Natural killer (NK) cell effector functions include natural cytotoxicity, antibody-dependent cell-mediated cytotoxicity (ADCC), and secretion of cytokines such as interferon γ (IFN-γ) and tumor necrosis factor α. Two subsets of human NK cells depending on the density of CD56 expression have been identified. CD56^dimCD16⁺CD3⁻ cells usually account for more than 90% of the NK cells in peripheral blood, whereas CD56^brightCD16^dim/−CD3⁻ cells are more common in lymphoid organs (1–3). The major effector function of the former cells may be cytotoxicity, whereas the latter may act mainly via cytokine secretion (2). However, most NK cells that are cytotoxic and/or produce IFN-γ on stimulation with different types of target cells (4–7), including K562 and antibody-coated target cells (5), belong to the CD56^dim subset. In contrast, NK cells that readily respond to cytokines such as IL-12 and IL-15, belong to the CD56^bright NK cell subset (2, 5). CD56^dim and CD56^bright NK cells may be more appropriately defined as “target cell-responsive” and “cytokine-responsive,” respectively (5).

The regulation of NK cell functions depends on a very fine balance between signals mediated by activating receptors (ARs) and inhibitory receptors (IRs) (6, 8). ARs mainly include the natural cytotoxicity receptors (NKp46/CD335, NKp44/CD336, NKp30/CD337), NKG2D/CD314, 2B4/CD244, and FcγRIIIa/CD16a, one of the low-affinity immunoglobulin G (IgG) receptors involved in ADCC (8, 9). IRs mainly include the C-type lectin NKG2A/CD94 heterodimer receptor, which recognizes human leukocyte antigen (HLA)-E molecules and killer Ig-like receptors (KIR) such as KIR2DL1 (CD158a), specific to the HLA-C group C2 allotype, and KIR2DL2/3 (CD158b), specific to the HLA-C group C1 allotype (10, 11).

According to the process referred to as “education or licensing” of NK cells, acquisition of functional responses depends on the engagement of IRs with self-ligands during their development (5, 12, 13). Remarkably, the vast phenotypic diversity in the human NK cell repertoire is related to the broad range of possible combinations of phenotypes on a single cell from a given donor. Thus, all NKG2A and KIR expression patterns are represented, including NK cells lacking IRs for self, which remain hyporesponsive (5, 12, 13).

Activating receptors involved in natural cytotoxicity such as NCR, NKG2D, and 2B4 can signal independently, but functional responses, including cytotoxicity and cytokine synthesis, require a combination of signals resulting from two or more interactions between different receptor–ligand pairs (14–16). By contrast, the FcγRIIIa receptor is unique in its ability to induce both responses without additional signal provided by co-engagement of other ARs (14–16). A partial dichotomy between IFN-γ-producing and degranulating NK cells upon FcγRIIIa engagement by anti-CD16-sensitized P815 cells (5) or by CD20⁺ cells opsonized with the therapeutic anti-CD20 monoclonal antibody (mAbs) rituximab (RTX) or obinituzumab (17) was previously reported. How a given AR induces different functional responses within the polyclonal NK cells of a given donor was not specifically discussed.

A stepwise differentiation/maturation of NK cells from the immature CD56^brightCD16⁻ (NKG2A⁺⁺KIR⁻) cells through the intermediate CD56^brightCD16^dim stage to the mature CD56^dimCD16⁺ (NKG2A^±KIR^±) population is usually admitted (18–21). A further differentiation/maturation of the CD56^dimCD16⁺ subset according to the gradual loss of NKG2A and CD62L and/or the gradual gain of KIRs and CD57 (21–26) has been demonstrated, supporting the concept of a continuous process starting from CD56^brightNKG2A⁺⁺KIR⁻CD62L⁺CD57⁻cells and ending with the CD56^dimNKG2A⁻KIR⁺CD62L⁻CD57⁺ phenotype. This phenotype change is associated with a shift in functionality from cytotoxicity/degranulation toward IFN and TNF secretion in response to ARs stimulation (27). While this effect is most strikingly observed in CD57⁺ NK cells, it has also been observed when comparing NKG2A⁺KIR⁻ with NKG2A⁻KIR⁺ NK cells stimulated by target cells in the context of NK cell transplantation (7). In addition, it has been shown that activation of CD56^dim NK cells results in the down-modulation of FcγRIIIa by ADAM17-mediated shedding or internalization (16, 28), resulting in the presence of CD56^dimCD16⁻ cells in peripheral blood. Finally, the FCGR3A gene, which encodes FcγRIIIa, displays a functional allelic dimorphism generating allotypes with either a phenylalanine (F) or a valine (V) at amino acid position 158 (29, 30). The V158F polymorphism of FcγRIIIA, which is associated with higher therapeutic response to RTX (31–33) or to the anti-ErB-2 mAb trastuzumab (TTZ) used in brain cancer (34), has also been related to interindividual variations in FcγRIIIA expression (35, 36), although this is not confirmed (9, 37). In the course of flow cytometry (FCM) analysis of human NK cells, we also noted a substantial intraindividual heterogeneity of FcγRIIIa expression on CD56^dim NK cells in accordance with a recent study, reporting the presence of an individualized subset of CD56^dim cells expressing low level of CD16 in human peripheral blood (38).

Here, we aimed to investigate whether individual differentiation/maturation of polyclonal CD56^dim NK cells defined by expression of NKG2A/KIR2DL is related to FcγRIIIa expression and to the heterogeneity of NK cell responses upon FcγRIIIa engagement. We used multi-color FCM (25, 39) to simultaneously evaluate the individual expression of NKG2A, CD158a, and CD158b on CD56^dim NK cells and (1) the level of FcγRIIIa expression on unstimulated cells and (2) the degranulation and IFN-γ production in response to FcγRIIIa engagement by plate-bound anti-CD16 mAb, or IgG1 therapeutic mAbs RTX and TTZ.

In this study, we show that within a given individual a gradual increase of FcγRIIIa expression is associated with the differentiation/maturation of CD56^dim NK cells from NKG2A⁺CD158a,h/b,j⁻ toward NKG2A⁻CD158/a,h/b,j⁺. We show, in addition, that FcγRIIIa engagement by using a murine anti-CD16 mAb or by the Fc portion of human therapeutic mAbs resulted in donor-dependent partial functional segregation of IFN-γ-producing and/or degranulating CD56^dim cells. Importantly, the proportion of CD158a,h/b,j⁺ cells and that of NKG2A⁺ cells was increased and decreased, respectively, in IFN-γ-producing cells, whereas the frequency of CD158a,h⁺, CD158b,j⁺, and NKG2A⁺ cells proportions were poorly modified in degranulating cells. Similar results were observed after engagement of ARs involved in natural cytotoxicity. Our results further support the notion of a continuous differentiation/maturation of CD56^dim NK cells defined phenotypically by a gradual increase of FcγRIIIa expression and associated with a shift in functionality toward IFN-γ secretion observed upon both FcγRIIIa-independent and FcγRIIIa-dependent stimulation.

The level of FcγRIIIa expression on NK cells has been shown to depend on several factors. For instance, we and others have demonstrated that the activation of CD56^dim NK cells results in the down-modulation of FcγRIIIa by ADAM17-mediated shedding or internalization (16, 25). Moreover, interindividual variations in FcγRIIIA expression related to the V158F polymorphism of FcγRIIIA has been reported (35, 36), although this is not confirmed (9, 37). Our results are unrelated to these observations. Indeed, we evaluated the correlation of FcγRIIIa and IR expression on unstimulated cells and compared cells of given individuals, excluding any effect of FcγRIIIa polymorphism. Variation of FcγRIIIa expression has been originally related to the existence of different subsets or maturation stages of NK cells defined by the level of CD56 expression: CD56^bright cells lack FcγRIIIa expression or exhibit low-density expression, whereas most CD56^dim cells are considered to express high level of FcγRIIIa (1, 2, 40). Thus, the relationship between expression of FcγRIIIa and the diffenciation/maturation of the latter cells has been sparsely studied. The differentiation/maturation of CD56^dim NK cells is associated with gain (KIRs and CD57) and loss (NKG2A and CD62L) of protein expression on the cell surface (21–26). On the one hand, Amand et al. have recently described a new subset of CD56^dim NK cells with low level of FcγRIIIa (38). Phenotypically, the new subset contained a high percentage of relatively immature cells, as reflected by a significantly stronger representation of NKG2A⁺ and CD57⁻ cells. It is likely that this population overlaps, at least partially, with the 20% of CD56^dim cells expressing the lower level of FcγRIIIa shown in Figure 1, which contained increased and decreased proportion of NKG2A⁺ and CD158a,h/b,j⁺ cells, respectively. On the other hand, the terminally differentiated CD57⁺ subset representing 30–60% cells has been shown to express a slightly higher (1.2-fold) level of FcγRIIIa as compared with the CD57⁻ subset and to contain a higher proportion of CD158⁺ cells and lower proportion of NKG2A⁺ cells (25, 26). In humans, a clonal or oligoclonal expansion of a subset of NK cells expressing or not the CD94-NKG2C receptor has been observed after human cytomegalovirus (HCMV) infection. These adaptive NK cells exhibit a surface receptor signature a mature phenotype (i.e., showing progressive gain of CD57 and KIRs and loss of NKG2A) and epigenetic remodeling. Furthermore, some adaptive NK cells display deficient expression of FcεRγ. These NK cells exhibit more robust IFN-γ and TNF production, but not degranulation after FcγRIIIa engagement. It is, however, of note that FcγRIIIa is equally expressed by conventional and adaptive NK cells. We also observed an association between FcγRIIIa, CD158a,h/b,j, and NKG2A expression. However, the association was a continuum rather than a positive/negative association: the greater the expression of FcγRIIIa on unstimulated CD56^dim cells, the greater and the lower the probability to coexpress CD158a,h/b,j and NKG2A, respectively (Figure 1B). Moreover, the magnitude of FcγRIIIa expression was substantial: the highest level of FcγRIIIa expression on CD56^dim NK cells from a given individual was 5- to 10-fold that of the lowest level (Figure 1A; Figure S1 in Supplementary Material). Therefore, the overexpression of FcγRIIIa we observed on CD158a,h/b,j⁺CD56^dim cells cannot be explained by the increased frequency of CD57⁺ cells observed within the latter population, indicating that it occurs before terminal differentiation marked by the acquisition of CD57. Thus, the gradual increase of FcγRIIIa expression is an important feature of the differentiation/maturation of CD56^dim cells, and we propose to use it as a new marker of this process.

A functional segregation related to CD56 expression has been previously reported: CD56^bright cells act mainly by secretion of cytokines, whereas CD56^dim cells are more prone to exert cytotoxicity (2, 3). CD56^bright cells respond primarily to monokine stimulation, whereas CD56^dim cells respond primarily to target cells (2, 5), so this functional dichotomy may depend mainly on the stimulation conditions. With CD56^dim cells, several reports have shown that IFN-γ⁺ and CD107a⁺ NK cells may be expressed in a mutually exclusive manner on simulation with different target cells including K562 cells (5, 7), P815 cells sensitized with anti-CD16 mAb (5), CD20⁺ cells sensitized with RTX or obinituzumab (17), Plasmodium falciparum-infected red blood cells (4) and Drosophila cells expressing several AR ligands (6). Our results, obtained in specific response to FcγRIIIa engagement by plate-bound anti-CD16 mAb or therapeutic mAbs or mAbs directed to ARs involved in natural cytotoxicity (i.e., in the absence of target cells), agree with these observations. However, Foley et al. observed a high frequency of CD107a-expressing cells not producing IFN-γ, an intermediate frequency of cells exhibiting both responses, and a low frequency of NK cells producing IFN-γ but not expressing CD107a (7). In our experiments, the subset that produced IFN-γ but did not degranulate usually dominated the subset of double-positive cells. This discrepancy may be related to the stimulation condition. In line with this is our observation that degranulating cells but not IFN-γ-producing cells from a given donors were substantially higher after stimulation with 3G8 than after stimulation with RTX or TTZ. The discrepancy may also result from our use of isolated NK cells vs thawed PBMCs in the previous study (7).

Fauriat et al. assumed that functional segregation may reflect differences in the kinetics of the responses (6). Indeed, in agreement with our results, degranulation occurred earlier than IFN-γ production. However, in our study, the dichotomy was observed as soon as IFN-γ production was detected and persisted for up to 20 h. This finding ruled out that the functional segregation we observed was related to the fact that cells that had already degranulated might produce IFN-γ later on. This conclusion is supported by our finding that the functional responses of CD56^dim NK cells were associated with different phenotypes: CD158a,h/b,j⁺ cells were more prone than NKG2A⁺ cells to produce IFN-γ in response to FcγRIIIa-dependent and FcγRIIIa-independent stimulation. These results are consistent with those reporting KIR expression associated with target cell-induced IFN-γ production by NK cells, but NKG2A was sufficient for degranulation (7, 25–27).

Distinct intracellular signaling pathways originating from different ARs lead to cytokine secretion or cytotoxicity (41–45). By contrast, the ability of one AR such as FcγRIIIa to preferentially induce one or the other response in individual NK cells demonstrates that a given pathway may lead to different responses in different NK cells from a given donor. Modest differences in MFI from donor to donor for any given AR may have a profound effect on the functional response to stimulation. IFN-γ production requires a higher level of activation than does degranulation (6). It may, therefore, be assumed that the high level of FcγRIIIa expression observed on CD158a,h/b,j⁺ cells, accounts for their tendency to preferentially produce cytokines in response to FcγRIIIa engagement. However, a clonal or oligoclonal expansion of a subset of NK cells expressing or not the CD94-NKG2C receptor has been observed after HCMV infection. These cells, which are called adaptive NK cells, exhibit a mature phenotype (NKG2A⁻CD57⁺KIRs⁺) and epigenetic remodeling. Although adaptive and conventional mature NK cells express equally FcγRIIIa, adaptive NK cells display more robust IFN-γ and TNF production, but not degranulation in response to FcγRIIIa engagement [recently reviewed in Ref. (46, 47)]. In addition, it has been shown that the ability of NK cells to produce IFN-γ in response to engagement of ARs involved in natural cytotocicity is gradually acquired and related to epigenetic remodeling of the IFNG promoter, during their terminal differentiation (27). It is, therefore, more likely that this epigenetic remodeling is a general feature of more differentiated cells and is responsible for the shift toward IFN secretion observed in the present study in response to FcγRIIIa engagement.

Finally, which function (killing or cytokine production) mediates clinical responses to therapeutic mAbs is not known. Among the most convincing evidence that ADCC plays a role in mediating the clinically relevant antitumor response to therapeutic mAbs is the demonstration by our group (31) and others (32–34) that FCGR3A gene polymorphism is associated with clinical responses to different cytolytic mAbs such as RTX (31–33), TTZ (34), and cetuximab (48). In accordance with this, an in vitro genotype–phenotype association has been observed: the FCGR3A polymorphism affects the concentration-effect association of rituximab-mediated ADCC by NK cells (37). These studies show that FcγRIIIa-expressing cells are involved in the mechanism of action of these mAbs, but they did not demonstrate which FcγRIIIa-expressing cells or which effector functions are involved in the in vivo situation. Although NK cells are considered to act through ADCC to mediate the mechanism of action of the different cytolytic mAbs, an indirect mechanism, whereby NK cells act by recruiting cells via FcγRIIIa-dependent cytokine production, remains possible. Following this hypothesis, functional heterogeneity related to the differentiation/maturation of CD56^dim NK cells could be involved in the variability of the clinical response observed in patients treated with cytolytic mAbs such as RTX or TTZ.

Peripheral blood mononuclear cells were exclusively obtained from the blood of healthy volunteers (i.e., blood donors from the Etablissement français du Sang Centre-Atlantique, who had given their written informed consent) according to institutional research protection guidelines (Agreement No IMMUNOUMR6239/37/12/01).

LL, NC-J, and AB designed and performed the experiments. LL performed the statistical analysis. LL, NC-J, and GT analyzed the results. LL and GT wrote the manuscript, GT supervised the study conception and design. All authors critically revised the work, provided substantial input, and gave final approval to the version to be published.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer SV and handling editor declared their shared affiliation.