Activated, But Not Resting, T Cells Can Be Recognized and Killed by Syngeneic NK Cells ¹

Natural killer cells are large granular lymphocytes that express multiple activation and inhibitory receptors. These receptors allow NK cells to rapidly survey their environment for danger. When an imbalance in signaling favors activation, secretion of cytokines and/or release of cytotoxic granules occur (1, 2). For example, the rapid secretion of IFN-γ by NK cells after viral infection (e.g., murine CMV (3), HSV (4, 5)) promotes a Th1 response necessary for the clearance of intracellular pathogens (3, 6). Activated NK cells also kill immature dendritic cells (DC) ³ (7), thus regulating T cell expansion and decreasing the risk of proinflammatory tissue damage. In this study, we demonstrate that NK cells can directly limit the survival of activated T cells.

NKG2D is a highly conserved (8, 9), homodimeric, lectin-like, type II transmembrane protein expressed on virtually all rodent and human NK cells (10, 11). It is highly promiscuous, and its ligands include a variety of distantly related MHC-I homologues that are often up-regulated on stressed and activated cells (12, 13). These include MICA and MICB (humans) (10), the ULBPs (humans) (14), Rae-1 (mice), and H-60 (BALB/c mice) (12, 13).

In this study, we demonstrate that activated T cells from BALB/c and B6 mice are killed by syngeneic activated NK cells or lymphokine-activated killer cells (LAK). This is mediated by NKG2D in BALB/c.

C57BL/6 (B6), BALB/c, BALB/c-dm2 (H-2L^d− mutant), and F₁ (B6 × BALB/c) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). B6 and BALB/c FcRIIIγ knockout mice were purchased from Taconic Labs (Germantown, NY). Perforin knockout mice (PKO), generously provided by H. Hentgartner (Zurich, Switzerland), were engineered on the B6 background (15) and bred at the Ontario Cancer Institute. D011.10 (D011) C.B-17-SCID TCR trangenic mice were engineered at the Ontario Cancer Institute by breeding D011 TCR transgenic mice (16) (The Jackson Laboratory) onto the C.B-17 SCID background. Most T cells (>99%) in these mice are CD4⁺ and specific for aa residues 323–339 of chicken OVA (in the context of I-A^d) (16). The 2C TCR transgenic mice were obtained from D. Y. Loh (Nippon Roche Research Centre, Kamakura, Japan) (17). The 2C T cell is CD8⁺ and specific for L^d (presenting the ubiquitous p2Ca mitochondrial peptide) (18, 19). Mice were kept in a specific pathogen-free animal colony in the Ontario Cancer Institute. Animal experimental protocols were monitored and approved by the Ontario Cancer Institute Animal Care Committee. In most experiments, 5- to 7-wk-old mice were used.

Hybridomas 2.4G2 (rat anti-mouse FcγRIII-α) and 1B2 (rat anti-mouse 2C TCR) were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and mAbs purified from tissue culture supernatants, as previously described (20). Fluorescently labeled mAbs anti-NK1.1 (PK136), anti H2-K^b, anti-B220, anti-CD4, anti-CD8, and anti-TCRβ were purchased from BD PharMingen (San Diego, CA). An anti-mouse NKG2D mAb (M315) was made from splenocytes of Lewis rats immunized with mouse NKG2D-Fc fusion proteins (D. Cosman, unpublished observations). Cosman et al. (14) have previously reported that in a murine system, ULBP2-Fc binds only to NKG2D. We found that ULBP2-Fc completely blocked binding of the M315 mAb to murine NK cells and vice versa. The Fc fusion protein, ULBP2-Fc, and a leucine zipper ULBP2 fusion (ULBP2-LZ) were made, as described (14). Labeling of ULBP2-Fc was performed with EZ-link biotin from Pierce (Rockford, IL). Labeling of purified Abs with FITC was performed, as previously described (20). Brefeldin A (BFA), Con A, PMA, and ionomycin were purchased from Sigma-Aldrich (St. Louis, MO). Flow cytometric acquisition and analysis were performed using a BD Biosciences (San Jose, CA) FACSCalibur. FCS Express software (Novosoft; Toronto, Ontario, Canada) was used to analyze flow cytometry data.

Mouse NKG2D-Fc (mNKG2D-Fc) was made in the mammalian expression vector pDC409 (14) by cloning of mNKG2D cDNA in frame with the carboxyl terminus of a human IgG1 Fc region to generate a type II fusion protein. Plasmids encoding the Fc fusion proteins were transfected into CV-1/EBNA (ATCC CRL-10478) cells, and the fusion proteins were purified from culture supernatants by chromatography on a protein A-Poros column (Applied Biosystems, Foster City, CA), as described (14).

NK cells were prepared from spleen cell suspensions from normal mice via the removal of B and T cells, as previously described (20). Purified NK cells (3 × 10⁶) were suspended in 5 ml α-MEM/10% FCS/150 μM 2-ME containing human rIL-2 (Proleukin; Chiron, Emeryville, CA) at a concentration equivalent to 350 mouse IL-2 U/ml and plated in a single well of a six-well flat-bottom plate (Nunc, Scarborough, Ontario, Canada). A conversion factor of 1 mouse unit/21 human IU was obtained by generating a titration curve for CTLL-2 tritiated thymidine uptake over a range of Proleukin concentrations and calculating the concentration for which 50% maximum tritiated thymidine uptake occurred. NK cell-enriched splenocytes were expanded at 37°C/7% CO₂ for 4–5 days before use in the cytotoxicity assay. Activated NK cell cultures were always greater than 90% NK1.1⁺ (B6) or NKG2D⁺ (BALB/c) and less than 5% TCR⁺.

T cells were isolated from splenocyte suspensions via the removal of B cells with anti-mouse B220 Dynabeads (Dynal, Oslo, Norway). For activation of 2C T cells, 2–3 × 10⁶ T cells were suspended in 5 ml complete medium containing 5–15 mouse units of Proleukin/ml and plated onto BALB/c or F₁ splenic adherent cells irradiated with 20 Gy γ-radiation in one well of a six-well plate. Plates were spun for 3 min at 700 rpm and incubated for various times at 37°C, 7% CO₂. If APC were allogeneic to the NK used in the killing assay, they were removed following T cell activation by staining with mAbs anti-H2-D^d (HB87; ATCC) and anti-H2-K^d (K9.18.10s; kind gift of W. Jefferies, University of British Columbia, Vancouver, British Columbia, Canada) and negative sorting with anti-mouse IgG Dynabeads (Dynal). For activation of D011 T cells, BALB/c plastic adherent spleen cells were first pulsed overnight with 1 mg/ml chicken OVA before addition of T cells in the same manner as that described above for 2C. Activation of polyclonal T cells with Con A was performed, as previously described (20), and cells were washed with 200 mM α-methyl-mannoside/α-methyl-glucoside before the cytotoxicity assay. To inhibit protein transport in T cells during activation so that APC were not affected, we used the same method described below for addition of BFA to the cytotoxicity assay.

Chromium release assays (CRA) were performed, as described previously (20). For assays in which Abs or Fc-fusion proteins were added that recognized ligands on the target cells, Ab-dependent cellular cytotoxicity was controlled by coating LAK with anti-FcRII/III Fab, or via the generation of LAK from FcRIIIγ^−/− animals (21). The use of BFA to disrupt the Golgi apparatus in target cells without affecting LAK effector function was performed, as previously described (22). Briefly, targets were first incubated in a high concentration of BFA (5 μg/ml for 45 min), a dose sufficient to disrupt the Golgi. Subsequent manipulation of targets and the cytotoxicity assay was performed at 0.5 μg/ml BFA. This concentration prevents the Golgi from reforming in the target cells, but does not affect the ability of LAK to kill. Conversion to lytic units (LU) was performed according to the procedure described by Bryant et al. (23).

Total RNA was prepared using RNeasy (Qiagen, Venlo, The Netherlands) and then reverse transcribed in a 20-μl reaction using an oligo(dT)_12–18 primer and SuperScript II (Life Technologies, Carlsbad, CA). Ten percent of the cDNA reaction was PCR amplified using the Expand High Fidelity PCR System (Roche, Laval, QC). The following primers were used: H60-F (5′-GTGTGATGACGATTTGTTGAG-3′) and H60-R (5′-ATTGATGGATTCTGGGCCATA-3′) (24). The β-actin control primer pair was purchased from BioSource International (Camarillo, CA). Cycling was completed as follows: 94°C for 30 s, and then 30 cycles of 94°C for 30 s, 52°C for 30 s, and 72°C for 45 s.

Murine NKG2D, corresponding to GenBank accession AF030313, was cloned into pMXs (generous gift of H. Arase and L. Lanier, University of California, San Francisco, CA) (25), a variant of pMX (26). This construct was transfected into the packaging cell line Plat-E (kind gift of T. Kitamura, University of Tokyo, Tokyo, Japan) (27) using Lipofectamine Plus (Life Technologies). Supernatants were harvested 48 h later and used to infect the human NK cell line YTSeco (generous gift of J. C. Zuniga-Pflucker, University of Toronto, Toronto, Ontario, Canada, and G. Cohen, MIT, Cambridge, MA) (28). This infection was conducted in the presence of polybrene at 10 μg/ml (Sigma-Aldrich). Infected cells were stained with PE-labeled αNKG2D mAb, and the positive fraction (YT-NKG2D) was sorted on a FACSVantage (BD Biosciences). The sequence of the integrated cDNA was verified using PCR primers specific for pMXs, as described (25).

All experiments performed in this study were reproduced at least twice.

Both CD4⁺ and CD8⁺ T cells of known specificity were separately activated by incubation for varying times on splenic adherent APCs and examined for their susceptibility to lysis by syngeneic LAK in standard CRA. OVA-specific CD4⁺ transgenic T cells were isolated from D011.10 (D011) TCR transgenic mice on the C.B-17 SCID background (identical with BALB/c, except for IgH locus). Similarly, transgenic CD8⁺ T cells specific for H-2L^d were isolated from 2C trangenic mice on the B6 background. Resting T cells (both CD4⁺ and CD8⁺) were highly resistant to lysis by syngeneic LAK at all E:T tested. However, upon activation on APC for 15 h, both CD4⁺ and CD8⁺ T cells became highly sensitive to syngeneic LAK (Fig. 1,A). When 2C T cells were activated on BALB/c APC for varying numbers of days, sensitivity was high on day 1, declined by day 4, and was nearly absent by day 7 (Fig. 1,B). If these T cells were subsequently reactivated on fresh APC for 24 h, sensitivity re-emerged (Fig. 1 C) and was quantitatively similar to that observed on day 1 of primary activation (≥14 LU for both 1° and 2°, day 1).

Acquisition of sensitivity to lysis appeared to require TCR engagement: when 2C T cells were activated on APC from DM2 mice that lack H-2L^d so that they could not be signaled through the 2C TCR, there was, at 6 h, no difference in sensitivity between these and resting T cells. In contrast, 2C T cells activated on wild-type H-2L^d+ APC (BALB/c) were 5-fold more sensitive to syngeneic LAK-mediated killing than resting controls (Fig. 1 D). The 2C T cells incubated on DM2 APC showed signs of viability loss after 6 h, so the experiment was not extended beyond this time point. D011.10 CD4⁺ T cells incubated on BALB/c APC that had not been pulsed with OVA (therefore no Ag) were also not lysed (unpublished observations).

MHC class I is well described as a pivotal NK inhibitory ligand (29). We therefore measured H2-K^b expression on resting and activated 2C T cells to test whether acquisition of sensitivity to lysis correlated with reduced expression of MHC-I. We found that MHC-I mean fluorescent intensity (MFI) increased from a relative value of 100 on day 0 to 257, 297, 248, and 214 on days 1, 2, 3, and 4 after activation, respectively.

LAK normally kill their targets by the perforin-dependent granule exocytosis pathway. To test whether killing of activated T cells was perforin dependent, we generated LAK from PKO mice (B6/H2^b) and used them as effectors against activated 2C T cells. Although lacking perforin, PKO LAK activate normally and have the capacity to kill through Fas (30). As predicted from previous experiments, 2C T cells activated for 20 h were >22 times more sensitive to syngeneic B6 LAK than resting controls. In contrast, lysis of these activated T cells by PKO LAK was only ∼1.2 times greater than for resting controls (Fig. 1 E).

To this point, we have shown that activation of T cells of known specificity on APC carrying the appropriate Ag induces sensitivity to lysis by syngeneic LAK. To address whether wild-type T cells would behave similarly, we activated T cells from B6 mice with Con A, PMA/ionomycin, or immobilized anti-TCR mAb. Compared with resting T cells (LU ≤0.5), T cells activated for 24 h with Con A, PMA/ionomycin, or immobilized anti-TCR mAb were highly sensitive to lysis (LU ≥18, 20, and 32, respectively). We found that coating of T cells with Con A resulted in very minor lectin-mediated killing, and this was reversed by standard treatment with α-methyl-mannoside (unpublished observations).

We hypothesized that sensitization might be due to transport of an NK activation structure to the cell surface through the Golgi apparatus. To test this directly, we blocked protein trafficking to the plasma membrane during activation of 2C T cells with the Golgi disrupting reagent, BFA, using a method that affected T cells, but neither APC nor LAK (22) (see Materials and Methods). After 20 h of activation, T cells activated in the presence of BFA were at least 2-fold less sensitive to lysis by syngeneic LAK than activated T cell controls (Fig. 1 F).

The activation receptor NKG2D is known to recognize several different inducible nonclassical MHC-I molecules and has been implicated in the lysis of syngeneic cells bearing normal levels of classical MHC-I. We tested D011 (BALB/c) T cells for staining with a soluble mouse NKG2D/human IgG1-Fc fusion protein. As seen in Fig. 2, A–C, they stained weakly when resting and intensely (>10-fold increase in MFI) after 24 h of activation, the latter largely inhibited when the cells were activated in the presence of BFA, consistent with Fig. 1,F. Staining intensity with NKG2D-Fc correlated with susceptibility to lysis by BALB/c LAK for D011 cells activated for 0 or 24 h in the absence or presence of BFA (Fig. 2 D). In particular, incubation with BFA reduced both sensitivity to lysis and binding of NKG2D-Fc. When D011 T cells were activated for up to 10 days, staining with NKG2D-Fc correlated with killing and predicted the transient nature of sensitivity (unpublished observations).

To test directly whether NKG2D was responsible for recognition of activated D011 T cells, we added anti-NKG2D mAb or the soluble NKG2D ligand, ULBP2 (14), to the CRA to test whether they would block lysis. ULBP2 was added conjugated to a leucine zipper motif (ULBP2-LZ). ULBP3-Fc was used as a negative control protein because it does not bind to mouse NKG2D (data not shown). Fig. 2 E illustrates the results obtained using syngeneic BALB/c LAK as effectors and D011 T cells, activated on Ag-pulsed APC (24 h), as targets at an E:T = 10:1. Activated T cells were highly sensitive to lysis when nothing was added to the CRA and at all concentrations of ULBP3-Fc. When either anti-NKG2D mAb or ULBP2-LZ was added to the CRA, lysis of activated T cells was abrogated. Resting T cells were highly resistant to lysis under all conditions.

As an even more direct test of whether mouse NKG2D was responsible for recognition of activated D011 T cells, we expressed a mouse NKG2D gene (GenBank accession AF030313) in the human NK cell line, YTSeco, using a pMX-derived retroviral vector, pMXs (25, 26). Independent experiments determined that YTSeco transduced with mouse NKG2D (YT-NKG2D) stained intensely with anti-mouse NKG2D mAb (MFI = 411), whereas parental YTSeco did not (MFI = 5). Fig. 2 F shows that YT-NKG2D lysed activated D011 T cells >25 times more effectively than the parental cell line. Addition of anti-NKG2D mAb to the CRA abolished lysis.

A known NKG2D ligand, H-60, can be expressed in BALB/c (but not B6). To test for its expression, we performed semiquantitative RT-PCR on resting and activated D011 T cells. As seen in Fig. 2 G, activated D011 T cells expressed higher transcript levels of H-60 compared with resting controls. Densitometric analysis of this and a duplicate experiment indicated that the level of H-60 message was 2- to 3-fold higher in activated cells.

We conclude that NKG2D is the NK receptor that recognizes activated BALB/c D011 cells and that H-60 may be its ligand. However, the situation is not so simple for lysis of activated B6 2C cells. B6 LAK express NKG2D (31) and stain with our anti-NKG2D mAb similarly to BALB/c, but addition of the mAb into the lysis assay had little effect (unpublished observations). Furthermore, addition of mAb against NKR-P1C, a known B6 NK activation receptor, had no effect (unpublished observations). Similarly, YT-NKG2D could not lyse activated 2C cells (unpublished observations). See Discussion.

In this study, we report that IL-2-activated NK cells recognize and lyse syngeneic T cells in a perforin-dependent manner within 15 h of activation on APC in both primary and secondary cultures. Sensitivity persisted for up to 4 days, required T-specific Ag, and reverted to resistance by day 7. For studies with APC, we used T cells from TCR transgenic animals specific for OVA (CD4) or H-2L^d (CD8) on the C.B-17 (identical with BALB/c, except for IgH locus) and B6 backgrounds, respectively. T cells became maximally sensitive to lysis before signs of polarization to the Th1 or Th2 phenotypes were detectable, and we have no data suggesting that this polarization is involved in the phenomenon. When nontransgenic T cells from BALB/c and B6 mice were activated with Con A, anti-TCR mAb, or PMA/ionomycin, sensitivity to lysis was acquired with similar temporal kinetics.

Lysis of activated T cells could be due to either loss of an inhibitory ligand or gain of an activating ligand. We first checked whether expression of MHC-I, the major inhibitory ligand for NK, decreased on activated T cells, but instead found slightly elevated levels relative to resting controls. This suggested that up-regulation of an activation structure was more likely responsible for lysis than down-regulation of an unknown inhibitory ligand. Consistent with transport of an NK target structure to the cell surface in response to activation, we found that inhibition of protein transport with BFA during T cell activation, in a manner that did not affect APC or LAK (see Materials and Methods) (22), largely inhibited the acquisition of LAK sensitivity (12-fold lower than activated control).

In several cases, it has been found that NK cells can kill targets expressing normal levels of MHC-I. Under these circumstances, activation signals received through NKG2D have been reported as responsible for tipping the balance toward lysis (10, 13). We found that staining of T cells from BALB/c (D011 or wild type) with a soluble mouse NKG2D-Fc fusion protein increased markedly on activation, correlated with sensitivity to lysis, and was inhibited by BFA. When we added either anti-NKG2D mAb or the soluble NKG2D ligand, ULBP2-leucine zipper fusion protein, to the CRA, lysis of activated T cells was reduced to levels observed for the resting controls, suggesting strongly that NKG2D is in fact the sole receptor responsible for killing of activated T cells in BALB/c, although we cannot rule out that other receptors may still be involved in conjugation (e.g., adhesion molecules) and augmentation of the activation signal (e.g., CD28). As further confirmation for the role of NKG2D, we transduced a mouse NKG2D gene (32) into a human NK cell line (YTSeco) with a pMX-based retrovirus (25, 26). Only cells transduced with mouse NKG2D killed activated D011 T cells, confirming directly that NKG2D can recognize activated BALB/c T cells.

We attempted to identify the NKG2D ligand involved in this system. The MHC-I homologue, H-60, was a good candidate because: 1) it is the only NKG2D ligand reported to be expressed on adult tissues, and 2) its expression has only been found in BALB/c (13). We measured H-60 transcripts by RT-PCR in resting and activated D011 T cells (Fig. 2 G). We found transcripts in both populations with an approximate 2- to 3-fold increase after activation. This may not be enough to explain the observed sensitivity to lysis because a 10-fold increase in NKG2D-Fc staining was seen after activation. We cannot rule out that other NKG2D ligands are involved.

Activated B6 T cells, 2C or wild type, were also extremely sensitive to lysis by syngeneic LAK. We tested whether NKG2D plays a role in the B6 system. Although H-60 is not expressed in B6, Carayannopoulos et al. (33) have recently described a novel ligand for NKG2D in B6, MULT1, murine ULBP-like transcript-1. We found that addition of neither anti-NKG2D mAb nor ULBP2-Fc to the CRA had an appreciable effect. Moreover, NKG2D-Fc did not bind B6-activated T cells, indicating that MULT1 is not involved. Nevertheless, NKG2D is expressed by B6 LAK at levels comparable to BALB/c (unpublished observations) and mediates redirected lysis of the FcR⁺ target P815 (unpublished observations). This suggests that B6 and BALB/c LAK use different receptors to recognize activated T cells. This may not be as unusual as it seems. NK receptor use differs broadly among inbred mouse strains in a manner that does not obviously correlate with the distribution of their known ligands. The apparent paradox between B6 and BALB/c, found in this study, becomes just one more of a growing number.

We observed that neither resting NK cells nor NKG2D⁺CD8⁺ T cells from either B6 or BALB/c could lyse syngeneic activated T cells (unpublished observations). This may be explained by the recent findings of Diefenbach et al. (34), who found that resting splenic NK cells predominantly express a long splice variant of NKG2D that associates exclusively with DAP10 and is not capable of activating NK cells without coligation of a second NK activation receptor (e.g., NKRP-1C in B6). A second, short splice variant of NKG2D was found to be expressed by poly(I:C) and IL-2-activated NK cells. This short form of NKG2D can associate not only with DAP10, but also DAP12, and appears to act as a bona fide activation receptor. In NKG2D⁺CD8⁺ T cells, the absence of DAP12 may render NKG2D unable to activate these cells without coligation of the TCR (34).

If a T cell can become a LAK target following activation, how might this affect an adaptive immune response? The answer to this question may lie within the realm of tissue distribution. In a recent report by Dokun et al. (35), the location of NK cells was examined during an immune response to murine CMV. NK cells were found excluded from T cell zones in the spleen and lymph nodes. This should protect T cells during and immediately after activation. In the liver and thymus, NK and T cells can potentially interact. This is interesting because following infection, NK cell numbers decrease in the spleen and peak in the liver after day 5 (35). Furthermore, the liver has been reported to trap activated CD8⁺ (36, 37) and CD4⁺ T cells (38, 39), both of which then undergo apoptosis by a still undefined mechanism, which we would speculate is NK cell mediated.

Several reports have indicated that NK cells activated by DC go on to limit immune responses by killing immature DC (7, 40). We suggest that these same NK cells can also directly limit the T cell arm of the adaptive immune response. If this occurs in vivo, our work has important implications for autoimmune and lymphoproliferative diseases (e.g., graft-vs-host disease (41)). SCID or RAG-deficient animals, which receive syngeneic CD4⁺CD45RB⁺ T cells, develop a proinflammatory autoimmune-like colitis. As a strong in vivo correlate of our findings, when NK cells are depleted from these animals or when recipients are perforin deficient, the colitis score increases >5-fold (42).

In conclusion, we believe the data presented in this work provide a framework for the pursuit of further study in this new area of NK cell biology.