NK cell-mediated effector functions are regulated by a delicate balance between positive and negative signals. Receptors transmitting negative signals upon engagement with target cell MHC class I molecules have been characterized in detail in recent years. In contrast, less information is available about receptor-ligand interactions involved in the transmission of positive or “triggering” signals to NK cells. Recently, it has been described that murine NK cells are triggered by the costimulatory molecules CD80, CD86, and CD40. Using NK cell lines derived from PBMC as effectors, we demonstrate that the human CD80 and CD86 gene products can function as triggering molecules for NK cell-mediated cytotoxicity. Expression of human CD80 or CD86 molecules in murine B16.F1 melanoma cells rendered these significantly more susceptible to lysis by human NK cell lines. Blocking of the transfected gene products with specific mAb reduced lysis levels to that of nontransfected control cell lines. Triggering of human NK cells by CD80 and CD86 appeared to be independent of CD28 and CTLA-4, at least as determined by the reagents used in the present study, because the expression of these molecules could not be detected on the NK cell lines by either flow cytometry or in redirected lysis assays. Thus, human NK cells may use receptors other than CD28 and CTLA-4 in their interactions with CD80 and CD86 molecules. Alternatively, interactions may involve variants of CD28 (and possibly CTLA-4) that are not recognized by certain anti-CD28 mAb.
The outcome of NK cell-mediated effector functions, including “spontaneous” cytotoxicity and cytokine release, elicited upon contact with other cells depends on a delicate balance between inhibitory and stimulatory signals (1, 2, 3). NK cells express MHC class I-specific receptors that suppress NK cell-mediated cytotoxicity upon interaction with their ligands. Several inhibitory class I binding receptors have been identified. Human NK cells express Ig superfamily members of the killer inhibitory receptors (KIR)3 family (4) and the lectin-like receptor CD94, which forms disulfide-linked heterodimers with members of the NKG2 family (5). The latter CD94/NKG2 receptors are most likely also expressed by murine NK cells (6, 7), which also express the relatively well-characterized lectin-like Ly49 inhibitory receptors (8).
Relatively less is known about activating or “triggering” receptors on NK cells and their corresponding ligands. Probably the best defined activation receptor is the FcγRIII (CD16) molecule, through which NK cells mediate Ab-dependent cellular cytotoxicity (ADCC) against targets coated with IgG (9, 10). Several other cell-surface molecules have been shown to activate NK cells upon specific stimulation, though the biological significance of these responses have been less clear. Perhaps the best studied molecule in this respect is the rodent NKR-P1 molecule (11). Other candidate activation molecules include the mouse 2B4 molecule and LAG-3 (12, 13). The NK-TR1 molecule may play a significant role in signal transduction during natural killing but not in ADCC (14). Several other NK cell-expressed molecules have been reported to activate NK cells. Of these, the best characterized are CD2, mouse CD69 and Ly6, and rat gp42 molecules (15, 16, 17).
Most of the identified KIR have immunoreceptor tyrosine inhibitory motifs (ITIM) in their cytoplasmic tail that prevent NK cell activation (18). However, additional KIR lacking ITIM have been identified that trigger NK cell-mediated lysis upon interaction with appropriate HLA alleles (19, 20). In a similar manner, CD94/NKG2A heterodimers contain an ITIM and inhibit NK cell lysis upon interaction with its ligand HLA-E (21, 22). In contrast, CD94/NKG2C heterodimers lack a corresponding ITIM and may function as activating receptors upon association with KARAP/DAP12 (23). The biological role of these MHC class I binding activating receptors, and corresponding Ly49 receptors in mice that lack ITIM (24), is not yet defined.
Two recently identified candidates for non-MHC binding receptors are the human NKp44 and NKp46 molecules described by the Moretta groups (25, 26, 27). NKp44 is expressed by NK cells upon in vitro cultivation in IL-2. NKp46 is expressed by both resting and activated NK cells. The available data strongly support the notion that NKp44 and NKp46 trigger NK cell-mediated cytotoxicity (25, 26, 27). NKp44 and NKp46 are coupled to the intracytoplasmic signal transduction machinery via association with KARAP/DAP12 or CD3ξ, respectively (27).
Previous reports from our own and other laboratories have demonstrated that murine NK cells are triggered to lyse tumor target cells expressing the costimulatory molecules CD80, CD86, and CD40 (28, 29, 30, 31) and that human peripheral blood-derived NK cells can be triggered by CD40 (32). In addition, it has been demonstrated that the human NK leukemia cell line YT2C2 is capable of killing B7-1-transfected tumor cell lines in a CD28-dependent fashion (33). In the present study, we have addressed the ability of human peripheral blood-derived NK cell lines to be triggered by human CD80 and CD86 costimulatory molecules expressed on tumor cell lines. We demonstrate that the expression of human CD80, CD86, or CD80/CD86 molecules in murine B16.F1 melanoma cells increases their susceptibility to lysis by human NK cell lines. Blocking of the transfected gene products with specific mAb reduced susceptibility to lysis to levels comparable to that of nontransfected control lines. Triggering of human NK cells by CD80 and CD86 appeared to be independent of CD28 and CTLA-4, at least as determined by the reagents used within this study, because expression of these molecules could not be detected on the cell surface of the NK cells used as assessed by either flow cytometry or in redirected lysis assays. However, these data do not exclude the possibility that NK cells may express variants of CD28 (or possibly CTLA-4) that are not recognized by the mAb used within this study (34).
Materials and Methods
K562 (human erythroleukemia), P815 (murine mastocytoma), RPMI 8866 cells (human EBV-transformed B cell line, kindly provided by G. Trinchieri, Philadelphia), and the B16.F1 murine melanoma (C57BL/6) were cultured in complete medium consisting of RPMI 1640 media (Life Technologies, Täby, Sweden) supplemented with 10% heat-inactivated FCS, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. B16.F1 cells transfected with vectors containing the human CD80 and CD86 molecules have been described (35). B16.F1 cells transfected with human CD80 (B16.F1-hCD80) were maintained in complete media supplemented with 500 μg/ml G418 (Saveen, Malmö, Sweden); and B16.F1 cells transfected with human CD86 (B16.F1-hCD86) were maintained in media supplemented with 20 μg/ml mycophenolic acid (Sigma, Stockholm, Sweden), 20 μg/ml xanthine (Sigma), and hypoxanthine, aminopterin, and thymidine (1× HAT; Life Technologies). The B16.F1 cell lines transfected with both CD80 and CD86 (B16.F1-hCD80/CD86) were maintained in media containing 500 μg/ml G418, 20 μg/ml mycophenolic acid, 20 μg/ml xanthine, and 1× HAT.
Monoclonal Abs and flow cytometry
FITC- and PE-labeled mAbs against CD3 (clone HIT3a), CD16 (clone 3G8), CD28 (L293), CTLA-4 (BNI3), CD56 (B159), CD80 (L307.4), CD86 (2331), HLA-DR (L243), and isotype-matched labeled controls were obtained from Becton Dickinson (Stockholm, Sweden). An additional anti-CD28 mAb (YTH913.12; Serotec, Novakemi, Stockholm, Sweden) was also used. Abs were used to routinely screen transfected cell lines for expression of the transgene product by flow cytometry. Purity of NK lines was also assessed by flow cytometry. For blocking studies, purified anti-human CD80 (DAL-1) and anti-human CD86 (BU63) mAb were used (Serotec). For staining, cells were washed and resuspended in PBS with 1% heat-inactivated FCS and 0.01% NaN3. Abs were diluted in this buffer and used at a final concentration of between 2 and 20 μg/ml. Incubations with Abs were conducted for 30 min on ice. Following the final washing, labeled cells were fixed with 1% formaldehyde solution (Sigma), and 10,000 cells were analyzed by flow cytometry on a FACScan flow cytometer using CellQuest software (Becton Dickinson).
NK polyclonal populations and NK cell lines
Lymphokine-activated killer (LAK) cells and NK cell lines were generated from human blood. Briefly, human PBMC were isolated from heparinized venous blood or cytopheresis buffy coats from healthy adult volunteer donors by Ficoll-Hypaque density centrifugation (Lymphoprep, Nycomed, Oslo, Norway). PBMC were depleted of adherent cells by plastic adherence for 45 min at 37°C. The recovered cells were resuspended at a concentration of 1 × 106 cells/ml in complete medium supplemented with 1000 U/ml recombinant human IL-2 (rhIL-2, Peprotech, London, U.K.) or 2000 U/ml rhIFN-α (Peprotech) and cultured at 37°C for 48 h. NK cell lines were generated by resuspending nonadherent cells in IMDM medium (Life Technologies) supplemented with 5% heat-inactivated FCS, 5% heat-inactivated AB+ human serum, and 10% leukocyt- conditioned medium (36). Then, 2.5 × 105 cells/well were cultured in 24-well plates at 37°C, in the presence of 0.5 × 105 irradiated (5000 rad) RPMI 8866 cells (37). Following 7 days of culture, cells were collected and enriched for NK cells by negative depletion. Cells were incubated for 45 min at 4°C with saturating amounts of anti-CD3 (OKT3), anti-CD4 (OKT4), and anti-CD8 (OKT8) hybridoma supernatants. Cells were washed twice with PBS before resuspending in rabbit complement (Pel-Freez, Rodgers, AR) and incubated at 37°C for 1 h. After washing, 1 × 105 recovered cells were further expanded in 24-well plates in complete medium supplemented with 200 U/ml rhIL-2 in the presence of 1 × 105 RPMI 8866 cells to generate NK cell lines. Typical lines were >99% positive for the human NK cell marker CD56 and negative for CD3 and other T cell markers. In some experiments, the NK cell lines were cultured for 48 h in media supplemented with 100 ng/ml rhIL-12 (Peprotech) before use as effector cells in cytotoxicity assays.
NK cell-mediated cytotoxicity
Cytotoxicity of cells was measured in a standard 4-h 51Cr release assay using Na251CrO4-labeled cells as targets. Experiments were conducted in triplicate at various E:T ratios. The percentage specific 51Cr release (specific lysis) was calculated according to the formula: % specific lysis = [(experimental release − spontaneous release)/(maximum release − spontaneous release)] × 100.
In Ab blocking studies, 1 × 106 target cells were preincubated for 30 min at room temperature with 25 μg/ml of either anti-CD80 or anti-CD86 or isotype control mAb. Cells were washed in complete media before use in the cytotoxicity assay.
Purified NK cells were incubated, in round-bottom 96-well plates, together with 51Cr-labeled P815 cells to achieve a final E:T ratio of 40:1 in a total volume of 150 μl. Monoclonal Ab specific for CD28, CTLA-4, CD16, and isotype-matched controls were added at a final concentration of 200 ng/ml. Cultures were incubated for 4 h and harvested as in a standard chromium release assay.
CD80-, CD86-, and CD80/CD86-transfected B16.F1 cell lines
Transfection of murine melanoma B16.F1 cells with cDNA for human CD80 (hCD80), human CD86 (hCD86), or both hCD80 and hCD86 resulted in stable cell surface expression of these molecules (35). Table I shows the expression of hCD80 and hCD86 on control and transfected B16.F1 cells. Expression of MHC class I molecules was not affected by the hCD80 or hCD86 gene products (data not shown). The levels of CD80 and CD86 expressed by the transfected cell lines were found to correspond roughly to those expressed by monocyte-derived dendritic cells (data not shown).
Lysis of CD80-, CD86-, and CD80/CD86-transfected B16.F1 cells by human LAK cells
Human LAK cells, stimulated with either rhIFN-α or rhIL-2, do not efficiently lyse B16.F1 melanoma cells. In contrast, similar effector cells readily killed B16.F1 cells transfected with the hCD80, hCD86, or hCD80/hCD86 costimulatory molecules. Significantly higher levels of lysis were observed for the transfected cells when compared with corresponding untransfected cell lines (Fig. 1) or vector-transfected control cell lines (data not shown).
Lysis of CD80-, CD86-, and CD80/CD86-transfected B16.F1 cells by human NK cell lines
Because LAK cell cultures stimulated by high levels of IL-2 are a potential source of cytotoxic T cells, we wished to address the lytic activity of pure NK cells. Thus, long-term CD16+CD56+CD3− NK cell lines were generated. These lines readily killed targets transfected with the hCD80, hCD86, and hCD80/hCD86 costimulatory molecules while the wild-type parent cell line or vector control cells were killed at significantly lower levels (Table II, and data not shown). Notably, however, the hCD80/hCD86 double transfectant cells were not killed better than any of the single transfectants (Table II). It was reported that murine NK cells stimulated with IL-12 exhibited augmented responses to murine CD80 (mCD80)- and murine CD86 (mCD86)-transfected tumor cell lines (30). Therefore, additional experiments were performed in which the IL-2-propagated NK cell lines were stimulated with rhIL-12. These effector cells did not display any detectable increased ability to lyse hCD80- or hCD86-transfected target cells compared with cell lines stimulated with IL-2 alone (data not shown).
Blocking of CD80 and CD86 with specific mAbs impaired NK cell line triggering of lysis
To verify that the increased sensitivity of the CD80- and CD86-transfected cell lines to NK cell line-mediated lysis was a direct consequence of cell-surface costimulatory molecule expression, experiments were performed in which the hCD80, hCD86, or hCD80/hCD86 molecules were blocked with specific mAb. The increased lysis of the transfected target cells could be inhibited completely by addition of specific mAb against hCD80 or hCD86, demonstrating that the observed effect was not an indirect consequence of the expression of the transgene (Fig. 2).
NK cell-mediated killing of CD80/CD86-expressing targets occurred in the absence of CD28 and CTLA-4
The NK cell lines used in the present study were analyzed phenotypically using flow cytometry and were found to be CD3−CD14−CD19−CD56+. The IL-2-propagated CD56+ NK cell lines did not express any detectable levels of CD28 or CTLA-4 (data not shown). As T cells up-regulate these molecules upon activation, it could be argued that this is also the case with NK cells. Thus, the NK cell lines were analyzed for expression of CD28 and CTLA-4 after stimulation with high doses of rhIL-2 (1,000 U/ml) (Fig. 3, A and B) or with rhIL-2 plus rhIL-12 (1000 U/ml and 100 ng/ml, respectively) (Fig. 3, C and D). Using two separate anti-CD28 mAb (L293 and YTH913.12), we could not detect any levels of expression of CD28 and CTLA-4 following activation (Fig. 3, A and C; data not shown). Control experiments using PHA and Con A T cell blasts demonstrated that these Abs were able to recognize appropriate cell-surface molecules (data not shown).
NK cells can mediate reverse ADCC against NK-resistant P815 cells following CD16 cross-linking. Addition of anti-CD28 and anti-CTLA-4 mAb into reverse ADCC assays did not lead to any induction of lysis of the P815 cells. This absence of lysis provides further support for the notion that NK cell recognition of hCD80 or hCD86 is not mediated by either CD28 or CTLA4 (Fig. 4).
There is an emerging consensus that NK cell recognition and subsequent effector mechanisms are regulated by both activating and inhibitory signals. In this study, using purified populations of human NK cells, we demonstrate that the human CD80 and CD86 gene products function as triggering signals for NK cell-mediated cytotoxicity.
The present results are in line with observations in the murine system, demonstrating that CD80, CD86, and CD40 can trigger mouse NK cells to mediate lysis (28, 29, 30, 31). Furthermore, it has been demonstrated that human NK cells can be triggered by interactions between CD40 on the target cell and CD40L expressed by NK cells (32). However, it should be noted that introduction of the human CD80 gene into a previously negative human squamous cell carcinoma of the head and neck was found not to be associated with increased susceptibility to lysis by human NK cells (38). This led the authors to conclude that CD80 molecules are not involved in triggering of adult NK cells. While this conclusion may be valid for this particular cell line, it may not exclude a triggering effect of CD80, or other costimulatory molecules, when introduced into other tumors. NK cell insensitivity despite expression of costimulatory molecules can be explained in many ways, e.g., lack of appropriate expression of relevant adhesion molecules, inhibition of lysis by MHC class I, or lack of other necessary triggering molecules.
Previous data from our groups have suggested that murine costimulatory molecules trigger NK cells in a CD28- and CTLA-4-independent way (29, 31). However, murine (39, 40) and human studies (33, 41, 42) suggest that when CD28 is expressed by NK cells, such as the human YT cell line, then this molecule may transmit triggering signals through ligation with CD80. Thus, for CD28-expressing NK cell effectors, killing of target cells expressing costimulatory molecules may be in part dependent upon triggering via this receptor. However, this does not exclude a potential CD28/CTLA-4-independent component operating in parallel. The present results demonstrate that triggering of NK cells by CD80- and CD86-expressing target cells may occur independently of CD28. The failure to detect cell-surface expression of both CD28 and CTLA-4 by either redirected lysis experiments or direct flow cytometric analysis argues against the involvement of these molecules, at least in a form detectable by the present reagents used. The absence of detectable levels of CD28 and CTLA-4 on human NK cell lines derived from peripheral blood is also in line with observations by others (38). However, we cannot exclude that very low expression levels of CD28 and/or CTLA-4 (undetectable by the present means of analysis) or the existence of variants of CD28 and/or CTLA-4 on NK cells could account for some of the observed effects. Indeed, a recent publication by Galea-Lauri et al. has suggested that detection of the CD28 molecule on NK cells depends entirely upon the mAb used (34). The authors suggested that NK cells may express a variant isotype of CD28, as two of four anti-CD28 mAbs positively stained NK cells whereas all four stained T cells. Incidentally, the two mAbs that failed to stain NK cells were also the two mAb used throughout this study. This difference of mAb binding could be due to posttranslational modifications, splice variants of the CD28 molecule, or other alterations. Galea-Lauri et al. observed that there was considerable variation in the levels of CD28 expression between NK lines and also individuals despite mRNA levels remaining steady. NK cells may also express novel CD80/CD86 binding receptors. In this respect, it should be mentioned that a novel receptor, functionally related to CD28, has been described recently (43). It is possible that such CD28-related receptors could account for the observed effects. Alternatively, it cannot be excluded that the CD80 and CD86 molecules are recognized by several low-affinity “pattern recognition receptors” that may not easily be identified using conventional flow cytometric techniques.
It is known that NK cells can be triggered by several different receptors (2), thus the triggering observed with targets expressing CD80 and CD86 could operate in parallel with other activating molecules. Susceptible targets may interact with NK cells through multiple different receptors simultaneously. It has been proposed that NK cell activation results from signals through one or more receptors that are not NK cell specific such as adhesion molecules, including e.g., CD2, LFA-1, and CD28 (18). However, NK cell specificity has been claimed for the recently identified NKp44 and NKp46 receptors (26, 27). Furthermore, NKR-P1 expression appears to be NK cell specific, with the exception of the expression of NK1.1 on murine “NKT” cells (44). Thus, NK cell triggering may be mediated by NK cell specific as well as more widely used receptor-ligand interactions.
In conclusion, the present, as well as recently published (28, 29, 30, 31, 32, 33), observations on NK cell interaction with costimulatory molecules suggest that these molecules should be put on the list for candidate NK cell-activating ligands on target cells. As to the costimulatory molecules, the biological relevance of the ability of NK cells to recognize these molecules remains to be explored. Perhaps NK cells communicate in a specific manner with professional APC via these molecules (45). Indeed, a recent study in a murine model suggested that cell-to-cell contact between DC and resting NK cells resulted in a substantial increase in both NK cell cytotoxicity and IFN-γ production (46). Moreover, we have observed that monocyte-derived dendritic cells and Langerhans-like cells are efficiently lysed by autologous NK cells (J.L.W. et al., unpublished observations).
We thank Drs. E. Carbone, K. Kärre, and K. Söderström for fruitful discussions.
This work was supported by grants from the Swedish Cancer Society, the Swedish Medical Research Council, the Petrus and Augusta Hedlunds Stiftelse, the Tobias Stiftelsen, the Lars Hiertas Stiftelse, and the Karolinska Institutet. J.L.W. has been supported by a long-term Marie Curie Biotechnology Fellowship. A.M.-F. has been supported by a Cancerfonden Postdoctoral Fellowship.
Abbreviations used in this paper: KIR, killer cell inhibitory receptor; ITIM, immunoreceptor tyrosine inhibitory motif; LAK, lymphokine activated killer cell; ADCC, Ab-dependent cellular cytotoxicity; rh, recombinant human; h, human; m, murine.