Ovarian Cancer Exosomes Trigger Differential Biophysical Response in Tumor-Derived Fibroblasts

Exosomes are cell-secreted microvesicles that play important roles in epithelial ovarian cancer (EOC) progression, as they are constantly secreted into ascites fluids. While cells spontaneously release exosomes, alterations in intracellular calcium or extracellular pH can release additional exosomes. Yet, little is known about how these exosomes compare to those that are continuously released without stimulation and how they mediate cellular activities important in cancer progression. Here, we demonstrate that chelation of extracellular calcium leads to release of chelation-induced exosomes (CI-exosomes) from OVCAR-3 EOC cells. CI-exosomes display a unique miRNA profile compared to naturally secreted exosomes (SEC-exosomes). Furthermore, treatment with CI- and SEC-exosomes leads to differential biophysical and functional changes including, adhesion and migration in EOC-derived fibroblasts that suggest the development of a malignant tumor microenvironment. This result highlights how tumor environmental factors contribute to heterogeneity in exosome populations and how different exosome populations mediate diversity in stromal cell behavior.

Next, we confirmed that CAFs rapidly internalize either exosome population using a combinatorial approach of flow cytometry and fluorescence microscopy (Fig. 1C,D). Flow cytometry and fluorescence microscopy highlight exosome uptake at short (1-hour) and long (24-hour) time points, respectively. We also sought to determine if endocytosis was primarily responsible for this internalization. Therefore, we treated fibroblasts with Dynasore, a dynamin inhibitor, to block the endocytic pathway 27 . Fibroblasts treated with 10 nM Dynasore were no longer able to uptake exosomes as effectively, suggesting that endocytosis could be a primary mechanism for internalization (Fig. 1D).
To further characterize CI-exosomes, we investigated differences in surface protein expression between these exosome populations. We ran immunoblots probing for CD63 (Fig. 1E) and secretory phospholipase A2 Group IIA (sPLA2) (Fig. 1F). Phospholipases, including sPLA2, are proteins that are found in the lipid rafts on cholesterol-rich cell and exosome membranes 28,29 . Western blot results showed that both populations of exosomes positively expressed sPLA2 and CD63. Interestingly, CI-exosomes expressed higher levels of sPLA2 compared to SEC-exosomes, suggesting that CI-exosomes may be sequestered in lipid rafts. Together, these data reveal that chelating extracellular calcium elicits the release of a subpopulation of exosomes that exhibit varying physical and molecular characteristics.
Comprehensive differences in miRNA expression in exosome populations. We next sought to determine if either population of exosomes presented unique miRNA profiles. This is important because exosome-secreted miRNAs play crucial roles in regulating post-transcriptional gene expression important in cancer progression 30,31 . Therefore, we performed microarray analysis to determine the miRNA content in CI-and SEC-exosome populations, along with OVCAR-3 cell lysates that served as a miRNA control. From the total 2,578 human miRNA probes the genechip miRNA 4.0 array identified, we limited our screening to miRNAs with log2 fold differences in expression levels and p-values <0.05 between CI-and SEC-exosomes. Hierarchical clustering and two-dimensional principal component analysis (PCA) of CI-and SEC-exosomes ( Fig. 2A,B) suggest specific differences in miRNA content between exosome populations and cell lysate. PCA also showed decreased heterogeneity in CI-exosome compared to SEC-exosome miRNA content. We then examined the number of miRNAs that were differentially expressed between each group (Fig. 2C). This analysis showed the greatest overlap in miRNA content between CI-exosomes and cell lysate (with only 450 differentially expressed miRNAs). This was in contrast to SEC-exosomes and cell lysate, which had the largest variation in miRNA content with 2,063 differentially expressed miRNAs. We also identified 1,019 miRNAs were differentially expressed between CI-and SEC-exosomes; this included 79 upregulated and 940 downregulated miRNAs. size distribution was performed using dynamic light scattering. Although average diameter was similar, the size of SEC-exosomes compared to CI-exosomes was more distributed. (C) Flow cytometry histograms of fluorescence intensity for untreated fibroblasts (gray area) and fibroblasts treated with DiI labeled exosomes (green and red areas) for one hour. The rightward histogram shift showed that fibroblasts uptake either population of exosomes. (D) Untreated and Dynasore treated CAFs were treated with DiI labeled exosomes for 24 hours at 37 °C. CAFs were then stained with DAPI and merged with 10X magnification. Representative fluorescent images show fibroblast uptake DiI labeled exosomes (top panels) is diminished upon treatment with 10 nM Dynasore (bottom panels). Scale Bar: 10 μm. Immunoblots probing for (E) CD63 and (F) sPLA2 expression in equal number of exosomes. CI-exosomes displayed a higher expression of sPLA2 compared to SEC-exosomes. Expression was normalized to cell lysate expression. Full-length blots are presented in Supplement Fig. 5.

Figure 2.
Exosome miRNA Profiling. miRNA from CI-and SEC-exosomes and OVCAR-3 cell lysates (serving as a control) were collected and analyzed. (A) Hierarchical clustering analysis and (B) PCA mapping were performed for CI-exosome, SEC-exosome, and cell lysate samples. For the PCA plot (cell lysates-red, CIexosome-green, and SEC-exosome-blue), each point represents a biological sample, the x-axis represents first principal component, and the y-axis represents second principal component. In our miRNA data, the first two principal components account for 71.9% and 16.6% of the total variability in miRNA expression, respectively. The first two principal components explain 88.5% of the variance. The observed principal components consequently led to separate groups between exosomes and cell lysates. (C) Table representing the number of miRNAs that are differentially upregulated and downregulated between population of exosomes and cell lysates. (Fig. 3)]. Both exosome populations exhibited different miRNA expression compared to cell lysates with the largest variation in SEC-exosomes (Sup. Figure 2B). SEC-exosomes also displayed a greater number of downregulated miRNAs with larger fold differences in miRNA expression compared to cell lysates.
Next, we performed more in-depth analysis of select miRNAs involved in mechanosensitive cellular responses in SEC-vs. CI-exosomes. We report expression fold changes in key miRNAs in these pathways (Fig. 3A,B). We found several overlapping miRNAs involved in these mechanosensitive pathways, including miR-429, miR-200a-c, and miR-1290. SEC-exosomes had a reduced expression of miR-429 compared to CI-exosomes. This is significant because miR-429 plays a critical role in EOC progression 33 ; downregulation of miR-429 induces tumorigenesis and patients with lower levels of miR-429 exhibit reduced overall survival 34 . We further compiled a list of miRNAs that serve as biomarkers for poor prognosis in ovarian cancer (Fig. 3A). In particular, the expression of the miR-200 family has been associated to many ovarian cancers and has been reported to serve as a biomarker of poor survival and early relapse in stage I ovarian cancers [35][36][37] . We observed that CI-exosomes have an elevated expression of miR-200a-c compared to SEC-exosomes. Taken together, our miRNA array reveals that CI-exosomes and SEC-exosomes exhibit unique miRNA expression profiles differentially regulating key cytoskeletal and adhesion pathways implicated in cancer progression. Comprehensive miRNA microarray analysis between SEC-exosome vs. CI-exosome showed that various miRNAs were differentially upand downregulated between exosome populations. (A) Analysis was used to determine Log2 fold changes after FDR correction of representative, specific miRNAs that are associated to focal adhesions, actin cytoskeleton, and epithelial ovarian cancer. The compiled list of associated miRNAs was analyzed using DIANA TOOLS. Several miRNAs that were highly regulated overlapped in presented pathways. (B) A representative list of differentially regulated miRNAs between exosome populations were plotted in the volcano plot (gray) and examples of up-or downregulated miRNAs are highlighted for specific pathways, including focal adhesions (red), actin cytoskeleton (blue), and miRNAs in EOC (yellow).
www.nature.com/scientificreports www.nature.com/scientificreports/ Exosome treatments induce actin cytoskeletal remodeling. Since tumor-derived exosomes have been implicated in cancer progression by altering stromal cell behavior in tissue and tumor microenvironments, we next sought to determine how CI-and SEC-exosomes affect the biophysical properties of resident fibroblasts 38 . Fibroblast differentiation into CAFs has been associated with morphological elongation and actin stress fiber formation so we examined these biophysical properties 39,40 . Our patient-derived CAFs naturally exhibited a spindle-like morphology consistent with other literature; 5 however, after 24-hour exosome treatment, many of these CAFs became even more elongated (Fig. 4A). The cell shape factor was quantified in ImageJ, where cell shape factors of 0 and 1 represent a line and circle, respectively. Both CI-(p = 0.0254) and SEC-(p = 0.0047) were stained for actin (red) and vinculin (green) to determine actin and focal adhesion protein organization. Scale Bar: 10 μm. (B) CAF shape factor, (C) actin fiber length, (D) actin fiber width, and (E) vinculin area were analyzed for each exosome condition (N = 3). CAF morphology was measured using ImageJ, actin fiber lengths and widths were measured using CT-Fire, and vinculin area was measured using Cell Profiler. Exosome treated CAFs revealed more elongated morphology. Quantification of actin fiber parameters showed that both exosome populations induced longer actin fibers, whereas CI-exosomes induced the development of thicker actin fibers. Vinculin area decreased upon exosome treatment and dispersed at CAF edges. Statistics calculated using Student's t-test and reported as values +/− SEM. *p < 0.05 **p < 0.005, ***p < 0.001. (2020) 10:8686 | https://doi.org/10.1038/s41598-020-65628-3 www.nature.com/scientificreports www.nature.com/scientificreports/ exosome treatments resulted in morphological elongation compared to control as indicated by significant reductions in their CAF shape factors (Fig. 4B). There was no significant difference in CSFs between exosome treatment groups (p = 0.443).
Because we quantified morphological changes in CAFs treated with CI-or SEC-exosomes, we projected that cytoskeletal fibers and focal adhesion proteins would reorganize to mechanically support changes in cell shape 41 . CAFs were stained for filamentous actin using Phalloidin and focal adhesions using vinculin antibody (Fig. 4A). Treatment with either exosome population resulted in increased alignment of actin filaments and the formation of long actin stress fibers (CI-exosomes, p = 0.0125; SEC-exosomes, p = 0.0005; CI-exosomes vs. SEC-exosomes, p = 0.3583). The longest actin stress fibers were seen in CAFs treated with SEC-exosomes (Fig. 4C). This result was consistent with their lower fibroblast shape factor, and likely indicates more polarized actin cytoskeleton after exosome treatment. CAFs treated with CI-exosomes also developed longer actin stress fibers; however, these actin stress fibers were also thicker (CI-exosomes vs. Control, p = 0.0007; CI-exosomes vs. SEC-exosomes, p = 0.0082) (Fig. 4D), suggesting that CI-exosome treatment induces actin filament bundling. These results reveal that CAF treatment with either exosome population induces more polarized actin cytoskeletal structures with differences in cytoskeletal organization.
Focal adhesions are dynamic integrin-based adhesion complexes that anchor the actin cytoskeleton to the ECM; they play a critical role in transferring environmental stimuli to the cell to alter cell shape, adhesion, and motility 42 . Based on the observed changes in cytoskeletal actin, we hypothesized that exosome treatment would alter the distribution of focal adhesions complexes. In CAFs treated with either exosome population we observed focal adhesions that more clearly aligned with actin stress fibers. These focal adhesions were often localized to the ends of actin stress fibers after exosome treatment (Fig. 4A). Using Cell Profiler, we quantified the size of focal adhesions based on the area of vinculin punctae. We observed smaller focal adhesions in fibroblasts treated with either exosome population (CI-exosomes vs. Control, p = 0.0001; SEC-exosomes vs. Control, p = 0.0011; CI-exosomes vs. SEC-exosomes, p = 0.0046) (Fig. 4E). This may indicate that untreated CAFs have more stable actin cytoskeleton with larger and more mature focal adhesions; whereas, exosome treatment leads to more dynamic actin cytoskeleton with smaller and more nascent focal adhesions.

Exosome treatments enhance random and directional migration. Previous studies demonstrated
increased migration of exosome treated cells 43 Actin myosin (actomyosin) contractility plays a critical role in actin cytoskeletal reorganization and cell motility 47 . Thus, we performed gene expression analysis of key molecules involved in this pathway 48,49 . This included myosin regulating protein MYL9, Rho family small GTPase RhoA, and Rho-associated kinases ROCK1 and ROCK2. RhoA and ROCK have both been associated with matrix-stiffness induced malignancy 50,51 and play key roles in mechanotransduction pathways 52 . We observed similarities in gene expression for CI-exosome treated CAFs and control ( Fig. 5F), suggesting that these genes are not responsible for observed biophysical differences in CAFs treated with CI-exosomes. The expression of RhoA (p = 0.0460) and ROCK1 (p = 0.0304) were both significantly reduced in CAFs treated with SEC-exosomes compared to control, suggesting these genes are more important in regulating cytoskeletal alterations in this population. Previous studies demonstrated that CAFs treated with ROCK inhibitor have smaller focal adhesions 53 and less vinculin expression; 54 these focal adhesions were less stable resulting in faster migration 54 . Therefore, the reduced ROCK1 and ROCK2 expression may contribute to the smaller vinculin punctae and increased migration observed in CAFs treated with SEC-exosomes.
Exosome treatments result in differences in CAF adhesion strength and spreading. The rapid assembly and turnover of focal adhesions are critical in cancer progression as cells migrate, adhere, and spread on multiple ECM-coated surfaces 42,55 . Focal adhesion complexes also cluster integrins to the cell surface to modulate adhesion strength to promote cell attachment and spreading on ECM-coated surfaces 56 . However, too much adhesion strength from large focal adhesions may limit the rate of focal adhesion turnover and cell migration. Consistent with this idea, less migratory control cells had larger focal adhesions and migrated more slowly than exosome treated cells, which displayed smaller but more diffuse patterns of focal adhesions (Figs. 4A,D; 5A-E). We used a centrifuge-based assay to probe for differences in adhesion strength for control and exosome treated CAFs on FN-coated surfaces. Treatment with CI-exosomes resulted in increased adhesion strength (p = 0.0350); this was surprising since the focal adhesion area per cell was actually diminished (Fig. 6A). However, CAFs treated with CI-exosomes formed thick actin stress fibers with small focal adhesions that were distributed throughout the cells; this pattern of focal adhesions may support Velcro-like interactions between actin stress fibers and the FN-coated surfaces. CAFs treated with SEC-exosomes displayed the lowest adhesion strength (Fig. 6A) (p = 0.086); this was not surprising since these cells exhibited smaller focal adhesions that were often localized to the ends of long actin stress fibers, providing fewer points of attachment with the FN-coated surfaces. www.nature.com/scientificreports www.nature.com/scientificreports/ To further investigate this dynamic adhesion process, we used time-lapsed imaging to monitor cell attachment and spreading on FN-coated surfaces. CAFs were treated with exosome populations for 24 hours, then detached, and re-seeded on ECM-coated surfaces; we report the initial adhesion rates and spreading profiles measured at multiple time points over a 12-hour period. Exosome treatment resulted in faster adhesion to FN-coated surfaces; www.nature.com/scientificreports www.nature.com/scientificreports/ this is based on increased fraction of spread cells at the initial time point (Fig. 6B). The fraction of spread cells was increased in exosome treatment groups until the 12-hour time point when the majority of cells in all groups had adhered (Fig. 6B); this data demonstrates that exosome pre-treatment enhances CAF adhesion rates. Next we probed for morphological differences in CAFs at the 12-hour time point. Interestingly, CAFs treated with CI-exosomes exhibited markedly greater spread areas (p = 0.0004), while SEC-exosome treated CAFs displayed similar spread areas as the untreated CAFs (Fig. 6C-F). This increase in spread area for CI-exosome treated cells www.nature.com/scientificreports www.nature.com/scientificreports/ could correlate to the increase in fibroblast adhesion strength, suggesting that the focal adhesions in this CAF populations are more stable and can withstand greater detachment forces. Although exosome treatments led to differences in spread area, CAFs treated with either exosome population displayed more elongated phenotypes (Fig. 6G), consistent with more pronounced CAF phenotype.

Exosomes alter crosstalk between CAFs and Ovarian Cancer Cells.
We previously showed that crosstalk with stromal cells mediates the adhesion and spreading of cancer cells and induces a more elongated mesenchymal phenotype in epithelial cancer cells 57,58 . In co-culture studies with equal numbers of OVCAR-3 cells and CAFs, we further examined how crosstalk between these two cell types is modified by exosome treatment. Time-lapsed microscopy was used to monitor how exosome treatments affect the attachment and spreading of OVCAR-3 cells over a 16-hour period in the presence of CAFs (Sup. Figure 3A-C). The spread area of OVCAR-3 cells increased with time from the initial (I) to final (F) time point for all conditions (Fig. 7A). However, OVCAR-3 cells attached more rapidly and spread more completely in co-cultures treated with CI-exosomes compared to control; this was demonstrated by significant increases in the initial (p = 0.012) and final (p = 0.0036) OVCAR-3 cell spread areas (Fig. 7Ai,ii), respectively. In contrast, treatment with SEC-exosomes did not significantly alter the initial and final spread areas of OVCAR-3 cells relative to control (Fig. 7Ai,iii). We next looked for changes in cancer cell morphology (Fig. 7B). The average cancer cell shape factor was similar for all groups (CSF ~0.8), indicating the majority of OVCAR-3 cells remain round regardless of time or exosome treatment. In part to the heterogeneity in epithelial cancer cells, we used heterogeneity analysis to examine the distribution in cancer cell shape factors for all conditions. In control cells, the distribution in cancer cell shape factors increased with time ((Coefficient of Variation) CV, I = 0.126; CV, F = 0.144), demonstrating the natural heterogeneity in cancer cell morphology after spreading (Fig. 7Bi, Sup. Table 2). The increased variation in cancer cell shape factor after spreading was also seen in co-cultures treated with SEC-exosomes (CV, I = 0.132; CV, F = 0.238) (Fig. 7Biii, Sup. Table 2). In CI-exosome treated cells, cancer cell shape factors were more distributed initially compared to the final time point. At the initial time point, the cancer cells were also much larger than in the other conditions, likely indicating that CI-exosomes prime cancer cells for more rapid adhesion (Fig. 7Aii). Immediately after CI-exosome treatment, we also observe an increase in the percentage of cells with an intermediate morphology (0.55 < CSF < 0.75); however, as these cells continue to spread, their morphology begins to become more round (CV is reduced and CSF increased). Although the majority of OVCAR-3 cells remain round for all conditions, a small number of cells develop a much more elongated phenotype after spreading (CSF < 0.5). The % of more elongated "outlier" cells was increased from ~3% in untreated controls to ~10% in co-cultures treated with SEC-exosomes (Fig. 7Biii). Taken together, our co-culture studies indicate that CI-exosomes more widely promote adhesion and spreading of cancer cells, whereas SEC-exosomes increase the percentage of highly elongated "outlier" cells.

Discussion
Paracrine interactions between EOC cells and the surrounding tissue microenvironment, with its diversity in cell types and matrix mechanics, play a critical role in directing EOC metastasis 59 . Tumor derived exosomes transfer nucleic acids and proteins between various cell types to mediate cell-cell communication and signaling pathways important in cancer progression. EOC cells are shed from the primary tumor into the peritoneum as individual cells or multi-cellular spheroids and often adhere and metastasize to the omentum, a common site for EOC metastasis 2 . Exosomes shed from the primary tumor into the ascites fluid may play an important role in preparing this niche for EOC metastasis 60 . Here, we demonstrate that an extracellular calcium perturbation, using EDTA, led to the release of a unique population of exosomes, referred as CI-exosomes, which possessed unique miRNA signatures compared to the traditionally studied secreted exosomes, referred as SEC-exosomes. Furthermore, these populations of EOC-derived exosomes elicited more pronounced CAF phenotypes characterized by biophysical and molecular alterations. Importantly, treatment with CI-exosomes resulted in critical differences in adhesion strength, important in mediating critical steps in EOC metastasis.
CAFs are myofibroblast-like cells found abundantly in tumor tissues. They play a central role in tumor growth and matrix remodeling 61 by increasing matrix deposition, cross-linking, and bundling for increased tissue stiffness 62,63 , activating mechanosensitive signaling pathways important in cancer 64 , and generating force and protease-mediated tracks that cancer cells follow during metastasis 65 . They arise from normal fibroblasts that have been activated by tumor-secreted factors to form CAFs 66 . A CAF-like phenotype is characterized by changes in cytoskeletal architecture, motility, and adhesion patterns, along with increased expression of a-smooth muscle actin (a-SMA), platelet derived growth factor receptor beta (PDGFR-B), and fibroblast activated protein (FAP) 67 . CAFs treated with either population of exosomes increased in FAP and PDGFR-B expression (data not shown). The patient-derived CAFs naturally expressed high levels of a-SMA; therefore, noticeable increases in a-SMA expression were difficult to observe after exosome treatment. However, our data demonstrated undifferentiated lung fibroblasts that had a low basal level of a-SMA showed upregulated expression upon exosome treatment (Sup. Figure 4). In our study we demonstrate that CAFs treated with exosomes isolated from EOC cells, including both CI-and SEC-exosome populations, develop more striking CAF phenotypes. This was demonstrated through changes in the actin cytoskeleton, including increased actin stress fiber formation and alterations in the patterns of focal adhesion proteins (Fig. 6). Increased actin fiber length and bundling are hallmarks of a more activated CAF-phenotype; these structural changes in tumor stromal cells are important in generating aligned tracks in the ECM that allow cancer cells to invade during metastasis 68 .
Dynamic cytoskeletal proteins, such as actin, can also directly affect the turnover of focal adhesions that anchor actin filaments to the ECM 69 . Vinculin orientation has been shown to directly regulate cell polarization, which then dictates the direction of cell migration 70,71 . Consistent with studies that link vinculin orientation and migration, SEC-exosome treatment induced more brush-like pattern of focal adhesions that decorated long www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ actin filaments that were polarized for enhanced cell migration (Fig. 4A, Fig. 5C). Fibroblast treatment with SEC-exosomes also affected the molecular expression of actin myosin contractility genes (including RhoA and ROCK1) important in cell migration (Fig. 5F); these changes were consistent with previous ROCK inhibitor studies resulting in increased stromal cell migration and wound closure 53,54 .
Fibroblast and cancer cell interactions are particularly important in EOC metastasis since ovarian cancer cells often metastasize through ascites as spheroid aggregates containing stromal cells 7,72 . Fibroblasts in these spheroids have been shown to enhance the survival of cancer cells and guide peritoneal invasion during metastasis 7 . Dynamic adhesion processes regulate the shedding of cancer cells at the primary site, as well as their attachment at the secondary site 73 . We hypothesized that exosome crosstalk between cancer cells and CAFs would also be critical in mediating these dynamic adhesion processes. We show that CAFs treated with SEC-exosomes detach faster and adhere more rapidly to FN-coated surfaces (Fig. 6A,B). The decrease in adhesion strength with the treatment of SEC-exosomes could assist CAFs to readily detach from the primary site while the more rapid adhesion turnover may promote reattachment at the secondary site. CI-exosome treatment strengthened fibroblast adhesion and promoted greater cell spreading to FN-coated surfaces (Fig. 6F), suggesting this exosome population may be important in allowing fibroblasts to firmly attach at the metastatic site. Our data suggests that tumor-derived exosomes would elicit critical biophysical changes in CAFs to guide their tissue interactions through the metastatic process.
Previous studies have illustrated how exosome miRNAs directly affect cell functional properties [74][75][76] . For instance, the increased expression of miRNA-1290 has previously been associated with enhanced cell motility and invasion 77,78 . Similar to this result, our microarray data show miRNA-1290 is highly expressed in SEC-exosomes. CAFs treated with SEC-exosomes also showed elevated motility (Fig. 3A). Conversely, our microarray results show that SEC-exosomes had lower expression of miRNA-200a-c in comparison to CI-exosomes (Fig. 3A). The miRNA-200 family is used as a diagnostic marker for EOC 36,79 . High grade EOCs often displays elevated levels of miRNA-200, which is associated with increased expression of E-cadherin 80 . Increases in E-cadherin expression in epithelial surface invaginations of the ovary have been shown to induce metaplastic changes and tumor cell proliferation 80 . A previous study has reported that the knockdown of the miRNA-200 family resulted in reduced adhesion with increased motility 81 . Based on the miRNA 200 family expression, our data similarly suggests that SEC-exosomes may be more likely to induce more migratory phenotypes in stromal cells and an epithelial-mesenchymal transition (EMT) response in EOC cells; whereas, the CI-exosomes would be more likely to mediate strong cell adhesions and mesenchymal-epithelial transition (MET) in cancer cells. MET is thought to be critically important in EOC; however, this process is not well understood and remains highly understudied.
To examine the effects of exosome-mediated crosstalk, we investigated the biophysical effects of cells that largely constitute the primary TME: ovarian cancer cells and CAFs (Fig. 7). Our co-culture model showed that the addition of either population of exosomes altered the biophysical phenotype of OVCAR-3 cells (Sup. Figure 3). Specifically, CI-exosomes promoted rapid attachment and spreading of cancer cells; whereas, SEC-exosomes increased the percentage of highly elongated "outlier" cancer cells. Studies have shown cancer cells elongating as a hallmark of EMT. Our co-culture model further suggests that SEC-exosomes could induce more distal metastasis, as cancer cells elongate and take on more migratory phenotypes as they become more mesenchymal 82 . Conversely, the increase of OVCAR-3 spread area with CI-exosomes indicates more epithelial signatures and suggests that CI-exosomes may contribute to MET and local TME reorganization.
To further elucidate the differences in exosome populations, we characterized the physical and molecular properties of CI-and SEC-exosomes. Interestingly, physical properties, such as exosome diameter (Fig. 1A,B), and molecular profiles, such as miRNA make-up, were more heterogeneous in our SEC-exosomes (Fig. 2B) in comparison to CI-exosomes (Fig. 2B). The observed homogeny in CI-exosomes could be in part to the fact that CI-exosomes remain sequestered on the cell membrane until extracellular calcium is chelated to stimulate their release; whereas our SEC-exosome population includes all exosomes spontaneously released by cells over a 48-hour period. Previous work has shown that cells can use exosomes as autocrine signaling molecules 83 and use cell adhesion molecules to dock and adhere to the plasma membranes 84 . It is well documented that cell adhesion molecules, including cadherins and integrins, are reliant on calcium signaling for activation 85 . Therefore, the downregulation of extracellular divalent cations could destabilize exosome attachments to the membrane and trigger their release. Thus, CI-exosomes could be a sub-population of SEC-exosomes released prematurely in response to perturbations in extracellular calcium.
Through biophysical functional assays, we were able to show that different exosome populations mediate functional cellular responses to direct the multistep metastatic process. These studies highlight that exosome heterogeneity, along with exosome miRNA content, contributes to the development of a more invasive EOC TME. 1% L-Glutamine, and 1% penicillin streptomycin and expanded in T-75 flasks. We performed flow cytometry for specific surface markers to confirm fibroblast phenotype according to previously published studies (Sup. Figure 1) 85,86] . CAFs received changes of media every 3 days and were split at 80% confluence. Isolated CAFs were maintained for no more than 6 passages to minimize the effects of cellular senescence and population drift, which can occur with extended culture.

Exosome collection.
To harvest SEC-exosomes, serum-free cell culture media was collected after a 48-hour culture period with 70-90% confluent OVCAR-3 cells. Exosomes were collected using standard ultracentrifugation protocol 26 . The media was centrifuged at 2,000 × g for 20 minutes at 4 °C and at 10,000 × g for 30 minutes at 4 °C to remove cell debris; the remaining media was ultracentrifuged at 100,000 × g twice for 70 minutes to pellet exosomes. After ultracentrifugation, the exosome pellet was resuspended in 100 μl PBS and labeled with DiI (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanin, Biotium, USA). To harvest CI-exosomes, cells were briefly washed with PBS to remove any residual media. OVCAR-3 cells were then treated with 10 mM of EDTA (Promega, USA) for 10 minutes in the incubator. The EDTA solution was subsequently collected and the CI-exosomes were harvested using the same standard ultracentrifugation methods to collect the SEC-exosomes. Briefly, the EDTA solution was centrifuged at 2,000 × g for 20 minutes and 10,000 × g for 30 minutes at 4 °C to remove any debris. The supernatant was then ultracentrifuged at 100, 000 × g twice for 70 minutes to pellet CI-exosomes. The CI-exosome pellet was subsequently resuspended in 100 μl PBS and labeled with DiI.
Exosome treatment. CAFs were treated at an exosome concentration of 3.8 ×10 8 exosomes/mL for each biophysical assay. CAFs were pretreated with either exosome population for 24 hours prior to experimentation. Immunoblotting. Equal number of CI-and SEC-exosomes were collected (counted using ZetaView) and immediately lysed. Lysed exosome samples were heated for 5 minutes and separated by sodium dodecyl sulfate polyacrylamide gel (SDS-PAGE) methods. Gels were then transferred onto polyvinylidene difluoride (PVDF) membranes (Thermo Fisher, USA) and probed using primary antibodies targeting for phospholipase A2 Group IIA (Abcam) and CD63 (Abcam). Bands were then visualized using Clarity ECL Western Substrate (BioRad, Hercules, CA, USA). Band intensities were measured and analyzed using Quantity One Analysis Software (BioRad).

Dynamic light scattering (ZetaView
Exosome miRNA microarray. Total exosome RNA was isolated according to the manufacturer's published protocol (Total Exosome RNA and Protein Isolation Kit, Invitrogen, USA). Briefly, exosomes were resuspended in the provided Exosome Resuspension Buffer. After acid-phenol chloroform extraction, the exosome RNA solution was washed several times and eluted through provided filter cartridges. Total RNA was also collected from OVCAR-3 cell lysates to serve as a control. 10 μg of exosome miRNA was labeled for Affymetrix array analysis using Encore Biotin Module (NuGEN Technologies, USA). The Genomics Core Facility at the Center for Genomics and Proteomics (Brown University, USA) carried out the array hybridization and analysis according to Affymetrix protocols.
Exosome uptake studies. CAFs were seeded on 12 mm glass coverslips at 50% confluency. Fibroblasts were treated with or without CI-and SEC-exosomes for 24 hours. Cells were then fixed with 4% paraformaldehyde (PFA) and stained with DAPI. Coverslips were then mounted and fluorescence images were taken at 40X magnification using an inverted Nikon Eclipse Ti microscope. Dynasore treatment. CAFs were seeded on coverslips at 50% confluency and were treated with CI-exosome and SEC-exosome. Concurrently with exosome treatment, fibroblasts were treated with 10 nM of Dynasore (Selleckchem). CAFs were incubated with the exosome-dynasore solution for 24 hours and were fixed with 4% PFA and stained with DAPI. Coverslips were then mounted and fluorescence images were taken at 40X magnification using an inverted Nikon Eclipse Ti microscope.
Flow cytometry. CAFs were treated with exosomes for one hour. Cells were detached, pelleted, suspended in FACS buffer (2% FBS and 1 mM EDTA in PBS), and fixed in 2% PFA. Samples were run on an easyCyte HT (Guava instruments) flow cytometer.
Single-cell motility. CAFs were cultured to 20-40% confluency on wells coated with 20 ng/mL of fibronectin (FN) (Alfa Aesar, USA). Wells were coated with FN because there is increased deposition of FN in ovarian cancer ECM. Fibroblasts were pre-treated with exosomes for 24 hours before collecting time-lapsed cell migration videos. To track random cell motility, nuclei were labeled with Hoechst-a fluorescent dye used to stain DNA. An inverted Nikon Eclipse Ti microscope at 10X magnification was used to collect time-lapsed images of cells over a 12-hour period at 10-minute intervals. Individual cell tracks were analyzed from microscopy videos using custom-written MATLAB code.
Quantitative real time-polymerase chain reaction (qRT-PCR). Total RNA was isolated using Purezol (BioRad) following manufacturer's recommendation. 1 μg of total RNA was converted to cDNA using the iScript cDNA synthesis kit (BioRad). cDNA was used for qRT-PCR to analyze the expression of genes listed in Sup. Actin and Vinculin Immunocytochemistry. Cells were cultured at 50-60% confluency on FN coated coverslips. CAFs were then treated with CI-or SEC-exosomes for 24 hours. Cells were then fixed with 4% PFA, permeabilized with 0.5% Triton X-100 (Fisher Bioreagents, USA), blocked with 5% bovine serum albumin (BSA) (Alfa Aesar, USA), and stained with 1:200 anti-vinculin (Invitrogen, USA) in 2% BSA. CAFs were then washed and stained with 1:200 Rhodamine Phalloidin (Invitrogen, USA) and 1:500 of Alexa Flour 488 goat anti-rabbit (Invitrogen, USA). Cells were then visualized using an inverted Nikon Eclipse Ti to obtain fluorescence images at 40X and 100X magnification.
Cell morphology. CAF shape factor was measured from 10X bright-field images. Data was quantified using ImageJ shape factor parameter defined as 4π*Area/Perimeter 2 . CAF shape factor of 1 and 0 represent a circle and line, respectively.
Adhesion strength studies. CAFs were seeded on FN coated surfaces overnight and pre-treated with CI-or SEC-exosomes for 24 hours. Cells were then labeled with 1 µM Calcein-AM (BioLegends USA), a cell-permeant fluorescent dye. The CAF attachment fraction was measured using an established centrifugal force-based adhesion assay. Briefly, the supernatant (containing cell culture media and exosomes) was replaced with adhesion buffer before an initial fluorescence reading was taken. Plates were then inverted and centrifuged using a TS-5.1-500 rotor (Beckman Coulter) at 500 rcf for 5 minutes to remove loosely adherent cells before the final fluorescence reading. The attached fraction was reported as the final reading over the initial reading.
Cell spreading studies. CAFs were cultured on FN coated surfaces to 50% confluency and were then pretreated with exosomes for 24 hours. Cells were then detached from surfaces, stained with Calcein-AM, and reseeded on FN coated surfaces. Images were taken every 30 minutes over a 12-hour period to analyze fibroblast-spreading profiles. Various cell shape parameters were measured using Image J. The cell shape factor was measured for the 12-hour time point.
Coculture studies. OVCAR-3 cells were stained with 1 µM CFSE (BioRad, USA) according to manufacturer's protocol. Equal number of CAFs and CFSE-labeled OVCAR-3 were mixed and seeded on surfaces. Coculture models were treated with or without exosomes. Time-lapsed 10X microscopy images were taken every 30 minutes over a 16-hour period to analyze various OVCAR-3 biophysical phenotypes. Initial and final cell shape factor and area measurements were done using ImageJ.
Statistics. Data are reported as mean ± standard error of the mean (SEM). At least three experiments were performed for all biophysical experiments. Student's t-test was calculated to determine experimental significance, where p < 0.05 was considered statistically significant (*p < 0.05, **p < 0.005, ##p < 0.005 ***p < 0.001). Experimental groups were compared to untreated (control) groups or time corresponding control groups. For normalized data, Kruskal-Wallis test was used to determine significance, where p < 0.05 was considered statistically significant. (*p < 0.05, **p < 0.005, ***p < 0.001).