5- (and 6-) Carboxyfluorescein diacetate succinimidyl ester
5- (and 6-) (((4-Chloromethyl) benzoyl) amino) tetramethylrhodamine
A prediction from the "missing self" hypothesis is that down-regulation of MHC class I on resting hematopoietic cells should be sufficient to make them susceptible to NK cell killing. Using a method enabling kinetic and quantitative assessments of NK cell-mediated rejection responses in vivo, we here show that resting hematopoietic cells from β2-microglobulin-deficient (β2m–/–) mice were rapidly rejected in unmanipulated C57BL/6 (B6) mice. In situations of allelic MHC class I mismatches rejection occurred but required longer time. β2m–/– donor cells pre-activated with concanavalin A were more efficiently eliminated compared to resting cells, as were MHC– tumor cells. When recipient mice were pretreatedwith an IFN inducer to activate NK cells, rejection was also enhanced. The signaling adaptor KARAP/DAP12 was dispensable for rejection of β2m–/– cells (lacking MHC) but critical for rejection of BALB/c cells (mismatched MHC) in unmanipulated B6 recipients. In contrast, B6 recipients with pre-activated NK cells rejected BALB/c cells in a KARAP/DAP12-independent fashion. Loss or mismatch of MHC class I in resting cells was thus sufficient to convey susceptibility to NK cell rejection. However, activation of the effector or the target enhanced rejection and shifted the balance between different signaling pathways involved.
NK cells are part of the innate immune system and provide rapid protection against infections. NK cells are also efficient killers of cells lacking MHC class I expression 1.The reason for increased killing of MHC class I-deficient cells is their inability to deliver inhibitory signals to the NK cells. Negative signals are transmitted primarily by MHC class I-specific receptors containing immunoreceptor tyrosine-based inhibitory motifs (ITIM) 2, 3. In the mouse, those include but are not limited to Ly49 family receptors expressed on NK cells and subsets of T cells 4, 5. Inhibitory signals counteract triggering signals delivered by a diverse set of activating NK cell receptors. A few Ly49 receptors lack ITIM and instead have the capacity to deliver activating signals to the NK cells. One such example is Ly49D, which recognizes the MHC class I molecule H-2Dd and activates NK cells via the adaptor molecule KARAP/DAP12 6–9. Other activating receptors on NK cells are NKR-P1, NKG2D, natural cytotoxicity receptors (NKp30, NKp44, NKp46), Ly49H, and adhesion molecules 10, 11. Ly49H and NKp44 also signal via KARAP/DAP12, while other activating receptors associate with other adaptor proteins such as DAP10, CD3ζ, and FcϵRIγ 9, 11, 12. Further adding to the complexity, some receptors, such as 2B4 and NTB-A, can be both activating and inhibitory depending on the maturation state of the NK cell 13, 14.
Tumor cells, stressed cells, and virally infected cells are particularly rich in ligands for NK cell activating receptors, explaining their increased NK sensitivity compared to normal cells 15, 16. In light of the potential use of NK cells in clinical settings 17 it is becoming increasingly important to understand in more detail how NK cells differentiate between tumor cells and normal surrounding cells and which role missing self recognition may have in this process. An initial prediction from the missing self model is that normal cells would be protected from NK cell killing as long as they expressed self MHC class I molecules 1. Previous studies supporting this concept all have their specific limitations. Tumor transplantation experiments are limited to cancer cells, while BM transplantation experiments require irradiation of the recipient and measures mainly rejection of activated, proliferating cells 18–21. Allogeneic lymphocyte cytotoxicity reveals NK-mediated rejection responses against normal lymphocytes but still depends on radioactive labeling of the donor cells and is limited to allogeneic situations 22, 23. Thus, the question of whether unmanipulated hosts display missing self reactivity against healthy resting cells with low levels of self MHC class I has not been critically studied.
In this study, we have used and further developed a quantitative fluorescence-based technique previously used to study rejection of infected lymphocytes by CD8+ T cells in vivo24 for studies of NK cell responses. Rejection of resting spleen cells and BM cells in unmanipulated hosts was compared to the NK cell-mediated killing of pre-activated T cell blasts and lymphoma cells. We also used hosts in which NK cells had been activated by administration of an IFN inducer. Our results demonstrate that resting hematopoietic cells were indeed rejected according to the principles of missing self by unmanipulated hosts, implying their constitutive expression of ligands for activating NK cell receptors. This rejection response was augmented in situations where either the grafted cells or the NK cells were pre-activated. Our experiments also showed that NK cells may use different signal transduction pathways for different types of MHC-based recognition; pathways that are not static but dependent on the state of activation of NK cells at the time of rejection.
2.1 Rejection of β2-microglobulin-deficient bone marrow and spleen cells in vivo
The intracellular dye 5- (and 6-) carboxyfluorescein diacetate succinimidyl ester (CFSE) easily labels all cell types, including normal resting cells. The intensity of the labeling is proportional to the concentration of dye used in the labeling reaction, opening up for simultaneous detection of more than one cell type based on the intensities of the CFSE signals. Freshly isolated BM cells from C57Bl/6 (B6) and β2-microglobulin-deficient (β2m–/–) mice were labeled with different concentrations of CFSE, mixed in a 1:1 ratio and injected intravenously into B6 mice. One day later, spleen and liver cells were analyzed by flow cytometry. To obtain a quantitative estimate of the rejection, we compared the percentage of remaining β2m–/– cells to injected control B6 cells, calculated a ratio of β2m–/– to B6 cells, and compared it with the ratio of these cells in the injection cell mix at day 0.
Freshly isolated β2m–/– BM cells were rapidly rejected into B6 mice; only about 30–50% of β2m–/– cells localizing in the spleen or liver remained after 18 h (Fig. 1A). The rejection was NK cell-dependent since mice depleted of NK cells had a ratio of β2m–/– cells to B6 cells similar to the inoculation mixture (Fig. 1A, bar graph). Equivalent results were obtained when spleen cells from B6 and β2m–/– mice were injected (Fig. 1B). CFSE-labeled BM and spleen cells could be recovered from blood, lung, BM, spleen, lymph node, and liver with similar results regarding rejection of target cells, suggesting that the preferential loss of β2m–/– cells was not due to differential homing properties compared to syngeneic cells (data not shown). These results demonstrated that NK cells in unmanipulated hosts rapidly sensed lack of MHC class I molecules on spleen and BM cells in vivo and selectively removed them without affecting MHC+ cells.
We next took advantage of the quantitative nature of the CFSE-based method and asked whether the number of target cells reaching a specific organ would influence the percentage of cells being killed. Under such circumstances, an increased number of inoculated cells reaching an organ would lead to a decreased elimination ratio of β2m–/– cells. This was indeed the case. The more inoculated cells had reached the spleen, the less relative rejection of β2m–/– cells compared to B6 cells was seen (Fig. 1C). Our interpretation of this finding is that resting NK cells are able to eliminate a certain number of susceptible target cells per time unit. This may provide an explanation for the less efficient rejection of BM in the liver compared to the spleen (Fig. 1A, bar graph). BM cells home better to the liver compared to the spleen, which results in a larger amount of injected cells in this organ.
An alternative explanation for a decreased rejection ratio when more target cells were present in an organ is a tolerizing effect on the NK cells. If that was the case, NK cells would be less likely to reject targets lacking MHC class I after an earlier encounter with such targets. However, we did not observe increased or decreased rejection efficiency of CFSE-labeled β2m–/– spleen cells after a previous inoculation of unlabeled β2m–/– spleen cells or RMA-S tumor cells into B6 mice (data not shown). This indicated that the NK cells, after exposure to cells lacking MHC molecules, became neither more reactive, nor tolerant to such targets.
2.2 MHC class I-deficient T cells and B cells are rejected with equal efficiency
We next evaluated whether missing self reactivity against normal cells was differently effective against specific cell populations. Since surviving cells are detected using flow cytometry, we were able to monitor NK cell-mediated elimination of different cell populations using antibodies. When the rejection of β2m–/– T cells versus non-T cells as well as β2m–/– B cells versus non-B cells was determined, we found no differences in relative rejection (Fig. 2), suggesting that all naive β2m–/– lymphocytes were eliminated with equal efficiency by B6 NK cells.
2.3 NK cell-mediated rejection is more efficient against tumor cells, activated target cells, and in pre-activated recipients
Our data showed that normal resting cells could be killed if they represented a missing self phenotype, but they did not exclude that this rejection process would be more efficient against tumor cells and "activated" normal NK cells. We therefore compared the rejection of normal, naive spleen cells from β2m–/– mice with rejection of RMA-S tumor cells or β2m–/– Con A blasts. The tumor cells (RMA and RMA-S) as well as the Con A blasts (B6 and β2m–/–) mainly localized to the lungs rather than to the liver and spleen, presumably due to their larger sizes. Despite the larger number of RMA-S cells and β2m–/– Con A blasts reaching the lungs, both types of "activated" MHC– targets were significantly better rejected compared to resting β2m–/– spleen cells in this organ (Fig. 3A). This suggests that the response by NK cells in vivo was enhanced by activation or transformation of the target cells. Conversely, NK cell-mediated rejection of lymphocytes was significantly increased in recipients that had been treated with the IFN inducer tilorone before the transfer (Fig. 3B). Thus, activation of either the target cell or the NK cell led to a significant increase in rejection efficiency, presumably due to up-regulation or induction of new activating pathways.
2.4 Faster rejection of β2m–/– cells compared to B6 cells in D8 mice
The rejection of β2m–/– cells represents an extreme case of missing self recognition, where all MHC class I molecules are lacking. Since NK cell activity is regulated by several different Ly49 receptors sensing different or overlapping MHC molecules 4, 5, we tested whether rejection of resting cells occurred even in a case of a single MHC class I mismatch relative to the host. By usage of several fluorescent dyes, the method enables to quantitatively estimate three different targets simultaneously. We used D8 mice (B6 mice transgenic for H-2Dd) as recipients, and injected B6, β2m–/–, and D8 spleen cells labeled with 0.4 μM of CFSE, 4 μM of CFSE, and 10 μM of 5- (and 6-) (((4-chloromethyl) benzoyl) amino) tetramethylrhodamine (CMTMR), respectively.
We followed the rejection kinetics by measuring the percentages of remaining cells from the respective donor populations in the blood at the indicated time points (Fig. 4A, B). If the injected spleen cells lacked all MHC molecules (β2m–/–), rejection was very fast and few cells remained after 2 days (Fig. 4B). Cells lacking only one of the MHC molecules (B6) were also rejected; however, this process was slower and it took several days for the NK cells to remove all injected B6 cells (Fig. 4B). The slower rejection of B6 cells was not due to the fact that the NK cells first sensed cells lacking all MHC molecules and that these acted like "cold targets" with respect to the B6 targets, since similar results were obtained when β2m–/– cells and B6 cells were inoculated separately (data not shown).
2.5 The role of KARAP/DAP12 in NK cell rejection triggered by MHC mismatches
The activating receptors providing the initial triggering signals in missing self recognition are not known. Rejection of cells lacking β2m-associated molecules in vivo was not dependent on activating NK cell receptors that signal via the adaptor molecule KARAP/DAP12 in our experiment, since rejection of β2m–/– spleen cells in vivo was as efficient in KARAP/DAP12-loss-of-function mice as in B6 control mice (Fig. 5A). We next evaluated the role of missing self recognition in rejection of fully mismatched allogeneic cells by measuring the rejection of BALB/c cells in B6 mice. NK cells could here reject the target cells because they would sense either "missing" H-2Kb and H-2Db, or the "presence" of H-2Kd, H-2Dd or HL-2Ld via activating receptors such as Ly49D.
In contrast to the elimination of β2m–/– cells, rejection of naive BALB/c cells was completely defective in KARAP/ DAP12–/– mice even after 2 days in vivo (Fig. 5B). Interestingly, when the experiments were repeated using mice injected intraperitoneally with the synthetic IFN inducer tilorone, rejection of BALB/c spleen cells in B6 recipients no longer required KARAP/DAP12 (Fig. 5B). Thus, although KARAP/DAP12 was important for NK cells to recognize and kill allogeneic cells in the naive mouse, this pathway was no longer dominant in mice with pre-activated NK cells. We consistently observed some degree of rejection of β2m–/– cells in tilorone-activated, NK1.1-depleted mice. This could be due to a few remaining, strongly activated NK cells or possibly another cell type. However, the difference between NK-depleted and untreated mice was still present confirming a KARAP/DAP12-independent NK cell-mediated rejection.
NK cells are prevented from killing resting, healthy cells through the expression of inhibitory receptors, which bind to overlapping sets of alleles of MHC class I molecules 25, 26. For example, in the B6 mouse Ly49C binds to H-2Kb and turns the NK cells off 27. Implicit in this reasoning is that cells that lack the appropriate MHC molecules would automatically turn into susceptible targets for killing by NK cells. The in vivo assay utilized here tests this implication for the first time. We show that freshly isolated, resting target cells are eliminated within a short time frame in vivo. Both BM and spleen cells were rapidly rejected by the NK cells if they lacked β2m-associated MHC molecules.
Another advantage of a fluorescence-based rejection assay is that the presence of target cells in different organs can be recorded. Injected cells could be found in several different organs after 1 or 2 days, but typically different cell types located preferentially to different organs. This may be due to different expression of adhesion molecules, chemokine receptors, as well as to the size of different cell types 28. Thus, BM cells were mainly found in the liver and spleen cells mainly in the spleen. However, both types of cells were efficiently killed in both organs. We also compared the rejection ratio of different lymphocyte populations such as B cells and T cells, which again were killed with equivalent efficiency. This result suggests that most, if not all, normal hematopoietic cells express ligands for activating NK cell receptors. The use of intracellular dyes makes it possible to study a wide range of NK-sensitive targets in vivo, during a short time frame, and in many different situations. The method is not restricted to missing self activity, but could be used for measuring cytotoxicity against all NK-sensitive targets that have an NK-resistant counterpart, for example transfected versus non-transfected cells.
Activated cells or tumor targets showed enhanced recognition and elimination by NK cells, possibly due to higher cell surface levels of activating ligands. In support of a quantitative relationship between the extent of activating ligands and NK cell killing, ectopic expression of ligands for the activating receptor NKG2D, such as Rae-1 and H60, results in enhanced NK cell-mediated rejection of those tumors despite the fact that those tumors express normal levels of MHC molecules 29, 30. For the comparisons we used in this study, this explanationmust invoke other activating ligands, since RMA, RMA-S, and B6 Con A blasts do not express ligands for the NKG2D receptor 29.
Previous studies have shown that NK cells in D8 mice are inhibited when they sense the presence of H-2Dd through the interaction with Ly49A and Ly49G2 31, 32. We observed a kinetically slower rejection of B6 target cells compared to β2m–/– cells in the D8 mouse. We believe that this is due to the fact that most NK cells in D8 mice express not only the H-2Dd-binding receptors Ly49A and/or Ly49G2, but co-express and are inhibited via the Ly49C receptor for H-2Kb. Therefore, a smaller NK cell subset was actively rejecting the B6 cells compared to the β2m–/– cells. However, during a longer time frame, it was possible for those few NK cells to completely eliminate cells lacking only one MHC molecule. Different kinetics for rejecting cells with different degrees of "missing self" disparities in the same recipient argue for independent control of different NK cell subsetsand against the idea that the presence of a strong missing self barrier in one direction (missing H-2Kb and H-2Db) would influence rejection against another (missing H-2Dd only).
The so far identified activating NK cell receptors use different proximal signaling transduction pathways. One example is Ly49D, which recognizes H-2Dd and gives activating signals via the adaptor molecule KARAP/DAP12 8, 9. KARAP/DAP12–/– mice can recognize and kill MHC– tumor targets, such as RMA-S and C-44.25–, butare dysfunctional in natural cytotoxicity against the macrophage cell lines J774 and IC-21 33. We found no difference in rejection of β2m–/– cells between B6 and KARAP/DAP12–/– mice, indicating that NK cells did not critically depend on a KARAP/DAP12-associated receptor to sense missing self. However, the rejection of fully MHC-mismatched cells (BALB/c) was impaired. Previous results have demonstrated that BALB/c BM allografts are normally rejected in B6 mice, and administration of an anti-Ly49D antibody reduces rejection significantly 6, 34. We conclude that in naive B6 mice, the major cause for rejection of BALB/c cells was a direct recognition of H-2Dd by one or several KARAP/DAP12-associated receptors such as Ly49D. Intriguingly, B6 NK cells did not appear to sense missing self under those circumstances. One explanation might be that B6 mice express high levels of inhibitory Ly49 receptors for H-2d molecules. Another possibility could be that the activating receptor responsible for missing self in B6 mice does not recognize its ligand on BALB/c cells.
Intriguingly, NK cells activated with an IFN inducer were able to reject BALB/c cells also when the mouse was deficient in KARAP/DAP12 signaling. This indicated that a novel triggering receptor, acting independently of KARAP/DAP12, had been up-regulated. Alternatively, Ly49D or other H-2Dd-sensing activating receptors normally signaling through KARAP/DAP12 used another signalingmachinery after activation via IFN. Ly49D is capable of associating with and signaling via CD3ζ in Jurkat cells 35, but the biological relevance of this pathway is uncertain 33. Another activating receptor, NKG2D, associates with DAP10 in naive NK cells. When the NK cells are activated, a splice variant of NKG2D, which associates with KARAP/DAP12, appears 36. It is possible that Ly49D or another activating receptor may function in a similar manner.
We conclude that NK cells rapidly recognize and react against missing MHC class I molecules in vivo, on naive cells. This process does not require pre-activation of the NK cells, such as wouldbe expected during inflammation, or of the target cells, as during tumorigenesis or infection. Different MHC– lymphocyte populations are eliminated by NK cells with similar kinetics, but pre-activated or transformed targets are better recognized by NK cells in vivo compared to naive targets. In naive mice, a KARAP/DAP12 signaling deficiency abolishes recognition of MHC class I-mismatched cells while missing self recognition remains intact. Interestingly, a KARAP/DAP12-independent activation pathway appears to be dominant in mice with activated NK cells to sense allogeneic targets. The differential efficiency by which missing self recognition operates against resting cells and activated cells suggests a way by which NK cells could act on tumor cells and infected cells but still spare normal cells and maintain self tolerance. This knowledge may help improve targeting of NK cells in clinical situations.
4 Materials and methods
B6 (H-2b), B6β2m–/– (H-2b), D8 (H-2b, H-2Dd), BALB/c (H-2d), and KARAP/DAP12-loss-of-function (KARAP/DAP12–/–, H-2b) 33 mice were all maintained and bred at the animal facility at the Microbiology and Tumor Biology Center, Karolinska Institutet, Stockholm, Sweden. All mice were 5–12 weeks old at the time of experiment. When indicated, mice were given 2 mg tilorone analogue R10,874DA orally 1 day before the experiment (T-8014; Sigma-Aldrich Chemie, Steinheim, Germany) and/or 200 μg PK136 (anti-NK1.1, mouse IgG2a; Mabtech, Stockholm, Sweden) intraperitoneally 2 days before the experiment. The Local Animal Ethical Committee approved all animal work.
4.2 Antibodies, FACS analysis
The mAb B220-biotin (RA3–6B2), CD3-PerCP (145–2C11; PharMingen, Stockholm, Sweden) and streptavidin-Alexa 633 (Molecular Probes, Eugene, OR) were used in flow cytometry. To inhibit nonspecificbinding of antibodies to the FcRγ, the cells were incubated with anti-FcRγ (2.4G2) prior to staining with specific antibodies. The cells were stained using standard protocols, gated for viable lymphocytes on forward/side scatter, and analyzed using a FACScan flow cytometer (Becton Dickinson).
4.3 Cell cultures
Con A blasts were made by culturing 107 splenocytes for 2 days in α-minimum essential medium containing 10 % FCS, 10 mM Hepes, 100 U/ml penicillin, 100 μg/ml streptomycin, 50 μM β-mercaptoethanol, and 2 μg/ml Con A (Sigma-Aldrich).
4.4 CFSE labeling and in vivo killing of CFSE-labeled cells
Ascites-grown RMA and RMA-S cells, Con A blasts, and erythrocyte-depleted spleen cells or BM cells were washed and resuspended in PBS supplemented with 0.4 or 4 μM of the dye 5- (and 6-) carboxyfluorescein diacetate (transformed when bound to proteins into CFSE; Molecular Probes) or 10 μM CMTMR (Molecular Probes). The cells were incubated at 37°C for 10 min and washed in RPMI medium containing 5–10% FCS. Cells (1×107) of each type were mixed, and the mixture (200 μl) injected intravenously into recipient mice. A small sample of the injection mix was analyzed by flow cytometry for reference. The concentrations of CFSE used were not toxic to the cells, as the survival was not decreased after incubation and the results were similar whether the β2m–/– or wild type cell population was stained with the higher concentration of the dye (data not shown).
After 18–24 h, blood, spleen, liver, and/or lungs were harvested. CFSE-labeled BM and spleen cells could also be recovered from lymph nodes with similar results (data not shown). Livers were perfused with PBS, cut into pieces, and pressed through a cell dispenser. After washing with PBS, leukocytes were isolated by centrifugation on a Percoll gradient (Amersham Pharmacia Biotech). Lungs were cut in pieces and incubated at 37°C for 1 h in RPMI with 400 U/ml collagenase type IV (Sigma-Aldrich). Erythrocytes were lysed in all cell populations. The relative percentage of cells in each CFSE population was measured with a FACScan. The relative survival of CFSEhigh cells compared to CFSElow cells was calculated as follows: (% of CFSEhigh cells in sample / % of CFSEhigh cells in injection mix) / (% of CFSElow cells in sample / % of CFSElow cells in injection mix). At least 3,000 CFSE+ cells were acquired in each sample.
The authors wish to thank Margareta Hagelin and Maj-Britt Alter for assistance with animal experiments, and the members of K. Kärre's and P. Höglund's groups for fruitful discussions. This study was supported by grants from the Swedish Cancer Society, Karolinska Institutet, the Swedish Foundation for Strategic Research, Cancer Research Institute (USA), the Human Frontiers Science Foundation, the Swedish Research Council, and the Åke Wiberg Foundation. The work was conducted within The Strategic Research Center for Immunoregulation (IRIS), funded by the Swedish Foundation for Strategic Research.