Mesenchymal Stem Cell Exosomes as Immunomodulatory Therapy for Corneal Scarring

Corneal scarring is a leading cause of worldwide blindness. Human mesenchymal stem cells (MSC) have been reported to promote corneal wound healing through secreted exosomes. This study investigated the wound healing and immunomodulatory effects of MSC-derived exosomes (MSC-exo) in corneal injury through an established rat model of corneal scarring. After induction of corneal scarring by irregular phototherapeutic keratectomy (irrPTK), MSC exosome preparations (MSC-exo) or PBS vehicle as controls were applied to the injured rat corneas for five days. The animals were assessed for corneal clarity using a validated slit-lamp haze grading score. Stromal haze intensity was quantified using in-vivo confocal microscopy imaging. Corneal vascularization, fibrosis, variations in macrophage phenotypes, and inflammatory cytokines were evaluated using immunohistochemistry techniques and enzyme-linked immunosorbent assays (ELISA) of the excised corneas. Compared to the PBS control group, MSC-exo treatment group had faster epithelial wound closure (0.041), lower corneal haze score (p = 0.002), and reduced haze intensity (p = 0.004) throughout the follow-up period. Attenuation of corneal vascularisation based on CD31 and LYVE-1 staining and reduced fibrosis as measured by fibronectin and collagen 3A1 staining was also observed in the MSC-exo group. MSC-exo treated corneas also displayed a regenerative immune phenotype characterized by a higher infiltration of CD163+, CD206+ M2 macrophages over CD80+, CD86+ M1 macrophages (p = 0.023), reduced levels of pro-inflammatory IL-1β, IL-8, and TNF-α, and increased levels of anti-inflammatory IL-10. In conclusion, topical MSC-exo could alleviate corneal insults by promoting wound closure and reducing scar development, possibly through anti-angiogenesis and immunomodulation towards a regenerative and anti-inflammatory phenotype.


Introduction
The cornea is the transparent part of the eye, essential in refracting light for normal vision. Corneal scarring, caused by various aetiologies such as infections, inflammation, trauma, and degenerative diseases [1][2][3][4][5][6][7][8], interrupts the normal passage of light into the eye, causing visual loss. Over 10 million people worldwide suffer from corneal blindness, among which 7% are due to corneal scarring [9]. Sufficient vision can be restored if the opacities in corneal scars are removed, or the density of the opacities is reduced. Visual protective effects [38], and small MSC-EVs of 50-200 nm, such as exosomes, are the most active EVs [39].
MSC exosomes (MSC-exo) are small EVs that carry a wide range of cellular molecules such as proteins, small nucleic acid fragments, lipids, and cytokines to mediate cell-cell communications [40]. Of the varied EV constituents, proteins are presently considered the main components [41]. Various studies have also reported the efficacy of MSC-exo in pre-clinical models of ocular diseases or injuries, such as experimental autoimmune uveitis [42,43], laser-induced retinal injuries [44], hyperglycemic-induced retinal inflammation [45], and choroidal neovascularization [46]. MSC-exo have also been shown to promote ocular repair and regeneration, such as retinal ganglion cell growth [47]. Such exosome-mediated effects could represent a non-cell-based (cell-free), therapeutic strategy.
It is now widely recognized that the cell source, culture, conditioning, and EV enrichment process, influence the identity and potency of MSC EV/exosome preparations [39,48]. To ensure reproducible and consistent performance, we have been using an immortalized monoclonal MSC line, cultured and conditioned under identical conditions for all our MSC-exo preparations [49]. These MSC-exo preparations have been shown to be non-tumorigenic [50] and therapeutic in various pre-clinical models of diseases such as myocardial ischemia, [35,51] liver injury [52], survival of allogeneic skin grafts [53], acute GVHD [54], cartilage injuries and osteoarthritis [28,[55][56][57], aging [58], acute irradiation toxicity [59], and psoriasis [60]. More importantly, these preparations also reportedly possess many immune modulating attributes and properties that could mitigate the stromal insult-induced inflammation. For example, MSC-exo has been shown to inhibit the assembly of terminal C5b-9 complement complex through CD59, a known inhibitor of terminal complement activation complex formation and an abundant MSC exosomal protein [61]. This inhibition attenuates complement-mediated activation of neutrophils manifested as NETosis and IL-17 production. Exosomal EDA-fibronectin has also been reported to activate TLR4 and polarize anti-inflammatory M2-like monocytes, which is implicated in MSC exosome-mediated polarization of Tregs [53,54]. Likewise, exosomal CD73 activity has been implicated in MSC exosome-induced proliferation and migration of chondrocytes, and synthesis of collagen II and TGF-β1, after cartilage injury or osteoarthritis [28,56].
We hypothesized that MSC-exo can exhibit similar wound healing and immunomodulatory properties following corneal injuries. In this study, we evaluated if our MSC-exo preparation would attenuate corneal scarring, inflammation, and neovascularization in a rat model of anterior corneal stromal injury induced by irregular phototherapeutic keratectomy (irrPTK).

In Vitro Cellular Uptake and Immunological Effects of Mesenchymal Stem Cells-Derived Exosomes
Corneal wound healing is initiated by the transition of surviving keratocytes to the repair-type fibroblasts and, at a later stage, into scar-associated myofibroblasts [62]. Although fibroblasts are not specialized immune cells, they trigger the recruitment of inflammatory cells through the secretion of chemokines and adhesion molecules [63]. Hence, we first studied the effect of MSC-exo on these two cell types. Using fluorescence microscopy and Incucyte live-cell imaging system, we observed that when fed with Alexa Fluor 488-labeled MSC-exo, CD90-positive fibroblasts, and α-SMA positive myofibroblasts were fluorescent at 4 h and reached maximum fluorescence at 72 h ( Figure 1A,B).
Next, we investigated the effects of the MSC-exo on LPS-challenged fibroblasts and myofibroblasts. The LPS stimulation increased CD90 and ACTA2 expressions in fibroblasts and myofibroblasts ( Figure 1C,D). Treatment with MSC-exo did not change the CD90 expression in both cell types at 4,8,24, and 48 h ( Figure 1C). Similarly, the MSC-exo did not alter ACTA2 expression in both cell types at 4,8,24, and 48 h ( Figure 1D).  Next, we investigated the effects of the MSC-exo on LPS-challenged fibroblasts and myofibroblasts. The LPS stimulation increased CD90 and ACTA2 expressions in fibroblasts and myofibroblasts ( Figure 1C,D). Treatment with MSC-exo did not change the CD90 expression in both cell types at 4, 8, 24, and 48 h ( Figure 1C). Similarly, the MSC-exo did not alter ACTA2 expression in both cell types at 4, 8, 24, and 48 h ( Figure 1D).
We found that the MSC-exo treatment significantly reduced the secretion of IL-8 and MCP-1 from the fibroblasts at 24 and 48 h after LPS stimulation ( Figure 1E,F), whereas CXCL1 from fibroblast was significantly suppressed at a later timepoint (48 h) in the presence of MSC-exo ( Figure 1G). We did not detect the presence of IL-1β, IL-10, and TNF-α from the fibroblasts in any treatment group. The multiplex assay and ELISA did not detect IL-1β, IL-8, IL-10, MCP-1, CXCL1, and TNF-α secretion from the myofibroblasts, even in the presence of LPS.

Dosing and Retention of Topically Applied AlexaFluo-488-Labeled Exosomes on Corneal Stroma
A similar dose of AlexaFluor 488-labeled MSC-exo (4 µg protein in 8 µL volume) was applied to epithelium-on and -off corneas. At 10 min, the epithelium-on corneas retained <25% of the fluorescence on epithelium-off corneas, and the fluorescence was not detectable by 3 h (Figure 2A,B). In contrast, the epithelium-off corneas retained not only more fluorescence but also for a much longer time of 24 h (Figure 2A,B). Cryosections of epithelium-off corneas revealed that at 24 h, the fluorescence was restricted to the bare stromal surface. No fluorescent signal was seen for the epithelium-on corneas ( Figure 2C). Together, this study demonstrated that the corneal epithelium forms an effective barrier against topically applied MSC-exo and prevents distribution of the exosomes into the corneal layer. Removal of the epithelium barriers allows topically applied MSC exosomes to enter and persist in the stroma for at least 24 h. However, the fluorescence was limited to the topmost layer of the stroma even after 24 h, suggesting the MSC-exo did not permeate the depth of the stroma and exit the stroma. and 48 h treatments with (+exo) or without MSC-exo (-exo). Myofibroblast culture did not express any of the analyzed cytokines and chemokines. * p < 0.05; ** p < 0.001.
We found that the MSC-exo treatment significantly reduced the secretion of IL-8 and MCP-1 from the fibroblasts at 24 and 48 h after LPS stimulation ( Figure 1E,F), whereas CXCL1 from fibroblast was significantly suppressed at a later timepoint (48 h) in the presence of MSC-exo ( Figure 1G). We did not detect the presence of IL-1β, IL-10, and TNF-α from the fibroblasts in any treatment group. The multiplex assay and ELISA did not detect IL-1β, IL-8, IL-10, MCP-1, CXCL1, and TNF-α secretion from the myofibroblasts, even in the presence of LPS.

Dosing and Retention of Topically Applied AlexaFluo-488-Labeled Exosomes on Corneal Stroma
A similar dose of AlexaFluor 488-labeled MSC-exo (4 µg protein in 8 µL volume) was applied to epithelium-on and -off corneas. At 10 min, the epithelium-on corneas retained <25% of the fluorescence on epithelium-off corneas, and the fluorescence was not detectable by 3 h (Figure 2A,B). In contrast, the epithelium-off corneas retained not only more fluorescence but also for a much longer time of 24 h (Figure 2A,B). Cryosections of epithelium-off corneas revealed that at 24 h, the fluorescence was restricted to the bare stromal surface. No fluorescent signal was seen for the epithelium-on corneas ( Figure 2C). Together, this study demonstrated that the corneal epithelium forms an effective barrier against topically applied MSC-exo and prevents distribution of the exosomes into the corneal layer. Removal of the epithelium barriers allows topically applied MSC exosomes to enter and persist in the stroma for at least 24 h. However, the fluorescence was limited to the topmost layer of the stroma even after 24 h, suggesting the MSC-exo did not permeate the depth of the stroma and exit the stroma.

Effects of Mesenchymal Stem Cells-Derived Exosome Treatment on Epithelial Wound Closure
The wound healing effects of MSC-exo were first assessed by an in-vitro scratch wound assay on HCET cell culture. The wound closed faster in the presence of MSC-exo treatment ( Figure 3A), with a statistically significant difference at 14-h (3.0 ± 0.3% wound area in the MSC-exo treated group versus the control, 20.9 ± 10.4%) (p = 0.041) ( Figure 3B). To confirm the in vitro results, we tested MSC-exo treatment in a rat model of acute corneal injury caused by irrPTK ( Figure 3C). The corneal epithelial wound closed more rapidly after topical treatment with MSC-exo, compared to PBS-treated control corneas ( Figure 3D). Statistically significant greater wound closure in MSC-exo group was observed on day 5 (2.3 ± 3.6% vs. 8.3 ± 5.3% wound areas; p = 0.045) ( Figure 3E). exo (green fluorescence) on epithelium-off corneal cryosections at 24 h. No fluorescent signal was detected from the epithelium-on corneas. DAPI (blue fluorescence) stained the nuclei. Scale bars: 100 µm.

Effects of Mesenchymal Stem Cells-Derived Exosome Treatment on Epithelial Wound Closure
The wound healing effects of MSC-exo were first assessed by an in-vitro scratch wound assay on HCET cell culture. The wound closed faster in the presence of MSC-exo treatment ( Figure 3A), with a statistically significant difference at 14-h (3.0 ± 0.3% wound area in the MSC-exo treated group versus the control, 20.9 ± 10.4%) (p = 0.041) ( Figure 3B). To confirm the in vitro results, we tested MSC-exo treatment in a rat model of acute corneal injury caused by irrPTK ( Figure 3C). The corneal epithelial wound closed more rapidly after topical treatment with MSC-exo, compared to PBS-treated control corneas (Figure 3D). Statistically significant greater wound closure in MSC-exo group was observed on day 5 (2.3 ± 3.6% vs. 8.3 ± 5.3% wound areas; p = 0.045) ( Figure 3E).

Effects of Topical Mesenchymal Stem Cells-Derived Exosome Treatment on Corneal Stromal Haze Development in Rat Corneas Injured by Irregular Phototherapeutic Keratectomy
After irrPTK injury, rat corneas topically treated with MSC-exo remained clear throughout the follow-up period, whereas PBS-treated control corneas showed progressive haze development ( Figure 4A). On day 5 post-injury and treatment, the MSC-exo treated group had significantly lower haze score (median = 2; IQR = 0.5) than the PBS control (median = 3; IQR = 2) (p = 0.002; Figure 4B). There was no statistically significant difference in central corneal thickness (CCT), measured on AS-OCT, between both groups on day 2 (p = 0.588) and on day 5 (the end of examination) (p = 0.654) ( Figure 4A,E). Stromal haze intensity, as measured with in vivo confocal microscopy, was reduced in MSC-exo treatment group from 65.2 ± 11.7 pixels on day 2 to 61.9 ± 16.3 pixels on day 5 but was increased in PBS control group from 81.7 ± 20.9 pixels on day 2 to 87.5 ± 18.3 pixels on day 5 ( Figure 4F). On both day 2 and 5, the differences in stromal haze intensity between the treatment and control groups were statistically significant at p = 0.017 and p = 0.004, respectively ( Figure 4F).
After irrPTK injury, rat corneas topically treated with MSC-exo remained clear throughout the follow-up period, whereas PBS-treated control corneas showed progressive haze development ( Figure 4A). On day 5 post-injury and treatment, the MSC-exo treated group had significantly lower haze score (median = 2; IQR = 0.5) than the PBS control (median = 3; IQR = 2) (p = 0.002; Figure 4B). There was no statistically significant difference in central corneal thickness (CCT), measured on AS-OCT, between both groups on day 2 (p = 0.588) and on day 5 (the end of examination) (p = 0.654) ( Figure 4A,E). Stromal haze intensity, as measured with in vivo confocal microscopy, was reduced in MSC-exo treatment group from 65.2 ± 11.7 pixels on day 2 to 61.9 ± 16.3 pixels on day 5 but was increased in PBS control group from 81.7 ± 20.9 pixels on day 2 to 87.5 ± 18.3 pixels on day 5 ( Figure 4F). On both day 2 and 5, the differences in stromal haze intensity between the treatment and control groups were statistically significant at p = 0.017 and p = 0.004, respectively ( Figure 4F).
Neovascularization score in the MSC-exo treated group (median = 1; IQR = 1) was also lower than the PBS control group (median =1; IQR = 0) (p = 0.338; Figure 4C), but this was not statistically significant. Taking both the corneal haze and neovascularization scores into consideration, the corneas that received MSC-exo had substantially better clarity and were less vascularized (median total score = 3; IQR = 1.5) compared to the injured corneas receiving PBS only (median total score = 4; IQR = 1.5) (p = 0.004; Figure 4D).  Neovascularization score in the MSC-exo treated group (median = 1; IQR = 1) was also lower than the PBS control group (median =1; IQR = 0) (p = 0.338; Figure 4C), but this was not statistically significant. Taking both the corneal haze and neovascularization scores into consideration, the corneas that received MSC-exo had substantially better clarity and were less vascularized (median total score = 3; IQR = 1.5) compared to the injured corneas receiving PBS only (median total score = 4; IQR = 1.5) (p = 0.004; Figure 4D).

Immunohistochemistry to Assess the Relative Molecular Changes in the Injured Corneas
To identify the molecular changes in MSC-exo-and PBS-treated corneas after irrPTK, the corneas were isolated. Consistent with the above observations, corneas from the MSCexo-treated group had visibly reduced haze compared to the PBS controls ( Figure 5A; top panel). Immunohistochemistry revealed that PBS-treated control corneas but not MSCexo-treated corneas were highly immunoreactive for corneal fibrosis markers, fibronectin, collagen 3A1, and α-SMA. The naive corneas were not immunoreactive for fibronectin, collagen 3A1, and α-SMA. a neovascularization (NV) score (0-2, with 2 having vascularization >2 cm over the cornea) (C), and a total score (0-6, with 6 indicating the most severe opacification and neovascularization) (D). (E) Central corneal thickness (CCT) of the corneas preoperatively and at D2 and D5 postoperatively, measured from the AS-OCT images. (F) Comparative relative haze intensity at 10-15 µm-stromal depth quantified by stromal reflectivity using IVCM, showing a significantly lower haze intensity in the group receiving MSC-exo compared to the control. * p < 0.05.

In Vivo Immunomodulatory Effects of Mesenchymal Stem Cells-Derived Exosome Treatment
Using Bio-Plex Multiplex Immunoassay, the tissue lysates of MSC-exo treated corneas treated statistically significant lower expression of pro-inflammatory cytokines, IL-1β, IL-8, and TNF-α ( Figure 6A-C, Table S2) than PBS control group on day 2 but not day 5. Anti-inflammatory cytokine, IL-10, was statistically significantly higher in MSC-exo treated corneas relative to PBS control corneas on day 2 but not day 5 ( Figure 6D, Table S2). The anti-inflammatory responses after topical MSC-exo treatment were also demonstrated by the downregulated expression of pro-inflammatory M1 macrophage-associated genes (CD80, CD86, and NOS2) on day 5, when compared to PBS control group ( Figure  7A-C). In contrast, the anti-inflammatory M2 macrophage-associated genes (CD163, CD206, and ARG1) were strongly expressed in corneas treated with MSC-exo ( Figure 7D-F). Individually, the differences in the M1 and M2 polarized macrophage gene expressions were not significantly different between MSC-exo and PBS groups. However, collectively, the M2/M1 gene ratio was significantly higher in the MSC-exo treated group compared to the PBS control group (3.4 ± 0.3 vs. 1.7 ± 0.3; p = 0.002) ( Figure 7G).
The RNA expression results were supported by the immunostaining experiments of M1 and M2 macrophage markers on corneal sections. In Figure 7H, the quantitative results showed a lower number of CD80-positive cells (27 ± 4 cells vs. 33 ± 5 cells; p = 0.006) Unlike the other cytokines, a statistically significant difference in CXCL1 chemokine between the MSC-exo treated corneas and PBS control corneas was observed only on day 5 when CXCL1 was downregulated in the MSC-exo group ( Figure 6E, Table S2). Likewise, α-MPO expression was also lower on day 5 in the MSC-exo group ( Figure 6F, Table S2). Overall, the expressions of cytokines, chemokines, and neutrophils were significantly lower or negligible in the non-injured corneas ( Figure 6A-F).
The anti-inflammatory responses after topical MSC-exo treatment were also demonstrated by the downregulated expression of pro-inflammatory M1 macrophage-associated genes (CD80, CD86, and NOS2) on day 5, when compared to PBS control group ( Figure 7A-C). In contrast, the anti-inflammatory M2 macrophage-associated genes (CD163, CD206, and ARG1) were strongly expressed in corneas treated with MSC-exo ( Figure 7D-F). Individually, the differences in the M1 and M2 polarized macrophage gene expressions were not significantly different between MSC-exo and PBS groups. However, collectively, the M2/M1 gene ratio was significantly higher in the MSC-exo treated group compared to the PBS control group (3.4 ± 0.3 vs. 1.7 ± 0.3; p = 0.002) ( Figure 7G).

Discussion
Our study demonstrated that human MSC exosomes are therapeutically efficacious against corneal injury in both in vitro and in vivo models. Specifically, we showed that topically applied MSC-exo in an excimer laser-induced rat corneal injury model improved the rate of corneal epithelial wound healing, reduced corneal haze development, and suppressed corneal neovascularization. These observations were consistent with reduced immunoreactivity for molecular markers specific for corneal fibrosis and angiogenesis in the MSC-exo treated corneas, namely, fibronectin, collagen 3A1, α-SMA, CD31, and LYVE-1. As inflammation is a major driver of fibrosis and neovascularization, we observed that there was a downregulation of pro-inflammatory M1 macrophage-associated genes with corresponding upregulation of the anti-inflammatory M2 markers in MSC-exo treated corneas. In addition, the level of pro-inflammatory cytokines IL-1β, IL-8, and TNF-α in the MSC-exo-treated corneal tissue lysate was reduced with a concomitant increase in anti-inflammatory IL-10. Overall, our results demonstrate that topically applied MSC-exo alleviates corneal insults by promoting wound healing and preventing scar development through the modulation of injury-induced inflammation towards a regenerative immune phenotype.
In our animal study, we referenced the dosage of MSC-exo for topical application on rat corneas to an earlier study where topically applied exosomes were used to reduce psoriatic inflammation [60]. The dosage was a QD topical application of 16 µg protein/cm 2 . For rat corneas with 5.5 mm diameter, an equivalent QD dosage would be 4 µg protein in 8 µL to cover the entire surface. To assess the biodistribution of topically applied exosomes on corneas, fluorescent exosomes were applied on intact corneas ("epithelium on") or injured corneas ('epithelium-off'). In 'epithelium-on' corneas, the fluorescent signals of MSC-exo were cleared within an hour, indicating that topical application of MSC-exo on normal corneas with rapid tear fluid turnover corneal surface will not persist or permeate the stroma. In the 'epithelium-off' corneas, the fluorescence permeated the stroma and decayed slowly over 24 h. As 50% fluorescence remained after 3 h, topical MSC-exo treatment was applied at 3 hourly dosing intervals for 5 days in our study. This is to ensure MSC-exo bioavailability in the corneal stromal during the acute phase of corneal injuries prior to epithelial wound closure, which usually occurs 4-5 days following corneal injuries.
When AlexaFluor 488-labeled MSC-exo was applied to cell cultures, we also observed that MSC-exo was uptaken by cultured fibroblasts and myofibroblasts, and the intracellular detection of fluorescent signals were apparent from 4 h to over 120 h, with intracellular signals peaking at 72 h. The MSC-exo treatment did not alter fibroblast phenotypes but exerted a direct immunomodulatory effect by modulating the expression and secretion of chemo-attractants, IL-8, MCP-1, and CXCL1 at around 24 to 48 h after insult. Interestingly, MSC-exo did not affect the secretion of cytokines or chemo-attractants from cultured myofibroblasts. Taking the aforementioned findings into consideration, MSC-exo therapies may thus be most beneficial clinically when applied in the acute phase of corneal insults prior to epithelium healing to modulate the production of chemo-attractants by the stromal fibroblasts, an early event committing to fibrosis development and neovascularization, prior to late myofibroblast transdifferentiation and scar formation. This also formed the basis of our in vivo study using an acute corneal haze model. It must be pointed out that in addition to corneal fibroblasts, it has been observed that myofibroblasts can also develop from other progenitors such as bone marrow-derived cells (fibrocytes) [64], Schwann cells [65], and epithelial cells [66]. Nevertheless, understanding the roles of these latter progenitors in corneal scar formation is still largely unknown, and it was beyond the scope of our study to evaluate the effects of MSC-exo on these cell types.
A key finding in this study is the dominant effect of MSC-exo in modulating the immune phenotype during a corneal injury in vivo. The injured corneas treated with MSC-exo exhibited significantly lower expressions of pro-inflammatory cytokines, IL-8 on day 2 and CXCL1 on day 5. Interestingly, the downregulated CXCL1 on day 5 may have attenuated a neutrophil influx, indicated by a correspondingly lower level of α-MPO expression in the MSC-exo treated corneas. In addition, the reduced expressions of IL-1β and TNF-α, on day 2 may have also played a role in modulating corneal fibrosis and neovascularization [67]. Furthermore, we showed that MSC-exo may enhance wound healing through immunomodulation as evidenced by (1) a macrophage polarization towards regenerative M2 macrophages with an observed higher M2/M1 macrophage-associated gene expression ratio and amplified presence of M2 over M1 macrophages, (2) reduced levels of M1 macrophage-associated pro-inflammatory cytokines, IL-1β and TNF-α, and (3) increased levels of anti-inflammatory cytokine, IL-10, in MSC-exo treated corneas compared to PBS-treated controls. The capacity of MSC exosomes to polarize macrophages toward a M2 phenotype was first reported in 2014 [53]. This effect of tissue repair and regeneration has since been implicated in animal models of osteochondral defects and osteoarthritis [28,29].
The attenuation of corneal stromal inflammation with MSC-exo treatment may have also resulted in the observed enhancement of corneal epithelial wound closure. Corneal injuries that involve the disruption of the epithelium and Bowman's membrane trigger stromal inflammatory cascades due to the permeation of cytokines and chemokines (e.g., IL-1, IGF-1, and TGF-β1) from the corneal epithelium and tear film into the stroma [67]. These factors are thought to activate fibroblast activity to recruit inflammatory cells into the stromal tissue and myofibroblasts to produce abnormal ECM, resulting in corneal opacities. The earlier closure of epithelial wound after MSC-exo treatment might reduce such 'cytokine and chemokine floodgates', thereby further modulating the corneal inflammation and haze formation. Such enhancement of epithelial wound closure, promoted by exosomes derived from corneal epithelial cells and corneal MSCs, has also been observed by other investigators [68,69].
Although designed to limit tissue injury and promote repair, the inflammatory responses of any injured tissue can be a double-edged sword, as excessive and persistent inflammation damages healthy neighboring tissue [70]. Furthermore, scarring is the pathological end-stage complication of inflammation. At present, the mainstay of treatment of inflammation is non-specific immunosuppressive medications (e.g., corticosteroids). However, such medications are associated with significant adverse effects, which include secondary infections as a result of non-specific suppression of host immune defenses. This is important in cases of corneal infections, especially in fungal keratitis, where corticosteroids are often not administered at the outset due to the risks of non-specific immunosuppression and worsening infections. Such planned delays in starting corticosteroids, however, may result in uncontrolled inflammation and ultimately, fibrosis. By immunomodulation and effecting a regenerative phenotype within the injured tissues, MSC-exo could potentially be a more effective alternative therapy for corneal insults compared to traditional immunosuppressive therapies.
There are limitations to this study which the authors acknowledge. In our experiments where we evaluated the signal retention of labeled exosomes, applied to 'epithelium on' versus 'epithelium off' naive corneas, we were working on the assumption that the emitted signals were from functional intact exosomes. We felt that signal retention was a good starting point for the determination of the dosing frequency of this novel therapy for the cornea. We also assumed that the continuous exosome exposure would maximize the potential of the MSC-exo within the relatively short but most effective therapeutic window (4-5 days before complete epithelial closure). In our in vitro experiments, we observed internalization of exosomes in the fibroblasts and myofibroblasts. Once again, this worked on the assumption that the uptake of exosomes into the target cells was important in inducing their therapeutic effects. However, we cannot rule out that the MSC-exo could also affect target cells extracellularly and sometimes even indirectly through mediator cells. Despite these assumptions, the wound-healing effects of the MSC-exo using the inferred dosing were demonstrated in our subsequent experiments. Nevertheless, we acknowledge that the optimal dosing and potentially more effective exosome delivery routes to the corneal stroma require further investigation.

Fluorescence Labeling of Exosomes
Exosomes were labeled by incubating 1 mg exosomes in 0.8 mL PBS with 1 mg AlexaFluor 488 amine-reactive probe (Thermo Fisher) in a final 1 mL volume of 0.1 M sodium bicarbonate buffer (Sigma-Aldrich) with gentle agitation and protection from light for 1 h at room temperature (~25 • C). Excess unreacted probes were removed by passing the mixture through Bio-Gel P30 gel columns (#7326231, Bio-Rad Laboratories, Hercules, CA, USA). The labeled exosomes in the flow-through were sterile filtered with 0.22 µm filters (Merck Millipore).

Primary Corneal Stromal Fibroblast and Myofibroblast Culture
Research grade cadaveric human corneal tissues procured from Lions Eye Institute for Transplant and Research (Tampa, FL, USA) were transported in Optisol-GS (Bausch and Lomb, Irvine, CA, USA) at 4 • C and cultured immediately on arrival. The central button was trephined, followed by gentle scraping to remove the corneal epithelium and endothelium. The corneal stroma was digested with collagenase I (1 mg/mL; Worthington Biochemical Corp., Lakewood, NJ, USA) in DMEM/F12 (Thermo Fisher) overnight at 37 • C. Isolated cells were harvested and cultured in DMEM/F-12 containing fetal bovine serum (FBS; 10%; Thermo Fisher) and 1% penicillin and streptomycin sulfate (Thermo Fisher) to generate stromal fibroblasts. To generate myofibroblasts, fibroblasts at passages 3 to 6 were cultured in the presence of recombinant human TGF-β1 (1 ng/mL; R&D Systems, Minneapolis, MN, USA) for 3 days.

In Vitro Inflammatory Assay by Lipopolysaccharide Treatment
Cultures of fibroblasts (n = 3) and myofibroblasts (n = 3) seeded at a density of 10 4 cells/cm 2 in a 48-well plate were incubated in SFM containing lipopolysaccharide (LPS; 100 ng/mL; Sigma-Aldrich) for 24 h, as previously described [73]. After washes, the cells were either treated with MSC-exo (4 µg protein/mL) or PBS for 120 h. At 0, 4, 8, 24, 48, and 120 h, the conditioned media were collected for enzyme-linked immunosorbent assays (ELISA) or multiplex assay to detect inflammatory marker expression, and the cells were harvested for RNA analysis.

Corneal Epithelial Scratch Wound Assay
Human SV-40 immortalized corneal epithelial cell line HCET (RCB1384, Riken Cell Bank, Ibaraki, Japan) were seeded in 24-well culture plates and grown to confluency in DMEM/F-12, supplemented with 5% FBS [74]. Scratches were made with a P200 pipette tip. After washes, the cells were either treated with MSC-exo (4 µg protein/mL; n = 4) or with PBS (n = 4). Every 2 h, for a total of 14 h, the denuded area at fixed positions was captured using a Carl Zeiss Axioplan 2 microscope (Carl Zeiss, Oberkochen, Germany), and the area was quantified with ImageJ software version 1.54a.

Time-Lapse Tracing Studies: Dosing and Retention of AlexaFluo-488-Labeled Exosomes on Corneal Stroma
All animal experimentation followed the guidelines of the Use of Animals in Ophthalmic and Vision Research, The Association for Research in Vision and Ophthalmology (ARVO) Statement and was approved by the Institutional Animal Care and Use Committee (IACUC) of SingHealth, Singapore (protocol 2018/SHS/1446). To determine the optimal dosing frequency by topically administration, the retention of MSC-exo within normal corneas of Sprague-Dawley rats (6 to 8 weeks old) were examined in both intact (corneal epithelium-on, n = 5) and de-epithelialized (epithelium-off, n = 5) conditions. Rats were anesthetized by intraperitoneal ketamine hydrochloride (80 mg/kg; Parnell Lab., Alexandria, Australia) and xylazine (12 mg/kg; Troy Lab., Glendenning, Australia). The corneal epithelium was wetted with 5% ethanol for 10 s, followed by saline rinses and scraping using a surgical blade (#15; BD Pharmingen, Franklin Lakes, NJ, USA) sparing the limbus. AlexaFluor 488-labeled MSC-exo (4 µg protein in 8 µL volume) was topically applied at time 0. Corneas were examined using a confocal laser scanning ophthalmoscope (Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany) with a lens of 30 • field of view and an excitation filter under fluorescein mode and intensity setting to 90. Time-lapse corneal pictures with labeled exosomes were recorded at time 0, and every 10 min for the first hour, then hourly for the next six hours, and then every six hours until the fluorescent signals disappeared. At least five in-focused images were chosen at each time point for Quantity One 1-D Analysis (Bio-Rad). The grey-scale images were imported, and signal area and density were measured. After background subtraction (images before treatment), the percentages of fluorescence intensity were calculated with reference to the peak intensity reading at the time of 10 min (after topical exosomes). Following rat euthanization at 24 h after topical application, the corneas were dissected and embedded in an optimal cutting temperature (OCT) compound. Serial cross-sections with a thickness of 6 µm were viewed under a Zeiss Axioplan 2 fluorescence microscope (Carl Zeiss) to detect AlexaFluor 488-tagged MSC-exo in the epithelium-on and epithelium-off corneas.

Rat Corneal Opacity Model by Irregular Phototherapeutic Keratectomy (irrPTK) and Exosome Eyedrops
Sprague-Dawley rats (6 to 8 weeks old, n = 76) were treated under general anaesthesia. The surgeries were performed by HSO and JSM. Only one eye of each rat was used for the experiments. The rat model of corneal opacity was created as previously described [75]. The breakdown of the number of rats used in each experiment was tabulated in Table S1. The eyes first received topical analgesic, lignocaine hydrochloride (1%; Pfizer, Brooklyn, NY, USA). The corneal epithelium was removed by ethanol treatment and scraping sparing the limbus. Anterior central stroma was ablated by irrPTK using a Technolas 217z excimer laser (Bausch and Lomb, Rochester, NY, USA) with the settings of a 3 mm ablation zone and an ablation depth of 15 µm. IrrPTK was performed by placing a fine mesh screen in the path of the laser after firing 50% of the pulses to induce stromal irregularity in the corneal stroma. After saline rinsing for 10 s, the corneas received topical tobramycin (1%; Alcon, Geneva, Switzerland).
MSC-exo were topically applied to the mouse corneas one hour after irrPTK. Lyophilized MSC-exo were reconstituted at a concentration of 0.5 µg protein/µL with injection water. A volume of 8 µL (representing 4 µg protein) was applied on the injured corneal surface every 3 hourly (determined by a half-life study as described below), 6 times a day for 5 days until the corneal epithelium healed. PBS drops (8 µL volume) were administered to the injured corneas of the control group.

Ophthalmic Examinations and Measurements
Rat corneas were examined 3 days prior to corneal injury or treatment to obtain the naïve reference and on days 2 and 5 postoperatively. Corneal changes were assessed using a Zoom Slit Lamp NS-2D (Righton, Tokyo, Japan). The manifestation of corneal haze and neovascularization at postoperative day 5 was graded according to the modified Hackett-McDonald cornea scoring system [76]. Corneal re-epithelialization was assessed with cobalt blue light of the slit-lamp apparatus after instillation of 2% sodium fluorescein (Bausch and Lomb). Corneal cross-section visualization was done using an RTVue anterior segmentoptical coherence tomographer (AS-OCT; Optovue Inc., Fremont, CA, USA) and the central corneal thickness (CCT) was measured as a mean of 3 measurements taken at the center and at 0.5 mm on either side, respectively [77]. Optical sections along the corneal depth were obtained by in vivo confocal microscopy using Heidelberg retinal tomography HRT3 with Rostock corneal module (Heidelberg Engineering GmbH, Germany). All corneas were examined centrally with at least 3 z-axis scans from the corneal epithelium to endothelium. Semi-quantitative analysis of stromal reflectivity at 10-15 µm depth was performed using ImageJ (National Institute of Health, Bethesda, MD, USA) after the images were tonaladjusted to map the actual pixel values using PhotoShop CC (Adobe Systems Inc., San Jose, CA, USA). Rats were sacrificed by overdosed intraperitoneal pentobarbital (Jurox, Rutherford, Australia). Both eyes were enucleated on day 2 (n = 18) and 5 (n = 58). The isolated corneas were imaged under a stereomicroscope with indirect illumination and processed for immunohistochemistry, ELISA, multiplex immunoassay, or RT-PCR. mean CT value to either housekeeping human glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or rat hypoxanthine guanine phosphoriboxyl transferase (HPRT1). Macrophage polarization M1/M2 ratio was used to evaluate the ratio of gene expression of rat M1 (CD80, CD86, and NOS2) and M2 macrophage markers (CD163, CD206, and ARG1). M2/M1 ratio is defined as follows: M2/M1 ratio = ∑ (M2 marker fold change/HPRT1)/ ∑ (M1 marker fold change/HPRT1) (2)

Enzyme-Linked Immunosorbent Assay and Multiplex Immunoassay
Rat corneas (n = 9 in each treatment group) without residual iris and scleral tissues were trimmed into small pieces and placed in 300 ul tissue lysis buffer (Thermo Fisher) containing protease inhibitor cocktail (Complete TM ; Roche, Basel, Switzerland) and 1 mM phenylmethyl sulfonylfluoride (PMSF; Sigma-Aldrich) and disrupted by sonication with 4 times 30-s bursts on ice. Three rat corneas were pooled to generate 1 biological sample. The lysate was centrifuged at 14,000× g for 15 min at 4 • C. The supernatant was collected for the expression assay of CXCL1 and α-myeloperoxidase (α-MPO) was determined using a fluorometric immunoassay (R&D Systems) following the manufacturer's instructions. The expression of IL-1β, IL-8, IL-10, MCP-1, and TNF-α was analyzed using Bio-Plex Multiplex Immunoassay (Bio-Rad). Three rat corneas were pooled to generate one biological sample. The samples were analyzed in triplicates, and the expression values were determined from standard curves. Separately, the conditioned media from fibroblast and myofibroblast cultures, challenged with LPS or not, were subjected to ELISA and multiplex immunoassay of the same markers (human-specific).

Statistical Analyses
Data were managed in Excel (Microsoft, Redmond, WA, USA) and analyzed using Statistical Program for Social Sciences (SPSS) Version 23 (IBM, Armonk, NY, USA). Differences in the distribution of continuous variables between groups were analyzed using the non-parametric Mann-Whitney U test and one-way ANOVA. The intra-class correlation coefficient (ICC) was used to evaluate inter-observer agreement levels of stromal haze grading. The significance level was set at p < 0.05

Conclusions
This pre-clinical study provides a scientific rationale for the use of MSC-exo to promote healing and prevent scarring in corneal injuries. This approach of using exosomes has the additional advantage of being acellular, negating the potential risks of allogenic tissue transplantations. From a translational perspective, the use of MSC-exo has the benefits of being standardized when manufactured and, compared to the use of other cell-based therapies, safer as they are immunologically inert and stable.