Oncogenic H-Ras Reprograms Madin-Darby Canine Kidney (MDCK) Cell-derived Exosomal Proteins Following Epithelial-Mesenchymal Transition*

Epithelial-mesenchymal transition (EMT) is a highly conserved morphogenic process defined by the loss of epithelial characteristics and the acquisition of a mesenchymal phenotype. EMT is associated with increased aggressiveness, invasiveness, and metastatic potential in carcinoma cells. To assess the contribution of extracellular vesicles following EMT, we conducted a proteomic analysis of exosomes released from Madin-Darby canine kidney (MDCK) cells, and MDCK cells transformed with oncogenic H-Ras (21D1 cells). Exosomes are 40–100 nm membranous vesicles originating from the inward budding of late endosomes and multivesicular bodies and are released from cells on fusion of multivesicular bodies with the plasma membrane. Exosomes from MDCK cells (MDCK-Exos) and 21D1 cells (21D1-Exos) were purified from cell culture media using density gradient centrifugation (OptiPrep™), and protein content identified by GeLC-MS/MS proteomic profiling. Both MDCK- and 21D1-Exos populations were morphologically similar by cryo-electron microscopy and contained stereotypical exosome marker proteins such as TSG101, Alix, and CD63. In this study we show that the expression levels of typical EMT hallmark proteins seen in whole cells correlate with those observed in MDCK- and 21D1-Exos, i.e. reduction of characteristic inhibitor of angiogenesis, thrombospondin-1, and epithelial markers E-cadherin, and EpCAM, with a concomitant up-regulation of mesenchymal makers such as vimentin. Further, we reveal that 21D1-Exos are enriched with several proteases (e.g. MMP-1, -14, -19, ADAM-10, and ADAMTS1), and integrins (e.g. ITGB1, ITGA3, and ITGA6) that have been recently implicated in regulating the tumor microenvironment to promote metastatic progression. A salient finding of this study was the unique presence of key transcriptional regulators (e.g. the master transcriptional regulator YBX1) and core splicing complex components (e.g. SF3B1, SF3B3, and SFRS1) in mesenchymal 21D1-Exos. Taken together, our findings reveal that exosomes from Ras-transformed MDCK cells are reprogrammed with factors which may be capable of inducing EMT in recipient cells.

ases (7). Established as a central process during the early stages of development (8,9), EMT also has implications in wound healing, fibrosis and, more recently, cancer progression (10 -12). In the latter, EMT is thought to promote metastasis by triggering invasive and anti-apoptotic mechanisms in tumor cells, stimulate the cancer stem cell phenotype, and activate the tumor microenvironment via structural and biochemical modifications (13). Although, crosstalk between numerous intracellular signaling pathways are known to regulate EMT (14), it is now emerging that the EMT process can modulate the tumor microenvironment (15).
The complexity of the tumor microenvironment goes far beyond occupant epithelial cancer cells containing several nonmalignant, albeit genetically altered, heterotypic cell types (e.g. fibroblasts, endothelial cells, and immune cells) (16). Crosstalk is possible, either physically or via secretion of components such as extracellular matrix (ECM) proteins, enzymes, or paracrine signaling molecules such as growth factors and inflammatory cytokines (collectively referred to as the secretome) (17)(18)(19). Given that cancer cells at the leading tumor edge can undergo EMT and initiate metastatic lesion formation in response to signals from the microenvironment (11,20), considerable effort has been directed toward characterizing the tumor secretome (21,22). To identify extracellular modulators of EMT, which may influence tumor cell state and invasive potential, we have previously analyzed the secretome (soluble-secreted proteins) from Madin-Darby canine kidney (MDCK) and Ras-transformed MDCK (21D1) cells (23,24). This proteomic-based approach enabled an unbiased global overview of events occurring in the extracellular microenvironment. The expression of components mediating cell-cell and cell-matrix adhesion (collagen XVII, IV, and laminin 5) were attenuated, with concordant up-regulation of proteases and ECM constituents promoting cell motility and invasion (MMP-1, TIMP-1 kallikrein-6, -7, fibronectin, collagen I, fibulin-1, -3, biglycan, decorin, S100A4 and SPARC) (23,24). It is becoming increasingly clear that in addition to the soluble-secreted cytokines and chemokines that mediate cell communication at primary and secondary tumor sites (25), extracellular membranous vesicles, including exosomes, are important regulators of the tumor microenvironment (19,26,27).
Extracellular vesicles (EVs) are capable of enhancing the invasive potential of breast cancer and induce angiogenesis and metastasis in lung cancer (28,29). In addition, transfer of oncogenic potential to a recipient cell through activation of MAPK and Akt signaling pathways highlights new mechanisms of intercellular communication via EVs in the tumor microenvironment (30,31). EVs can be categorized by size with apoptotic bodies ranging up to 4000 nm in diameter, shed microvesicles and ectosomes 100 -1000 nm, and 40 -100 nm exosomes (32,33). Importantly, exosomes have been associated with modulating the immune response, controlling tumor stroma in the metastatic niche, activating signaling pathways, and transferring genetic and oncogenic information to neighboring cells (32, 34 -38). Although many functional activities have been ascribed to exosomes, it should be noted that the majority of sample preparations used for functional studies are heterogeneous in nature containing several EV types including shed microvesicles, exosomes, and apoptotic blebs. As a first step toward characterizing the specific contribution of exosomes to the tumor microenvironment, we report in this study the first protein analysis of highly purified exosomes before and after the EMT process. Comparison of MDCK exosome protein profiles following oncogenic Rasinduced EMT revealed extensive reprogramming in favor of components promoting metastatic niche formation. Additionally, enrichment of transcription and splicing factors known to induce EMT were observed in 21D1 exosomes, suggesting that a recipient cell may undergo EMT following exosome uptake.

EXPERIMENTAL PROCEDURES
Cell Culture and CCM Preparation-MDCK cells (39) and oncogenic H-Ras-transformed MDCK derivative 21D1 cells (23,24) were routinely cultured in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen, NY, USA) supplemented with 10% FCS (Invitrogen), at 37°C with 10% CO 2 . MDCK and 21D1 cells were grown to 70% confluence in DMEM containing 10% fetal calf serum (FCS), washed three times with serum-free DMEM, and left to culture in this medium at 37°C with 10% CO 2 for 24 h. Culture medium (CM) from 60 dishes of each cell line (a total of 900 ml from ϳ3 ϫ 10 8 cells) was harvested and centrifuged twice (480 ϫ g 5 min, 2000 ϫ g 10 min) to sediment floating cells and remove cellular debris. CM was centrifuged at 10,000 ϫ g for 30 min to remove shed microvesicles. The resultant supernatant was filtered using a VacuCap ® 60 filter unit fitted with a 0.1 m Supor ® Membrane (Pall Life Sciences, Port Washington, NY) and then concentrated to 1 ml concentrated culture medium (CCM) using Amicon ® Ultracel-15 centrifugal filter devices with a 5K nominal molecular weight limit (NMWL) (Merck-Millipore, MA).
Exosome Isolation Using OptiPrep™ Density Gradient-Exosomes were isolated as previously described (40). Briefly, to prepare the discontinuous iodixanol gradient, 40% (w/v), 20% (w/v), 10% (w/v) and 5% (w/v) solutions of iodixanol were made by diluting a stock solution of OptiPrep™ (60% (w/v) aqueous iodixanol from Axis-Shield PoC, Norway) with 0.25 M sucrose/10 mM Tris, pH 7.5. The gradient was formed by adding 3 ml of 40% iodixanol solution to a 14 ϫ 89 mm polyallomer tube (Microfuge ® Tube, Beckman Coulter), followed by careful layering of 3 ml each of 20 and 10% solutions, and 2 ml of the 5% solution. For each exosome preparation, CCM (1 ml) was overlaid on the gradient, and centrifugation performed at 100,000 ϫ g for 18 h at 4°C. Twelve individual 1 ml gradient fractions were collected manually (with increasing density). Fractions were diluted with 2 ml PBS and centrifuged at 100,000 ϫ g for 3 h at 4°C followed by washing with 1 ml PBS, and resuspended in 50 l PBS. Fractions were monitored for the expression of exosomal markers Alix and TSG101 by Western blotting. To determine the density of each fraction, a control OptiPrep™ gradient containing 1 ml of 0.25 M sucrose/10 mM Tris, pH 7.5 was run in parallel. Fractions were collected as described, serially diluted 1:10,000 with water, and the iodixanol concentration determined by absorbance at 244 nm using a molar extinction coefficient of 320 L g Ϫ1 cm Ϫ1 (41).
Protein Quantitation-The protein content of exosome preparations was estimated by 1D-SDS-PAGE/SYPRO ® Ruby protein staining densitometry. This method is reproducible, has a linear quantitation range over three orders of magnitude (42), and is compatible with GeLC-MS/MS (43). Briefly, 5 l sample aliquots were solubilized in SDS sample buffer (2% (w/v) sodium dodecyl sulfate, 125 mM Tris-HCl, pH 6.8, 12.5% (v/v) glycerol, 0.02% (w/v) bromphenol blue) and loaded into 1 mm, 10-well NuPAGE™ 4 -12% (w/v) Bis-Tris Precast gels (Invitrogen). Electrophoresis was performed at 150 V for 1 h in NuPAGE™ 1 ϫ MES running buffer (Invitrogen) using an XCell Sure-lock™ gel tank (Invitrogen). After electrophoresis, gels were removed from the tank and fixed in 50 ml fixing solution (40% (v/v) methanol, 10% (v/v) acetic acid in water) for 30 min on an orbital shaker and stained with 30 ml SYPRO ® Ruby (Invitrogen, NY, USA) for 30 min, followed by destaining twice in 50 ml of 10% (v/v) methanol with 6% (v/v) acetic acid in water for 1 h. Gels were imaged on a Typhoon 9410 variable mode imager (Molecular Dynamics, Sunnyvale, USA), using a green (532 nm) excitation laser and a 610BP30 emission filter at 100 m resolution. Densitometry quantitation was performed using Im-ageQuant software (Molecular Dynamics) to determine protein concentration relative to a BenchMark™ Protein Ladder standard of known protein concentration (1.7 g/l) (Invitrogen). The yield of purified exosomes was ϳ60 g from 3 ϫ 10 8 cells for both MDCKand 21D1-Exos.
GeLC-MS/MS-MDCK-and 21D1-Exos (20 g) were lysed in SDS sample buffer, and proteins separated by SDS-PAGE and visualized by Imperial™ Protein Stain (Thermo Fisher Scientific), according to manufacturer's instructions. Gel lanes were cut into equal slices (33 ϫ 2 mm) using a GridCutter (The Gel Company, San Francisco, CA) and individual gel slices were subjected to in-gel reduction, alkylation and trypsinization (45). Briefly, gel bands were reduced with 10 mM DTT (Calbiochem, San Diego, CA) for 30 min, alkylated for 20 min with 25 mM iodoacetic acid (Fluka, St. Louis, MO), and digested with 150 ng trypsin (Worthington Biochemical Corp, Freehold, NJ) for 4.5 h at 37°C. Tryptic peptides were extracted with 50 l 50% (v/v) acetonitrile, 50 mM ammonium bicarbonate, concentrated to ϳ10 l by centrifugal lyophilization and one technical replicate analyzed by LC-MS/MS. RP-HPLC was performed on a nanoAcquity ® (C18) 150 ϫ 0.15-mm-internal diameter reversed phase UPLC column (Waters, Milford, MA) using an Agilent 1200 HPLC coupled online to an LTQ-Orbitrap mass spectrometer equipped with a nanoelectrospray ion source (Thermo Fisher Scientific). The column was developed with a linear 60 min gradient with a flow rate of 0.8 l/min at 45°C from 0 -100% solvent B where solvent A was 0.1% (v/v) aqueous formic acid and solvent B was 0.1% (v/v) aqueous formic acid/60% acetonitrile. Survey MS scans were acquired with the resolution set to a value of 30,000. Real time recalibration was performed using a background ion from ambient air in the C-trap (46). Up to five selected target ions were dynamically excluded from further analysis for 3 min. An additional biological replicate of MDCK-and 21D1-Exos (20 g) was analyzed on an LTQ-Orbitrap mass spectrometer (supplemental Data) to validate our primary findings. Raw mass spectrometry data is deposited in the PeptideAtlas and can be accessed at http://www. peptideatlas.org/PASS/PASS00225 (47)(48)(49).
Database Searching and Protein Identification-Peak lists were extracted using extract-msn as part of Bioworks 3.3.1 (Thermo Fisher Scientific). The parameters used to generate the peak lists were as follows: minimum mass 700, maximum mass 5000, grouping tolerance 0.01 Da, intermediate scans 200, minimum group count 1, 10 peaks minimum and total ion current of 100. Peak lists for each LC-MS/MS run were merged into a single MGF file for Mascot searches. Automatic charge state recognition was used because of the high-resolution survey scan (30,000). MGF files were searched using the Mascot v2.2.01 search algorithm (Matrix Science) against the LudwigNR_Q410 database with a taxonomy filter for human, cow, and dog, comprising 13112897 entries (http://www.ludwig.edu.au/ archive/LudwigNR/LudwigNR.pdf). The search parameters consisted of carboxymethylation of cysteine as a fixed modification (ϩ58 Da), NH 2 -terminal acetylation (ϩ42 Da) and oxidation of methionine (ϩ16 Da) as variable modifications. A peptide mass tolerance of Ϯ20 ppm, #13C defined as 1, fragment ion mass tolerance of Ϯ0.8 Da, and an allowance was made for up to two missed tryptic cleavages. Protein identifications were firstly clustered and analyzed by an in-house developed program MSPro (50). Briefly, peptide identifications were deemed significant if the Ion score was Ն the Homology score. False-positive protein identifications were estimated by searching MS/MS spectra against the corresponding reverse-sequence (decoy) database (50). MDCK-and 21D1-exosome protein identifications were based on a protein score above the 1% false discovery rate cut-off of 48, and with at least two significant peptides. The BioMart data-mining tool (http://www.ensembl.org/biomart/index.html) was used to obtain Ensembl protein description and gene name as described (51). UniProt (http://www.uniprot.org) and Protein Information Resource (http://pir.georgetown.edu) were used to obtain gene ontology (GO) annotation.
Semiquantitative Label-free Spectral Counting-Significant spectral count fold change ratios (R SC ) were determined using a modified formula from a previous serial analysis of gene expression study by Beissbarth et al. (52).
where, n is the significant protein spectral count (a peptide spectrum is deemed significant when the Ion score Ն the Homology score), t is the total number of significant spectra in the sample, and f a correction factor set to 1.25 (53). Total number of spectra was only counted for significant peptides identified (Ion score Ն Homology score). When Rsc is less than 1, the negative inverse Rsc value was used. The number of significant assigned spectra for each protein was used to determine whether protein abundances between the two categories (MDCK-and 21D1-Exos). For each protein the Fisher's Exact test was applied to significant assigned spectra. The resulting p values were corrected for multiple testing using the Benjamini-Hochberg procedure (54) and computations carried out in R (55).

RESULTS AND DISCUSSION
Exosomes are Released from MDCK Cells Following Oncogenic H-Ras Induced EMT-Previously, we established that cultured MDCK cells undergoing oncogenic H-Ras mediated EMT (21D1 cells) secrete protein components that extensively remodel the extracellular microenvironment, e.g. increased expression of ECM proteins, migration factors, and proteases that promote cell motility and invasion (23,24). MDCK cells exhibit cobblestone-like morphology while 21D1 cells displayed a spindle-shaped mesenchymal phenotype (Fig. 1A). To maintain the mesenchymal phenotype, 21D1 cells require coculture with their own culture medium (supplemental Fig.  S1). To isolate exosomes, MDCK and 21D1 cells were cultured to ϳ70% confluence, washed with DMEM, and then left to culture in serum-free medium for 24 h. We have previously shown that both cell lines remain greater than 96% viable during this time (23). Culture medium (CM) from ϳ3 ϫ 10 8 cells was harvested, concentrated (CCM) by centrifugal membrane ultrafiltration and crude exosomes were fractionated based upon their buoyant density into 12 fractions using iodixanol density gradient centrifugation (40) as outlined in Fig. 1B. Western blot analysis of these fractions revealed enrichment of exosomes (based on exosome markers Alix/ PDCD6IP and TSG101) in a fraction with buoyant density 1.09 g/ml ( Fig. 2A, Supplemental Fig. S2). Interestingly, H-Ras was found in both MDCK-and 21D1-Exos, but with much higher levels in 21D1-Exos ( Fig. 2A). The existence of H-Ras in the MDCK-Exos suggests that endogenously expressed Ras is implicated in secretory exosomal trafficking; however, it is not clear which form of H-Ras this is (inactive Ras-ADP or active Ras-ATP). Given that v-H-Ras expressed in 21D1 cells is the mutated active form, and that higher levels are observed in 21D1-Exos, it is suggestive of the involvement of the membrane-bound active Ras-ATP form in the secretory 21D1-Exos. This is consistent with the findings regarding K-Ras by Demory Beckler et al., (56). The yield of purified exosomes was ϳ60 g from 3 ϫ 10 8 cells for both MDCK-and 21D1-Exos. Cryo-EM of purified exosomes revealed a relatively homogenous population of round membranous vesicles 40 - 100 nm in size, which is in accordance with the typical size reported for exosomes (Fig. 2B) (33).
EMT Hallmark Proteins are Observed in Exosomes Following Oncogenic H-Ras Induced EMT-We next examined whether the pattern of EMT hallmarks typically seen in whole cells (7) are reflected in exosomes released from MDCK cells following H-Ras modulated EMT. For this purpose we used relative spectral count ratios (Rsc) and Western immunoblotting to indicate differential protein expression between samples. Proteins mediating cell-cell contact, cell-matrix contact and cell polarity displayed decreased expression levels in 21D1-Exos (Table I and Fig. 2A), correlating with typical EMT hallmarks seen in whole cells (7). Foremost of these were the adhesive glycoprotein and inhibitor of angiogenesis thrombospondin 1 (THBS1 Rsc -482.8), and the epithelial cell markers E-cadherin (CDH1 Rsc -34.4) and EpCAM (Rsc -16.5). Consistent with these findings were the elevated protein expression levels in 21D1-Exos of vimentin (VIM, Rsc 8.1) and matrix metalloproteins, MMP-1 (Rsc 7.3), MMP-19 (Rsc 11.3) and MMP-14 (Rsc 3.4), typically observed in mesenchymal cells. Confirmatory data for the different abundance levels of CDH1, EpCAM, VIM, and MMP-1 observed in MDCK-and 21D1-Exos was obtained by Western blot analysis ( Fig. 2A).
c Protein abundance ratio (ratio of spectral counts; Rsc) reveals differential protein abundance between MDCK-and 21D1-Exos based on Eq.1. The use of zero spectra is overcome using an arbitrary correction factor (1.25) in Eq. 1. The use of this correction factor allows relative quantitation of all proteins within both normalized datasets to be performed, based upon Old et al. ͓53͔. Positive Rsc values reflect increased protein abundance in 21D1-Exos relative to MDCK-Exos; negative values indicate decreased abundance in 21D1-Exos relative to MDCK-Exos.
* Differential expression with p values Ͻ0.05 as reported in supplemental Table S1. genesis, tumor invasion, and establishment of metastatic regions at secondary sites (78). Presence of MMP-1 in human colorectal carcinomas correlates with the depth grading of tumor invasion, lymphatic invasion, and lymph node metastasis (79). MMP-14 promotes cell invasion and motility by pericellular ECM degradation, shedding of CD44 (also detected in 21D1-Exos) and syndecan 1, and through activation of ERK (80). Expression of MMP-19 is associated with increased invasion, migratory behavior and early metastasis of melanoma cells (81), and localization of MMP-14 and -19 at the invasive tumor front is characteristic of highly motile invading tumor cells (81,82). The finding that MMP-14 and -19 are unique to 21D1-Exos and not observed in our previously published MDCK/21D1 secretome analysis (24), may represent a mechanism that allows exosome-bound proteases to traffic and function at distant/metastatic sites. ADAM proteases contain MMP-like catalytic domains (83) and are important mediators of cell surface protein shedding during tumor progression (84). Interestingly, ADAM10 has been shown to be an active vesicle-based protease, cleaving cell adhesion molecule L1 at the cell surface, and subsequently promoting cell migration (85). Given that other substrates of ADAM10 include components of the ECM, epidermal growth factors, chemokines, cytokines, and Notch receptor when bound to its ligands Delta-like 1 or Jagged-1 (also unique to 21D1-Exos, Rsc 4.9) (84), ADAM10 has the ability to extensively modify the tumor microenvironment. Likewise, ADAMTS1 is also capable of degrading various ECM components (86), and increased expression promotes pulmonary metastasis of mammary carcinoma and Lewis lung carcinoma cells (87). ADAMTS1 has also been shown to modulate the metastatic tumor microenvironment by promoting angiogenesis and invasion in osteoclastogenesis (88). These findings suggest that addition to soluble proteases, exosome-associated proteases ADAM10 and ADAMTS1 may also contribute to the EMT process (78) and, additionally, play a role in premetastatic niche formation (19).
Integrins-Integrins represent another class of metastatic niche components that were enriched in 21D1-Exos (Table II). Integrins facilitate cell attachment to surrounding ECM, initiating intracellular signaling cascades that maintain cell survival, proliferation, adhesion, migration and invasion (89). The finding of enriched protein levels of integrins in 21D1-Exos is of particular significance given that a study of ovarian carcinoma identified that collagen-induced activation of integrin receptors caused Ras, Erk and Akt pathway activation (90). In particular, integrins subunits ␣3, ␣6, ␣V and ␤1, all of which were enriched in 21D1-Exos, have been associated with modulating ECM-induced signaling leading to proliferation, adhesion, migration and invasion of the ovarian cancer cells (90). Moreover, ␣V mediates latent TGF-␤ activation, which is required for the maintenance of EMT and tumor cell invasion and dissemination (91). These findings are in accord with our earlier studies showing plasma membrane bound integrins  Table S1).
c Protein abundance ratio (ratio of spectral counts; Rsc) reveals differential protein abundance between MDCK-and 21D1-Exos based on Eq.1. The use of zero spectra is overcome using an arbitrary correction factor (1.25) in Eq. 1. The use of this correction factor allows relative quantitation of all proteins within both normalized datasets to be performed, based upon Old et al. ͓53͔. Positive Rsc values reflect increased protein abundance in 21D1-Exos relative to MDCK-Exos; negative values indicate decreased abundance in 21D1-Exos relative to MDCK-Exos.
* Differential expression with p values Ͻ0.05 as reported in supplemental Table S1.
␣6␤1 and ␣3␤1 were significantly enriched in cell membrane preparations of H-Ras transformed MDCK cells (21D1 cells) when compared with parental MDCK cells (51), further studies are required to ascertain whether these integrins are integral components of the 21D1-Exos membrane.
Tetraspanins-Tetraspanins are characterized by four transmembrane domains, intracellular N-and C termini and two extracellular domains. They are reported to function as scaffolding proteins, which interact with integrins; many tetraspanins have been implicated in tumor progression (92)(93)(94). In this study, we observed an enrichment in 21D1-Exos of tetraspanins involved in cancer progression including CD81, CD82, and CD151 (Table II). Interestingly, it has been previously shown that interactions between ␣6␤1 (both integrin components identified in this study) and CD81 may up-regulate cell motility, affecting migration mediated by other integrins (95). Recently, CD81-positive fibroblast-derived exosomes, isolated using differential ultracentrifugation, were reported to regulate breast cancer cell protrusions and motility through Wnt-planar cell polarity signaling (96). Further, CD82 has been implicated in integrin-mediated functions including cell motility and invasiveness (97), while CD151 has been shown to promote cancer cell metastasis via integrins ␣3␤1 and ␣6␤1 (also seen in our study) in vitro (98).
Annexins-Annexins are involved in a diverse array of cellular functions and physiological processes including membrane scaffolding, trafficking and organization of vesicles, exocytosis, endocytosis, and cell migration (99). In this study, we observed increased expression levels of annexins A1, A2, A4, A7, A8, and A11 (Rsc 1.3-2.3) in 21D1-Exos (Table II). In particular, annexin A2 (Rsc 1.9), has been shown to regulate the tumor microenvironment by inducing the remodelling of cytoskeletal structures and actin of breast and colorectal cancer cells (100). siRNA-based experiments have recently demonstrated that annexin A2 is critical in determining the invasive potential of cancer cells, and regulates secretion of pro-angiogenic factors including MMP-14 (101). The precise functional roles played by other annexins during metastatic progression remain to be defined.

Transcriptional Regulators and Splicing Factors are Enriched in Exosomes Following H-Ras-induced EMT-It is well
recognized that splicing events and transcription regulation drive critical aspects of EMT-associated phenotypic change (102,103). For example, the EMT transcription factor twist altered global changes in mRNA splicing in a human mammary epithelial cell line (HMLE cells) resulting in many alternatively spliced genes that are implicated in processes such as cell migration, actin cytoskeletal regulation and cell-cell junction formation, all of which contribute to EMT phenotypic change (102). We report, for the first time, the presence of key transcriptional regulators (e.g. the master transcriptional regulator YBX1) and core splicing complex components in highlypurified exosomes.
Splicing Factors-Recent studies have highlighted an important contribution of alternative splicing to the metastatic cascade, including regulation of EMT at the post-transcriptional level (104,105). Alternative splicing results in the expression of protein isoforms with distinct structural and functional characteristics, and can even give rise to proteins with opposite properties (106). The involvement of alternative splicing in EMT was first reported in relation to the fibroblast growth factor receptor 2 (FGFR2) (107), and since then, several splicing factors and spliced genes involved in cell migration, actin cytoskeletal regulation and cell-cell junction formation during EMT have been discovered (102,108,109). Several splicing factors were identified in 21D1-Exos (Table  III) including the splicing regulator protein SRP20 (Rsc 2.7), and SF3B1 (Rsc 8.9) and SF3B3 (Rsc 2.6), which are compo-  Table S1).
c Protein abundance ratio (ratio of spectral counts; Rsc) reveals differential protein abundance between MDCK-and 21D1-Exos based on Eq.1. The use of zero spectra is overcome using an arbitrary correction factor (1.25) in Eq. 1. The use of this correction factor allows relative quantitation of all proteins within both normalized datasets to be performed, based upon Old et al. ͓53͔. Positive Rsc values reflect increased protein abundance in 21D1-Exos relative to MDCK-Exos; negative values indicate decreased abundance in 21D1-Exos relative to MDCK-Exos.
* Differential expression with p values Ͻ0.05 as reported in supplemental Table S1. nents of the SF3b complex that interacts with U2 small nuclear ribonucleoprotein (snRNP) complex at the catalytic center of the spliceosome (110). Increased expression levels of splicing factor, arginine/serine-rich 1 (SFRS1/SRSF1) (Rsc 23.2), previously known as (SF2/ASF), in 21D1-Exos is of particular significance given its ability to induce EMT (111). SRSF1 has been shown to regulate the splicing of the tyrosine kinase receptor Ron which is synthesized as a single chain precursor, and is comprised of an extracellular 40 kDa ␣-subunit and a 145 kDa transmembrane ␤-subunit (112). SRSF1 promotes the production of ⌬Ron 165, which is an isoform lacking 49 amino acids in the extracellular ␤-subunit generated through the skipping of exon 11 (111,113). ⌬Ron 165 is unable to undergo proteolytic processing and as a consequence accumulates in the cytoplasm in a constitutively phosphorylated form which induces invasive properties (114). By these means, SRSF1 affects the Ron/⌬Ron ratio, which in turn, promotes the morphological and molecular hallmarks of EMT (111). SRSF1 is frequently up-regulated in various human tumors (115). Our finding of the proto-oncogene SRSF1 in H-Ras induced 21D1-Exos may represent a mechanism by which a recipient cell upon uptake of an SRSF1-containing exosome may induce the recipient cell to undergo EMT. Further studies are required to examine this hypothesis. Transcription Factors-A salient finding of this analysis was the identification of Y-box-binding protein (YBX1), a DNA-and RNA-binding protein that has properties of a nucleic acid chaperone (116), in 21D1-Exos (Table III). YBX1 was the most up-regulated protein in exosomes following EMT (Rsc 37.5), and its unique expression in 21D1-Exos was validated by Western blotting (Fig. 2A). YBX1 is known to be involved in almost all DNA-and mRNA-dependent processes including DNA replication and repair, transcription, pre-mRNA splicing, and mRNA translation (116), and is considered to be a master transcriptional regulator. YBX1 can bind RNA to limit protein synthesis, or bind DNA through the Y-box promoter element containing an inverted CCAAT box to either activate or repress transcription (117,118). YBX1 is known to interact with other DNA binding proteins such as PUR␣ (PURA), also uniquely present in 21D1-Exos (Rsc 2.6) (Table III). PUR␣ regulates cell proliferation through the activation of growthassociated gene transcription (119,120). MMP-13 expression is also known to be regulated by YBX1 (121), and given that MMP-13 was uniquely identified in MDCK-Exos (Rsc -79.8) it is possible that its diminished expression in 21D1-Exos is because of elevated YBX1 expression. MMP-13, also known as collagenase-3, is an ECM-degrading proteinase (122) that has been reported to be selectively down-regulated in conjunction with MMP-9, by the transcription factor SPDEF during prostate tumor metastasis (123). Given that YBX1 is known to promote an epithelial-mesenchymal transition through translational activation of snail1, it is interesting to hypothesize that 21D1-Exos may also induce EMT via YBX1 in recipient cells (124).
In summary, proteomic profiling of highly-purified exosomes has revealed new insights into the contribution of exosomes to the extracellular microenvironment after oncogenic H-Ras-induced EMT. We show that exosomes released from epithelial MDCK cells undergo extensive reprogramming causing exosome-mediated release of several factors associated with modifying the extracellular tumor microenvironment including proteases, annexins, integrins and secreted ECM components. It is possible that these factors may positively feedback on themselves to maintain the EMT process, or induce neighboring cells to undergo EMT. In addition, our findings reveal for the first time that oncogenic H-Ras transformation induces the packaging and release of mediators associated with nuclear assembly, transcription, splicing, and protein translation. Given that 21D1-Exos contain several features known to induce EMT, it is tempting to speculate that Ras-transformed exosomes are functionally capable of initiating EMT in recipient cells.
* This work was supported by the National Health & Medical Research Council (NHMRC) of Australia for program grant #487922 (RJS, JH, DWG), grants #280913 and #433619 (H-JZ), grants #628946 and #400202 (AFH). AFH is also supported by an Australian Research Council (www.arc.gov.au) Future Fellowship (FT100100560). RAM is supported by an Early Career CJ Martin Fellowship #APP1037043, and BMC by an NHMRC Dora Lush Biomedical Postgraduate Scholarship #628959. BJT is supported by The University of Melbourne Research Scholarship. Analysis of proteomic data described in this work was supported using the Australian Proteomics Computational Facility funded by the National Health & Medical Research Council of Australia grant #381413. Electron microscopy was performed at the Advanced Microscopy Facility at the Bio21 Molecular Science and Biotechnology Institute, The University of Melbourne. This work was also supported, in part, by American Recovery and Reinvestment Act funds through National Institutes of Health Grant R01 HG005805 (RLM), the NIGMS Grant 2P50 GM076547 from Center for Systems Biology, the Luxembourg Centre for Systems Biomedicine and the University of Luxembourg, and from the National Science Foundation (MRI Grant 0923536). UK was supported by a fellowship from the German Academic Exchange Service. We thank the NCI of the NIH for support (Grant #1R03CA156667 to RLM). □ S This article contains supplemental Figs. S1 and S2 and Tables S1 to S4.