Circulating Small Extracellular Vesicles Activate TYRO3 to Drive Cancer Metastasis and Chemoresistance

Extracellular vesicles (EV), including exosomes and microvesicles from tumor tissues, play a pivotal role in cancer progression. In particular, exosomes (30–100 nm) derived from the multivesicular body have recently garnered considerable interest. Over the past decade, tumor-derived EVs (TdEV) have been extensively studied, and it has been demonstrated that they lead to tumor progression including tumor microenvironment (TME) remodeling, vascular leakiness, metastasis, and chemoresistance (1).

The vascular system is critical for maintaining cancer cell survival, because tumors undergo necrosis or apoptosis without a blood supply. Furthermore, blood is an inevitable space for circulating tumor cells (CTC) to begin metastasizing (2), and the levels of circulating angiogenic factors correlate with the aggressiveness of cancer (3, 4). Interestingly, there are many more platelets and erythrocytes compared with CTCs (∼10⁸ fold) in the blood (5), suggesting that most circulating small EV (csEV) do not originate from tumors even though TdEVs are known to be a factor in Paget's “seed and soil” theory (6). Although several reports have noted the importance of csEVs in cancer, most have focused on the clinical potential of csEVs as a biomarker (7, 8), and there is little information about their role in cancer progression.

TAM receptors, including TYRO3, AXL, and MerTK, are members of a family of receptor tyrosine kinases and are promising therapeutic targets in various cancers (9). TAM receptors are involved in the proliferation and invasiveness of cancer cells and immunosuppressive activities in the TME (10). The activation of TAM receptors requires both a bridging protein ligand (e.g., Gas6 and PROS1) and phosphatidylserine (PS) lipid moiety. Because the exosome is composed of PS, it has been suggested that it functions as a ligand for TAM receptors (10, 11).

Given the unknown role of csEVs in cancer progression and the potential link between csEVs and TAM receptors, we determined whether csEV promotes the migration and survival of aggressive cancer cells. We found that csEVs mainly activated TYRO3 among the TAM receptors, which facilitated the epithelial–mesenchymal transition (EMT) and RhoA activation, thereby enhancing cancer cell migration and metastasis. In addition, TYRO3 was involved in the chemoresistance of epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor (TKI)-resistant lung cancer cells via YAP activation. Our findings provide a novel role for csEV-TYRO3 in aggressive cancers, and suggest that a selective TYRO3 inhibitor could be used with standard chemotherapy to overcome metastasis and chemoresistance in the clinic.

Human prostate cancer cells (LNCaP and LNCaP-SL), colon cancer cells (HCT116 and SW480), breast cancer cells (TAMR-MCF-7 and MDA-MB-231), and non–small cell lung cancer (NSCLC) cells (HCC827, GR-HCC827, H1975, H292, ER-H292, GR-H292, H1993, GR-H1993, and ER-H1993) were cultured at 37°C in 5% CO₂/95% air either Dulbecco's modified Eagle's medium or RPMI medium supplemented with 10% fetal bovine serum (FBS) and penicillin/streptomycin. LNCaP cells were obtained from ATCC 2019. LNCaP-SL cells were donated from Dr. Hyungshik Kim (Sungkyunkwan University, Suwon, Korea). Human non–small lung cancer cells, including H292, ER-H292, GR-H292, H1993, ER-H1993, and GR-H1993, and colon cancer cells (HCT116, SW480) were kindly donated from Dr. Sang Kook Lee (Natural Products Research Institute, Seoul National University, Seoul, Korea). HCC827 and GR-HCC827 cells were obtained from Dr. Cheol Ho Jeong (Kyemyung University, Daegu, Korea). GR-H1993 and ER-H1993 cells were subcultured in the presence of 10 μmol/L gefitinib or erlotinib. GR-H292, GR-HCC827, and ER-H292 were subcultured in the presence of 1 μmol/L gefitinib or erlotinib. Ba/F3 cells were maintained in RPMI supplemented with 10% FBS (Merck Millipore). Parental Ba/F3 cells were grown in the presence of 5 ng/mL IL3 (PeproTech). Ba/F3-MerTK cells were cultured in the presence of 1 μg/mL puromycin. Ba/F3-AXL, Ba/F3-TYRO3 cells were grown in IL3-free, puromycin-free condition. Cell cultures were routinely tested for the Mycoplasma detection kit (PlasmoTest, Invivogen). Frozen stocks were made for all cell lines and cultured for no more than 1 month for after each thawing.

Antibodies recognizing Albumin (#4929), AXL (#8661), MERTK (#4319), TYRO3 (#5585), p-TAM (#44363), p-SMAD3 (#9520), SMAD3 (#9523), AKT (#9272), pAKT (#9271), ERK (#9102), pERK (#4376), p-YAP (#4911), YAP/TAZ (#8418), alpha-tubulin (#2144), Zeb-1 (#3396), horseradish peroxidase–conjugated donkey anti-rabbit (#7074), and anti-mouse IgGs (#7076) were purchased from Cell Signaling Technology. Anti-MLC, p-MLC (ab2480), and CD63 (ab68418) antibody was purchased from Abcam. Antibodies for ApoB (sc-393636), ApoE (sc-390925), CD63 (sc-15363), RHOA (sc-418), SLUG (sc-166476), TWIST (sc-15393), TYRO3 (sc-166360, used for IF), cofilin (sc-376476), p-Cofilin (sc-271921), c-MYC (sc-40), connective tissue growth factor (CTGF; sc-14939), vimentin (sc-32322), and YAP (sc-376830, used for IF) were supplied from Santa Cruz Biotech. Alexa Fluor 488 donkey anti-mouse IgG, Alexa Fluor 568 goat anti-rabbit IgG, and Alexa Fluor 488-conjugated phalloidin antibodies were purchased from Life Technologies. Other antibodies specific for the following proteins were used: N-Cadherin (#610920, BD Biosciences), β-actin (#a2228, Sigma-Aldrich), glyceraldehyde 3-phosphate dehydrogenase (#CB1001, Merck), Bovine-CD63 (#MCA2042GA, Bio-Rad), TYRO3 (#AF859, R&D Systems). Also, gefitinib was obtained from Medchemexpress (Monmouth Junction). Phophatidylcholine (PC), cholesterol (CH), phosphatidylserine (PS), and phosphatidic acid (PA) were purchased from Avanti Polar Lipids.

csEV originated from bovine or porcine species were isolated from commercially used bovine or porcine serum (Invitrogen). To minimize intraspecies difference, the same lot number of sera was used as possible. Serum was diluted with PBS (1:2) and serially centrifuged to remove dead cells, debris, and large size of extracellular vesicles (300 × g for 10 minutes, 2,500 × g for 20 minutes, 10,000 × g for 30 minutes). 200-nm filtration was performed to remove vesicles and contaminants larger than the csEV size. The resulting supernatant was subjected to ultracentrifugation (120,000 × g for 90 minutes, Optima XE-100 with SW 32Ti rotor, Beckman Coulter). The supernatants were completely discarded for removal of contaminating proteins, and then the tube was filled with PBS and subjected to ultracentrifugation (120,000 × g for 90 minutes). The resident sEV pellet was solubilized in sterile PBS and filtrated using 200-nm filter again to remove aggregates for enrichment of sEV. For the isolation of human csEV, blood collection was conducted in healthy non-smoking volunteers. Fresh blood samples from the donors were centrifuged to obtain serum and the serum was immediately centrifuged for the isolation of csEV following protocol as previously described. Ethical clearance was obtained from the Institutional Review Board of Seoul National University (#SNU 19-02-040). Intensity, volume, and numerical distribution of the isolated csEV were analyzed by ELSZ-1000 (Photal Otsuka Electronics). Obtained csEV from serum were stored at 4°C and used in all experiments within 5 days.

Iodixanol solutions [40% (w/v), 20% (w/v), 10% (w/v), 5% (w/v)] were prepared by diluting a stock solution of OptiPrep (60% aqueous iodixanol; Axis-Shield, Oslo, Norway) in 0.25 M sucrose buffer (10 mmol/L Tris-HCl, pH 7.4). To form the discontinuous gradient, 6 mL (for 40%, 20%, 10%) or 4 mL (for 5%) was layered serially in centrifugation tubes, and 6 mL serum was then layered on to these gradients after processing 300 × g for 10 minutes, 2,500 × g for 20 minutes, 10,000 × g for 30 minutes and 200 nm filtration. Afterward, the sample was subjected to ultracentrifugation in SW32 Ti rotor for 15 hours (at 120,000 × g, 4°C). Fractions of density gradient layers were collected (F1-F10) from the top of the tubes, and absorbance was measured at 340 nm to determine fraction density. Each fraction was diluted with PBS and subjected to ultracentrifugation again for 150 minutes (at 120,000 × g, 4°C). The supernatant was completely discarded, and the pellet from each fraction was solubilized in sterile PBS.

A sampling of csEV was performed as previously described by Thery and colleagues with modifications (12). Briefly, isolated csEV was fixed with 4% paraformaldehyde and a diluted solution of csEV was dried on formvar-carbon-coated copper grids for 20 minutes. After washing with PBS, the grid was post-fixed with 1% glutaraldehyde for 5 minutes and washed 7–8 times again with distilled water. csEVs were negatively stained with 2% aqueous uranyl acetate for 5 minutes, dehydrated and completely dried.

As for immunogold labeling with PROS1 and CD63, fixed csEV preparations (20 μL) were adsorbed to a carbon-Formvar-coated 200 mesh nickel grids for 30 minutes in a dry environment. Samples were transferred droplets of PBS, droplets of 50 mmol/L ammonium chloride (4 times, for 3 minutes), and 1% BSA blocking solution (for 30 minutes). Grids were then incubated with either blocking buffer (negative control) or primary antibody in 1% BSA in PBS. After washing with 0.1% BSA in PBS (6 times, for 3 minutes), grids were floated on drops of 1.4 nm anti-mouse nanogold (Nanoprobes Inc.) diluted 1:500 in 1% BSA for 1 hours and washed with PBS. Further, grids were post-fixed by 1% glutaraldehyde for 5 minutes followed by rinsing in distilled water (7 times, for 3 minutes). Enhancing of grids was performed using silver enhancement solution (0.1% silver nitrate, 1% hydroquinone in 0.2 M citrate buffer) for 5 minutes, followed by rinsing in deionized water. Finally, grids were negatively stained with 2% aqueous uranyl acetate for 5 minutes and air-dried. The sample images were obtained by transmission electron microscope (LIBRA 120, Carl Zeiss).

Total cell lysates were prepared as previously reported (13). After centrifugation of total cell lysates (13,000 × g for 15 minutes), the supernatants were applied for immunoblottings. Enhanced chemiluminescence (ECL) system reagent (EMD Millipore) was used for band detection.

Phospho-tyrosine immunoprecipitation was performed using p-tyrosine antibody conjugated with sepharose bead (#9419, Cell Signaling Technology) described in the manufacture's protocol. Briefly, cells were washed in cold PBS and immediately lysed with cell lysis buffer. After centrifuging the lysate to discard pellets, 150 μg protein extracts were incubated with 5 μL P-tyrosine sepharose beads overnight at 4°C. P-tyrosine beads were washed 5 times with cold cell lysis buffer. The sepharose beads were then eluted by 3 × SDS sample buffer and boiled for 5 minutes to disassociate the immune complex from beads. RhoA pulldown activation assay was conducted as described in the manufacture's protocol (#BK036, Cytoskeleton Inc.).

Flag-YAP wild-type and flag-YAP5SA cDNAs cloned into retroviral pMSCV-puro vector were kindly provided by Dr. Joon Kim (KAIST, Daejeon, Korea). For retroviral particle assembly, retroviral constructs were transfected into Amphophoenix cells, and retroviral supernatant was collected 48 hours after transfection. The supernatants were filtered through a 0.45-μm filter, and infected into the target cells with 4 μg/mL polybrene for 2 weeks.

Plasmid DNA (U6-gRNA/CMV-Cas9-RFP plasmid for TYRO3, U6-gRNA/CMV-Cas9-GFP plasmid for negative control) were purchased from Sigma. LNCaP-SL and HCT116 cells were transfected using Lipofectamine 2000, followed by sorting of GFP- or RFP-labeled cells with FACSAria II cell sorter (BD Biosciences) installed at the National Center for Inter-University Research Facilities at Seoul National University. Single-cell clone was isolated by limiting dilution and screened for genetic and functional deletion of TYRO3 and YAP1.

The 1 × 10³ cells (LNCaP-SL) or 3 × 10³ cells (all cell lines except LNCaP-SL) were seeded in 96-well plate, and the phase confluence of cells was scanned every 4 hours by using the IncuCyte ZOOM or S3 Live Cell Analysis System (Essen Bioscience). For transwell migration assay, 3–5 × 10³ cells were seeded in the upper chamber of the IncuCyte clearview chemotaxis plate (#4582, Essen Bioscience) and the lower chamber was filled with media. The cells were incubated for 24 hours, and the migrated cell numbers were counted using IncuCyte Zoom Chemotaxis Migration Assay for Live-Cell Analysis (Essen bioscience). Relative migration ratio was calculated by migrated cells or area normalized to 0 hours total cell area of the upper well.

LNCaP-SL or HCT-116 cells were cultured overnight on coverslips. After treatment with vehicle or compound for 24 hours, the cells were fixed with 4% paraformaldehyde for 20 minutes at room temperature and incubated with 0.1% Triton X-100 for 15 minutes. The fixed cells were blocked with 10% horse serum for 1 hour and incubated with indicated primary antibody (1:200) at 4°C overnight. After washing with PBS, the cells were incubated with fluorophore-conjugated secondary antibody (1:1000). Finally, the coverslips were washed thoroughly with PBS and then mounted with ProLong Gold Antifade reagent with 4′,6-diamidino-2-phenylindole (DAPI; Invitrogen). Images were obtained using confocal microscope (Leica TCS SP8 MP, Leica Microsystems).

The negatively charged liposomes were prepared by thin lipid film hydration and extrusion technique with slight modification (14, 15). Briefly, the lipid mixture of PC, CH, and PS with different molar ratio (PC:CH:PS = 2:1:0, 1.5:1:0.5, 1:1:1, and 0.5:1:1.5) was dissolved in chloroform and then evaporated under reduced pressure using a rotary evaporator (Rotavapor R-3; Büchi Labortechnik AG). The resulting thin lipid film was further dried by purging a stream of nitrogen gas for 30 minutes to remove the trace of organic solvent. The completely dried lipid film was hydrated with a 5% dextrose solution maintained at 40°C above the phase transition of lipids. The obtained liposomal dispersion was sonicated for 30 minutes in a bath type sonicator operated at 40°C and then extruded 10 times through 100-nm polycarbonate membrane using a Mini-Extruder (Avanti Polar Lipids, Inc.). PA liposomes were prepared by the same method except that PA was used instead of PS.

The physical properties of negatively charged liposomes were characterized in terms of particle size, polydispersity index (PDI), and zeta potential. The average particle size and PDI of negatively charged liposomes were determined by photon correlation spectroscopy using a Zetasizer Nano ZS (Malvern Instruments). The zeta potential of negatively charged liposomes was measured by electrophoretic light scattering using the same instrument. The negatively charged liposomes were suitably diluted with filtered deionized water prior to all measurements.

The 3 × 10⁴ LNCaP-SL and HCT116 cells were plated on 24-well plates and cotransfected with 8 × GTIIC-luciferase vector (#34615, Addgene) or Smad-binding element (SBE) luciferase reporter and pRL-TK plasmids (a plasmid that encodes Renilla luciferase for normalization of transfection efficacy) using Lipofectamine 2000 (Life Technologies). The firefly and Renilla luciferase activities were determined by a Dual-Luciferase Reporter Assay System (Promega) using a luminometer (LB960, Berthold Technologies). The relative luciferase activities were calculated by normalizing the promoter-driven firefly luciferase activity to pRL-TK (Renilla) luciferase activity.

All experiments and methods were performed in accordance with relevant guidelines and regulations. All animal procedures were approved by the Institutional Animal Care and Use Committee of Seoul National University (Approval #: SNU-190919-1 and SNU-191022-8). LNCaP-SL, HCT116, GR-H1993 cells (5 × 10⁵, 5 × 10⁶, and 5 × 10⁶ cells, respectively) were suspended in 50 μL Cultrex pathclear basement membrane extract (50% v/v in PBS, R&D Systems) and subcutaneously injected into 5-week-old male athymic BALB/c-nu mice (Raon Bio Inc.). For GR-H1993 tumor model, mice were grouped based on tumor volume to ensure similar starting tumor volume distribution before treatment once tumors became palpable (>100 mm³). Gefitinib (25 mg/kg, PO) or KRCT-6j (20 mg/kg, IP) was administered daily for 28 days. Tumor growth was measured every three days. Tumor length and width were detected by calipers, and tumor volume was calculated using the formula (length × width²) × π/6. Mice were humanely euthanized when tumor volume reached around 1,000 mm³.

All animal procedures were approved by the Institutional Animal Care and Use Committee of Seoul National University (Approval #: SNU-190114-6). Five-week-old BALB/c-nu mice were anesthetized, and the mice were injected with 5 × 10⁵ sgCtrl LNCaP-SL, sgTYRO3 #1 LNCaP, or sgTYRO3 #2 LNCaP-SL cells diluted in 100 μL PBS into the spleen. After 3 weeks, the mice were sacrificed and the liver samples were analyzed by the hematoxylin and eosin staining method. For pathologic assessment, specimens were cut into 2-mm-thick sections after formalin fixation. All abnormal regions (nodules or suspicious regions) were observed under a microscope, and the absolute metastatic tumor area was analyzed. All abnormal regions were analyzed (3–4 slides per liver tissue) when nodules are found.

All statistical analyses were performed using GraphPad prism (GraphPad software). Densitometry scanning was performed using Multi gauge software (Fujifilm). The values were presented as means SD. Unpaired Student t test or one-way analysis of variance (ANOVA) were used to assess statistical significance. Results were considered significant when P < 0.05 (*, P < 0.05; **, P < 0.01; ***, P < 0.005; ****, P < 0.001).

Using serial centrifugation, csEVs were isolated from human, bovine, and porcine sera (Fig. 1A). To minimize intraspecies variance, we used csEV originating from the same lot number of bovine serum in similar experiments, and key data were confirmed using csEVs from human serum. As shown by direct light scattering, the size of the isolated EVs was below 100 nm and classified as small EVs (Fig. 1B). Moreover, the expression of cluster of differentiation 63 (CD63), a representative EV and exosome marker (Supplementary Fig. S1A), transmission electron microscopy (Supplementary Fig. S1B), and the uptake of DiD-labeled csEVs observed by confocal microscopy (Fig. 1C) also validated that the isolated vesicles were small EVs. In addition, we detected cytochrome c to confirm the purity of the isolated csEVs (Supplementary Fig. S1C).

Considering that EVs originating from tumor cells enhance metastasis (1), we determined whether csEV also promoted the migration of cancer cells. Interestingly, csEVs originating from bovine, porcine, or human sera promoted the migration of LNCaP-SL cancer cells, an aggressive subline of LNCaP prostate cancer cells. In particular, csEVs originating from human serum had the strongest migratory capability (38.1-fold; Fig. 1D). We also assessed the migratory effect of csEVs in other aggressive cancer cell lines [HCT116 (colon cancer), SW480 (colon cancer), gefitinib-resistant-HCC-827 (GR-HCC827, NSCLC), H1975 (NSCLC), and patient-derived primary breast cancer cells]. Expectedly, migration of the five cancer cell lines was increased by csEVs in a concentration-dependent manner (Fig. 1E; Supplementary Fig. S1D). To exclude the role of lipoprotein particles and non-EV proteins in the csEV-driven cell migration, we performed iodixanol density gradient ultracentrifugation (DGUC; Fig. 1F). The migratory capability was significantly enhanced by EV-rich fraction (F6, CD63 most abundant fraction) than lipoprotein-rich fraction (F2; Fig. 1G).

Next, we determined which component of csEV led to the migration of cancer cells. We first divided csEVs with charge distribution by electronic stimulation (Fig. 1H; Supplementary Fig. S1E), and observed differential migratory ability according to the “surface charge” of the csEVs. csEVs possessing a relatively negative charge had the most potent migratory effect (Fig. 1I). Because lipids with a negative charge are the main surface components of EVs, we evaluated the involvement of the lipid component of EVs in cell migration. Among the various lipid components, we focused on PS, which is the fourth most abundant and most negatively charged lipid of small EVs (16). The data showing csEV-induced cell migration were significantly inhibited by Annexin V, a PS-binding protein, indicating the importance of PS in the effects of csEV on cancer cell migration (Fig. 1J; Supplementary Fig. S1F).

To ascertain the role of PS, negatively charged liposomes were prepared and optimized with different ratios of phosphatidylcholine, cholesterol, and PS (or PA) using the simple hydration and extrusion method. Physical characterization of negatively charged liposomes including particle size, PDI, and zeta potential is presented in Table 1. All of the liposomes showed an average particle size below 100 nm with PDI values < 0.2, demonstrating the narrow size distribution of liposomal vesicles. As shown in Fig. 1K, the migration of LNCaP-SL cells was enhanced as the PS ratio of the tested artificial liposomes increased. Additionally, we tested if other negatively charged lipids could also promote the migration of cancer cells. However, liposomes consisting of PA, which have a similar zeta potential to PS, did not cause cell migration (Fig. 1L), suggesting that PS is a unique surface lipid component of csEVs that induces cell migration.

We observed the expression of TAM receptors in various cancer cell lines, as PS is sensed by TAM receptors (Supplementary Fig. S2A; ref. 10). To investigate whether csEV could activate TAM receptors, we used Ba/F3 cells selectively expressing the three different TAM receptors. Usually, Ba/F3 cells can only proliferate in the presence of interleukin 3, but Ba/F3 overexpressing a specific TK receptor can proliferate when the overexpressed TK is activated. Because TYRO3-overexpressing Ba/F3 cells were most sensitive to csEVs (Fig. 2A), we examined the tyrosine phosphorylation of TYRO3 receptor in LNCaP-SL cells under csEV stimulation. Phosphorylation of TYRO3 was found to increase at 0.5–1.5 hours after csEV exposure (Fig. 2B; Supplementary Fig. S2B). Because LNCaP-SL cells only express TYRO3 (Supplementary Fig. S2A), we observed the phosphorylation of AXL and TYRO3 in HCT 116 cells expressing three kinds of TAM receptors, and found that the tyrosine phosphorylation of AXL and TYRO3 was increased by csEVs (Supplementary Fig. S2C). Interestingly, transmission electron microscopy imaging also confirmed that PROS1, a bridging ligand of TYRO3, was detected in the csEV (Fig. 2C). Furthermore, PROS1 was mainly detected in CD63-abundant F6 fraction (Fig. 2D), indicating that csEV possesses PS and PROS1. Additionally, PROS1 or Gas6 were detected in the serum-free culture media obtained from incubation of LNCaP-SL or HCT116 cells, indicating that secreted Gas6 and PROS1 from cancer cells can act as ligands with csEV for the TYRO3 receptor in the TME (Supplementary Fig. S2D). Because Annexin V treatment can reverse the binding to TAM receptor (17), we used Annexin V-treated csEV to assess the direct interaction of TYRO3 with csEV. As shown in Fig. 2E, Annexin V treatment significantly decreased colocalization of Did-EV with TYRO3 immunofluorescence. Moreover, csEV or EV-rich fraction (F6) pretreated with proteinase K showed the reduced migration ability compared with its naïve form (Supplementary Fig. S2E), indicating the involvement of bound protein(s) in csEV-induced cell migration. To confirm the role of PROS1 in csEV-dependent migration, csEV was exposed to warfarin, a vitamin K antagonist. Because warfarin makes PROS1 non-γ-carboxylated form, TAM receptor–mediated activities can be blocked by warfarin (18, 19). Pretreatment of warfarin to csEV decreased cell migration of LNCaP-SL cells (Supplementary Fig. S2F).

Next, to examine the role of TAM receptor, predicted to be the main receptors for csEV, we checked the level of TAM receptor in metastatic tumor compared with primary tumor. Analysis of Gene-Expression Omnibus (GEO) data sets (GSE6919) showed that TYRO3 expression was increased in metastatic prostate tumor compared with normal prostate tissues or primary prostate tumor (Fig. 2F). Unexpectedly, AXL was significantly downregulated in metastatic prostate tumor compared with normal prostate tissues or primary prostate tumor, whereas MERTK expression remained unchanged (Supplementary Fig. S2G). To determine whether TYRO3 is an essential receptor for csEV-mediated migration and metastasis, we introduced the Crispr-based gene editing system to inactivate TYRO3. We confirmed that TYRO3 was successfully knocked down in LNCaP-SL cells by Western blotting and PCR assays (Supplementary Fig. S3A and S3B). The migration promoting ability of csEV was significantly downregulated in sgTYRO3 LNCaP-SL cells, while there was no significant difference of cell migration between sgCtrl and sgTYRO3 under 10% FBS stimulating condition (Fig. 2G; Supplementary Fig. S3C). Additionally, the expression patterns of TAM receptors in human colon cancer samples (GSE28702) showed that TYRO3 and MERTK expression upregulated in metastatic tissues (Supplementary Fig. S3D). The csEV-induced migratory effect was also decreased in sgTYRO3 HCT116 cells (Supplementary Fig. S3E).

N-cyclohexylpyrimidin-4-amine derivatives have selective inhibitory effects on TYRO3 (20), particularly KRCT-6j (Supplementary Fig. S4A), a potent selective inhibitor with 300-fold selectivity for TYRO3 over both AXL and MerTK. In fact, KRCT-6j selectively inhibited the proliferation of TYRO3-overexpressing Ba/F3 cells (Supplementary Fig. S4B and S4C). Because LNCaP-SL cells express only TYRO3 with very low levels of MerTK and AXL compared with other tested cell lines expressing at least two kinds of TAM receptors (Supplementary Figs. S2A and S5A), we used LNCaP-SL cells in subsequent TYRO3 inhibitor experiments. csEV-stimulated cell migration of LNCaP-SL cells was also suppressed by KRCT-6j (20) in a concentration-dependent manner (Fig. 2H). To identify the pathophysiologic role of TYRO3 in the migration and metastasis of invasive cancer cells, we utilized a splenic injection model for liver metastasis (21). Mice implanted with sgCtrl LNCaP-SL cells had metastatic foci in the liver, while metastatic foci were not found in the liver from mice implanted with the TYRO3 knockout cells (Fig. 2I). When we quantified the metastasis area, the metastatic tumor area was not detected in all of the liver tissues from TYRO3 knockout LNCaP-SL-implanted mice (Fig. 2J). These results suggest that the csEV-TYRO3 interaction plays a pivotal role in the metastasis of invasive cancer cells.

To elucidate the mechanism(s) underlying the csEV/TYRO3 interaction-induced migration and metastasis of cancer cells, comparative transcriptomic analysis was performed in LNCaP-SL cells. Heat map analysis of RNA-seq data identified global changes in the transcriptome by csEVs (Fig. 3A). Among the top upregulated gene sets, we focused on the EMT, which is closely related to the initiation of metastasis and invasiveness acquisition (Fig. 3B). In addition, CTCs, frequently exposed to csEV, actively operate the EMT process known as intermediate EMT during blood circulation (22). Gene set enrichment analysis (GSEA) revealed that EMT-related signatures were enriched in csEV-treated LNCaP-SL cells (Fig. 3C), and the expression of EMT-related genes was regulated by csEV treatment (Fig. 3D). Western blotting and immunocytochemistry confirmed that several EMT markers including ZEB1, TWIST, and SLUG were significantly increased 48 hours after csEV exposure, indicating that csEV is critical to maintaining the EMT of cancer cells (Fig. 3E). Moreover, increased EMT markers by csEV were decreased in sgTYRO3 LNCaP-SL cells compared with sgCtrl LNCaP-SL cells (Fig. 3F). When we performed F-actin staining with phalloidin, csEV-induced membrane localization of F-actin was diminished by KRCT-6j (Fig. 3G).

Transforming growth factor-β (TGFβ) pathway is partly involved in TdEV-mediated metastasis and acts as a vital mechanism in EMT (23). To reveal the molecular basis of csEV-mediated EMT, we explored the relationship between the TGFβ signaling pathway and EMT. Because we revealed that TYRO3 is a crucial receptor for cell migration and metastasis by csEV, we evaluated whether TYRO3 contributes to invasiveness of cancer cells by regulating the phosphorylation of Smad3. When we detected the luciferase activity from SBE-driven reporter transfection, TYRO3 knockout in LNCaP-SL cells repressed SBE reporter activity in both basal and csEV-stimulated conditions (Fig. 3H). Additionally, TYRO3 knockout also inhibited SBE reporter activation and smad3 phosphorylation by TGFβ (Supplementary Fig. S6A and S6B). These data suggest that TYRO3 is linked to the TGFβ Smad2/3 signaling pathway.

Although csEV-induced EMT is a relatively late-response process after csEV treatment (48–72 hours), cell migration induced by csEV began 4–8 hours after csEV exposure. Hence, we identified other signaling pathway(s) that control csEV-mediated cell migration. Based on the reports that the RhoA/Rho-associated protein kinase (ROCK) signaling pathway is rapidly activated independently of Smad2/3, and regulates cell movement by controlling actin filament (24, 25), we hypothesized that RhoA activation might be linked to csEV-induced cell migration.

Analysis of the GEO data set (GEO21304) showed that RHOA expression was highly correlated with TYRO3 in transcripts of human prostate tumor (150 samples), indicating a possible relationship between the two genes (Fig. 4A). To illustrate the connection between csEV and RhoA activation, we measured GTP-bound RhoA (active form). GTP-RhoA was significantly increased 3 hours after csEV treatment (Fig. 4B; Supplementary Fig. S6C). We further assessed the role of ROCK signaling in csEV-mediated cell migration and invasion using cell-based transwell assays. Y27632, a pan ROCK inhibitor, suppressed the increased cell migration by csEV (Fig. 4C). Additionally, active RhoA was notably decreased in sgTYRO3 LNCaP cells, confirming the correlation between csEV/TYRO3 and RhoA (Fig. 4D).

Next, we explored the downstream signaling of RhoA for csEV-induced cell migration. ROCK consists of two isoforms, ROCK1-activating myosin light chain 2 and ROCK2 phosphorylating cofilin; both isoforms act as major downstream effectors contributing to the progression of RhoA-mediated diseases (26). The phosphorylation of cofilin was time dependently determined after csEV or phosphate-buffered saline (vehicle) exposure. The phosphorylation of cofilin was dynamically changed by csEV treatment (Fig. 4E), indicating that the RhoA–ROCK2–cofilin pathway is involved in csEV-driven cell migration.

To further investigate the role of csEV in the proliferation of cancer cells, we examined the effect of csEV on the proliferation of LNCaP-SL cells. The proliferation of LNCaP-SL was not affected by csEV in medium supplemented with FBS. Because LNCaP-SL cells showed FBS-independent cell growth up to initial 30 hours, we used serum-deprived condition to guarantee to exclude the potential interaction of serum factors (e.g., growth factors) with csEV in mitogenic signaling. Interestingly, csEV treatment significantly promoted cell proliferation (7.8- vs. 12.8-fold at 96 hours) in serum-deprived conditions (Fig. 5A and B). Moreover, the proliferation rate of TYRO3 knockout cells was markedly lower in serum-deprived conditions compared with control cells (Fig. 5C), suggesting that TYRO3 activation may act as a key player for cancer cell survival in growth factor-restricted conditions. To elucidate the molecular basis of csEV-mediated proliferation, we examined the activities of representative downstream signaling pathways (PI3K–AKT and MEK1/2–ERK) of growth factor receptors. It has also been reported that TYRO3 is coupled to the PI3K–AKT and MEK1/2–ERK pathways in cancer cells (27). However, the phosphorylation of AKT or ERK did not distinctly change after csEV treatment (Supplementary Fig. S7A). Moreover, csEV-induced proliferation of LNCaP-SL cells was not inhibited by U0126 (MEK1/2 inhibitor), LY294002 (PI3-kinase inhibitor) or LY364947 (TGF-β receptor inhibitor), suggesting that csEV-stimulated cell growth is not dependent on typical cell proliferation signals (Supplementary Fig. S7B).

YAP/TAZ has recently generated considerable interest due to its ability to promote proliferation in various cancers including chemoresistant cancer (28, 29). In particular, the activation of RhoA is a potent activator of YAP, inducing the nuclear localization of YAP (30). Because YAP responds to serum starvations (31), we hypothesized that it might facilitate csEV-mediated cell proliferation. Notably, GSEA revealed that a YAP-conserved signature was significantly enriched and the expression of YAP-target genes was increased in csEV-treated LNCaP-SL cells compared with vehicle-treated LNCaP-SL cells (Fig. 5D). As expected, Ser127 phosphorylation of YAP (inactive form) was decreased in response to csEV under serum starvation (Fig. 5E). We established YAP knockout LNCaP-SL cells (sgYAP LNCaP-SL cells) using the CRISPR system (Supplementary Fig. S7C), and assessed the proliferation of sgYAP LNCaP-SL cells in the presence of csEV. sgYAP LNCaP-SL cells had lower responsiveness to csEV than sgCtrl LNCaP-SL cells, indicating that YAP is involved in csEV-induced cell proliferation (Fig. 5F). Immunocytochemistry confirmed that the expression and nuclear localization of YAP were increased after a 24-hour incubation with csEVs, and were reversed by KRCT-6j treatment in LNCaP-SL cells (Fig. 5G). We also examined YAP/TAZ-dependent gene transcription activity using the YAP/TAZ-responsive TEAD reporter luciferase assay. csEV exposure increased TEAD reporter activity, which was attenuated by TYRO3 knockout in both HCT-116 cells (Fig. 5H) and LNCaP-SL cells (Supplementary Fig. S7D).

To confirm our findings, we transfected HCT116 and LNCaP-SL cells with YAP-5SA, a constitutively active mutant of YAP that prevents its cytoplasmic location and phosphorylation. We confirmed that CTGF, the representative downstream target gene of YAP, was significantly upregulated in YAP5SA-overexpressing HCT116 and LNCaP-SL cells (Supplementary Fig. S7E). When we compared cell proliferation under BMS-777607 treatment (a representative pan-TAM receptor inhibitor), LNCaP-SL cells expressing YAP5SA had decreased sensitivity to BMS-777607 treatment (Supplementary Fig. S7F). Additionally, csEV did not affect the proliferation of YAP5SA-overexpressing HCT116 cells, whereas treatment of MOCK HCT116 cells with csEV significantly increased cell proliferation (Fig. 5I). These results indicated that csEV-TYRO3–mediated cell proliferation in serum-deprived condition was derived from YAP activation. To evaluate further the role of TYRO3 in tumor growth in vivo, we adopted a tumor xenografts model using LNCaP-SL and HCT116 cells. However, both tumor volume and tumor weight were not affected by TYRO3 knockout, indicating that TYRO3 is not a critical receptor for the growth of cancer cells under normal conditions (Fig. 5J).

Recent studies have shown that YAP activity is closely related to various chemoresistant cancer types, including resistance to EGFR-TKIs (32, 33). Hence, we determined whether the TYRO3–YAP axis may affect the chemosensitivity to anticancer agents. We evaluated the proliferation of LNCaP-SL cells in EV-containing FBS or EV-free FBS and in combination with 50 nmol/L docetaxel, a representative chemotherapeutic agent against metastatic prostate cancer. LNCaP-SL cells were significantly more resistant to docetaxel in EV-containing 10% FBS compared with EV-free FBS 10% conditions (Fig. 6A). We further evaluated the combination effects of docetaxel with BMS-777607 on the proliferation of LNCaP-SL cells. Although a low concentration of docetaxel alone was moderately cytostatic in LNCaP-SL cells, combinatorial treatment with BMS-777607 synergistically increased the cytostatic effects compared with each treatment group (Fig. 6B).

Enhanced YAP expression leads to EGFR-TKI resistance in lung cancer cells (34). When we compared the relative expression of TYRO3 in normal lung tissues and lung tumor tissues in the GEO database (GSE18842), TYRO3 was significantly upregulated in human lung tumor compared with normal lung tissue (Supplementary Fig. S8A). We estimated the role of the TYRO3–YAP axis in the chemoresistance of lung adenocarcinoma. We used various acquired EGFR-TKI–resistant lung cancer cell lines that were established by continuous exposure with gefitinib or erlotinib (35). Sensitivities to gefitinib or erlotinib were decreased in all of the resistant cell lines (GR or ER) compared with the parental cells (Supplementary Fig. S8B). We found that TYRO3 was elevated in three of the tested EGFR-TKI–resistant lung cancer cells (GR-H1993, GR-HCC827, and ER-H292 cells) compared with parental cells (Fig. 6C). Because GR-H1993 cells have higher TYRO3 expression than parental cells (2.4-fold), we selected GR-H1993 cells as a model cell to explore the association between TYRO3 and EGFR-TKI resistance. Moreover, we confirmed that GR-H1993 has higher p-TYRO3 (tyr749) expression than H1993 cells (Supplementary Fig. S8C). csEV treatment increased the proliferation of GR-H1993 cells, but not of parental H1993 or ER-H1993 cells showing low levels of TYRO3. Besides, KRCT-6j only inhibited the proliferative effect of csEV in GR-H1993 cells (Fig. 6D and E). Comparing TYRO3 expression in H1993 and LNCaP-SL (representative csEV-responsive cell line), TYRO3 level was higher in LNCaP-SL cells than H1993 cells (Supplementary Fig. S8D). We next monitored the expression level and phosphorylation intensity of YAP in H1993 and GR-H1993 cells. YAP mRNA and protein levels of YAP and nuclear localization were increased in GR-H1993 cells, compared with parental H1993 cells (Fig. 6F; Supplementary Fig. S8E). Moreover, treatment of GR-H1993 cells with KRCT-6j suppressed the csEV-induced nuclear localization of YAP (Fig. 6G).

Next, we performed real-time proliferation assays to assess the synergistic effects of gefitinib with KRCT-6j in GR-H1993 cells. The synergism was evaluated by the Chou–Talalay method, which calculates a combination index (CI; ref. 36). The CI calculated by Compusyn was used to calculate synergism (CI<1), additive effect (CI = 1), and antagonism (CI>1). We used two different concentrations of gefitinib and KRCT-6j, and observed modest antiproliferative effects on GR-H1993 cells. When we assessed the CI of the combination treatment of KRCT-6j and gefitinib, the two agents showed synergistic effects (CI<1; Fig. 6H).

To determine the synergistic effect of gefitinib and KRCT-6j on tumor growth in vivo, we established xenografts by implanting GR-H1993 cells. Although treatment with either KRCT-6j or gefitinib alone did not affect tumor growth, coadministration of KRCT-6j and gefitinib significantly inhibited tumor growth in xenografts implanted with GR-H1993 cells, consistent with the in vitro data (Fig. 6I; Supplementary Fig. S8F). The isolated tumor tissues were additionally stained with Ki67, a representative proliferation marker. The Ki67 immunopositive cell population was markedly diminished in the coadministration group (Supplementary Fig. S8G). Our results provide evidence that the combination treatment of TYRO3 inhibitor with EGFR-TKI effectively inhibits the growth of aggressive cancers as a consequence of YAP inactivation.

EV is considered a crucial mediator of communication among cells involved in many pathophysiologic processes, including cancer. Because they contain several markers of the originator cell, EVs derived from biological fluids have been exploited as a diagnostic tool. Many studies have described the pathologic functions of EVs derived from tumors. Despite the known effects of EVs from tumors, the basis for csEV in cancer biology has not been fully identified. Given that csEVs are more abundant than TdEVs in the blood [median value of plasma EV particle number = ∼2 × 10¹⁰/mL measured by Nanosight (37)], they could interact with CTCs and be powerful effectors in the invasive area near vessels. Although tumor microvasculature is abnormal and heterogeneous depending on tumor types (38), blood supply in the TME is fundamentally required for sustaining tumor growth.

Our results showed that csEVs collected from healthy donors promoted cancer cell migration and chemoresistance via the TYRO3 receptor. We assume that physiologic activity of csEVs could be more important for aggressive cancer cells showing abnormal TYRO3 expression than normal tissues. Because EGFR expression in fully differentiated normal cells is limited and its expression is frequently overexpressed in NSCLC, EGFR–TKI give new insights for the treatment of NSCLC as targeted therapy, though significant amounts of EGF are detected in urine, salvia, and plasma from healthy volunteers (39).

We found that csEVs containing both PtdSer and PROS1 maximized the effector coupling of TAM receptors. PROS1 is a ligand for TYRO3 and exists in plasma at 346 nmol/L, circulating as a free active form (30%–40%) in the blood (40). Because PROS1 can bind to PtdSer moieties with their N-terminal γ-carboxylglutamic acid-rich domain, we assume that PtdSer of the csEV membrane binds to the free form of PROS1 when circulating in blood, and the PROS1-bound form of csEVs may interact with the TYRO3 receptor in cancer cells. Furthermore, GEO analysis showed that metastatic human tumor tissues had significantly higher TYRO3 expression compared with normal tissues or primary tumor (Fig. 2F; Supplementary Fig. S3D), suggesting that the TYRO3 could be a crucial receptor mediating metastasis by responding to csEVs in aggressive cancer. Even though TYRO3 receptor is considered as a key TAM receptor for csEV-induced metastasis and chemoresistance in this study, previous reports showing AXL-driven EGFR-TKI resistance (41, 42) and involvement of overexpressed MerTK in bone metastasis (43) indicate that all members of the TAM family have capabilities to progress cancer.

The observation that csEVs facilitate the EMT through TYRO3 activation also confirmed the regulatory role of csEV in cancer cell migration and metastasis. Most circulating EVs are derived from platelets (44). Several reports have indicated that platelets contribute to the EMT of CTCs, and CTCs exhibit a hybrid EMT phenotype to survive in the blood. However, little is known about the specific interaction between csEV and cancer cells. Here, we suggest that the csEV/TYRO3 interaction is critical to inducing mesenchymal markers including ZEB1 and TWIST in CTCs. Although csEV-induced Smad activation was suppressed in TYRO3 knockout cells, the basal and TGFβ-induced Smad reporter activities were also downregulated in the cell type. Hence, csEV-derived EMT seems to be partially dependent on Samd2/3 activity. However, future study is needed to clarify direct interaction between the TYRO3 and Smad pathway.

Another important finding of our study was the elucidation of two novel pathways involved in csEV-TYRO3–mediated cancer progression. RhoA promotes stress fibers and strong adhesion through focal adhesion formation (45, 46). Here, we demonstrated that RhoA was activated by the csEV/TYRO3 interaction, and ROCK inhibition attenuated csEV-mediated cell migration. Moreover, the phosphorylation of cofilin, a substrate of ROCK2, was actively changed by csEV. These results indicate that the RhoA–ROCK2–cofilin axis is downstream of csEV-driven TYRO3 activation, in accordance with a recent report showing that TGFβ-stimulated active RhoA is responsible for invasiveness and metastasis via the regulation of cofilin phosphorylation in invasive prostate cancer cells (47). Even though little is known about downstream TYRO3 signaling cascades, it has been reported that cytoskeletal rearrangement occurs in the brain endothelium by PROS1–TYRO3 activation (48). Hence, csEV-induced RhoA–cofilin activation would be a key step for csEV-induced cancer cell migration.

We also demonstrated that the Hippo–YAP signaling pathway may be another modulator of csEV-TYRO3–mediated cancer progression. Here, we showed that YAP activation was correlated with csEV-driven cell proliferation and chemoresistance. Diverse tyrosine kinases, including the TAM receptor family, regulate the Hippo signaling pathway via regulation of large tumor suppressor kinase 1 activity (35). Consistent with our results, phosphorylation of YAP-S127 was inhibited after transfection of TAM receptors including TYRO3 (49). Unexpectedly, our in vivo xenograft assays with TYRO3-knockout-LNCaP-SL and TYRO3-knockout-HCT116 cells showed that tumor growth was not altered by TYRO3 deletion. In fact, TAM receptors did not have much attention firstly owing to their mild oncogenic action. Additionally, MERTK and AXL affects survival rather than proliferation of cancer cells (10). These concepts correspond to our results showing that csEV does not affect proliferation of cancer cells under 10% FBS condition (Fig. 5A), while EV elimination in FBS makes cancer cell vulnerable to chemotherapeutic agent (Fig. 6A). Hence, TYRO3 may confer survival advantage to cancer cells only under stressful conditions (deprivation of growth factors or nutrients and chemotherapy exposure).

The balance between proliferation and apoptosis is often disrupted in tumors, and the acquisition of chemoresistance advances tumor malignancy. In our findings, analysis of clinical GEO data and the immunoblot results using several cancer cell line panels indicated that lung cancer cells resistant to EGFR-TKI or metastatic tumor tissues showed relatively higher TYRO3 expression. Moreover, we found that metastatic prostate cancer cells were more susceptible to docetaxel under EV-free FBS compared with the EV-containing FBS condition, and a combination of docetaxel with BMS-777607 (pan-TAM receptor inhibitor) synergistically increased cytotoxicity compared with each treatment alone. This concept is supported by the finding that the Gas6/TYRO3 axis helps the invasiveness of sorafenib-resistant hepatocellular carcinoma cells (50). Although YAP plays a marginal role in the xenograft model (51), it is actively involved in the growth of chemoresistant tumors, especially EGFR-TKI–resistant cancer cells (32). Our results obtained from the tumor xenograft study with KRCT-6j and gefitinib clearly demonstrated the active involvement of csEV–TYRO3–YAP signaling in the growth of EGFR-TKI–resistant cancer. Moreover, these data indicate the feasibility of using a TYRO3-selective inhibitor with EGFR-TKI for synergistic anticancer effects.

The function of TdEV as an endogenous immunosuppressive factor has garnered much attention (52). Circulating cancer cells have a chance to come into contact with immune cells such as monocyte/macrophage and leukocytes in the blood. Because the TAM receptor family is differentially expressed in several types of immune cells, csEVs may help cancer cells evade immune cells through the activation of TAM receptors. MerTK, which is expressed in most subsets of immune cells, has been extensively noted because of its immunosuppressive activity (10). In addition, PROS1–TYRO3 or –MerTK interaction inhibited macrophage M1 polarization required for immune stimulation (53). Hence, our results raise the possibility that csEV can induce immune suppression via TYRO3 in a vascular network surrounding the metastatic spread of cancer.

In conclusion, we showed that csEVs promote cancer cell migration and proliferation, and that TYRO3 exerts a key role in mediating the effects of csEV through Rho/ROCK and/or YAP signaling. When carcinoma acquires aggressive and chemoresistant phenotypes, TYRO3 expression can be upregulated. Elevated TYRO3 is activated by abundant csEV in the TME, thereby sustaining survival under growth-limiting conditions and promoting metastasis. These findings provide a novel therapeutic target for the prevention or treatment of cancer.

M. Park reports a patent for composition for treatment of circulating small extracellular vesicles-mediated cancer pending. K.W. Kang reports a patent for composition for treatment of circulating small extracellular vesicles-mediated cancer pending. No disclosures were reported by the other authors.

M. Park: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. J.W. Kim: Investigation, methodology, writing–review and editing. K.M. Kim: Writing–review and editing. S. Kang: Data curation, software, formal analysis, methodology. W. Kim: Methodology. J.-K. Kim: Resources, investigation, visualization, methodology. Y. Cho: Investigation, methodology. H. Lee: Investigation. M.C. Baek: Conceptualization. J.-H. Bae: Investigation. S.H. Lee: Validation, investigation. S.B. Jeong: Formal analysis, investigation, visualization. S.C. Lim: Visualization, methodology. D.W. Jun: Resources. S.Y. Cho: Resources, investigation. Y. Kim: Investigation. Y.J. Choi: Investigation. K.W. Kang: Conceptualization, resources, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing.

The authors appreciate members of our laboratories for their kind suggestions. They thank Dr. Sang Kook Lee and Dr. Cheol Ho Jeong for their kind donation of EGFR-TKI–resistant lung cancer cell lines. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (NRF-2018R1A2B2003590). M. Park was a recipient of Basic Science research fellowship through NRF funded by the Ministry of Education (NRF-2020R1A6A3A13070908).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.