50 and 1,500 µm, which can span hundreds of microns between two cells. These structures allow immediate intercellular communication. Moreover, the fibers formed from the cytoskeleton allow the transfer of cytosolic material. Mitochondria can be transferred from MSCs to injured target cells to restore their function (Spees et al., 2006;Jiang et al., 2016;Vignais et al., 2017). In an inflammation-related RGC degeneration mouse model (NADH dehydrogenase ubiquinone Fe-S protein 4, or Ndufs4 knockout mice), intravitreally injected human iPSC-MSCs could transfer functional mitochondria across the inner limiting membrane to RGCs and restore the RGC function against mitochondrial dysfunction-induced inflammation (Jiang et al., 2019). While the results from these pre-clinical studies support the usage of MSCs in the treatment of glaucoma, there are currently no reports regarding their usage in clinical trials (Harrell et al., 2019).

B-Photoreceptors
Subretinal transplanted mouse GFP-labeled BM-MSCs into RP model Rho −/− mice integrated into the retinal pigment epithelium layer and protected the photoreceptor from death (Arnhold et al., 2007). Similarly, subretinal transplantation (trans-scleral and trans-choroidal approach) of mouse BM-MSCs into the Royal College of Surgeons (RCS) rat delayed retinal degeneration and preserved retinal function (Inoue et al., 2007). It has been demonstrated in vitro that the conditioned medium of the MSCs delays photoreceptor cell apoptosis, suggesting that secreted factor(s) from MSCs promote photoreceptor cell survival (Inoue et al., 2007). Subretinal or intravitreally injected human BM-MSCs into RCS rat can delay photoreceptor death for about 12-20 weeks (Tzameret et al., 2014). Subretinal transplantation of rat MSCs or engineered erythropoietin (EPO)expression rat MSCs into a sodium iodate (SI)-induced rat model of retinal degeneration protected RPE and retinal neurons; EPO expression MSCs had an even greater effect (Guan et al., 2013). Subretinal transplantation of human adipose derived stem cell (hADSCs) (Li et al., 2016a) and human periodontal ligamentderived stem cells (hPDLSCs) (Huang et al., 2017) also protected the photoreceptors in RCS rats. It has been suggested that hADSCs can suppress the expressions of Bax, Bak, and Caspase 3 and produce VEGF, HGF, and pigment epithelium-derived factor (PEDF), all of which may contribute to their neuroprotective effects (Li et al., 2016a).

C-RPE Cell Function
Subretinal injection of human umbilical tissue-derived cells (hUTCs) in the RCS rat model of retinal degeneration can preserve photoreceptors and visual function (Lund et al., 2007), as hUTCs can rescue the phagocytic dysfunction in RCS RPE cells in vitro by secreting several trophic factors-including BDNF, HGF, and GDNF-as well as opsonizing bridge molecules MFG-E8, Gas6, TSP-1, and TSP-2 (Cao et al., 2016). These trophic factors-derived from the conditioned medium of hUTCs-are also beneficial to the phagocytic function of human RPE cells isolated from the post-mortem eyes of AMD-affected subjects (Inana et al., 2018). In a phase 2b clinic trial, while hUTCs (palucorcel) were delivered successfully to the targeted subretinal space for most participants, improvements in GA (geographic atrophy of AMD) area or visual acuity were not demonstrated; thus, no apparent therapeutic effect was observed (Heier et al., 2020).

D-Multiple Cell Types in Diabetic Retinopathy
Intravitreal injection of human ASCs or cytokine-primed ASCs conditioned media (ASC-CM) into STZ-induced diabetic athymic nude rats (Rajashekhar et al., 2014) and diabetic Ins2 Akita mice (Elshaer et al., 2018), improved ERG b-wave amplitudes and vascular leakage, and reduced apoptotic cells around the retinal vessels. ASC-CM (but not ASCs itself) can improve retinal gliosis, DR-related gene expression profile, and mouse visual acuity. ASC-CM had high levels of anti-inflammatory proteins, including indoleamine 2, 3-dioxygenase 1 (IDO-1), IDO-2, and TSG-6 (Elshaer et al., 2018). Intravitreally injected ASCs also reduced oxidative damage and increased the intraocular levels of several potent neurotrophic factors-including NGF, bFGF, and GDNF-in a diabetic mouse model, thus preventing RGC loss (Ezquer et al., 2016). Interestingly, intravitreally injected BM-MSCs were found to integrate into the inner retina, differentiate into retinal glial cells, and improve ERG amplitude, thereby protecting vision in a STZ-induced mouse model (Çerman et al., 2016). Excitingly, intravenously administrated autologous BM-MSCs were found to be beneficial in non-proliferative DR (NPDR) patients, showing significant improvements in macular thickness and best-corrected visual acuity (BCVA) from baseline (Gu et al., 2018).

MSCs Regulate Retinal Inflammation and Immune Responses
When exposed to an inflammatory environment, MSCs can modulate local and systemic, innate, and adaptive immune responses through the release of various mediators, which include cytokines, chemokines, and some metabolites, such as IDO, IL-6, PGE2, and TGF-β1. While immunosuppression is mainly mediated by IDO in human MSCs, it is mediated by inducible nitric oxide synthase (iNOS) in mouse MSCs (Ren et al., 2009). Interestingly, apoptotic MSCs also have some immunosuppressive functions in vivo. MSCs can be actively induced to undergo perforin-dependent apoptosis by recipient cytotoxic cells; apoptotic MSCs are engulfed by host phagocytic cells that, by producing IDO, became the ultimate effectors of immunosuppression (Galleu et al., 2017). Thus, host phagocytes are crucial for generating an immunosuppressive environment by clearing apoptotic MSCs (De Witte et al., 2018). MSCs can suppress the proliferation of T cells, B cells, and natural killer cells, inhibit the differentiation and maturation of monocytederived dendritic cells, and promote the generation of regulatory T cells (Melief et al., 2013). TNT-mediated mitochondria transfer also plays a role in immunomodulation. Mitochondrial transfer from MSCs to CD4 + T cells increases the expression of factors involved in T-cell activation and differentiation, including FoxP3, IL2Ra, CTLA4, and TGF-β1, leading to an increase in the suppressive regulatory T cell population (Court et al., 2020). Both pro-and anti-inflammatory effects can be observed in the application of MSCs to the retina (Figure 1B).

A-MSCs Suppress Retinal Inflammation
Inflammation is a major mechanism in the pathogenesis of RDDs, including DR, AMD, and RP (Akhtar-Schafer et al., 2018;Arroba et al., 2018). The upregulated pro-inflammatory mediators (such as cytokines) induce breakdown of the bloodretinal barrier (BRB) and immune cell infiltration (such as macrophages and microglia), which form an inflammatory retinal microenvironment (Kauppinen et al., 2016;Mesquida et al., 2019). MSCs can regulate this inflammatory retinal environment through their immunomodulation properties (Shi et al., 2018). For instance, intravitreal injection of proinflammatory cytokines (IL-1β, TNF-α, and IFN-γ) into mice induces retinal edema, vessel dilatation, and microglial accumulation in the retina, resulting in a mouse model of retinal inflammation (Mugisho et al., 2018). Intravitreal injection of BM-MSCs into this mouse model reduced the retinal expression of pro-inflammatory molecules such as IL-1α, IL-6, iNOS, TNF-α, and VEGF and reduced macrophage infiltration . MSC-derived exosomes can also suppress laser injury-induced mouse retinal inflammation; MCP-1 is likely the major cytokine mediating this effect (Yu et al., 2016). TNT-mediated mitochondria transfer also plays a role in the immunomodulatory effect in the Ndufs4 knockout mouse model to reduce inflammation and RGC degeneration (Jiang et al., 2019).

B-MSCs Promote Retinal Inflammation
The immunomodulatory capabilities of MSCs are not constitutive but rather are licensed by inflammatory cytokines (Boland et al., 2018); the effects of MSCs may vary depending on the local inflammatory environment within the targeted tissues Shi et al., 2018). Indeed, intraocular injection of MSCs generally induces retinal inflammation in wild-type animals. For instance, Wharton's jelly MSCs (hWJMSCs), injected intravitreally into a rat model of optic nerve axotomyinduced RGC degeneration, secrete anti-inflammatory molecules and trophic factors, but in naive retinas, they instead induce a massive migration of microglial/macrophages from the choroid to the inner retina, disrupting the retinal architecture-a typical retinal inflammatory response (Millán-Rivero et al., 2018).

C-NF-κB/Rap1 Inhibition and Pre-licensing
The method by which to simultaneously enhance the antiinflammatory properties and inhibit the pro-inflammatory capabilities of MSCs is a pertinent issue to be addressed for the therapeutic use of MSCs in the retina. There are two possible strategies (Figure 1B). The first one is based on the NF-κB signaling pathway, a pivotal mediator of inflammatory responses . NF-κB mediates cytokine/growth factor secretion by MSCs (Mutt et al., 2012); thus, inhibition of NF-κB pathways may suppress the pro-inflammatory capability of MSCs. However, as this pathway is crucial in maintaining normal host defense and generating innate immune responses, complete blockade of NF-κB activity is not feasible (Poon et al., 2015;Liu et al., 2018). Rap1, a member of the telomeric shelterin complex, is a novel modulator involved in the NF-κB pathway (Teo et al., 2010). Selective inhibition of Rap1 in BM-MSCs can decrease NF-κB sensitivity to pro-inflammatory cytokines and enhance MSC-based therapeutic efficacy for myocardial infarction (Zhang et al., 2015). It would be interesting to test if Rap1 suppression can improve the effects of MSCs application in the retina.
Another strategy is based on the fact that the inflammatory environment can actually enhance the immunosuppression function of MSCs (Naji et al., 2019). MSCs can be polarized to pro-or anti-inflammatory phenotypes by pre-conditioning with cytokines, including IFN-γ, TNF-α, or IL-17 (Hemeda et al., 2010), and by signaling through Toll-like receptors (Mastri et al., 2012). Similarly, pre-stimulation of MSCs with anti-inflammatory factors such as TGF-β reversed the immunosuppressive effect of MSCs and conferred a proinflammatory phenotype . It has been suggested that in vitro pre-stimulation of MSCs with appropriate pro-inflammatory factors (pre-licensing) may obtain optimal therapeutic effects in vivo (Boland et al., 2018;Naji et al., 2019). IFN-γ is the most commonly used cytokine for pre-licensing or priming. While IFN-γ pre-licensing enhances IDO expression of cryopreserved human MSCs in vitro, surprisingly these prelicensed MSCs lose effectiveness in vivo; they rescued fewer RGCs than either fresh or unlicensed cryopreserved MSCs in a mouse model of retinal ischemia/reperfusion injury (Burand et al., 2017). While MSCs licensed with IFN-γ are known to increase expression of immunosuppressive factors, the expression of MHC-I and MHC-II molecules was also enhanced in the surface of these cells, which may induce strong immune rejection (Ankrum et al., 2014a). Further investigations are needed to optimize the pre-licensing procedures (Boyt et al., 2020).

D-Immune Rejection and MSC Transplantation
As most clinical applications of MSCs in the retina are allogeneic transplantations, immune rejection is an issue that needs to be considered. The vitreous cavity and subretinal space are advantageous for stem cell transplantation as they are immune privileged (Mead et al., 2015). The low immunogenicity and immunosuppressive property of MSCs also reduce the chance of immune rejection (Li et al., 2015).
Indeed, several pre-clinical studies injecting human MSCs into the eyes of rodent disease models (xenotransplantation) without using immunosuppressant have not observed obvious rejection (Lund et al., 2007;Tzameret et al., 2014;Li et al., 2016a;Elshaer et al., 2018). Subretinally administered human adult bone marrow-derived somatic cells (hABM-SCs) can achieve similar therapeutic benefits to protect the rods with or without cyclosporine A in the RCS rats (Lu et al., 2010). Interestingly, these effects do not require the cells to survive for a long period of time (Lu et al., 2010). This observed therapeutic benefit in the absence of long-term survival of transplanted cells is consistent with the so-called hit-and-run mechanism, mediated by the production of exosomes or secretion of trophic and immunomodulatory factors during the initial days following MSC injection (Von Bahr et al., 2012;Ankrum et al., 2014b) and may also be related to the immunomodulation effects mediated by apoptotic MSCs (Galleu et al., 2017).
However, immune privilege of the subretinal space and vitreous is determined by the retinal microenvironment and the integrity of BRB, which may be disrupted by RDDs or surgical procedures (Jiang et al., 1993). It has also been demonstrated that MSCs, like all somatic tissues, express MHC class I molecules constitutively and have the ability to express MHC class II when exposed to inflammatory cues such as INF-γ (Galipeau and Sensébé, 2018). Indeed, several studies used immunosuppressive drugs after subretinal transplantation of MSCs and showed that these treatments are efficacious in the prevention of immune rejection and increase the survival rates of MSCs, indicating that immune rejection does exist in these immune-privileged tissues (Francis et al., 2009;Xian and Huang, 2015). It is still controversial whether or not immunosuppression should be used after MSC transplantation; it may depend on the integrity of the RPE layer and BRB (Xian and Huang, 2015). If the RPE layer or BRB is intact, immune suppression is not necessary; if not, as in the RDDs, an immunosuppressant is needed (Oner et al., 2016). While it is not clear if extended MSC persistence or immune tolerance to MSCs will translate to a sustained therapeutic effect and improve clinical outcomes, conventional wisdom suggests that the beneficial effects of MSC therapy could be boosted by extending their persistence after injection (Ankrum et al., 2014b). In order to minimize cell rejection until the BRB is reestablished, many clinical trials of allogeneic MSC application for retinal diseases required systemic immunosuppressive therapy for the first 2-3 months after cell transplantation surgery (Oner et al., 2016;Park et al., 2017). However, the Palucorcel clinical trial (clinicaltrials.gov identifier NCT01226628) has no protocolspecified systemic immunosuppression included in the study, and no apparent effect of treatment was observed in the study (Ho et al., 2017;Heier et al., 2020).

CHALLENGES FOR MSC THERAPIES OF RDDs
Early encouraging pre-clinical animal results in the therapeutic use of MSCs have prompted great interest in exploring their potential for promoting retinal cell survival and modulating retinal inflammation. There are currently many MSC clinical trials for retinal diseases around the world (Shen, 2020;Singh et al., 2020). While the safety profiles shown in clinical trials are generally promising, some outcomes of advanced clinical trials have fallen short of expectations (Heier et al., 2020). Overenthusiasm and optimism surrounding stem cell therapies have also caused a surge in for-profit "stem cell" clinics globally, forming a direct-to-consumer marketplace for autologous stem cell interventions (Turner, 2018;Nirwan et al., 2019;Hwang et al., 2020). These clinics are run without the oversight of regulatory agencies and offer scientifically unproven retinal cell therapy products, which are not prepared in a standard and rigorous manner (Shen, 2020). The surgical procedures involved in the delivery of MSCs to vitreous, subretinal space, or periocular space also have significant risks, such as retinal detachment, PVR, and retinal perforations (Singh et al., 2020). Overall, there are still challenges with regard to the heterogeneity in the MSC product and the routes of delivery for MSCs (Galipeau and Sensébé, 2018;Shen, 2020;Singh et al., 2020).

Overcoming the Heterogeneity in the MSC Product
MSCs are an inherently heterogeneous population of cells whose therapeutic potency varies with the characteristics of the donor, tissue of origin, isolation method, and in vitro preparation methods (e.g., cell culture protocol and scale-up). The common sources of MSCs for clinical trials of retinal degeneration therapy are bone marrow (BM) stem cells (Cotrim et al., 2017;Park et al., 2017), adipose-derived MSCs (ASC) (Oner et al., 2016(Oner et al., , 2018 and umbilical cord-derived MSCs (such as UC-MSC or palucorcel) (Heier et al., 2020). They differ in the ease and efficiency of harvesting, proliferative ability and senescence, and paracrine activities, which greatly affect their therapeutic potency (Fričová et al., 2020). For example, BM aspirate is difficult to harvest and contains only 0.001-0.01% MSCs in the overall cell population; an intensive culturing process to expand the MSCs is required for clinical use. ASCs can be harvested easily from a small area under local anesthesia-an ASC harvest obtains a 500 times greater yield of MSCs than an equivalent amount of BM aspirate; thus, ASCs are a better source for autologous clinical use. UC-MSC harvesting is the least invasive and can produce a large amount of MSCs which can minimize the need to extensively expand the cells for allogenic use. In general, UC-MSCs proliferate faster than ASCs, and ASCs proliferate faster than BM-MSCs. UC-MSCs, ASCs, and BM-MSCs have been found to have senescence landmarks starting at passage 16, 8, and 7, respectively. Senescence can influence the therapeutic effectiveness, number, and maximum lifespan of the MSCs. ASCs have a relatively low secretion rate of proangiogenic molecules and cytokines, and thus they might be less suited to suppressing inflammation. UC-MSCs can secrete more angiogenic, neuroprotective factors, which make them an attractive option in RDD therapy. The potential concern for UC-MSC transplantation is that, as a type of non-neuronal or non-terminally differentiated cell, they may retain a capacity for continued proliferation inside the eye after transplantation (Fričová et al., 2020;Singh et al., 2020).
The obvious heterogeneity of the MSC product introduced during the manufacturing process emphasizes the need for characterizing and controlling the therapeutic potency of MSCs. Quality control protocols to standardize MSC product potency can reduce the risk of clinical failure (Trivedi et al., 2019). The most widely used potency assay for the MSC product is based on in vitro inhibition of T cell proliferation using activated CD4 + T cells (Bloom et al., 2015). While potency assays may improve product quality by excluding MSCs with low potential therapeutic efficacy, alternative strategies are needed to generate high-quality MSCs in sufficient quantities for large clinical trials (Levy et al., 2020).
In order to overcome MSC product heterogeneity and generate homogeneous, standardized high-quality MSCs, human-induced pluripotent stem cell (iPSC)-derived MSCs have been proposed as an unlimited source of cells for therapeutic applications in regenerative medicine. These cells possess better cell quality with batch-to-batch consistency and higher proliferative potential and display stronger immunomodulation effects (Zhang et al., 2012;Sabapathy and Kumar, 2016). Cymerus TM iPSC-MSCs significantly prolong survival in a pre-clinical humanized mouse model of graft-vs-host disease (GVHD) (Ozay et al., 2019). Good manufacturing practice (GMP)-grade iPSC-MSCs have already been used in clinical trials for refractory GVHD (Bloor et al., 2020). The iPSC-MSC approach serves as an excellent solution for scaling MSC manufacturing without sacrificing therapeutic potency through the passage and expansion of cells. One major concern regarding iPSC-MSCs is that the viral vector-based strategy for reprogramming might present a potential for tumorigenic transformation. However, recent developments in non-viral-based technologies, such as non-integrating episomal plasmid-based reprogramming (Slamecka et al., 2016), might present safer strategies for the generation of iPSC-MSCs suitable for use in a clinical setting (Sabapathy and Kumar, 2016). Human embryonic stem cells and iPSCs have already been extensively investigated for use as potential retinal stem cell treatments (Takagi et al., 2019); GMP-grade iPSC-MSCs may improve the efficacy of cell therapy for retinal degeneration.
Another strategy to overcome MSC product heterogeneity is using primed MSCs or boosting the innate therapeutic efficacy by other bioengineering methods (Levy et al., 2020). For instance, using a medium-based approach (similar to pre-licensing), MSCs can be induced to secrete elevated levels of neurotropic factors, including GDNF, BDNF, VEGF, and HGF, which have been shown to have protective effects in animal models of neurodegenerative diseases (Gothelf et al., 2014). When administered to patients with neurodegenerative diseases, these primed MSCs have been demonstrated to simultaneously deliver multiple neurotrophins and immunomodulatory components (Gothelf et al., 2017). While some pre-clinical studies have shown promising results of primed MSCs for RDDs, such as erythropoietin (EPO)-expressed rat MSCs (Guan et al., 2013), BDNF-expressed MSCs (Levkovitch-Verbin et al., 2010;Harper et al., 2011), it remains to be seen whether or not these engineered MSCs improve therapeutic outcomes in a clinical setting.

Optimization of Delivery Methods
A pre-clinical study indicated that systemically administered MSCs did not migrate to the injured eyes and had no effects on RDDs, as they are trapped in lung capillary beds (Johnson et al., 2010;Ge et al., 2014). Thus, local administration is the major route of delivery of MSCs for RDD therapy, as this route can deliver paracrine factors directly to the retina. Most clinical trials use intravitreal or subretinal transplantation, but subtenon's or retrobulbar injections are also used in the non-FDA registered, patient-funded clinical trial "Stem cell ophthalmology treatment study (SCOTS)" (Singh et al., 2020). It is imperative that the transplantation techniques have a reasonably low risk, so that the safety and efficacy of the MSCs can be evaluated properly.
Subretinal injection allows direct contact of the host and MSCs, but the injection procedure may disturb the retina. There are two main surgical approaches, internal and external. The internal approach requires an initial vitrectomy, after which a small-gauge needle is introduced into the eye via the pars plana and is passed through a retinotomy at the injection site and the cell suspension is deposited in the subretinal space near the fovea. The external approach involves subretinal delivery of cells via a cannula delivered outside the eye through the suprachoroidal space. This technique avoids exposure of the cells to the vitreous cavity but can induce retinal perforation and detachment (Ho et al., 2017). These severe adverse effects (retinal detachment or perforation) can be avoided by optimizing the procedures, as shown in a phase 2b study (Heier et al., 2020).
The intravitreal delivery approach is straightforward and widely used in anti-VEGF therapy in the ophthalmic clinics. The MSCs are injected as a suspension into the vitreous cavity via a needle through the pars plana. The cells do not gain access to the subretinal space and remain in the vitreous. Cellular cluster formation in the vitreous occurred in some cases of MSC intravitreal injection (Tzameret et al., 2014). As MSCs generally do not integrate into the retina, intravitreal injection is safer than subretinal injection. However, MSCs injected into vitreous cavity can still cause secondary glaucoma, epiretinal membrane and PVR (Kuriyan et al., 2017;Park et al., 2017;Khine et al., 2020). MSCs may locate on the retinal surface and differentiate improperly into myofibroblast-like cells which may induce fibrosis, PVR, and tractional retinal detachment; they can also increase oxygen demand, inducing ischemic microvascular changes and subsequent ocular neovascularization (Kuriyan et al., 2017;Khine et al., 2020). There are no definitive answers about which delivery approach leads to the best therapeutic effect. Thus, these surgery-related severe adverse effects will need to be carefully monitored in clinical trials.
Both intravitreal and subretinal injection can disturb the host BRB and expose the grafted cells to the host retinal immune system and thus raise concerns about long-term donor cell survival. Indeed, both the retention and survival of MSCs following local administration are important factors affecting the therapeutic outcome (Levy et al., 2020), despite pre-clinical study suggesting that the therapeutic potency to protect the retinal cells is not dependent upon the long-term survival of the donor cells (Lu et al., 2010). Nevertheless, multiple strategies have been investigated to improve the local administration of MSCs, including priming MSCs in vitro (Levy et al., 2020). For example, hypoxic priming upregulated the expression of prosurvival factors such as hypoxia-inducible factor 1, which can help MSCs adapt to the typically hypoxic disease site (Hu et al., 2008). Similarly, human MSCs from Wharton's jelly (hWJMSCs) expressing erythropoietin enhance the survival of retinal neurons against oxidative stress (Shirley Ding et al., 2018). Using biomaterials to encapsulate MSCs is another promising strategy to overcome challenges associated with local administration (Führmann et al., 2016). For example, HGFoverexpressing MSCs in a synthetic peptide-based hydrogel survived much longer and showed superior reduction in scar formation compared with native MSCs (Wu et al., 2017). It will be interesting to test these strategies in the MSC clinical trials for RDDs treatment in the future.

CONCLUSIONS
Retinal degeneration diseases are a leading cause of blindness worldwide and there is no effective treatment. MSCs can be easily isolated from multiple tissues and have shown promise in treating many diseases by restoring organ homeostasis in inflamed, injured, or diseased tissues. Pre-clinical animal studies suggested that their trophic and immunomodulatory properties can protect retinal neurons and enhance the function of retinal pigment epithelium cells from an array of retinal degeneration diseases. Recent clinical trials have demonstrated promising safety profiles for the application of MSCs in retinal diseases and have provided valuable data for future exploration. However, significant complications that arise from poorly designed clinical trials, questionable practices or regulatory shortcuts could substantially hinder research in MSC-based retinal therapies. Nevertheless, MSCs have great therapeutic potential for the treatment of retinal degeneration. A better understanding of the interaction between MSCs and host retinal degenerative environments is the key to yield an optimal benefit. Overcoming the heterogeneity in the MSC product and optimizing ocular surgical delivery to avoid adverse sequela are major challenges to translational use of MSCs to treat retinal degeneration.

AUTHOR CONTRIBUTIONS
YL and DC were involved in the concept and design. YL and XR wrote the first version of the manuscript. YL, YC, and DC reviewed and revised the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This work was supported by grants from the National Natural Science Foundation of China (81870665) to DC.