De-identified whole globes or corneal rims from human donors were obtained from Miracles in Sight (Winston-Salem, NC, USA) in accordance with the Declaration of Helsinki on research involving human tissue and with the approval of the Duke University Health System Institutional Review Board. The demographic characteristics of the human donors that contributed to this study are provided in Table 1 . Human TM cells were isolated using a blunt dissection technique, characterized, and cultured in our laboratory as previously described (30, 31). For this study, nine separate TM strains were used: six isolated from non-glaucomatous human tissue and three isolated from glaucomatous human tissue between passages 3 and 6. For sEV collection, TM monolayers were differentiated in DMEM supplemented with 1% fetal bovine serum (FBS; Thermo Fisher Scientific, Waltham, MA, USA) and maintained by culturing in DMEM supplemented with 1% exosome-depleted FBS (Thermo Fisher Scientific) for approximately 90 days. The conditioned media from TM monolayers were collected every 48 h during media exchanges and stored at −80°C before further processing.

For this study, sEVs were isolated using a gentle double iodixanol (OptiPrep™; Sigma, USA) cushion ultracentrifugation, followed by an iodixanol cushioned-density gradient ultracentrifugation (C-DGUC), as described by Li et al. (32). In brief, collected conditioned media were centrifuged at 2,000 g for 10 minutes to remove cellular debris. The supernatants were centrifuged at 10,000 g for 30 minutes at 8°C. The resulting supernatant was carefully collected and layered onto a cushion of 60% iodixanol medium. Sedimented EVs were extracted from the iodixanol cushion interfaces and diluted using particle-free Dulbecco’s phosphate-buffered saline (PBS), layered over a 60% iodixanol cushion, and centrifuged at 100,000 g. The iodixanol cushion containing isolated sEVs was collected and used as the base layer for the iodixanol gradient. Density gradient ultracentrifugation was performed, and the medium was collected from the top in 1-mL increments to create 12 fractions. The fractions were then diluted in PBS and washed. The supernatant was discarded, and the remaining pellet containing purified EVs was resuspended in lysis buffer (100 mM tris, pH 6.8, 2% sodium dodecyl sulfate (SDS) and stored at −80°C until further analysis. The sEVs are typically found in fractions in a density range of approximately 1.07 g/mL–1.11 g/mL (33); however, we also detected the enrichment of EV markers in denser fractions and opted to assess these as a separate dataset.

For each sample, approximately 8 μg of total protein was used to prepare peptide mixtures for proteomic profiling. The proteins were cleaved with the trypsin/endoproteinase Lys-C mixture (V5072; Promega, Madison, WI, USA) using the paramagnetic beads-based method (34). Each digest was dissolved in 15 μL of 1%/2%/97% (by volume) of trifluoroacetic acid/acetonitrile/water solution, and 5 μL was injected into a 5 μm×5 mm PepMap™ Neo C18 column (Thermo Scientific™) in 1% acetonitrile in water for 3 minutes at a rate of 5 μL/minute. The analytical separation was then performed using an EasySpray PepMap Neo 75 μm × 150 mm, 2 μm, C18 column (Thermo Scientific) over 90 minutes at a flow rate of 0.3 μL/minute at 35°C using the Vanquish™ Neo ultra-high-performance liquid chromatography (UHPLC) system (Thermo Scientific). The 5%–35% mobile phase B gradient was used, where phase A was 0.1% formic acid in water and phase B was 0.1% formic acid in 80% acetonitrile. The peptides separated by LC were introduced into the Q Exactive™ HF Orbitrap mass spectrometer (Thermo Scientific) using positive electrospray ionization at 1900 V and with a capillary temperature of 275°C. The data collection was performed in the data-dependent acquisition (DDA) mode with 120,000 resolutions (at m/z 200) for MS1 precursor measurements. The MS1 analysis utilized a scan from 375 m/z to 1500 m/z, with a target automatic gain control (AGC) value of 1.0e6 ions, the radiofrequency (RF) lens set at 30%, and a maximum injection time of 50 ms. Advanced peak detection and internal calibration (EIC) were enabled during data acquisition. The peptides were selected for MS/MS using charge state filtering, monoisotopic peak detection, and a dynamic exclusion time of 25 seconds with a mass tolerance of 10 ppm. MS/MS was performed using higher-energy C-trap dissociation (HCD) with a collision energy of 30% ± 0.5%, detection in the ion trap using a rapid scanning rate, an AGC target value of 5.0e4 ions, a maximum injection time of 150 ms, and ion injection for all available parallelizable time enabled.

For label-free relative protein quantification, raw mass spectral data files (.raw) were imported into Progenesis QI for Proteomics 4.2 software (Nonlinear Dynamics) for alignment of technical replicate data and peak area calculations. The peptides were identified using Mascot version 2.5.1 (Matrix Science) for searching the UniProt 2019-reviewed human database, which contains 20,243 entries. The Mascot search parameters were as follows: 10 ppm mass tolerance for precursor ions; 0.025 Da for fragment-ion mass tolerance; one missed cleavage by trypsin; a fixed modification of carbamidomethylation of cysteine; and a variable modification of oxidized methionine. Only the proteins identified with two or more peptides (i.e., Mascot scores > 15 for a peptide and > 50 for a protein corresponding to a protein confidence of p < 0.05), were included in the protein quantification analysis. To account for variations in experimental conditions and amounts of protein material in individual LC-MS/MS runs, the integrated peak area for each identified peptide was corrected using the factors calculated by the automatic Progenesis algorithm, utilizing the total intensities for all peaks in each run. The values representing protein amounts were calculated based on the sum of ion intensities for all identified constituent non-conflicting peptides. Protein abundances were averaged across the two duplicate runs for each sample.

To assess the different protein classes, we used the Protein Analysis Through Evolutionary Relationships (PANTHER) software. We pooled the proteomic datasets from each biological replicate and sorted these by abundance. We then filtered out duplicate proteins to obtain the top 100 proteins from the pooled datasets. These were then inputted into the PANTHER website and analyzed for protein class. The data exported from this was then transferred to GraphPad Prism (GraphPad Software Inc., San Diego, CA, USA) to create charts used in figures.

A standard Western blotting protocol was followed. Briefly, EV samples were solubilized in Laemmli buffer and approximately 5 µg of each sample was loaded onto SDS-polyacrylamide (PAGE) gels, separated electophoretically, and transferred to nitrocellulose membranes. The membranes were blocked with 5% bovine serum albumin (BSA) in tris-buffered saline with 0.01% Tween-20 (TBST) for 1 hour at room temperature on a rocking platform. After blocking, membranes were incubated with primary antibodies in blocking buffer at 4°C overnight. The antibodies used were CD9 (ab263019; Abcam, Cambridge, UK), TSG101 (ab125011; Abcam), Calnexin (ab133615; Abcam), albumin (ab207327; Abcam), epidermal growth factor (EGF)-like repeats and discoidin domains 3 (EDIL3; ab190692; Abcam), and fibronectin (ab6328; Abcam). Primary antibodies were removed, and membranes were washed three times for 10 minutes at room temperature in TBST. Horseradish peroxidase (HRP)-conjugated secondary antibodies [goat anti-mouse (#115-035-146) and goat anti-rabbit (#111-035-144); Jackson ImmunoResearch, West Grove, PA, USA] were added and incubated at room temperature for 1 hour. The membranes were washed again, as before, and developed using chemiluminescent reagents (SuperSignal™ West Atto; A38555; Thermo Fisher Scientific). The membranes were then imaged using the ChemiDoc™ Imaging System (BioRad) or the iBright Imaging System (Invitrogen). Protein band intensity was normalized to the concentration of particles (NTA) per protein (ug), which was measured by bicinchoninic acid (BCA) assay, as previously described (21).

The ZetaView nanoparticle tracking analysis instrument (Particle Metrix, Ammersee, Germany) was used to determine exosome vesicle diameter and estimated particle concentration. For analysis, the instrument was calibrated for size using 100-nm polystyrene beads and the sample material was diluted to a concentration of 1: 5,000 in EV-free PBS. Averages of measurements taken in triplicate from eight positions within the imaging chamber at 25°C under a 405-nm laser were used to estimate vesicle size and concentration.

Data are presented as the average (confidence interval range). The Student’s t-test was used to assess statistical significance between groups, with a p-value < 0.05 determined as being statistically significant.

The EVs were isolated from conditioned media from glaucomatous TM (GTM) cells and non-glaucomatous or "normal" TM (NTM) cells. The gradient fractions with densities of approximately 1.05 g/mL–1.10 g/mL (fractions 5–8) and approximately 1.11 g/mL–1.17 g/mL (fractions 9–10) were analyzed.

Nanoparticle tracking analysis (NTA) was used to determine the size distribution and concentration of released EVs. The results showed no significant difference in the number of EVs released from the NTM and GTM samples—although there was some variability seen between the cell strains ( Figure 1A ). The size of EVs from both the NTM and GTM cells were within the expected size range (30 nm–150 nm) for sEVs and there was no significant difference in the size distribution between the groups ( Figure 1B ).

Western blotting was conducted to determine the presence of EV markers CD9 and TSG101 in the EV preparations—both CD9 and TSG101 were found in fractions 5–8, but not as consistently in fractions 9 and 10 ( Figure 1C ). The preparations were negative for albumin and calnexin, indicating that they were free from cellular debris and thus pure sEV preparations ( Figure 1C ).

To assess the proteomic cargo of EVs isolated from TM cells, we used mass spectrometry and validated target proteins by Western blotting. Mass spectrometry was conducted on isolated EVs and the top 100 most abundant proteins from the proteomic datasets were analyzed ( Figure 2 ). The summaries of the top 100 proteins from the NTM EVs and the GTM EVs are presented in Tables 2 , 3 , respectively. The complete proteomic datasets for the NTM and GTM groups are shown in Supplementary Tables 1, 2 , respectively. The Venn diagrams demonstrate the distinct proteomic cargos of the NTM and GTM EVs—only 35 the 100 most abundant proteins in each group were shared between the groups. In both groups, fibronectin was the most abundant protein found on the sEVs from the TM cells. Many of the other most abundant proteins in each group were ECM or cytoskeleton related; however, they differed between groups. On the NTM sEVs, collagen isoforms, laminin isoforms, emilin, and other ECM proteins constituted the most abundant proteins. In contrast, on the GTM sEVs, fibrillin, plectin, and actins were the most abundant proteins. Collectively, this demonstrates the presence of ECM glycoproteins on NTM sEVs, compared with cytoskeleton- and actin-related proteins on GTM sEVs.

Next, using the PANTHER database, we determined the protein class differences between the NTM and GTM groups ( Figure 3 ) (35, 36). We analyzed the differences between all isolated sEVs ( Figures 3A, D ), as well as separating each of our sEV sub-populations into fractions 5–8 ( Figures 3B, E ) and fractions 9 and10 ( Figures 3C, F ).

From the PANTHER database, the protein class analysis in the total NTM vs. GTM dataset ( Figures 3A, D ) showed a lower percentage of extracellular matrix proteins associated with sEVs from GTM cells than from NTM cells (5.7% vs. 13.1%, respectively). In the sEV subpopulation consisting of fractions 5–8 ( Figures 3B, E ), there was an increased percentage of ECM proteins found in the GTM EVs compared with the NTM EVs (4.9% vs. 2.8%). In the sEV subpopulation consisting of fractions 9 and 10 ( Figures 3C, F ), there was a decreased percentage of ECM proteins associated with the GTM EVs compared with the NTM EVs (5.7% vs. 15.1%).

There was also a lower percentage of cell adhesion molecules (CAMs) on the GTM EVs than on the NTM EVs (3.8% vs. 11.2%) This was also demonstrated in the 5–8 sEV fraction group (GTM 3.9% vs. NTM 8.3%) and in the 9 and 10 fraction group (GTM 3.8% vs. NTM 8.5%).

However, the GTM EVs had a higher percentage of cytoskeletal proteins associated with them than NTM EVs (24.7% vs. 11.2%, respectively). This increase in cytoskeletal proteins was maintained in both the 5–8 sEV subpopulation (GTM 20.4% vs. NTM 11.1%) and in the 9 and 10 sEV subpopulation (GTM 21.0% vs. NTM 12.3%).

Two of the most abundant proteins found in sEVs across all biological replicates from NTM and GTM cells were fibronectin and EDIL3. Fibronectin is a major component of the ECM in the TM (37), and EDIL3 is a ligand of integrin αV/β3 that promotes endothelial cell adhesion and migration, promotes epithelial-to-mesenchymal transition, and is associated with both endothelial cells and the extracellular matrix (38–44). When validating these proteins, we used two unique populations of EVs from both the NTM and GTM groups—sEVs from fractions 5–8 and sEVs from fractions 9 and 10.

We validated the presence of these ECM-related proteins within the EV cargo by Western blotting. Quantification of the bands was conducted, and the band intensity was normalized to the number of particles per sample. Figure 4 shows a decrease in Fn associated with EVs from GTM compared with EVs from NTM cells in both EV populations, i.e., fractions 5–8 [fold change 0.21 (−0.18, 0.6); p = 0.1476] and fractions 9 and 10 [0.03 (−0.01, 0.06); p = 0.0011]. Figure 4 shows the decreased abundance of EDIL3 in EVs from GTM cells compared with NTM cells in both EV populations, i.e., fractions 5–8 [0.14 (−0.16, 0.44); p = 0.0707] and fractions 9–10 [0.11 (−0.06, 0.29); p = 0.0195].