Transfer of NKG2D and MICB at the cytotoxic NK cell immune synapse correlates with a reduction in NK cell cytotoxic function

Abstract

Although transfer of membrane proteins has been shown to occur during immune cell interactions, the functional significance of this process is not well understood. Here we describe the intercellular transfer of NKG2D and MHC class I chain-related molecule (MIC) B proteins at the cytotoxic natural killer cell immune synapse (cNK-IS). MICB expressed on the 721.221 cell line induced clustering of NKG2D at the central supramolecular activation cluster, surrounded by a peripheral supramolecular activation cluster containing F-actin. Moreover, natural killer (NK) cell membrane-connective structures formed during cytotoxic interactions contained F-actin, perforin, and NKG2D. NKG2D transfer depended on binding to MICB and was specific because transfer of other molecules not involved in NK-IS formation was not observed. Transfer of MICB to NK cells also was noted, suggesting a bidirectional exchange of receptor/ligand pairs at cNK-IS. Experiments designed to test the functional significance of these observations revealed that brief interactions between NK cells and MICB expressing target cells led to a reduction in NKG2D-dependent NK cytotoxicity. These data demonstrate interchange of an activating receptor and its ligand at the cNK-IS and document a correlation between synapse organization, intercellular protein transfer, and compromised NK cell function after interaction with a susceptible target cell.

NKG2D, a member of the killer cell lectin-like receptor family, is one of the best characterized natural killer (NK) cell-activating receptors, but it also is expressed on TCRγδ⁺ and CD8⁺ TCRαβ⁺ T cells, where it can act as a costimulatory molecule for T cell activation. NKG2D associates with DAP10, a transmembrane adaptor molecule containing a YINM motif that binds and activates via phosphatidylinositol 3-kinase and Grb2 (1, 2). In humans, NKG2D-immune activation can be triggered by interaction with its ligands: the polymorphic MHC class I chain-related molecule (MIC) A and MICB and the UL16-binding proteins (3). NKG2D ligands are absent, or expressed only at low levels, on normal cells, but their expression is often enhanced/induced on tumors and cells infected by various pathogens. Dysregulation of NKG2D ligand expression with inappropriate activation of NKG2D⁺ T cells also has been reported recently to be a feature of various autoimmune diseases (4).

Receptor–ligand interactions between cells do not occur randomly, rather the groups of interacting molecules are found in ordered structures, named immune synapses (IS), providing a platform for intercellular communication among cells of the immune system (5). Upon NK cell–target cell interaction, highly organized supramolecular clusters are formed at the contact area of both cells, termed the NK cell IS (NK-IS) (6, 7). When NK cells interact with susceptible target cells, the redistribution of signaling molecules at the central (c) SMAC (supramolecular activation cluster, SMAC), as well as LFA-1 and talin at the peripheral SMAC, resembles the T cell IS (8, 9). However, the mechanisms by which activating receptors redistribute to the NK/target cell interface remain ill understood. Only polarization of CD2 and 2B4 to the contact area of NK cell–target cell conjugates has been studied in detail (10, 11), although NKG2D has been observed to accumulate at the contact site between NK cells and MIC-expressing colorectal tumors in vivo (12).

One consequence of synapse formation is transfer of molecules from antigen-presenting cells to T or B cells (13, 14). For NK cells, CD96-mediated acquisition of poliovirus receptor has been documented (15), and bidirectional transfer of human and murine NK cell inhibitory receptors and MHC class I molecules has been reported (16–18). Recently, the transfer of 2B4-GFP from NK cell membrane connective structures (MCS) to susceptible target cells also has been observed, leading to the hypothesis that cytotoxic NK/target cell interactions might result in a synaptic transfer of receptors to target cells (11). However, little is known regarding the functional consequences of protein transfer.

In an attempt to gain additional insights into the organization of activating receptors at the cytotoxic NK cell IS (cNK-IS), we have analyzed the distribution of NKG2D in conjugates formed between NK cells and the susceptible human B cell line 721.221 stably transfected with the NKG2D ligand MICB. During these experiments we observed that brief incubation of NK cells with MICB-expressing targets rapidly induces both clustering of receptor and ligand and interchange of NKG2D and MICB proteins. Interestingly, these processes are associated with a reduction in the cytotoxic capacity of NK cells in subsequent encounters with MICB-expressing target cells.

Results

Distribution of NKG2D at cNK-IS.

To study the behavior of NKG2D during specific NK recognition of susceptible target cells, 721.221 cells (referred to as 221) were transfected with MICB (referred to as 221B) because they do not express any known ligand of NKG2D (ref. 19; Fig. 7A, which is published as supporting information on the PNAS web site). We used the NKL cell line (NCR-negative, NKG2D-positive; Fig. 7B) as the effector cell in cytotoxic assays with 221 or 221B cells as targets. 221B cells were much more susceptible to lysis by NKL than 221 cells and blocking experiments with an NKG2D-specific mAb showed that this enhanced lysis depended on NKG2D (Fig. 7C). Immunofluorescence experiments revealed that NKG2D clustered at the NKL/221B contact site in ≈60% of conjugates analyzed, but clusters were almost undetectable at NKL/221 interfaces (Fig. 1A; see also Fig. 8A, which is published as supporting information on the PNAS web site). Thus, there is an association between cytotoxicity triggered by NKG2D and clustering of this receptor at the cNK-IS.

Fig. 1.

MICB-induced clustering and cSMAC segregation of NKG2D at cNK-IS. (A) NKL/221 and NKL/221B conjugates were stained for NKG2D. Fluorescence and bright-field images (a) and quantification (%) of cell conjugates in which NKG2D was localized at the contact site (b) are shown. One representative experiment of at least three independent ones is shown. (B–D) Conjugates formed with NKL (B) or primary-activated NK cells (C and D) and 221 or 221B cells were costained for NKG2D (red) and F-actin (green). (C) NKG2D distribution patterns shown in NK/221B or NK/221 interface are representative of 11 conjugates analyzed over three independent donors. (D) Four confocal sections of two representative NK/221B conjugates showing intracellular NKG2D-containing vesicles (arrows). In all experiments, target cells were identified by their bigger size or by labeling with CMAC (blue) in experiments done with the NKL cell line. (Scale bars: 5 μm.)

To further characterize the organization of the NK-IS formed in NKL/221B conjugates, cells were costained for NKG2D and F-actin and analyzed by confocal microscopy. NKG2D clustered at the cSMAC surrounded by F-actin at the peripheral SMAC. In contrast, both molecules codistributed through the synapse in NKL/221 conjugates (Figs. 1B and 8B, see Movies 1 and 2, which are published as supporting information on the PNAS web site). A similar NKG2D segregation pattern at the synapse to that observed with NKL cells was found when conjugates formed between primary human activated NK cells and 221 or 221B cells were studied (Fig. 1C). CD56 did not cluster at the site of contact between NK and target cells, but perforin polarized to the contact site in agreement with observation of mature synapses (ref. 10; Fig. 8C). Remarkably, polarization to NK-IS of intracellular NKG2D-containing vesicles, some of them colocalizing with perforin, also was observed (Fig. 8D). Moreover, NKG2D-containing vesicles at both sides of synaptic F-actin-rich structures were observed in confocal planes obtained from NK/221B conjugates (Fig. 1D).

Together, these data show that target cells expressing MICB induce clustering and segregation of NKG2D at the cSMAC of the cNK-IS. The data further suggest that at least some NKG2D traffics to the NK-IS in intracellular vesicles.

Transfer of NKG2D to MICB-Expressing Target Cells.

Another feature seen during the inspection of NKL/221B and NK/221B conjugates was punctuated NKG2D staining on the surface of target cells (Fig. 2 A and B) at the cNK-IS (Fig. 2B, arrow), and at the edge of NK cell F-actin containing lamelipodia spread over the target cells (Fig. 2C; see Movie 3, which is published as supporting information on the PNAS web site). NKG2D colocalized with the adaptor molecule DAP10, not only at the cNK-IS, but also on the target cell surface at sites distant from the synapse (Fig. 2D; see Movie 4, which is published as supporting information on the PNAS web site), which suggested transfer of the activating NKG2D/DAP10 receptor complex to target cells. Although NKG2D clusters could be observed clearly on the surface of isolated 221B cells in NK/221B samples, anti-NKG2D mAb did not stain 221B cells plated alone (Fig. 2E). Although NKG2D clusters were found in >80% of NK/221B conjugates, only ≈10% of NK/221 ones contained clusters that were weaker in fluorescence intensity. Clusters corresponding to the coactivating receptor 2B4, whose ligand CD48 is expressed on both 221 and 221B cells, were frequently found in both types of conjugates (Fig. 2F). Finally, a clear-cut colocalization of membrane NKG2D and 2B4 spots could be found on interacting NK-221B cells (Fig. 2G).

Fig. 2.

NKG2D transfer to 221B cells. (A–C) NKL/221B (A) or NK/221B (B and C) cell conjugates were costained for NKG2D (red) and F-actin (green). A confocal section (merged channels in A and B) shows NKG2D clusters on target cells (see arrowheads). (B) Three-dimensional max projection shows the synapse face of the NK cell pointed by an arrow. (C) Zoom of merged channels shows NKG2D clusters at the edge of F-actin rich lamelipodia. (D) NK/221B conjugates costained for NKG2D (red) and DAP10 (green). One confocal section and the 3D max projection are shown. An arrowhead points to an NKG2D/DAP10-containing cluster distant to the synapse face. (E) Confocal microscopy analysis of NK/221B samples shows NKG2D-containing clusters transferred to an isolated 221B target cell (Left, arrow). As a negative control, NKG2D staining obtained in 221B plated alone on coverslips is shown (Right). (Scale bars: Left, 5 μm; Right, 20 μm.) (F) Quantification of target cells containing NKG2D or 2B4 membrane clusters in NK/221 or NK/221B conjugates. Histogram bars represent the arithmetic mean ± SD (n = 100). (G) NK/221B conjugates were costained for 2B4 (green) and NKG2D (red). Arrowheads point to colocalization of NKG2D and 2B4 on interacting cells. One confocal section and the differential interference contrast image are shown. 221B cells were labeled with CMAC (blue) in experiments done with NKL cells. (Scale bars: 5 μm except E Right.)

Interestingly, NKG2D clusters were found specifically in NK/221B conjugates communicated by MCS but not in NK/221 ones. These clusters are observed at MCS and transferred to 221B target cells (Fig. 3A, arrowheads and arrows; see Fig. 9A, which is published as supporting information on the PNAS web site). In contrast, 2B4 clusters of both types of conjugates were found at MCS, consistent with previous data (ref. 11; Fig. 9A). Similarly, DAP10-GFP transfected in NKL cells also was found at MCS and on 221B cells (Fig. 9B).

Fig. 3.

MCS formed in NK/221B conjugates contain NKG2D. (A) The indicated cell conjugates were stained for NKG2D. Confocal microscopy analysis shows NKG2D containing MCS (arrowheads) and clusters transferred to 221B target cells but not to 221 ones (arrows). (B–D) NK/221B conjugates were costained for NKG2D (red)/F-actin (green) (B), perforin (red)/F-actin (green) (C), and perforin (green)/NKG2D (red) (D). (B) Arrowheads and yellow arrows inside the zoom of merged channels indicate NKG2D spots and overlapping NKG2D and F-actin, respectively. Two arrows point to the MCS in the differential interference contrast image. (C) A magnified view of the area of the MCS is shown. (D) A yellow arrow points to the area magnified that shows the partial overlapping between NKG2D and perforin. Confocal sections and differential interference contrast images are shown. (Scale bars: 5 μm.)

We further characterized these MCS by staining conjugates with F-actin. MCS contained F-actin up to the distal zone where NKG2D spots were concentrated (Figs. 3B and 9C, see also Movie 5, which is published as supporting information on the PNAS web site). Large-diameter MCS containing perforin inside peripheral F-actin also were observed (Fig. 3C). Strikingly, the NKG2D staining partially overlapped with staining for F-actin and perforin in these MCS (Fig. 3 B and D). Finally, MCS communicating NK cells and target cells also were tracked by live-cell time-lapse confocal microscopy (Fig. 10A and Movies 6 and 7, which are published as supporting information on the PNAS web site). Remarkably, movement of bulges was observed in these stable structures (lasting for >40 min), suggesting that there is traffic of organelles between cells. Staining of fixed samples for NKG2D suggested the presence of the receptor within bulges in these structures (Fig. 10B, arrows).

Together all these data strongly suggested the specific transfer of NKG2D to 221B cells after synapse formation and at the end of MCS communicating NK cells and targets.

Specific Transfer of NKG2D to MICB-Expressing Targets Depends on Ligand Binding.

To further analyze the transfer of NKG2D to 221B targets, mixtures of NK and 221 or 221B cells were prepared, stained, and analyzed by FACS. 221 and 221B cells were gated in forward scatter/side scatter dot plots (data not shown) and NK cells present in these gates were excluded by MHC class I staining. The 221B population was detected further by positive staining with MICB (data not shown). After incubation at 37°C, a clear staining for NKG2D was observed in 221B cells but not in 221 ones, or in 221B cells mixed with NK cells and then directly separated at 4°C (Fig. 4A). Staining of CD56, a molecule not involved in synapse formation, was not specifically detected on 221 targets in any case studied (Fig. 4B). NKG2D transfer to 221B cells was completely abolished by preincubation of NK cells with a blocking anti-NKG2D mAb, whereas a negative control anti-CD56 mAb did not have this effect (Fig. 4C). Similar results were obtained in experiments realized with the NKL cell line (data not shown). Together, these data demonstrate the rapid and specific transfer of NKG2D to 221B cells after binding to MICB expressed on target cells.

Fig. 4.

NKG2D transfer to 221B cells depends on MICB binding. (A) Shown are NKG2D (phycoerythrin) and MHC class I (FITC) double staining of cell mixtures containing NK and 221B cells at 4°C (Left) and NK cells previously incubated with 221 or 221B at 37°C (Center and Right). Numbers within plots indicate the FL2 geomean fluorescence of the 221 or 221B gated population. (B) Quantification of the NKG2D transfer in NK/221 or NK/221B conjugates. Histogram bars represent the arithmetic mean ± SD of NKG2D/negative staining geomean ratio obtained in five experiments done with five different donors. Statistical significance of data was analyzed by the Mann–Whitney U test. (C) Abrogation of NKG2D transfer by anti-NKG2D blocking mAb in NK/221B conjugates. As a negative control, an anti-CD56 mAb was used. Concentrations of the antibodies are indicated. One representative experiment of at least three independent ones is shown. (D) Down-modulation of NKG2D surface expression in NK cells exposed to 221B cells. Means ± SD of three independent experiments are compared by the Student t test. P > 0.05 was considered nonsignificant (ns). (E) Histograms of one representative experiment showing the surface expression of NKG2D in NK cells before (a) and after an encounter with either 221 (b) or 221B (c) cells. Numbers indicate the geomean fluorescence of NKG2D staining. Thin lines represent the negative staining. (Ed) NKG2D histograms of a (thin lane) and c (thick lane) are overlays.

The percentage of NK cell NKG2D transferred to 221B in experiments with five donors analyzed in Fig. 4B was 16, 4, 11, 3.5, and 3.6 (average 8%) (calculated as described in Supporting Text, which is published as supporting information on the PNAS web site). NKG2D expression on the NK cell surface after a 221 or 221B encounter also was analyzed. An ≈20% decrease in surface NKG2D was found in NK cells that had been exposed to 221B cells compared with those that had seen only 221 cells (Fig. 4 D and E). Although these data are consistent with NKG2D transfer to 221B targets, in general, protein transfer does not explain all of the NKG2D down-modulation observed.

Transfer of MICB to NK Cells.

We also studied the distribution of MICB at the cNK-IS. Costaining for F-actin and MICB revealed clustering of this NKG2D ligand at the cNK-IS and on the side of NK cell MCS (Fig. 5A). MICB also was found in MCS, on the surface of interacting NK cells and target cells (Fig. 5B), and in isolated NK cells (Fig. 5C). Interestingly, colocalization of MICB and NKG2D on the NK cell surface could be observed (Fig. 5 B and C, see z sections). Importantly, primary NK cells plated alone were not stained with anti-MICB mAb (Fig. 5D). MICB transfer also was detected by flow cytometry (Fig. 5E) and in experiments done with 221 cells transfected with a MICB-GFP chimaeric protein, where colocalization of NKG2D and transferred MICB at the NK cell surface was observed (Fig. 5F, arrowheads and zoom). Approximately 5% of target cell MICB-GFP is transferred to NK cells as a consequence of the cytotoxic interaction (Fig. 11, which is published as supporting information on the PNAS web site).

Fig. 5.

MICB transfers to NK cells. (A–C) The indicated samples were costained for MICB in red and F-actin (A) or NKG2D (B and C) in green. Arrowheads point to MICB at the contact interface (A) and at MCS and interacting NK cells (B). Yellow arrows point to NKG2D/MICB colocalization on NK cells (B and C). (C) Cross-sections (zx and zy) show the colocalization between NKG2D and MICB on the cell surface of an isolated NK cell. (D) Negative MICB staining on activated NK cells plated alone. (A–D) One confocal section and differential interference contrast images are shown. (E Left) A representative example of MICB staining on NK cells previously incubated with CSFE-labeled 221 or 221B cells is shown (Materials and Methods). (E Right) The quantification of the MICB transfer to NK cells in NK/221 or NK/221B cell mixtures is shown. Histogram bars represent the arithmetic mean ± SD of MICB geomean fluorescence obtained from three independent experiments done with three different donors. NK/221 (MICB-GFP) samples were stained for NKG2D (red). A confocal plane and 3D max projection are shown. Arrowheads point to NKG2D/MICB-GFP colocalization on NK cell surface. (Scale bars: A–C, E, and F, 5 μm; D, 10 μm.)

Previous Encounter with MICB-Expressing Cells Provokes a Reduction of NKL Cell Line Cytotoxic Activity.

The data presented above showed that after NK-221B interaction, a synaptic interchange of NKG2D/MICB pair and a reduction in NKG2D surface levels occurred. We therefore tested whether a first interaction with 221B had any effect on NK cytotoxic activity on a subsequent encounter with 221B cells. These experiments were carried out with the NKL cell line as efficient lysis of 221B cells by this effector depends principally on NKG2D (Fig. 7). N1 or NB effectors were obtained by mixing NKL cells with 221 or 221B cells at 37°C during 5 min (ratio 1:1), purifying the NKL cells and using them in cytotoxic assays against ⁵¹Cr-labeled 221B cells. As expected, NKL cells killed 221B cells more efficiently than 221 ones (lysis triggered by NKG2D). A specific reduction of this NKG2D-dependent cytotoxic activity was found in NB effector cells but not in N1 ones (Fig. 6A). Importantly, perforin levels were similar in NKL, N1, and NB effector cells, as assessed by FACS (Fig. 6B).

Fig. 6.

MICB expressed on 721.221 cells reduce NKL cytotoxic function. (A) Cytotoxic activity, against ⁵¹Cr-labeled 221B cells, of NKL cells previously incubated with 221 or 221B cells at 37°C for 5 min (E:T ratio 1:1) was evaluated (N1/221B and NB/221B) and compared with the cytotoxic activity of NKL (not incubated with any target) against 221 and 221B cells (NKL/221 and NKL/221B). Cytotoxic assays were done at an E:T ratio of 5:1. Histogram bars represent the arithmetic mean ± SD of three independent experiments done in triplicate. Means were compared by the Student t test. (B) FACS analysis of perforin expression in effector cells. Numbers in histogram plots indicate the geomean fluorescence of the positive population under the line. Dotted lines represent the isotype control staining.

Discussion

In the in vitro model used in this work, we show ligand-induced clustering and transfer of NKG2D from NK cells to target cells in the context of a cNK-IS. Transfer of MICB from targets to NK cells also is observed in this synapse. Finally, our data show that bidirectional synaptic transfer of NKG2D and MICB is a rapid process that occurs at the same time as a marked reduction in the capacity of the NK cells for NKG2D-dependent cytotoxicity.

Little is known regarding the organization of NK cell-activating receptors at cNK-IS (10–12). Segregation of NKG2D to the cSMAC surrounded by peripheral F-actin at the interface of NKL/221B conjugates supports the same observation obtained when we analyze primary-activated NK cells. These data are consistent with the segregation of NKG2D ligands at the central region of ICAM1 rings in antigen-independent junctions formed between CTLs and planar lipid bilayers (20, 21), and with antigen-independent recruitment of NKG2D to the γδ T cell IS (22). Thus, it seems that NKG2D is organized similarly at both the NK and T cell IS.

Polarization to the synapse of intracellular TCR in transit through recycling endosomes has been described during T cell activation (23). Strikingly, NKG2D also appears to polarize to the cNK-IS in intracellular vesicles. Colocalization of perforin and NKG2D in vesicular structures suggests that NKG2D could be sorted to secretory lysosomes as occurs with other molecules involved in NK cell killing (24). However, the observation of vesicles containing NKG2D but not perforin suggests intracellular traffic of the receptor through other compartments also. Thus, NKG2D may traffic to the synapse both by lateral diffusion in the membrane and via intracellular routes.

Multiple previous studies have shown intercellular transfer of surface proteins from the target to the effector cell: MHC molecules and membrane fragments transfer from antigen-presenting cells to T cells, and this process can lead to TCR internalization and fratricidal killing (25). Acquisition of antigen by B cells from targets leads to enhanced presentation of these antigens to T cells (14). For NK cells, receptor-dependent acquisition of MHC class I and transfer of inhibitory receptors (Ly-49 and KIR) in an inhibitory NK-IS context has been reported in refs. 16–18. The functional consequences of bidirectional KIR/MHC transfer have not been studied.

NKG2D/MICB bidirectional transfer in the cNK-IS reported here resembles the KIR/MHC behavior at the inhibitory synapse (18), because clustering and transfer of NKG2D and MICB are general population phenomena that occur rapidly because they can be detected after only 5 min of cell interaction. Moreover, the NKG2D transfer depends on the binding of MICB expressed on target cells. Although transfer of NKG2D to target cells correlates with the cytotoxic process, whether it is involved in this process or a consequence is an interesting question.

The mechanisms of protein transfer at the synapse are not clear, but observations such as DAP10/NKG2D colocalization on the target cell and transfer of DAP10 and MICB tagged at the C terminus with GFP, argue indirectly against transfer of proteolytic extracellular fragments of the receptor or the ligand and are consistent with previous data showing transfer of 2B4 also tagged at the C terminus with GFP (11). Perhaps diffusion in membrane bridges similar to those that form in CTL cytotoxic interactions (9) could be one way these molecules transfer between cells.

After cNK-IS disassembly, the formation of MCS (11, 26) (elsewhere termed nanotubes; ref. 27) has been described, and transfer of 2B4-GFP from these MCS to target cells has been observed (11). Our data strongly suggest that NKG2D also is transferred to the target cell surface both in stable conjugates (in which a mature synapse could be seen) and at the distal zone of MCS communicating NK cells and target cells. These data and those described in ref. 26 indicate that activating receptors are clustered at cNK-IS and left on target cells after synapse disassembly and formation of MCS/nanotubes. This hypothesis is supported by the colocalization of 2B4 and NKG2D on the surface of interacting NK cells and target cells suggesting that both molecules could be transferred together. MCS also may provide a mechanism for intercellular transfer of NKG2D, as has been suggested for other cell-surface proteins (27). We have observed that MCS contain F-actin, as described for nanotubular highways for intercellular organelle transport (28). Moreover, NKG2D spots partially overlapping F-actin and perforin also are observed within these tubular structures and specifically detected at bulges formed in MCS. Finally, transport of organelles between NK and target cells is supported by the movement of these bulges inside MCS communicating NK cells and target cells migrating on fibronectin. These observations support the traffic of vesicles and/or surface proteins between NK cells and target cells connected by MCS/nanotubes, and they represent evidence of intracellular transport of material via these structures during cytotoxic interactions.

An important issue to study is the effect that synapse organization and bidirectional transfer of membrane proteins between NK and target cells has on NK cell function. In our experiments, a prior encounter with MICB-expressing target cells provokes a marked reduction in NKG2D-mediated NKL cytotoxicity. Interactions of only 5 min result in a decreased NKG2D surface expression and a 40–50% reduction of cytotoxic activity, indicating that exposure to NKG2D ligands rapidly can cause a reduction in the cytotoxic function of human NK cells.

Previously, loss of NK cell cytotoxic activity has been described as a phenomenon happening after NK/susceptible target cell interactions of several hours (29). Recently, prolonged (24 h) exposure to NKG2D ligand in vitro has been shown to impair cytotoxicity via NKG2D, and this phenomenon correlated with complete loss of the adaptor molecule DAP10 (30). In this system, killing via a different activating receptor, Ly-49D, was conserved substantially. In contrast, chronic exposure of murine NK cells to NKG2D ligands in vivo, by expression of Rae-1ε and MICA transgenes, resulted in down-regulation of surface NKG2D and a global reduction in NK cell function (31, 32). In our model, lysis by NKL depends principally on NKG2D/NKG2D ligand interactions; thus it is not adequate to determine with confidence whether the loss of function observed is specific for NKG2D or is a more general effect. Importantly, perforin levels in NB cells are not decreased in comparison with NKL and N1 cells, which suggest that NK cells retain all of their cytotoxic machinery after these brief interactions and excludes exhaustion of cytolytic granules as an explanation for the compromised cytotoxicity via NKG2D.

It seems likely that in our experiments the reduction in cell surface NKG2D is one element of the compromised NK function. Quantitative analyses indicate that both transfer of NK cell NKG2D to the target cell and internalization of NKG2D contribute to receptor down-modulation after interaction. These data are consistent with the observation that binding to MIC can induce endocytosis of NKG2D and loss of cytotoxic activity (33). Our data also show that synaptic NK cell acquisition of MICB also correlates with the reduced cytotoxicity and that the transferred MICB (or MICB-GFP) colocalizes with NKG2D on the NK cell surface, perhaps “blocking” NKG2D, and, thus, further reducing the amount of available receptor. Acquisition of MICB by NK cells also might serve to dampen immune responses by provoking fratricidal interactions, as has been observed for activated T cells (25).

In any event, the observation that exposure to NKG2D ligands, although leading to lysis of some target cells, also can lead to reduction in NK cytotoxic capacity perhaps explains why many tumors maintain expression of NKG2D ligands like MIC. This phenomenon may represent a previously undescribed mechanism for evasion of the NK cell immune response and favor tumor growth.

In summary, these observations contribute insights into the distribution of NK cell-activating receptors during assembly and after disassembly of cNK-IS. The consequences that synapse organization and/or the bidirectional transfer of receptor/ligand pairs may have on NK cell immune surveillance in physiological and pathological situations are important questions that deserve further investigation.

Materials and Methods

Cells, Antibodies, and Reagents.

Primary polyclonal NK cells were grown as described in ref. 34. Pure NK cell populations (>97% CD56⁺ CD3⁻) were obtained by negative selection (Miltenyi Biotec, Germany). The 721.221 cell line (American Type Culture Collection) (referred to as 221) was transfected with MICB (referred to as 221B) or MICB-GFP by electroporation (250V and 500 μF) in a Bio-Rad gene pulser II (Supporting Text). NKL was transfected with DAP10-GFP via retroviral infection (Supporting Text). Anti-NKG2D and -MICA/B specific mAbs were obtained from R & D Systems. Other Abs and reagents are listed in Supporting Text.

Conjugate Formation and Microscopy Experiments.

Conjugates of NK cells and 221 or 221B cells (formed by centrifugation at 100 × g for 1 min) were incubated for 5 min at 37°C, fixed, stained, and analyzed by confocal microscopy (11). CMAC labeling was done by following the manufacturers instructions (Molecular Probes). Time-lapse experiments also were performed and analyzed as described in ref. 11 by using the Leica DM IRE2 confocal microscope. A detailed description of these protocols is placed in Supporting Text.

FACS Analysis.

Conjugates were prepared as described above at an E:T cell ratio of 6:1, incubated for 5 min at 37°C, and disrupted by washing with cold PBS/0.5 mM EDTA. As a negative control, conjugates were washed in cold PBS/0.5 mM EDTA immediately after centrifugation. Cell mixtures then were stained for NKG2D and MHC class I and analyzed by FACS (Supporting Text). Quantification of NKG2D transfer is described in Supporting Text. Analysis of NKG2D and MICB on NK cells after 221 or 221B encounters and the quantitation of MICB-GFP transfer is detailed in Supporting Text. Ca-AM and CSFE labeling were done by following Molecular Probes’ instructions.

Cytotoxicity Assays.

Four-hour ⁵¹Cr release cytotoxic assays were carried out at the indicated E:T ratios and analyzed as described in ref. 34. To study the cytotoxic activity after exposure to 221 or 221B cells, NKL cells were incubated for 5 min at 37°C with Ca-AM-labeled 221 or 221B cells, separated with cold PBS/0.5 mM EDTA, purified by FACS sorting at low pressure, and used as effector cells (N1 or NB effectors, respectively) in assays (see also Supporting Text). Viability of NKL, N1, and NB effector cells was tested by trypan blue exclusion.

Abbreviations

cNK-IS

cytotoxic natural killer cell immune synapse

IS

immune synapse

MCS

membrane connective structures

MIC

MHC class I chain-related molecule

NK

natural killer

SMAC

supramolecular activation cluster.