Terminal differentiation of NK cells is crucial in maintaining broad responsiveness to pathogens and discriminating normal cells from cells in distress. Although it is well established that KIRs, in conjunction with NKG2A, play a major role in the NK cell education that determines whether cells will end up competent or hyporesponsive, the events underlying the differentiation are still debated.
A combination of complementary approaches to assess the kinetics of the appearance of each subset during development allowed us to obtain new insights into these terminal stages of differentiation, characterising their gene expression profiles at a pan-genomic level, their distinct surface receptor patterns and their prototypic effector functions. The present study supports the hypothesis that CD56dim cells derive from the CD56bright subset and suggests that NK cell responsiveness is determined by persistent inhibitory signals received during their education. We report here the inverse correlation of NKG2A expression with KIR expression and explore whether this correlation bestows functional competence on NK cells. We show that CD56dimNKG2A−KIR+ cells display the most differentiated phenotype associated to their unique ability to respond against HLA-E+ target cells. Importantly, after IL-12 + IL-18 stimulation, reacquisition of NKG2A strongly correlates with IFN-γ production in CD56dimNKG2A− NK cells.
Citation: Béziat V, Descours B, Parizot C, Debré P, Vieillard V (2010) NK Cell Terminal Differentiation: Correlated Stepwise Decrease of NKG2A and Acquisition of KIRs. PLoS ONE 5(8): e11966. https://doi.org/10.1371/journal.pone.0011966
Editor: Jacques Zimmer, Centre de Recherche Public de la Santé (CRP-Santé), Luxembourg
Received: June 28, 2010; Accepted: July 12, 2010; Published: August 6, 2010
Copyright: © 2010 Béziat et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported in part by national funds from the Institut National de la Santé et de la Recherche Médicale (INSERM) under contract PHRC “Minicord” (P060206-AOM06206), by the association Cent pour Sang la Vie and the Agence de la BioMédecine. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
NK cells are one component of the innate immune system, able to recognise various targets without specific sensitisation. They constitute a first line of defence and can kill infected and tumour cells. Mature NK cells are heterogeneous and differ in their proliferative potential, homing characteristics, functional capacities and responses to a wide range of cytokines , . The complexity of the NK cell compartment is attributable to the vast network of receptors that are expressed on NK cell surfaces, which allow them to recognise target cells. These receptors can be either inhibitory or activating, and it is the balance between these opposing signals that dictates the functional status of the cells . Around 10% of the NK cells in peripheral blood are CD56bright; this immunoregulatory subset produces a wide range of cytokines and chemokines after monokine stimulation, but its ability to kill target cells spontaneously is poor. In contrast, the remaining 90% of the NK cells, the CD56dim cytotoxic NK cell subset, produces relatively lower cytokine levels, but possesses an abundance of cytolytic granules and can spontaneously lyse susceptible target cells .
The precise relationship between these two major NK cell subsets, CD56bright and CD56dim, has long been debated. Several recent reports have finally shown quite convincingly that CD56bright NK cells are very likely precursors of the CD56dim subset: the shorter telomeres of the latter, in both peripheral blood and lymph nodes, imply that they are older than the former , . In addition, Romagnani et al have shown in vivo that efferent lymph nodes contain a substantial proportion of CD56dim KIR+ NK cells . Very recently, Yu et al reinforced the concept of a transition from CD56bright to CD56dim cells, proposing CD94 as a developmental intermediary between these NK cell subsets . Along with others, we have demonstrated that NK cells are the first lymphocytes appearing in patients recovering from hematopoietic stem cell transplantation (HSCT) and that these cells are constituted mainly from CD56bright NK cells , , , .
Other and we have shown that CD56dim NK cells from patients who have recently had hematopoietic transplantations transiently overexpress NKG2A, compared with cells from healthy donors , , . In addition, CD56dim NK cells from cord blood highly over-expressed NKG2A . Altogether, these data suggest that the modulation of NKG2A expression among CD56dim cells may define a crucial step of their differentiation. NKG2A is an inhibitory NK receptor that has been reported to form disulfide-linked heterodimers with invariant CD94. The NKG2A ligand has been identified as HLA-E, a non-classical MHC class-Ib molecule that is widely distributed among various tissues, exhibits relatively low surface expression, and has limited polymorphism . The cross-linking of NKG2A by target cells bearing HLA-E results in NK cells inhibition through the phosphorylation of the two immunoreceptor Tyr-based inhibitory motifs (ITIMs) present in the intracellular tail of NKG2A .
Although it is well established that KIRs, in conjunction with NKG2A, play a major role in the NK cell education that determines whether cells will end up competent or hyporesponsive, the mechanisms underlying this education are still debated . One hypothesis proposes that the engagement of self-MHC-specific inhibitory receptors directly renders NK cells functional, “licensing” them, so to speak. The “disarming” model, on the other hand, suggests that the hyporesponsiveness of NK cells lacking self-MHC-specific inhibitory receptors is a consequence of the persistent stimulation of previously competent cells. More recently a new “rheostat” model argues that the responsiveness of each NK cell is tuned by the input of inhibitory receptors , , .
Here we used a combination of complementary approaches that assess the kinetics of the appearance of each subset during development to examine the terminal stages of differentiation in the light of these hypotheses and to develop a detailed model that integrates our new insights with recent findings by others. All the observations from this study support the previous hypothesis that CD56dim cells derive from the CD56bright NK cell subset. Furthermore, this study provides the first phenotypic and functional analysis of mature CD56dim subsets that focuses on the progressive decrease in NKG2A expression and the sequential acquisition of KIRs. We thus put forward a new model of terminal differentiation that starts at the CD56bright stage and leads through the following stages: CD56dimNKG2A+KIR−CD56dimNKG2A+KIR+ CD56dimNKG2A−KIR+. Our data support the concept that terminally differentiated CD56dimNKG2A−KIR+ NK cells are fully competents, able to kill HLA-E+ targets, or alternatively CD56dimNKG2A−KIR− hyporesponsive NK cells, due to impaired KIR acquisition.
Materials and Methods
Samples collection and treatment
All samples used in this study derived from anonymous individuals at the EFS (Etablissement Français du Sang), the National Blood Service. Written Informed consent was obtained for each donor. This study was approved by the institutional review board at Pitié-Salpêtrière Hospital (Paris, France). PBMC were isolated using Ficoll gradient immediately following collection. Cells were thereafter used for different tests.
Five-colour fluorescent activated cell-sorting analysis was performed on freshly harvested PBMCs, or whole blood cells for post-UCBT. Cells were stained with the appropriate antibody cocktail: anti-CD3 (UCHT1; Beckman Coulter), anti-CD56 (N901; Beckman Coulter), anti-CD16 (3G8; Beckman Coulter or VEP13; Miltenyi biotec), anti-CD94 (HP-3D9; Becton Dickinson), anti-CD159a/NKG2A (Z199; Beckman Coulter), anti-NKG2C (134591; R&D systems), anti-CD85j/ILT2 (HP-F1; Beckman Coulter), anti-CD27 (B1.49.9; Beckman Coulter), anti-CD62L (MEL-14; Becton Dickinson), anti-CD57 (S-HCL-1; Becton Dickinson), anti-Siglec-9 (191240; R&D Systems) and anti-FCRL6 (2H3; BioLegend). Global KIR analysis was performed with an anti-KIR antibody mix: anti-KIR2DL1/KIR2DS1 (EB6B; Beckman Coulter), anti-KIR2DL2/KIR2DL3/KIR2DS2 (GL183; Beckman Coulter) and anti-KIR3DL1/KIR3DS1 (Z27; Beckman Coulter). Single KIR analysis was performed by subtraction to discriminate between KIR2DS1 and KIR2DL1, in the presence of anti-KIR2DL1 (143211; R&D Systems) and anti-KIR2DL1/KIR2DS1. Similarly, KIR2DL2 and KIR2DL3 were discriminated with anti-KIR2DL3, (180701; R&D Systems) and anti-KIR2DL2/DL3 (DX27, Miltenyi Biotec). Other single KIRs were detected with anti-KIR3DL1 (DX9; R&D Systems), anti-KIR2DS4 (179315; R&D Systems) or anti-KIR2DL5 (UP-R1; E-Biosciences). FACS lysing solution (Becton Dickinson) was used to lyse erythrocytes for whole blood staining. Cells were permeabilised for intracellular staining with cytofix/cytoperm kit (Becton Dickinson) and then stained with anti-granzyme-K (GM6C3; Santa Cruz) or anti-IFN-γ (B27, BD Biosciences) mAbs. IFN-γ staining was performed after overnight incubation in the presence of 10 ng/ml IL-12 (R&D Systems) plus 100 ng/ml IL-18 (R&D Systems).
104 cells of CD56brightCD16−, CD56dimNKG2A+ and CD56dimNKG2A− NK cell subsets from four different healthy donors were sorted by FACS. Purified cells were collected and then resuspended in SuperAmp™ Lysis Buffer (Miltenyi Biotec, Cologne, Germany). The Agilent 2100 Bioanalyzer System (Miltenyi Biotec) was used to control total RNA quality. Gene expression analysis was performed with the Human Whole Genome Oligo Microarray (44 K), as a commercial service by Miltenyi Biotec, to reveal a gene expression profile for each NK cell subset of every individual donor. The Agilent Feature Extraction Software was used to read out and process the microarray image. Data were then normalised by dividing each reporter signal with the median signal intensity of the whole microarray, and then filtered to remove noisily or arbitrarily regulated genes with expression levels close to background. Reporters with expression intensities of less than 10% of the array median were removed from the dataset. Thus 33,867 of the 41,000 genes with strong reliable signals were sorted and then analysed with TM4 software . Complete linkage hierarchical clustering was then performed on the significantly modulated genes (p<0.01), with one-way ANOVA. Each cluster was then individually used for ontogenetic analysis using Panther-BP (in the gene ontology section) of DAVID software , . All microarray data are MIAME compliant and raw data have been deposited in the MIAME compliant database ArrayExpress (accession number: E-MEXP-2798).
Whole NK cells from freshly isolated PBMCs were negatively purified with a MACS NK cell isolation Kit (Miltenyi biotec). The various subsets of NK cells were sorted with a BD FACS-Aria (Becton Dickinson) cell sorter from purified NK cells in the presence of anti-CD3 (UCHT1; Beckman Coulter), anti-CD56 (N901; Beckman Coulter), anti-NKG2A (Z199; Beckman Coulter), and anti-CD16 (VEP13, Miltenyi). Post-sorting purities were routinely >96%. NK subsets were then used for the microarrays and for IFN-γ production assays.
Degranulation assays used CD107a detection, as previously described . PBMC were resuspended at a 1∶1 ratio in the presence of target cells and anti-CD107a mAb (H4A3; Becton Dickinson). After 1 h incubation, monensin was added at 2 mM for an additional 4 h of incubation. Degranulation was assayed against MHC class-I-deficient K562, 721.221 target cells, and LCL-221-AEA, which expresses the HLA-E*0101 allele . Antibody-dependant cell cytotoxicity assays were performed against CD20+ RAJI cells in the presence of 1 µg/ml of anti-CD20 (rituximab, Roche).
Mann Whitney tests were performed for individual comparisons of two independent groups. Wilcoxon tests were performed for individual comparison of the paired groups. Repeated measures ANOVA with Tukey post-test for P-value calculation were performed for multiple comparisons assays. Statistical analysis used Prism 5 software (GraphPad Software, San Diego, CA). P value less than 0.05 were considered as significant. 1* = p<0.05; 2* = p<0.01; 3* = p<0.001.
Phenotypic and transcriptomic profile of CD56bright, CD56dimNKG2A+ and CD56dimNKG2A− subsets
In order to better characterize the last steps of NK cell differentiation, we analysed the expression patterns of NK receptors and other maturation molecules on CD56bright, CD56dimNKG2A+, and CD56dimNKG2A− cells from healthy controls by flow cytometry. As shown in Fig. 1, CD62L, granzyme-K and CD27 were highly expressed in CD56bright NK cells, but more importantly their expression progressively decreased in CD56dimNKG2A+ and CD56dimNKG2A− NK cells. Inversely, CD57, ILT2, FCRL6 and Siglec-9 progressively accumulated when looking sequentially these subsets. In contrast, other NK markers distinguished only CD56bright and CD56dim. These included Pen5, CD8α, perforin and granzyme-B, which were expressed mainly on CD56dim cells, independently of NKG2A expression. NKp44, CCR7, CD127 and CD117 were expressed almost exclusively on CD56bright NK cells (data not shown). Finally, these three NK cell subsets were indistinguishable for the other receptors tested, including NKp30, NKp46, NKG2D, 2B4, LAIR-1, DNAM, CD11a and CD11b (data not shown). Of note, NKG2A is covalently associated with CD94 to form heterodimers and, as expected, expression of NKG2A and of CD94 was highly correlated (Figure S1). However, CD94 expression is also linked to that of NKG2C, an activating NK receptor that, like NKG2A, recognises HLA-E . NKG2C is poorly expressed on NK cells from healthy donors. The CD56dim NK subset expressed NKG2C and NKG2A mainly separately while the CD56brightNKG2C+ subset co-expressed NKG2A (Figure S1) .
Representative samples (left panels) and box and whiskers plots (right panels) are presented for each marker from at least 16 healthy controls. *: p<0.05; **: p<0.01; ***: p<0.001.
In an attempt to better characterize the different NK cell subsets, we compared the gene expression profiles of CD56bright, CD56dimNKG2A+, and CD56dimNKG2A− NK cells. The analysis of variance revealed that 723 of the 41,000 genes tested showed significant changes in gene expression between the subsets at a p value (significance level) of p<0.01 (Table S1). The hierarchical clustering analysis of the differentially expressed genes revealed only six clusters (Fig. 2A). As expected, among the genes downregulated in cluster 1 (30 genes) and overexpressed in cluster 5 (125 genes) during the transition between CD56bright to CD56dim, regardless of NKG2A expression, were some KIRs (Kir3DL1, Kir2DS4 and Kir2DL5A), other NK cell-surface receptors: (CD6, CD16a, 2B4, CD48 and CD62L), chemokines (CCL3, CCL4, CCL5, CCL3L3 and CCL23), and signalling molecules (STAT2, STAT4 and LAT2) (Table S2).
CD56brightCD16− (Bright), CD56dimNKG2A+ (D2A+) and CD56dimNKG2A− (D2A−) NK cell subsets from 4 healthy donors (#1, #2, #3 and #4) were sorted and analysed using pan-genomic microarrays. (A) Left panel, Hierarchical clustering analysis of significantly modulated genes. The six clusters were numbered C1 to C6. Right Panel, alternative representation of each cluster showing the relative intensity of genes expression. Median relative intensities are represented by a red line. (B) Gene ontogeny analysis was performed with DAVID software on genes of clusters 2, 4 and 5. Highly significant terms are shown with the associated genes (boxes) and their expression profile in the three NK cell subsets. Median relative intensities are represented by a red line.
More interestingly, this gene profiling analysis largely confirmed that the CD56dimNKG2A+ subset mainly display an intermediary profile between CD56bright and CD56dimNKG2A− subsets. Indeed, the largest clusters, cluster 2 (212 genes) and cluster 4 (155 genes), corresponded to genes that were downregulated or overexpressed, respectively, according to the sequential CD56bright, CD56dimNKG2A+ and CD56dimNKG2A− NK cell subsets. Focussing on these two major clusters, we used DAVID software (Table S2) to conduct a gene ontogeny analysis , . As Fig. 2B shows, the most representative group of genes in cluster 2 involved mRNA transcription regulation. It contained notably GATA3, TCF7, VAV1 and SOX4, all described in T cells or NK activation or differentiation pathways , , . Moreover, looking specifically at NK cell markers, we found substantial downregulation of both CD117, a key marker of immature NK cells, and of granzyme-K, a cytotoxic component of CD56bright NK cells, thereby confirming our flow cytometry results (Fig. 1) (Table S2).
Conversely, cluster 4 was mainly characterised by the presence of genes involved in signal transduction and cell structure, including a number of tumour suppressor genes (FAT-2, DCC, RASSF-4 and BRAF) that play a role in the control of cell proliferation. In addition, expression increased for several genes in this cluster that are involved in the regulation of NK cell functional activities, such as KIRs (Kir2DL4, Kir3DL2, Kir2DS2, Kir2DL2 and Kir2DS1), granzyme-B and Ksp37 for cytotoxicity, and CARD11 (CARMA1) for the regulation of cytokine/chemokine production (Fig. 2B) (Table S2).
Although KIRs control the reactivity of mature NK cells, their expression also influences the functional maturation of developing NK cells. Fig. 3A shows that expression of whole KIRs, monitored by a cocktail of KIR antibodies, linearly increases among CD56bright, CD56dimNKG2A+ and CD56dimNKG2A− subsets. More importantly, cell-surface expression of individual KIRs on all three NK cell subsets revealed that inhibitory (KIR2DL1, KIR2DL2, KIR2DL3, KIR3DL1, KIR2DL5) and activating (KIR2DS1) receptors increased significantly (p<0.001, for each marker) from the CD56bright to CD56dimNKG2A+ subset and from the CD56dimNKG2A+ to CD56dimNKG2A− subset (Fig. 3B). In contrast, KIR2DS4 rose only from the CD56bright to CD56dim independently of NKG2A expression. These data were consistent with the results obtained by gene array analysis (Fig. 2).
Cell surface expression of whole KIRs (A), or single KIR (B) on CD56bright (open bars), CD56dimNKG2A+ (grey bars) and CD56dimNKG2A− (black bars) NK cells subsets from 32 healthy controls. *: p<0,05; **: p<0.01; ***: p<0.001.
Together these results clearly show that specific combinations of NKG2A and CD56 expression permit to differentiate three phenotypicaly different subsets, including a CD56dimNKG2A+ subset displaying an intermediary profile between CD56bright and CD56dimNKG2A− subsets.
NKG2A acquisition correlates with IFN-γ production in CD56dimNKG2A− NK cells after IL12/IL18 stimulation
To determine the functional significance of these findings, we assessed intracellular IFN-γ production in highly purified CD56bright, CD56dimNKG2A+, and CD56dimNKG2A− NK cell subsets after IL12+IL18 stimulation (Fig. 4). A significantly higher frequency of CD56bright (compared with CD56dim) NK cells produced IFN-γ, regardless of whether or not the latter expressed NKG2A (p<0.001 in both cases). As Fig. 4A shows, around 80.9±6.2% of CD56bright NK cells expressed IFN-γ, whereas only 49.7±20.8% of CD56dimNKG2A+ and 33.8±20.9% CD56dimNKG2A− NK cells did so. Moreover, the number of IFN-γ positive cells decreased significantly in the CD56dimNKG2A− NK cells, compared with the CD56dimNKG2A+ subset (p<0.05).
(A) Intracellular expression of IFN-γ by CD56bright (Bright), CD56dimNKG2A+ (D2A+) and CD56dimNKG2A− (D2A−) NK cell subsets from 7 healthy donors, after cell sorting and IL-12/IL-18 overnight activation. *: p<0.05; **: p<0.01; ***: p<0.001. (B) Representative example of NKG2A expression and IFN-γ production, before (UT) and after stimulation with IL-12/IL-18 in CD56bright (Bright), CD56dimNKG2A+ (D2A+) and CD56dimNKG2A− (D2A−) NK cell subsets.
Next, we evaluated the effect of IL12+IL18 stimulation on the CD56dimNKG2A− NK subset and found that a fraction of these cells re-expressed NKG2A (Fig. 4B). More intriguingly, up to 95% of this de novo subpopulation expressed IFN-γ, compared with 49.7±20.8% of the CD56dimNKG2A− subset (Fig. 4B). Similar results were observed with CD94, but NKG2C, the activating counterpart of NKG2A, did not increase after IL12+IL18 stimulation (data not shown). In addition, CD56bright and CD56dimNKG2A+ cells still expressed NKG2A after IL12+IL18 stimulation, regardless of their IFN-γ production (Fig. 4B). These results revealed that NKG2A could be specifically re-expressed on IFN-γ-producing-NKG2A− NK cells, after IL-12/IL-18 stimulation.
HLA-E expression on target cells restricts NK responsivity to the CD56dimNKG2A−KIR+ subset
Next we compared the degranulation capacity of these three subsets by assessing their ability to express CD107a at the cell surface. The response of CD56bright NK cells to antibody-dependent cell mediated cytotoxicity (ADCC) was poor, but increased strongly in the CD56dim subset, regardless of NKG2A expression (Fig. 5A). When we checked the degranulation of these NK cell subsets against HLA class-I negative K562 target cells, we observed that the direct cytotoxic capacity of CD56bright NK cells was much lower than that of the CD56dim subsets. More importantly, the level of CD107 expression was higher in CD56dimNKG2A+ than CD56dimNKG2A− NK cells against K562 target cells (Fig. 5A).
CD107a degranulation assays of CD56bright (Bright), CD56dimNKG2A+ (D2A+) and CD56dimNKG2A− (D2A−) NK cell subsets were accessed without target, or against K562 or RAJI cells in the presence of 1 µg of rituximab (ADCC) (A), or against 721.221 target cells that do or do not express HLA-E (LCL-221-AEH) (C). Box and whiskers plots of CD107a degranulation capacities of CD56bright (Bright), CD56dimNKG2A+KIR− (D2A+KIR−), CD56dimNKG2A+KIR+ (D2A+KIR+), CD56dimNKG2A−KIR+ (D2A−KIR+) and CD56dimNKG2A−KIR− (D2A−KIR−) NK cell subsets from 11 healthy controls. NK cells were tested without target cells, against RAJI cells in the presence of 1 µg of rituximab (ADCC), or K562 cells (B) or 721.221 target cells that do or do not express HLA-E (D). *: p<0.05; **: p<0.01; ***: p<0.001.
Because KIRs play an established role in the education of NK cells, it was crucial to test their relative functional effects. Accordingly, we assessed the degranulation capacity of the following five NK cell subsets: CD56bright, CD56dimNKG2A+KIR−, CD56dimNKG2A+KIR+, CD56dimNKG2A−KIR+ and CD56dimNKG2A−KIR−. As Fig. 5A and 5C show, CD56dimNKG2A−KIR− NK cells had significantly lower degranulation rates against K562 target cells than the other CD56dim NK cell subsets (p<0.001), explaining that CD56dimNKG2A− cells, regardless of KIR expression, are less able to degranulate against HLA class-I negative targets. Nonetheless, CD56dimNKG2A−KIR− NK cells retained a significant capacity to respond to ADCC compared to the other CD56dim subsets (Fig. 5A and 5C).
We next investigated whether degranulation of CD56bright, CD56dimNKG2A+KIR−, and CD56dimNKG2A+KIR+ NK cells was significantly inhibited in the presence of HLA-E-transfected 721.221 target cells, compared with the same cells against non-transfected 721.211 cells. Interestingly, CD56dimNKG2A−KIR+ cells were more functional against 721.221 target cells expressing HLA-E than against their non-transfected counterparts (Fig. 5B and 5D), which suggest preferential expression of NKG2C on CD56dimNKG2A− cells. Taken together, functional data show that each CD56dim NK cell expressing at least one inhibitory receptor is competent against HLA-I negative target cells. However, only the CD56dimNKG2A−KIR+ NK cells are competent to kill HLA-E-expressing target cells.
Phenotypic changing correlate with stepwise decrease of NKG2A and acquisition of KIR, suggesting a new complex model of NK cells terminal differentiation
Showing that KIRs coupled to NKG2A highlight functional subset specialization, we explored precisely the phenotype of these subsets. We looked markers that might provide clues about NK cell differentiation in the following five NK subsets: CD56bright, CD56dimNKG2A+KIR−, CD56dimNKG2A+KIR+ and CD56dimNKG2A−KIR+. Interestingly, the comparison of these subsets showed a continuum in the expression of some NK receptors, with a significant decrease of CD62L, CD27 and granzyme-K levels (Fig. 6A), with the progressive increase in markers associated with maturation, including ILT-2, CD57, Siglec-9 and FCRL6 (Fig. 6B). Finally, we analyzed in more detail the specific features of the hyporesponsive CD56dimNKG2A−KIR− NK cell subset. Comparison between competent (NKG2A−KIR+) and hyporesponsive (NKG2A−KIR−) CD56dim NK cells revealed that hyporesponsive cells expressed significantly more CD62L, and CD27 than competent cells. In contrast these hyporesponsive cells expressed less CD57, Siglec-9 and FCRL6 than competent NK cells (Fig. 6B). Together, these data support a new complex model of NK cell differentiation in which NKG2A expression, correlate inversely with KIRs expression (Fig. 7).
Expression of CD62L, granzyme-K and CD27 (A) or ILT2, CD57, Siglec-9 and FCRL6 (B) on CD56bright (Bright), CD56dimNKG2A+KIR− (D2A+KIR−), CD56dimNKG2A+KIR+ (D2A+KIR+), CD56dimNKG2A−KIR+ (D2A−KIR+) and CD56dimNKG2A−KIR− (D2A−KIR−) NK cell subsets from at least 16 healthy donors. *: p<0.05; **: p<0.01; ***: p<0.001.
NK cell repertoire gradually changed from CD56bright towards a CD56dim dominant response; CD56dimNKG2A+KIR− cells then develop into CD56dimNKG2A+KIR+ and then terminally differentiate into CD56dimNKG2A−KIR+ cells. Expression of KIR in a few CD56bright NK cells also suggested that a part of those cells give rise directly to CD56dimNKG2A+KIR+. This model also includes the potential re-expression of NKG2A on CD56dimNKG2A− NK cells after stimulation. Hyporesponsive CD56dimNKG2A−KIR− NK cells are integrated as cells that failed to acquire KIR before NKG2A loss due to their incompletely differentiate phenotype.
In this study, we set out to investigate the terminal differentiation of NK cells, focussing on NKG2A and KIRs as key markers during this dynamic process. Our data strongly endorse a new model of NK cell differentiation that begins with CD56bright cells, all expressing high-density NKG2A+ and few KIRs, and evolves to a set of CD56dimNKG2A−KIR+ cells, passing through intermediate stages, each characterised by unique phenotypic and functional features. A part of our results agree with those in a recent report presenting a simple linear differentiation model based on the expression of CD94 . However, differences in the expression pattern of other NK receptors have been noticed when comparing the inhibitory NKG2A+ and the activating NKG2C+ NK-cell subsets, which cannot be discriminated in CD94+ cells . Furthermore, CD94 does not permit to conclude about specific activities, because NKG2A and NKG2C have opposite functions. The association reported between elevated numbers of circulating NKG2C+ NK cells and positive serology findings for viruses suggests that these NKG2A and NKG2C receptors might be involved in different stages of NK cell ontogenetic development , , . Therefore this dichotomy, impossible when using CD94, is necessary for NK cells differentiation studies, particularly following pathogens encountering.
It is well established that CD56bright NK cells produce more cytokines than CD56dim NK cells in response to exogenous cytokine stimulation , . We found that treatment with IL12/IL18 increased the frequency of IFN-γ production significantly more in CD56bright NK cells than in CD56dimNKG2A− NK cells: CD56dimNKG2A+ cells responded more vigorously in term of IFN-γ production than did CD56dimNKG2A−, but still significantly far less than CD56bright NK cells, consistent with the recent data of Yu et al  on CD56dimCD94+ NK cells. Also, these results were in line with previous reports showing that CD56dimNKG2A−KIR+/− NK cells have the developmental potential to reacquire NKG2A after cytokines stimulation , . More importantly, our data provide evidence that the partial recovery of NKG2A correlates with IFN-γ production in CD56dimNKG2A− cells after IL-12/IL-18 stimulation. Interestingly, Gustafson et al demonstrated that IFN-γ induces, through STAT-1 pathway, a high increase of HLA-E expression in exposed cells . Therefore, IFN-γ production might induce its own regulation through HLA-E induction resulting in the inhibition of NKG2A expressing cells. These findings also indicate that CD56dimNKG2A− phenotype does not represent an irreversible state for NK cells; they also highlight the robust plasticity of NK cell response to various stimuli.
Previous evidences support the hypothesis that NK cell development proceeds sequentially from CD56bright to CD56dim NK cells , , . This model is buttressed by our data suggesting a robust link between the acquisition of functional capacities and both the stepwise decrease in NKG2A and the correlated acquisition of KIRs. Notably, we compared the CD56bright, CD56dimNKG2A+KIR−, CD56dimNKG2A+KIR+, and CD56dimNKG2A−KIR+ subsets and observed a gradient in the expression of NK markers between the three subsets, with a progressive decrease in CD62L, CD27 and granzyme-K, all highly expressed on CD56bright cells from peripheral blood, together with a progressive increase in receptors associated with maturity, such as ILT-2, FCRL6, Siglec-9 and CD57. This notion of a continuum or gradient along a specific pathway of differentiation is perfectly illustrated by CD57 expression, which has been recognised as a marker of in vitro replicative senescence of T lymphocytes  and which is known to be absent on cord blood NK cells and subsequently increase with elderly . Similarly, Cooley et al have shown that CD56dimKIR+ NK cells, which express high levels of CD57, proliferate more slowly . This supports our model of NK cells terminal differentiation proposed in Fig. 7, in which the NK cell repertoire gradually changes from a CD56bright toward a CD56dim dominant response, whereby the CD56dimNKG2A+KIR− subset develops first into a CD56dimNKG2A+KIR+ phenotype and then terminally differentiates into CD56dimNKG2A−KIR+ cells. As show by our functional data, these three CD56dim subsets display the same ability to degranulate against HLA-I− targets. More importantly, the acquisition of the terminal differentiated CD56dimNKG2A−KIR+ subset is crucial to acquire the capacity to respond against HLA-E+ target cells. Indeed, the degranulation of this subset is even increased by HLA-E expression on target cells. This particularity might be used in order to fight viruses like HCMV, which increase HLA-E expression on infected cells to escape immune system .
We also report the presence of a small population of CD56brightKIR+ NK cells, which has previously been described and which overlaps these developmental stages . It has been suggested that this KIR expression on CD56bright subset reflects their activation status in response to cytokines such as IL-15, which is responsible for homeostatic proliferation . We propose as an alternative that a fraction of the CD56dimNKG2A+KIR+ cells might be directly derived from CD56bright cells. Our results also show that a substantial fraction of CD56dim NK cells that lacked both NKG2A and KIRs were hyporesponsive, even against the HLA class I-deficient K562 target cells; however, the paucity of available KIR mAb limited our study and prevents us from concluding definitively that KIR− NK cells do not express any KIR. Intriguingly, we observed that these CD56dimNKG2A−KIR− hyporesponsive cells had partially preserved their capacity for ADCC. This is consistent with their normal levels of perforin and granzyme-B cytotoxic components, as initially described . It is likely that other receptors expressed by NK cells contribute to their education and may explain this activity. Recently Carrega et al  showed that hyporesponsive NK cells are fully responsive against autologous melanoma cells upon activation. In our study, expression of CD62L, CD27, CD57, Siglec-9 and FCRL6, suggests that CD56dimNKG2A−KIR− NK cells are not fully differentiated, but can be re-educated under specific stimulation to become fully functional, in line with our model (Fig. 7). This is strong evidence that at least part of NK cell education might occur in the periphery during immune response, as suggest by Juelke et al .
Not surprisingly, multiple levels of tolerance keep fully functional NK cells in check. It is now well established that interaction between inhibitory receptors and their self-MHC class-I ligands affects NK cell education. Different mechanisms to maintain self-tolerance may accompany the different expression and effector function profiles of the various NK cell subsets. The mechanism underlying education of NK cells via inhibitory KIRs and NKG2A receptors remains unknown, and the ontogenetic development of tolerance is poorly understood. As shown in Fig. 7, our model, which is based on phenotypic and functional characteristics of the different NK cell subsets, neither refutes nor supports any particular mechanism for the regulation of self-tolerance, including licensing, arming, disarming or tuning/rheostat , , .
Taken together, our data shed new light on NK cell physiology and suggest a continuum of NK cell differentiation. The dynamics of this process seems to be closely linked to KIR acquisition and simultaneous NKG2A loss, and gives rise to NK cells responsive against HLA-E+ target cells. Our findings provide detailed insight into the plasticity of the NK cell repertoire and allow our model to include the failure of KIR acquisition before NKG2A loss in hyporesponsive NK cells.
NKG2C is expressed mainly separately from NKG2A among CD56dimCD94+ NK cells. (A) Expression of NKG2C, NKG2A and CD94 on NK cells from two representative healthy donors (#7, #8). (B) Expression of NKG2C on CD56bright (Bright), CD56dimNKG2A+ (D2A+) and CD56dimNKG2A− (D2A−) NK cell subsets from 38 healthy donors. Statistical analysis used one-way ANOVA with Tukey post-test. *: p<0.05; **: p<0.01; ***: p<0.001.
(8.37 MB TIF)
Genes with significant modulations over the whole genome micro-arrays.
(0.22 MB XLS)
We thank A. Achour (INSERM UMR-S 945, Paris, France) for technical assistance. We also thank the personnel from the Etablissement Français du Sang (EFS) for the healthy adult blood samples.
Conceived and designed the experiments: VB PD VV. Performed the experiments: VB BD CP. Analyzed the data: VB PD VV. Wrote the paper: VB PD VV.
- 1. Vivier E, Tomasello E, Baratin M, Walzer T, Ugolini S (2008) Functions of natural killer cells. Nat Immunol 9: 503–510.E. VivierE. TomaselloM. BaratinT. WalzerS. Ugolini2008Functions of natural killer cells.Nat Immunol9503510
- 2. Moretta A, Marcenaro E, Parolini S, Ferlazzo G, Moretta L (2008) NK cells at the interface between innate and adaptive immunity. Cell Death Differ 15: 226–233.A. MorettaE. MarcenaroS. ParoliniG. FerlazzoL. Moretta2008NK cells at the interface between innate and adaptive immunity.Cell Death Differ15226233
- 3. Lanier LL (2003) Natural killer cell receptor signaling. Curr Opin Immunol 15: 308–314.LL Lanier2003Natural killer cell receptor signaling.Curr Opin Immunol15308314
- 4. Poli A, Michel T, Theresine M, Andres E, Hentges F, et al. (2009) CD56bright natural killer (NK) cells: an important NK cell subset. Immunology 126: 458–465.A. PoliT. MichelM. TheresineE. AndresF. Hentges2009CD56bright natural killer (NK) cells: an important NK cell subset.Immunology126458465
- 5. Ouyang Q, Baerlocher G, Vulto I, Lansdorp PM (2007) Telomere length in human natural killer cell subsets. Ann N Y Acad Sci 1106: 240–252.Q. OuyangG. BaerlocherI. VultoPM Lansdorp2007Telomere length in human natural killer cell subsets.Ann N Y Acad Sci1106240252
- 6. Romagnani C, Juelke K, Falco M, Morandi B, D'Agostino A, et al. (2007) CD56brightCD16- killer Ig-like receptor- NK cells display longer telomeres and acquire features of CD56dim NK cells upon activation. J Immunol 178: 4947–4955.C. RomagnaniK. JuelkeM. FalcoB. MorandiA. D'Agostino2007CD56brightCD16- killer Ig-like receptor- NK cells display longer telomeres and acquire features of CD56dim NK cells upon activation.J Immunol17849474955
- 7. Yu J, Mao HC, Wei M, Hughes T, Zhang J, et al. (2009) CD94 surface density identifies a functional intermediary between the CD56bright and CD56dim human NK cell subsets. Blood 115: 274–281.J. YuHC MaoM. WeiT. HughesJ. Zhang2009CD94 surface density identifies a functional intermediary between the CD56bright and CD56dim human NK cell subsets.Blood115274281
- 8. Vago L, Forno B, Sormani MP, Crocchiolo R, Zino E, et al. (2008) Temporal, quantitative, and functional characteristics of single-KIR-positive alloreactive natural killer cell recovery account for impaired graft-versus-leukemia activity after haploidentical hematopoietic stem cell transplantation. Blood 112: 3488–3499.L. VagoB. FornoMP SormaniR. CrocchioloE. Zino2008Temporal, quantitative, and functional characteristics of single-KIR-positive alloreactive natural killer cell recovery account for impaired graft-versus-leukemia activity after haploidentical hematopoietic stem cell transplantation.Blood11234883499
- 9. Nguyen S, Dhedin N, Vernant JP, Kuentz M, Al Jijakli A, et al. (2005) NK-cell reconstitution after haploidentical hematopoietic stem-cell transplantations: immaturity of NK cells and inhibitory effect of NKG2A override GvL effect. Blood 105: 4135–4142.S. NguyenN. DhedinJP VernantM. KuentzA. Al Jijakli2005NK-cell reconstitution after haploidentical hematopoietic stem-cell transplantations: immaturity of NK cells and inhibitory effect of NKG2A override GvL effect.Blood10541354142
- 10. Dulphy N, Haas P, Busson M, Belhadj S, Peffault de Latour R, et al. (2008) An unusual CD56(bright) CD16(low) NK cell subset dominates the early posttransplant period following HLA-matched hematopoietic stem cell transplantation. J Immunol 181: 2227–2237.N. DulphyP. HaasM. BussonS. BelhadjR. Peffault de Latour2008An unusual CD56(bright) CD16(low) NK cell subset dominates the early posttransplant period following HLA-matched hematopoietic stem cell transplantation.J Immunol18122272237
- 11. Beziat V, Nguyen S, Lapusan S, Hervier B, Dhedin N, et al. (2009) Fully functional NK cells after unrelated cord blood transplantation. Leukemia 23: 721–728.V. BeziatS. NguyenS. LapusanB. HervierN. Dhedin2009Fully functional NK cells after unrelated cord blood transplantation.Leukemia23721728
- 12. Le Garff-Tavernier M, Beziat V, Decocq J, Siguret V, Gandjbakhch F, et al. (2010) Human NK cells display major phenotypic and functional changes over the life span. Aging Cell 9: 527–535.M. Le Garff-TavernierV. BeziatJ. DecocqV. SiguretF. Gandjbakhch2010Human NK cells display major phenotypic and functional changes over the life span.Aging Cell9527535
- 13. Sullivan LC, Clements CS, Rossjohn J, Brooks AG (2008) The major histocompatibility complex class Ib molecule HLA-E at the interface between innate and adaptive immunity. Tissue Antigens 72: 415–424.LC SullivanCS ClementsJ. RossjohnAG Brooks2008The major histocompatibility complex class Ib molecule HLA-E at the interface between innate and adaptive immunity.Tissue Antigens72415424
- 14. Borrego F, Masilamani M, Kabat J, Sanni TB, Coligan JE (2005) The cell biology of the human natural killer cell CD94/NKG2A inhibitory receptor. Mol Immunol 42: 485–488.F. BorregoM. MasilamaniJ. KabatTB SanniJE Coligan2005The cell biology of the human natural killer cell CD94/NKG2A inhibitory receptor.Mol Immunol42485488
- 15. Parham P (2006) Taking license with natural killer cell maturation and repertoire development. Immunol Rev 214: 155–160.P. Parham2006Taking license with natural killer cell maturation and repertoire development.Immunol Rev214155160
- 16. Joncker NT, Raulet DH (2008) Regulation of NK cell responsiveness to achieve self-tolerance and maximal responses to diseased target cells. Immunol Rev 224: 85–97.NT JonckerDH Raulet2008Regulation of NK cell responsiveness to achieve self-tolerance and maximal responses to diseased target cells.Immunol Rev2248597
- 17. Held W (2008) Tolerance and reactivity of NK cells: two sides of the same coin? Eur J Immunol 38: 2930–2933.W. Held2008Tolerance and reactivity of NK cells: two sides of the same coin?Eur J Immunol3829302933
- 18. Brodin P, Karre K, Hoglund P (2009) NK cell education: not an on-off switch but a tunable rheostat. Trends Immunol 30: 143–149.P. BrodinK. KarreP. Hoglund2009NK cell education: not an on-off switch but a tunable rheostat.Trends Immunol30143149
- 19. Saeed AI, Sharov V, White J, Li J, Liang W, et al. (2003) TM4: a free, open-source system for microarray data management and analysis. Biotechniques 34: 374–378.AI SaeedV. SharovJ. WhiteJ. LiW. Liang2003TM4: a free, open-source system for microarray data management and analysis.Biotechniques34374378
- 20. Huang da W, Sherman BT, Lempicki RA (2009) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat Protoc 4: 44–57.W. Huang daBT ShermanRA Lempicki2009Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources.Nat Protoc44457
- 21. Dennis G Jr, Sherman BT, Hosack DA, Yang J, Gao W, et al. (2003) DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 4: P3.G. Dennis JrBT ShermanDA HosackJ. YangW. Gao2003DAVID: Database for Annotation, Visualization, and Integrated Discovery.Genome Biol4P3
- 22. Lee N, Llano M, Carretero M, Ishitani A, Navarro F, et al. (1998) HLA-E is a major ligand for the natural killer inhibitory receptor CD94/NKG2A. Proc Natl Acad Sci U S A 95: 5199–5204.N. LeeM. LlanoM. CarreteroA. IshitaniF. Navarro1998HLA-E is a major ligand for the natural killer inhibitory receptor CD94/NKG2A.Proc Natl Acad Sci U S A9551995204
- 23. Braud VM, Allan DS, O'Callaghan CA, Soderstrom K, D'Andrea A, et al. (1998) HLA-E binds to natural killer cell receptors CD94/NKG2A, B and C. Nature 391: 795–799.VM BraudDS AllanCA O'CallaghanK. SoderstromA. D'Andrea1998HLA-E binds to natural killer cell receptors CD94/NKG2A, B and C.Nature391795799
- 24. Saez-Borderias A, Romo N, Magri G, Guma M, Angulo A, et al. (2009) IL-12-dependent inducible expression of the CD94/NKG2A inhibitory receptor regulates CD94/NKG2C+ NK cell function. J Immunol 182: 829–836.A. Saez-BorderiasN. RomoG. MagriM. GumaA. Angulo2009IL-12-dependent inducible expression of the CD94/NKG2A inhibitory receptor regulates CD94/NKG2C+ NK cell function.J Immunol182829836
- 25. Yu Q, Sharma A, Oh SY, Moon HG, Hossain MZ, et al. (2009) T cell factor 1 initiates the T helper type 2 fate by inducing the transcription factor GATA-3 and repressing interferon-gamma. Nat Immunol 10: 992–999.Q. YuA. SharmaSY OhHG MoonMZ Hossain2009T cell factor 1 initiates the T helper type 2 fate by inducing the transcription factor GATA-3 and repressing interferon-gamma.Nat Immunol10992999
- 26. Upshaw JL, Arneson LN, Schoon RA, Dick CJ, Billadeau DD, et al. (2006) NKG2D-mediated signaling requires a DAP10-bound Grb2-Vav1 intermediate and phosphatidylinositol-3-kinase in human natural killer cells. Nat Immunol 7: 524–532.JL UpshawLN ArnesonRA SchoonCJ DickDD Billadeau2006NKG2D-mediated signaling requires a DAP10-bound Grb2-Vav1 intermediate and phosphatidylinositol-3-kinase in human natural killer cells.Nat Immunol7524532
- 27. Kuo CT, Leiden JM (1999) Transcriptional regulation of T lymphocyte development and function. Annu Rev Immunol 17: 149–187.CT KuoJM Leiden1999Transcriptional regulation of T lymphocyte development and function.Annu Rev Immunol17149187
- 28. Lopez-Botet M, Bellon T, Llano M, Navarro F, Garcia P, et al. (2000) Paired inhibitory and triggering NK cell receptors for HLA class I molecules. Hum Immunol 61: 7–17.M. Lopez-BotetT. BellonM. LlanoF. NavarroP. Garcia2000Paired inhibitory and triggering NK cell receptors for HLA class I molecules.Hum Immunol61717
- 29. Guma M, Cabrera C, Erkizia I, Bofill M, Clotet B, et al. (2006) Human cytomegalovirus infection is associated with increased proportions of NK cells that express the CD94/NKG2C receptor in aviremic HIV-1-positive patients. J Infect Dis 194: 38–41.M. GumaC. CabreraI. ErkiziaM. BofillB. Clotet2006Human cytomegalovirus infection is associated with increased proportions of NK cells that express the CD94/NKG2C receptor in aviremic HIV-1-positive patients.J Infect Dis1943841
- 30. Ahlenstiel G, Titerence RH, Koh C, Edlich B, Feld JJ, et al. (2009) Natural Killer Cells Are Polarized Toward Cytotoxicity in Chronic Hepatitis C in an Interferon-Alfa-Dependent Manner. Gastroenterology 138: 325–335.G. AhlenstielRH TiterenceC. KohB. EdlichJJ Feld2009Natural Killer Cells Are Polarized Toward Cytotoxicity in Chronic Hepatitis C in an Interferon-Alfa-Dependent Manner.Gastroenterology138325335
- 31. Oliviero B, Varchetta S, Paudice E, Michelone G, Zaramella M, et al. (2009) Natural killer cell functional dichotomy in chronic hepatitis B and chronic hepatitis C virus infections. Gastroenterology 137: 1151–1160, e1151-1157.B. OlivieroS. VarchettaE. PaudiceG. MicheloneM. Zaramella2009Natural killer cell functional dichotomy in chronic hepatitis B and chronic hepatitis C virus infections.Gastroenterology13711511160, e1151-1157
- 32. Cooper MA, Fehniger TA, Turner SC, Chen KS, Ghaheri BA, et al. (2001) Human natural killer cells: a unique innate immunoregulatory role for the CD56(bright) subset. Blood 97: 3146–3151.MA CooperTA FehnigerSC TurnerKS ChenBA Ghaheri2001Human natural killer cells: a unique innate immunoregulatory role for the CD56(bright) subset.Blood9731463151
- 33. Fauriat C, Long EO, Ljunggren HG, Bryceson YT (2009) Regulation of human NK cell cytokine and chemokine productionby target cell recognition. Blood 115: 2167–2176.C. FauriatEO LongHG LjunggrenYT Bryceson2009Regulation of human NK cell cytokine and chemokine productionby target cell recognition.Blood11521672176
- 34. Cooley S, Xiao F, Pitt M, Gleason M, McCullar V, et al. (2007) A subpopulation of human peripheral blood NK cells that lacks inhibitory receptors for self-MHC is developmentally immature. Blood 110: 578–586.S. CooleyF. XiaoM. PittM. GleasonV. McCullar2007A subpopulation of human peripheral blood NK cells that lacks inhibitory receptors for self-MHC is developmentally immature.Blood110578586
- 35. Juelke K, Killig M, Thiel A, Dong J, Romagnani C (2009) Education of hyporesponsive NK cells by cytokines. Eur J Immunol 39: 2548–2555.K. JuelkeM. KilligA. ThielJ. DongC. Romagnani2009Education of hyporesponsive NK cells by cytokines.Eur J Immunol3925482555
- 36. Gustafson KS, Ginder GD (1996) Interferon-gamma induction of the human leukocyte antigen-E gene is mediated through binding of a complex containing STAT1alpha to a distinct interferon-gamma-responsive element. J Biol Chem 271: 20035–20046.KS GustafsonGD Ginder1996Interferon-gamma induction of the human leukocyte antigen-E gene is mediated through binding of a complex containing STAT1alpha to a distinct interferon-gamma-responsive element.J Biol Chem2712003520046
- 37. Brenchley JM, Karandikar NJ, Betts MR, Ambrozak DR, Hill BJ, et al. (2003) Expression of CD57 defines replicative senescence and antigen-induced apoptotic death of CD8+ T cells. Blood 101: 2711–2720.JM BrenchleyNJ KarandikarMR BettsDR AmbrozakBJ Hill2003Expression of CD57 defines replicative senescence and antigen-induced apoptotic death of CD8+ T cells.Blood10127112720
- 38. Tomasec P, Braud VM, Rickards C, Powell MB, McSharry BP, et al. (2000) Surface expression of HLA-E, an inhibitor of natural killer cells, enhanced by human cytomegalovirus gpUL40. Science 287: 1031.P. TomasecVM BraudC. RickardsMB PowellBP McSharry2000Surface expression of HLA-E, an inhibitor of natural killer cells, enhanced by human cytomegalovirus gpUL40.Science2871031
- 39. Freud AG, Yokohama A, Becknell B, Lee MT, Mao HC, et al. (2006) Evidence for discrete stages of human natural killer cell differentiation in vivo. J Exp Med 203: 1033–1043.AG FreudA. YokohamaB. BecknellMT LeeHC Mao2006Evidence for discrete stages of human natural killer cell differentiation in vivo.J Exp Med20310331043
- 40. Anfossi N, Andre P, Guia S, Falk CS, Roetynck S, et al. (2006) Human NK cell education by inhibitory receptors for MHC class I. Immunity 25: 331–342.N. AnfossiP. AndreS. GuiaCS FalkS. Roetynck2006Human NK cell education by inhibitory receptors for MHC class I.Immunity25331342
- 41. Carrega P, Pezzino G, Queirolo P, Bonaccorsi I, Falco M, et al. (2009) Susceptibility of human melanoma cells to autologous natural killer (NK) cell killing: HLA-related effector mechanisms and role of unlicensed NK cells. PLoS One 4: e8132.P. CarregaG. PezzinoP. QueiroloI. BonaccorsiM. Falco2009Susceptibility of human melanoma cells to autologous natural killer (NK) cell killing: HLA-related effector mechanisms and role of unlicensed NK cells.PLoS One4e8132