Phosphatidylserine-positive extracellular vesicles boost effector CD8⁺ T cell responses during viral infection

Significance

Many in vitro studies have shown a role of extracellular vesicles (EVs) in priming of naive T cells. However, if EVs really play a relevant role in T cell responses in vivo remained highly controversial. We have recently developed a method, allowing us to identify and characterize cells binding naturally occurring EVs in vivo. With this method, we demonstrate here that activated effector cells, but not naive T cells, extensively engage with EVs during viral infections. We provide convincing evidence from in vivo studies that this interaction increases the expression of effector genes and proliferation of CD8⁺T cells in (lymphocytic choriomeningitis virus) LCMV-infected mice. This, in turn, boosts T cell effector functions.

Abstract

CD8⁺ T cells are crucial for the clearance of viral infections. During the acute phase, proinflammatory conditions increase the amount of circulating phosphatidylserine⁺ (PS) extracellular vesicles (EVs). These EVs interact especially with CD8⁺ T cells; however, it remains unclear whether they can actively modulate CD8⁺ T cell responses. In this study, we have developed a method to analyze cell-bound PS⁺ EVs and their target cells in vivo. We show that EV⁺ cell abundance increases during viral infection and that EVs preferentially bind to activated, but not naive, CD8⁺ T cells. Superresolution imaging revealed that PS⁺ EVs attach to clusters of CD8 molecules on the T cell surface. Furthermore, EV-binding induces antigen (Ag)-specific TCR signaling and increased nuclear translocation of the transcription factor Nuclear factor of activated T-cells (NFATc1) in vivo. EV-decorated but not EV-free CD8⁺ T cells are enriched for gene signatures associated with T-cell receptor signaling, early effector differentiation, and proliferation. Our data thus demonstrate that PS⁺ EVs provide Ag-specific adjuvant effects to activated CD8⁺ T cells in vivo.

T cells confer protection against pathogens and tumors. The initiation and maintenance of T cell responses require specific signals delivered by cell-to-cell interactions and secreted soluble cytokines. In addition, extracellular vesicles (EVs) constitute a mechanism of intercellular communication during immune responses (1). However, the precise mechanisms of how EVs can affect T cells in vivo remain unclear (2, 3).

EVs are lipid-bilayer enclosed, spherical structures released by virtually all cell types. They carry various bioactive molecules, including proteins, lipids, and nucleic acids, able to exert functional modifications and phenotypic changes upon interaction with recipient cells (4). Due to their high heterogeneity in origin, size, and composition, several subtypes of EVs have been described, including exosomes and microvesicles (5).

Early studies first demonstrated that exosomes from B cell lines bear peptide/major histocompatibility class (MHC) II complexes and could directly activate CD4⁺ T cells in vitro (6). The finding that EVs derived from tumor peptide-pulsed dendritic cells (DCs) can elicit strong CD8⁺ T cell responses and tumor suppression in vivo further supported the idea that EVs were involved in antigen (Ag) presentation (7). Numerous follow-up studies showed that Ag-presenting cell (APC)-derived EVs, particularly those released by mature DCs, can serve as sources of Ag and induce T cell proliferation, memory development, and antitumor responses in vitro and in vivo (1, 7–14). DC-derived EVs carry the relevant MHC-I and -II complexes and costimulatory molecules such as CD86 and CD54 for productive interaction with T cells. However, the precise mechanisms of T cell stimulation are still unknown.

While some reports showed that EVs alone can activate T cells in vitro (6, 13–15), other studies suggested that an indirect mode of action as DCs was required for the stimulation of T cells by EVs (1, 9, 12, 16–22). Therefore, the capacity of free EVs to regulate T cell function and differentiation as cell-independent biological modifiers in vivo remained unclear.

Like apoptotic cells, also EVs display phosphatidylserine (PS) on their outer membrane layer (23–26). Based on this marker, we previously described a sensitive and robust method to analyze EVs in lymphocytic choriomeningitis virus (LCMV)-infected mice (27) and SARS-CoV-2-infected humans (28). During these acute infections, high frequencies of activated CD8⁺ T cells were associated with PS⁺ EVs. Here, we demonstrate that the binding of PS⁺ EVs to Ag-specific CD8⁺ T cells triggers TCR-signaling in vivo, as demonstrated by nuclear accumulation of NFATc1. Specifically, EV-associated CD8⁺ T cells showed enhanced proliferation, effector gene expression signatures, including genes like Ifng and Tnf. Therefore, our findings establish that EVs can boost effector gene expression by directly associating with activated Ag-specific CD8⁺ T cells during acute infection.

Results

In Vivo Detection of Naturally Occurring EVs Associated with CD8⁺ T cells.

Previously, we found that APC-derived PS⁺ exosomes bind to activated CD8⁺ T cells during LCMV infection (27). However, neither the kinetics of EV appearance nor the T cell subset interacting with EVs is known. Furthermore, it is unclear whether T cell/EV interaction would affect the functional differentiation of T cells.

To address these questions, we used intravenously injected PS-binding Milk fat globule-EGF factor 8 protein (MFG-E8)-eGFP (27) or biotinylated murine MFG-E8 C1 domains tetramerized using Streptavidin (C1 tetramers) (29). Both reagents detect naturally occurring PS⁺ EVs in spleens of mice acutely infected with LCMV Armstrong (LCMV_Arm) in vivo at different time points after infection (Fig. 1A). We examined single-cell suspensions of live cells (SI Appendix, Fig. S1A) by imaging flow cytometry (IFC) using an ImageStream cytometer (Fig. 1B). PS⁺ dying cells (live/dead dye⁻, with intact cell membrane) and PS⁺EV⁺ living cells were digitally sorted using a convolutional autoencoder (CAE) module (Fig. 1 B, Right) (27). The percentage of PS⁺ cells increased sharply in virus-infected mice compared to the noninfected control group (Fig. 1B). The frequency of EV⁺ live cells and EV⁺CD8⁺ T cells peaked on day five post infection (p.i.) and declined from day 10 to day 15 p.i. (Fig. 1 B and C). Similarly, apoptosis of total live and CD8⁺ T cells reached its maximum on day five (Fig. 1C). While the frequency of bona fide apoptotic PS⁺CD8⁺ T cells remained low (1 to 2% of CD8⁺ T cells) even at the peak response, up to 35% of CD8 T cells were associated with PS⁺ EVs at day five p.i. (Fig. 1C). Notably, on day 15, p.i. EV-decoration and apoptosis dropped below the levels of noninfected mice (Fig. 1C). These data indicate that EV-decoration of CD8⁺ T cells is transient and correlate with the published kinetics of LCMV_Arm titers (30, 31). Furthermore, we detected the peak of EV decoration at the maximum expansion of LCMV-specific CD8⁺ T cells in the spleen, which occurs between day five and day eight after infection (32).

Fig. 1.

MFG-E8-eGFP detects PS⁺ apoptotic and PS⁺ EV-decorated cells in vivo. (A) Flow chart shows experimental setup for analysis of PS⁺ cells in mice. Splenocytes were analyzed by IFC using an ImageStream cytometer. Dot plots display the gating strategy of live/dead⁻PS⁺ cells from the spleens of noninfected and LCMV_Arm-infected mice. Area and aspect ratio of the bright field (BF) channel were used to identify single cells (SI Appendix, Fig. S1). Numbers in the gate indicate the mean percentages ±SD of PS⁺ cells from MFG-E8-eGFP-injected mice (n = 3) and control (PBS)-injected mice (n = 1). Only live cells were analyzed (SI Appendix, Fig. S1A) with the convolutional autoencoder (CAE) module as described previously (Kranich et al. 27). Representative images of bright field (BF), PS, and BF/PS overlay channels of PS⁻ (SI Appendix, Fig. S1B) and PS⁺ cells are shown. (Scale bar: 7 µm.) Shown are representative results from two independent experiments. (B) Mice were infected with LCMV_Arm (n = 3 per timepoint), and frequencies of EV⁺ (red) and apoptotic (blue) total live cells and CD8⁺ T cells were determined on days 5, 10, and day 15 p.i. using the CAE. Gating see SI Appendix, Fig. S1C. Representative results from two independent experiments are shown. (C) Representative dot plots of different T cell subsets binding naturally occurring PS⁺EVs after i.v. administration of MFG-E8-eGFP are shown. Gates show total live CD8⁺ (Left, black) and live/CD8⁺/PS⁺EV⁺ (Right, red) CD62L⁺CD44^– naive (T_N), CD62L⁺CD44⁺ central memory (T_CM), and CD62L⁻CD44⁺ effector (T_E) T cells in spleens of LCMV-infected (Upper) and noninfected (Lower) mice. Values next to the gate indicate the frequencies ±SD of cells within the respective gate. Bar graphs visualize frequencies ±SD of T_N, T_CM, and T_E CD8⁺ T cells that have bound PS⁺EVs. Bar graphs show the results from nine LCMV-infected and nine noninfected mice pooled from three independent experiments. (D) Analysis of serum EVs isolated from noninfected and LCMV_Arm-infected mice (day five p.i.). Upper panel shows nanoparticle tracker (NTA) analysis of EV fractions 2 to 7 isolated by size exclusion chromatography, n = 4, representative results from three independent experiments are shown. Bar graphs show particles/mL in noninfected and infected sera, n = 11, pooled results from three independent experiments are shown. Lower panel shows ImageStream analysis of serum EVs. Gating strategy, unstained, dye only, and detergent controls see SI Appendix, Fig. S1D. Dot plots show PS⁺ EVs present in sera of infected and noninfected mice. Mean number (±SD) of PS⁺ particles are indicated next to the gate. Bar graph shows percentage of PS⁺ EVs (n = 7 and 8 for noninfected and infected, respectively). Pooled results of two independent experiments are shown. Unpaired Student’s t test was used to determine statistical significance, with *P < 0.05, **P < 0.01, and ***P < 0.001.

Next, we determined if EVs preferentially associate with a specific T cell subset in LCMV-infected mice. On day five, CD8⁺ effector T cells (CD62L⁻CD44⁺, T_E) showed the highest levels of EV-decoration (Fig. 1D). While less than 10% of central memory (CD62L⁺CD44⁺, T_CM) CD8⁺ T cells were associated with PS⁺ vesicles, EVs were absent on naive (CD62L⁺CD44⁻, T_N) CD8⁺ T cells. These data reveal that EVs bind differently to distinct CD8⁺ T cell subsets, with a preference for activated T_E cells during acute LCMV infection.

The observed enhanced binding of EVs to CD8⁺ T cells could be caused by increased concentration of EVs in infected animals or altered binding properties of activated T cells, or both. To distinguish between these two factors, we next sought to determine whether the concentration of EVs increases in infected animals. Marker-independent particle analysis indicated that the size of EV particles did not change after infection in serum (Fig. 1D). Surprisingly, the serum concentration of particles decreased significantly after virus infection (Fig. 1D, NTA). To rule out the possibility that by focusing on PS⁺ EVs we would miss a significant proportion of PS^- EVs, we analyzed the frequencies of PS⁺ EVs in both situations (Fig. 1D, image stream). Yet, in infected and noninfected animals, the largest proportion (over 95%) of all EVs was PS⁺.

These data suggest that in the case of infection, EV binding to lymphocytes may lead to a statistically significant decrease in the concentration of free EVs in serum. Furthermore, PS is an excellent marker for EVs as most EVs are PS⁺.

EV-Associated CD8⁺ T Cells Show Increased Levels of Nuclear NFATc1.

With the above-described approach, we are able to stain naturally occurring PS⁺ EVs in vivo. We showed in a previous study that EVs associated with activated CD8⁺ T cells during acute LCMV infection in vivo carry exosome-markers CD9/CD63 as well as MHC-II, CD86, and CD54, demonstrating their APC origin (27). Several studies have also shown that APC-derived EVs carry MHC-I (1). The presence of these molecules allows EVs potentially to stimulate T cells. T cell receptor (TCR)-triggering induces a calcium-dependent, rapid translocation of NFAT from the cytoplasm to the nucleus, a process which is essential for CD8⁺ T cell cytotoxicity during viral infection (33). Therefore, nuclear translocation of NFAT is a sensitive read-out for TCR signaling. The NFAT-family member NFATc1 is an important regulator of CD8⁺ T_E cell activation and cytotoxicity (34). To investigate whether CD8⁺ T cells having bound naturally occurring PS⁺ EVs display different TCR signaling compared to EV^- CD8⁺ T cells, we analyzed the nuclear translocation of NFATc1 in EV⁺ and EV⁻ CD8⁺ T cells during LCMV infection in vivo (Fig. 2A). We found significantly higher levels of nuclear NFATc1 as demonstrated by a higher median similarity score (ss) and an increased frequency of nuclear NFATc1 in EV⁺ CD8⁺ T cells compared to EV⁻ CD8⁺ T cells (Fig. 2A). As NFAT is rapidly exported from the nucleus (after ~15 min), when TCR-mediated signaling stops (35), these results support the idea that EV–T-cell interactions could cause NFAT translocation.

Fig. 2.

Increased nuclear translocation of NFATc1 in EV⁺CD44⁺CD8⁺ T cells. Mice were infected with LCMV_Arm (n = 6, pooled from two independent experiments), on day five p.i., naturally occurring PS⁺ EVs were stained in vivo by injecting MFG-E8-eGFP (100 µg/mouse). Isolated splenocytes were stained for surface markers, NFATc1 and DRAQ5 (DNA dye) and analyzed by IFC. (A) Nuclear translocation of NFATc1 in Draq5⁺NFATc1⁺PS⁺ and PS⁻CD44⁺CD8⁺ T cells (gating strategy SI Appendix, Fig. S2) was determined using the similarity feature of the IDEAS software. Cells with a similarity score (SS)>1 have nuclear NFATc1, cells with a SS<1 have cytoplasmic NFATc1. Representative histograms show median SS of PS⁺ and PS⁻CD44⁺CD8⁺ T cells and frequencies of cells with cytoplasmic and nuclear NFATc1. Representative IFC images display cells with cytoplasmic NFATc1 (Left) and nuclear NFATc1 (Right). (Scale bar 7 m.) Bar graphs (Left) show the median SS ±SD (Left) and the frequencies ±SD (Right) of PS⁺ and PS⁻CD44⁺CD8⁺ T cells. Representative results from four independent experiments are shown. For statistical analysis, unpaired Student’s t test was used with *P < 0.05, **P < 0.01, and ***P < 0.001. (B) Experimental setup for analyzing NFATc1 translocation in transferred CD45.1^-CD90.1⁺ P14 and CD45.1⁺CD90.1⁻ OT-I CD44⁺CD8⁺ T cells. (C) Splenic P14 and OT-I CD8⁺ T cells were stimulated in vitro and transferred into day two LCMV-infected mice. After 3 d, MFG-E8-eGFP was injected, and NFATc1 translocation was analyzed. Dot plots show gating of PS⁻ and PS⁺ CD90.1⁺CD45.1⁻ P14 and PS⁻ and PS⁺ CD45.1⁺CD90.1⁻ OT-I T cells with average percentages ±SD (n = 6). Bar graph displays average frequencies ±SD of PS⁺ P14, OT-I cells. (D) Representative histograms show SS of PS⁻ and PS⁺ CD44⁺ P14 and CD44⁺ OT-I T cells with frequencies of cells with a SS>1 and <1. Bar graphs show frequencies of cells with nuclear NFAT and median SS ±SD. Results from six mice were pooled from two independent experiments. Dotted horizontal lines represent the SS of in vitro-activated P14 and OT-I before transfer into mice (n = 1). (E) Experimental setup for (F): infected mice were injected with 50 µg C1-SA-AF647 to stain PS⁺ EVs in vivo. One hour later, spleens were removed and stained with CD8-CF488 and CD9-CF568. After fixation, PS⁺ and PS⁻ CD8⁺ T cells were sorted and analyzed by dSTORM superresolution microscopy. (F) Upper and Lower panels show representative superresolution images of PS⁺, and PS⁻ CD8 T cells, respectively (PS green, CD9 red, CD8 white). Areas labeled with I, II, III, and IV are shown with higher magnifications in the micrographs labeled accordingly. Total CD8⁺ clusters and CD8⁺CD9⁺PS⁺ triple-positive clusters were quantified using the cluster analysis tool of the CODI software (Oxford Nanoimaging). Upper bar graph shows the number of total CD8⁺ and triple-positive clusters per sorted PS⁻ (n = 17) or PS⁺ (n = 23) cell. Each datapoint represents one cell. Lower bar graph shows the percentage of triple-positive clusters of all CD8⁺ clusters per cell. Pie charts indicate the number of cells with triple-positive clusters. For statistical significance, unpaired Student’s t test was used with *P < 0.05, **P < 0.01, and ***P < 0.001.

To investigate whether NFAT translocation in EV⁺ CD8⁺ T cells is driven by TCR stimulation and hence Ag-dependent, we adoptively transferred in vitro activated TCR-transgenic LCMV-specific, P14 CD8⁺ T effector cells [specific for the LCMVgp_33-41 peptide/MHC-I D^b, (36)] together with activated LCMV-irrelevant OT-I CD8⁺ T effector cells [specific for ovalbumin (OVA)_257-264 peptide/MHC-I K^b, (37)] into mice that had been infected with LCMV 2 d before (Fig. 2B). In this setting, OT-I cells do not receive Ag-specific TCR stimulation due to the absence of their cognate Ag from the system.

As expected, 3 d after transfer, due to the presence of their cognate LCMV-Ag P14 T cells had accumulated to higher frequencies as compared to “bystander” OT-I T cells (Fig. 2 C, Left). Additionally, P14 cells also showed a significantly higher degree of EV association as compared to nonspecific OT-I cells (Fig. 2 C, Right). This finding suggests that Ag specificity of the TCR may significantly contribute to the binding of EVs to T cells. Most importantly, we found a significantly higher ss and higher levels of intranuclear NFATc1 in EV⁺CD8⁺ P14 T cells compared to their EV⁻ CD8⁺ P14 T cell counterparts within the same animals (Fig. 2D). Notably, the amount of nuclear NFATc1 of EV⁺ P14 cells exceeded that of P14 cells stimulated in vitro with antibodies to CD3ε and CD28 (Fig. 2D, dotted line). The frequency of nonspecific OT-I cells with nuclear NFATc1 was about half of that of P14 T cells (Fig. 2D).

Nevertheless, at rather low levels, EV⁺ OT-I T cells had slightly but significantly more intranuclear NFATc1 than EV⁻ OT-I T cells (Fig. 2D). Together, these results further support the idea that EVs may promote the continuous Ag-specific TCR stimulation of activated CD8⁺ T cells during acute LCMV infection.

To confirm EV–TCR interaction, we performed direct stochastic optical reconstruction microscopy (dSTORM) on fluorescence-activated cell sorting (FACS)-sorted PS⁺ and PS⁻ CD8⁺ T cells from LCMV-infected mice. As we were limited to three dSTORM compatible fluorochromes, we used PS [stained in vivo with C1 tetramers (C1-SA-AF647)] and CD9-CF568 to identify EVs and CD8-CF488 to identify the TCR complex (Fig. 2E), as CD8 colocalizes with TCRab in activated T cells (38). Indeed, on PS⁺ CD8 T cells, we could readily identify PS⁺ CD9⁺ double-positive clusters of EV size (approximately 200 nm, Fig. 2 F, Upper), which were nearly absent on PS⁻ CD8 T cells (Fig. 2 F, Lower). These PS⁺CD9⁺ EVs colocalized with CD8 (Fig. 2F, I.I, I.II, II.I) and therefore with the TCR complex. Quantification of total CD8⁺ clusters revealed a significant increase of CD8⁺ clusters on PS⁺ CD8 T cells (Fig. 2 F, Upper bar graph). While more than 75% of the sorted PS⁺CD8⁺ T cells had CD8⁺CD9⁺PS⁺ triple-positive clusters (Fig. 2 F, Lower, pie chart), approximately 30% of all observed CD8⁺ clusters of each cell were CD8⁺CD9⁺PS⁺ triple positive and therefore colocalized with CD9 and PS. (Fig. 2 F, Lower, bar graph). The few PS⁺ triple-positive clusters on PS⁻ cells were of low intensity and likely due to sort impurities. These data show that CD8⁺ T cells associated with PS⁺ EVs had i) more aggregated CD8⁺ clusters as compared to PS⁻ CD8⁺ T cells and ii) that PS⁺ EVs locate to the CD8/TCR-complex on the T cell surface, making EV-mediated TCR-signaling a likely event.

Transcriptional Profiling Reveals Increased Proliferation, Effector Function, and Reduced Memory Potential of EV⁺ CD8⁺ T Cells.

For an in-depth comparison of EV⁻ and EV⁺ CD8 T_E cells, we performed RNAseq analysis of these cells sorted from LCMV-infected mice at days five and 10 after infection. We were specifically interested in whether EV-binding to CD8⁺ T cells influences specific CD8⁺ memory or effector differentiation programs. For this, we performed gene expression analyses (RNAseq, data available at GEO accession number GSE201507) and subsequent gene set enrichment analyses (GSEA) against 10 well-described gene clusters (39). These clusters define gene-expression signatures of naive, short-term, early, and late effector cells, as well as memory precursor, and memory CD8⁺ T cells. On day five, three gene sets (cluster I—initial cytokine or effector response, cluster II—preparation for cell division, and cluster III—cell cycle and cell division; Fig. 3A) were significantly enriched in EV⁺ CD8⁺ T cells (Fig. 3 A and B). All differentially expressed genes (DEGs) on day five (Fig. 3 B and C and SI Appendix, Fig. S4) and day 10 (SI Appendix, Fig. S5) with a false discovery rate (FDR) of <0.05 are highlighted in fold change plots. Expressions of the main CD8⁺ T_E cell cytokine Ifng and the orphan nuclear receptor Nra4a1 (encoding for NUR77), which is induced by TCR signals in a dose-dependent manner (40) were significantly up-regulated in EV⁺ CD8⁺ T cells (Fig. 3C and SI Appendix, Fig. S4A).

Fig. 3.

EV⁺ CD8⁺ T cells have a gene signature of proliferating effector cells. PS^- and PS⁺ CD44⁺CD62L⁻ CD8+ T cells were sorted from LCMV_Arm-infected mice on day five after infection (sorting strategy see SI Appendix, Fig. S3). Transcriptomic analysis was performed by RNAseq on sorted cells. (A) Gene set enrichment analysis (GSEA) was performed against 10 gene clusters characterizing different CD8⁺ effector and memory subsets during infection and memory formation (39). Bar graph shows summary of GSEA results. x axis shows normalized enrichment score, color code indicates adjusted P values. (B) Heat maps show differentially expressed genes (DEGs, padj < 0.05) between PS⁻ and PS⁺ CD8+ T cells from clusters I, II, III, and VII using Pearson correlation as distance measure. Color code indicates z score. (C) GSEA plots from clusters I, II, III, and VII. (D) Cells carrying PS⁺ EVs from LCMV_Arm-infected mice were stained in vivo by injecting MFG-E8-eGFP on day five p.i. (D) G₀/G₁, S, and G₂/M cell cycle stages of PS⁺ and PS⁻ CD44⁺CD8⁺ T cells were analyzed using the DNA dye DRAQ5. Representative histograms of PS⁺ (upper histogram) and PS⁻ (lower histogram) CD44⁺CD8⁺ are shown. Bar graphs show frequencies of cells in G₂/M stage of PS⁺ (red) and PS⁻ (gray) CD44⁺CD8⁺ T cells from nine mice pooled from three independent experiments. Samples from the same experiment are shown with the same symbol. Student’s t test was used to determine statistical significance, with **P > 0.01, ****P < 0.0001.

Also, EV⁺ CD8⁺ T cells showed significantly elevated Irf4 and Irf8 transcription factors, which are both crucial for CD8⁺ T cells to develop effector functions (41, 42) (Fig. 3C and SI Appendix, Fig. S4A). Hence, we conclude that EV⁺ CD8⁺ T cells receive additional TCR signals for proliferation and effector cytokine production than their EV⁻ counterparts. In addition, EV⁺ CD8⁺ T cells showed five significantly down-regulated gene sets (cluster IV—naive and late memory, cluster VI—short-term effector and memory, cluster VII—memory precursor, cluster VIII—naive or late effector or memory, and cluster X—late effector or memory) and therefore may have a reduced potential to differentiate into memory T cells (Fig. 3 A–C and SI Appendix, Fig. S4B)—a process that is normally initiated around days four to six p.i. (43, 44). Specifically, genes such as the transcriptional repressors Id2 (cluster VI, SI Appendix, Fig. S4B) and Id3 (cluster II, Fig. 3C and SI Appendix, Fig. S4B), which regulate memory differentiation (45), the transcription factor TCF1 (cluster VII, encoded by Tcf7) which is essential for the formation of central memory CD8⁺ T cells (46–48), and Il7r (cluster VII), which is highly expressed on memory cells (49), were significantly down-regulated in EV⁺ CD8⁺ T cells (Fig. 3C and SI Appendix, Fig. S4B).

We then assessed differences between EV⁺ and EV^- CD8⁺ T cells on day 10 of the LCMV infection and performed a GSEA analysis against the same gene sets (Fig. 3A and SI Appendix, Fig. S5). Also, on day 10, clusters I, II, and III showed clear enrichment in EV⁺ CD8⁺ T cells, indicating that these cells are still proliferating and have effector phenotypes with Irf4, Irf8, and Nra4a1, but not Ifng being significantly up-regulated (Fig. 3A and SI Appendix, Fig. S5A). However, at this later point and in contrast to day five, cluster IV (naive and late memory) was also enriched in EV⁺CD8⁺ T cells (Fig. 3A and SI Appendix, Fig. S5A). Clusters VI, VII, VIII, and X were, as seen on day 5, down-regulated in day 10 EV⁺CD8⁺ T cells (Fig. 3A and SI Appendix, Fig. S5B). In contrast to Id2 and Id3, essential memory signature genes, such as Il7r and Tcf7, were not significantly reduced on day 10 (SI Appendix, Fig. S5 A and B). Overall, EV⁺CD8⁺ T cells on day five and day 10 showed a very similar gene signature indicative of proliferating effector cells. However, one must consider that a much smaller number of EV⁺CD8⁺ T cells were present on day 10 than on day five (Fig. 1B). Therefore, the potential influence of EVs during the immune response probably fades after the early peak of T cell expansion and would therefore primarily influence critical amounts of CD8⁺ T_E cells early during adaptive immune responses.

As gene clusters regulating cell cycle and cell division were enriched in PS⁺ cells (Fig. 3 A–C), we set out to confirm these results by assessing cell cycle stages of PS⁺ and PS^- CD44⁺CD8⁺ cells in LCMV_Arm-infected mice on day five using nuclear staining with DRAQ5. Cells with the highest DNA content represent cells in G₂/M stage. In line with the transcriptome analysis, T cells carrying in vivo stained PS⁺ EVs showed significantly higher DRAQ5 levels than EV-free T cells. While approximately 19 ± 6% of PS⁺ CD44⁺CD8⁺ T cells were in G₂/M stage, only approximately 5 ± 2% of their PS⁻ counterparts were in G₂/M stage (Fig. 3E), confirming that EV⁺ T cells show more proliferation.

BMDC-Derived EVs Stimulate Activated T Cells Independently of DCs in an Ag-Dependent Manner.

The above results demonstrate that T cells that bind Ag-specific EVs have increased TCR signaling as demonstrated by more NFAT translocation to the nucleus. Furthermore, EV⁺ T cells show a more robust effector phenotype than EV⁻ T cells. However, it still remains to be clarified if EVs cause these changes by directly triggering the TCR and/or costimulatory receptors on T cells or whether a specific T_E subset, e.g., one that has just recently been activated by APCs, is particularly good at binding EVs. Although many studies in the past have aimed to demonstrate T cell stimulation or even priming of naive T cells by EVs (14, 50), clear and convincing evidence for such a scenario is still lacking (2, 3). Considering the negligible binding of EVs to naive T cells found during viral infection (Fig. 1D), we decided to concentrate on activated T cells and wanted to convincingly determine if these are directly modulated by EV binding in vivo. To do so, we set up the following experiment: We adoptively transferred in vitro activated TCR-transgenic OVA-specific OT-I and LCMV-specific P14 CD8⁺ T cells into H2-K^bm1 mice (Fig. 4A). These recipient mice cannot present the cognate OT-I peptide OVA_257-264 on MHC-I (51), ruling out participation of endogenous APCs in specific OT-I TCR triggering. After the adoptive transfer, we injected PKH26-labeled EVs derived from OVA-pulsed BMDCs (Fig. 4A). We measured DC activation markers by flow cytometry and confirmed successful activation of EV-producing BMDCs (SI Appendix, Fig. S7A). We also determined the presence of molecules needed for T cell stimulation (MHC-I, CD86, and CD54) on EVs using IFC (SI Appendix, Fig. S7 B and C). The cotransferred P14 cells do not recognize the OT-I peptide and thus served as an internal control. To test whether T cells received a stimulatory signal from the EVs, we analyzed PKH26⁺ EV⁺ and EV^- OT-1 and P14 cells 1 h after the EV injection for nuclear translocation of NFATc1 (Fig. 4A). To exclude that PKH26 aggregates forming during the preparation bind to activated CD8⁺ T cells and render them falsely PKH26⁺, we followed the EV staining protocol using PKH26 without adding EVs and injected the same fractions as PKH26 only controls into LCMV-infected mice. No aggregate binding occurred by CD8⁺ T cells (SI Appendix, Fig. S7D).

Fig. 4.

Ag-dependent stimulation of CD8⁺ T cells by BMDC-derived EVs. (A) Illustration of experimental setup. At day six of the culture, BMDCs were stimulated with LPS and pulsed with SIINFEKL peptide. PKH26-labeled EVs from BMDCs and injected into H2k^bm1 mice that had received in vitro activated OT-I and P14 cells 3 d earlier. One hour later, isolated splenocytes were stained for surface markers, NFATc1 and DRAQ5, and analyzed by IFC. (B) Dot plots show gating of transferred CD90.1⁺CD45.1⁻ P14 and CD90.1⁺CD45.1⁺ OT-I CD8⁺ T cells with average percentages ±SD (n = 3). The bar graph displays average frequencies ±SD of P14 and OT-I CD8⁺ T cells in H2k^bm1 mice injected with BMDC-derived EVs (+) or noninjected (−). (C) Dot plots show gating of EV⁺P14 and EV⁺OT-I CD8⁺ T cells with average percentages ±SD (n = 3). Bar graph displays average frequencies ±SD of EV⁺P14 and EV⁺OT-I CD8⁺ T cells in H2k^bm1 mice injected with BMDC-derived EVs. (D) Representative images depict CD90.1/PKH, CD45.1/PKH, and NFATc1/DRAQ5 overlays of EV⁺OT-1 and EV⁺ P14 cells. (E) Representative histograms show the SS of EV⁺ OT-I and P14 CD8⁺ T cells with average frequencies ±SD of cells with a SS>1. The bar graphs show the frequencies and the median similarity scores ±SD of EV⁺, EV⁻ OT-I and P14 T cells. Results from OT-I and P14 cells from mice that did not receive EVs are labeled “no EV inj”. Individual mice in different graphs are represented by the same symbol. Results from three mice were pooled from three independent experiments. For statistical significance paired t test was used with *P < 0.05, **P < 0.01, and ***P < 0.001.

Both, OT-I and P14 T cells were present in similar frequencies of approximately 15% of all CD8⁺ T cells (Fig. 4B). This frequency was irrespective of whether mice had received EVs or not, indicating that EVs could neither induce specific T cell proliferation, expansion, nor survival during this 1-h in vivo incubation (Fig. 4B). However, when we analyzed OT-I and “Bystander” P14 T cells for EV-binding, about twofold more OT-I T cells were associated with OVA-EVs than nonspecific P14 T cells (Fig. 4C). This observation indicates a contribution of cognate TCR–peptide/MHC interaction to EV-binding, corroborating our findings in Fig. 2C.

To determine if EV association could induce TCR signaling, we next analyzed nuclear NFATc1 translocation in the T cells by IFC (Fig. 4D and SI Appendix, Fig. S8). Bystander P14 cells had comparable frequencies of nuclear NFATc1 (Fig. 4 D and E), regardless of whether they were bound to EVs or not or were isolated from control mice that had not received EVs (Fig. 4E). This suggested that activated T cells which bind EVs with unspecific peptide/MHC complexes do not receive TCR-downstream signals via the NFAT-signaling pathway. However, in marked contrast to P14 cells, we observed statistically significant NFATc1-translocation from the cytoplasm to the nucleus in EV⁺ vs. EV⁻ OT-I T cells and relative to OT-I T cells in mice that had not received EVs at all. (Fig. 4E). This clearly indicated that recognition of peptide/MHCI complexes on EVs indeed triggered Ag-specific TCR signaling in OT-I T cells. Moreover, the strength of the NFATc1 signal in EV⁺ OT-I cells was much higher than in EV⁻ OT-I cells, as measured by the colocalization (similarity) score with DRAQ5 (ss of 1.5, Fig. 4E). These results demonstrate that TCR stimulation as assessed by NFAT translocation is strongly enhanced when cognate Ag is present on EVs. Importantly, this TCR stimulation was independent of DCs since the OT-I peptide cannot be presented on MHC-I of H2-K^bm1 mice. To our knowledge, this is the first direct demonstration of Ag-dependent TCR signaling by EVs in vivo. Direct TCR stimulation by EVs is also supported by our demonstration of PS⁺CD9⁺ EVs colocalizing with the TCR complex (Fig. 2E).

Discussion

Especially activated CD8⁺ T cells associate with naturally occurring EVs during viral infection. These EVs induced TCR signaling in Ag-specific CD8⁺ T cells in vivo. As a result, CD8⁺ T cells receiving signals from EVs had enhanced T_E cell gene signatures, while T cells without EVs showed more memory T cell gene signatures. Our data suggest that EVs attached to the surface of CD8⁺ T cells act as adjuvants for virus-specific activated CD8⁺ T function to enhance antiviral T effector cell responses.

Our approach of staining naturally occurring EVs is limited to PS⁺ EVs. The presence of PS in the membrane of EVs is very common (4, 5), and PS-affinity based methods are used to isolate almost all EVs from bodily fluids, suggesting that the PS⁺ EVs represent the majority of all EVs. Yet, PS⁻ or PS^low EVs seem to exist as a minor fraction and a recent study showed that injected PS⁻ EVs persist longer in the blood circulation of mice than PS⁺ EVs (52). However, in our own analysis of serum-derived EVs from healthy and infected mice, virtually all EVs (>95%) were PS⁺. Additionally, we confirmed our result with a PS-independent EV-detection approach using PKH26-labeled BMDC-derived EVs.

We initially developed our in vivo PS-labeling approach to detect apoptotic cells in vivo but realized that >90% of the stained cells were not apoptotic but live cells with EVs attached to their surface (27). To identify the rare apoptotic cells, we used a deep learning-assisted image interpretation approach. However, PS-staining of cells is not limited to apoptotic or EV⁺ cells. Highly phagocytic cells may also become PS⁺ by our method, although most of the eGFP signal is quenched after phagocytosis of the MFG-E8-eGFP-stained material. We also excluded that MFG-E8-eGFP is bound via integrins, as MFG-E8 constructs lacking the RGD-motif gave identical results (27). While there are also reports that B cell receptor and TCR signaling induces transient PS exposure on the cell surface (53, 54), our results from this and previous studies (27) suggest that this rather increases their capacity to bind PS⁺ EVs and that they therefore become PS⁺.

Previous publications using in vitro generated exosomal EVs from APC such as B cells or DCs reported activation of naive CD4⁺ (6, 16) and naive CD8⁺ T cells (7). However, in these landmark studies, naive T cells were not activated by direct exosome–T cell interactions but instead required bystander DCs, which indirectly presented specific exosome-borne Ag (16, 20). During LCMV infection, most T cell-bound EVs were of APC origin as they carried the corresponding surface molecules MHC-II, CD54, and CD86 and exosome markers CD9 and CD63 (27). These EVs caused NFATc1 nuclear translocation in activated EV⁺ but not EV⁻ CD8⁺ T cells arguing for direct effects rather than a role for DCs in this setting. As we found that EVs associated with CD44⁺ activated CD8⁺ T_E cells but not naive CD8⁺ T cells, it is, therefore, rather unlikely that EVs can activate naive CD8⁺ T cells through direct interaction in vivo, confirming previous findings (50).

These findings correlate with reports that activated LFA1 expressed by T cells (11) and CD54 on EVs (55) mediate T cell-EV interaction. As activated LFA-1 is not present on naive CD8⁺ T cells, it is likely that they neither can bind CD54⁺ EVs nor get activated directly by EVs generated in vivo. Due to the widespread usage of in vitro generated EVs in most previous studies, it remained unclear whether one could also transfer these findings to EVs in vivo (reviewed in refs. 1–3). Possible contributions of the amount, purity, Ag-density, bioavailability, and half-life of artificially generated EVs from cell cultures could not be excluded and might have influenced these former studies. We show for the first time the kinetics of EV interaction with CD8⁺ T cells in vivo during acute viral infection, which correlated roughly to the expansion kinetics of CD8⁺ T cells and peaked between day 5 and 10 p.i. (30, 56). Such EV–CD8⁺ T cell association is likely not dependent on a temporal increase in generation of EVs as observed in various types of pathogen infections such as HIV (57), SARS-CoV-2 (58), mycobacteria (59), or plasmodium (60). On the contrary, we found a significant drop in EV serum numbers in LCMV-infected mice. However, LCMV-associated destruction of the splenic microarchitecture, including the marginal zone (61) might lead to increased EV influx into the spleen and hence facilitate EV binding by T cells resulting in the reduction of serum EV numbers.

TCR specificity did contribute to EV–T cell binding, as activated specific CD8⁺ T cells bound higher frequencies of EVs in vivo than activated nonspecific T cells. This is also supported by the superresolution microscopy data showing clear colocalization of EVs with CD8, which is part of the TCR complex. Furthermore, EVs carry CD86 and CD54 and could engage costimulatory molecules and therefore facilitate EV binding by all activated CD8⁺ T cells, independently of TCR specificity.

The continuous stimulation of T cells and NFAT activation by EVs might lead to T cell exhaustion in certain settings. Indeed, NFAT has been shown to induce exhaustion in CD8⁺ T cells, when not complexed with the transcription factor AP-1 (62). Indeed, the MAP kinase pathway is sensitive to tuning by persistent Ag and overexpression of its target c-Jun makes CAR-T cell products more resistant to exhaustion (63–65) Such unbalanced TCR signaling is unlikely to happen during an acute viral infection like LCMV_Arm, as we have observed a rapid drop of EV binding after viral clearance. However, it is conceivable that EVs could contribute to T cell exhaustion in other settings, such as chronic viral infections and cancer. There is considerable evidence that specifically tumor-derived exosomal EVs carry PD-L1 on their surfaces and can thereby suppress both activation and proliferation of CD8⁺ T cells in an Ag-independent way (66–73). As PD-1 signaling attenuates T cell activation by interfering with TCR/MHC and CD28/CD86 signaling pathways, tumor-derived PD-L1⁺ EVs also inhibited PD-1⁺ CD8⁺ T cells with other TCR specificity than for tumors (66). Therefore, it is unlikely that the TCR needs to receive direct signals from MHC on tumor-derived EVs in this setting.

In marked contrast, we show that APC-derived exosomal EVs influence T cell function and differentiation during antiviral immune responses by directly engaging the TCR. The extent to which costimulatory pathways are necessary for these positive effects is unknown, however CD28 and ICOS induce sustained PI3K activity, leading to the upregulation of NFATc1 transcription (74). But it is unclear, whether costimulatory molecules alone have the potential to increase nuclear NFATc1 translocation, as unique, TCR-independent roles for CD28 have been challenging to dissect (75). It has been shown that CD28 signals alone can inhibit the kinase GSK3β, which promotes NFAT export from the nucleus (76, 77), causing trapping of NFAT in the nucleus. Potentially, such NFAT-trapping may cause slight effects on NFATc1 translocation, as we have observed in nonspecific EV⁺ T cells in LCMV-infected mice.

NFATc1 positively regulates cell proliferation and represses cell death (78), and EV⁺ CD8⁺ T cells showed upregulation of corresponding gene clusters II (preparation for cell cycle) and III (cell cycle and division), linking NFATc1 translocation to gene expression of EV⁺ CD8⁺ T cells. Among known genes regulated by NFATc1, we found upregulation of Egr2, Ifng, Irf4, Tnf, and Nr4a1 in EV⁺ CD8⁺ T cells. Especially the effector cytokines IFN-γ and TNF-α are NFATc1 dependent (79). CD8⁺ T_E cells are central to fight viral infection by cytotoxic lysis of infected cells (80). In addition, noncytocidal effector mechanisms mediated by CD8⁺ T cell-derived IFN-γ and TNF-α critically contribute to viral clearance of LCMV and other viruses (81–86). Therefore, a potential boost of such effector cytokine gene expression in EV⁺ CD8⁺ T cells suggests an additional, previously unrecognized EV/exosome-mediated adjuvant effect for Ag-specific effector CD8⁺ T cells.

Materials and Methods

Mice.

All mice were housed and bred under specific pathogen-free conditions at the Core Facility Animal Models of the Biomedical Center of the Ludwig-Maximilians-University, Munich. All protocols were approved by the Government of Oberbayern. Age and sex-matched mice of both sexes were used at 6 to 12 wk of age. C57BL/6NRj mice were bred in-house or purchased from Janvier (strain C57BL/6), P14/CD90.1 (B6-Tg (TcrLCMV)327Sdz-Thy1^a) mice express a transgenic Va2/Vb8.1 TCR that is specific for specific for the LCMVgp33-41 peptide, presented by H-2D^b (36LCMV). OT-I/CD45.1 and OT-I/CD90.1 (B6-Tg (TcraTcrb)1100Mjb-ptprc^a) express a transgenic Vα2/Vβ5 TCR recognizing the OVA peptide SIINFEKL (OVA257-264) by H-2K^b (37). To reduce the number of mice needed for this study and to apply the 3R principles, OT-I mice used were either CD45.1⁺CD90.1⁻ (Fig. 2) or CD45.1⁺CD90.1⁺ (Fig. 4). H-2K^bm1 mice carrying a mutant H-2K allele are incapable of presenting the OVA_257-264 peptide (51). To keep animal numbers at a minimum and to determine the optimal sample size, the software G*Power (87) was used.

Antibodies/Reagents.

See SI Appendix, Table S1.

LCMV.

Mice were infected with 2 × 10⁵ p.f.u. LCMV Armstrong intraperitoneally. Virus was propagated in L929 cells. Viral titers were determined by focus-forming assays on Vero cells followed as described previously (88).

Preparation of Single-Cell Suspensions.

Single-cell suspensions of the spleen were prepared by meshing the organs through a nylon mesh followed by erythrocyte lysis or by centrifugation of cells through Pancoll (PAN Biotech). For NFAT translocation experiments, cells were immediately fixed after organ removal by preparing the single-cell suspension using ice-cold FACS buffer (PBS+2% FCS) and 4% PFA mixed 1:1 as described previously (65).

Serum EV Isolation.

Serum samples were diluted with PBS + Protease inhibitors (Complete tablets, EDTA-free, Roche). After centrifugation (1,500 g and 10,000 g, 10′, RT), serum was loaded onto qEV 35-nm columns (Izon Science) and the flow through was collected (500 μL fractions). Fractions 2 to 7 were pooled and concentrated to 300 µL. Particle number and size distribution of serum EVs were determined by NTA using the ZetaView PMX110 instrument (ParticleMetrix). Eleven positions were measured with three reading cycles. Preacquisition parameters were sensitivity = 75, shutter speed = 50, frame rate = 30 fps, trace length = 15. Postacquisition parameters were minimum brightness = 20, pixels size = 5 to 1,000. For ImageStream analysis serum, EVs were stained with mC1 for 1 h at RT followed by the addition of 5% (v/v) NP-40 to detergent controls or the corresponding volume of filtered PBS. Prior to analysis, EV samples were fixed with 2% (v/v) filtered PFA. All data were acquired at 60× magnification at low flow rate and with the removed beads option deactivated, as described previously (89).

Adoptive Transfer.

For adoptive transfer of OT-I and P14 CD8⁺ T cells, single-cell suspensions of the spleen were prepared. Non-CD8⁺ T cells were removed using the CD8⁺ T cell negative selection isolation kit (Miltenyi Biotech). Isolated CD8⁺ T cells were activated in vitro for 2 d using 10 µg/mL anti-CD3 and anti-CD28 mAbs coated 96-well plates (Sarstedt) and then 3.0 to 3.5 × 10⁶ cells were adoptively transferred into recipient mice and analyzed 3 d later (90).

Flow Cytometry/IFC/FACS Sort.

First, 1 × 10⁶ and 5 × 10⁶ cells were stained for flow cytometry of IFC, respectively, with appropriate antibody mixes (20 min on ice) and analyzed on a FACSCanto (BD Biosciences) or an ImageStream^X MKII imaging flow cytometer (Luminex). For digitally sorting PS⁺ cells into apoptotic and EV⁺ subsets, a CAE was used as described in ref. 27.

For sorting PS⁺ CD8⁺ T cells for dSTORM microscopy, mice injected with 50 µg C1-SA-AF647 to stain PS in vivo. For C1-SA-AF647 conjugation, C1 monomers (Apo-Monomer, Cat. #480157, Biolegend) were mixed with Streptavidin-AF647 in a 1:5 ratio. Mice were killed 1 h after the injections. Spleens were removed and surface staining with CD8-CF488 and CD9-CF586 (antibodies were labeled in-house with dSTORM compatible dyes using Mix-n-Stain™ CF™488 and CF™568 kits according to the manufacturer’s instructions) was performed on splenocytes. To exclude unwanted and dead cells, LIVE/DEAD™ violet, CD11c-BV421, IgM-BV421, and CD4-BV421 were added to the dump channel. After fixation with 4% PFA (20 min), cells were sorted on a FACSAriaIII cell sorter (BD Biosciences) using a 130-µm nozzle and low flow rate to minimize shearing forces resulting in EV loss.

For sorting PS⁺ CD8⁺ T cells for RNAseq analysis, spleens from mice injected with 100 µg MFG-E8-eGFP were stained with LIVE/DEAD™ violet anti-CD8 PE-Cy7, anti-CD44 APC/Fire750, and anti-CD62L PE mAbs. Cells were sorted on a FACSAriaIII cell sorter (BD Biosciences) using a 130-µm nozzle. After sorting cells were lysed in Trizol.

dSTORM Superresolution Microscopy.

For dSTORM, sorted and labeled cells (see section Flow cytometry/IFC/FACS sort) were immobilized on 0.1% (w/v) poly-L-lysine (P8920, Sigma-Aldrich)-coated 76 × 26 mm slides (03-0001, Langenbrinck) with a 22 × 22 mm 1.5H cover glass (80-2222/5, Marienfeld). Freshly prepared BCubed STORM-imaging buffer (ONI, Oxford Nanoimaging) was added prior to image acquisition on a temperature-controlled Nanoimager S Mark II microscope from ONI. Images were taken in dSTORM mode acquired sequentially using the total reflection fluorescence (TIRF) illumination (calculated evanescent field penetration depth was >200 nm). Before imaging, channel mapping was calibrated using 0.1 µm TetraSpeck beads (T7279, Thermo Fisher Scientific). Superresolution images were filtered using the NimOS software (v.1.18.3, ONI) and data have been further processed with the Collaborative Discovery (CODI) online analysis platform from ONI.

RNAseq.

RNA from sorted cells was isolated by Vertis Biotechnology AG (Freising, Germany) with Monarch RNA Cleanup Kit (New England Biolabs). First-strand cDNA was synthesized using an oligo(dT)-adapter primer and M-MLV reverse transcriptase. The Illumina TruSeq sequencing adapter was ligated and the cDNA was PCR amplified to about 10 to 20 ng/μL using a high-fidelity DNA polymerase followed by cDNA purification. Samples were pooled in approximately equimolar amounts. The cDNA pool in the size range of 300 to 400 bp was eluted from an agarose gel and sequenced in four runs on an Illumina NextSeq 500 system using 1 × 75 bp read length. Sequencing reads were aligned to the mouse reference genome (version GRCm38.99) with STAR (version 2.6.1d). Expression values (TPM) were calculated with the software package RSEM (version 1.3.0). Post-processing was performed in R/bioconductor (version 4.0.0) using default parameters if not indicated otherwise. Differential gene expression analysis was performed with “DEseq2” (version 1.26.0). An adjusted P value (FDR) of less than 0.1 was used to classify significantly changed expression. GSEA were conducted with “clusterProfiler” (version 3.18.1) on the statistic reported by DEseq2. Data are available at Gene Expression Omnibus GSE201507.

Production of BMDC-derived EVs.

Bone marrow cells from femur and tibia were seeded at a density of 1 × 10⁶/mL in RPMI 1640 Glutamax (Fisher Scientific) medium with 10% EV-free FCS, 1% penicillin/streptomycin, 20 ng/mL GM-CSF (produced in-house, centrifuged at 100,000 g at 4 °C for 18 h to remove EVs), and 0.05 mM 2-mercaptoethanol. EV-free FCS was produced as described (91). In short, FCS was diluted 1:1 with pure RPMI and centrifuged at 100,000 g at 4 °C for 18 h. The supernatant was filtered (0.2 µm and stored at −20 °C).

After 6 d, adherend cells were stimulated with 30 ng/mL LPS (Sigma-Aldrich) and 2 µg/mL SIINFEKL peptide (NeoBiotech) for 72 h. Then, the supernatant from activated BMDCs was collected (from approximately 70 T175 bottles per experiment), debris was removed (2,000 g, 20 min), filtered (0.2 µm), and centrifuged [90 min, 100,000 g, 4 °C using a type 45 Ti fixed-angle rotor (Beckman Coulter)]. EV-containing pellet was resuspended in PBS containing protease inhibitors (Complete tablets, EDTA-free, Roche), pooled, and concentrated using Amicon spin columns. For PKH26 labeling, Diluent C and PKH26 were incubated for 5 min. Staining was stopped by adding 2 mL of sterile-filtered 1% BSA. Excess dye was removed using Amicon spin columns. After washing, labeled EVs were mixed with iodixanol (Optiprep) (60%) and layered onto the bottom of ultraclear centrifuge tubes (Beckman Coulter; Cat no 344062). Another layer of 30% iodixanol diluted in PBS was placed on top, followed by a final layer consisting of filtered PBS. The discontinuous density gradient was then centrifuged for 160,000 g at 4 °C for 18 h in an SW 55 Ti swinging-bucket rotor (Beckman Coulter).

Then, 500 µL fractions were collected from the top of the gradient. Fractions 3 to 8 were pooled, washed, and concentrated using Amicon spin columns. Concentration of EVs was determined using Nanoparticle Tracking Analysis (NTA). Then, 2 to 7 × 10¹¹ EVs were injected into recipient mice. For PKH26-only controls, PKH26 and Diluent C were added to PBS, containing protease inhibitor, without EVs. Subsequently, the same protocol as for the EV-containing samples was adhered to and fractions 3 to 8 were pooled after the iodinaxol gradient ultracentrifugation. Concentrated fractions were injected into recipient mice at day five of LCMV infection.

Statistical Analysis.

For statistical analysis, the PRISM software (GraphPad Software) was used. Significance was analyzed using Student’s t test or one-way ANOVA test unless stated otherwise, with *P < 0.05, **P < 0.01, and ***P < 0.001. Bar graphs show average ±SD.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNAseq data have been deposited in Gene Expression Omnibus (GSE201507) (92).