CD94/NKG2 Expression Does Not Inhibit Cytotoxic Function of Lymphocytic Choriomeningitis Virus-Specific CD8⁺ T Cells¹

Murine Ag-specific CD8⁺ T cells express various NK markers and Ly49 NK inhibitory receptors (NK-IR).⁴ For instance, following infection of mice with either lymphocytic choriomeningitis virus (LCMV) or influenza A strain virus, subpopulations of CD8⁺ T cells express NK1.1, DX-5, asialo GM1, and various Ly49 inhibitory receptors (1, 2, 3, 4). In addition, Lohwasser et al. (5) recently showed that murine CD8⁺ T cells also express the NK receptor CD94/NKG2, the receptor that is the focus of this study.

CD94 covalently pairs with members of the NKG2 family (A, B, C, or E) via a disulfide bond to form CD94/NKG2 heterodimers (6, 7, 8, 9, 10). Immunoreceptor tyrosine-based inhibitory motifs possessed by NKG2A and NKG2B mediate signals that inhibit NK killing (11). In contrast, positively charged intracellular residues of NKG2C and NKG2E receptors bind to DAP-12 and transduce signals that induce NK killing when bound to its receptor ligand (11, 12, 13).

The ligand for CD94/NKG2 is Qa-1^b (8, 10). Qa-1^b is a nonpolymorphic (14), nonclassical MHC class I molecule encoded by the t23 gene of the murine H-2 complex (15). All tissues express Qa-1^b, though at much lower levels on the cell surface than classical class I molecules (16). Similar to classical MHC class I molecules, Qa-1^b binds peptides and β₂-microglobulin (16, 17). The preferred peptide ligand for Qa-1^b is the peptide Qa-1^b determinant modifier (Qdm), which is derived from the leader sequence of some, but not all, classical class I MHC alleles (17, 18). Despite its preference for binding Qdm, Qa-1^b can also bind to Salmonella and insulin peptides (19, 20).

Multiple laboratories have demonstrated that CD94/NKG2 expression on human CTL clones and lines inhibits cytolysis and cytokine secretion. Mingari et al. (21) first showed that anti-CD94 blocking Abs increased CTL killing and TNF-α secretion of a CD94-expressing T cell clone, BC2.4. Similarly, Noppen et al. (22) showed that killing by CD94^high but not CD94^low subclones of the CTL clone TRP-2 was enhanced by an anti-CD94 mAb. Concurrently, Le Drean et al. (23) showed that engagement of CD94/NKG2A inhibits TNF-α release and CTL activity using melanoma-specific human T cell clones. Finally, recent studies by Speiser et al. (24) yielded similar results to those above using tumor-specific CTL ex vivo.

Current studies in mice largely corroborate human research. Lohwasser et al. (5) found that Qa-1^b expressed on target cells partially inhibits target cell killing by allospecific CTL derived from a MLR. More recently, Moser et al. (25) showed that polyoma virus (PyV)-specific CTL that express CD94/NKG2 do not kill peptide-pulsed targets cells. Treating the target cells with anti-Qa-1^b Abs in combination with goat anti-mouse secondary Abs in the presence of brefeldin A reversed this inhibition at high E:T ratios.

Our study of CD94/NKG2 used the LCMV system. The murine LCMV infection model yields immune responses especially suited for the study of Ag-specific CD8⁺ T cells. Following acute LCMV infection, Ag-specific CD8⁺ T cells expand ∼10,000-fold, and the virus is cleared from most tissues within 8 days (26). CD8⁺ T cells respond against four major epitopes in C57BL/6 mice, glycoprotein-1 33–41 (gp33), glycoprotein-2 276–286 (gp276), nucleoprotein 396–404 (NP396), and glycoprotein 34–42 (gp34) (26, 27, 28). In BALB/c mice, most activated CD8⁺ T cells respond to the L^d-restricted peptide epitope nucleoprotein 118–126 (NP118) (26). With these models, we sought to determine the dynamics of CD94/NKG2 expression on CD8⁺ T cells following viral infection and to examine the effects of expression of this receptor on CD8⁺ T cell function. In contrast to previous studies (5, 21, 22, 23, 24, 25), we show that CD94/NKG2 expression is not sufficient to inhibit CTL killing or cytokine secretion in the LCMV model.

Six- to 8-wk-old C57BL/6, BALB/cJ, and C3H/HeJ mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Animal care was provided according to the guidelines of the Institutional Animal Care and Use Committee and Emory University (Atlanta, GA).

BALB clone 7 and Ag104 cells were grown in DMEM containing 10% FCS and penicillin/streptomycin antibiotics. Cells were grown at 37°C in 5% CO₂.

Naive mice were infected i.p. with 2 × 10⁵ PFU LCMV Armstrong strain. Following protocols that are standard in the field, LCMV-immune mice were subsequently challenged with 2 × 10⁶ PFU LCMV clone 13 i.v. (26, 29). To infect mice with Listeria monocytogenes (strain 45231; American Type Culture Collection, Manassas, VA), 2 × 10³ PFU were injected i.v. (30). For influenza virus infections, BALB/cJ mice were anesthetized and subsequently infected by intranasal inoculation with 50 μl of A/PR/8/34 (H1N1) virus diluted in PBS/2% BSA (0.005–0.01 LD₅₀; 0.00025 hemagglutinating units) (31). Polyoma infections were performed by injecting 2 × 10⁶ PFU of virus s.c. into the hind footpads of C3H/HeJ mice (25). To generate splenocytes, spleens from infected mice were ground on a metal screen and treated with ammonium chloride to lyse RBCs. Lung-infiltrating lymphocytes were isolated by collagenase D treatment of total lungs (32) followed by ammonium chloride treatment.

MHC tetramers were made as previously described (33). For Qa-1^b tetramer stains, single-cell splenocyte suspensions were stained with PerCP-labeled anti-CD8α (clone 57-6.7; BD PharMingen, San Diego, CA), allophycocyanin-labeled MHC class I tetramers, FITC-labeled anti-CD4 (negative gate; Coulter, Fullerton, CA), and PE-labeled Qa-1^b tetramers in 1× PBS containing 1% BSA fraction V (Sigma-Aldrich, St. Louis, MO). Excess free d-biotin (0.5 μM; Sigma-Aldrich) and unlabeled CT-CD8α (1/1600 dilution; Caltag Laboratories, Burlingame, CA) were added to staining solutions to eliminate binding of free biotinylated MHC monomers to free streptavidin and to inhibit binding of Qa-1^b to CD8, respectively (data not shown). All other cytometry reagents were purchased from BD PharMingen. Flow cytometry data were acquired on a FACSCalibur cytometer (BD Biosciences, Mountain View, CA) and analyzed using FlowJo software (Treestar, San Carlos, CA). Sorted T cells were obtained using a FACSVantage Cell Sorter (BD Biosciences) or a MoFlo (Cytomation, Fort Collins, CO).

Peptides were synthesized by B. Evavold (Department of Microbiology and Immunology Peptide Core Facility, Emory University) using F-moc chemistry using a Symphony/Multiplex Peptide Synthesizer and purified by HPLC (Rainin Instrument, Woburn, MA).

CD8⁺NP118⁺ T cells were sorted by flow cytometry. Total RNA was obtained using an RNAeasy kit (Qiagen, Valencia, CA). Total RNA was treated with DNase I (Roche, Basel, Switzerland) and reverse transcribed with Superscript II (Invitrogen, Carlsbad, CA) and a mixture of random hexamers and oligo(dT) primers. The resulting cDNA was amplified in an iCycler PCR machine (Bio-Rad, Hercules, CA) using the following NKG2 isoform-specific primer pairs: NKG2A/forward (FWD), GTTGTAATTACTACAGTTGCCACACCATATAACG; NKG2A/reverse (REV), CTGTGCTGAAGATAGAGTGTAGTTTATATCTCT; NKG2B/FWD, CAGAGAAACCTCATTGCTGGTACCCTGGGCCT; NKG2B/REV, CTGAAGATAGAGTGTAGTTTATATATGG; NKG2C/FWD, AATCTTGGAATGACAGTTTGGGGTCCTGCC; NKG2C/REV, CGGAAAATCCTGCTCCTGTTCACTATCTATGTG; NKG2E/FWD, TATTCTCACAATTGTTATTACATTGGCATGGAA; NKG2E/REV, GTCCATGAGACCAGTGAAAGGGATTGCAGAAAG; GAPDH/FWD, GGATGCAGGGATGATGTTC; and GAPDH/REV, TGCACCACCAACTGCTTAG. Each 50-μl quantitative PCR contained 50 nM forward and reverse primers, 0.1% Tween (Sigma-Aldrich), 5% DMSO (Sigma-Aldrich), 5 μg BSA (New England Biolabs, Beverly, MA), 2.25 mM MgCl₂, 0.1% SYBR green (BioWhittaker, Walkersville, MD), and 2 U of Taq polymerase (Promega, Madison, WI) (34). The amount of starting cDNA was determined using a standard curve derived from cloned plasmids. Results are expressed as a percentage of GAPDH amplification.

Intracellular IFN-γ staining was performed as previously described (26). For Qdm cross-titration experiments, 1 × 10⁶ splenocytes from day-30 LCMV-infected BALB/cJ mice were incubated in flat-bottom 96-well plates with different concentrations of the antigenic peptide LCMV NP118 (RPQASGVYM) and either the Qdm (AMAPRTLLL) or the Qdm variant peptide, Qdm·R5K (AMAPKTLLL), as indicated. Cultures were then incubated for 1 h at 37°C without brefeldin A, followed by a 5-h incubation in the presence of brefeldin A. After transfer to round-bottom 96-well plates, we stained for cell surface and intracellular Ags using an intracellular staining kit (BD PharMingen). Chronic LCMV and polyoma experiments were performed similarly using gp33 (KAVYNFATM), NP396 (FQPQNGQFI), or MT389 (RRLGRTLLL) peptides in place of NP118.

CTL assays and blocking of Qa-1^b on target cells were performed as previously described (25, 35). Briefly, target cells were loaded with chromium and incubated with antigenic peptide. For Qa-1^b blocking experiments, target cells were incubated with 20 μg/ml anti-Qa-1^b (BD PharMingen), washed, and then incubated with goat anti-mouse IgG (The Jackson Laboratory) in the presence of brefeldin A. Washed targets were incubated with dilutions of effector cells for 5 h at 37°C. CTL supernatants were spotted on 96-well ytrium silicate scintillator plates (Packard Instrument, Meriden, CT), dried overnight, and assayed using a 96-well plate gamma counter (Wallac, Turku, Finland).

Virus plaque assays were performed as described (29).

To test the specificity of the Qa-1^b/Qdm tetramer for CD94/NKG2 receptors expressed on CD8⁺ T cells, splenocytes from LCMV-infected mice were stained with Qa-1^b/Qdm tetramers and costained with either 18d3 or 20d5 mAbs. Previous studies showed that the 20d5 mAb blocks Qa-1^b/Qdm tetramer binding to NK cells (10). Consistent with this previous finding, the 20d5 mAb inhibited binding of the Qa-1^b/Qdm to CD8⁺ T cells (Fig. 1,B), while the CD94-specific mAb 18d3 did not (Fig. 1 A). This demonstrates that the Qa-1^b/Qdm tetramer binds specifically to CD94/NKG2 receptors expressed on CD8⁺ T cells. Identical results were observed in experiments using NK cells (data not shown).

To explore the dynamics of CD94/NKG2 expression on Ag-specific CD8⁺ T cells following viral infection, we harvested splenocytes from LCMV-infected BALB/cJ and C57BL/6 mice at multiple time points postinfection and assayed them for CD94/NKG2 expression using Qa-1^b/Qdm tetramers (Fig. 2, A and B). The frequency of Ag-specific CD8⁺ T cells that express CD94/NKG2 increases until the peak of the CD8⁺ T cell response at day 8 postinfection, at which point nearly all of the specific cells express CD94/NKG2. CD94/NKG2 expression on LCMV-specific CD8⁺ T cells in BALB/c and C57BL/6 mice is then uniformly maintained for at least 1 year. At all time points, CD94/NKG2 expression was restricted to CD44^high cells (Fig. 2 C), indicating that only activated T cells express this receptor following LCMV infection. CD4⁺ T cells did not express CD94/NKG2 at any time point following infection (data not shown). The demonstration that LCMV-specific CD8⁺ T cells express a molecule associated with inhibition of lytic function (at least in NK cells) begs the question of whether CD94/NKG2 receptors play a similar functional role at any stage of the in vivo T cell response.

The Qa-1^b tetramer binds to at least four NKG2 isoforms, but there are no NKG2 isoform-specific Abs available in the BALB/cJ model. To determine which NKG2 isoforms are expressed on LCMV-specific T cells, we performed a real-time quantitative PCR assay. NP118 tetramer-positive, Qa-1^b tetramer-positive T cells were sorted by flow cytometry at day 8 post-LCMV infection. Isolated total RNA was treated with DNase I, reverse transcribed into cDNA, and PCR amplified using NKG2 isoform-specific primer pairs. The amount of cDNA starting material for each NKG2 isoform was quantified using a standard curve derived from cloned NKG2 PCR fragments. Results are expressed as the percentage of GAPDH cDNA (Fig. 3). The results of this experiment show that LCMV-specific CD8⁺ T cells transcribe NKG2A, NKG2B, and some NKG2C transcripts but few, if any, NKG2E transcripts. This experiment demonstrates that LCMV-specific CD8⁺ T cells express more CD94/NKG2 inhibitory receptors than activating receptors.

Previous studies showed that LCMV- and influenza-specific CD8⁺ T cell populations express Ly49A or Ly49G2 NK receptors, respectively (2, 36). To determine whether LCMV-specific CD8⁺ T cells express Ly49 receptors, we stained splenocytes with mAbs specific for Ly49 A, D, F, G2, I/C, and I receptor isoforms at multiple time points following acute LCMV infection. We used the C57BL/6 mouse model for this experiment because most commercially available Ly49-specific Abs recognize the C57BL/6 Ags. In contrast to the previous results, we observed only slight staining of LCMV-specific CD8⁺ T cells with Abs specific for Ly49A or Ly49I/C, although each Ab effectively stained DX5⁺ NK cells from naive C57BL/6 mice (Fig. 4). We did observe a clear but low-frequency population of Ly49G2⁺CD8⁺ T cells, but only a small fraction (<1%) of the cells that stained with any of the three D^b tetramers also stained for Ly49G2. Previous studies of Ly49G2 expression on LCMV-specific CD8⁺ T cells did not report the fraction of LCMV-specific cells that were also Ly49G2⁺ (36). Likewise, memory CD8⁺ T cells did not appreciably express Ly49 receptors. In summary, the primary NK-IR expressed on LCMV-specific CD8⁺ T cells appears to be CD94/NKG2.

The newly characterized U5A2-13 mAb recognizes a complex comprised of three proteins (65, 34, and 31 kDa) expressed on NK and NKT cells (37, 38). We also stained splenocytes from LCMV-infected mice for the U5A2-13 Ag to determine whether Ag-specific CD8⁺ T cells also express this Ag following infection. We detected the NKT cell marker on Ag-specific cells at the peak and memory phase of the CD8⁺ T cell immune response (Fig. 4). The dynamics of the U5A2-13 Ag are similar to CD94/NKG2, demonstrating that activated CD8⁺ T cells express multiple Ags normally associated with the NK lineage. However, due to a dearth of information regarding the U5A2-13 Ag (37, 38), the significance of this result remains unknown.

To ascertain whether CD94/NKG2 expression is a common characteristic of Ag-specific CD8⁺ T cells, BALB/cJ mice were infected with L. monocytogenes and influenza virus. At multiple time points post-L. monocytogenes infection, K^d/LLO91 tetramer-positive CD8⁺ T cells stained positive for the Qa-1^b/Qdm tetramer (Fig. 5,A). Similarly, at day 9 post-influenza infection, K^d/HA533 tetramer-positive CD8⁺ T cells expressed CD94/NKG2 as detected by Qa-1^b/Qdm tetramer staining (Fig. 5,B). These results are analogous to previous experiments using LCMV virus (Fig. 2, A and B). Moreover, experiments using the vesicular stomatitis virus/C57BL/6 (data not shown) and PyV/C3H/HeJ viral infection models showed that CD8⁺ T cells specific for these viruses also express CD94/NKG2. These data imply that CD94/NKG2 expression may be a common characteristic of Ag-specific T cells following acute infection with different viral or bacterial pathogens.

To determine whether CD94/NKG2 expression is altered during CD8⁺ T cell secondary immune responses, LCMV-immune BALB/cJ mice were challenged with LCMV clone 13. Splenocytes from these mice were stained with L^d/NP118 and Qa-1^b/Qdm tetramers at multiple time points post-clone 13 challenge. At days 2 and 3 postchallenge, all Ag-specific CD8⁺ T cells continued to express CD94/NKG2 (Fig. 6,A). During this time, resting memory LCMV-specific CD8⁺ T cells are reactivated and regain the ability to lyse NP118 peptide-coated target cells (Fig. 6 B), which express Qa-1^b by both Ab staining and bioassay (data not shown) (20). In contrast to previous studies in the polyoma model system (25), these results clearly show that CD94/NKG2 levels are not altered by LCMV rechallenge, suggesting that down-regulation of CD94/NKG2 expression is not necessary to achieve re-expansion of memory populations and the development of effector functions.

Next, we sought to determine whether signaling through CD94/NKG2 inhibits production of IFN-γ by Ag-stimulated cells. At day 30 post-LCMV infection, we incubated BALB/cJ splenocytes with the NP118 peptide and with cross-titrations of either Qdm or the Qdm variant peptide, Qdm·R5K. The Qdm and Qdm·R5K peptides bind to Qa-1^b with similar affinity (9), but the Qa-1^b/Qdm·R5K complex is not recognized by CD94/NKG2 (Fig. 7,A). Because Qdm and Qdm R5K peptides quickly dissociate from Qa-1^b complexes (P. Jensen and J. Kraft, personal communication), we used excess concentrations of each Qdm peptide to ensure binding to Qa-1^b throughout the assay. Splenocytes were incubated for 6 h in the presence of brefeldin A and subsequently stained for intracellular IFN-γ. At all concentrations of NP118 peptide, splenocytes stimulated with Qdm or Qdm R5K-pulsed target cells yielded identical frequencies of responders with equal mean fluorescence intensities (Fig. 7 B). Similar results were obtained at day 8 post-LCMV infection (data not shown). Thus, CD94/NKG2 expression on LCMV-specific CD8⁺ T cells does not inhibit IFN-γ production in the LCMV system, even when the LCMV-specific cells are stimulated with suboptimal concentrations of their peptide Ag.

Moser et al. (25) recently showed that expression of CD94/NKG2 renders MT389-specific CD8⁺ T cells incapable of killing peptide-pulsed targets at day 8 post-PyV infection, while previous work by Lukacher et al. (39) showed that all D^k/MT389 tetramer-positive T cells are capable of producing IFN-γ at day 8 post-PyV infection. To explore this apparent paradox more closely, we stimulated splenocytes from day-8 PyV-infected C3H/HeJ mice with 10 μM MT389 peptide, in the presence of excess Qdm peptide to stabilize Qa-1^b in the presence of brefeldin A, and then stained cells for intracellular IFN-γ and surface NKG2A/C/E. As expected, all MT389-specific CD8⁺ T produced IFN-γ (Fig. 8,A) when compared with unstimulated splenocytes stained with D^k/MT389 tetramers (Fig. 8,B). We performed this experiment with the same concentration of MT389 peptide that was previously used to demonstrate that CD94/NKG2 inhibits killing by MT389-specific CD8⁺ T cells (25). Thus, we were not using a supraoptimal concentration of MT389 that would overwhelm CD94/NKG2 inhibition. This result shows that CD94/NKG2 does not inhibit IFN-γ production by PyV-specific CD8⁺ T cells. Together with studies demonstrating that CD94/NKG2 inhibits PyV CD8⁺ T cell killing (25), this result suggests that CD94/NKG2 differentially regulates CD8⁺ T cell effector functions in the acute polyoma model. We could not perform a peptide stimulation experiment as described in Fig. 7, because high concentrations of Qdm (AMAPRTLLL) or Qdm·R5K (AMAPKTLLL) peptide can bind to D^k and thereby compete with MT389 (RRLGRTLLL) due to peptide sequence homology (A. Lukacher and A. Byers, personal communication), yielding uninterpretable results.

To determine whether CD94/NKG2 expression affects the cytotoxic function of LCMV-specific T cells, we prepared splenocytes from BALB/cJ mice at 7 days post-LCMV infection, and FACS sorted L^d/NP118-specific CD8⁺ T cells into Qa-1^b/Qdm^high and Qa-1^b/Qdm^low populations. Next, we assayed the sorted T cell populations for their ability to lyse Qa-1^b+ BALB clone 7 target cells in a 5-h chromium release assay. At all E:T ratios, the Qa-1^b/Qdm tetramer-high and -low populations lysed target cells equally (Fig. 9). None of the T cell populations lysed uninfected target cells. These results were repeated in five separate experiments. Experiments using peptide-coated targets in lieu of virus-infected targets and experiments using targets cultured overnight in the presence or absence of IFN-γ yielded identical results (data not shown). These data show that CD94/NKG2 expression by CD8⁺ T cells is not sufficient to inhibit killing of virus-infected targets following acute LCMV infection.

Previous experiments demonstrated that blocking recognition of Qa-1^b on target cells dramatically increased PyV-specific CTL killing at high E:T ratios (25). We used this recently developed method for blocking recognition of Qa-1^b to determine whether CD94/NKG2 affects target cell killing by LCMV-specific T cells. At day 12 post-LCMV infection, splenocytes were isolated from LCMV-infected mice and added to Qa-1^b Ab-treated BALB clone 7 target cells in a 5-h CTL assay. The NP118-specific T cell response is restricted by H-2L^d. Therefore, we also used an anti-K^d mAb (SF1.1.1), which binds to BALB clone 7 cells but does not block the NP 118 response, as an Ab control. Experiments with Qa-1^b transfected human cells demonstrate SF1.1.1 does not cross-react with Qa-1^b (data not shown); experiments using other anti-K^d control Abs yielded identical results (data not shown). Our results show that LCMV-specific T cells lysed Qa-1^b-blocked target cells as efficiently as targets treated with the control SF1.1.1 mAb (Fig. 10). This result is consistent with earlier experiments in which CD94/NKG2 expression did not correlate with differences in CTL killing (Fig. 9), and suggests that CD94/NKG2 expression by CD8⁺ T cells is not sufficient to inhibit CTL killing in the LCMV model. Similar to previous studies (25), splenocytes from PyV-infected mice displayed an increased ability to kill Qa-1^b-blocked target cells as compared with control target cells treated with anti-K^k mAbs (Fig. 10).

Prior studies observed that chronically stimulated CD8⁺ T cells from tumor-bearing or HIV-infected patients express CD94/NKG2 (24, 40, 41, 42, 43). In the LCMV system, some strains of LCMV are capable of causing persistent infections (29). In the absence of CD4 T cell help, LCMV clone 13 infection results in the deletion of NP396-specific CD8⁺ T cells and the induction of gp33-specific CD8⁺ T cells that lack detectable function (44). To determine whether chronic infection causes up-regulation of CD94/NKG2, we assayed CD8⁺ T cells from chronically infected C57BL/6 for the expression of CD94/NKG2. At 30 days post-clone 13 infection, the majority of D^b/gp33-specific CD8⁺ T cells lacked CD94/NKG2 expression when compared with immune controls (Fig. 11,A). As shown previously (44), these cells lack the ability to produce IFN-γ following peptide stimulation (Fig. 11 B) when compared with immune controls. This result suggests that CD94/NKG2 is not responsible for the phenotype of anergic LCMV-specific CD8⁺ T cells and implies that chronic viral infection alone is not sufficient for CD94/NKG2 expression.

The literature has recently been flooded with reports of the expression of NK-IRs on Ag-specific CD8⁺ T cells (2, 21, 23, 24, 25, 36, 41, 43, 45, 46, 47). Most of these reports confirm the prediction that the inhibitory receptors exist to inhibit the effector functions of cytotoxic T cells, perhaps as a mechanism to reduce immunopathology following clearance of a pathogen or during chronic infection (40, 41). Most of these reports used established T cell lines or clones (21, 22, 23, 24, 41, 43, 47). This raises the possibility that the observations might be a reflection of in vitro culture conditions, but Moser et al. (25) recently reported that expression of CD94/NKG2 on polyoma-specific CD8⁺ T cells inhibited their cytotoxic function directly ex vivo. Using the well-characterized LCMV system, we have now shown that inhibition of cytotoxic function does not generally follow from expression of CD94/NKG2 on Ag-specific CD8⁺ T cells and that chronic infection does not necessarily lead to expression of CD94/NKG2 on Ag-specific CD8⁺ T cells. Raulet et al.⁶ have recently found similar results.

Our surprising data raise at least two conundrums. What advantage is gained by expressing inhibitory receptors on Ag-specific CD8⁺ T cells? Why do CD94/NKG2 inhibitory receptors not appear to affect function in the LCMV model, in contrast to the observations in the polyoma and other models?

Although we have observed that CD94/NKG2 expressed on LCMV-specific CD8⁺ T cells does not inhibit their lytic and IFN-γ-producing functions, we must still consider what potential advantage is gained by expressing inhibitory receptors on Ag-specific T cells, perhaps even using assumptions that CD94/NKG2 inhibits some function of these cells that we have not detected. CD8⁺ T cells that are chronically exposed to Ag can be functionally deficient (44, 48), and it has been proposed that this is a mechanism for avoiding immunopathology. Although the mechanisms by which this impotent phenotype is induced often aren’t known, expression of inhibitory receptors is one of a number of possible causes. The attraction of this model is enhanced by the observation that, until recently, most reports of NK-IR on CD8⁺ T cells were on cells that were responding to chronic stimuli, such as melanoma Ags or HIV (21, 22, 24, 41). However, our data, together with the recent data from the polyoma model (25), show that CD94/NKG2 inhibitory receptors can be up-regulated on murine CD8⁺ T cells during an acute primary immune response, demonstrating that chronic Ag exposure is not required for killer inhibitory receptor (KIR) expression. Furthermore, using the same chronic LCMV infection model in which Zajac et al. (44) found impotent Ag-specific CD8⁺ T cells, we found, to our surprise, that the D^b/gp33-specific CD8⁺ cells were present and lacked function, but they also lacked CD94/NKG2 expression. Of course, it remains possible that other, unidentified, inhibitory receptors are responsible for the unusual phenotype of the LCMV-specific cells seen in this model, but that is well beyond the scope of our present work.

In the context of an acute infection, it is difficult to understand what advantage might be gained by inhibiting the function of T cells that recognize a foreign Ag. Perhaps one solution to this problem follows from the observations that NK-IR tend to inhibit CTL function only at low doses of activating Ag (21, 22, 23), and even then KIR-mediated inhibition of lytic function is often only partial (21, 22, 23, 24). From these observations, a number of investigators have concluded that the function of KIR on CTL is to raise their threshold of TCR-mediated activation. Although we were unable to detect CD94/NKG2-mediated inhibition of LCMV-specific CD8⁺ T cell function, even at suboptimal LCMV-peptide doses, we cannot rule out the possibility that CD94/NKG2 acts to inhibit virus-specific T cells from killing cells that present low-affinity, otherwise cross-reactive self-peptides on their surfaces, thereby reducing potential immunopathology. In addition, although Ag-specific T cells predominantly express inhibitory receptor transcripts, we cannot discount a possible role of the CD94/NKG2C activating receptors. It is possible, although unlikely, that the net result of CD94/NKG2C activating signals and CD94/NKG2 A and B inhibitory signals is no signal to the cell. The development of assays that would allow determination of the expression of individual NKG2 isoforms at the single cell level would shed light on this problem.

Because we were already using the polyoma model as a positive control for our attempts to increase LCMV-specific killing by blocking the CD94/NKG2 ligand on target cells with anti-Qa-1^b plus a cross-linking secondary Ab (25), we also looked at the effect of CD94/NKG2 expression on IFN-γ production by peptide-stimulated, polyoma-specific CD8⁺ T cells. Previously, we had shown that at days 7, 9, and 12 postpolyoma infection of C3H/HeN mice, all D^k/MT389 tetramer⁺CD8⁺ T cells produce IFN-γ following peptide stimulation (39). Together with the more recent data from Moser et al. (39), which demonstrated that between 50 and 80% of D^k/MT389-specific CD8⁺ T cells express CD94/NKG2 at days 7–13 postinfection, we expected that CD94/NKG2 would not inhibit IFN-γ production by these cells, which is what we indeed found. In contrast to our previous IFN-γ production experiments, we performed these experiments in the presence of a vast excess of the Qdm peptide, counteracting the potential loss of the CD94/NKG2 ligand that might occur by dissociation of Qdm from Qa-1^b. Our IFN-γ production experiments in the polyoma model were performed at the same concentration of MT389 peptide that was previously used to demonstrate that CD94/NKG2 inhibits killing by MT389-specific CD8⁺ T cells (25) and that we repeated here (Fig. 10), demonstrating that we were not using a superoptimal concentration of MT389 in our experiments that would overcome CD94/NKG2 inhibition. In the polyoma model, we were unable to perform the same MT389/Qdm cross-titration experiment that we performed in the LCMV model, because at low concentrations of MT389 we cannot rule out that the Qdm peptide competes with MT389 binding to D^k. The discordant results that we obtained in the polyoma model (CD94/NKG2 mediated inhibition of lytic function by not cytokine production) are puzzling. Slifka et al. (49) have shown that direct ex vivo CTL retain their lytic function in the presence of cycloheximide, indicating that lytic function does not require de novo protein synthesis but that cytokine production by Ag-stimulated CD8⁺ T cells was inhibited by treatment with actinomycin D, indicating that cytokine production does require de novo mRNA synthesis. It will be interesting to determine how CD94/NKG2 can transduce signals that inhibit an effector function (lysis of target cells) that does not require de novo RNA and protein synthesis, while leaving intact an effector function (cytokine production) that does.

How, then, do we account for the differences between the ability of CD94/NKG2 to inhibit lytic function in the polyoma and LCMV systems? We have no satisfying resolution to this problem, but we can offer some speculations. The possibilities start with differences in virology, and different viruses (or different epitopes in the context of different viruses) may induce specific CD8⁺ T cells that have different phenotypes. This may also account for our observation that, during a secondary immune response to LCMV, CD94/NKG2 down-regulation on Ag-specific cells is not observed during viral clearance (Fig. 6,B, inset), in contrast to previous observations in which CD94/NKG2 down-regulation is observed on PyV-specific CD8⁺ T cells that are recalled with a recombinant vaccinia virus (25). It is also possible that the differences lie in the strains of mice that are used in the two systems. For example, CD94/NKG2-mediated inhibition of CD8⁺ CTL may be particularly effective in C3H/HeJ (or N) mice, as both recent reports of CTL inhibition—the polyoma system used by Moser et al. (25) and the allospecific CTL used by Lohwasser et al. (5)—used D^k-restricted T cells. The obvious experiments to perform include testing the ability of CD94/NKG2 to inhibit lytic function of 1) polyoma-specific CD8⁺ T cells in B6 and BALB/c mice, and 2) LCMV-specific CD8⁺ T cells in C3H/HeN or CBA/J mice. However, in both cases the relevant CTL epitopes have not been mapped, making it difficult to characterize the expression of CD94/NKG2 on virus-specific CD8⁺ T cells using MHC tetramer methods, as in Fig. 1. We are currently working with the A. Lukacher laboratory to map polyoma CTL epitopes in B6 mice, and we will soon begin to map LCMV epitopes in H-2^k mice (50, 51, 52). In addition, experiments in MHC congenic mice might shed light on whether the different susceptibility to NK-IR-mediated inhibition of function lies inside or outside the MHC.

We thank Sue Kaech for help with plaque assays; Charles Maris, Joshy Jacob, and John Wherry for help with chronic LCMV infections; Lisa Reed for Qa-1^b bioassays; Harriet Robinson for influenza virus; Michael Hulsey and Robert Karaffa for cell sorting; David Willer for GAPDH primers; and Mary Ann Skeen for L. monocytogenes. We also thank Peter Jensen, Rafi Ahmed, Janice Moser, and Aron Lukacher for providing both reagents and critical advice.