Extracellular vesicle therapy for retinal diseases

Abstract

Extracellular vesicles (EV), which include exosomes and microvesicles, are secreted from virtually every cell. EV contain mRNA, miRNA, lipids and proteins and can deliver this expansive cargo into nearby cells as well as over long distances via the blood stream. Great interest has been given to them for their role in cell to cell communication, disease progression, or as biomarkers, and more recent studies have interrogated their potential as a therapeutic that may replace paracrine-acting cell therapies. The retina is a conveniently accessible component of the central nervous system and the proposed paradigm for the testing of many cell therapies. Recently, several studies have been published demonstrating that the delivery of EV/exosomes into the eye can elicit significant therapeutic effects in several models of retinal disease. We summarize results from currently available studies, demonstrating their efficacy in multiple eye disease models as well as highlighting where future research efforts should be directed.

1. Introduction

The retina, due to its diencephalic origin, is part of the central nervous system (CNS) and converts photons into an electrochemical signal in a process known as phototransduction, allowing organisms to see. As typical with the CNS, damage, which can arise through a variety of traumatic and degenerative reasons, is permanent alongside the subsequent visual loss (Berry et al., 2008, 2019). Retinal diseases have multiple (non-mutually exclusive) theories explaining their cause and progression, owing to their complicated and multifactorial nature. It can be argued that a successful therapy must consider these multiple mechanisms rather than focusing on one pathway or molecule. In the example of glaucomatous damage, the majority of studies and preclinical therapies however target just one particular mechanism or signaling pathway e.g. glutamate-mediated excitotoxicity, tumor necrosis factor-α (TNF-α)-mediated inflammation, neurotrophic factor (NTF) deprivation etc. Several combinatorial therapies have been devised to address the multifactorial nature of retinal disease, from the literal combination and delivery of multiple NTF (Logan et al., 2006) to cellular therapy, whereby transplanted mesenchymal/neural stem cells (MSC/NSC) secrete a combination of said NTF (Flachsbarth et al., 2018; Johnson et al., 2014; Lu et al., 2013; Mead et al., 2014, 2015; Mesentier-Louro et al., 2014) and anti/pro-inflammatory cytokines (Millan-Rivero et al., 2018; Uccelli et al., 2008). Our earlier studies demonstrated the potential for MSC (bone marrow- and dental pulpderived) transplantation in a model of glaucoma (Johnson et al., 2010a; Mead et al., 2016), with significant neuroprotection of degenerating retinal ganglion cells (RGC) observed. MSC remained in the vitreous with their therapeutic efficacy resulting from paracrine-mediated mechanisms (Fig. 1).

Fig. 1.

Retinal ganglion cell (RGC) counts in a rat model of glaucoma after mesenchymal stem cell (MSC) treatment. Glaucoma was modeled through intracameral injections of transforming growth factor-β for 35d. Treatments consisted of intravitreal transplantation of dental pulp stem cells (DPSC), bone marrow MSC (BMSC), adipose-derived stem cells (ADSC) and dead DPSC (sham-treated; A). Retinae were stained with the phenotypic RGC marker BRN3A (red) and the nuclear marker DAPI (blue; scale bar: 50 μm). In (B), GFP⁺ MSC stained for the MSC marker STRO1 are identified in the vitreous, adhering to the inner limiting membrane. In (C), the mean number of BRN3A⁺ RGC in a 1 mm region of retina either side of the optic nerve head is shown from each of the above groups. Note the significant neuroprotective effect elicited by the transplanted MSC. Black lines indicate significant difference between groups (p < 0.01). Modified Fig. 4 from Mead et al. (2016), re-used under the Creative Commons Attribution 4.0 International (CCBY4.0) licence.

However, another secreted product of MSC that is suggested to mediate the paracrine benefit has recently gained a large amount of research interest. Known as extracellular vesicles (EV), their strong therapeutic potential derives from their expansive cargo and ability to deliver said cargo into cells and act on multiple signaling pathways (Kalluri and LeBleu, 2020). EV are becoming established signaling mediators between cells, including in the eye (Reviewed in Klingeborn et al., 2017a), but only recently are they gaining traction as a candidate treatment for ocular disease. This review discusses current progress in utilizing EV as a therapy for retinal diseases.

2. Extracellular vesicles

EV is the collective term for secreted vesicles and includes exosomes, microvesicles, and apoptotic bodies. They have distinct biogenesis pathways (Fig. 2) and are often distinguished by their size, internal cargo, and surface proteins.

Fig. 2.

Schematic diagram detailing exosomal treatment of the retina. Exosomes and microvesicles are isolated through ultracentrifugation of culture medium, conditioned by the proposed cell source. Lower speeds of centrifugation can be used in protocols that utilize polyethylene glycol while other techniques such as passing through a sucrose gradient are employed to further specify the vesicle size obtained. To partially-purify the 30–150 nm exosomes from the 100–1000 nm microvesicles, passage through a 0.22 μm filter is utilized. Following purification, exosome identity can be confirmed with Nanoparticle Tracking Analysis and Western blot before injection into the eye (vitreous or subretinal).

Exosomes form via the fusion of multivesicular bodies (an intracellular vesicular structure derived from endosomes) with the cell membrane and their subsequent release into the extracellular space (Mathieu et al., 2019; Thery et al., 2006). Microvesicles are instead formed due to outward budding of the plasma membrane. Exosomes are typically 30–150 nm whereas microvesicles are 100–1000 nm, although the exact values vary greatly between studies. Analysis of their size can be done using electron microscopy (Osteikoetxea et al., 2015) or a nanoparticle tracking analysis instrument (Fig. 3). The third class of EV, apoptotic bodies, are >1000 nm and are released through the membrane blebbing of cells undergoing apoptosis (Battistelli and Falcieri, 2020; Caruso and Poon, 2018; Jiang et al., 2017), although smaller EV from apoptotic cells (termed apoptotic microvesicles) have also been suggested. Apoptotic bodies have so far not seen any therapeutic use in the eye and appear to mainly function as signals to recruit macrophages to aid in cell debris clearing, as well as antigen presentation (Caruso and Poon, 2018). They do however contain miRNA and proteins as well as represent a heterogeneous population of subtypes and further study into their therapeutic and/or deleterious effects should be explored. In contrast, exosomes have shown remarkable therapeutic potential in many diseases throughout the body (Reviewed in Keshtkar et al., 2018) including Alzheimer’s disease (de Godoy et al., 2018), spinal cord injury (Sun et al., 2018; Wang et al., 2018a) and stroke (Xin et al., 2013, 2017), amongst others.

Fig. 3.

Electron microscopy images of exosomes before and after filtration through a 0.22 μm filter along with corresponding Nanosight/Nanoparticle Tracking Analysis of quantity and size. Modified Fig. 2 from Mead et al. (2018) and Mead et al., 2017, re-used under the Creative Commons Attribution 4.0 International (CCBY4.0) licence. The figure inset shows a higher quality electron microscopy image of an exosome (EXO), microvesicle (MV), and apoptotic body (APO). Reused from Osteikoetxea et al. (2015) with permission under the Creative Commons Attribution 4.0 International (CCBY4.0) licence.

Despite these differing subpopulations of EV, much overlap exists in the literature, with the term “exosome” used interchangeably with “extracellular vesicle”. Interestingly, “exosome” has been referenced more often than “extracellular vesicle” in published manuscripts over the last several decades, reflecting its popularity, yet this gap has narrowed significantly in 2019, perhaps reflecting research groups new to the field adopting the correct terminology (Fig. 4). The Minimum Information for Studies of Extracellular Vesicles (MISEV2018) (Théry and Witwer, 2018) are a recent published set of guidelines that, in summary, state that the term “EV” should be used exclusively unless they are confirmed to originate from the exosome biogenesis pathway (Fig. 2). While we agree with the above guidelines, given that we are reviewing past and present literature that has yet to take these into consideration, certain concessions were made. Thus, while many studies refer to their preparations as “exosomes” as opposed to EV, we will refer to them as EV unless they fulfill the definition of exosomes detailed recently (Klingeborn et al., 2017a) which are 30–150 nm vesicles loaded with at least some of the exosomal proteins CD63, CD9, CD81, syntenin-1, and TSG101. We also add a further requirement to the definition that is often included in most recent studies which is the use of a 0.22 μm filter, removing contaminating microvesicles (albeit at the cost of a reduced overall yield) that are often isolated with exosomes in most techniques such as ultracentrifugation or polyethylene glycol precipitation (Konoshenko et al., 2018; Ma et al., 2020; Mead and Tomarev, 2017; Pan et al., 2019). Isolation of exosomes with more intricate techniques such as sucrose gradients, relying on their buoyant density of about 1.10–1.19 g/ml, are also employed but this is largely restricted to specialized studies into vesicle mechanics and not therapeutic assessment (Shurtleff et al., 2016).

Fig. 4.

Publications with the keyword “exosome” or “extracellular vesicle” in the abstract/title from Jan 1st, 1980–Jan 1st, 2020. Note the exponential rise in publications referencing exosomes along with the historical popularity of “exosome” over “extracellular vesicle”, with the gap narrowing significantly in 2019.

A key detail regarding EV and the reason for their great research interest is that their cargo is expansive, containing proteins, mRNA, miRNA, and lipids. Furthermore, following secretion from a cell, these vesicles act as mediators of cell signaling, delivering their cargo into recipient cells and in the case of mRNA/miRNA, lead to the translation of new proteins/modulation of gene expression, respectively (Ratajczak et al., 2006; Skog et al., 2008; Valadi et al., 2007). By delivering multiple proteins, mRNA and miRNA (all of which target multiple different mRNA), EV are intrinsically a multifactorial treatment.

While the different types of EV are distinct, another variable that defines them is the source that the EV is secreted from. For example, we recently sequenced miRNA of exosomes from human BMSC and fibroblasts and identified over 40 candidates that were distinct between the two samples (Mead et al., 2018b). Several studies have been published detailing the miRNA within exosomes from multiple MSC types (Table 1). While many of the most abundant miRNA are consistent between studies, it is apparent that some variability even from the same cell type exists, demonstrating the heterogeneity of MSC cultures as well as the variability between miRNA library construction/analysis methodologies. Regarding the proteome and lipidome, distinct differences have been observed between exosomes isolated for U87 glioblastoma cells, Huh7 hepatocellular carcinoma cells, and BMSC (Haraszti et al., 2016). Exosomes from distinct retinal cells such as retinal astrocytes (Hajrasouliha et al., 2013), retinal pigment epithelium (RPE) cells (Klingeborn et al., 2017b), and retinal progenitor cells (Zhou et al., 2018) also contain distinct proteomes and this is reflected in their therapeutic efficacy (discussed below). It has equally been shown that cells secrete exosomes with variable cargo depending on the stage of differentiation they are in. For example, osteogenic differentiation of MSC leads to exosome secretion with extracellular matrix mineralization properties, but only in the late, not early phase of differentiation (Wang et al., 2018b). Even cell passage has been shown to have an effect on the neuroprotective efficacy of MSC exosomes, with it diminishing with increasing cell passage of the source (Venugopal et al., 2017). With the cargo of EV varying depending on their cellular origin, it could be thus assumed that EV cargo is just a sample of the cellular cytoplasm. However, proteomic analysis of adipose-derived MSC and EV demonstrated that over 100 proteins were more abundant in EV in comparison to the cell, suggesting that the loading of EV involves active and specific trafficking (Eirin et al., 2016). Interestingly, when comparing exosomes to microvesicles, it appears that this specific trafficking mechanism is more evident in exosomes as their protein population is more distinct from the host cell’s than the protein population of microvesicles (Haraszti et al., 2016). Comparison of mRNA, miRNA, and transfer RNA abundance between adipose-derived MSC/T cells and their EV leads to similar conclusions, select mRNA and miRNA enriched in EV in comparison to the cell (Chiou et al., 2018; Eirin et al., 2014). Likewise, exosomes isolated from HEK293 cells contain selectively packaged miRNAs compared with HEK293 cells and it was found that the RNA-binding Y-box protein (YBX1) binds to and is required for the sorting of certain miRNAs (mir-223) in exosomes (Shurtleff et al., 2016).

Table 1.

The ten most abundant miRNA in exosomes isolated from human mesenchymal stem cells (bone marrow-derived mesenchymal stem cells, BMSC; umbilical cord blood-derived mesenchymal stem cells, UCMSC; adipose-derived stem cells, ADSC).

Study	Mead et al. (2018b)	Ferguson et al. (2018)	Baglio et al. (2015)	Wang et al. (2018)	Sun et al. (2017)	Qian et al. (2016)	Fang et al. (2016)	Baglio et al. (2015)
Source of exosomes (human)	BMSC	BMSC	BMSC	BMSC	UCMSC	UCMSC	UCMSC	ADSC
Ten most abundant miRNA	miR-221-3p	miR-1246	miR-143-3p	miR-21-5p	miR-125b-5p	miR-21	miR-21-5p	miR-486-5p
	let-7a-5p	miR-23a-3p	miR-10b-5p	miR-125b-5p	miR-21-5p	miR-125b	miR-125b-5p	miR-10a-5p
	miR-21-5p	miR-451a	miR-486-5p	miR-221-3p	miR-24-5p	miR-23a	miR-23a-3p	miR-10b-5p
	miR-320a	miR-125b-5p	miR-22-3p	miR-16-5p	miR-16-5p	miR-100	miR-100-5p	miR-191-5p
	miR-486-5p	miR-199a/b-3p	miR-21-5p	let-7a-5p	miR-92a-3p	let-7f-5p	miR-145-5p	miR-222-3p
	miR-423-5p	let-7a-5p	miR-222-3p	miR-23a-3p	miR-100-5p	let-7a-5p	let-7f-5p	miR-22-3p
	miR-21-5p	miR-4454/7975	miR-191-5p	miR-100-5p	miR-106a-5p	miR-145	let-7a-5p	let-7a-5p
	miR-1246	miR-21-5p	miR-100-5p	miR-142-3p	miR-19b-3p	miR-1260b	miR-1260a	miR-21-5p
	miR-122-5p	let-7b-5p	let-7a-5p	miR-222-3p	miR-145-5p	miR-1260a	miR-1260b	miR-127-3p
	miR-92a-3p	miR-100-5p	miR-99b-5p	miR-24-3p	miR-25-3p	miR-199a	miR-199a-3p	miR-143-3p

Along with their cargo, the number of exosomes released by cells can also vary greatly. In a comparison of several cell types including myoblasts and HEK cells, MSC secreted a significantly greater number (>10-fold) of exosomes (Yeo et al., 2013). Finally, along with the differences between the three types of EV, within each type they may be further divided into subtypes. Exosomal surface proteins can be analyzed using flow cytometry and antibody-bead conjugates which, although cannot quantify exosomes due to an inability in knowing how many exosomes have bound to each bead, does demonstrate considerable differences between sub populations that possess or lack exosomal proteins such as CD63 (Mead and Tomarev, 2017). To analyse these subtypes, more homogeneous population of exosomes are obtained through the inclusion of additional purification steps such as a flotation of exosomes to an interface between 20 and 40% sucrose and immunoprecipitation with CD63 antibody-immobilized beads (Shurtleff et al., 2016) or high-resolution iodixanol density gradient fractionation (Jeppesen et al., 2019). A distinct difference in the RNA cargo between the “high density” and “low density” exosomes (Jeppesen et al., 2019; Shurtleff et al., 2016) and distinct functional differences between them (Willms et al., 2016) has been reported, yet it remains to be seen if these distinctions are relevant when it comes to developing potential therapies for the eye. These additional purification and subtype separation techniques are not typically employed in research outside exosome-focused research groups and thus their therapeutic and biological relevance is largely unexplored.

One considerable benefit EV offer over cellular therapy as a treatment is their storage properties (Kusuma et al., 2018). EV RNA quality showed little to no deterioration after storage for 5 years at −20 °C in comparison to fresh EV while storage at 4 °C for 2 weeks led to significant degradation in some RNA (Ge et al., 2014). This can be partly explained by their bi-lipid membranes, which protects the cargo from enzymatic/chemical degradation. Another benefit is the safety in comparison to injections of dividing/differentiating cells into the eye. A recent report detailed three patients who received intravitreal injections of adipose-derived MSC as a treatment for age-related macular degeneration (AMD). Unfortunately, these patients subsequently went blind due to a variety of complications associated with the stem cell transplant including retinal detachment and hemorrhage (Kuriyan et al., 2017). The formation of a monolayer of cells on the inner limiting membrane of the retina is heavily involved in the pathology of retinal detachment and proliferative vitreoretinopathy (Yang et al., 2015) and these cells originate from the epithelial-mesenchymal transition of RPE cells. Given that transplanted MSC adhere and cluster to the inner limiting membrane (Mead et al., 2016), it is possible that these transplanted cells formed epiretinal membranes as observed in one patient receiving MSC (Kim et al., 2017). A more recent study transplanting MSC into the vitreous of rats also demonstrated significant vascular damage alongside glial activation and an inflammatory response (Huang et al., 2019). Cell therapy is also compounded by an unknown division rate and rate of death after transplantation, meaning a known number of cells quickly becomes unknown after administration. While the above does not mean cellular therapy is unfeasible, these complications are avoided by purifying the active secreted compound, believed to be EV (as well as neurotrophic proteins), and administering this in the place of the cells. It can be argued that EV therapy is more controlled regarding its dose since there is no risk of division occurring post-transplantation. However, both cells and EV share a dosing problem that is intrinsic to their role as deliverers of a multifactorial cargo. Given that this cargo varies between passages (Mead et al., 2014) and donors (Table 1), using the same number of cells/EV does not guarantee that the therapeutic cargo is being correctly dosed. Finally, a large disadvantage with retinal cell therapy is the lack of integration of transplanted cells into the retina (Emre et al., 2015; Johnson et al., 2010a; Mead et al., 2013) unless further measures are taken such as digestion of the inner limiting membrane and modulation of retinal glial activity (Johnson et al., 2010b), which may itself damage the retina. EV therapy avoids this complication and can pass through the inner limiting membrane with ease (Mead and Tomarev, 2017).

3. Retinal disease

EV are a strong candidate as a cell free therapy and below we discuss current evidence for their use in various diseases affecting the retina.

3.1. Optic nerve crush

Optic nerve crush is a model of traumatic optic neuropathy, a severe acute condition in which the delicate optic nerve, on its path from the retina to the lateral geniculate nucleus/superior colliculus, is physically injured. Crushing of the optic nerve in mice and rats leads to a 50% loss of RGC by 7 days and 90% loss by 14 days (Berkelaar et al., 1994; Leung et al., 2008; Rodriguez et al., 2014). Not only is it characterized by the selective loss of RGC but also the Wallerian degeneration of RGC axons that fail to regenerate (Berry et al., 2008). Finally, the optic nerve crush model appears to selectively kill certain RGC subtypes while largely preserving others, and in particular, α-RGC and melanopsin-expressing M1-RGC demonstrating robust survival in comparison to other RGC subtypes (Duan et al., 2015; Tran et al., 2019).

Recently we transplanted exosomes derived from BMSC into the vitreous of rats after optic nerve crush (Mead and Tomarev, 2017). Exosomes delivered their cargo into RGC, as shown by preloading the exosomes with a fluorescent marker, and provided significant neuroprotection and functional preservation, whereas long-distance axon regeneration was not observed. Fibroblast exosomes, which were used as control exosomes, provided no therapeutic effects. BMSC exosomes also preserved RGC function by over 50%, as measured by electroretinography. Since preventing RGC death does not inherently mean a prevention of RGC dysfunction (Fry et al., 2018), this result suggests exosomes work through multiple pathways to not only protect RGC but also distinctly preserve their function. Interestingly, exosomes appeared to be the therapeutically efficacious EV whereas microvesicles were not and were even toxic to RGC at higher concentrations (Mead and Tomarev, 2017) (Fig. 5b), an observation seen also in a retinal ischemic model (discussed below, van der Merwe et al., 2019) as well as in cortical neuron cultures (Lopez-Verrilli et al., 2016). The mechanism of action was determined to be, at least partially, due to the miRNA evident by the ablation of therapeutic efficacy if argonaute-2 (AGO2) is knocked down in BMSC. AGO2 is a protein that forms part of the miRNA complex and is necessary for their ability to inhibit mRNA translation. Knocking down AGO2 in cells prior to EV isolation leads to EV lacking in mature miRNA (Lv et al., 2014; Zhang et al., 2016). We can speculate that differences between the cargo packaged in exosomes and microvesicles (e.g. proteins and/or RNA) is the reason for their opposing effects on neurons but further investigations into the mechanism of action are needed before this can be corroborated.

Fig. 5.

Differential effects of exosomes and microvesicles on retinal ganglion cells (RGC)/neurons. In three separate studies, one in cortical neurons (A) and 2 in RGC (B/C), exosomes demonstrated a neuritogenic/neuroprotective effect with microvesicles exerting the opposite. The first study (A) showed that exosomes were neuritogenic whereas the effect of microvesicles was worse than untreated controls. The second (B) demonstrated the efficacy of extracellular vesicles diminished at higher doses and this was due to the contamination of microvesicles. A third study (C) showed the same but did not confirm the effect was due to contaminating microvesicles. Modified Fig. 3 from Loppez-Verrilli et al. (2016) (A), Fig. 3 from Mead et al. (2017) (B), and Fig. 1 from van der Merwe et al. (2019), re-used under the Creative Commons Attribution 4.0 International (CCBY4.0) licence.

Further studies by Pan and coauthors transplanted exosomes derived from umbilical cord MSC into the vitreous of rats after optic nerve crush (Pan et al., 2019). As we had previously defined, by removing microvesicles through the use of a 0.22 μm filter, significant RGC neuroprotection was achieved and similarly, RGC axon regeneration was not. Authors also demonstrated significant glial activation. Interestingly the effect was not as significant as seen with BMSC exosomes and authors reasoned that this is due to the reported differences in exosomal miRNA between those isolated from BMSC (Baglio et al., 2015; Mead et al., 2018b) and UMSC (Fang et al., 2016).

A separate study utilized exosomes isolated from L-cells, a fibroblast cell line (Tassew et al., 2017). It is worth noting that in this study, authors did not filter their EV or fully define their preparation. Thus, their preparation is more accurately referred to as EV, a mixture of exosomes and microvesicles. Authors did not observe any significant neuroprotection of RGC but interestingly, observed significant regeneration of RGC axons. The mechanism of action appears to be due to the recruitment of Wnt10b to lipid rafts and subsequent activation of the axogenic mTOR pathway via GSK3β. This distinction between L cell exosomes/EV and MSC exosomes in the axogenic effect is likely due to a difference in their internal cargo. We recently sequenced L cell exosome miRNA and performed a comparison between them and BMSC/fibroblast exosomes. (Fig. 6). Their miRNA profile is shown with the majority distinct from that found in BMSC exosomes, although some similarities were observed (Fig. 6b). The difference in RGC neuroprotection observed could also be explained by the exosome isolation technique. Microvesicles were included in the authors preparation (i.e. not filtered out), and our observation of their toxicity on RGC (Mead and Tomarev, 2017) suggests L cell exosomes may indeed be neuroprotective but is obfuscated by microvesicle-induced RGC death.

Fig. 6.

miRNA in L cell exosomes. miRNAseq was performed on exosomes derived from L cells with those detected displayed (A) as mean estimated abundance (derived from the reads) ± standard error mean (SEM). Mouse L cell exosome miRNA that are homologues to their human miRNA counterpart were selected and compared to human bone marrow mesenchymal stem cell (BMSC) and dermal fibroblast exosome miRNA. Those miRNA also shown to be abundant in BMSC exosomes and L cell exosomes in comparison to fibroblast exosomes are displayed (B) as mean estimated abundance (derived from the reads) ± SEM. Comparative data for miRNA expression in BMSC exosomes/fibroblast exosomes is from a previous publication (Mead et al., 2018b).

In the above studies, BMSC (Mead and Tomarev, 2017) and umbilical cord MSC (Pan et al., 2019) exosomes both promoted neuroprotection without axon regeneration whereas L-cell exosomes (Tassew et al., 2017) did the opposite. This confirms the distinction between the pathways involved in neuroprotection and those for axon regeneration. It has been shown that Sox11 expression promotes axonal regeneration for some RGC subtypes yet for some subtypes promotes their death (Norsworthy et al., 2017). It is possible that despite the expansive cargo of MSC exosomes, they do not properly activate regeneration pathways which also include pten/socs3 modulation (Sun et al., 2011) and induction of neural activity (Lim et al., 2016). This may represent a benefit of cell therapy over EV as MSC have been demonstrated to reliably stimulate both regeneration and survival (Mesentier-Louro et al., 2014, 2019; Tan et al., 2015).

3.2. Glaucoma

Glaucoma bares some similarities to optic nerve crush in that it is also characterized by the selective death of RGC (Almasieh et al., 2012). In contrast, the death is a slow, progressive degeneration as opposed to acute loss and thus, is a more sinister condition. The principle risk factor is an elevation in intraocular pressure (IOP) which is believed to cause compression of the optic nerve at the lamina cribrosa. IOP is only a risk factor not a cause however, owing to the fact glaucoma can occur with normal IOP values (Coleman and Miglior, 2008). The mechanism by which RGC die in glaucoma is still not fully understood and studies demonstrate a myriad of processes responsible including NTF deprivation, excitotoxicity, inflammation, oxidative stress, and antero/retrograde axon transport dysfunction (Reviewed in Almasieh et al., 2012; Syc-Mazurek and Libby, 2019). For a treatment to be effective in preventing RGC death and dysfunction, it must be equally multifactorial to address these injury processes. Previous success has been found through transplantation of MSC (Fig. 1) which secrete a multitude of beneficial factors (Emre et al., 2015; Harrell et al., 2019; Johnson et al., 2010a; Mead et al., 2013; Mesentier-Louro et al., 2014, 2019).

We recently transplanted BMSC exosomes into the vitreous of three separate animal models of glaucoma: laser and microbead rat models (Mead et al., 2018b), and a genetic DBA/2J mouse model (Mead et al., 2018a). In all three models, BMSC exosomes promoted significant survival of RGC along with preventing their functional decline that is characteristic of glaucoma models. In the DBA/2J model, we also observed a protective effect on RGC axons. As with the optic nerve crush model, we used fibroblast exosomes as a negative control as they elicited no therapeutic effect in these three models of glaucoma. One interesting finding was that the efficacy of exosomes was maintained even when delivered on a monthly basis but failed to elicit neuroprotection if the treatment was delivered more infrequently. The DBA/2J mice are a 12-month model of glaucoma and exosomes were still efficacious over this time period.

The mechanism of action appeared to be, as before, due to the miRNA cargo they delivered into RGC. This was confirmed through AGO2 knockdown and the ablation of neuroprotection (Mead et al., 2018b). To determine which miRNA were responsible for these therapeutic effects, miRNAseq was performed, comparing miRNA in the efficacious MSC exosomes to the ineffective fibroblast exosomes. Previous studies have already profiled MSC EV/exosomes and mapped out the most abundant miRNA (Ferguson et al., 2018; Qian et al., 2016; Sun et al., 2017) (Table 1), and we identified 43 miRNA that were abundant in BMSC exosomes in comparison to fibroblast exosomes (Mead et al., 2018b). Given that miRNA target a great many different mRNA, it is difficult to determine which molecules and pathways are responsible for the therapeutic effects observed. Many of these targets are still only predictions with only a fraction tested and experimentally observed. Within these targets however, well known instigators of RGC death including the bcl2 family (Maes et al., 2017), tnf (Tezel, 2008), and pten/mtor (Morgan-Warren et al., 2016) exist and further study will determine to what extent exosome-derived miRNA is acting through these pathways.

This data suggests that exosomes may serve as a suitable neuroprotective strategy, both in glaucoma that is not amenable to IOP lowering therapies, or as an adjunctive treatment. Another important conclusion is that long-term exosome treatment could be developed that requires only a monthly injection, as is done with anti-vascular endothelial growth factor (VEGF) treatments for AMD. This is likely based on a combination of the stability of exosomes as well as miRNA whose stability is reported to be over several days (Bartel, 2018).

In an effort to determine if the therapeutic effects we and others have observed is also applicable to human retina, we tested exosomes in a human in vitro retinal culture (Sluch et al., 2015, 2017). Human embryonic stem cell lines were differentiated into retinal cells, which included RGC, and were injured using the microtubule poison colchicine (Mead et al., 2020). Delivery of BMSC-derived exosomes provided significant neuroprotection of human RGC (Fig. 7). While we would certainly not argue that this in vitro system models glaucoma, it does provide evidence that the efficacy we are seeing in animal models may indeed be translatable to the human condition. More studies are needed using human tissue to strengthen this argument.

Fig. 7.

Exosome treatment of human retina. Heterogeneous retinal cultures were generated from a human embryonic stem cell line expressing a fluorescent marker under the brn3b (retinal ganglion cell (RGC) specific) promoter. To induce RGC degeneration, colchicine, a microtubule poison, was added to cultures (B) and led to significant RGC loss compared to uninjured controls (A). Ciliary neurotrophic factor (CNTF) led to significant neuroprotection of RGC (positive control, C), as did mesenchymal stem cell (MSC) exosomes (D), and tumor necrosis factor-α (TNF-α) primed MSC exosomes (E; scale bar: 250 μm). The quantified number of BRN3B⁺ RGC is shown in F. Fig. 2 from Mead et al. (2020) re-used under the Creative Commons Attribution 4.0 International (CCBY4.0) licence.

3.3. Retinal ischemia

Retinal ischemia, such as due to occlusion of the retinal artery or detachment of the retina, causes significant and irreversible damage. As with glaucoma, transplantation of MSC has shown efficacy at preventing retinal cell loss and dysfunction (Dreixler et al., 2014) and also, as with glaucoma, exosomes isolated from BMSC were able to recapitulate the effects of BMSC when transplanted into the vitreous of retinal ischemic mice, induced by hyperoxic conditioning (Moisseiev et al., 2017). These therapeutic effects included a significant reduction in retinal thinning and neovascularization and were present 14 days after the treatment. The ability of exosomes to prevent neovascularization is also seen in the choroid following delivery of retinal astrocyte-derived exosomes but is not seen when using RPE-derived exosomes (Hajrasouliha et al., 2013), again demonstrating the importance of the exosome source.

A more recent study utilized a brief elevation in IOP (15–150 mmHg for 60 min) to induce retinal ischemia in rats (van der Merwe et al., 2019). EV were isolated from bioscaffolds and in particular, decellularized porcine urinary bladder matrix. These EV, known as matrix bound nanovesicles, are similar to exosomes in that they are lipid membrane bound, containing protein and RNA, although their exact cargo profile may differ. The characterization of these matrix bound nanovesicles, including size or RNA/protein abundance was however not shown and thus it is unknown if these matrix bound nanovesicles are indeed just EV that have become associated with the scaffold following secretion. Evidence for this is shown when RGC in cultures are treated with membrane bound nanovesicles, which promoted neuritogenesis with increasing dosage, but a bi-phasic effect was observed with the neuritogenic effect dissipating at very high doses (Fig. 5c). This observation mirrored what we observed whereby MSC EV promoted neuritogenesis of RGC in a bi-phasic dose responsive manner (Mead and Tomarev, 2017). We had confirmed that this negative effect at increasing doses was due to microvesicles, and their removal from the EV sample, leaving just exosomes, ablated the bi-phasic dose response effect (Fig. 5b). Lopez-Verrilli and coauthors (Fig. 5a) also demonstrated a similar effect on cortical neurons with exosomes eliciting neuritogenesis while microvesicles did not (Lopez-Verrilli et al., 2016). Thus, it is possible microvesicles were present in the authors preparation. Despite this, authors demonstrated that EV treatment prevented microglia/astrocyte activation-induced release of the pro-inflammatory cytokines interleukin (IL)-1β, IL-6, and TNF-α, significantly reducing subsequent RGC degeneration in vitro and in vivo (van der Merwe et al., 2019). Finally, authors also demonstrated that the intravitreal delivery of these EV reduced loss of cholera toxin b-subunit⁺ RGC axons as well as dysfunction in RGC, as measured by the photopic negative response.

Retinal ischemia can also occur when the retina becomes detached from the choroid, from which it depends on for its blood supply. In a rat model of retinal detachment, injection of BMSC-derived exosomes reduced the expression of pro inflammatory cytokines such as TNF-α while upregulating autophagy (Ma et al., 2020). Authors demonstrated a subsequent neuroprotective effect on photoreceptors, reducing cell loss despite the detached retina. While a mechanism of action was not deduced, authors did note the abundance of exosomal proteins with neuroprotective and anti-inflammatory properties.

3.4. Retinal laser injury

A separate model of retinal injury utilizes a laser, not to burn the outflow pathways like in glaucoma but to directly burn the retina. Several laser burn spots are delivered to the retina, which initiates indiscriminate rather than specific cellular degeneration alongside inflammation.

Delivery of MSC EV (unfiltered exosomes) into cultures of retinal cells after heat induced injury, or into the vitreous of mice after laser injury provided significant neuroprotection of retinal cells to the same efficacy as the MSC themselves (Yu et al., 2016). Along with a reduction in TUNEL⁺ retinal cells/thinning of retinal layers, MSC EV also prevented declines in a- and b-wave amplitudes, suggesting a preservation of photoreceptor and bipolar cell function, respectively. MSC EV diffused throughout the retina and RPE within 1 h. One mechanism identified by the authors was the exosome-induced down-regulation of MCP-1 retinal expression, whose upregulation is usually a consequence of retinal injury. MCP-1 is a chemotactic cytokine that attracts macrophages and microglial cells into the injury site, leading to further damage and degeneration. MSC EV reduced MCP-1 expression in vitro and in vivo, reducing macrophage infiltration and this effect was abolished if MCP-1 was delivered into animals. This study reveals another mechanism of action for EV, an anti-inflammatory one, yet did not determine if the down-regulation of MCP-1 was due to the EV protein or RNA cargo.

A separate study focused on the effects of laser damage to RPE, which causes choroidal neovascularization, a characteristic feature of wet AMD (He et al., 2018). In vitro, laser damage to RPE cells induced the production of VEGF, the principal growth factor responsible for the neo-vascularization and the basis for the anti-VEGF drugs used in the clinic. Treatment of laser-injured RPE cells with umbilical cord MSC-derived EV reduced the transcription and translation of VEGF whereas in vivo, delivery of EV did the same while reducing retinal damage as measured by fundus fluorescein angiography.

In the same in vivo model of choroidal neovascularization, Hajrasouliha and coauthors demonstrated that exosomes from retinal astrocytes can inhibit the formation of new blood vessels as well as suppress retinal vascular leakage (Hajrasouliha et al., 2013). Similar to other studies, a 0.22 μm filter was employed to filter out microvesicles. The authors state that the mechanism of action is likely through the inhibition of macrophage migration which is a major source of inflammatory cytokines as well as VEGF. Interestingly, authors attributed the therapeutic effects to the protein content of exosomes and identified several anti-angiogenic candidates abundantly found in retinal astrocyte-derived exosomes. By inhibiting the MMP-induced production of endostatin and utilizing the subsequently generated endostatin-free exosomes, suppression of vascular leakage was no longer observed, demonstrating a role for exosome-delivered endostatin.

3.5. Autoimmune uveitis

Uveitis is an inflammatory condition of the eye requiring immunosuppressive treatment. Since long-term use of immunosuppression comes with several side effects, there is still a need for new treatments. Interphotoreceptor retinol-binding protein immunization induces experimental autoimmune uveitis, and the inflammatory cell retinal infiltration (granulocytes, natural killer cells, macrophages, and T cells) is ameliorated after treatment with umbilical cord MSC exosomes (filtered and characterized) (Bai et al., 2017). MSC exosomes also prevented a loss in a- and b-wave amplitude, suggesting photoreceptor and bipolar cell function was preserved. Authors found that the MSC exosomes anti-inflammatory effects were specifically on T cell migration and not proliferation/apoptosis. While the study did not determine their mechanism of action, previous studies have shown that MSC exosomes inhibit macrophage activation through miRNA-mediated down-regulation of the toll-like receptor and nuclear factor kappa B (NF-kB) pathway (Phinney et al., 2015). Other mechanisms such as the polarization of CD4⁺ T cells to regulatory T cells has also been described (Bin et al., 2014).

A separate study testing the effects of MSC exosomes in experimental autoimmune uveitis delivered exosomes into the tail vein (Shigemoto-Kuroda et al., 2017). Exosomes were isolated by column fractionation and characterized using exosomal markers CD63 and CD81. MSC exosomes performed just as well as MSC in preventing photoreceptor layer disruption and inflammatory cell infiltration. Interestingly, only a single injection was administered at the beginning of the 21 day study, corroborating our own reports of MSC exosomes remaining efficacious for up to 1 month in the eye (Mead et al., 2018a, 2018b). MSC exosomes/MSC also reduced the transcription of many pro-inflammatory cytokines including interferon gamma (IFN-g), IL-17A, IL-2, IL-1b, IL-6, and IL-12A (Shigemoto-Kuroda et al., 2017). Unlike the previous study however, authors demonstrated that MSC exosomes suppressed T cell proliferation. One possible explanation for this discrepancy is that authors cultured their MSC in serum free medium designed to activate/prime the MSC prior to exosomes isolation, which would likely have changed their internal cargo and thus, therapeutic action.

The anti-inflammatory properties of EV have also been demonstrated by RPE, whose secreted EV provide immunomodulatory effects on monocytes and even induce their death (Knickelbein et al., 2016). Currently however, they have not been utilized as a potential therapeutic.

These studies suggest that MSC EV and in particular exosomes have potential as a treatment in inflammatory diseases of the eye. Further studies on their long-term efficacy, dose and ideal source of said exosomes are needed to improve the treatment. One exciting observation is MSC EV efficacy is still present when delivered into the blood stream rather than the eye, suggesting that they can home into an injured environment (Shigemoto-Kuroda et al., 2017). While this would be a more ideal route of administration from the patient’s perspective, the potential for off-target effects with pernicious consequences would need to be considered. The anti-inflammatory properties of EV are not just relevant to uveitis but also the retinal injury models discussed above. Retinal/optic nerve injury is followed by a polarization of microglia to a M1 pro-inflammatory phenotype, which secrete various inflammatory cytokines including TNF-α. These can not only directly induce the neurodegeneration of RGC (Tezel, 2008) but polarize astrocytes to a neurotoxic A1 phenotype which itself leads to RGC neurodegeneration (Liddelow et al., 2017). Further studies are required to determine if these anti-inflammatory effects are a relevant mechanism behind the EV-mediated neuroprotection previously discussed.

3.6. Diabetic retinopathy

Diabetic retinopathy, a consequence of diabetes mellitus that involves inflammation, microaneurysms, vasculature damage and subsequent neo-vascularization (Stitt et al., 2016) has also shown preliminary promise as an eye disease amenable to EV therapy. Delivery of MSC (adipose-derived) EV into the eye, either subconjunctival or intravitreous (but not intravenous) prevented significant retinal degeneration (Safwat et al., 2018) in a streptozotocin-induced model of diabetic retinopathy. Authors demonstrated that exosomes delivered miRNA-222 into the retina and restored falling levels typically associated with diabetic retinopathy. The discrepancy between this study’s inability to obtain a clinical effect after intravenous administration, and the positive effects seen in the above study (Shigemoto-Kuroda et al., 2017) emphasize the need for further investigation on this potential route of administration.

A separate study utilized the same model and delivered umbilical cord MSC exosomes intravitreally (Zhang et al., 2019). Hyperglycemia-induced inflammation is ameliorated by MSC exosomes in comparison to fibroblast exosomes, as measured by ELISA for the inflammatory markers IL-1β, IL-18, and caspase-1 in the vitreous. The mechanism of action appears to be miR-126-mediated inhibition of the high mobility group box 1 (HMGB1) signaling pathway. Diabetic retinopathy is associated with decreased miR-126 and over expression of miR-126 in MSC exosomes further augmented the therapeutic efficacy.

3.7. Clinical trials

As of this review 148 clinical trials have been listed looking at “exosomes” and 39 mentioning “extracellular vesicles”. However, very few are utilizing them as a therapy with the rest mostly focusing on the use of exosomes as biomarkers of disease.

Two of these clinical trials testing MSC EV therapies that have been published include steroid refractory graft-versus-host disease (Kordelas et al., 2014) and chronic kidney disease (Nassar et al., 2016). MSC EV reduced pro-inflammatory cytokine secretions including TNF-α, increased anti-inflammatory cytokines secretions including TGF-β, and improved patient recovery and kidney function.

Two clinical trials are listed using exosomes as a treatment for eye disease, one in diabetic retinopathy which is not yet recruiting (ClinicalTrials.gov Identifier: NCT03264976) and another for the treatment of macular holes which is still recruiting and has already published preliminary results (ClinicalTrials.gov Identifier: NCT03437759). Five patients with large and refractory macular holes were treated with an intravitreal delivery of MSC-derived exosomes (Zhang et al., 2018). Exosomal presence was confirmed using Western blot, staining for exosome markers such as CD63, CD9 and CD81. Since no size exclusion (e.g. 0.22 μm filter) was utilized, the preparation undoubtedly also included microvesicles and is thus more accurately described as MSC EV. Results of the study suggest that MSC EV stimulate the closure of macular holes although the mechanism of action was not elucidated, and control groups not included. The intravitreal MSC EV therapy was well tolerated with only one patient experiencing an inflammatory reaction which was not present when the dose was reduced.

As more studies demonstrate that EV have an active and potentially therapeutic role in the body, as opposed to only a passive one (Joo et al., 2020; Tieu et al., 2019), it is anticipated that there will be more clinical trials focusing on their clinical potential rather than their role solely as biomarkers.

4. Future considerations

While EV show great promise, many questions still remain unanswered.

4.1. Toxicology and dosing

While no evidence exists for any complications arising from delivery of EV into the eye, extensive toxicology studies are still needed. Some in vitro (Maji et al., 2017) and in vivo (Zhu et al., 2017) toxicology reports have been published detailing their safety after culture treatment or systemic delivery, but how true this is for ocular delivery is still not known. They also report toxicological differences between different cellular sources of EV which although is unsurprising given what we know, emphasizes the importance of treating EV from different cells as distinct agents.

Secondly, the large-scale production of clinical-grade EV represents a significant barrier to moving this experimental treatment into the clinic. Issues such as ensuring the batch-to-batch variability remains minimal as well as the detection of any viruses that will likely be enriched alongside EV remains paramount when moving forward (Rohde et al., 2019). Variations in the length of time in culture may also affect the cells and subsequently, the EV, increasing variability. For EV to be effectively dosed, it is not enough to simply consider their quantity but instead to dose for their cargo, ensuring that a controlled amount of the therapeutically efficacious elements are delivered irrespective of the number of EV particles. It is also important to consider that the therapeutically efficacious component of the isolate may indeed be an EV subtype that can be further purified, however techniques to achieve this are still lacking (Greening and Simpson, 2018) and the benefits would need to be balanced against the added cost. Along with the EV subtype, the subtypes of the target cells should also be taken into consideration. Using RGC as an example, just as different injuries affect different RGC subtype, it is also possible that EV treatment only protects specific RGC subtypes and given that over 40 subtypes have been identified (Reviewed in Sanes and Masland, 2015; Tran et al., 2019), these potential differential effects warrant investigation. Regarding large scale EV production, one research focus has been to target the MSC themselves, modifying them in such a way as to improve the isolated EV yield and efficacy (Phan et al., 2018).

4.2. Targeting EV to cells

For EV to exert their effects on the injured retina, they must be targeted to the correct cells and subsequently internalized. The above studies have demonstrated that EV deliver cargo into a whole range of retinal cells including RGC (Mead and Tomarev, 2017), microglia, astrocytes (van der Merwe et al., 2019), and RPE cells (He et al., 2018). However, many studies do not interrogate the exact cellular target, only referencing global changes in retinal expression, function, or morphology. Future studies should pay special attention to this aspect of EV, particularly as it is becoming apparent EV can preferentially bind to specific cells based on their protein cargo (Murphy et al., 2019). Thus, particular EV can be selected depending on the desired retinal cell target.

4.3. Mechanisms of action and the discovery of novel pathways

It is clear that EV contain an expansive cargo while unclear which of this cargo is responsible for the therapeutic effects observed in the above retinal diseases. It is tempting, and perhaps more feasible, to focus on clearly established pathways and delineate from this which of the EV cargo is likely responsible. However, EV also represent an opportunity to discover novel targets, particularly given most miRNA targets are untested and remain predicted rather than observed (Mead et al., 2018b).

Research has often used EV in a cross-species manner, in particular, human-derived EV in rodent models. It is unclear what interactions and effects are being excluded due to, for example, particular human-miRNA being incompatible with rodent mRNA. More studies using human EV on human cells may help refine the mechanisms or yield new candidates. If the mechanism of action can be limited to just a select few miRNA/mRNA/proteins, the treatment could be further simplified just using these particular candidates. Finally, it is currently unknown what the miRNA landscape of RGC (and their subtypes) is, as well as other specific retinal cells. This is important information considering the delivery of miRNA is one important mechanism of EV. It would be equally important to know the retinal mRNA/miRNA changes before and after EV treatment as well as under different injury conditions. Additionally, knowing the EV signaling that occurs to maintain eye homeostasis will help shape future EV therapies.

4.4. EV modification, priming, and loading

While it is clear EV are therapeutically efficacious in several disease models, how this effect can be improved further is of strong interest and may allow lower doses or less frequent administrations to be utilized. Modifying EV to better target cells of interest is one such approach and is demonstrated in a previous study involving the fusion of the exosomal protein lysosome-associated membrane protein 2 (Lamp2b) with the brain targeting peptide rabies viral glycoprotein peptide (Alvarez-Erviti et al., 2011). Subsequently generated EV were able to selectively target neurons, microglia, and oligodendrocytes in the brain after systemic administration. Priming or modifying the EV is another approach and we have recently demonstrated that by exposing MSC to the inflammatory cytokine TNF-α, the EV they release are more efficacious in the context of retinal neuroprotection (Mead et al., 2020) (Fig. 7). These “primed” EV warrant further investigation as it is expected that a cocktail of factors is required to maximally prime MSC and their EV. EV themselves can be also be modified directly, such as loaded with an abundance of a particular miRNA to increase their efficacy. This was achieved in a study described above, loading EV with miR-126 and increasing their efficacy further in a model of diabetic retinopathy (Zhang et al., 2019).

5. Conclusions

Exosomes/EV are strong candidates as a treatment for the injured retina. They circumnavigate the risk factors associated with delivering dividing cells into the eye while still possessing their multifactorial mechanism of action due to their expansive cargo. Further work is needed to characterize their mechanism of action including the mRNA, miRNA and proteins responsible alongside the myriad of therapeutic targets.