Synaptic Release of CCL5 Storage Vesicles Triggers CXCR4 Surface Expression Promoting CTL Migration in Response to CXCL12

CD8⁺ T lymphocytes are a key component of the adaptive immune response to tumors and viral infections. Ag-specific CTL exhibit a wide range of functions, including migration to inflammatory sites, TCR-mediated target cell recognition, production of effector cytokines and chemokines, exocytosis of cytotoxic granules (CG), and thereby killing of target cells. Poor or inadequate CD8⁺ T cell immunity is, at least in part, responsible for tumor progression and persistent chronic infection. The CD8⁺ T lymphocyte response is initiated by the interaction of TCR with peptide–MHC class I (pMHC-I) complexes, resulting in rearrangement of surface receptors and signaling components to the contact zone between the T cell and target cell, referred to as an immune synapse (IS). Formation of a cytotoxic IS (cIS) is associated with polarization of the T cell microtubule-organizing center (MTOC) toward the cell–cell interface, which enables the directional delivery of CG, specialized secretory lysosomes that contain perforin and granzymes, toward the target and maintains specificity of the T lymphocyte cytolytic function. Indeed, release of CG content is confined to secretory clefts that provide a limited space in which cytotoxic agents are kept concentrated for target cell killing without any bystander effect (1–3). Rab27a and hMunc13-4 trafficking proteins are essential for the polarized exocytosis of CG by driving their docking at and fusion with the plasma membrane (4, 5). They are localized on vesicular structures distinct from CG, and both types of granules coalesce together near the plasma membrane to finally deliver the “lethal hit” (6).

Dynamic interactions of T cells with APC are important for the onset of immune responses (7–9). The migratory behavior of T lymphocytes is under the control of the chemokine–chemokine receptor network that regulates their spatiotemporal distribution within lymphoid tissues and recruitment within inflammatory areas and the tumor microenvironment (10, 11). In this context, CCL5 was one of the first chemokines implicated in regulating antitumor immune responses (12). Local production of CCR5 ligands CCL5 and CCL3 was shown to induce mobilization of CD8⁺ T cells and CTL-dependent tumor growth suppression (13). Moreover, recruitment of CCR5 at the IS between tumor-infiltrating lymphocytes and malignant cells contributes to T cell retention at the tumor site by inhibiting T cell responsiveness to CCL5 (14). In addition to their chemotactic properties, chemokines have been suggested to directly regulate T cell development, priming, and effector functions (15). CCL5 has been reported to induce T cell proliferation and lymphokine production and to enhance CTL-mediated cytolysis (16–18). Chemokines have also been implicated in T cell costimulation, through engagement of CCR5 and CXCR4 at the IS (19, 20). However, the potential contribution of chemokines to fine-tuning specific cytotoxic immune responses has not been systematically addressed. In this study, we provide further evidence that CTL store CCL5 in cytoplasmic vesicles (CCL5V) distinct from CG (21), and we show that their probable coalescence with the Rab27a-hMunc13-4 compartment at the IS with target cells, triggers, concomitantly with CG exocytosis, synaptic secretion of the chemokine. Remarkably, synaptic, but not multidirectional, secretion of CCL5 induces CXCR4 expression on the T cell surface, thereby promoting CTL detachment from target cells and migration in response to CXCL12. These results emphasize a mechanism regulating IS resolution by CTL following successful tumor cell killing. They are compatible with a model of CTL serial encounter and conjugation with multiple targets.

Tumor cell line IGR-Heu and autologous T cell clone Heu171, recognizing a mutated α-actinin-4 (actn-4) tumor Ag, were established as described (22).

CD8⁺ T cell blasts were generated from PBMC of two healthy donors (HD1 and HD2) and two patients displaying genetic deficiencies in either UNC13D (hMunc13-4^−/−) or RAB27A (Rab27a^−/−). Briefly, CD8⁺ T cells were selected using Isolation Kit II (Miltenyi Biotec) and cultured on irradiated allogeneic feeder cells in the presence of IL-2 (100 U/ml) and 3% conditioned medium from PHA-activated PBMC (22).

CD8⁺ T cells were freshly isolated from two additional HD (HD3 and HD4) and were transiently transfected with pEGFP-C1-Rab27a-GFP plasmid (23) in an Amaxa system using V solution and the X-001 program (Lonza).

The human HLA-A2–transfected Gerl-A2 melanoma cell line (kindly provided by Pierre Coulie, University of Louvain, Brussels, Belgium) and the murine FcγR-expressing P815 mastocytoma cell line were used as APC. SEB superAg was purchased from Toxin Technologies.

T cell phenotypic analysis was performed by flow cytometry (24) using anti-CD8 (Immunotech), anti–granzyme B (Caltag Laboratories), anti-CCR5, and anti-CXCR4 (R&D Systems) mAb. For the degranulation assay, T cells were stimulated in flat-bottom 96-well plates precoated with 5 μg/ml anti-CD3 mAb (UCHT1) in the presence of anti-CD107a mAb (Becton-Dickinson), as described (25). The time-course of CG and chemokine release was assessed in T cells stimulated with plastic-bound anti-CD3 mAb (10 μg/ml) in the absence or presence of 3 μM brefeldin A, 10 nM monensin or 10 μg/ml nocodazole (Sigma-Aldrich). Cells were fixed with 2% paraformaldehyde, permeabilized with 0.2% saponin, and then stained with anti-granzyme B, anti-CCL3, anti-CCL4 (R&D Systems), or anti-CCL5 (BD Biosciences) mAb (22). CCL5 secretion in cell culture supernatants was analyzed by ELISA (eBioscience). The induction of CXCR4 expression was assessed on T cells stimulated with plastic-bound anti-CD3 mAb, pretreated or not with 10 μg/ml nocodazole (Sigma-Alrich), in the absence or presence of 100 nM human rCCL5 (PeproTech), using anti-CXCR4 biotinylated mAb (R&D Systems) followed by streptavidin-PE–labeled secondary mAb (Molecular Probes). For inhibition of CXCR4 expression, T cells were incubated for 30 min with 1 μM of the CCR5 antagonist maraviroc (kindly provided by Serge Adnot, Créteil, France).

T cells were seeded alone or with tumor cells at a 1:5 E:T ratio in the upper chambers of Transwell plates, as described (14). After 30 min to 2 h of incubation, 50 nM human rCXCL12 was added to the lower chambers to trigger T cell migration. After an additional 2 h, the number of cells that had migrated into the lower chambers was counted by flow cytometry. Results were expressed as chemotaxis index (14). Chemokine receptor activity was inhibited by preincubating T cells in the upper chambers with anti-CXCR4 blocking mAb or maraviroc.

Tumor cells and T cells, untreated or pretreated with nocodazole, were plated on poly-(l-lysine)–coated coverslips at a 2:1 E:T ratio, as described (24). Staining was performed using rabbit anti-CCL5 (PeproTech) and mouse anti-granzyme B (Invitrogen), anti–α-tubulin (Sigma-Aldrich), anti-CXCR4, or anti-CCR5 mAb followed by a secondary mAb coupled to anti-mouse–Alexa Fluor 488 or anti-rabbit–Alexa Fluor 546 (Molecular Probes). Coverslips were then mounted with Fluoromount-G (Southern Biotech) and analyzed by a confocal microscope (LSM-510; Carl Zeiss Microimaging) with a ×63 lens. Polarization of CCL5V and CG was defined by the accumulation of CCL5 and granzyme B staining, respectively, at the contact area between effector and tumor cells, which was defined by a tight junction, characterized by ≥10 Z stacks. For each stack, the optimal Z interval was calculated by LSM Image Examiner software (Zeiss). A minimal fraction of three-fourths of CCL5 and granzyme B staining needed to be localized at the IS to be considered polarized. The colocalization of CCL5 and granzyme B and/or Rab27a was evaluated by the Manders coefficients determined by IMARIS software, which estimate the degree of colocalization of fluorescence A in fluorescence B within the region of interest. The microtubule network and MTOC were visualized by α-tubulin staining. Z-projection of slices was performed using LSM Image Examiner software (Zeiss).

For CCL5V relocalization in TCR-engaged T cells, poly-l-lysine slides were precoated with anti-CD3 (10 μg/ml) or anti-CD3 mAb (5 μg/ml) plus rE-cadherin–Fc or rICAM-1–Fc (5 μg/ml) overnight at 4°C. T cells were either left untreated or treated with nocodazole and plated on precoated slides for 30 min at 37°C. Cells were then fixed, permeabilized as described above, and stained with rabbit anti-CCL5 mAb followed by a secondary anti-rabbit–Alexa Fluor 488 Ab to monitor relocalization of the chemokine. Polymerized F-actin was visualized with rhodamine phalloidin (Molecular Probes).

The FluoroSpot assay was adapted from Ref. 26. T cells were plated on plastic slides coated with anti-CD3 mAb for T cell activation and anti-CCL5 mAb (7 μg/ml, R&D Systems) for chemokine capture. After 45 min incubation at 37°C, T cells were washed with PBS Tween 20 (0.05%) and stained with biotinylated anti-CCL5 mAb (144 ng/ml) for 2 h at room temperature, followed by streptavidin–Alexa 488. Slides were mounted and analyzed the following day using the confocal microscope. The areas and mean fluorescence intensities (MFI) of chemokine spots were quantified with ImageJ software (Fiji program).

Data were compared using the unpaired two-tailed Student t test. Two groups were considered significantly different if p < 0.05.

CTL exert their lytic function through exocytosis of perforin- and granzyme-containing CG during interaction with the specific target. CTL possess cytoplasmic stores of CCL5, but whether these vesicles are distinct from CG remains controversial (21, 27). Experiments were therefore performed to assess the intracellular distribution of CCL5V and CG in CTL clone Heu171, unconjugated or conjugated with autologous IGR-Heu tumor cells. Confocal microscopy analysis indicated that the majority of T cells contain CCL5 in cytoplasmic compartments distinct from CG (Fig. 1A). Moreover, both types of granules are rapidly recruited at the contact area with target cells with high degree of colocalization (Fig. 1B, 1C). Similar results were obtained with conjugates between HD1 or HD2 T cell blasts and anti-CD3–coated P815 targets (Supplemental Fig. 1A). Notably, blocking of de novo protein synthesis with cycloheximide had only a marginal effect on CCL5 and granzyme B relocalization in T cells engaged in synapse-like interactions when plated on plastic slides coated with anti-CD3 mAb. Moreover, quantitative RT-PCR analysis did not reveal any significant increase in CCL5 mRNA expression levels at 30- and 60-min T cell stimulation with anti-CD3 mAb (data not shown).

We then followed the kinetics of CCL5 and granzyme B release by CTL activated with anti-CD3 mAb. High amounts of both proteins were initially observed in the T cell clone cytoplasm and rapidly decreased starting 15 min after TCR triggering (Fig. 2A), which correlated with chemokine secretion (Fig. 2B). In contrast, low levels of the two other CCR5 ligands, CCL3 and CCL4, were detected in the T cell cytoplasm and increased progressively after lymphocyte stimulation in the presence of brefeldin A, suggesting de novo synthesis (Fig. 2C). This was further confirmed by confocal microscopy showing accumulation of both chemokines (Fig 2D). Notably, blockage of chemokine secretion with brefeldin A had no effect on CCL5 expression levels (Fig. 2A). Similar results were obtained with HD1 and HD2 CD8⁺ T cell blasts treated with brefeldin A or monensin, another inhibitor of intracellular protein transport (Supplemental Fig. 1B, 1C). These data demonstrate that CCL5 is stored in cytoplasmic compartments of effector CD8⁺ T cells. Despite being distinct from CG, CCL5V relocalized at the IS and were released after TCR triggering.

MTOC polarization has been functionally associated with T cell synaptic secretion of CG (3, 28, 29). To determine whether MTOC polarization is required for CCL5V recruitment at the IS, we treated Heu171 lymphocytes with nocodazole, a microtubule-depolymerizing drug that does not affect IS formation. Nocodazole also had no side effects on T cell viability and activation, as certified by the low percentages of annexin V⁺/propidium iodide⁻ apoptotic cells after lymphocyte treatment with the drug and unaffected Ca²⁺ response to stimulation with anti-CD3 mAb (data not shown). CCL5V remained associated with the MTOC of untreated cells and were transported to the synaptic region close to the T cell membrane (Fig. 3A). In contrast, treatment of the CTL clone with nocodazole inhibited relocalization of CCL5V and CG at the IS with IGR-Heu target cells (Fig. 3A, 3B).

Next, we monitored CCL5V polarization in TCR-engaged lymphocytes plated on coverslips coated with anti-CD3 plus rE-cadherin–Fc, the ligand of integrin CD103 (24). Results revealed recruitment of CCL5V into the central region of the cell surrounded by an actin ring in the proximity of the cell-support contact surface. In contrast, in T cells treated with nocodazole, CCL5V remained dispersed in the cytoplasm (Fig. 3C, 3D). MTOC polarization was also required for directional mobilization of CCL5V at the IS between HD1 or HD2 CD8⁺ T cell blasts and anti-CD3–coated P815 targets (Supplemental Fig. 2A). These data demonstrate that polarization of CCL5V is dependent on MTOC.

Recruitment of CCL5V into the IS raised the question of the polarity of chemokine secretion. We therefore evaluated CCL5 and granzyme B contents in CTL, untreated or pretreated with nocodazole, stimulated with plastic-bound anti-CD3 mAb. Intracytoplasmic staining indicated that both proteins were released from untreated Heu171 CTL, but even though nocodazole completely inhibited release of granzyme B, it had no effect on CCL5 secretion, as revealed by intracytoplasmic immunofluorescence analyses (Fig. 4A) and ELISA (Supplemental Fig. 2B). This result suggests that, contrary to CG, CCL5 release is not restricted to the IS.

We thus assessed CCL5 secretion polarity using a chemokine capture (FluoroSpot) assay adapted from a previously described procedure (26). Results showed that CCL5 accumulates on outlined spots smaller than the average of the T cell contact area, suggesting polarized secretion (Fig. 4B). In contrast, treatment of the Heu171 clone with nocodazole resulted in multifocal localization of the chemokine, indicating multidirectional release. This was further confirmed by quantification of the spot median surface and MFI of CCL5 (Fig. 4C), which were, respectively, 9 μm² ± 1 and 215 ± 4 in untreated cells and reached 47 μm² ± 6 and 70 ± 5 in nocodazole-treated lymphocytes. These data support the conclusion that CCL5 is released at the IS when microtubule integrity is preserved.

We then tested whether, in situations in which MTOC polarization is preserved but CG exocytosis is abrogated, CCL5 focal release will be re-established. For this purpose, we generated CD8⁺ T cell blasts from two patients displaying genetic deficiencies in the hMunc13-4-Rab27a complex, and we tested their ability to secrete CCL5 in a polarized manner. Initial experiments confirmed that, as opposed to HD1 and HD2 T cells stimulated with anti-CD3 mAb, T cells generated from an Rab27a^−/− patient and an hMunc13-4^−/− patient displayed a defect in CD107a surface expression and CG release (Supplemental Fig. 3A, 3B). In contrast, CCL5 was secreted by both mutation-bearing lymphocytes upon TCR triggering (Fig. 4D, Supplemental Fig. 3B). We then analyzed the CCL5 cytoplasmic localization in conjugates between Rab27a^−/− or hMunc13-4^−/− CD8⁺ T cells and anti-CD3–coated P815. Confocal microscopy studies indicated that both CCL5- and granzyme B–containing granules are recruited at the IS (Supplemental Fig. 3C, 3D). Concordant results were obtained with Rab27a^−/− and hMunc13-4^−/− mutation-bearing donor T cells plated on coverslips coated with the anti-CD3 mAb and rICAM-1–Fc molecule (Supplemental Fig. 4A, 4B). To evaluate the relationship between CG exocytosis and the CCL5 secretion pattern, we applied the FluoroSpot assay to T cells from patients. Results indicated that contrary to HD1 and HD2, hMunc13-4^−/− and Rab27a^−/− T cell blasts displayed a multidirectional secretion of CCL5 (Fig. 4B, 4C). This was further confirmed by quantification of median surfaces of the FluoroSpot assays, displaying average sizes of 129 ± 13 μm² and 163 ± 26 μm² for hMunc13-4^−/− and Rab27a^−/− T cells, respectively, and only 31 ± 4 μm² and 29 ± 5 μm² for HD1 and HD2 T cells (Fig. 4C, left panel). CCL5 MFI of the FluoroSpot assays corresponded to 88 ± 5 and 73 ± 5 for hMunc13-4^−/− and Rab27a^−/− T lymphocytes, respectively, and 216 ± 5 and 205 ± 6 for HD1 and HD2 lymphocytes, respectively (Fig. 4C, right panel). These data suggest that, similarly to CG exocytosis, polarized release of CCL5V requires a functional hMunc13-4-Rab27a complex.

We therefore tested whether CCL5V are localized in the same compartment as the Rab27a-hMunc13-4 complex by transfecting HD3 and HD4 CD8⁺ T cell blasts with a plasmid encoding a rRab27a-GFP molecule and staining of CCL5V with specific mAb. Anti-granzyme B mAb were included as controls. Results indicated that the three proteins are located in distinct types of granules (Fig. 5A). However, stimulation of Rab27a-GFP–transfected T cells with anti-CD3 plus rICAM-1–Fc resulted in rapid polarization and likely coalescence of Rab27a and CCL5V at cell–plastic contact zones (Fig. 5B, 5C). These data suggest that CCL5V release at the IS uses the same mechanism as CG exocytosis, resulting in focal chemokine secretion concomitant with target cell lysis.

Next, we investigated the consequence of polarized release of CCL5 on the chemokine receptor expression profile. Heu171 CTL stimulated with anti-CD3 displayed increased surface expression of CXCR4 that was abrogated by nocodazole and re-established with high concentrations of rCCL5 (Fig. 6A). Moreover, induction of CXCR4 by rCCL5 was inhibited by the CCR5 antagonist maraviroc (Fig. 6B). These results suggest that CXCR4 upregulation is dependent on high local levels of CCL5 provided upon synaptic release of the chemokine. Accordingly, Rab27a- and hMunc13-4–deficient CD8⁺ T cells stimulated with anti-CD3 mAb failed to express CXCR4 unless exogenous rCCL5 was added (Fig. 6C, Supplemental Fig. 4C). In contrast, anti–CD3-stimulated HD1 and HD2 T cells displayed a high expression level of CXCR4, which was further upregulated with rCCL5 (Fig. 6C).

We then evaluated the capacity of Heu171 T cells conjugated with IGR-Heu tumor cells to respond to CXCR4 ligand CXCL12 in an in vitro transmigration assay. Previous confocal microscopy analysis indicated that, in contrast to CCR5, which is engaged at the IS, CXCR4 is uniformly expressed on T cells but is excluded from the IS with target cells (Fig. 7A). Results indicated that CTL engaged with either autologous (Fig. 7B) or peptide-pulsed Gerl-A2 (Fig. 7C) target cells are able to detach and migrate in response to CXCL12 more efficiently than are unconjugated T cells. Moreover, anti-CXCR4 neutralizing mAb (Fig. 7B) or CCR5 antagonist (Fig. 7D, Supplemental Fig. 4D) inhibited lymphocyte migration. Notably, only a marginal migration of unconjugated or conjugated targets was observed (data not shown). With regard to Rab27a^−/− or hMunc13-4^−/− T cell blasts, conjugated with Gerl-A2 unpulsed or pulsed with SEB superantigen and exposed to CXCL12, they migrated less efficiently than HD1 T cells conjugated with the same target (Fig. 7E). Similar results were obtained with T cells conjugated with anti-CD3 plus rICAM-1–Fc–coated beads (Supplemental Fig. 4E). These data strongly suggest that synaptic release of CCL5V triggers T cell surface expression of CXCR4, which regulates IS duration by controlling T cell detachment and migration toward a novel CXCL12-producing target.

In the current study, we demonstrate that activated CTL secrete CCL5 into the IS with target cells, resulting in autocrine induction of CXCR4. This function leads to enhanced responsiveness of T cells to CXCL12 and represents a mechanism regulating the duration of the T cell–target cell interaction during effective killing. TCR-mediated activation of CTL triggers secretion of effector cytokines and chemokines, including CCR5 ligands CCL3, CCL4, and CCL5. In agreement with previous studies (21), we show that CCL3 and CCL4 are synthesized and promptly secreted after effector CD8⁺ T cell stimulation. In contrast, CCL5 is stored in post-Golgi vesicles and released after TCR triggering by a secretory pathway typical of CG exocytosis (30). Indeed, our results indicate that, as with CG, CCL5 secretion occurs minutes after T cell activation and is not affected by Golgi-dependent protein transport inhibitors. Our results are also in good agreement with previous studies showing that CCL5V are distinct from granzyme B–containing granules (21), but both are released with similar kinetics. Notably, our results revealed that CCL5V and CG are localized close to each other even when away from the IS. Although the functional relevance of this finding is not clear, this tethering of CCL5V and CG may facilitate synaptic release of the chemokine. A Vti1b-dependent tethering of CG and CD3-containing endosomes has been previously reported to determine efficient lytic granule secretion at the IS (31). Munc13-4–mediated cooperation of CG and Rab11⁺ endosomal “exocytic vesicles” has also been reported to be required for the cytotoxic function of CTL (6). Our data suggest a functional relationship between cytotoxicity and CCL5 secretion, most likely for regulating CTL effector functions rapidly after recognition and killing of specific targets. In contrast, protracted secretion of CCL3 and CCL4 may further shape the tumor microenvironment by recruiting multiple effectors to sites of lymphocyte activation (32).

Synaptic secretion of some cytokines is though to provide a mechanism for Ag specificity of effector T cell functions (33). It has been reported that CD4⁺ T cells secrete IFNγ and IL-2 through the IS, whereas CCL3 and CCL5 are secreted multidirectionally (26). In contrast, CD8⁺ T cells contain CCL5V that polarize rapidly in response to TCR engagement (21). This discrepancy might be due to the analysis in the first study (26) of newly synthesized CCL5 in helper T cells, whereas in the latter (21) and present studies we focused on CCL5 present in CTL cytoplasmic vesicles. We show in this article that, similarly to CG, CCL5V are rapidly recruited at the IS and require T cell MTOC polarization. Indeed, CCL5 was enriched around the MTOC after TCR triggering, whereas disorganization of the microtubule network altered polarization of both CCL5V and CG. Remarkably, whereas T cell microtubule depolymerization abrogated CG exocytosis, it did not inhibit CCL5 secretion. However, instead of being released into the IS, CCL5 was secreted multidirectionally. The ability of T cells to use both multidirectional and synaptic pathways to secrete CCL5 may reflect different context-dependent functions. For instance, multidirectional secretion may permit recruiting effector cells to sites where Ag was recognized but CG exocytosis had not yet been achieved whereas synaptic secretion may act locally upon effective target cell killing.

Rab27a and hMunc13-4 proteins interact directly and are essential effectors of CG exocytosis (6, 34, 35). Although defects in either of the two proteins do not affect CG polarization, they dramatically impair exocytosis-dependent target cell lysis (4, 36). Likewise, we show that loss of functions of Rab27a or hMunc13-4 does not affect CCL5V polarization but inhibits synaptic release of the chemokine. We also show that Rab27a and hMunc13-4 endosomal exocytic vesicles are likely distinct from CCL5V, and that both types of granules strongly colocalize and potentially fuse at the IS in a manner similar to that with CG. Moreover, our results indicate that Rab27a and hMunc13-4 deficiencies lead to multifocal secretion of CCL5, most likely in a piecemeal manner by shuttling the chemokine, through a pool of smaller secretory vesicles, from CCL5V throughout the T cell surface (37, 38). These data suggest that CCL5V and CG use the same Rab27a-hMunc13-4–dependent mechanism for their synaptic release, resulting in a high local concentration of the chemokine, concomitant with target cell killing.

The duration of adhesive contacts between CTL and target cells determines the outcome of T cell responses and is critical for eliminating all malignant cells. Tumor-infiltrating CD8⁺ T lymphocytes display random migration and a transient Ag-independent interaction with transformed cells until they encounter specific pMHC-I–expressing targets to arrest their movements and establish stable contacts (39). Subsequent binding of integrin LFA-1 or CD103 on T cells to their respective ligands ICAM-1 or E-cadherin on tumor cells results in maturation of the cIS and promotes polarized exocytosis of CG leading to target cell death (24, 40, 41). Once transformed cells are cleared, CTL resume motility to further search for new targets (42). In this report, we show that polarized release of CCL5V at the IS induces expression of T cell surface–distributed CXCR4, promoting CTL detachment from target cells and migration in response to CXCL12. CXCR4 induction is dramatically compromised by microtubule cytoskeleton disorganization and loss of Rab27a or hMunc13-4 function, and is re-established in the presence of high concentrations of exogenous rCCL5. Moreover, neutralizing CCR5, which is relocalized at the IS with target cells, inhibited expression of CXCR4 on T cell surface. These results suggest that only synaptic release of CCL5V, most likely through increased local concentration of the chemokine, and then binding to CCR5 at the IS induces CXCR4 expression.

Crosstalk between CCR5 ligands and CXCL12 provides a mechanism that underlies the movement of hematopoietic cells (43). CXCL12, produced in lymphoid organs and in many tumors, induces T lymphocyte migration that is arrested upon TCR engagement (44). In this article, we show that synaptic release of CCL5V triggers CXCR4 expression in an autocrine manner, thereby promoting CD8⁺ T cell migration in response to CXCL12. It is tempting to speculate that successful engagement of a target cell results in synaptic release of CCL5 that, through triggering of polarized CCR5 (14), induces CXCR4 surface expression and thus promotes T cell detachment in a CXCL12-rich tumor microenvironment. These data are compatible with a model of CTL serial conjugation and killing of multiple targets (45) and suggest that CTL can regulate cIS duration through rapid synaptic release of CCL5V concomitant with exocytosis of CG.

The concept of serial killing has been primarily based on in vitro studies (45). Imaging of IS between live CTL and specific target cells revealed the accumulation of CG within minutes after initial cell–cell contact and rapid target cell death (2). Studies on intact lymph nodes have also shown that CTL-mediated killing of specific APC and subsequent T cell detachment requires <25 min (46). In contrast, two-photon imaging of murine tumors revealed a long-lasting interaction between CTL and target cells (47). One possible explanation for these unexpected results, apart from the impact of the tumor microenvironment, as suggested by the authors, is abnormal nonphysiological overstimulation of CTL. The present study relies on a cancer patient–derived tumor cell–specific CTL couple model. We had also defined a threshold concentration of the antigenic peptide above which, when presented by the MHC-I–overexpressing Gerl-A2 APC, CXCL12 was no longer able to override the TCR-mediated stop signal. Importantly, this threshold was never reached with autologous tumor cells, which express low levels of specific pMHC-I complexes. Further analyses of more relevant human tumors or xenografts expressing physiological tumor Ag, and that are infiltrated by specific effector T cells often bearing low-affinity TCR, are required to shed more light on the in vivo dynamics of malignant cell destruction by CTL and the role of synaptic release of CCL5.

In conclusion, our data reveal a mechanism of control of cIS duration through synaptic secretion of CCL5. A better understanding of the dynamics of antitumor CTL functions may offer new opportunities for optimizing the design of cancer immunotherapy strategies.

We thank Sophie Salomé-Desmoulez and Emmanuel Bourrin for help with confocal microscopy.

The authors have no financial conflicts of interest.