Melanoma-Derived Exosomes Induce PD-1 Overexpression and Tumor Progression via Mesenchymal Stem Cell Oncogenic Reprogramming

Recently, it has been described that programmed cell death protein 1 (PD-1) overexpressing melanoma cells are highly aggressive. However, until now it has not been defined which factors lead to the generation of PD-1 overexpressing subpopulations. Here, we present that melanoma-derived exosomes, conveying oncogenic molecular reprogramming, induce the formation of a melanoma-like, PD-1 overexpressing cell population (mMSCPD-1+) from naïve mesenchymal stem cells (MSCs). Exosomes and mMSCPD-1+ cells induce tumor progression and expression of oncogenic factors in vivo. Finally, we revealed a characteristic, tumorigenic signaling network combining the upregulated molecules (e.g., PD-1, MET, RAF1, BCL2, MTOR) and their upstream exosomal regulating proteins and miRNAs. Our study highlights the complexity of exosomal communication during tumor progression and contributes to the detailed understanding of metastatic processes.


INTRODUCTION
Heterogeneous tumor tissue is comprised of a wide variety of collocated cells. Their spatiotemporal co-existence facilitates direct communication between them. Cancer cells contribute to tumor niche formation not only by soluble factor production and receptor-ligand interactions (1), but also by releasing vesicles whose molecular contents add up to a complex information package. Previous studies demonstrated that cultured human tumor cells release extracellular vesicles such as exosomes of 20-120 nm diameters (2). Among others, exosomes carry structural and signaling proteins, MHC molecules, cell surface molecules typically associated with apoptosis, and mRNAs/miRNAs with multiple functions (3). Therefore, these exosomal-molecularpatterns, as unique entities of the complex intercellular communication, are not independent of the quality or state of the mother cell (4,5). Exosomes have been recognized long ago, but their identification, characterization, and isolation are still under intense investigation. Further, whereas the definition of exosomes is based on the pathway of biogenesis, oncosomes form a functional class of extracellular vesicles. Indeed, oncosomes are suggested being capable of carrying and conveying tumor-related information (6) and might have a significant role in formation of tumor microenvironment (7).
Mesenchymal stem cells (MSCs) were first described as stromal cells of the bone marrow with multipotent differentiation potential and characteristic immunomodulatory effects (8). In relation to their immunological and differentiation properties, there is a debate about the role of MSCs in tumor progression (9). Indeed, the cellular fate could depend on the cancer type and also on the status of the affected MSCs. Activated MSCs can secrete pro-angiogenic soluble factors and are able to differentiate to vessel wall associated cells (10) or to cancer associated fibroblasts (CAFs) (11). Furthermore, Baglio et al. showed that tumor secreted extracellular vesicles promote osteosarcoma progression via TGFβ signaling induced IL-6 production by MSCs (12). Moreover, Peinado et al. demonstrated that exosomes from highly metastatic melanomas increase the metastatic behavior of primary tumors by permanently educating bone marrow progenitors through the receptor tyrosine kinase MET (13).
Malignant melanoma is a highly invasive, metastatic cancer with poor prognosis and survival rate. Since melanoma has been long recognized as a highly immunogenic tumor, therapeutic approaches target different immunological mechanisms to treat patients with this aggressive skin cancer. In the past decade, besides IL-2 and CTLA-4 therapies, the PD-1:PD-L1 blockade proved to be an effective treatment in metastatic melanoma (14). Programmed cell death protein 1 (PD-1), mainly expressed by immune cells such as activated T cells, dendritic cells and macrophages is a cell surface receptor with a central role in modulation of T cell responses. PD-1 binding to its ligand, PD-L1, can result in apoptosis and anergy of immune cells.
Notably, PD-1 expression is not a default property of melanoma cells. Of further importance, recently Kleffel et al. demonstrated that PD-1 overexpressing melanoma cell subpopulations are especially aggressive, and that melanoma PD-1:PD-L1 interactions modulate mTOR signaling, whereby they interfere with programmed cell death (15). This work is remarkably instrumental with respect to the consequences of PD-1 overexpression and experimental inhibition of PD-1 functions (Pdcd1-shRNA or anti-PD-1 antibody). However, it does not define the factors the activation of which lead to the generation of PD-1 overexpressing subpopulations. This question, though, is an especially intriguing one, if one considers that one of the most effective immunotherapies available today is based on the blockade of the aforementioned PD-1:PD-L1 interaction (16).
Based on the above, in the current study, we aimed at specifically and systematically addressing the following questions: 1. Can we unambiguously define the cellular and molecular signs of melanoma-derived exosome-induced, intercellular communication-mediated malignant transformation of MSC cultures? 2. Can we detect the melanoma-derived exosome-induced tumor progression in vivo? 3. What alteration can be detected in the expression of PD-1, a melanoma progression marker and therapeutic target, upon exposure to melanoma-derived exosomes?
Here, we present the first piece of evidence that melanoma exosomes, by conveying a complex oncogenic molecular reprogramming, induce the formation of a PD-1 overexpressing cell population (melanoma-like MSC PD-1+ ; mMSC PD-1+ ) from naïve MSCs. These mMSC PD-1+ cells represent a new entity with melanoma-like gene expression profile and phenotypic properties. Further, exosomes and exosomeactivated mMSC PD-1+ cells induce rapid tumor progression in vivo, due to their strong expression of oncogenic dominance factors and decreased susceptibility to programmed cell death.

MATERIALS AND METHODS
Cell Culture and Cell Line B16F1 mouse melanoma cell line was obtained from ECACC and cultured in Dulbecco's modified Eagle's medium (DMEM, Lonza, Basel, Switzerland) containing 10% fetal bovine serum (FBS) (Euroclone, Milano, Italy), 1% sodium pyruvate, 1% MEM non-essential amino acids, 1% MEM vitamin solution, 2 mM L-glutamine, and 1% Penicillin-Streptomycin-Amphotericin B Mixture (all from Lonza). Murine MSCs were isolated from C57BL/6N (Charles River Laboratories, Sulzfeld) 8 week-old male mice (17). Abdominal inguinal fat pads were excised, rinsed with RPMI 1640 medium, transferred to sterile tissue culture dishes and mechanically dissociated. The dissociated tissue was resuspended in RPMI 1640 containing 100 µg/ml collagenase (Sigma-Aldrich, St. Louis, MO) and incubated at 37 • C for 50 min. Collagenase was neutralized with a growth medium containing 10% FBS. After centrifugation at 470 g for 15 min, cell pellets were resuspended and washed in the culture medium. After the centrifugation, cell pellets were resuspended in complete MesenCult medium (Stemcell Technologies, Vancouver, British Columbia, Canada) and filtered through a 100 µm cell strainer (BD Biosciences, Franklin Lakes, NJ) to tissue culture dishes, and cultured for 48 h at 37 • C in 5% CO 2 and 90% humidity. Unattached cells and debris were then removed and fresh medium was added to the adherent cells. The cells were cultured to 80% confluence before being released with trypsin-EDTA and sub-cultured.
Purity of MSC cultures were checked by flow cytometry analysis using the Mouse Multipotent Mesenchymal Stromal Cell Marker Antibody Panel (R&D Systems, Minneapolis, MN) according to the manufacturer's instructions.

Exosome Isolation by Filtration and Differential Centrifugation
Exosomes were isolated by adapting the protocol of Peinado et al. (13) from melanoma cell culture supernatant. B16F1 supernatants were harvested, supplemented by complete protease inhibitor cocktail (Roche, Mannheim, Germany) and centrifuged at 780 g for 5 min at 4 • C to remove intact cells. Then, the supernatants were centrifuged at 3,900 g for 15 min at 4 • C and filtered through a 0.2 µm membrane (Millipore, Billerica, MA) to remove larger cell fragments and microvesicles. Exosomes were pelleted by ultracentrifugation at 150,000 g (T-1270 rotor at 40,500 rpm) for 1 h at 4 • C. The pellet was washed twice and resuspended in DPBS and stored at −80 • C.
The concentration of exosomal proteins was determined using a Pierce BCA Protein assay kit (Thermo Fisher Scientific, Waltham, MA) and a Benchmark Microplate Reader (Bio-Rad, Hercules, CA) according to the manufacturer's instructions.

Scanning electron microscopy (SEM)
Five microliters of isolated exosomes in DPBS were spotted on glass coverslips coated with 0.01% Poly-L-Lysine (Sigma-Aldrich) and incubated overnight at 4 • C. Using sterile forceps, coverslips were transferred into a plastic plate. Exosomes were fixed by gently adding 2% paraformaldehyde buffer pH 7.2 diluted in DPBS for 30 min. The coverslips were washed twice with DPBS and dehydrated with a graded ethanol series (40,60,80, and 100% ethanol, each for 5 min). The samples were dried with a critical point dryer (QUORUM K850, Quorum Technologies Ltd, Laughton, UK) and the coverslips were mounted onto a microscope stub at a time using carbon tape, followed by 7 nm gold coating (QUORUM Q150, Quorum Technologies Ltd) and observed under a field-emission scanning electron microscope (JEOL JSM-7100F/LV).

Atomic force microscopy (AFM)
Exosomes were let to adsorb to the freshly cleaved muscovite mica (SPI-ChemTM Mica Sheets, Structure Probe, Inc., West Chester, PA) surface directly from DPBS. All AFM measurements were carried out with an Asylum MFP-3D head and Molecular Force Probe Controller (Asylum Research, Santa Barbara, CA), using Asylum Research MFP-3D program (version 15.09.112) written in Igor Pro software (version 6.37, Wavemetrics, Lake Oswego, OR). Image procession and data calculation were made using the same software. Silicon rectangular cantilevers (OMLC-AC240TS, Olympus Optical Co., Ltd., Tokyo, Japan) with the typical spring constant of 2 N/m were used in dry condition and silicon nitride rectangular cantilevers with "V" shaped tips (Bio-Lever BL-RC150VB, Olympus Optical Co. Ltd.) with the typical spring constant of 0.03 N/m in buffer solution. Typically, 512 × 512 points was taken at 1 line/s scan rate. The measurements presented here are amplitude images and height profile.

Detection of Exosomal Markers by Western Blotting
Protein samples were resuspended in 4x sample buffer (NuPAGE LDS Sample Buffer (4X), NuPAGE Sample Reducing Agent (10X), Thermo Fisher Scientific), boiled at 96 • C for 10 min, and immediately cooled on ice. Electrophoresis of the proteins was performed using 4-12% Bis-Tris Protein Gels (NuPAGE Novex, Thermo Fisher Scientific), ProSieve Color Protein Markers (Lonza), and electrophoresis buffer (NuPAGE MOPS SDS Running Buffer, Thermo Fisher Scientific) at 200 V and 0.03 A for 40 min.
The proteins of electrophoresed gels were transferred to Immobilon transfer membrane (Millipore, Darmstadt, Germany) using transfer buffer [NuPAGE Transfer Buffer (20X)] at 30 V and 170-110 mA for 60 min. Membranes were blocked in TBST buffer (25 mM Tris-HCl, 150 mM NaCl, 0.05% Tween-20, pH 7.2) containing 5% non-fat milk for 60 min at room temperature (RT). After blocking, membranes were incubated with each primary antibody in TBST buffer containing 1% non-fat milk overnight at 4 • C. Membranes were washed three times for 10 min with TBST buffer and incubated for 60 min at room temperature with secondary antibody in TBST buffer containing 1% nonfat milk. Membranes were washed three times for 10 min with TBST buffer.
The high voltage (1.2 kV) was applied through liquid junction between the chromatographic column and the non-coated silica nanospray emitter (NewObjective, Woburn, MA, 10 µm tip ID). The mass spectrometer operated in data-dependent mode: the survey mass spectra were detected in the orbitrap with high resolution (R = 60 k @ m/z: 400, mass range m/z: 380-1,400) and the most abundant multiply charged 20 peaks were selected for ion-trap fragmentation (NCE: 35%; activation q: 0.25; activation time: 10 ms; minimum signal intensity: 5,000 counts). The MSMS spectra were detected in the ion trap. Dynamic exclusion was used, the precursors were excluded for 15 s after the first fragmentation event.
Data analysis: searchable peaklists (mgf format) were extracted using ProteomeDiscoverer (ver:1.4 Thermo Fisher Scientific) and subjected to database search on our in-house ProteinProspector (ver: 5.14.1) search engine using the following parameters: parent ion tolerant: 5 ppm; fragment ion tolerant: 0.6 Da; Cys carbamidomethylation was set as constant and Met oxidation, cyclisation of peptide N-terminal Glu to pyroglutamic acid, protein N terminal acetylation were set as variable modifications. Only fully tryptic peptides were considered with maximum of 2 missed cleavage sites. The Mus musculus and Bos taurus protein sequences of the Uniprot (UniProtKB.06.11.2014) database completed with human keratins and pig trypsin, altogether 106,330 protein sequences were searched. For the false discovery rate (FDR) estimation, the searches were performed on the database concatenated with the randomized sequences. Protein identification was accepted if the ProteinProspector expectation value was <0.01 and the protein was identified with at least 2 unique peptides (expectation value <0.05 and score higher than 15). FDR values were <1% in all cases.
For functional validation, the resulted protein list was analyzed by the "Core Analysis" function included in Ingenuity Pathway Analysis (IPA, Quiagen Bioinformatics) software.

Isolation and Detection of Exosomal miRNAs
miRNA sequencing was performed using SOLiD Total RNA-Seq lit for Small RNA Libraries (Applied Biosystems now part of Thermo Fisher Scientific) according to the manufacturer's instructions. Purification was performed on 10% TBE-Urea gels stained with Sybr Gold nucleic acid gel stain (both from Invitrogen now part of Thermo Fisher Scientific). Final purification was performed using PureLink PCR Micro Kit (Invitrogen). Final libraries were quality checked using High Sense DNA kit on Bioanalyzer (all from Agilent, Santa Clara, CA). Concentration of each library was determined using the SOLID Library TaqMan Quantitation Kit (Life Technologies now part of Thermo Fisher Scientific). Each library was clonally amplified on SOLiD P1 DNA Beads by emulsion PCR (ePCR). Emulsions were broken with butanol, and ePCR beads enriched for template-positive beads by hybridization with magnetic enrichment beads. Template-enriched beads were extended at the 3 ′ end in the presence of terminal transferase and 3 ′ bead linker. Beads with the clonally amplified DNA were deposited onto SOLiD sequencing slide and sequenced on SOLiD 5500 Instrument using the 50-base sequencing chemistry.
Bioinformatic Analysis Raw data quality assessment, read trimming read mapping and miRNA expression profiling were carried out in CLC Genomics Workbench tool version 8.0.2 (CLC Bio now part of Qiagen, Venlo, Netherlands) using annotated Mus musculus miRNA sequences according to the miRBase release 21 as a mapping reference.

Visualization of Labeled Exosome Internalization in MSCs
To examine the uptake of exosomes by MSCs, cells were plated to black 24-well Visiplates (1 × 10 4 cells/well) and incubated for 24 h. The exosomes were labeled with Dil dye (1,1 ′ -dioctadecyl-3,3,3 ′ ,3 ′ -tetramethylindocarbocyanine perchlorate, PromoKine, Heidelberg, Germany) and the MSC cultures were labeled with DiO dye (3,3 ′ -dioctadecyloxacarbo-cyanine perchlorate, PromoKine) according to the manufacturer's instructions. Dillabeled exosomes were washed in DPBS by ultracentrifugation (at 150,000 g for 1 h at 4 • C). Forty micrograms per milliliter DiL-labeled exosomes were added to DiO-labeled MSC cultures and the exosome uptake was followed for 24 h in the Celldiscoverer 7 automated live cell imaging system (Zeiss, Oberkochen, Germany). After 24 h, the cells were fixed with 4% paraformaldehyde solution and a nucleus staining was performed using DAPI (Life Technologies now part of Thermo Fisher Scientific). Then, 5 image z-stacks were acquired for both channels by Operetta High Content Screening System (Perkin Elmer, Waltham, MA). The stacks were maximum intensity projected and then analyzed automatically using a customized version of CellProfiler (18). Nuclei were detected with Otsuadaptive threshold combined with diameter based filtering, then cytoplasms were identified with propagation method seeded from the nuclei and using the exosome channel. Exosomes were located with a customized version of A-trous wavelet transform based spot detection (19). Several wavelet levels were used to ensure the detection of exosomes with various size and then the overlaps were removed based on circularity measures. Finally, the exosome numbers per cell were identified using MATLAB programming, the diagrams were created in Microsoft Excel.

Cell Proliferation
After 72 h incubation, exosome-exposed and control MSC cultures were dissociated with trypsin from the culture surface. Cells were washed in medium and counted in a Bürker chamber and a cell counter (Bio-Rad, TC10 Automated Cell Counter).

Detection of Apoptosis
Exosome-exposed MSCs and control cells were treated with 100 ng/ml mouse TNFα (R&D Systems). After 24 h incubation, QRT-PCR was performed on a RotorGene 3000 instrument (Corbett Research, Sydney, Australia) with gene-specific primers and SybrGreen protocol to monitor gene expression. One microgram of total RNA was reverse transcribed with random primers using the High-Capacity cDNA Archive Kit (Applied Biosystems) according to the manufacturer's instructions in final volume of 30 µl. The temperature profile of the reverse transcription was the following: 10 min at room temperature, 2 h at 37 • C, 5 min on ice, and finally 10 min at 75 • C for enzyme inactivation. These steps were carried out in a Thermal Cycler machine (MJ Research, Marshall Scientific, Hampton, NH). After dilution with 30 µl of water, 1 µl of the diluted reaction mix was used as template in the QRT-PCR. Reactions were done with FastStart SYBR Green Master mix (Roche) according to the manufacturer's instructions at a final primer concentration of 250 nM under the following conditions: 15 min at 95 • C, 40 cycles of 95 • C for 15 s, 60 • C for 25 s, and 72 • C for 25 s. The fluorescence intensity of SybrGreen dye was detected after each amplification step. Melting temperature analysis was done after each reaction to check the quality of the products. Primers were designed using the online Roche Universal Probe Library Assay Design Center. The quality of the primers was verified by MS analysis provided by Bioneer (Daejeon, Republic of Korea). Relative expression ratios were calculated as normalized ratios to MmRpl27 (Mus musculus ribosomal protein L27) gene. Nontemplate control sample was used for each PCR run to check the primer-dimer formation. The final relative gene expression ratios were calculated as delta-delta Ct values. Information about the genes and the primers is collected in Table S1.

RNA Isolation, Reverse Transcription, and TaqMan Panel
RNAs were isolated from MSC cells with Qiagen RNeasy mini kit (Qiagen) based on the manufacturer's instruction. RNA concentrations were measured by NanoDrop spectrophotometer (NanoDrop). cDNAs were reverse transcribed with TaqMan Reverse Transcription Reagents (Thermo Fisher Scientific) following the manufacturer's instruction. For the TaqMan QRT-PCR panel, cDNA mixture of three parallel samples was used in case of each condition: control, 6, 24, and 72 h exosome treatment applied on MSC cells derived from four mice. Eighty nanogram cDNA and TaqMan Gene Expression Master Mix (Thermo Fisher Scientific) were used for the qPCR experiment. Expression of the examined 44 genes was calculated by Ct method and normalized to the average Ct values of 4 internal controls (PPIA, 18S RNA, ACTB, and GAPDH).

PD-1 detection by Western blotting
The cells were washed three times in PBS, then were lysed in TENT Buffer (50 mM Tris-HCl, 2 mM EDTA, 150 mM NaCl, 1% TritonX-100, completed with 1x protease inhibitor cocktail (Roche). Protein samples were separated and blotted as described above.

PD-1 detection by STORM super-resolution imaging
All dSTORM super-resolution experiments were performed on a custom-made inverted microscope based on a Nikon Eclipse Ti-E frame. After being conditioned (through spatial filtering via fiber coupling and beam expansion) the applied laser beams were focused into the back focal plane of the microscope objective (Nikon CFI Apo 100x, NA = 1.49), which produced a collimated beam on the sample. The angle of illumination was set via a tilting mirror mounted into a motorized gimbal holder and placed into the conjugate plane of the sample. All the dSTORM images were captured under EPI illumination at an excitation wavelength of 647 nm (Nikon: 647 nm, 300 mW). The laser power, controlled via an acousto-optic tunable filter (AOTF), was set to 4 kW/cm 2 on the sample plane. An additional laser (Nichia, Anan, Tokushima, Japan, 405 nm, 60 mW) was used for both reactivation and reference measurements. Images were captured by an Andor iXon3 897 BV EMCCD digital camera (512 × 512 pixels with pixel size of 16 µm). The size of the illuminated region of the sample was matched to the size of the detector, which determined the field of view (FOV = 80 × 80 µm 2 ). Frame stacks for dSTORM super-resolution imaging were typically captured at a reduced image size (crop mode), when only the central 128 × 128 pixel region was selected. A fluorescence filter set (Semrock, Rochester, NY, LF405/488/561/635-A-000) was used to select and separate the excitation and emission lights in the microscope. Additional emission filters (Semrock, BLP01-647R-25) were used in the detection path to further clean the fluorescent light spectrally for the reduction of spectral crosstalk.
During the measurements, the perfect focus system of the microscope was used to keep the sample in focus with a precision of <30 nm. The storage buffer on the sample was replaced with a special switching buffer (20). Typically, 10,000 frames were captured with an exposure time of 30 ms. Reference images with full size FOV were captured at low intensity when the majority of fluorescent molecules were still active and the overall structure of the labeled sample could be visualized.
The captured and stored image stacks were evaluated and analyzed by rainSTORM localization software (21). The individual images of single molecules were fitted with a Gaussian point spread function and their center positions were associated with the position of the fluorescent molecule. Localizations were filtered via their intensity, ellipticity and standard deviation values. Localizations with precisions of <45 nm were only used to form the final image. The estimated mean precision of the accepted localizations was 19 nm. Mechanical drift introduced by either the mechanical movement of the sample or thermal effects was analyzed and reduced by means of a blind drift correction algorithm. Spatial coordinates of the localized molecules were stored and the final super-resolved image was visualized. The multicolor merged images were generated by ImageJ software.

In vivo Experiments
Mouse Model B16F1 melanoma cells were administrated intravenously (1 × 10 5 cell/100 µl) to 6-8 week old female C57BL/6N mice (Charles River Laboratories). One week later, tumor bearing mice were randomized and divided into 3 groups (n = 10/group). Mice were injected intravenously with control buffer (100 µl), exosome exposed MSCs (1 × 10 5 cell/100 µl) or exosomes (40 µg/100 µl) 7, 8, 9, 10, 11 days after the tumor injection (Table S2). One week after the first MSC administration, 3 animals/group were euthanized, their lungs were removed, photographed and stored at −80 • C for further protein, mRNA and histological analyses. The remained mice were observed for 10 more days. At the end point, the animals were euthanized, and the tumor metastases were investigated not only in their lung, but also in their entire body and removed for histological analysis.

Tumor Coverage
Tumor coverage of lungs was determined by the analysis of acquired images using the ImageJ software. The area of tumors and the healthy regions was measured and the mean percentage, SD, and p values were calculated in Microsoft Excel.

Quantification of Metastases Associated Vessel Diameters
From native animal lungs 4 µm criostat sections were made on silanized slides, than fixed and retrieved by Fix and Perm A-B solution (Thermo Fisher Scientific, USA) for 20 min. Sections were counterstained by conventional hematoxilin for 30 min than washed in tap water and coverslipped. Sections were digitalized by automatic slide-scanner (3DHistech, Hungary), using the software 3DHISTECH Pannoramic Viewer (3DHistech, Hungary). Strict tumor border was carefully marked then vessel diameters were measured.

Cytokine and Chemokine Array
Lung samples were lysed in NP40 cell lysis buffer (Thermo Fisher Scientific, USA) and protein content was measured by the Pierce BCA Protein Assay kit (Thermo Fisher Scientific, USA). Expression levels of different cytokines in pooled lung specimens were determined using Mouse Cytokine Array Panel A (R&D Systems, Cat. no. ARY006), according to the manufacturer's instructions. Immunoreactive signals were detected by using LI-COR ODYSSEY Fc (Dual-mode imaging system) imager followed by analysis with Image Studio Lite v5.2.

Custom TaqMan Array Panels
QRT-PCR was performed on an ABI Prism 7000 sequence detection system (Applied Biosystems) using specific predesigned customized 96-well TaqMan Array oncogene panels (Thermo Fisher Scientific) containing 44 specifically selected primers and probes according to extensive literature work. Total RNA was isolated using TRIzol (Thermo Fisher Scientific), DNase treatment was performed according to the manufacturer's protocol, and then total RNA was transcribed into cDNA using High Capacity cDNA Kit (Thermo Fisher Scientific). PCR amplification was performed using TaqMan primers and probes and thermal cycle conditions were set as follows: 2 min at 50 • C, 10 min at 95 • C, and 40 cycles of 15 s at 95 • C and 1 min at 60 • C. As internal control transcripts of ACTB (Assay ID: Mm00607939_s1), GAPDH (Assay ID: Mm99999915_g1), and PPIA (Assay ID: Mm02342430_g1) were determined. The amount of the transcripts was normalized to those of the housekeeping genes using the Ct method. Finally the results were normalized to the expression of the vehicle control ( Ct method).
For relationship discovery Hierarchical Cluster Analysis was performed by R software. In detail, the "bottom up" agglomerative hierarchical clustering strategy was used and results represented in a tree-based dendrogram [For refer the R software: (22)].

Network Representation by Ingenuity Pathway Analysis
According to literature data, we established a protein network from the in vivo overexpressed genes. A custom graphical representation of this network was generated using the Path Explorer tool of the IPA Path Designer. Genes are represented as red nodes, using various shapes that represent the functional class of the gene product. In a few cases proteins are substituted with the complex, which they are involved in. To identify the potential exosomal factors, which can induce the activation of the network, a list was generated from the exosomal proteins detected by LC-MS/MS and the exosomal miRNAs identified by SOLiD sequencing. The Grow tool in the IPA Path Designer revealed significant interactions between the network and the generated list (the interacting exosomal factors are listed in gray boxes and the types of relationships are indicated in parentheses). During construction in the IPA, the significance level was set to "experimentally observed" data to avoid the representation of predicted, unproven interactions.

Isolated Vesicles Show Exosomal Properties
First, we isolated extracellular vesicles from B16F1 mouse melanoma cells. As it was shown by SEM and AFM (Figures 1A,B), the isolated fraction indeed contained exosomes as the particles were cap-shaped, and their size was within the 40-120 nm range. Then, presence of molecules (CD9, CD63, CD81, and HSP70), characteristic for exosomes (4,17) was assessed by Western blotting (Figure 1C). Exosomes were then subjected to large-scale analysis to determine their protein and miRNA profiles. Whole proteome analysis (using LC-MS/MS) and bioinformatics tools (used to identify elements of the exosome's proteomics spectrum in the UniProt database, and to compare the findings with entries of the ExoCarta database) revealed that 95 distinct proteins were identified in melanoma exosomes ( Table 1). These proteins exhibited 86.3% overlap with molecules listed as characteristic exosome protein markers in ExoCarta. Further, miRNA sequencing (using SOLiD 5500xl technology) identified 168 known miRNA elements ( Table 2) which, similarly to the proteomics data, exhibited a large (93.5%) overlap with molecules listed as characteristic exosome miRNA markers in ExoCarta.
To uncover the functional significance of the proteomics and miRNA sequencing data, IPA was employed. This analysis has shown that the identified proteins most probably participated in cellular and molecular processes such as "Cell Death and Survival, " "Cellular Movement, " "Cell-to-Cell Signaling and Interaction, " "Cellular Growth and Proliferation, " and "Cell Morphology" (P range = 7.53 × 10 −15 -9.32 × 10 −4 significance range) (Figure 1D). Very similarly to these data, functions of the identified miRNAs were suggested to be linked to mechanisms of "Cellular Development, " "Cellular Growth and Proliferation, " "Cellular Movement, " "Cell Cycle, " and "Cell Death and Proliferation" (P range = 1.25 × 10 −12 -4.88 × 10 −2 significance range) (Figure 1E).

Tumor Exosome Exposure Resulted in Oncogenic Reprogramming of MSCs in vitro
Then, we investigated the effect of exosomes on biological processes (e.g., proliferation, survival, malignant transformation, etc.) of MSCs, which are generally considered as proper in vitro models of tumor stroma (12). For these experiments, MSC     (3) lipid dye] as early as 1-2 h after application ( Figure S1). Importantly, after 24 h, the majority of MSCs were loaded by exosomes (Figures 2A,B). Indeed, image analysis and statistical evaluation revealed a 91% internalization efficacy. This suggests that functional alterations demonstrated by this study were due to exosome-induced cell-population, and not individual cell level effects.
Then, we determined whether the internalized exosomes could induce a melanoma-like malignant transformation of the MSCs. By employing two complementary cell-counting methods, we found that proliferation rate of the MSCs significantly accelerated 72 h after exosome exposure ( Figure 2C). By flow cytometry, we also showed that the exosome-treated MSCs exhibited a partial resistance to the cell death-inducing effects of 100 ng/ml tumor necrosis factor-α (TNFα) as the fraction of the dead cells was significantly decreased in these cultures ( Figure 2D).
Since the exosomes were isolated from melanoma cells, we were then intrigued to uncover whether the above alterations (which all argue for the malignant transformation of the susceptible cells) also resulted in de novo appearance of melanoma-specific features in the transformed MSCs. To answer this question, expression of the melanoma-specific markers MLANA and MITF were investigated. By QRT-PCR, we found that mRNA transcript levels of both markers elevated markedly in MSCs upon exosome treatment (Figure 2E and Figure S2), albeit the kinetics of elevation of the two molecules were slightly different. Moreover, in good accordance with the mRNA data, immunofluorescence labeling showed that exosome exposure markedly increased the expression of MLANA at the protein level as well ( Figure 2F).
Next, we assessed whether the above effects of exosomes inducing a malignant-like transformation of the MSCs was accompanied by a cellular-molecular oncogenic reprogramming of the target cells. Naïve MSCs were exposed to a standardized volume of exosomes for various time intervals (to avoid the experimental fluctuations, we collected pooled samples from multiple independent in vitro experiments). Then, samples were subjected to QRT-PCR analysis using a self-designed panel of 40 oncogenes and tumor suppressor genes which were previously suggested to play a role in melanoma progression ( Table 3).
As shown in Figure 3A, gene expression pattern of MSCs exposed to melanoma exosomes, exhibited a clear oncogenic dominance (compared to the non-exposed cells). This was verified by statistical analysis of the mean relative gene expression levels of all molecules investigated. Statistically higher values were obtained in the case of exosome-treated cells (p = 1.9 × 10 −5 , p = 0.031 and p = 2.3 × 10 −8 for the 6, 24, and 72 h time points, respectively).
As mentioned above, Kleffel et al. has recently shown that melanoma cell subpopulations which overexpress PD-1, quite intriguingly exhibit remarkably increased invasiveness and aggressive growth properties (15). However, the authors did not define the factor(s) which induced the above PD-1 overexpression. Since the above finding strongly suggested an "MSC re-education" capacity of melanoma exosomes to induce malignant-like behavior, we next assessed the expression of PD-1 in MSC cultures.
As expected, only insignificant PD-1 expression (both at the mRNA and protein levels) could be identified in control, nontreated MSCs. In contrast, a marked, significant, and timedependent elevation of PD-1 expression was detected upon exosome treatment by QRT-PCR (Figure 3B), Western blot and an immunocytochemical analysis (Figures 3C,D). Further, by employing super-resolution microscopy, we were able to identify a dramatic upregulation of PD-1 at the single molecular level in exosome-treated MSCs ( Figure 3E).
Importantly, since proteomics analysis did not identify the presence of PD-1 in exosomes, these data suggest that the high PD-1 protein content in exosome-exposed MSCs was a result of de novo induction and not of exosome-mediated molecular transfer. Our findings therefore suggest that melanoma exosomemediated "re-education" of the cells resulted in a novel MSC population which could be identified as MSC PD-1+ .

B16F1 Exosomes Augment in vivo Tumorigenesis and Tumor Progression
After presenting evidence for the in vitro tumorigenic induction potential of exosomes on cultured MSCs, we hypothesized that this phenomenon could be identified in vivo as well. To probe this assumption, we employed the well-known animal model, routinely used in our laboratories (46), in which tumors, developed mostly in the lungs, are induced in mice by intravenous administration of mouse B16F1 melanoma cells (to the tail vein). Then, tumor-bearing mice received buffer, or exosomes isolated from the same B16F1 melanoma cells, or exosome-induced MSC PD-1+ cells.
Notably, the exosome-related groups (i.e., exosome, MSC PD-1+ ) were characterized by a markedly increased size of tumor-covered lung tissues (the increase proved to be significant in the MSC PD-1+ group) (Figure 4A). Of further (green lipid dye)-labeled cells were exposed to Dil (red lipid dye)-labeled exosomes for 24 h and fixed in 4% PFA. Control cells were treated in the same manner without exosomes. Pictures were acquired by an Operetta high content screening system (Perkin Elmer). Lower two images represent an experiment, where nuclei were counterstained with DAPI. Exosomes were detected with a customized version of A-trous wavelet transform, and were highlighted with yellow. (B) Quantitative analysis of exosome uptake. The graph shows the number of detected exosomes (Y axis) in each cell (X axis) in three exosome-exposed (24 h) and three control cell cultures (ctrl). (C) Cell proliferation assay of exosome-exposed MSCs. Cells were plated at 1 × 10 4 cell/cm 2 density in control and exosome-exposed cultures, and 72 h after the exosome treatment, the cell number was determined by manual counting and by an automated cell counter. Both methods showed a significantly increased cell proliferation of exosome-treated cells. Results are presented as mean ± SD (n = 3). (D) Apoptosis analysis of exosome-exposed MSCs by flow cytometry. Exosome-pre-exposed cells were treated by 100 ng/ml TNFα for 24 h, stained with Annexin V-FITC and propidium iodide. They were analyzed by flow cytometry both in the case of the TNFα-treated and the untreated control cell cultures. The graph represents the percentage of early apoptotic, late apoptotic and necrotic cells. Both TNFα-induced total cell death (Σ) and necrosis were significantly lower (Σp = 0.03 and p = 0.02, respectively) in exosome-pretreated cell cultures compared to the corresponding control cells. Results are presented as mean + SD (n = 3). (E) QRT-PCR analysis of Mlana and Mitf in MSC cultures treated by exosomes in every 24 h indicated by arrows in the graph. The expression of both mRNAs increased after exosome exposure, but they showed different kinetics. (F) Fluorescent immunocytochemistry of MLANA in exosome-exposed MSC cultures using a primary rabbit antibody to MLANA and a secondary AlexaFluor555-conjugated antibody to rabbit IgG (red). α-tubulin network of cells was directly labeled by an AlexaFluor488-conjugated antibody (green) and the nuclei were stained with DAPI (blue). Can play multiple roles in diverse processes related to melanoma development, dormancy, migration/invasion, and angiogenesis Ets1 ETS1 E26 transformation-specific Is required for migration in cell lines with an active RAS/ERK signaling pathway (29) Hgf HGF Hepatocyte growth factor Can activate the MAP-kinase pathway, which is upregulated in the majority of melanoma, through the proto-oncogene c-MET Jak2 JAK2 Janus kinase 2 Is an activator of transcription (STAT) pathway is thought to play a central role in melanoma cell biology Met MET Hepatocyte growth factor receptor Induces several biological responses that collectively give rise to a program known as invasive growth  Mtor mTOR Serine/threonine-protein kinase Mechanistic target of rapamycin, is a serine/threonine protein kinase that regulates cell growth, cell proliferation, cell motility, cell survival (29) Akt1 AKT Protein kinase B Plays a key role in multiple cellular processes such as glucose metabolism, apoptosis, cell proliferation, transcription, and cell migration (29) Map2k1 MEK1 Dual specificity mitogen-activated protein kinase kinase 1 Is an essential component of the MAP kinase signal transduction pathway, this kinase is involved in many cellular processes such as proliferation, differentiation, transcription regulation, and development (29) Map2k2 MEK2 Dual specificity mitogen-activated protein kinase kinase 2 Plays a critical role in mitogen growth factor signal transduction. It phosphorylates and thus activates MAPK1/ERK2 and MAPK2/ERK3 Mapk3 ERK1 Extracellular-signal-regulated kinases Ras-Erk1/2 is a key regulator pathway in melanoma cell proliferation (29) Mapk1 ERK2 Mitogen-activated protein kinase 1 Ras-Erk1/2 is a key regulator pathway in melanoma cell proliferation (29) Rac1 RAC1 Ras-related C3 botulinum toxin substrate 1 Functions in multiple signaling pathways are leading to cell adhesion, migration, proliferation, and transformation (45) importance, we also found that in both exosome-related groups, the numbers of distant metastases were significantly elevated when compared to the control (Figure 4B), and diameter of metastases associated blood vessels significantly increased in exosome and MSC PD-1+ injected groups of tumor bearing mice ( Figure S3). Using of cytokine and chemokine arrays, tumor supportive cytokine and chemokine levels were elevated in exosome related groups ( Figure S4). As we have seen during our previous studies (46), these metastases were mostly localized to the ovaries and kidneys (and very rarely to the lymph nodes) of control tumor-bearing animals. However, besides these sites, the presence of exosomes resulted in frequent metastases in the lymph nodes and, as a new location, in the liver. Interestingly, in MSC PD-1+ treated mice, exosome-transformed MSCs could be identified in the para-aortic lymph nodes by FISH (Figure 4C) verifying the successful in vivo adherence of MSC PD-1+ cells. Lung tissues of the different groups were then subjected to indepth expressional profiling 14 days after injection of exosomes or MSC PD-1+ cells. Namely, QRT-PCR analysis was performed using a self-designed panel of 40 genes; besides housekeeping genes, we assessed expression of (i) proto-oncogenes; (ii) genes reportedly involved in malignant transformation; and/or (iii) genes found to be involved in melanoma development and progression.
Furthermore, we showed that, besides the above genes, expression of PD-1 was also significantly increased in both exosome-related groups ( Figure 4E). Notably, although mRNA transcript level of PD-1 was close to two-fold in MSC PD-1+ lung tissues in comparison to samples in the exosome group, the difference was not significant (most probably due to the large inter-animal variability and standard error).
We constructed a Venn-diagram ( Figure 4F) to show all possible logical connections between the various gene expression patterns presented in Figure 4D. Importantly, according to cluster analysis, we could not identify a single gene which was missing from the exosome-related groups in comparison to the control tumor-bearing mice. In other words, whereas these exosome-related groups did exhibit gene expression profiles which were characteristic only for them, such individual profiles could not be detected in the control group.
Finally, it should be noted that the dramatic gene expression alterations seen in the exosome-related groups were exclusively due to the presence of the exosomes as the "MSC PD-1+ only" cluster contained no genes ( Figure 4F).

Melanoma-Derived Exosomes Promote
Tumorigenic and Cell Survival Signaling Pathway(s) Figure 5 shows the map of interaction pathways based on overexpressed molecules (red symbols) in in vivo experiments. We detected overexpressed elements of three main pathways h exosome-exposed MSC cultures, using a primary rat antibody to PD-1 and a secondary AlexaFluor647-conjugated antibody to rat IgG (red). Nuclei were stained with DAPI. (D) α-tubulin network of cells was directly labeled by an AlexaFluor488-conjugated antibody (green). Images were acquired by confocal microscopy. (E) Images were taken by STORM super-resolution microscopy. STORM super-resolution imaging of PD-1 revealed that PD-1 was localized mostly around the nucleus, which were blurred and were not resolvable using diffraction-limited confocal microscopy (D) of the same region.
which do participate in tumor progression and metastasis formation (47). Using the IPA Path Designer Grow tool, we generated an interaction map which contains proteins (or their established complexes) encoded by the overexpressed genes, and exosomal miRNAs and proteins (gray boxes) which were previously shown to control or affect the marked signaling molecules.
To demonstrate these results and to explore the potential causative exosomal factors, we again performed an IPA. The analysis allowed further refinement of the underlying molecular  built from the overexpressed genes, and their complexes known to be related to tumor progression using the Path Explorer tool in the IPA Path Designer. Then a list was generated that contained the exosomal proteins detected by MS and the exosomal miRNAs identified by SOLiD sequencing. The Grow tool in the IPA Path Designer revealed significant interactions between the exosome-induced expression network and the molecular content of these vesicles. The resulting interaction network shows both direct and indirect interactions, but it contains only experimentally observed relationships based on the Ingenuity Knowledge Base. This network demonstrates that 61 exosomal molecules ( Table 4)

DISCUSSION
Considering the fact that metastatic complications are responsible for almost 90% of cancer-driven mortality (48), a deep understanding of tumor metastatic processes is one of the most important challenges in both fundamental and applied research directions in tumor biology. Whilst certain organotropic aspects of tumor cells were addressed by Paget's "seed-and-soil" hypothesis in the early nineteenth century (49), the detailed description of the spatiotemporal characteristics of the metastatic processes requires an extended model. MSCs, as multipotent stromal cells, are important players in dynamic tumor microenvironment (12). MSCs could stimulate angiogenesis via paracrine signaling and participate in metastasis formation to promote the epithelial-mesenchymal transition (12). Consequently, in the last decade, numerous studies suggested the importance of soluble and vesicle-like components of the tumor microenvironment in epithelial-mesenchymal transition as crucial factors in the metastatic establishment (50). Among these, recent findings of intercellular communication mechanisms unquestionably highlighted the crucial role of EVs (e.g., exosomes) serving as "information packages" within the coordinated cell-to-cell signaling (51).
In the current study, we demonstrate that MSCs, widely abundant in solid tumors as well as in healthy tissues, undergo a marked re-education process upon communication with metastatic cancer cells via exosome-mediated information transfer. This transformation process results in characteristic response-patterns corresponding to a given cancerous cell lines (12).
First, we intended to precisely define the EV population which we later employed via in vitro and in vivo experimental approaches. Besides applying the experimental requirements of the International Society for Extracellular Vesicles (ISEV) (i.e., determination of shape, size distribution and protein markers) (52), we thoroughly defined the miRNA and protein content of melanoma exosomes. These systematic assays concluded that the isolated vesicles contained factors, which may regulate signaling pathways related to cell death and survival, cellular movement, development and cell proliferation. Given the appropriate uptake of these vesicles by the targeted, multipotent recipient cells, it is highly probable that most signaling pathways mentioned above were significantly affected. In accordance with previous results in the literature, we showed successful internalization of exosomes by recipient MSCs. As expected, following 72 h of exposure, MSCs displayed accelerated proliferation presumably initiated by the transforming exosome cargo. Moreover, melanoma exosomes were able to hinder the induction of the cell death response commonly triggered by TNFα. TNFα is a frequently identified, soluble, tumor-suppressing factor in cancerous microenvironments (53). This inhibition strongly suggested that exosomes initialize and sustain malignant transformation of MSCs.
Further, we described numerous, exosome-induced phenotypic manifestation of the above malignant re-education process in the recipient cells. Apart from identifying tumor markers, the transformation of exposed cells is characterized by markedly modified gene expression patterns. As discussed previously, specific gene expression alterations are often associated with specific steps of cancerous transformation. Malignant cell lines possess altered expression of genes responsible for apoptotic regulation (Bax, Bcl2, caspases), regulating metastasis (integrins, ITGA2-6, ITGB1, KIT) or various growth factors (ETS1, HGF, MET, etc.). Importantly, we successfully identified two melanoma-specific markers, i.e., MITF (Microphthalmia-associated transcription factor) and MLANA (Melanoma antigen recognized by T-cells) in the exosome-exposed MSCs. MLANA and its regulator MITF play a fundamental role in melanocyte development, tumor progression and are overexpressed in melanoma cells (54). The presence of these markers in exosome-exposed MSCs shows effective transfer of the encoded tumorigenic information, and may have a pivotal role in disease progression. Besides, we also identified a characteristic set of genes, belonging mostly to protooncogenes, whose expression was significantly elevated upon exosome treatment. Among these molecules, we emphasize the increased expression of CD44 which was previously assigned as cancer stem cell marker and of mTOR, a characteristic indicator of the engagement of the PD-1:PD-L1 pathway (15). Indeed, activation of mTOR signaling is particularly important. As it was previously demonstrated by Kleffel et al., overexpression of PD-1 in exceptionally aggressive melanoma subpopulations facilitated tumor progression by activating the mTOR signaling pathway (actually, this response was highly unexpected compared to that of the classic PD-1:PD-L1 dependent T-cell anergy, a crucial aspect of immunotherapy in the clinics) (15). Quite strikingly, in exosome treated MSCs, we detected a significant induction of PD-1 and mTOR. These novel data strongly support that these vesicles might trigger the formation of an aggressive, melanoma-like subpopulation of re-educated recipient MSCs.
Our findings therefore propose that the re-educated melanoma-like MSC PD-1+ subpopulation, originating from multipotent tissue-derived MSCs, is an autonomous entity. It is characterized by typical properties of cancerous transformation such as hyperproliferation, resistance to apoptosis, expression of melanoma markers, proto-oncogenic gene expression patterns, and PD-1 over-expression. Indeed, those melanoma exosomes that induced in vitro generation of the melanoma-like MSC PD-1+ cells, also facilitated in vivo metastasis formation in the lungs or other distant organs of treated animals. Actually, our findings that melanoma exosomes or exosome re-educated MSC PD-1+ promote tumor progression, are in good accordance with previous finding of Kleffel et al. describing that shRNA or antibody mediated inhibition of PD-1 signaling inhibits metastasis formation and tumor progression in experimental animals (15). Assessment of the expression patterns of genes related to tumor progression in the exposed animal tissues revealed numerous cases of exosome-induced overexpression; this supports our hypothesis that melanoma exosomes should be considered as oncosomes (6).
Our results also demonstrated that melanoma exosomes also generated a characteristic signaling pattern in MSC PD-1+ recipient population. As it was shown in numerous studies PI3K/Akt, Ras/MAPK, and STAT3 pathways are commonly activated in tumor cells (47,55). In this study, based on our in vitro and in vivo results and literature data, we introduce a novel, comprehensive network of common tumorrelated proteins, such as PD-1, MET, RAF1, STAT3, BCL2, or mTOR. The network highlights not only the relationship of these elements, but also contains upstream exosomal regulating factors which may contribute to the activation of tumorigenic signaling, and hence, fast tumor progression.
According to our best knowledge, this associative network between overexpressed genes and the potential exosomal inducers, is the first tumor progression signaling pattern which connected experimental response-patterns with experimentally detected exosomal-molecular-patterns. Further, our data also indicate that the above complexity of exosomal communication requires system-level approaches.
Finally, it is important to note that our conclusions are not based on in silico predictions exclusively, but rather on carefully designed and systematically executed in vitro and in vivo experiments, all suggesting that the specific, robust, molecular content of the isolated exosomes can indeed generate a unique intercellular niche responsible for the re-education of neighboring cells via oncogenic transformation. These reeducated melanoma-like MSC PD-1+ cells, in turn, facilitate metastatic disease progression via ignition of a complex series of subsequent events both locally and systematically. Thus, based on our results and recently published, additional evidences about exosomes, there is an urgent demand to supplement and extend the outdated "seed-and-soil" hypothesis with the oncosomedriven re-education process.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the manuscript/Supplementary Files.