Cell contact-dependent acquisition of cellular and viral nonautonomously encoded small RNAs

Abstract

In some organisms, small RNA pathways can act nonautonomously, with responses spreading from cell to cell. Dedicated intercellular RNA delivery pathways have not yet been characterized in mammals, although secretory compartments have been found to contain RNA. Here we show that, upon cell contact, T cells acquire from B cells small RNAs that can impact the expression of target genes in the recipient T cells. Synthetic microRNA (miRNA) mimetics, viral miRNAs expressed by infected B cells, and endogenous miRNAs could all be transferred into T cells. These mechanisms may allow small RNA-mediated communication between immune cells. The documented transfer of viral miRNAs raises the possible exploitation of these pathways for viral manipulation of the host immune response.

Small RNAs can control the expression of genes at the post-transcriptional level by inhibiting translation or by promoting mRNA degradation (Dykxhoorn et al. 2003; Bartel 2004; He and Hannon 2004; Meister and Tuschl 2004; Mello and Conte 2004; Valencia-Sanchez et al. 2006; Selbach et al. 2008). In Caenorhabditis elegans, the introduction of dsRNAs via feeding can induce both systemic silencing throughout the organism and a heritable response that is transmitted through the germline (Mello and Conte 2004). This relies on a membrane protein, SID-1, that is conserved throughout multicellular animals (Feinberg and Hunter 2003). It was even recently shown that systemic silencing is required for antiviral immunity in Drosophila (Saleh et al. 2009).

In plants, viral silencing responses spread throughout the organism, even entering tissues engrafted onto the plant in which the response was initiated (Tijsterman et al. 2002). Plants' plasmodesmata present convenient channels for short-range spread of small RNAs (Carrington 2000; Hofmann et al. 2007), but the mediators of long-range communication via RNAi pathways in plants are not well understood (Tijsterman et al. 2002).

Thus far, nonautonomous silencing responses have not been apparent in mammals, although the secretory compartments of mammalian cells were found to contain both mRNAs and microRNAs (miRNAs) (Valadi et al. 2007; Dinger et al. 2008; Skog et al. 2008). Intercellular transfer of proteins and small molecules (e.g., cAMP) does occur between specific mammalian cell types. For example, transfer can occur at the immune synapse (IS) through connections formed between conjugated immune cells (Bromley et al. 2001). This process is mediated by structures such as tunneling nanotubes, gap junctions, “pores,” and plasma membrane (PM) bridges (Joly and Hudrisier 2003; Bopp et al. 2007; Davis 2007; Rechavi et al. 2007). We postulated that these intercellular links might also allow acquisition of RNAs, particularly small RNAs, by mammalian cells, allowing noncoding RNAs to extend their regulatory functions beyond their cell of origin. We therefore tested the ability of human lymphocytes to acquire functional, nonautonomously encoded small RNAs upon cell–cell contact. We tested B cells as potential small RNA donors to αβ T cells as their interacting partners.

Results

We began by probing the potential for transfer of fluorescently (Cy3) tagged oligonucleotides from antigen-presenting cells (APCs) to T cells. As the donor, we used the Burkitt lymphoma B-cell line (BL2) that had been transfected with either a 22-base-pair (bp) (22bpCy3) duplex, analogous to a mature miRNAs, or a 72-bp hairpin oligonucleotide, to mimic its precursor. The BL2 cells were transfected by electroporation to obtain high efficiency uptake of small RNAs while avoiding the use of lipid (liposome)-based reagents, which might persist in the media and artifactually transfer small RNAs to recipient cells. The transfected BL2 cells were also thoroughly washed prior to coculturing to remove surplus Cy3-dsRNAs left from the original transfection.

As a recipient, we used a herpes virus saimiri (HVS)-immortalized human T cells that were cocultured with the transfected BL2 cells for varying periods of time to allow conjugates to form. At the end of coculture, the conjugates were disrupted by vigorous agitation and treatment with chelating agents. In addition, the cells were treated with RNase to reduce potential surface contamination with Cy3-labeled RNA species that might have either remained in the media or been released from dying B cells. The HVS T cells were stained with fluorophore-conjugated anti-CD3 monoclonal antibodies (mAbs) and purified by FACS.

In as little as 1.5 h, a large fraction of the HVS T cells acquired synthetic small RNAs from BL2-22bpCy3 cells (Fig. 1) or BL2-72bpCy3 cells (Supplemental Fig. 1). Neither tryptic digestion nor acid washing of the cocultures prior to the FACS analysis significantly diminished Cy3 intensity in the HVS T-cell partners, indicating that the 22bpCy3-conjugated RNA molecules were present inside the adopting cells (Supplemental Fig. 2). In parallel experiments, we introduced a 22bpFITC-conjugated locked nucleic acid (LNA) into the donor cells and found that the LNA oligonucleotide was not transferred from B to T cells (Supplemental Fig. 3). This indicated that the intercellular transfer pathway displays some degree of specificity.

Figure 1.

Cy3-tagged 22-bp oligonucleotides are transferred from BL2 cells to HVS CD4⁺ T cells. Dot plots of FACS analysis depict total live-cell gate records of HVS T cells and donor BL2 cells transfected with 22bpCy3 alone and after coculturing (see the Materials and Methods for details). (A) HVS T cells alone are positive for CD3 and negative for Cy3. (B) BL-22bpCy3 transfectants alone are positive for Cy3 and negative for CD3. (C,D) HVS T cells and BL2-22bpCy3 were cocultured for 90 min either at 37°C (C) or at 4°C (D) and then labeled for CD3 and analyzed by FACS. Appearance of CD3⁺/Cy3⁺ cells indicates 22bpCy3 acquisition by HVS T cells. Numbers in the quadrants represent the percentages of cells within each quadrant. Results shown are of a typical experiment performed in duplicate (data collected from ∼10,000 single-cell events). Similar results were obtained in >10 experiments.

Throughout these experiments, we employed a stringent doublet discrimination algorithm (see Materials and Methods) for the analysis of the FACS data to ensure that the cells in the CD3⁺ T-cell gate did not include B–T-cell doublets. Moreover, we performed additional coculture experiments to conclusively exclude B–T cell doublets from the analysis (Supplemental Fig. 4). We generated a donor B-cell lines (B721) engineered to stably express GFP and also transiently transfected these cells with 22bpCy3. The B721-GFP cells were then cocultured for 1.5 h with Jurkat T cells, and the mixture was stained for CD3. The rationale was that if 22bpCy3 transfer simply reflected B-cell endocytosis or cell fusion, GFP would be detected in the “pure” CD3⁺ T-cell gate. This set of experiments demonstrated (Supplemental Fig. 4) that the Jurkat T cells that acquired 22bpCy3 were devoid of GFP.

Interestingly, small RNA transfer was completely blocked by low temperature (4°C) (Fig. 1), clearly distinguishing this mode of transfer from the mode of transfer of intracellular and cell-surface proteins known to occur at the IS (Davis 2007; Rechavi et al. 2007).

To test whether the transfer of 22bpCy3 occurred unidirectionally (from B to T cell) or bidirectionally (also from T to B cell), we cocultured Jurkat T cells transfected with 22bpCy3 oligonucleotides together with BL2 cells for 1.5 h. We noted the appearance of 22bpCy3 in BL2 cells (Supplemental Fig. 5), indicating that intercellular transfer could occur in either direction.

A previous report (Valadi et al. 2007) indicated that total RNA purified from exosomes secreted by cultured human mast cells contain miRNAs. It was therefore critical to determine whether the strong miRNA transfer we observed was chiefly contact-dependent (namely, via the IS) or alternatively mediated by secreted particles (e.g., exosomes). Thus, we performed time course experiments either in which the donor and the recipient cells were mixed in suspension to allow both contact-dependent and contact-independent transfer or in which donor and recipient cells were separated by means of a semipermeable membrane (0.4-μm pore size). The latter arrangement would allow only contact-independent transfer. These experiments indicated both contact-dependent and contact-independent modes of transfer (Fig. 2A,B). The contact-dependent transfer of 22bpCy3 from BL2 transfectants to HVS T cells was rapid, being evident after only 1 h of coculture (Fig. 2A,B). In contrast, the contact-independent acquisition of 22bpCy3 was slow and became evident only after 5 h of coculture (Fig. 2A,B). Contact-dependent transfer was more efficient at all time points than was contact-independent transfer (Fig. 2A,B).

Figure 2.

The vast majority of oligonucleotides are transferred from BL2 to T cells in a contact-dependent manner. (A) Overlaid dot plots of FACS analysis depict the populations of HVS T and donor BL2-22bpCy3 cells recorded after 1, 5, and 24 h of coculturing (red dots) or after similar coculturing in Transwell (0.4-μm-pore membrane) in which the two cell types are physically separated from each other (green dotes, only the HVS cells are shown). Cy3-positive/CD3⁺ HVS T cells (acquired 22bpCy3) are apparent already after 1 h of coculturing and only after 5 and 24 h of Transwell coculturing. (B) The mean fluorescence intensity (log MFI) of 22bpCy3 recorded in CD3⁺ HVS T cells and in the BL2-22bpCy3 donor cells under the coculture conditions described in A as a function of coculture time. The Cy3 log MFI of HVS T cells that were cultured alone is shown as well. Results of a typical experiment performed in duplicate (data from ∼10, 000 single-cell events) are shown. Similar results were obtained in two additional experiments. (C,D) The rate of formation and the quality of the cell contact play a role in the efficiency of 22bpCy3 transfer. 2E2 T or Jurkat cells were cocultured with 22bpCy3 transfer or nontransfected Mel-526 cells for either 0.5 h or 1.5 h under various conditions as diagrammed in C. The efficiency by which Mel-526 transfectants can transfer 22bpCy3 molecules to 2E2 or to Jurkat T cells were measured by FACS as described in Supplemental Figure 5. (C) Schematic representation of the experiments and the results. (D) The actual results of the experiment, histograms represent the Cy3 intensities in the acceptor T cells measured under the different conditions. Results shown are from a typical experiment out of 3 performed in duplicate (data collected from ∼10 000 single-cell events).

When we quantified the 22byCy3 signal in both the donor B cells and the acceptor T cells, we noticed a reduction of Cy3 label in the donor B cells as a function of coculture time that was directly proportional to the increased signal observed in T cells (Fig. 2B). This phenomenon was observed in all the experiments we performed and suggested that the cell-to-cell transfer of nucleic acid material is quite efficient.

The slower and less significant contact-independent transfer of RNA between lymphocytes was confirmed by incubating T cells with the conditions media from BL2 transfectants for various time periods without T cells (Supplemental Fig. 6). Again, the ability of BL2 supernatants to transfer small RNAs to T cells became evident only when we used >5-h incubation media. The RNA transfer from BL2 supernatants to T cells persisted despite washing of the BL2 transfectants prior to media collection, treatment of the collected media with RNase A and H and removal of cell debris by high-speed centrifugation and filtration through 0.4-μm filter (Supplemental Fig. 6).

It was unclear whether the transfected RNAs themselves were transferred or whether metabolic products of those RNAs moved between cells. Particularly in the case of 72bpCy3, we did not determine whether this partially dsRNA was cleaved by Dicer into a mature miRNA mimetic before transfer. Moreover, in neither case, was it clear whether uncomplexed RNAs were moving or whether RNAs moved as part of an RNAi effector complex (RISC). To address this issue, we used engineered donor cells that stably express myc-tagged Ago2, the component of RISC that binds miRNAs, and examined whether the Ago2 protein moved from the donor cells to recipient T cells. While we and others have reported previously the transfer of other tagged proteins (e.g., Ras) (Rechavi et al. 2007), we could not detect Ago2 movement between lymphocytes (data not shown). It thus appears the 22bpCy3 does not travel between cells as part of RISC. In a previous study, we cocultured B cells that stably express GFP with nonexpressing T cells. We did not detect any GFP signal in the T cells even after long periods of coculture (Rechavi et al. 2007), as we would have observed if mRNAs were moving between cell types. Thus, it appears that the transfer mechanism that we observe operates more efficiently on small RNA species.

In the experiments described thus far, we employed a standard coculturing procedure in which both the donor and acceptor cells were spun briefly to promote the formation of cell-conjugates. However, we were also able to demonstrate transfer of 22bpCy3 under conditions where cell contacts were not enhanced by centrifugation (Fig. 2C,D). We also explored more generally the impact of the nature and quality of cell contact on the efficiency of 22bpCy3 transfer. To this end, we took advantage of the 2E2 T cell clone (kindly provided by Dr. Lotem, Hebrew University of Jerusalem, Israel). This was derived from patient tumor infiltrating lymphocytes and specifically recognizes and kills the HLA-A2⁺ melanoma cell line, Mel-526. We examined the efficiency with which Mel-526 transfectants could transfer 22bpCy3 molecules to 2E2 as compared with transfer efficiency to Jurkat T cells. For these studies, a variety of conditions were employed. We examined transfer in mixtures that were cosedimented to promote cell–cell contact and in nonspun mixtures incubated for varying times and at varying cell concentrations. In accord with prior observations, acquisition of small RNAs by Jurkat cells was strongly dependent on conditions, with those designed to promote contact (e.g., spinning, high cell concentrations) giving the best results. In contrast, 2E2 T cells acquired small RNAs efficiently under all conditions tested (Fig. 2C,D). Thus, specific recognition of antigen by the T-cell receptor could promoter RNA transfer under conditions in which it was otherwise inefficient.

To test explicitly whether miRNAs could be transferred, we took advantage of miR-127, which is expressed at very low levels in T cells. BL2 cells were transfected with unlabeled synthetic pre-miR-127 along with 22bpCy3 as a marker. Cells were washed thoroughly and after 24 h were cocultured for 3 h with freshly isolated T cells. We then performed cell sorting using a stringent multiparameter gating strategy to physically purify from the cocultures two populations containing Cy3^high and Cy3^low singlet T cells (Fig. 3A). We thus obtained two populations of highly purified (>99.5%) singlet T cells, each with a significantly different content of Cy3 (marker) oligonucleotide. We analyzed the levels of miR-127 in the two T-cell populations using a mature miR-127-specific TaqMan RQ-PCR assay and found that the T cells that acquired high amounts of 22bpCy3 (Cy3^high) from donor B cells contained higher levels of miR-127 than did the Cy3^low T cells that acquired low amounts of the fluorescent marker oligonucleotide (Fig. 3B). Even T cells that had low Cy3 fluorescence had substantially more miR-127 than did T cells that had not been cocultured with transfected B cells (Fig. 3B). Since the transfer of miR-127 occurred within 3 h, a time point at which transfer of RNAs by secretion is undetectable (Fig. 2), it appears that miR-127 was transferred from B to T cells in a contact-dependent manner.

Figure 3.

T cells acquire nonautonomous miR-127 and EBV-viral miRNAs from B cells. (A) Dot plots demonstrating the isolation of highly purified human CD3⁺ T cells composed of >99.5% singlet cell from cocultures of CD3⁺ T cells and BL2- pre-miR-127/22bpCy3 cotransfectants using stringent FACS gating strategy are shown. The 22bpCy3 served as a reporter for small RNA transfer. Side scatter area (SSC-A)/forward scatter area characteristics (FSC-A) plot (panel i) and the FSC-H (height)/FSC-W (width) plot (panel ii) are shown where the arrow denotes the singlet CD3⁺ T cells consisting of two populations of low and high Cy3 content (panel iii) that were sorted into CD3⁺ cells Cy3^low (panel iv) and Cy3^high (panel v). (B) Levels of miR-127 in the sorted CD3⁺ Cy3^high and CD3⁺ Cy3^low cells and in CD3 Cy3-negative cells (from T cells that were not cocultured) as determined by real-time PCR (qPCR). Data are expressed in terms of fold increase in miR-127 content (means, n = 2) and represent a typical experiment out of three with similar results.

It remained formally possible that coculture with transfected B cells induced expression of miR-127 in recipient cells. It was therefore critical to examine the transfer of a natural miRNA that could not be encoded by the recipient T cells. Raji is a B-cell lymphoma line that carries the Epstein-Barr virus (EBV). The EBV genome encodes several small RNAs, including EBV-miR-BHRF-1-2 and EBV-miR-BHRF-1-1, that enter the miRNA pathway in infected cells (Cai et al. 2006). Raji cells and freshly isolated human T cells were cocultured for 3 h. A highly purified T-cell population (>99.5% purity) was subsequently recovered from the coculture. Parental Raji cells, parental T cells, and the cocultured T cells were used for the preparation of small RNA fractions (19–24 nucleotides). These were adapter ligated, reverse transcribed, and profiled by deep sequencing (Ibarra et al. 2007). We tested several normalization procedures to permit comparative analysis of the deep sequencing data (see the Materials and Methods).

As expected, donor Raji cells expressed several EBV-derived miRNAs, mainly EBV-miR-BHRF-1-2 and EBV-miR-BHRF-1-1 (Cai et al. 2006). Remarkably, a significant fraction of the two prominent EBV-derived miRNAs was also detected in cocultured T cells. Since Raji cells do not produce infectious virus, virally encoded miRNAs must have originated in the B-cell donor. Using a highly conservative bootstrapping method, we excluded the possibility that the detection of EBV-derived miRNAs in T cells resulted from simple contamination of T-cell fractions with Raji cells (P < 0.015) (Fig. 4A). Complementary experiments with a Raji cell line engineered to stably express GFP (GFP-Raji) provided additional support for our conclusions. In these experiments, T cells were cocultured for 1.5 h with Raji-GFP and then sorted to high purity (Supplemental Fig. 7) and probed by RT-qPCR for the presence of the EBV-derived BHRF-1-2 miRNAs (Supplemental Fig. 8). Again, T cells acquired EBV-derived miRNAs in a time-dependent manner. It should be noted that a very stringent doublet discrimination algorithm was employed (see the Materials and Methods), and as shown in Supplemental Figure 7, <0.5% GFP⁺ events were recorded in the purified T cells. Thus, these studies confirmed that the sorted CD3⁺ T cells did not include B–T-cell fusion events.

Figure 4.

Levels of nonautonomous miRNAs from deep sequencing under different normalization conditions. (A) Expression levels of EBV-BHRF-1/2 from CD3⁺ cells after a 3-h incubation with Raji cells. The red line denotes the normalized expression level of EBV-BHRF-1/2 found in the incubated CD3⁺ cells. The histogram shows the expected distribution of EBV-BHRF-1/2 that would be induced by 0.5% Raji contamination as calculated using bootstrapping. The blue line indicates the mean value of this distribution. The difference in standard deviations bounds the type I error. We analyzed this effect with five different normalization procedures, and reported the most conservative results: P < 0.015 (normalizations to tRNA levels). (B) Endogenous miR transfer. Expression levels of the miR-18a from deep sequencing under normalization conditions as in A. The most conservative results are normalization to miR-31 that corresponds to P < 0.0004.

Next we asked whether two endogenously encoded miRNAs expressed at high levels in resting Raji cells, miR-106b and miR-18a, were also transferred to cocultured T cells. Of these, we detected efficient transfer of miR-18a. Contamination alone cannot explain the increased abundance of miR-18a in cocultured T cells, which was statistically significant after all normalization procedures (worst P value, 0.0004) (Fig. 4B). In the case of miR-106, the increased signal in recipient cells did not meet our rigorous test for significance.

We next asked whether transferred small RNAs could exert a regulatory impact in the recipient cells (Fig. 5A). It can often be problematic to identify bona fide and highly regulated targets of endogenous miRNAs in a given cell type. We therefore created Raji cells stably expressing miRNA mimetic shRNAs engineered to specifically regulate luciferase, CD4, or mouse p53. Each of these Raji cell lines was cocultured with two different types of recipient cells. The first was a Jurkat T-cell line engineered to stably express a fluorescent Venus reporter fused to a portion of luciferase containing the target site (Jurkat-luciferase-venus). The use of such a fusion protein was necessary to enable monitoring and quantification of small RNA effects by FACS. The second type of recipient was a T-cell line that endogenously expressed the surface molecule CD4 (CD4⁺ T cells). Since each target could be repressed only by its specific shRNAs, these recipients served as reciprocal controls. The p53 shRNA potently enters the RNAi pathway but does not recognize either the Venus reporter or CD4, thus serving as a general control for the effects of a nonhomologous shRNA.

Figure 5.

Nonautonomously expressed shRNAs function in recipient T cells. (A) A schematic presentation of the experiment depicts donor Raji cells stably expressing an shRNA against luciferase, a CD4 shRNA, or a control mouse p53 shRNA. These cells were cocultured either with Jurkat-3′-luciferase-Venus cells or with normal human CD4⁺ T cells. Decreases in the relevant target sensor (luciferase, detected by the venus fluorophore or CD4, detected by red fluorescent mAbs) are depicted. (B) A graph presenting results of an actual experiment out of more than three with similar results is shown. Experiments were performed in duplicate (data from ∼10 000 single-cell events). Bars, mean values (MFI). (*) P < 0.05; (**) P < 0.05 ANOVA with Bonferroni post-hoc test.

After 72 h of coculture, we used FACS to determine the mean fluorescence intensity (MFI) of the Venus fluorophore in the Jurkat-luciferase-Venus recipient cells. CD4 expression was similarly quantified by FACS after staining of cells with a fluorescently labeled anti-CD4 antibody. A statistically significant reduction in Venus reporter expression was seen in the Jurkat-luciferase-Venus cells cocultured with Raji expressing anti-luciferase shRNAs as compared with cells cocultured with the Raji cells expressing CD4 or p53 shRNAs (Fig. 5). Additionally, we detected a significant (P < 0.05) down-regulation of CD4 expression in CD4⁺ T cells cocultured with Raji cells expressing a CD4 shRNA as compared with CD4⁺ T cells cocultured with Raji cells expressing either p53 or luciferase shRNAs (Fig. 5B).

We also tested whether the transfer of the endogenous, EBV-derived miRNAs, BHRF-1-2, could regulate gene expression in recipient T cells using a similar target sensor approach. We observed sensor silencing in T cells, the efficiency of which correlated with the number of coincubated Raji cells (Supplemental Fig. 9). A scrambled, control sensor was unaffected.

These experiments suggest that in mammals, transferred RNAs are able to spread functional silencing signals across cell boundaries. It should be noted that the short time of coculture may have somewhat underrepresented the eventual impact of the intercellular silencing signal, since our measurements were subject not only to the efficiency of mRNA targeting but also to the half-lives of the proteins that we were quantifying.

Discussion

Considered together, these studies demonstrated a hitherto unidentified, efficient, and primarily contact-dependent transfer of small RNAs between B and T lymphocytes. We also demonstrate the potential of these transferred RNAs to exert regulatory effects. Thus, small RNAs now join proteins and chemical messengers as potential means of communication between cells. Our results do place an explicit limit on the types of RNAs that can be transferred, since we focused only on canonical small RNAs and their immediate precursors. Although we could not detect intercellular transfer and expression in trans of mRNAs in our experiments, it remains possible that larger RNAs might migrate between cells under certain conditions.

Our observations do not challenge the many experiments suggesting that in mammals, small RNAs do not spread systemically as is observed in plants and in some nematodes. Instead, our studies suggest intercellular transfer only under a few restricted conditions when cells establish intimate contacts. This situation offers the opportunity not only for the use of small RNA transfer to regulate normal physiology but also for the exploitation of these pathways by pathogens. Indeed, we demonstrate that virally encoded miRNAs are transferred between infected B cells and T cells and can silence the expression of target transcripts in the recipient cells. Given the propensity of viruses to exploit such opportunities, it seems likely that the studies presented herein will prompt more investigation of the nonautonomous effects of small RNAs in contexts where an advantage may be gained by immune modulation. Moreover, our results also raise the possibility that transfer of small RNAs will be a general principle encountered when other cell types establish the types of contacts represented by example in immune cells.

Materials and methods

Human subjects

This study was approved by the Institutional Ethics Committee at the Chaim Sheba Medical Center. All peripheral blood samples were obtained from healthy subjects.

Antibodies and reagents

Allophycocyanin or FITC mAbs directed against CD3 and CD4 were from BD Biosciences.

shRNA, oligonucleotides, plasmids, and transfections

We purchased 22- and 72-bp oligonucleotides conjugated to Cy3 and pre-miR-127 from Ambion. The lentivirus and retrovirus expression vectors containing cDNAs encoding the 3′-luciferase-Venus target sensor, GFP-shRNAs against CD4, luciferase, and P53 were described previously (Chang et al. 2006), a sequence complementary to the BHRF1-2 sequence (UAUCUUUUGCGGCAGAAAUUGA) was cloned to the psiCHECK2 (Promega) luciferase sensor vector as described previously (Borchert et al. 2006). BL2, Raji, and Jurkat cells were selected to stably express the indicated vector as described elsewhere (Paddison et al. 2004). The 5′ FITC-labeled scrambled knockdown LNA 22-bp oligonucleotides was from Exiqon.

Isolation of T cells

Peripheral blood lymphocytes (PBLs) were isolated by density-gradient centrifugation on Histopaque 1077 (Sigma), as described previously (Rechavi et al. 2007). Primary CD4⁺ T cells were isolated from the PBLs by the use of anti-CD4 microbeads and the MACS cell separation system (Miltenyi Biotec). To obtain a population enriched with CD3⁺ T cells, CD56⁻ PBLs were cultured for 16 h in plastic flasks, and the nonadherent cells (∼85% CD3⁺ T cells) were collected.

T-cell cultures

Cells were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum (FBS), 2 mM L-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin, all from GIBCO, and maintained at 37°C in a humidified 5% CO₂ incubator. T cells were cultured for in a medium supplemented with 100 IU of rhIL-2.

Cell lines and transfection

The human cell lines Jurkat, HVS, BL2, and Raji were cultured in RPMI-1640 + 10% FBS; the CD4⁺ HVS T cells (generated as previously described) (Rechavi et al. 2007) were cultured in RPMI-1640 medium + 10% FBS supplemented with 100 IU of rhIL-2. BL2 cells were transfected by Amaxa electroporation (Program R-013), and infection of the BL2, Jurkat and Raji cells was done using retro- and lentiviruses as previously described (Paddison et al. 2004; Rechavi et al. 2007). The Mel526 cell line and the anti-melanoma 2E2 CD8⁺ T cell line were kindly provided by M. Lotem from Hebrew University.

Formation of cell conjugates and analysis of intercellular transfer

To examine the intercellular transfer of Cy3-tagged oligonucleotides, miRs, and shRNAs, the various BL2 transfectants were distributed into U-bottom tubes (0.5 × 10⁶ cells per tube in 1 mL) to which T cells or T-cell lines were added (0.25 × 10⁶ cells per tube in 1 mL) to obtain an effector to target ratio of 1:2. The culture tubes were centrifuged for 2 min at 1000 rpm to promote cell-conjugate formation and then cocultured at 37°C. For FACS-based analysis of Cy3-oligonucleotides (22/72 bp) transfer, the collected cells were resuspended vigorously in 5 mM EDTA/PBS and kept on ice for 30 min to allow cell conjugates to dissociate. Immunofluorescence staining with anti-CD3 or anti-CD4-allophycocyanin mAbs to label the T cells was performed for 30 min at 4°C. After labeling, the cells were washed and again resuspended in 5 mM PBS/EDTA. Data collected from ∼10,000 single-cell events were then analyzed by multiparametric FACS. Primary T cells or T-cell lines were distinguished from target cells by their smaller size (as defined by their FSC/SSC) and CD3 expression. To exclude B–T-cell conjugates from the analysis, we employed a very stringent state-of-the-art doublet discrimination algorithm using fluorescence height versus area and fluorescence width versus area pulse measurements to distinguish T-cell singlets from B–T-cell conjugates. This doublet discrimination model is considered very precise in cells, such as lymphocytes, as they are rather homogenous and spherical in shape.

Transwell assays and exosomes

Effector T cells were prevented from directly contacting the BL2 cells by a Transwell 0.4-μm-pore membrane (Costar). Briefly, 0.25 × 10⁶ effector cells (in 0.3 mL of medium) were either added to 0.5 × 10⁶ BL2-22bpCy3 cells (in 0.5 mL of medium) in the lower compartment (done in 12-well plates) or placed in the upper chamber separated from targets. The cells were incubated for different time periods at 37°C, and cells were then collected in 5 mM EDTA/PBS and analyzed for 22bpCy3 acquisition by the T cells as described above.

Acid and trypsin treatments

Cells were washed twice in PBS and resuspended for 4 min at 20°C in citrate buffer (0.13 M citric acid and 0.06 M Na₂HPO₄ at pH 3.3). The treatment was stopped by the addition of an excess of 5% FCS in PBS. For trypsin treatment, the cells were washed as above and then incubated for 5 min at 37°C in PBS containing 0.25% trypsin C−EDTA (Biological Industries). The treatment was stopped by the addition of an excess of 5% FCS in RPMI medium.

FACS analysis and cell sorting.

Cell samples were analyzed on a FACSCalibur using Cellquest software or on a FACSAria using FACSDiva software (all from BD Biosciences). FACS data analysis was done using FlowJo 7.2.1 software (Tree Star). All cell sorting experiments were performed on FACSAria. To obtain a single-cell suspension, cells were pretreated (as described above) so as to dissociate cell-conjugates and were sorted at 4°C. To sort out only viable T-singlet cells, we employed a stringent multiparametric gating strategy. Viable lymphocytes were identified by their distinct FSC and SSC (including pulse width, height, and area), propidium iodide exclusion, and expression of a distinct T-cell marker as indicated (e.g., CD3 or CD4).

Real-time qPCR

RNA was extracted with the mirVana miRNA isolation kit (Ambion). Mature miRNA quantification was performed by real-time qRT–PCR in an ABI PRISM 7900 Sequence Detection System (Applied Biosystems), using TaqMan miRNA Assays (Applied Biosystems) as described previously (Chen et al. 2005). Mature miR-127 and BHRF1-2 levels were normalized to small nuclear RNA RNU6B.

Deep sequencing

Small RNA libraries were prepared from Raji cells, CD3⁺ cells, and sorted CD3⁺ cells after 3-h incubation with Raji (iCD3⁺) and were cloned with Solexa adapters as described previously (Ibarra et al. 2007). The libraries were sequenced in three different lanes for 36 cycles, using the Illumina standard recipe. We obtained 4.2, 5.2, and 5.6 million reads for the Raji, CD3⁺ alone, and the iCD3⁺ cells, respectively.

Statistical analysis

P values were calculated as appropriate either by the Student's t-test or by ANOVA with Bonferroni post-hoc test. P < 0.05 was considered significant. Statistical procedures were carried out using SPSS 14.0 software (SPSS, Inc.).

Statistical analysis of small RNA expression profiles

We used bootstrapping to test the hypothesis that the detected levels of EBV-derived miRNAs in the iCD3⁺ library are due to 0.5% Raji contamination. We created an artificial Solexa library with 0.5% contamination level by randomly taking 1,000,000 reads from the CD3⁺ library and from the Raji with a ratio of 99.5% and 0.5%, respectively. This simulated library was aligned using BLAT (Kent 2002) to a database that contained human and EBV miRNA from miRBase (Griffiths-Jones et al. 2008) version 11.0 and human tRNA from the genomic tRNA database (Lowe and Eddy 1997). We repeated that process 100 times and saved the alignments. We aligned the iCD3⁺ library using the same parameters and saved the results. Since expression profiles that are obtained by Illumina GAII contain compositional data, one should use a priori normalization assumptions to conclude an absolute change in the expression level. We tested five different normalization conditions on the sum of the detection levels of EBV-miR-BHRF-1-1 and EBV-miR-BHRF-1-2. The five normalization conditions were total amounts of tRNA, miR-21 level, miR-103 level, miR-31 level, and total number of reads. miR-21, miR-103, and miR-31 were chosen since their expression values should increase after T-cell activation (Wu et al. 2007). We calculated the mean values and the variance of the EBV-derived miRNA on the 100 bootstrap simulations for each normalization condition (Fig. 4A). We found that EBV-derived miRNA expression levels in the iCD3⁺ libraries are 8.5 (tRNA), 138.6 (miR-21), 10.2 (miR-103), 11.4 (miR-31), and 15.75 (total RNA) standard deviation stronger than the observed one in the iCD3⁺. We used one-side Chebyshev inequality to calculate the confidence values (P): 0.014 (tRNA), 5.2E-05 (miR-21), 0.01 (miR-103), 0.008 (miR-31), and 0.004 (total reads). We reported the worst confidence value upon the normalization to tRNA. We repeated the same procedure for miR-18a and miR-106b. Note that Chebyshev inequality is a distribution-free method and therefore is a highly conservative test. Raw sequencing data and support programs are available on request.