Exosomes Associated with Human Ovarian Tumors Harbor a Reversible Checkpoint of T-cell Responses

Effector memory T cells present in the microenvironments of human tumors are hyporesponsive to activation via the T-cell receptor (TCR; refs. 1–5). Multiple cells and factors have been reported to contribute to the arrest of the antitumor response of T cells present in the microenvironment of human tumors (6). The unresponsiveness of T cells in tumors is due in part to an arrest in the T-cell signaling cascade that occurs following activation (7). The T cells are quiescent, but not functionally inert, as the tumor-associated T cells can be activated in tumor xenografts by treatment with IL12-loaded liposomes, a mechanism that bypasses TCR-induced activation (1). These activated T cells can kill tumor cells.

Previously, it was established that a noncellular component of ovarian tumor microenvironments induced the TCR signaling blockade in both tumor-associated T cells and in T cells isolated from normal donor peripheral blood lymphocytes (PBL; ref. 7). Many biologically active immunosuppressive soluble factors have been reported to be present in ovarian tumor ascites fluids (7). Ovarian tumor ascites fluids also contain extracellular vesicles that may impact tumor progression (8–10). The tumor-associated vesicles, often referred to as exosomes, are spherical membrane-bound particles with an average diameter of 50 nm with characteristic marker proteins (11–14). These exosomes have been reported to be both immunosuppressive and immunostimulatory, depending upon their surface phenotype and the intravesicular cargo (15, 16). Exosomes increase in number with tumor progression (15). A considerable controversy currently exists as to whether exosomes mediate a loss or gain of an antitumor immune response and how these vesicles function. Given their presence in immunosuppressive tumor microenvironments, most emphasis has been placed upon determining mechanisms by which exosomes (present in tumor ascites, solid tumors, and serum) inhibit immunocompetent cells in cancer patients.

Exosomes isolated from tumor microenvironments have been suggested to suppress antitumor responses indirectly by augmenting the function or preventing the apoptosis of T regulatory cells, generating myeloid-derived suppressor cells (MDSC), and by blocking the maturation of dendritic cells and macrophages (15, 17–20). Several direct mechanisms have also been proposed to explain how exosomes may arrest T-cell function. These include the induction of T-cell apoptosis that is mediated by exosomes expressing apoptosis-inducing ligands, such as FasL, PDL1, and TRAIL (21), and a time-dependent inhibition of CD3ζ chain in T cells (22). Cell culture–derived exosomes that bind to, but were not internalized by, T cells regulate expression of several genes that collectively result in a loss of T-cell function (23). Although some or all of these mechanisms may be contributing to the exosome-mediated immunosuppression of T cells, they point toward a relatively slow and irreversible arrest in T-cell functions. The actual mechanism by which the exosomes inhibit the activation of T cells remains poorly understood.

We report here that exosomes incubated with T cells rapidly (within 2 hours) bind to and internalize into the cells. T cells so treated lose the ability to respond to activation via their T-cell receptor. This immunosuppressive effect on T cells occurs in response to exosomes and without loss of T-cell viability. In view of differences with previous reports, we began by characterizing the exosomes with regard to their morphology, size, and composition and evaluating the immunosuppressive ability of exosomes derived from tumor ascites fluids of 12 patients with ovarian cancer and T cells derived from 8 different normal donor PBLs. To validate our findings, we have used multiple different and independent activation endpoints and have established the ability of the exosomes to inhibit the activation of virus-specific CD8⁺ T cells that are stimulated with viral peptides in the context of class I MHC. Finally, we demonstrate here that the exosome-mediated inhibition of T-cell activation is reversible, which makes this system function as a checkpoint that could be a useful immunotherapeutic target in ovarian cancer patients.

The study was designed to assess the effect of exosomes on T-cell function in response to antigen-specific as well as polyclonal stimuli. The source of exosomes was ascites fluids of ovarian cancer patients and lymphocytes isolated from several different healthy donors, who were randomly selected based on availability. Forty-one different ascites fluids and 13 different lymphocyte specimens were used in the study. A total of 300 to 500 μg exosomes (by protein weight) were used in the assays. All experiments reported were reproducibly repeated at least three times. Seven different independent endpoints of T-cell activation were monitored. For the analysis of transcription factor translocation by confocal microscopy, a minimum of 400 cells were counted. For flow cytometry and imaging cytometry experiments, data acquisition was stopped after acquiring 5 × 10⁴ lymphocytes or nucleated cells respectively.

Ascites fluids from stage III or stage IV ovarian cancer patients were received from the Roswell Park Cancer Institute (RPCI) Tissue Procurement Facility. Experiments were done using cell-free ascites fluids that had been stored at −80°C. Normal donor peripheral blood was provided by the Flow and Image Cytometry Facility at RPCI. Normal donor PBLs (NDPBL) were obtained by monocyte depletion and Ficoll–Hypaque density separation. Cells were frozen and stored in liquid nitrogen until use, as reported previously (1, 7). To perform MHC multimer studies (which are haplotype specific), we required peripheral blood from individuals that had been previously haplotyped and tested positive for a specific AST population. All specimens were obtained under sterile conditions and using IRB-approved protocols.

For reagents, see Supplementary Table S1.

Ascites fluids were first centrifuged at 300 × g to separate cells and large debris, followed by another round of centrifugation at 1,150 × g to remove smaller debris and membrane fragments. They were then diluted to 50% (with RPMI1640 or PBS), passed through a 0.22-μm PVDF filter (Millipore), and ultracentrifuged at 200,000 × g for 90 minutes. The pellet was resuspended in RPMI1640 + 1% HSA (for functional experiments) or PBS (for biophysical characterization).

For transmission electron microscopy (TEM) studies, exosomes were isolated and fixed using 2% paraformaldehyde. Ten microliters of exosome suspension was coated on formvar-carbon–coated grids and negatively stained with 2% uranyl oxalate. The grids were air dried for 5 minutes. The specimens were analyzed with a 100CX Transmission Electron Microscope (JEOL USA Inc.).

The size of exosomes was measured using nanoparticle tracking analysis (NTA; NanoSight NS300). The exosomes were diluted appropriately to give counts in the linear range of the instrument (i.e., 3 × 10⁸ to 10⁹/mL). Videos of the particles undergoing Brownian motion in the laser beam were recorded and analyzed using the NTA software, which determines the exosome concentration and size distribution. Three videos of 10-second duration each were recorded for each sample.

The exosome pellet was resuspended in 1 mL of PBS and labeled with 0.6 μmol/L of the membrane probe, diphenyl hexatriene (DPH; Invitrogen). Fluorescence anisotropy experiments were conducted on a PTI Quantamaster fluorescence spectrophotometer (Photon Technology International), fitted with a Peltier unit. The sample was excited at 355 nm and the emission monitored at 430 nm. Fluorescence polarization and anisotropy were calculated as described previously (24). The phase behavior and transition was monitored using fluorescence anisotropy as a function of temperature over a temperature range of 4°C to 50°C.

The identification of protein markers on the isolated exosomes was done using the commercially available Exo-Check Exosome Antibody Array Kit (System Biosciences) as described by the manufacturer. The membrane was developed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) and analyzed using ChemiDoc MP System (Bio-Rad).

The method for detection of NFAT translocation following T-cell activation with MHC dextramers was as described previously (25) with the following modifications specific to studying the effects of ascites fluid–derived exosomes. Whole blood from Epstein–Barr virus (EBV)- or cytomegalovirus (CMV)-positive donors was incubated with peptide-loaded dextramers with or without exosomes for 2 hours at room temperature or for 10 minutes on ice, after which cells were immunophenotyped.

Antibodies were immobilized on maxisorb 12 × 75 mm tubes (Nunc) by incubating 0.1 μg of purified anti-CD3 (Bio X Cell, catalog number BE001-2; clone OKT3) and 5 μg of purified anti-CD28 (Life Technologies, catalog number CD2800-4; clone 10F3) in 500 μL of PBS, at 4°C overnight. PBLs from normal donors were thawed, resuspended in RPMI1640 + 1% human serum albumin, and 5 × 10⁵ total cells were incubated in anti-CD3/anti-CD28 in coated tubes at 37°C/5% CO₂ for the duration of activation.

After activation, cells were attached to alcian blue coverslips in a humid chamber (10 minutes) and fixed in 2% formaldehyde in 1× PBS (40 minutes); the cells were permeabilized and blocked with 30 μg NMIgG in 5% normal mouse serum in 1× PBS + 0.4% Triton X-100. The cells were then stained for intracellular CD3 for 20 minutes. After washing once with NGS block (5% normal goat serum in 1× PBS), the cells were incubated with 2 μg/mL goat anti-mouse IgG-Alexa Fluor 568 for 15 minutes. This was followed by staining with purified rabbit anti-human NFκB p65 or NFAT in NGS block/perm for 1 hour. After washing twice with NGS block, the cells were incubated with 2 μg/mL goat anti-rabbit IgG-Alexa Fluor 488 in 100 μL NGS block/perm for 30 minutes. The cells were washed twice with NGS block and twice with 1× PBS before mounting the coverslips on glass slides with Vectashield Mounting Medium (Vector Laboratories). Cells were then observed on a Zeiss LSM 510 Confocal Microscope with at least 400 CD3⁺ cells counted per condition.

Human NDPBLs were activated for 2 hours at 37°C with immobilized anti-human CD3/CD28 with or without ovarian ascites fluid–derived exosomes. The percentage of activated T cells was determined by monitoring the translocation of NFAT or NFκB from the cytosol into the nucleus using fluorescence microscopy as reported previously (7).

Human NDPBLs were activated for 2 hours at 37°C with immobilized anti-human CD3/CD28 with or without exosomes derived from ovarian ascites fluid. The cells were then incubated for 18 hours in RPMI1640 + 1% HSA at 37°C/5% CO₂ in the absence of stimulation or exosomes. For flow cytometry, the cells were labeled with the recommended amounts of fluorochrome-conjugated antibodies to CD3, CD4, CD8, and CD69 for 30 minutes at 4°C. The cells were then washed with 2 mL of PBS, acquired on an LSR Fortessa (BD Biosciences) flow cytometer and analyzed using FlowJo software (Tree Star Inc.).

Human NDPBLs were activated for 6 hours at 37°C/5% CO₂ with immobilized anti-human CD3/CD28 in the presence of 1 μL/mL GolgiStop (BD Biosciences) and 20 μL/mL fluorochrome-labeled antibody to CD107a with or without exosomes derived from ovarian ascites fluid. For flow cytometry, the cells were labeled with fluorochrome-conjugated antibodies to CD3, CD4, and CD8 for 30 minutes at 4°C, washed, fluorescence emission acquired, and results analyzed as above.

Human NDPBLs were labeled with CellTrace Violet Proliferation Kit (Thermo Fisher Scientific) as recommended by the manufacturer. The labeled cells were incubated in the presence or absence of ascites fluid–derived exosomes in tubes that were coated with immobilized antibodies to human CD3 and CD28 for 7 days. Fresh medium was added after 3 days. On day 7, the cells were labeled with fluorochrome-conjugated anti-human CD3. Sytox Red was added 15 minutes before flow cytometry at a final concentration of 5 nmol/L to label the dead cells. The fluorescence was acquired on an LSR Fortessa (BD Biosciences) flow cytometer. The data were analyzed using FlowJo software (Tree Star Inc.) and ModFit software (Verity Software House) to calculate the proliferation index.

Human NDPBLs were activated for 6 hours at 37°C/5% CO₂ with immobilized anti-human CD3/CD28 in the presence of 1 μL/mL GolgiStop (BD Biosciences) with or without ovarian ascites fluid derived exosomes. For flow cytometry, the cells were labeled with fluorochrome-conjugated antibodies to CD3, CD4, and CD8 for 30 minutes at 4°C. The cells were then fixed and permeabilized with the fixation/permeabilization solution from the Cytofix/Cytoperm Kit (BD Biosciences) as described by the manufacturer and labeled with fluorochrome-conjugated antibodies to IL2 and IFNγ at 4°C for 30 minutes, washed, fluorescence emission acquired, and results analyzed as above.

Human NDPBLs were activated for 2 hours at 37°C with immobilized anti-human CD3/CD28 with or without ovarian ascites fluid–derived exosomes. The cells were then incubated for 18 hours in RPMI1640 + 1% HSA at 37°C/5% CO₂ in the absence of stimulation, but with exosomes present in a 24-well plate. The amount of IFNγ secreted in the supernatant was determined using ELISA as reported previously (26).

Exosomes were labeled with CellTrace Violet using the CTV Proliferation kit (Thermo Fisher Scientific), or with PKH67 using the PKH67 Cell Linker Kit (Sigma-Aldrich) as recommended by the respective manufacturer.

Imaging flow cytometry acquisition and analysis was performed as described previously (27). Data acquisition was performed on an imaging flow cytometer (ImageStreamX Mk-II; Amnis, part of EMD Millipore). The selected laser outputs prevented saturation of pixels in the relevant detection channels as monitored by the corresponding Raw Max Pixel features during acquisition. Cell classifiers were set for the lower limit of size of the bright field image to eliminate debris, the upper limit of size of the brightfield image to eliminate aggregates, and a minimum intensity classifier on the DAPI channel to exclude noncellular (DAPI negative) images.

Following compensation for spectral overlap based on single color controls, image analysis was performed with IDEAS software (Amnis, part of EMD Millipore). The internalization score is a standard feature available in the IDEAS image analysis software. The Internalization feature is defined as the ratio of the intensity inside the cell to the intensity of the entire cell. The so-called masked area (region of interest) to define the inside of the cell was created by eroding the object mask of the brightfield by 3 pixels (Erode (Object (M01, Ch01, Tight), 3), and the internalization features were calculated using this mask for the CD3 and exosome-specific channels (Ch3 and Ch2, respectively). Note that the internalization feature is invariant to cell size and accommodates concentrated bright regions and small dim spots. The ratio is mapped to a log scale to increase the dynamic range. The spatial relationship between the transcription factors and nuclear images was measured using the “similarity” feature in the IDEAS software, as described previously (28, 29). Briefly, a “morphology” mask is created to conform to the shape of the nuclear DAPI image, and a “similarity score” (SS) feature is defined. The SS is a log-transformed Pearson correlation coefficient between the pixel values of two image pairs and provides a measure of the degree of nuclear localization of a factor by measuring the pixel intensity correlation between the NFAT images and the DAPI images within the masked region. Cells with a low SS exhibit poor correlation between the images (corresponding with a predominant cytoplasmic distribution of NFAT or NFκB), whereas cells with a high SS exhibit positive correlation between the images (corresponding with a predominant nuclear distribution of the transcription factor).

All statistics were calculated using Excel 2013 (Microsoft). Paired or unpaired Student t test was applied to determine whether the differences between groups could be considered significant. A P value higher than 0.05 was not significant (NS), whereas *, P < 0.05; **, P < 0.01; and ***, P < 0.001 were considered significant.

Vesicles isolated from ovarian cancer patients' tumor ascites fluid by ultracentrifugation were examined for ultrastructural morphology and size by TEM. Uranyl oxalate–stained vesicles were homogeneously spherical, membrane-bound particles consistent with the morphology of exosomes (Fig. 1A).

Orthogonal biophysical techniques, such as NTA and fluorescence anisotropy, were employed to determine size and lamellarity of the vesicles. NTA analysis of the vesicles revealed a size distribution of 50 to 200 nm with a modal diameter of 60 to 80 nm (Fig. 1B). The lamellarity of these vesicles was analyzed by labeling these vesicles with DPH; lipid order and dynamics were measured at various temperatures using fluorescence anisotropy (Fig. 1C). At lower temperatures, anisotropy values were higher, consistent with a rigid acyl chain packing, but anisotropy values decreased with higher temperatures due to increased acyl chain mobility. The anisotropy values as a function of temperature showed a broad transition centered around 37°C suggesting lamellarity in lipid organization. We conclude that the vesicles present within ovarian tumors are surrounded by a lipid bilayer.

Vesicles isolated by ultracentrifugation from ovarian tumor ascites fluids were assayed for the presence of marker proteins that are typically found on exosomes (30) using a commercially available antibody platform called exosome antibody array. Five of the exosome marker proteins (CD81, Tsg-101, Flotillin-1, EpCAM, and Annexin V) were found to be abundant in the vesicles; two other markers, CD63 and Alix, were detected but less abundant (Fig. 1D). The absence of a positive spot for GM130 indicated that our exosome preparations were not contaminated with cellular material. We and others have previously reported that tumor-associated exosomes also express a negatively charged glycerophospholipid, phosphatidylserine (PS), representing a lipid marker expressed on the surface of exosomes (15, 31).

Based upon the morphology, size, and presence of relevant protein and lipid markers, we conclude that the extracellular vesicles we are isolating from ovarian cancer patients' tumor ascites fluids are exosomes.

Extracellular vesicles derived from cancer patients' sera/plasma or from patients' ovarian tumor ascites fluids have been reported previously to inhibit the activation of T cells (31, 32). However, those studies used a method to active the T cells that depended on antibodies to CD3 and CD28 immobilized on antibody-coated beads (32). Such a protocol represents an artificial stimulus for T cells of unknown specificity. Because exosomes may simply block CD3 and/or CD28 antibody binding to T cells, we asked whether tumor ascites–derived exosomes would similarly inhibit an antigen-induced activation of T cells.

To address this question, we utilized class I MHC multimers (dextramers) loaded with peptides known to bind to antigen receptors on either EBV- or CMV-specific T cells and activate them (25). T-cell activation is determined by a translocation of NFAT from the cytosol into the nucleus and has been confirmed by cells' production of cytokines (25). Peripheral blood from HLA-A2 donors known to have EBV- or CMV-specific T cells was incubated either on ice (nonpermissive for activation) or at room temperature (permissive for activation) with EBV peptide (Fig. 2A and B) or CMV peptide (Fig. 2C and D) loaded HLA-A2 dextramers with or without exosomes. The location of the transcription factor NFAT in CD3⁺ CD8⁺ T cells (either in the cytosol or nucleus) was determined using imaging flow cytometry as reported previously (25). Prior to activation, NFAT is present in the cytosol of the T cells. At the permissive temperature only, the virus-specific CD8⁺ T cells incubated without exosomes, but with the appropriate peptide-loaded dextramer, translocated NFAT from the cytosol into the nucleus. The presence of exosomes resulted in a significant inhibition of the activation of both EBV-specific (Fig. 2A and B) and CMV-specific (Fig. 2C and D) T cells incubated with the appropriate dextramer (82% and 42% inhibition, respectively). No significant activation was observed with cells incubated on ice (Fig. 2A–D). Although prolonged exposure to tumor-derived vesicles eliminates tumor-specific T cells by driving them to apoptosis (33), we demonstrate here that these exosomes can inhibit the activation of antigen-specific T cells with a brief (2 hours) exposure. We conclude that the tumor-derived exosomes induce an arrest in an early activation endpoint (a blockade in the activation of NFAT) of a proportion of the virus-specific T cells that are stimulated by their cognate antigen in the context of MHC.

To better understand the kinetics and durability of the exosome-mediated inhibition of T cells, we studied the effects of exosomes on additional early and later endpoints of T-cell activation. In these studies, PBLs derived from normal donors were incubated for 2 hours in the presence or absence of exosomes, and with a T-cell stimulus of immobilized antibodies specific for CD3 and CD28. Early activation was monitored by detecting the translocation of the transcription factors NFAT and NFκB into the nucleus of CD3⁺ T cells by confocal microscopy. We found that, similar to the inhibition of viral peptide–induced NFAT translocation to the nucleus, the translocation of NFAT in response to polyclonal stimulation was also inhibited by 42% in the presence of exosomes (Fig. 2E). The translocation of another key transcription factor downstream of TCR signaling, NFκB, was also inhibited by 59% (Fig. 2F), consistent with our previous report (31). The percentage of exosome-mediated inhibition in T cells varies with the patient from which the exosomes are derived (Supplementary Fig. S1). The mean inhibition for 41 different ascites fluid–derived exosomes was found to be 41% ± 6.4%.

We next tested the effect of the presence of exosomes during activation on a later activation endpoint, the upregulation of CD69. Following a brief pulse with the polyclonal activation stimulus and exosomes, cells were washed and incubated overnight in culture. Using flow cytometry with gates set on viable CD3⁺CD4⁺ or CD3⁺CD8⁺ T cells, expression of the activation marker CD69 was assessed. After a 2-hour activation without exosomes, followed by overnight culture without further stimulation, expression of CD69 on both CD4⁺ and CD8⁺ T cells was upregulated. However, after a 2-hour activation with exosomes, followed by overnight culture, expression of CD69 was significantly inhibited in both CD4⁺ T cells (Fig. 3A and B) and CD8⁺ T cells (Fig. 3C and D).

These results establish that both CD4⁺ and CD8⁺ T cells require only a 2-hour exposure to exosomes to achieve and observe an inhibition of an activation endpoint (CD69 upregulation) that occurs much later without a persistent presence of the exosomes. As the T cells were gated on viable cells, the exosomal inhibition of activation occurred without a loss of T-cell viability over the period of analysis. This was confirmed by experiments that demonstrated that the viability of T cells activated for 2 hours with or without exosomes was comparable following overnight culture (Supplementary Fig. S2).

A well-defined function of CD8⁺ cytotoxic T cells is the killing of target cells that is dependent upon the release of preformed cytotoxic granules (34). Upon activation of cytotoxic T cells, these cytoplasmic granules move to and fuse with the plasma membrane of the cells and release their lytic enzymes. Surface labeling of T cells with antibodies to CD107a following activation identifies human and mouse degranulating CD8⁺ T cells (34). Six hours after activation, 35% of the CD8⁺ T cells derived from the PBLs of normal donors were positive for the surface expression of CD107a (Fig. 3E and F). CD107a expression was inhibited when the T cells were incubated with tumor-associated exosomes (Fig. 3E and F).

Another function of T cells for both CD4⁺ and CD8⁺ cells is the production and secretion of cytokines following activation. An increase in the percentage of both CD4⁺ and CD8⁺ T cells that express IL2 in the cytoplasm following activation was significantly inhibited when the cells were incubated with exosomes (Supplementary Fig. S3A–S3D). Similarly, exosomes were also found to inhibit the production of IFNγ at the single-cell level in CD4⁺ and CD8⁺ T cells (Supplementary Fig. S4A–S4D), and in bulk cultures of PBLs following activation (Supplementary Fig. S4E).

The results presented above establish that multiple early and late endpoints of activation are inhibited in T cells briefly pulsed with exosomes. We next attempted to determine whether the inhibition of activation could be overcome or reversed by persistent activation. To address this, we monitored another endpoint of activation, the proliferation of T cells cultured with or without exosomes for 7 days in the presence of immobilized antibodies to CD3 and CD28. Cell proliferation was quantified by the generational reduction of fluorescence intensity of T cells labeled with CellTrace Violet (CTV; Fig. 4A). Proliferation modeling was done using the ModFit software, and the proliferation index, which represents the fold expansion during culture, was calculated. As expected, T cells cultured with persistent stimulation but without exosomes proliferated with nearly a 6-fold population expansion (Fig. 4A and B). In contrast, T cells that were persistently stimulated in the presence of exosomes proliferated, but with less than a 3-fold population expansion (Fig. 4B). We conclude that exosomes suppress but do not eliminate the proliferation of T cells in response to persistent stimulation.

The binding of CTV-labeled exosomes to CD3⁺ T cells was quantified by flow cytometry. We found that approximately 25% of the T cells showed an intermediate increase in the MFI (Exo intermediate/Exo^int) of CTV. About 5% of the T cells had a higher MFI (Exo^hi) suggesting that exosome binding to T cells was occurring, but at two different levels of intensity (Fig. 5A). We determined that there was no inhibition of activation in the 70% of the T cells showing no evidence of binding of CTV-labeled exosomes (Exo⁻). T cells showing moderate binding and those with high levels of exosome binding revealed 39% and 60% inhibition of activation, respectively (Fig. 5B and C).

The association between exosome binding and inhibition of activation was further addressed using imaging flow cytometry. T cells were pulsed for 2 hours with exosomes labeled with PKH67, which labels the exosome lipid bilayer. Following the activation of the cells with immobilized antibodies to CD3 and CD28, the CD3⁺ T cells were individually interrogated by imaging flow cytometry simultaneously for their (i) binding and internalization of the PKH67-stained exosomes into the T cells and (ii) activation status as indicated by the localization of Alexa Fluor 647–labeled NFκB either in the cytoplasm (for unactivated cells) or in the nucleus (for activated cells). Figure 6A shows examples of T cells with exosome clusters labeled with PKH67 (cyan) present within the cell cytoplasm, and with NFκB (red) staining also in the cytoplasm of unactivated T cells. In contrast, T cells activated in the absence of exosomes translocate NFκB (red) to the nucleus, marked by DAPI (green) (Fig. 6B). Internalization of exosomes, defined as the ratio of the intensity inside the cell to the intensity of the entire cell, was calculated for the CD3 signal and the exosome signal using the IDEAS software and is demonstrated in Fig. 6C. Higher scores indicate a greater concentration of intensity inside the cell. The data in Fig. 6B establish that for the CD3⁺/exosome⁺ cells, the CD3 internalization score is predominantly negative, whereas that of the exosome signal is positive consistent with the expected membrane localization of CD3 and an internalized exosome localization. The IDEAS software was also used to calculate the SS, which is a measure of nuclear NFκB (Fig. 6D). Unactivated T cells had a similarity score of −0.66, with most cells having NFκB in the cytosol, which increased to +0.21 on activation in the absence of exosomes. However, the SS of cells that had internalized exosomes was only +0.08, consistent with the notion that exosome binding and internalization was coincident with a blockade of activation. Together, these results confirm that a proportion of T cells do bind and internalize exosomes, rendering them unresponsive to activation.

To determine whether the inhibition of T-cell activation by tumor-associated exosomes was reversible, we incubated T cells with immobilized antibodies to CD3 and CD28 in the presence or absence of exosomes. Activation was measured by determining the nuclear translocation of NFκB (Fig. 7A) or the production of intracellular IFNγ (Fig. 7B). As expected, we saw a significant inhibition in both activation endpoints in these T cells in the presence of exosomes. These cells, and control cells that were activated without exosomes, were then rested for either 24 or 48 hours in the absence of stimulation and exosomes and then reactivated. When using NFκB translocation as an endpoint, we observed a complete recovery of activation potential in the T cells previously inhibited by the exosomes (Fig. 7A). We observed a similar recovery in the activation potential when using IFNγ as the activation endpoint, as the T cells that were initially inhibited by exosomes were now found to be reactivated to the same level as control T cells (Fig. 7B). The decrease in the percentage of IFNγ seen in the control T cells upon reactivation is typical when cells are reactivated after a brief recovery period. Because these T cells recovered their activation potential within 24 hours, we conclude that the inhibition of T-cell activation that occurs during a 2-hour pulse with the exosomes is reversible.

Collectively, these results show that tumor-associated exosomes bind to and are internalized rapidly by T cells and that the binding/internalization coincides with the arrest of the activation of the T cells and does not affect viability. This T-cell arrest is transient and can be reversed by removing the immunosuppressive exosomes. This suggests that tumor-associated exosomes represent a potential cancer therapeutic target.

We have previously reported that T cells present in the ascites fluids of patients with ovarian cancer are hyporesponsive to activation via the TCR (7) and that the suppression of these tumor-associated T cells appears to be mediated by small but uncharacterized extracellular vesicles (31). In this report, we have characterized these membrane-encapsulated vesicles by size, morphology, composition, and biophysical properties as exosomes. We have determined that the activation of T cells derived from normal donor PBLs is arrested during a 2-hour incubation of the cells with the tumor-associated exosomes. This inhibition, which is shown here to include multiple different activation endpoints, occurs coincidentally with the binding and internalization of the exosomes, and without loss of T-cell viability. The exosome-induced T-cell arrest is reversible as the exosome inhibited T cells lost their inhibition after incubation for 24 to 48 hours without exosomes. This recovery included two different activation endpoints: (i) the early translocation of NFκB and (ii) the later functional activation indicated by production of IFNγ. Our results establish that tumor-associated exosomes have the ability to arrest T cells during an activation stimulus. Once exosomes are removed, this arrest is reversed over 24 to 48 hours.

However, the ability to reverse the exosome-mediated downregulation of the T cells may well depend upon the duration of the exposure of the cells to the exosomes. Others have reported that the T-cell inhibition induced by tumor-associated extracellular vesicles, including vesicles characterized as exosomes, occurs gradually and appears to be irreversible (15). For example, extracellular vesicles isolated from tumors act over several days and can act indirectly by augmenting the function of T regulatory cells and MDSCs or blocking the maturation of dendritic cells and macrophages (15, 17–20). It has also been proposed that the tumor-associated vesicles act directly over a period of days to permanently suppress T cells by driving them into apoptosis that occurs as a result of the suppression of the CD3 ζ chain, or through the expression of apoptosis-inducing ligands on exosomes, including FasL, PDL, and TRAIL (21). These results suggest that a prolonged exposure of the T cells to the exosomes (1–4 days) may drive these cells into irreversible suppression. A similar gradual and progressive loss of T-cell function is observed with antigen-driven exhaustion of T cells with an accumulation of multiple checkpoint molecules, such as PD-1, CTLA4, LAG-3, TIM3, etc., leading to a deterioration of T-cell functions that ultimately become irreversible (35). However, our results establish that a brief (2 hours) exposure of the T cells to the exosomes during the activation of the T cells results in a rapid but reversible arrest in their response to activation. Recognition of differences in the dynamic and kinetic effects of the exosomes on T cells may help determine the mechanisms by which exosomes suppress T-cell function, and for the eventual design of therapeutic strategies to enhance the antitumor effects of T cells by reversing the immunosuppressive effects of the exosomes in tumor microenvironments.

We propose that exosome-induced T-cell arrest begins with the binding of the exosomes to a receptor on T cells that induces an immunosuppressive signal. A causal link of PS to the exosomal suppression of T cells was established by blocking this suppression with anti-PS antibodies and Annexin V as well as by a selective depletion of PS⁺ vesicles (31). A testable hypothesis relevant to the T-cell suppression mechanism is that PS⁺ exosomes bind to a PS receptor such as TIM3. Furthermore, as empty liposomes expressing PS on their surface are able to mimic the same T-cell arrest induced by exosomes, it is possible that PS by itself has the capacity to modulate T-cell function directly, and that PS on exosomes is capable of inducing a direct signaling arrest independent of an immunosuppressive exosome cargo (31).

PS enhances the metabolic activity of diacylglycerol kinase (DGK; ref. 36), a negative regulator of diacylglycerol (DAG), which is part of the TCR signaling cascade. Because PMA, a DAG analogue, reverses the T-cell inhibitory effects of tumor-associated vesicles (31), it is plausible that PS acts to inhibit the TCR signaling cascade by a DGK phosphorylation of DAG, converting it into inactive phosphatidic acid. This mechanism is supported by the finding that inhibitors of DGK block the inhibitory activity of exosomes derived from the ascites (31), and the regulation of DAG by DGK is critical in the induction of T-cell anergy (37–40).

The presence of immunosuppressive exosomes in the tumor microenvironment likely contributes to local blockade of an antitumor response in patients. Treatment strategies that block or reverse the effects of exosomes may be useful alone or in combination with other immunotherapeutic approaches. We have established here that the immunosuppression of T cells following a brief (2 hours) exposure to exosomes during their activation is reversible.

No potential conflicts of interest were disclosed.

Conception and design: G.N. Shenoy, R.J. Kelleher Jr, H. Minderman, R.B. Bankert

Development of methodology: G.N. Shenoy, J. Loyall, O. Maguire, V. Iyer, R.J. Kelleher Jr, H. Minderman, P.K. Wallace, R.B. Bankert

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.N. Shenoy, J. Loyall, O. Maguire, V. Iyer, R.J. Kelleher Jr, H. Minderman, K. Odunsi, S.V. Balu-Iyer, R.B. Bankert

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G.N. Shenoy, J. Loyall, O. Maguire, R.J. Kelleher Jr, H. Minderman, P.K. Wallace, S.V. Balu-Iyer, R.B. Bankert

Writing, review, and/or revision of the manuscript: G.N. Shenoy, O. Maguire, R.J. Kelleher Jr, H. Minderman, R.B. Bankert

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G.N. Shenoy, J. Loyall, R.J. Kelleher Jr, R.B. Bankert

Study supervision: R.J. Kelleher Jr, R.B. Bankert

Research reported in this article was supported by the NCI of the NIH under award numbers R01CA108970 and R01CA131407 (to R.B. Bankert), the National Heart, Lung, and Blood Institute of the NIH under award number R01HL70227 (to S. Balu-Iyer), the NIH under award numbers P50CA159981 and R01CA158318 (to K. Odunsi), and the NIH under award numbers 1S10OD018048 and 1R50CA211108 (to H. Minderman). The Flow and Image Cytometry Core facility at the RPCI is supported in part by the NCI Cancer Center Support Grant 5P30 CA016056.

The authors thank Anthony Miliotto and the Tissue Procurement Facility of RPCI for their assistance in providing tumor tissues and ascites fluid. Flow cytometry and confocal microscopy services were provided by the Confocal Microscopy and Flow Cytometry Core Facility at the University at Buffalo. Additional cytometry services were provided by the Flow and Image Cytometry Core facility at the RPCI. Electron microscopy services were provided by Dr. Thaddeus Szczesny at the Electron Microscopy Core Facility at the University at Buffalo.

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.