Extracellular Vesicles Secreted by Corneal Epithelial Cells Promote Myofibroblast Differentiation

The corneal epithelium mediates the initial response to injury of the ocular surface and secretes a number of profibrotic factors that promote corneal scar development within the stroma. Previous studies have shown that corneal epithelial cells also secrete small extracellular vesicles (EVs) in response to corneal wounding. In this paper, we hypothesized that EVs released from corneal epithelial cells in vitro contain protein cargo that promotes myofibroblast differentiation, the key cell responsible for scar development. We focused on the interplay between corneal epithelial-derived EVs and the stroma to determine if the corneal fibroblast phenotype, contraction, proliferation, or migration were promoted following vesicle uptake by corneal fibroblasts. Our results showed an increase in myofibroblast differentiation based on α-smooth muscle actin expression and elevated contractility following EV treatment compared to controls. Furthermore, we characterized the contents of epithelial cell-derived EVs using proteomic analysis and identified the presence of provisional matrix proteins, fibronectin and thrombospondin-1, as the dominant encapsulated protein cargo secreted by corneal epithelial cells in vitro. Proteins associated with the regulation of protein translation were also abundant in EVs. This paper reveals a novel role and function of EVs secreted by the corneal epithelium that may contribute to corneal scarring.


Introduction
Corneal scarring affects over 4.9 million people worldwide as a leading cause of blindness [1] and may develop in response to injury, infection, or genetic corneal dystrophies [2]. Corneal scarring decreases vision by inducing tissue opacity that blocks light from entering the eye, and frequently distorts the shape of the cornea, which results in irregular astigmatism and higher order aberrations. While these irregularities in the cornea can often be corrected by hard contact lenses, opacities within the tissue necessitate surgery, such as corneal transplantation, in order to restore the eye's visual potential. However, corneal transplants only have a 20-year life expectancy, on average [3], and are associated with comorbidities, such as cataracts [4] and glaucoma [5]. Moreover, tissue access for corneal transplantation is limited worldwide [6]; therefore, the development of novel treatments to inhibit corneal scar formation and potentially reverse this damage remains an unmet global need.
Resident corneal cell populations (e.g., epithelium, keratocyte, neuronal, and immune cells) mediate the collective corneal tissue response to injury via autocrine and exocrine signaling that influences clinical outcome, such as corneal scar development, persistent pain, and ultimately tissue regeneration [7]. As the anterior cell surface of the eye, the corneal epithelium expresses various soluble

Transmission Electron Microscopy (TEM) Analysis
EVs were fixed and imaged by TEM, as previously described [27]. Briefly, the EV pellet was resuspended in 4% w/v paraformaldehyde (PFA) in PBS (Polysciences Inc., Warrington, PA, USA) for 30 min at room temperature for fixation. A 5-µL solution of the fixed EV pellet was added to an EM grid followed by a 20-min incubation to allow EVs to adhere to the grid surface. The grids were then washed in drops of PBS to remove residual PFA (repeat 5×) followed by resuspension in 1% v/v glutaraldehyde in PBS for 5 min. Residual glutaraldehyde was removed by gently resuspending the grid in water (repeat 7×). The grids then were transferred to a uranyl oxalate solution followed by a 10-min incubation with a methyl cellulose solution for contrast. The grid was allowed to dry before TEM imaging (JEM-1220 TEM: JEOL USA, Peabody, MA, USA).

Stimulated Emission Depletion (STED) Nanoscopy
EVs isolated from GFP-labelled hCE-TJs were suspended in PBS, applied to the surface of a glass microscopy slide, and allowed to dry for 30 min at room temperature in the dark. A drop of mounting medium (ProLong Diamond, Invitrogen) was then added to the EV smear, covered with a 1.5 glass coverslip, and allowed to incubate overnight at room temperature to permit curing. Samples were imaged on an SP8 confocal microscope with an STED 592 or 660 laser (Leica Microsystems, Bannockburn, IL, USA).

EV Uptake
PKH26-stained EVs were resuspended in complete corneal fibroblast medium and added to hCFs in 2-D culture for 12-16 h at 37 • C/5% CO 2 . Cultures were then isolated, washed with PBS 3× to remove unbound EVs, stained, and imaged by confocal microscopy. Quantification of EV uptake was performed on 10× images of each well using ImageJ (ImageJ 1.52a, National Institutes of Health, Bethesda, MD, USA) [29] and particle analysis with an adjusted threshold, circularity (0.00-1.00), and no size exclusion (0-∞).

Analysis of F-Actin Organization
The ImageJ plug-in, FibrilTool [30], was utilized to quantify the average angle and anisotropy of F-actin bundles based on the fluorescence signal detected using rhodamin-phalloidin staining. Briefly, a 40× fluorescent image detected in the red channel (λ ex 550 nm, λ em 573 nm) was opened in Fiji [31] and the region of interest was defined using the polygon tool, excluding areas with no signal. The average orientation of fibers (−90 • to 90 • ) and anisotropy (1 to 0) detected within the sample were recorded using the 'FibrilTool' function [30].

Collagen Contraction Assay
Primary hCFs (passage 3) were trypsinized and resuspended in complete corneal fibroblast medium. A 7.2 × 10 4 cells/mL suspension was added to 1 mg/mL rat-tail collagen type I (3 mg/mL stock, Gibco) neutralized with 6 µL of 1 M sodium hydroxide (Sigma Aldrich), and mixed by trituration. For EV-suspended gels, 25 µg/mL of EVs were added to the collagen mixture and mixed by trituration. Each cell/collagen mixture was then quickly added to each well of a 24-well plate and incubated for 20 min at room temperature followed by 20 min at 37 • C/5% CO 2 to allow for gelation. The gel was disassociated from the sides of the well using a sterile pipette tip and incubated with complete corneal fibroblast medium overnight at 37 • C/5% CO 2 . Brightfield images of the collagen gels were taken on an LG Android (LG-D321: LG Electronics, Seoul, South Korea) from 0-334 h. For quantification, the relative diameter of the gel and well were measured for each image using ImageJ [29]. The percent change in the normalized ratio of well/gel area was calculated based on the change in gel area over time according to the following equation: [area at time (0 h) − area at time (X)]/[area at time (0 h)] × 100.

Proliferation Assay
Primary hCFs (passage 4) were grown to 30% confluence on glass slides (Lab-Tek II Chamber Slide, Lab-Tek) in complete corneal fibroblast medium. At 24 h post-seeding, cells were treated with PBS or hCE-TJ-derived EVs (30 µg total protein) for 24 h at 37 • C/5%CO 2 followed by isolation and staining for Ki67 expression. ImageJ was used to count the cells in each sample [29], and the percent of Ki67-positive cells was determined based on the number of cells containing Ki67 staining divided by the total number of cells within the region of interest.

Migration Assay
Primary hCFs (passage 4) were seeded into 24-well plates at a cell density of 2.5 × 10 4 cells/well in complete corneal fibroblast medium. At 24 h post-seeding, a scratch wound was induced using a sterile P20 pipette tip. Brightfield images were taken at 0 and 24 h post-scratching using an inverted microscope with a 10 × objective lens (EVOS XL Core Imaging System: Life Technologies, Bothell, WA, USA).

Statistical Analysis
Statistical significance was assessed for multiple groups using a one-way or two-way analysis of variance (ANOVA) for column or grouped analysis, respectively. A comparison of two groups was analyzed using a Mann-Whitney t-test. All statistical analyses were performed in GraphPad Prism 7 for Windows (GraphPad Software, Version 7.03, La Jolla, CA, USA, www.graphpad.com). A p-value of ≤ 0.05 was considered statistically significant.

Characterization of Epithelial-Derived EVs
We began by characterizing the biochemical and functional properties of EVs secreted by hCE-TJs in vitro in order to understand how epithelial-derived EVs influence wound healing. EVs were isolated from hCE-TJ-conditioned medium using established ultracentrifugation approaches [27] and identified EVs exhibiting the characteristic cup-like structure reminiscent of red blood cells ( Figure 1A). The average size of hCE-TJ-derived EVs was 58 nm ± 32 nm, with a range of 20-340 nm. The most abundant size range of hCE-TJ-derived EVs was 50-59 nm, which is consistent with the relative size of exosomes or small microvesicles [10] ( Figure 1B). We then compared EVs isolated from unwounded control versus debrided corneal epithelium, and verified the expression of a common EV marker, CD63 [39][40][41], in isolated EVs using Western blot ( Figure 1C). Multiple bands were detected primarily from 45-60 kDa, consistent with the reported variations in the glycosylation of CD63 [42] ( Figure 1C). We also applied a super-resolution microscopy approach, termed STED nanoscopy, as an additional method to visualize isolated EVs from GFP-expressing hCE-TJs. These GFP-loaded EVs appeared as single punctate particles, as well as in small aggregates with a strong fluorescent signal ( Figure 1D). To determine the cytosolic localization of EVs in hCE-TJs, we utilized CD63 as a common marker for the exosomal surface [39][40][41]. In 2-D cultures of hCE-TJs, we identified the localization of CD63 primarily in the perinuclear region, which is the major site of exosomal loading and secretion via the endocytic pathway [43,44] (Figure S3).

EV Uptake by hCFs
EVs pre-stained with the lipophilic marker, PKH26, were added to hCFs in 2-D conventional cultures to determine whether EV uptake could be observed using confocal microscopy ( Figure 2A). We identified the presence of endocytosed EVs by 12-16 h post-application. We compared the uptake of PKH26-labelled FBS-derived EVs and hCE-TJ-derived EVs, as well as GFP-hCE-TJ-derived EVs using particle analysis to determine if the cell source influenced the uptake by hCFs. The relative uptake of PKH26-labelled FBS and hCE-TJ EVs was similar; however, the number of endocytosed GFP-labelled EVs was significantly lower compared to FBS or hCE-TJ-derived EVs (36% and 39%, respectively; p ≤ 0.001, Figure 2B). This finding may be reflective of a lower percentage of total EVs isolated from GFP-hCE-TJs with detectable levels of GFP fluorescence.

EV Uptake by hCFs
EVs pre-stained with the lipophilic marker, PKH26, were added to hCFs in 2-D conventional cultures to determine whether EV uptake could be observed using confocal microscopy ( Figure 2A). We identified the presence of endocytosed EVs by 12-16 h post-application. We compared the uptake of PKH26-labelled FBS-derived EVs and hCE-TJ-derived EVs, as well as GFP-hCE-TJ-derived EVs using particle analysis to determine if the cell source influenced the uptake by hCFs. The relative uptake of PKH26-labelled FBS and hCE-TJ EVs was similar; however, the number of endocytosed GFP-labelled EVs was significantly lower compared to FBS or hCE-TJ-derived EVs (36% and 39%, respectively; p ≤ 0.001, Figure 2B). This finding may be reflective of a lower percentage of total EVs isolated from GFP-hCE-TJs with detectable levels of GFP fluorescence. of PKH26-labelled FBS-derived EVs and hCE-TJ-derived EVs, as well as GFP-hCE-TJ-derived EVs using particle analysis to determine if the cell source influenced the uptake by hCFs. The relative uptake of PKH26-labelled FBS and hCE-TJ EVs was similar; however, the number of endocytosed GFP-labelled EVs was significantly lower compared to FBS or hCE-TJ-derived EVs (36% and 39%, respectively; p ≤ 0.001, Figure 2B). This finding may be reflective of a lower percentage of total EVs isolated from GFP-hCE-TJs with detectable levels of GFP fluorescence.  (A) EVs isolated from FBS or hCE-TJ-conditioned medium were labelled with the lipophilic dye, PKH26, and applied topically to hCFs. EVs were also isolated from GFP-expressing hCE-TJ-conditioned medium and incubated with hCFs. Then, 25 µg/mL of EVs were applied to each sample. PKH26-labelled EVs = red; GFP-labelled EVs = green; Nuclei (DNA) = blue. Scale bars = 10 µm. (B) Quantification of EV uptake was performed using particle analysis in ImageJ based on the relative signal detected in the TRITC (λ ex 550 nm, λ em 573 nm) or FITC (λ ex 495 nm, λ em 517 nm) channel for hCFs containing PKH26or GFP-labelled EVs, respectively. n = 6. Error bars represent SEM. *** p ≤ 0.001, **** p ≤ 0.0001 based on a one-way ANOVA.

Functional Effects of hCE-TJ-Derived EVs on hCFs
To determine if hCE-TJ-derived EVs promoted morphological changes in the cell structure, we analyzed the F-actin organization in the hCF control and EV-treated groups ( Figure 3). We observed differences in the cell morphology in hCFs treated with hCE-TJ-derived EVs characterized by higher cell spreading compared to a more parallel cell alignment found in the hCF controls ( Figure 3A). Quantification of the actin fibril orientation showed no significant difference in the directionality relative to the plane of view (ranging from −90 • to 90 • with respect to the x-axis) between control and EV-treated hCFs (4.98 • ± 11.05 • and 20.22 • ± 13.11 • , respectively, p = 0.320, Figure 3B). We then assessed the fractional anisotropy within groups, which is a measure of the disorder within the image with a range from 1 to 0, with 1 being completely ordered and 0 being a random distribution of actin fibrils. We found that hCFs treated with hCE-TJ-derived EVs exhibited significantly lower fractional anisotropy values compared to controls, suggesting that EV application stimulates changes in the cell cytoskeletal organization, thus affecting the total uniform directionality (0.201 ± 0.033 and 0.119 ± 0.020 for control and EV-treated groups, respectively, p = 0.036, Figure 3C).
The development of corneal scarring is largely dependent on the appearance and persistence of myofibroblasts, the contractile cell responsible for the secretion of a disorganized extracellular matrix (ECM) following wounding [8]. Based on the acute changes in F-actin distributions, we hypothesized that hCE-TJ-derived EVs may promote activation of hCFs to a more contractile phenotype. To determine if hCE-TJ-derived EVs promoted myofibroblast differentiation, we assessed the expression of α-SMA at 3 days following EV application and identified the presence of increased stress fibers in hCFs treated with hCE-TJ-derived EVs in select cells, with low expression apparent in controls ( Figure 4A). Western blot analysis verified an increase in α-SMA expression in groups treated with hCE-TJ-conditioned medium, which contains both free-and EV-bound factors, exhibiting the highest increase in myofibroblast marker expression (1.88-fold (p = 0.19) and 2.33-fold (p = 0.032) for hCF + control and debrided hCE-TJ-conditioned medium, respectively, Figure 4B,C). assessed the fractional anisotropy within groups, which is a measure of the disorder within the image with a range from 1 to 0, with 1 being completely ordered and 0 being a random distribution of actin fibrils. We found that hCFs treated with hCE-TJ-derived EVs exhibited significantly lower fractional anisotropy values compared to controls, suggesting that EV application stimulates changes in the cell cytoskeletal organization, thus affecting the total uniform directionality (0.201 ± 0.033 and 0.119 ± 0.020 for control and EV-treated groups, respectively, p = 0.036, Figure 3C). The development of corneal scarring is largely dependent on the appearance and persistence of myofibroblasts, the contractile cell responsible for the secretion of a disorganized extracellular matrix (ECM) following wounding [8]. Based on the acute changes in F-actin distributions, we hypothesized that hCE-TJ-derived EVs may promote activation of hCFs to a more contractile phenotype. To determine if hCE-TJ-derived EVs promoted myofibroblast differentiation, we assessed the expression of α-SMA at 3 days following EV application and identified the presence of increased stress fibers in hCFs treated with hCE-TJ-derived EVs in select cells, with low expression apparent in controls ( Figure 4A). Western blot analysis verified an increase in α-SMA expression in groups treated with Given that hCE-TJ-derived EVs appeared to promote the conversion of fibroblasts to myofibroblasts, we sought to determine if cell contractility was affected by the presence of epithelial-derived EVs using a free-floating collagen gel. We compared three conditions: 1) Untreated hCFs; 2) hCF and hCE-TJ-derived EVs suspended within the collagen gel; and 3) hCFs with hCE-TJ-derived EVs present in the cell culture medium ( Figure 5A). We hypothesized that EVs that have already been encapsulated throughout the gel would promote a faster rate of contraction compared to EVs dispersed within the medium and require diffusion through the hydrogel. At t = 17 h post-seeding, our results showed an initial delay in contraction in hCFs with hCE-TJ-EVs included within the medium compared to the control (10% decrease, p = 0.02, Figure 5B). Moreover, the inclusion of EVs in the medium promoted more contraction at the periphery of the gel compared to hCFs treated with EVs suspended within the collagen gel, which yielded a more uniform contraction ( Figure 5A). Interestingly, hCFs with suspended EVs showed an increase in contractility that became apparent at t = 115 h and continued to 334 h (26%, p = 0.02, Figure 5B). These results are consistent in showing that hCE-TJ-derived EVs promote a contractile cell phenotype suggestive of a myofibroblast.
Corneal wound healing requires rapid cell proliferation and migration in order to repopulate the region of injury and reduce permanent tissue damage [7]. To determine if EVs that were taken up attenuated the proliferation of hCFs, we assessed the effects of hCE-TJ-derived EVs on Ki67 expression, which is a marker for cell proliferation [46], following 24 h post-EV stimulation. A modest, though not statistically significant, change in proliferation was observed by 24 h (% Ki67-positive cells ± SEM = 16% ± 3% and 22% ± 5% in control and EV-treated groups, respectively, p = 0.4, Figure 6A). Additional studies applying BrdU incorporation at a shorter timepoint (8 h) may be useful to identify if hCE-TJ-derived EVs promote entry into the S-phase of the cell cycle.
To determine if cell migration was mitigated in response to EVs, we performed a scratch assay and evaluated the cell layer recovery at 24 h post-wounding. Likewise, we found a slight increase in cell migration, though this effect appeared to be insignificant in hCFs treated with hCE-TJ-derived EVs compared to controls (% change in area ± SEM = 77% ± 11% and 94% ± 3% in control and EV-treated groups, respectively, p = 0.2, Figure 6B). Cell proliferation likely contributed to closure of the scratch wound; however, the Ki67 proliferation assay showed no significant difference between control and EV-treated groups, suggesting that EV treatment did not accelerate closure via cell division. hCE-TJ-conditioned medium, which contains both free-and EV-bound factors, exhibiting the highest increase in myofibroblast marker expression (1.88-fold (p = 0.19) and 2.33-fold (p = 0.032) for hCF + control and debrided hCE-TJ-conditioned medium, respectively, Figure 4B,C). Given that hCE-TJ-derived EVs appeared to promote the conversion of fibroblasts to myofibroblasts, we sought to determine if cell contractility was affected by the presence of epithelialderived EVs using a free-floating collagen gel. We compared three conditions: 1) Untreated hCFs; 2) hCF and hCE-TJ-derived EVs suspended within the collagen gel; and 3) hCFs with hCE-TJ-derived EVs present in the cell culture medium ( Figure 5A). We hypothesized that EVs that have already been encapsulated throughout the gel would promote a faster rate of contraction compared to EVs dispersed within the medium and require diffusion through the hydrogel. At t = 17 h post-seeding, our results showed an initial delay in contraction in hCFs with hCE-TJ-EVs included within the medium compared to the control (10% decrease, p = 0.02, Figure 5B). Moreover, the inclusion of EVs in the medium promoted more contraction at the periphery of the gel compared to hCFs treated with EVs suspended within the collagen gel, which yielded a more uniform contraction ( Figure 5A). Interestingly, hCFs with suspended EVs showed an increase in contractility that became apparent at  Figure 5B). These results are consistent in showing that hCE-TJ-derived EVs promote a contractile cell phenotype suggestive of a myofibroblast. Corneal wound healing requires rapid cell proliferation and migration in order to repopulate the region of injury and reduce permanent tissue damage [7]. To determine if EVs that were taken up attenuated the proliferation of hCFs, we assessed the effects of hCE-TJ-derived EVs on Ki67 expression, which is a marker for cell proliferation [46], following 24 h post-EV stimulation. A modest, though not statistically significant, change in proliferation was observed by 24 h (% Ki67positive cells ± SEM = 16% ± 3% and 22% ± 5% in control and EV-treated groups, respectively, p = 0.4, Figure 6A). Additional studies applying BrdU incorporation at a shorter timepoint (8 h) may be useful to identify if hCE-TJ-derived EVs promote entry into the S-phase of the cell cycle.
To determine if cell migration was mitigated in response to EVs, we performed a scratch assay and evaluated the cell layer recovery at 24 h post-wounding. Likewise, we found a slight increase in cell migration, though this effect appeared to be insignificant in hCFs treated with hCE-TJ-derived EVs compared to controls (% change in area ± SEM = 77% ± 11% and 94% ± 3% in control and EVtreated groups, respectively, p = 0.2, Figure 6B). Cell proliferation likely contributed to closure of the scratch wound; however, the Ki67 proliferation assay showed no significant difference between control and EV-treated groups, suggesting that EV treatment did not accelerate closure via cell division.

EV Protein Cargo
To determine the protein cargo found in corneal epithelial-derived EVs, we performed proteomic analysis on EVs derived from hCE-TJs cultured in vitro. We identified over 556 proteins isolated from each biological sample, suggesting that a broad distribution of factors were encapsulated in the secreted EVs ( Figure 7A, Table S1). In addition, our results validated the presence of various proteins previously associated as exosomal markers, including cadherin and CD9 ( Figure  7B). Of interest, ECM-associated proteins were the most abundant class identified in hCE-TJ-derived EVs, including fibronectin, thrombospondin, and laminin ( Figure 7B). Cytosolic heat-shock proteins also contributed to the dominant class of proteins, along with endocytic/vesicle trafficking proteins (e.g., the Rab proteins and dynein). Importantly, very few growth factors were identified in secreted EVs, suggesting that the majority of proteins secreted by the corneal epithelium in response to injury,

EV Protein Cargo
To determine the protein cargo found in corneal epithelial-derived EVs, we performed proteomic analysis on EVs derived from hCE-TJs cultured in vitro. We identified over 556 proteins isolated from each biological sample, suggesting that a broad distribution of factors were encapsulated in the secreted EVs ( Figure 7A, Table S1). In addition, our results validated the presence of various proteins previously associated as exosomal markers, including cadherin and CD9 ( Figure 7B). Of interest, ECM-associated proteins were the most abundant class identified in hCE-TJ-derived EVs, including fibronectin, thrombospondin, and laminin ( Figure 7B). Cytosolic heat-shock proteins also contributed to the dominant class of proteins, along with endocytic/vesicle trafficking proteins (e.g., the Rab proteins and dynein). Importantly, very few growth factors were identified in secreted EVs, suggesting that the majority of proteins secreted by the corneal epithelium in response to injury, such as transforming growth factor-β, platelet-derived growth factor, interleukin-1, and tumor necrosis factor-α, are likely secreted as soluble proteins [47][48][49]. This mechanism is in agreement with studies showing that these growth factors bind directly to extracellular receptors found on the receiving cell membrane [50,51], rather than direct encapsulation via endocytosis. Thrombospondin-1, which is a known profibrotic factor that may activate latent transforming growth factor-β1 by binding to the latency-activating protein [52], was highly abundant in hCE-TJ-derived EVs. This data suggests that EV-packaged thrombospondin-1 secreted by hCE-TJs may promote stromal fibroblasts to differentiate to myofibroblasts. Furthermore, we compared control and debrided hCE-TJ and identified no significant difference in the EV protein cargo between hCE-TJ controls and scratched hCE-TJ, suggesting that in vitro debridement alone does not induce significant changes in the EV content ( Figure S5).  [53]. A full list of the proteomics data can be found in Table S1. (B) Distribution of select proteins identified in hCE-TJ-derived EV proteome. n = 3. Error bars represent SEM.
We performed a post-processing examination of our data using a proteome analysis program known as Reactome [37,38] to determine a genome-wide overview of the pathways potentially regulated by hCE-TJ-derived EV protein cargo. We found that the following significant pathways  [53]. A full list of the proteomics data can be found in Table S1. (B) Distribution of select proteins identified in hCE-TJ-derived EV proteome. n = 3. Error bars represent SEM.
We performed a post-processing examination of our data using a proteome analysis program known as Reactome [37,38] to determine a genome-wide overview of the pathways potentially regulated by hCE-TJ-derived EV protein cargo. We found that the following significant pathways were also regulated by hCE-TJ-derived EV proteins: The formation of the 43S preinitiation complex, which is involved in the initial steps of protein translation; PAK-2p34 regulation; AUF1 binding and mRNA stabilization; and translation initiation complex formation (Table 1). Interestingly, pathways related to translation and elongation, peptide chain elongation, and translation termination were also regulated by the proteins detected in hCE-TJ-derived EVs (Table 1). A number of proteins (67 out of 294) associated with protein translation were detected in hCE-TJ-derived EVs. These proteins included amino acid-conjugated tRNA ligases, e.g., alanine-, asparagine-, aspartate-, glycine-, serine-, threonine-, and tryptophan-tRNA ligase, along with eukaryotic translation initiation factor and elongation factors ( Table 2). The most abundant amino acid-linked tRNAs were alanine-and glycine-tRNA ligase (1.0 × 10 5 ± 2.0 × 10 4 and 1.0 × 10 5 ± 9.4 × 10 3 , respectively) compared to the less abundant aspartate-and tryptophan-tRNA ligases (5.9 × 10 3 ± 8.8 × 10 2 and 6.4 × 10 3 ± 8.3 × 10 2 , respectively, Table 2).

hCE-TJ-Derived EV Protein Localization
The significant abundance of provisional matrix proteins, such as fibronectin, led us to the question of whether hCE-TJ-EV application promoted the accumulation of fibronectin within the 3-D construct. Therefore, we treated hCFs in 3-D constructs with hCE-TJ-derived EVs and isolated the samples at 3 days to observe fibronectin localization ( Figure 8). Fibronectin expression was 22% lower in controls, while hCE-TJ-derived EV-treated hCFs showed an accumulation of fibronectin along the external cell body with a fibrillar staining pattern. Previous literature has indicated that fibronectin is upregulated following corneal injury (e.g., a keratectomy) in vivo, with the accumulation of this provisional matrix protein along the epithelial-stromal interface [16,54]. Our results suggest that EVs containing fibronectin may accumulate on the surface of stromal hCFs or may also promote increased secretion of this protein.

Discussion
The results of our study indicate that EVs secreted by corneal epithelial cells are bioactive and may promote myofibroblast differentiation, which further supports the growing evidence for the importance of cell-cell communication between the corneal epithelium and stroma following injury. This study characterized the functional effects of hCE-TJ-derived EVs on corneal fibroblasts, showing the promotion of α-SMA expression and increased contractility with little change in proliferation or migration. We further identified the dominant protein cargo contained in hCE-TJ-derived EVs, showing the regulation of pathways associated with translation and cell cycle control. We identified no significant difference in the protein cargo of EVs isolated from unwounded versus debrided hCE-TJ cultures, suggesting that the process of partial removal of the cell layer in vitro does not appear to drastically alter EV packaging or secretion. However, debridement of a confluent epithelial layer in vitro involves the disruption of cell junctions and is a relatively modest wound compared to debridement of the corneal epithelium in vivo, which involves damage of the nerve fiber endings and stimulation of possible immune cell activation [55,56]. Further studies are required to determine the effects of wounding on EV protein cargo following corneal epithelial debridement and keratectomy

Discussion
The results of our study indicate that EVs secreted by corneal epithelial cells are bioactive and may promote myofibroblast differentiation, which further supports the growing evidence for the importance of cell-cell communication between the corneal epithelium and stroma following injury. This study characterized the functional effects of hCE-TJ-derived EVs on corneal fibroblasts, showing the promotion of α-SMA expression and increased contractility with little change in proliferation or migration. We further identified the dominant protein cargo contained in hCE-TJ-derived EVs, showing the regulation of pathways associated with translation and cell cycle control. We identified no significant difference in the protein cargo of EVs isolated from unwounded versus debrided hCE-TJ cultures, suggesting that the process of partial removal of the cell layer in vitro does not appear to drastically alter EV packaging or secretion. However, debridement of a confluent epithelial layer in vitro involves the disruption of cell junctions and is a relatively modest wound compared to debridement of the corneal epithelium in vivo, which involves damage of the nerve fiber endings and stimulation of possible immune cell activation [55,56]. Further studies are required to determine the effects of wounding on EV protein cargo following corneal epithelial debridement and keratectomy in the complex in vivo microenvironment.
Previous work identified that corneal epithelium-derived EVs are restricted by the basement membrane in vivo [17]. Our results suggest that EVs may be an important modality for regulating the stromal response during homeostasis or wounding by promoting the activation of resident corneal fibroblasts. Thus, limiting the diffusion of these EVs may reduce scar development in circumstances where the epithelial basement membrane is damaged [21].
Proteomic analyses of cancer models [57,58], leukocytes [59], and biological fluids [27] have identified a diverse distribution of proteins found in isolated exosomes. Interestingly, numerous reports have shown that EVs contain "essential" proteins reflective of the cytosolic environment, including actin, tubulin, adhesion-related molecules, heat shock proteins, metabolic enzymes, ribosomal, and Ras signaling proteins (reviewed in [60]). These results and the conservation between species have led to suggestions of a "life-preserving" function of EVs dependent on the cell type that may mediate the required activities for proper physiological function from the host to the recipient cell [61]. Our results showed an abundance of these proteins, in addition to exosomal trafficking proteins and provisional matrix proteins, suggesting that the protein cargo contained in EVs that are taken up may directly influence the cytosolic contents of the recipient stromal fibroblast. A question remains of whether some of the proteins identified, such as cytoskeletal or transcription factors, which are normally associated with cytosolic or nuclear localization, are encapsulated in vacuoles serendipitously as a result of adhesion to lipid molecules, or rather funneled into the endocytic pathway and multivesicular bodies for active secretion.
Previous studies have identified that fibronectin-associated EVs may play an important role in influencing cell migratory behavior (i.e., directionality and speed) via interactions with cell integrins and adhesive proteins [62,63]. Further studies to determine whether the EV-associated fibronectin detected in our study influences epithelial cell migration following corneal injury are warranted. It is clear that a diverse range of functions of EVs in the cornea are likely involved in wound repair. Our study focused on an important aspect in corneal wound healing involving factors that mediate tissue closure (via the generation of myofibroblasts). However, the EV-bound factor secreted by corneal epithelial cells involved in promoting myofibroblast differentiation remains uncertain. Of note, we detected significant amounts of thrombospondin-1, which is considered a profibrotic factor as an activator of latent transforming growth factor-β1 [52], and may serve to promote myofibroblast differentiation by corneal fibroblasts. It is possible that the stromal response promoted by corneal epithelial-derived EVs may also be influenced by microRNA or the combination with protein content (free forms and encapsulated in EVs), rather than being solely protein derived. Identification of the effects of EVs secreted by the corneal epithelium may allow for new treatments that aim to reduce scar development of the cornea following epithelial injury.

Conclusions
Our results identified a potentially novel role for epithelial cell (hCE-TJ)-derived EVs in promoting the differentiation of corneal fibroblasts to myofibroblasts. The dominant proteins present in hCE-TJ-derived EVs consist of provisional matrix proteins, e.g., fibronectin and thrombospondin-1, as well as pathways associated with protein translation initiation, elongation, and termination. Further studies are required to determine whether epithelial-derived EV protein cargo is altered in response to injury in an in vivo model.