|Home | About | Journals | Submit | Contact Us | Français|
Daniel Korbel, Centre for Gastroenterology, Institute of Cell & Molecular Science, Barts & The London School of Medicine & Dentistry, 4 Newark Street, London E1 2AT, UK. firstname.lastname@example.org
IFN-γ emanating from natural killer (NK) cells is an important component of innate defence against infection. Here we demonstrate that, following in vitro stimulation of human peripheral blood NK cells with a variety of microbial ligands, CD56dim as well as CD56bright NK cells contribute to the overall NK cell IFN-γ response with, for most cell donors, IFN-γ+ CD56dim NK cells outnumbering IFN-γ+ CD56bright NK cells. We also observe that the magnitude of the human NK IFN-γ response to microbial ligands varies between individuals; that the antimicrobial response of CD56bright, but not CD56dim, NK cells is highly correlated with that of myeloid accessory cells; and that the ratio of IFN-γ+ CD56dim to IFN-γ+ CD56bright NK cells following microbial stimulation differs between individuals but remains constant for a given donor over time. Furthermore, ratios of IFN-γ+ CD56dim to IFN-γ+ CD56bright NK cells for different microbial stimuli are highly correlated and the relative response of CD56dim and CD56bright NK cells is highly significantly associated with killer immunoglobulin-like receptor (KIR) genotype. These data reveal an influence of KIR genotype, possibly mediated via NK cell licensing, on the ability of NK cells to respond to non-viral infections and have implications for genetic regulation of susceptibility to infection in humans.
Innate immune mechanisms represent the first line of defence against invading pathogens, limiting acute infection as well as triggering and directing the adaptive response. NK cells can play a crucial role in the innate response to infection by lysis of infected cells and by secretion of pro-inflammatory cytokines, such as IFN-γ, that promote phagocytic clearance of microbes and guide T cell differentiation (1, 2). On the other hand, dysregulated innate inflammatory responses may predispose to onset of immune-mediated pathology (3–5). Knowledge of the mechanisms of induction and regulation of the NK IFN-γ response may aid our understanding of the immunopathological basis of disease, identify novel adjuvant strategies for immunisation or new targets for immunotherapy and inform approaches for controlling unwanted inflammation.
NK cell activity is regulated by an extensive repertoire of regulatory receptors which may be constitutively expressed or upregulated upon activation and which bind to constitutively expressed or stress-induced self ligands (6). The most polymorphic of these receptors belong to the killer immunoglobulin-like receptor (KIR) family (7) and a number of studies implicate KIR diversity (and concomitant diversity among their MHC Class I ligands) in susceptibility to both infectious and non-infectious diseases (7, 8). These studies have primarily been interpreted in the context of the “missing self”-model of NK activation (9) in which down-regulation of MHC Class I on diseased cells leads to their lysis by NK cells. However, many different classes of pathogens (bacteria, protozoa, fungi and helminths as well as viruses) are able to activate NK cells; downregulation of MHC Class I on host cells is not a consistent feature of many of these infections and in these cases NK cells are triggered by a cocktail of agonist signals from myeloid and/or lymphoid accessory cells, including dendritic cells and monocyte-macrophages (2). It is not known whether classical NK inhibitory and activating receptors play any role in regulating this indirect pathway of NK activation.
Human NK cells can be divided into two subsets according to their levels of expression of the adhesion molecule CD56 (NCAM-1) (10). CD56dim NK cells account for approximately 80% of NK cells in peripheral blood with CD56bright NK cells making up the remainder. Whether these subsets represent different maturational stages of NK cells or are terminally differentiated populations remains controversial (11–13). It has been widely assumed that the primary role of each of these subsets differs, with CD56dim cells (which express varying combinations of the many different, highly polymorphic, KIR genes) being the major cytotoxic population and CD56bright cells (which express only the relatively nonpolymorphic KIR2DL4 (14)) being the major source of cytokines (10, 15). If true, this would imply that cytokine responses of NK cells are relatively impervious to regulation by KIR-MHC interactions since the major producers of these cytokines would not express any highly polymorphic KIR. However, in apparent contradiction of this hypothesis, we have documented extensive heterogeneity of the NK cytokine response to Plasmodium falciparum (malaria)-infected red blood cells (iRBC) (16, 17) and in a small genetic association study have found evidence that this might be influenced by KIR genotype (16, 17).
Here we have analysed the human NK cell IFN-γ response to an assortment of microbial ligands and to mononuclear cell-derived cytokines. Contrary to the prevailing view, we find that both CD56dim and CD56bright NK cells can produce copious amounts of IFN-γ and that, given the overwhelming preponderance of CD56dim cells in the circulation, the vast majority of peripheral blood IFN-γ-producing NK cells are CD56dim and KIR+. Nevertheless, in many individuals, the proportion of CD56dim NK cells producing IFN-γ is markedly (and consistently) lower than the proportion of CD56bright NK cells producing IFN-γ.suggesting that CD56dim NK cells may be being regulated by KIR-MHC interactions. In support of this hypothesis, we find that the degree of inhibition of the CD56dim NK cell population is strongly associated with KIR genotype with CD56dim NK cells from individuals homozygous for the KIR “A” haplotype being significantly more inhibited than those of individuals with the KIR AB genotype. These data offer compelling evidence that genetic diversity within the KIR locus can moderate the NK response to a diverse array of microbial pathogens and thus suggest that KIR diversity may influence susceptibility to many different infections.
Healthy adult blood donors were recruited at the London School of Hygiene and Tropical Medicine through an anonymous blood donation system. All donors gave fully informed, written consent for their blood to be used in this study. Ethical approval was given by the London School of Hygiene and Tropical Medicine Ethics Committee, application #805.
Venous blood was collected into sodium heparin (10 iu/ml blood, CP Pharmaceuticals, Wrexham, UK) and peripheral blood mononuclear cells (PBMC) were isolated by Histopaque 1077 (Sigma) density gradient centrifugation as described previously (17). Cells were resuspended at a concentration of 106 cells/ml in RPMI 1640 containing 5% autologous serum, 100 Units/ml penicillin, 100µg/ml streptomycin and 2mM L-glutamine (all Gibco) and cultured in flat-bottomed 24-well plates for 24 hours.
HLA Class I-deficient 721.221 cells (18) or 721.221 cells stably transfected with HLA-Cw6-GFP constructs (6.2 and 6.4) (19, 20) (kindly provided by D. Davis, Imperial College, London) were cultured in RPMI 1640 medium supplemented with 10% heat inactivated foetal bovine serum (FBS), 2mM L-glutamine, 50U/ml penicillin/streptomycin, 1x non-essential amino acids, 1mM sodium pyruvate and 0.5mM β-mercaptoethanol (Invitrogen). Cultures of 721.221 HLA-Cw6-GFP transfectants were supplemented with 1.6mg/ml genticin (Gibco). Prior to use, expression of HLA Class I was confirmed by staining with an anti-HLA Class I antibody (W6/32).
CD56+ CD3− NK cells were enriched from PBMC by magnetic cell separation (NK Cell Enrichment kit; StemCell Technologies) according to the manufacturer’s instructions and using LD magnetic separation columns (Miltenyi Biotec). B cells, T cells, monocytes and erythrocytes were retained in the column and the effluent containing unlabelled NK cells was collected. NK cells were counted, tested for viability by Trypan blue exclusion and tested for purity by flow cytometry. Only preparations containing >95% NK cells were used.
PBMC were stained for KIR expression using the following cocktail of PE-conjugated monoclonal antibodies: DX9 (KIR3DL1), DX27 (KIR2DL3, KIR2DL2 and KIR2DS2) (BD Bioscience) Z27 (KIR3DL1) and EB6 (KIR2DL1 and KIR2DS1) (Immunotech). PBMC were stained for NKG2A using a PE conjugated monoclonal antibody clone Z199 (Beckman). Cells were subsequently stained with goat anti-PE MicroBeads (Miltenyi Biotec) and depletions carried out using LD columns (Miltenyi Biotec) according to the manufacturer’s instructions.
P. falciparum parasites (strain 3D7) were grown in ORh− human erythrocytes (National Blood Service, London, UK) in RPMI 1640 (Gibco, UK) supplemented with 25 mM HEPES (Sigma, UK), 28 mM sodium bicarbonate (BDH, UK), 20 µg/l hypoxanthine (Sigma) and 10% normal human AB serum (National Blood Service). Cultures were gassed with 3% O2, 4% CO2 and 93% N2 and incubated at 37°C. Parasite cultures were routinely shown to be free from mycoplasma/acholeplasma species contamination by ELISA (Mycoplasma Detection Kit; Roche) incorporating polyclonal antibodies against M. arginini, M. hyorhinis, A. laidlawii and M. orale. Highly pure (>95%) mature schizonts were harvested from cultures of 5–8% parasitaemia by adherence to a LD magnetic separation column (Miltenyi Biotec). Columns were washed thoroughly with PBS to remove uninfected erythrocytes before elution. Schizont-infected (iRBC) or uninfected (uRBC) erythrocytes were added to cell cultures at a ratio of 3 RBC per mononuclear cell.
Mycobacterium bovis Bacillus Calmette-Guérin (BCG, Pasteur strain; kind gift from U. Schaible) was grown in Difco Middlebrook 7H9 Broth (Becton Dickinson) complemented with 0.05% Polysorbate 80 (v/v) BDH, UK) and 10% (v/v) BBL oleic acid, bovine albumin, dextrose, and catalase (OADC, Becton Dickinson). Bacteria were harvested at a density of 109 bacteria/ml, homogenised using a 23G syringe needle and added to the cells using a multiplicity of infection (MOI) of 10:1.
Lipopolysaccharide (LPS) purified from Salmonella typhimurium (Sigma) was used at a concentration of 1 µg/ml. Human recombinant interleukin 12, human recombinant IFN-α (Peprotech) and human recombinant interleukin 18 (MBL), were each used at 0.1 µg/ml.
Surface and intracellular staining was performed as described previously (17). The antibodies used were: anti-CD3 PerCP, IgG1 PerCP, anti-KIR3DL1 (clone DX9) and anti-KIR3DL2 (clone DX27) (BD Bioscience, UK); anti-CD56 APC, IgG2a APC (Beckman Coulter, UK), anti-IFN-γ FITC, IgG1 FITC (Serotec, UK) and anti-CD158e1/e2 (clone Z27), anti-CD158a (clone EB6) (Immunotech). Flow cytometric analyses were performed using a Becton Dickinson FACSCalibur flow cytometer and FlowJo analysis software (TreeStar).
Genomic DNA was extracted from PBMC using the QIAamp DNA Blood Mini kit (Qiagen, Valencia, CA). Intermediary-resolution HLA-A, -B and -C genotypes were determined by sequence-specific oligonucleotide probes (PCR-SSOP), using LABType SSO (One Lambda; LA, California, US). Generic KIR typing was performed by PCR amplification of genomic DNA using primers targeted to regions specific for each KIR; two primer sets were used as described (21), (22). Allele-specific KIR3DL1/S1 genotypes were determined by pyrosequencing (23). The likely expressed and non-expressed allotypes of KIR2DS4 were distinguished using allele-specific primers (24). Normally-expressed (10A) and truncated (9A) allotypes if KIR2DL4 were distinguished using previously described primers (25) to generate amplicons for pyrosequencing. The number of expressible KIR genes present in each individual was calculated from their generic and allele-specific genotypes according to previously described haplotype segregation (24, 26) and by the 2DL4 and 3DL1/S1 copy-numbers that were determined by pyrosequencing. KIR3DL1*004, 3DS1*049N, 2DL5B and 2DS4*003–7 were counted as non-expressed alleles.
All statistical analyses were performed using Prism 4 software (Graph Pad Software Inc).
It is often assumed that CD56bright NK cells constitute the major IFN-γ-secreting subset of NK cells and that CD56dim NK cells primarily fulfil a cytotoxic function (10). Challenging this dichotomy was our observation that Plasmodium falciparum-infected red blood cells (iRBC) induce significant IFN-γ production by human CD56dim NK cells (17). To explore this phenomenon further we have compared, in a panel of healthy adult donors, the IFN-γ response of CD56dim and CD56bright NK cells within PBMC to stimulation by monokines and by several microbial products. Figure 1 shows the IFN-γ response of CD56dim and CD56bright NK cells from two donors after stimulation in vitro for 24 hours with iRBC, bacterial LPS, Mycobacterium bovis BCG, a cocktail of recombinant IL-12 and IL-18 or a cocktail of IL-18 and IFN-α; data from the entire donor cohort are summarised in Figure 2. It was immediately apparent that both CD56bright and CD56dim NK cells were able to make robust IFN-γ responses.
Although responses of some donors (e.g. Figure 1A) were consistent with the view that CD56bright cells rather than CD56dim cells are the major producers of IFN-γ, in other donors very similar proportions of CD56bright and CD56dim cells contained intracellular IFN-γ (e.g. Figure 1B). For four of the stimuli tested (i.e. all three microbial stimuli and IL-12+IL-18), the percentage of CD56bright NK cells producing IFN-γ was slightly (but significantly) higher than the percentage of CD56dim cells making IFN-γ (Figure 2A), and the mean MFI for IFN-γ staining was somewhat higher in CD56bright than in CD56dim NK cells (Figure 2B). However, as CD56dim NK cells represent more than 90% of the NK cells in peripheral blood (Supplementary Figure I), the vast majority (~80%) of IFN-γ+ peripheral blood NK cells are CD56dim (Figure 2C). It is notable however that, in contrast to the other stimuli, the cocktail of IL-18+IFN-α preferentially induces IFN-γ production by CD56dim rather than CD56bright NK cells.
For some donors, CD56dim and CD56bright populations of NK cells responded similarly to microbial or cytokine stimuli (Figure 1B) whereas for other donors the IFN-γ response of CD56dim NK cells response appeared inhibited by comparison to the response of their CD56bright cells (Figure 1A). To quantify this effect, we calculated the ratio between the percentage of CD56dim NK cells and the percentage of CD56bright producing IFN-γ in response to each stimulus (Figure 3A). This ratio, which we have termed the “DIM factor”, is a measure of the extent to which the CD56dim NK cells of an individual respond to a particular stimulus when compared to their CD56bright NK cells; a DIM factor of <1 indicates that the CD56dim NK cell population is relatively inhibited compared to the CD56bright population.
We calculated the DIM factor for the response of NK cells of up to 40 donors to the various stimuli (Figure 3B). For the three microbial stimuli and for IL-12+IL-18, CD56dim NK cells tended to respond less well than CD56bright cells (mean DIM factor < 1.0). Strikingly, however, DIM factors for cells cultured with IFN-α+IL-18 tended to be >1.0 and were significantly higher than the DIM factors for other stimuli (unpaired t test: IFN-α+IL-18 vs. any other stimulus, t ≥ 4.78, p < 0.0001).
For ten donors in the panel, DIM factors following IL-12 + IL-18 stimulation were analyzed from repeated blood samples obtained at intervals of more than two months. Although DIM factors varied between donors, the DIM factor of cells from individual donors remained remarkably constant over time (Figure 3C). Moreover, for an individual donor, DIM factors for NK cells stimulated with LPS, BCG or iRBC were highly correlated with DIM factors for cells stimulated with IL-12+IL-18 (Figure 4). Similarly pairwise comparisons between DIM factors for any two microbial stimuli were highly correlated (r2 ≥ 0.89 for all comparisons). DIM factors for IL-18+IFN-α stimulated NK cells were significantly correlated with DIM factors for IL-12+IL-18, however the correlation coefficient (r2 = 0.37) was lower than for other comparisons.
These analyses indicate that the relative response of CD56dim and CD56bright NK cells to different microbial stimuli is constant within an individual but varies within the human population.
Accessory cell-dependence is emerging as a general feature of NK cell responses to pathogens, the necessary cytokines including IL-12, IL-18 and IFN-α (2). We have previously shown that NK cell IFN-γ responses to iRBC depend upon contact-dependent and cytokine-mediated signals from myeloid accessory cells (27) and that the magnitude of the NK IFN-γ response (and especially of CD56bright NK cells) is correlated with the magnitude of the accessory cell response (27). To determine whether the DIM factor is a manifestation of the magnitude of the accessory cell response to iRBC, we compared the DIM factor with the amount of IL-12 secreted into the culture, but found no correlation (Supplementary figure II). Thus, although the magnitude of the accessory cell response to pathogen-derived ligands influences the absolute magnitude of the NK cell IFN-γ response (27), the DIM factor appears to be independent of the accessory cell response.
We conclude that two independent factors determine the magnitude of the NK cell IFN-γ response to microbial stimuli, namely the size of the accessory cell signal and the DIM factor. In Figure 5, data from three donors illustrate how these two factors interact to determine the overall NK response to iRBC. Donor 177 has both a high DIM factor (0.9) and high levels of accessory cell IL-12 secretion in response to iRBC. This combination leads to a potent NK cell response to iRBC in which 44% of the NK cells are induced to make IFN-γ. For donor 147 a moderate accessory cell IL-12 response to iRBC combines with a low DIM factor (0.56), to give a moderate IFN-γ response involving 27% of the NK cells. In contrast, in donor 59, a very low accessory cell IL-12 response combined with a low DIM factor (0.54) results in a failure of NK cells to produce IFN-γ.
Key features of the DIM factor are population diversity, stability within individuals and stability between stimuli. These features suggest that a person’s DIM factor may be genetically determined. One obvious candidate genetic locus for differential regulation of CD56bright and CD56dim NK cells is the KIR locus, for which individual haplotypes contain variable numbers of highly polymorphic genes (21, 28). It has been reported that CD56dim NK cells express many different KIRs (11, 15) whereas CD56bright cells express either no KIR or only KIR2DL4 (14). In confirmation of this, we analysed KIR surface expression on NK cells from 12 donors using monoclonal antibodies specific for four different groups of KIR (but not KIR2DL4) and found that 10–50% of CD56dim NK cells stained positively with each of the anti-KIR antibodies, and that on average 55.7% of CD56dim NK cells were labelled by a cocktail of all 4 KIR antibodies, whereas few (< 5%) CD56bright NK cells stained with any of the antibodies (Supplementary Figure III). We therefore considered the possibility that responses of KIR+ CD56dim NK cells to microbial and cytokine stimulation might be regulated via KIR-MHC Class I interactions.
KIR genotypes were determined for 81 individuals for whom the DIM factor was known and MHC Class I genotypes were determined for 47 of these. KIR genotypes were compiled from KIR gene-specific and allele-specific typing and assigned as AA, AB or BB according to those previously defined from haplotype segregation in families (21, 24, 26). The number of expressible KIR genes present in each individual was calculated from their allele-specific genotypes and by the 2DL4 and 3DL1/S1 copy-numbers that were determined by pyrosequencing. KIR3DL1*004, 3DS1*049N, 2DL5B and 2DS4*003–7 were counted as non-expressed alleles. Among our donor population, 27 (33%) were homozygous for A haplotypes, 17 (21%) were homozygous for B haplotypes and 37 (46%) were AB heterozygotes.
DIM factor varied significantly with KIR genotype (Figure 6). Cell donors were grouped according to whether they were KIR-A or KIR-B homozygotes or KIR-AB heterozygotes. The mean DIM factor did not differ between KIR-A homozygotes and KIR-B homozygotes [mean (SEM), 0.56 (0.04) and 0.58 (0.05) respectively] but the DIM factor for all KIR homozygotes combined was significantly lower than the DIM factor for KIR-AB heterozygotes [mean (SEM), 0.57 (0.03) and 0.72 (0.03) respectively]. Similarly, when donors were grouped by DIM factor (low: <0.5, medium: 0.5–0.7, high:>0.7) we observed that the proportion of AB individuals increased (and the proportion of AA individuals decreased) significantly with increasing DIM factor (P = 0.0007; data not shown).
No association was observed between DIM factor and the presence or absence of any particular KIR gene (Supplementary figure IV), the presence of any particular KIR-HLA ligand pair (Supplementary figure V) or the total number of different KIR genes in the genotype (not shown).
Similarly, although one explanation for the low DIM factor (i.e. highly inhibited CD56dim NK cells) in AA individuals might be that these individuals have a paucity of activating KIR (KIR2DS4 only) compared to AB and BB individuals, when we stratified our donor cohort by the total number of activating KIRs in their genotype, there was no systematic association with DIM factor (supplementary figure VI.A). There was a trend for the DIM factor to be higher (i.e. for CD56dim cells to respond more like CD56bright cells) for individuals with more inhibitory KIR in their genotype, however this did not reach statistical significance (One way ANOVA test for linear trend; P = 0.07).
We have considered two mechanisms by which KIR genotype might affect NK cell IFN-γ responses. Firstly, signalling resulting from ongoing ligation of KIR (by HLA Class I on accessory cells) during in vitro culture might modulate the degree of activation of NK cells by microbial or cytokine stimuli. Alternatively, binding of KIR (or other inhibitory receptors) by relevant MHC and MHC-like ligands during NK cell ontogeny might lead to greater or lesser licensing of NK cells, in which case the degree to which CD56dim NK cells are licensed, relative to CD56bright NK cells, would determine the DIM factor.
To investigate whether KIR-HLA Class I interactions occurring in vitro modify NK cell activation we stimulated PBMC with IL-12 + IL-18 for 24 hours in the presence of blocking anti-HLA Class I antibodies (DX17 or W6/32) or an isotype matched control antibody. As shown in Figures 7A and 7B, both antibodies bound efficiently to lymphocytes and monocytes and remained bound for the duration of the experiment but neither antibody had any significant or consistent effect on NK cell IFN-γ responses. These data indicate that interrupting KIR-HLA Class I interactions in real time does not influence NK cell IFN-γ production.
To explore this further, we examined the effect of a single KIR-HLA Class I interaction on NK cell activation by comparing IFN-γ responses in NK cells from 6 donors carrying at least one copy of KIR2DL1 (and no copies of KIR2DS1) when cultured in the presence of HLA Class I-deficient 721.221 cells (18) or 721.221 cells stably transfected with HLA-Cw6-GFP constructs, 6.2 and 6.4. Flow cytometric analysis confirmed that the 721.221 cells did not express HLA-Cw6-GFP while 7126.96.36.199 and 7188.8.131.52 cell lines expressed low and high levels of HLA-Cw6-GFP respectively (Figure 7C, first panel). Purified NK cells were cultured in a 1:1 ratio with HLA Class I+ 7184.108.40.206 or 7220.127.116.11 cells, or with HLA Class I− 721.221 cells, for 24 hours in the presence or absence of IL-12+ IL-18. In unstimulated cultures, at this NK to target cell ratio, there was no induction of IFN-γ in NK cells whether the target cells expressed or did not express HLA-Cw6 (data not shown). Importantly, the presence or absence of cells expressing HLA-Cw6 made no consistent or significant difference to the proportion of either CD56bright or CD56dim NK cells making IFN-γ in response to IL-12 + IL-18 and consequently the DIM factor was unchanged (Figure 7C final three panels). Thus, when NK cells and target cells are cultured together at an appropriate ratio and in the absence of any exogenous stimuli, the absence of KIR-MHC Class I interactions does not cause NK cells to be activated to make IFN-γ; furthermore the absence of MHC Class I ligands does not influence their ability to respond to accessory cell-derived/pathogen-induced signals. Taken together, these two independent sets of experiments suggest that KIR-MHC interactions are not influencing NK responses to accessory cell-derived signals in real time.
Understanding the means by which innate immune responses are regulated offers the possibility to design more effective adjuvants and anti-inflammatory agents and to better understand disease susceptibility. There is now a substantial body of evidence to suggest that KIR-HLA Class I interactions influence susceptibility to viral infections and neoplasia (7, 8) presumably by regulating the ability of NK cells to lyse host cells that express abnormally low levels of self MHC (29). By comparison, very little is known regarding the role of KIR-HLA Class I-mediated regulation of responses to classes of pathogens for which cytokine production rather than cytolysis is the more relevant effector response. We have therefore characterised NK IFN-γ production in response to accessory cell-derived cytokines and to three, non-viral, microbial ligands and examined the effect of KIR and HLA Class I genotype on these responses.
Our key observations are that although both CD56bright and CD56dim NK cells produce IFN-γ following microbial stimulation, responses of these two NK cell subsets are independently regulated and CD56dim cells tend to be relatively inhibited in their IFN-γ response compared to CD56bright cells. Importantly, we have shown that the relative response of CD56bright and CD56dim NK cells (i.e. the DIM factor) is associated with KIR genotype, providing the first direct evidence that KIR genotype may influence the NK response to a wide range of non-viral pathogens. Taking these findings together with our previous observation that the magnitude of the NK cell IFN-γ response, and of the CD56bright (KIR−) subset in particular, to iRBCs is highly correlated with the degree of myeloid accessory cell activation (as assessed by upregulation of CD40 and HLA-DR as well as secretion of IL-12) (27) we propose that NK cell IFN-γ responses to external pathogen-derived stimuli are regulated by (at least) two independent factors. One of these is the strength of the accessory cell (cytokine and or costimulatory molecule) signal and this affects all NK cells to the same degree. The other is the regulatory effect conferred by KIR gene expression. This differs between clones of NK cells (since they express slightly different combinations of KIR genes) and between individuals (since their KIR genotype differs). Since CD56bright NK cells are essentially KIR- their response is unaffected by the regulatory effects of KIRs and thus, for each donor, the response of the CD56bright NK cell population is an indication of the strength of the accessory cell stimulus and represents the unregulated baseline NK response. However, the response of the CD56dim (KIR+) subset depends upon both the strength of accessory cell signals and the degree to which this is regulated by KIR gene expression. Thus, the difference between the response of the CD56bright and CD56dim cells (the DIM factor) is an indication of the strength of the regulation imposed by the KIR genotype.
Whilst the KIR genotype effect is constant, resulting in DIM factors that are highly correlated for different ligands, accessory cell responses may vary from one pathogen to another depending on the affinity (and amount) of their ligands for the various pattern recognition receptors expressed on different accessory cell populations (2). Indeed, we provide direct evidence here that the nature of the accessory cell signals can profoundly affect that NK response since, in combination with IL-18, IFN-α much more efficiently induces IFN-γ production from CD56dim NK cells than does IL-12. It is possible, therefore, that infections that induce high levels of IFN-α and IL-18 may induce particularly potent NK IFN-γ responses. It is likely, however, that environmental factors, specifically prior or concurrent infections, might also contribute to the shaping of NK compartments – either by modifying the accessory cell stimulus or by modifying the frequency and functional maturity of specific NK cell subsets. In our study, such environmental factors may account for differences in NK cell responses between individuals with similar KIR genotypes.
The ability of NK cells to kill cells expressing abnormally low levels of MHC-Class I molecules is determined by their expression of inhibitory receptors for relevant MHC ligands during NK cell development: only NK cells which express appropriate receptors are “educated” (or “licensed”) for subsequent cytotoxic activity (30, 31). Whether education also affects an NK cell’s capacity to respond to pathogen-derived signals is less clear although educated NK cells have been shown to be more responsive (as judged by IFN-γ production) to IL-12+IL-18 than uneducated NK cells (11, 32). Conversely, it has been suggested that, in mouse cells at least, pathogen-induced signals can overcome the lack of licensing such that all NK cells, whatever their regulatory receptor phenotype, can respond to inflammatory signals (33). Our data, however, provide convincing evidence that an individual’s KIR genotype does in fact predict the capacity of their NK cells (or at least, the majority CD56dim subset of NK cells) to produce IFN-γ in response to numerous indirect stimuli. Although NK cells have been reported to make IFN-γ when cultured with HLA Class-I deficient 721.221 tumour cells (34) we found no evidence that NK cells were induced to make IFN-γ when cultured at a 1:1 ratio with 721.221 cells (in the absence of IL-2). Moreover, NK cells cultured with blocking antibodies to MHC Class I or with 721.221 cells were no more or less able to make IFN-γ in response to IL-12+IL-18, a classical accessory cell stimulus for NK cells, than were NK cells cultured with MHC sufficient cells. Thus, since blocking ongoing KIR-HLA interactions during NK activation does not appear to influence IFN-γ production, we suggest that KIR genotype may affect the NK cell at a much earlier stage of NK cell ontogeny; in other words, this may be an example of NK cell education.
Our data suggest that the overall KIR genotype - rather than any single KIR receptor or KIR-HLA interaction - determines the responsiveness of KIR+ NK cells. Indeed, our data appear to be a classical example of heterozygote advantage, with AB heterozygotes having the most educated NK cells. This is very much in line with the concept that NK cell activation depends upon integration of signals from numerous inhibitory and activating receptors; AB heterozygotes might be expected to express a wider range of KIR alleles than would AA or BB homozygotes and thus a greater proportion of their NK cells are likely to encounter an appropriate self ligand and be licensed. Our data thus offer a mechanistic explanation for the observation that balancing selection has maintained an even frequency of A and B haplotypes (and a correspondingly high frequency of AB heterozygotes) in isolated Amazonian Amerindian populations (35). However, we cannot formally rule out the possibility that specific KIR alleles determine responsiveness since one consequence of having a large number of KIR genes available for expression is that each KIR is likely to be expressed at a relatively low frequency, making it difficult to observe the effect of any one particular receptor.
Of interest, several clinical studies have observed associations between disease state and overall KIR genotype rather than with the presence, absence or allelic variation in any particular KIR gene. For example, individuals with AA KIR genotypes are reported to be relatively protected against chronic inflammatory diseases (36, 37) – our data suggest that this might be explained by these AA individuals having relatively inhibited innate IFN-γ responses. Similarly, among long-term HIV-exposed subjects, AB individuals were significantly more likely than AA individuals to remain seronegative (38), consistent with the notion that individuals with less inhibited CD56dim cells may make more effective anti-viral responses.
In summary, we have found that the ability of NK cells to respond to accessory cell-derived cytokines is markedly affected by KIR genotype. Since these cytokines are classically induced from accessory cells by microbial ligands and are essential for NK cells to respond to pathogens as diverse as malaria and mycobacteria (2), our data reveal a hitherto unsuspected influence of KIR genotype on the ability of NK cells to respond to non-viral infections. When taken together with our recent evidence for major differences in KIR genotype between populations living under differing levels of pathogen exposure (23) our data suggest that the ability of NK cells to respond appropriately to common bacterial or protozoal, as well as viral, pathogens may represent a major force that is driving evolution of the KIR locus and, conversely, that KIR genotype might be significant predictor of susceptibility or resistance to many common infections.
We would like to thank Elizabeth King and Carolynne Stanley for invaluable technical support. We are enormously grateful to Mostafa Ronaghi (Stanford University) for guidance regarding KIR pyrosequencing and access to the pyrosequencing facility.
1This work was funded by grants from the UK Medical Research Council (GO400225) and NIH (AI017892). PJN was a Lymphoma Research Foundation Fellow and DK was a Boehringer Ingelheim Fonds pre-doctoral research fellow.
Daniel S. Korbel designed the research, performed the research, analysed data and contributed to writing and editing of the manuscript. Paul J. Norman designed the research, performed the research and analysed data. Kirsty C. Newman designed the research, performed the research, analysed data and contributed to writing of paper. Amir Horowitz performed the research and analysed data. Ketevan Gendzekhadze performed the research. Peter Parham designed the research, analysed data and contributed to writing of paper. Eleanor M. Riley designed the research, analysed data and drafted and revised the paper.