Single-cell analysis of the human NK cell response to missing self and its inhibition by HLA class I

Natural killer (NK) cells are innate immune lymphocytes active in host defense against pathogens and tumors. They mediate these activities through direct killing of transformed or infected cells and production of cytokines such as interferon-γ (IFN-γ), tumor necrosis factor-α (TNF-α), and granulocyte-macrophage colony-stimulating factor (GM-CSF).¹  NK cell functions are regulated by cytokine feedback loops, contact with dendritic cells, immune complexes, and direct interactions with the infected or transformed cells that serve as targets for their cytotoxicity.^2-4  NK cell specificity for diseased cells is determined by a balance of signals generated by diverse stimulatory and inhibitory receptors.⁵  Several inhibitory NK cell receptors have specificity for major histocompatibility complex (MHC) class I allotypes. These receptors survey for MHC class I molecules on potential targets, the NK cells being inhibited by cells with normal expression but responding to cells with compromised levels of class I on their surface.^6,7  In the context of allogeneic hematopoietic cell transplantation, the specificity of inhibitory NK cell receptors for class I can have significant impact on graft rejection, graft versus host disease (GVHD), and graft versus leukemia (GVL) effects.^8-10 

A key feature of NK cell class I receptor families is their extensive diversity, which plays out within individuals and across populations of individuals and species.^11-13  The distribution of multiple receptor families and multiple receptors within a family on NK cell subsets accounts for the diversity within an individual's NK cell receptor repertoire.^13-15  In humans, families of lectinlike and immunoglobulin (Ig)–like membrane glycoproteins share the responsibility of detecting host cell class I expression, thereby maintaining NK cell tolerance.¹³  The inhibitory receptors include the lectinlike receptor CD94:NKG2A that recognizes the nonclassical HLA-E class I molecule and is encoded in the NK complex (NKC) on chromosome 12.¹⁶  HLA-E is brought to the surface by association with peptides cleaved from the leader sequences of most classical HLA-A, -B, and -C allotypes.^17,18  In this way HLA-E expression is used by CD94:NKG2A as a sensor for the overall HLA class I level of a cell.

The Ig-like receptors are encoded by genes in the leukocyte receptor complex (LRC) of chromosome 19 and are divided into 2 families: killer Ig-like receptors (KIRs) and the leukocyte Ig-like receptors (LILRs).¹⁹  The LILRs (previously called immunoglobulin-like transcript [ILT] or leukocyte immunoglobulin-like receptor [LIR]) are broadly reactive toward HLA-A, -B and -C allotypes and, similarly to CD94:NKG2A, allow NK cells to survey for overall class I expression.^20-22  By contrast, individual KIR family members exhibit finer specificity for HLA class I allotypes and can distinguish between groups of HLA-A, -B, and -C allotypes.^13,23-29  The more focused specificity of inhibitory KIR governs human NK cell alloreactions and likely renders NK cells responsive to selective alterations to class I expression on virally infected and/or transformed cells.³⁰ 

Stimulatory members of the lectinlike (CD94:NKG2C/D) and Ig-like (KIR2/3DS) receptors have also been described, some of which have overlapping class I specificity and antibody epitopes with their inhibitory counterparts.¹⁴  The exact role of these stimulatory receptors remains a subject for debate; however, their capacity to activate NK cells is usually dominated by the negative signals generated through inhibitory KIR and/or CD94:NKG2A class I recognition.¹³  Demonstration that NKG2D recognizes MHC class I related chain A/B (MICA/B) and Rae1, and that a stimulatory class I receptor of mice recognizes a murine cytomegalovirus (MCMV) class I–like decoy protein, suggests that the stimulatory class I receptors may have evolved to detect class I–like structures encoded by genes located within and away from the MHC.^31-34 

The specificities of both the inhibitory and stimulatory members of KIR, ILT-2, and CD94:NKG2 receptor families have mainly been assessed with assays of NK cell function, although direct binding studies have also been used.^35-39  The functional determination of receptor specificity has usually employed cytolytic assays with NK cell lines or clones as effectors and class I–deficient cells transfected with individual class I alleles as targets. This approach has allowed the identification of the several KIR, LILR, and CD94:NKG2 specificities for HLA class I. Limiting the approach is its need for activated NK cell clones and lines derived from long-term culture in high doses of interleukin-2 (IL-2).^13,24-29  Given that a defining aspect of NK cell biology is the ability to work early in the immune response, we have investigated the HLA class I specificity of inhibitory NK cell receptors in short-term (12 to 24 hours) cytokine-cultured NK cells purified from the peripheral blood of human donors. By adapting the enzyme-linked immunospot (ELISPOT) technique for IFN-γ and granzyme B secretion and intracellular cytokine staining (ICS) for IFN-γ, we studied the response of NK cells to class I–deficient cells and their inhibition by HLA class I allotypes under conditions more like those of an early-stage immune response.

Epstein-Barr virus (EBV)–transformed B-cell lines, MHC class I–deficient 721.221, and 221-B*5801, -Cw*0304, -Cw*0401, and -Cw*0702 class I transfectants were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), 100 μg/mL l-glutamine, and an antibiotic mixture of penicillin and streptomycin. The expression of class I molecules in 721.221 transfectants was checked weekly by using W6/32 (anticlass I) antibody. Recombinant human IL-2 was kindly provided by Dr R. E. Aurigemma, National Cancer Institute, Frederick, MD. Recombinant human IL-12 was purchased from Sigma Aldrich (St Louis, MO).

Peripheral blood mononuclear cells (PBMCs) were purified from buffy coats derived from the blood of healthy donors and obtained from Stanford Blood Center (Stanford, CA) by separation on a Ficoll-Hypaque gradient. Genomic DNA was prepared from PBMCs using a QIAamp Blood kit (Qiagen, Valencia, CA). KIR typing of genomic DNA was performed as described previously.^40,41  KIR3DL1 allele typing was performed as described previously⁴²  with the addition of one primer pair (forward: 5′-CCATCGGTCCCATGATGCT-3′, reverse: 5′-ACGTTCATGGGCTCCCCG-3′) that amplifies 3DL1*002.

Purified, resting NK cells were obtained from frozen PBMCs by magnetic depletion of T and B lymphocytes and monocytes using a cocktail of biotin-conjugated monoclonal antibodies against CD3, CD4, CD14, CD15, CD19, CD36, and CD123, antibiotin microbeads, LS separation columns, and midiMACS magnet (Miltenyi Biotec, Auburn, CA). After depletion, NK cell preparations (CD56⁺/CD3^–) were from 85% to 95% pure, and contamination with CD14⁺, CD19⁺, and CD3⁺ cells was less than 1% and was, for lineage-negative cells, 5% to 15%.

ELISPOT plates and reagents for IFN-γ and granzyme B were purchased from BD Biosciences (Mountain View, CA). For each set of conditions the number of NK cells per well was established in preliminary experiments. In the presence of IL-2 or IL-12, NK cells were 2 × 10³ to 3 × 10³ NK cells per well and 2 × 10⁴ to 3 × 10⁴ cells per well for IFN-γ and granzyme B ELISPOT, respectively. In the absence of cytokines, the NK cell number was 2 × 10⁴ to 3 × 10⁴ NK cells per well and 2 × 10⁵ NK cells per well for IFN-γ and granzyme B ELISPOT, respectively. NK cells were mixed with 721.221 target cells (parental or singly transfected with class I alleles) at a ratio of 1:10 and plated in duplicate or triplicate in 200 μL assay medium per well at 37°C and 7.5% CO₂ for 20 to 24 hours of stimulation. Control wells were either targets or effectors alone in media in the presence or absence of cytokines. In blocking experiments, target cells (221 or 221-transfectants) were first incubated for 1 hour with monoclonal antibody (mAb) DX17 (anti-HLA class I, kindly provided by Dr L. Lanier, University of California, San Francisco) before the ELISPOT assay was performed. When GL-183 (anti-KIR2DL2/3; Beckman-Coulter, Brea, CA) was used, NK cells were first incubated for 1 hour with mAb and then targets were added. ELISPOT assays were performed according to the manufacturer's instructions. The spots were counted off the images of each well that were obtained from scanning and analysis of the plates with an ImmunoSpot Series 2 Analyzer (Cellular Technology, Cleveland, OH). The ImmunoSpot Series 2 Analyzer consists of an analog camera (Sony DXC-390). The images were scanned with the ImmunoSpot Image Acquisition software version 3 and were analyzed with ImmunoSpot Analysis software version 2.08 with a resolution of 490 × 480 pixels per well.

For the analysis of NK cell receptor expression, mAbs EB6-PE (anti-KIR2DL1/KIRDS1), GL-183–phycoerythrin (GL-183–PE) (anti-KIR2DL2/KIR2DL3/KIR2DS2) (Beckman-Coulter-Immunotech), DX9-PE (anti-KIR3DL1; BD Biosciences), and DX31-PE (anti-KIR3DL2; generously provided by Dr J. H. Phillips, DNAX Research Institute, Palo Alto, CA) were used in combination with Z199 (anti-NKG2A; Beckman-Coulter-Immunotech) conjugated with Alexa Fluor 488 by using a monoclonal antibody labeling kit (Molecular Probes, Eugene, OR) and with CD85j-CyChrome (anti-LILRB1; BD Biosciences) in 3-color flow cytometry (FACScan; BD Biosciences) of purifed NK cells from each donor. Analysis was performed using CellQuest software (BD Biosciences). Subsets of KIR⁺ CD94:NKG2A⁺, KIR⁺ CD94:NKG2A^–, and KIR^– CD94:NKG2A⁺ NK cells were detected in FL1 and FL2 channels. Subsets of LILRB1⁺ KIR^– CD94:NKG2A^– NK cells were detected by gating on the KIR^– LILRB1⁺ (FL2 and FL3) and subsequent analysis of that population for expression of CD94:NKG2A and LILRB1 (FL1 and FL3). Depletion of CD94:NKG2A⁺ NK cells was obtained by sorting of purified NK cells using the Z199 antibody. Experiments with 5(and-6)–carboxyfluorescein diacetate succinimidyl ester (CFSE) were performed by labeling NK cells with 10 μM CFSE (Molecular Probes) followed by incubation at 37°C at a cell density of 2 × 10⁷/mL in Hanks balanced salt solution (HBSS) without FBS. After 10 minutes, cells were washed in cold HBSS supplemented with 20% serum and then cultured in RPMI, 10% FBS at 2 × 10⁵/200 μL per well. Positive controls for the assay were CFSE-labeled PBMCs from the same donor stimulated with 5 μg/mL phytohemagglutinin (PHA). Cells were harvested after 1, 3, and 6 days and analyzed by flow cytometry.

Purified NK cells were mixed with target cells (721.221 or 221 transfected with class I) or without target cells (as control) in medium alone or with IL-2 (1000 to 3000 U/mL) at an effector-target (E/T) ratio of 1:1 in sterile round-bottom 96-well plates, centrifuged at 100 g for 1 minute, and incubated at 37°C and 7.5% CO₂ for approximately 12 to 14 hours. Higher doses of IL-2 were required for ICS experiments than those used for ELISPOT studies. Golgi Plug (1:1000) containing brefeldin A (BD Biosciences) was added 1 hour after the mixing of effectors and targets to inhibit cytokine secretion. Cells were prepared for flow cytometry using the protocol developed by BD Biosciences with slight modification. Briefly, the dead cell discrimination reagent (Miltenyi Biotec) that distinguishes dead cells in fixed samples was added to the cells. After incubation for 10 minutes under a light source for photolabeling of dead cells, antibodies for NK cell receptors were added for 30 minutes in the dark at 4°C. Cells were fixed and permeabilized using Cytofix/Cytoperm solution (BD Biosciences), and IFN-γ–fluorescein isothiocyanate (IFN-γ–FITC) (clone B27; BD Biosciences) was added for 30 minutes at 4°C in the dark. Three-color flow cytometry (FACScan; BD Biosciences) was performed. The results were analyzed by using FlowJo software. A gate was drawn on NK cells with exclusion of dead cells (in FL3-positive channel) and target cells (based on side and forward scatter).

As a modification of the methods used to calculate inhibition of NK clone cytotoxicity by class I–bearing targets, we used 2 simple calculations for our ELISPOT and ICS results. (1) ELISPOT: 1 – (no. of spots following incubation with 221 class I transfected/no. of spots following incubation with 221 parental cells) × 100 = percent inhibition; and (2) ICS: 1 – (% of IFN-γ–positive NK cells following incubation with 221 transfectants/% of IFN-γ–positive NK cells following incubation with 221 parental) × 100 = percent inhibition.

As a new approach for analyzing the regulation of NK cells by MHC class I, we adapted the ELISPOT technique to measure the IFN-γ and granzyme B secreted by individual cells in a bulk NK cell response to the class I–deficient B-cell line 221. Granzyme B and IFN-γ were chosen as markers for the 2 major NK cell functions: cytotoxicity and cytokine secretion. The number of NK cells secreting IFN-γ or granzyme B was increased by inclusion of either IL-2 or IL-12 in the assay culture, with IL-12 consistently being more potent than IL-2. (Figure 1A-B).

Human NK cell clones and lines are frequently susceptible to inhibition by HLA-C, because members of all 3 inhibitory NK cell receptor families recognize HLA-C allotypes.⁴³  KIRs distinguish 2 groups of HLA-C allotypes defined by presence of either asparagine or lysine at position 80. KIR2DL1 recognizes allotypes having lysine at position 80, which are represented in this study by Cw*0401. KIR2DL2 and KIR2DL3 recognize allotypes having asparagine at position 80, which are represented here by Cw*0304 and Cw*0702.²⁴  Both LILRB1 (previously called LIR-1 or ILT2) and CD94:NKG2A recognize a broad range of HLA-C allotypes, the latter through HLA-C leader peptides bound to HLA-E.^17,20,21,44  Because of the prevalence of HLA-C–mediated inhibition, it was first examined using the ELISPOT system.

By comparison with untransfected cells, 221 cells transfected with HLA-Cw*0304 (221-Cw*0304) stimulated fewer NK cells to secrete IFN-γ (Figure 1C) or granzyme B (Figure 1D). For both the IFN-γ and granzyme B response, the HLA-C–mediated inhibition was slightly less for NK cells cultured with IL-2 than for those cultured with IL-12 (Figure 1D-E). IL-12 was also more effective than IL-2 in raising the frequency of NK cells that responded to 221 cells (Figure 1A-B). Analogous experiments, in which 221 transfected with HLA-Cw*0401 (221-Cw*0401) was used to inhibit the NK cells from several donors, gave similar results (data not shown). These observations show that cytokine production (eg, IFN-γ) and cytotoxicity (eg, granzyme B) are inhibited to similar effect when inhibitory NK cell receptors (KIR, CD94:NKG2A, LILRB1) engage HLA-C ligands on target cells. Because it required fewer NK cells than the granzyme B ELISPOT, the IFN-γ ELISPOT was used as the main assay of effector function for most subsequent experiments in this study.

Finding increased sensitivity of IL-12–treated NK cell populations to class I inhibition led us to seek for changes in inhibitory NK cell receptor expression that occur on culture with IL-12. Comparison of KIR, CD94:NKG2A, and LILRB1 expression by purified NK cells following incubation with IL-12 revealed an increased frequency of CD94:NKG2A expression (Figure 2A) but not for KIR or LILRB1 (data not shown). When NK cells were similarly incubated with IL-2 there was no significant change in CD94: NKG2A expression (Figure 2B) or in KIR or LILRB1 (data not shown).

To assess the phenomenon further, purified NK cells were first depleted of CD94:NKG2A⁺ cells and then incubated with increasing concentrations of IL-12 for 1 day. After this treatment, more than 20% of the cells expressed CD94:NKG2A (Figure 2C). Extending the IL-12 treatment for several days led to further increase in the frequency of CD94:NKG2A⁺ NK cells (data not shown). Uncertain was whether the effect of IL-12 was to induce expression of CD94:NKG2A in NK cells that lack this receptor or to drive the proliferation of residual CD94: NKG2A⁺ NK cells that escaped depletion. When CFSE labeling was used to assess for proliferation in cultures depleted of CD94:NKG2A⁺ NK cells, we obtained no evidence that IL-12 had induced cell division in the population as a whole or any particular subset (data not shown). In contrast, IL-2 did induce cell division in such cultures but without a major increase in the frequency of CD94:NKG2A⁺ NK cells. These results, and the short exposure (20 to 24 hours) necessary to see the effect, favor the interpretation that IL-12 induces CD94:NKG2A expression on a subset of NK cells lacking detectable levels of this receptor.

To assess the contribution of different receptors to HLA-C–mediated inhibition of NK cell function, we sought donors with KIR genotyes that lack the stimulatory KIRs (KIR2DS1, 2DS2, and 2DS3). These stimulatory KIRs could have complicated the analysis in 2 ways: first, by interacting with HLA-C to activate NK cells and, second, by cross-reacting with the monoclonal antibodies used to detect inhibitory HLA-C–specific KIRs.⁴⁵ 

NK cells from the selected donors (Figure 3) were purified and challenged with either 221, 221-Cw*0304 (Figure 4A), or 221-Cw*0401 (Figure 4C) in the presence of IL-2, IL-12, or no cytokine. NK cell secretion of IFN-γ in these cultures was assayed by ELISPOT (Figure 4A,C). The purified NK cells (CD3^–CD56⁺) from each donor were also subjected to 3-color flow cytometric analysis to determine the frequency of NK cells expressing HLA-C–reactive inhibitory receptors: KIR, CD94:NKG2A, and LILRB1 (data not shown). The proportion of cells having one or more inhibitory receptors that can bind to the inhibiting HLA-C allotype gave a prediction for the frequency of NK cells that would be inhibited when challenged by that allotype. Such predictions were made from analysis performed on freshly isolated NK cells as well as cells cultured for 20 to 24 hours with cytokines. The proportion of NK cells expressing at least one inhibitory receptor for class I did not change following incubation in the absence of cytokines or with IL-2, whereas culture with IL-12 gave higher values for some HLA class I allotypes (eg, for donor 11; see the legend to Figure 4) because of the increased frequency of expression of CD94:NKG2A induced by IL-12 (Figure 2).

For NK cells cultured with IL-2 or IL-12 there was fair agreement between the frequency of inhibited NK cells observed in the ELISPOT assay and the frequency of inhibitable cells predicted from the flow cytometric analysis (Figure 4B,D). In contrast, the observed inhibition of NK cells cultured without cytokine was greater than predicted from the distribution of inhibitory HLA-C–reactive receptors (Figure 4B,D).

In continuing to study this difference, we further simplified the experimental system so that the action of a single inhibitory receptor could be studied. This was achieved by using HLA-Cw*0702 as the inhibitory ligand; it interacts with inhibitory KIR but does not furnish an HLA-E binding peptide to make a CD94:NKG2A ligand.^13,17  In addition, LILRB1 does not interact with HLA-Cw*0702.⁴⁴  When 221 transfectants expressing this allotype were used as inhibitors for NK cells from donors selected for KIR genotype that lacked cross-reactive stimulatory receptors, inhibition was predicted to be due to a single inhibitory KIR.

HLA-Cw*0702 is a ligand for KIR2DL2 and 2DL3.^13,25,50  Because donor 17 lacked genes for KIR2DL2 and the stimulatory receptors KIR2DS2/2DS3, the GL-183 monoclonal antibody could be used to enumerate NK cells expressing KIR2DL3. An experiment analogous to that shown in Figure 4 was performed with 221-Cw*0702 cells as inhibitors, and similar results were obtained. For NK cells incubated with IL-2 or IL-12, the observed Cw*0702-mediated inhibition of NK cells corresponded well with that predicted from the frequency of KIR2DL3 expression (Figure 5). For NK cells incubated without cytokine the observed inhibition was much greater than that predicted. That inhibition in this assay was dependent upon both HLA class I and KIR2DL3 was demonstrated by adding mAb specific to either class I (Figure 5D) or KIR2DL3 (Figure 5E), both of which reversed the inhibitory effect. Our results with unstimulated (resting) NK cells indicate that, in the absence of proinflammatory cytokines, NK cell populations are more susceptible to inhibition by class I as compared with that observed previously for NK cell lines and clones.

A flow cytometric approach was also used to assess the frequency and phenotype of NK cells secreting IFN-γ following a 12-hour incubation with target cells. As shown in Figure 6, this method was used to examine the responses of KIR2DL3⁺ and KIR2DL3^– NK cells to 221 cells and their inhibition by HLA-Cw*0702. Similar proportions of KIR2DL3⁺ and KIR2DL3^– NK cells made IFN-γ in response to 221 cells (Figure 6A, center panels), whereas few cells responded in the absence of target cells (Figure 6A, left panels). NK cells incubated with (Figure 6A, lower panels) or without IL-2 (Figure 6A, upper panels) showed the effect, but the frequency of cells responding with IL-2 was higher, as expected from previous ELISPOT analysis (Figures 1, 4, and 5). On culture with 221-Cw*0702 cells, the frequency of responding KIR2DL3^– NK cells was comparable to that against 221, whereas the KIR2DL3⁺ NK cells were mostly inhibited (Figure 6A, right panels). The results for assays containing IL-2 demonstrate that Cw*0702-mediated inhibition of NK cells was restricted to the subpopulation expressing the cognate KIR2DL3 receptor (Figure 6B, lower panel). Although a similar bias was seen for NK cells challenged in the absence of cytokine, there was also some Cw*0702-mediated inhibition of KIR2DL3^– NK cells (Figure 6B, upper panel). In these assays the expression of KIR2DL3 was not affected by the presence or the nature of the target cell or by the presence or absence of IL-2 (Figure 6C). Neither was HLA class I expression by the target cells affected by the presence or absence of IL-2 (data not shown).

A feature of the polymorphism of KIR3DL1 is that some allotypes are distinguished by different levels of cell surface expression.^51,52  Flow cytometric analysis of IFN-γ production by NK cells provided a method for direct comparison of the functional potential of allotypes expressed at different levels. The analysis of 2 donors, both of whom are KIR3DL1 heterozygotes, is shown in Figure 7. Donor 35 is heterozygous for the KIR3DL1*001 and KIR3DL1*002 alleles, both of which are expressed at high level (Figure 6A, left panel). Donor 43 expresses KIR3DL1*002 at high levels and KIR3DL1*005 at low levels. For donor 43, the histogram of NK cell staining with anti-KIR3DL1 mAb (DX9 antibody) comprised 3 subpopulations of cells: negative cells, cells with low mAb binding that express KIR3DL1*005, and cells with high binding that express KIR3DL1*002 (Figure 7B, far left panel). In addition, a small fraction of the high-binding cells are expected to express both alleles.⁵¹ 

Incubation of purified NK cells with IL-2 or IL-12 had no effect on the pattern of DX9 binding, indicating that the frequencies and the levels of allelic expression were maintained (data not shown). NK cells were cultured with 221 or 221-B*5801 and assessed by flow cytometry for production of IFN-γ. Costaining was performed with DX9 and anti-LILRB1 (mAb CD85j) so that the analysis could be directed at NK cells that express KIR3DL1 but not LILRB1, to which HLA-B*5801 also binds.⁵³ 

On incubation of donor 35's NK cells with 221 cells, some 21.4% of the NK cells responded by making IFN-γ, a response not seen in the absence of target cells, and was inhibited by 90% when B*5801-221 was the target. On incubation of donor 43's NK cells with 221 cells, IFN-γ was made by NK cells with high and low levels of KIR3DL1 expression. The frequency of the response was higher for NK cells expressing a low level of KIR3DL1 (17.8%) than for those expressing a high level (10.6%) In contrast, few cells (less than 1%) made IFN-γ in the absence of a target cell. NK cells incubated with 221-B*5801 were largely inhibited in their IFN-γ response, and similar inhibitory effect was observed for cells with high (89% inhibition) and low (82% inhibition) levels of KIR3DL1 (Figure 7, far right panel). This result shows that both the high-expressed allotype (3DL1*002) and the low-expressed allotype (3DL1*005) bind to HLA-B*5801 and deliver inhibitory signals with comparable functional effect.

NK cells are nonresponsive to healthy autologous cells, because healthy cells lack the ligands for activating NK cell receptors and their MHC class I molecules engage inhibitory receptors that prevent NK cell activation.^7,54,55  Previously, we studied NK cell cytotoxicity and its inhibition by HLA class I using established clones of IL-2–cultured human NK cells clones.¹³  Here, we have used the ELISPOT and ICS techniques to study NK cell functions and their inhibition by HLA class I in short-term cultures (12 to 24 hours) of NK cells isolated from peripheral blood. Comparison of cytotoxicity and cytokine production, as assessed by granzyme B and IFN-γ secretion, respectively, gave similar results. When NK cells were cultured short-term with IL-2 or IL-12 they were found to have inhibitory specificities for HLA class I that were the same as those previously defined for NK cell clones in long-term cultures.^13,23-29  Thus there was concordance between the ELISPOT/ICS approach and the more established methods. That NK cell clones had the same HLA class I specificities as NK cells activated with a tempo closer to that of an in vivo innate immune response increases confidence in the results previously obtained with NK cell clones.

Previous reports have shown that IL-12, IL-15, and transforming growth factor-β (TGF-β) induce the expression of CD94/NKG2A on activated human T lymphocytes.^56-58  Analogous to this, we observed that culture with IL-12 increased the frequency of NK cells expressing the CD94:NKG2A receptor, while IL-2 had minimal effect on receptor expression. Although not definitive, our results point to this increase being due to induction of CD94:NKG2A expression rather than selective expansion of cells expressing this receptor. In contrast, no change in either KIR or LILRB expression was detected, as previously reported.⁵⁹  Both IL-2 and IL-12 activate NK cell functions in similar ways; however, IL-12 acts early on NK cell functions in vivo, whereas IL-2, a product of antigen-specific T_h cells, acts later.^3,60  IL-12 is produced by diverse populations of antigen-presenting cells, being secreted rapidly and in large quantities by dendritic cell (DC) subsets.^61-63  By increasing the abundance of NK cells expressing CD94: NKG2A, an inhibitory HLA class I receptor with broad allotype specificity, the IL-12 made early in the immune response might serve to activate NK cells while simultaneously increasing the inhibitory potential of certain subsets through their increased CD94:NKG2A expression. This dual effect of IL-12, a product of DCs, could impact the outcome of NK-DC encounters by altering the stimulatory/inhibitory balance of the “NK-DC control switch.”^64-66  Indeed, CD94:NKG2A-expressing NK cell populations have been shown to interact with autologous DCs differently from that observed for KIR⁺ subsets.⁶⁷ 

For the NK cells cultured with IL-2 or IL-12, the frequency of cells that were inhibited by a particular HLA class I allotype correlated well with the frequency of NK cells that carried an inhibitory KIR specific for the allotype. This was not the case for “control” NK cells cultured in the absence of cytokine. These NK cells secreted IFN-γ in response to class I–deficient 221 cells but less so than cells cultured with IL-2 or IL-12. In addition, these resting populations were also more sensitive to inhibition by HLA class I than their cytokine-treated counterparts. That the frequency of resting NK cells inhibited by a particular HLA class I type was in excess of the frequency of cells expressing an inhibitory KIR specific for the allotype—and this was altered by cytokine treatment—suggests that NK cell sensitivity to class I inhibiton is not fixed. This was seen most clearly in our flow cytometric analyses of NK cell IFN-γ production, where a proportion of resting NK cells lacking the HLA-Cw*0702 inhibitory receptor, KIR2DL3, were still inhibited by 221 target cells expressing HLA-Cw*0702 (Figure 6). Taken as a whole, the ELISPOT and ICS analyses of NK cell populations indicate that the balance of stimulatory and inhibitory signals delivered to NK cells upon target cell encounter is modified rapidly by IL-2 and IL-12. Indeed, after only 12 to 24 hours of culture both cytokines elicit NK cell populations exhibiting class I specificity remarkably similar to that observed for long-term–cultured NK cell lines and clones. From the perspective of NK cell responses in vivo, our results argue that resting NK cell interactions with inhibitory class I allotypes are more likely to maintain tolerance to autologous cells in the absence of pathologic conditions (eg, infection or malignant transformation) when cytokines are not produced. Moreover, resting NK cells might also be more sensitive to inhibition by low levels of class I expression such as those found on nonhematopoietic cells (hepatocytes, neurons).

KIR3DL1 is a highly polymorphic locus for which certain allotypes have different levels of cell surface expression, and the functional implications of these expression levels was not clear. Using ICS to analyze NK cells from KIR3DL1 heterozygotes, we could assess and compare the HLA-B–mediated inhibition for different KIR3DL1 allotypes. Both high-(3DL1*002) and low-expressing allotypes (3DL1*005) gave similar frequencies of NK cell inhibition upon engagement with HLA-B*5801. With the ELISPOT and ICS methods, HLA class I–mediated inhibition and other aspects of NK cell function and specificity can be determined at high resolution using fewer cells in mixed populations and without cloning. In particular, the low cell numbers and the ease and speed of the methods should be amenable to population analysis and clinical studies.