Cytotoxic functions and susceptibility to apoptosis of human CD56bright NK cells differentiated in vitro from CD34+ hematopoietic progenitors



Cytotoxic functions and susceptibility to apoptosis are crucial aspects of NK cells suitable to counter cancer after infusion in oncologic patients. To test the feasibility and the usefulness of infusing in vitro generated NK cells, these two features were investigated in NK cells developed in vitro from CD34+ hematopoietic progenitors. Purified CD34+ cells were cultured for 15–30 days with FLT-3 ligand (FLT3-L) and IL-15 with or without IL-21. To induce terminal differentiation, NK cells were cultured for further 15 days with IL-15, IL-21, or their combination. A CD56dim/CD16+ NK subset, expressing high level of perforin, granzymes, and LFA-1, appeared early in cultures with FLT3-L, IL-15, and IL-21, but it quickly died, indicating its predisposition to apoptosis. On the contrary, CD56bright NK cells generated after 30 days of culture with FLT3-L plus IL-15 did not show a considerable apoptosis, nevertheless only a subset of these cells expressed granzyme-B, perforin, LFA-1, and CD94-CD159a heterodimer, indicating a functional immaturity. Interestingly, further 15 days of culture with IL-21 plus IL-15 did not induce the generation of CD56dim cells from the CD56bright subset and actually inhibited IL-15-induced maturation/activation of this latter subset. In fact, IL-15 alone upregulated granzyme-B, TRAIL, Fas ligand, CD94-CD159a, LFA-1, CD16, KIRs, and TRAIL-R2 on CD56bright NK cells. Our results suggest that during differentiation CD56bright NK cells, similarly to mature activated NK cells, become highly cytotoxic and are relatively resistant to apoptosis induced by TNF family members. © 2012 International Society for Advancement of Cytometry.

The NK functional activities and the end of NK cell responses are finely controlled by NK cell receptors (NKRs) with activatory and inhibitory functions. To this regard, NK cells express inhibitory receptors for MHC class I (MHC-I) molecules, the killer cell immunoglobulin (Ig)-like receptors (KIRs) and C-type lectin CD94-CD159a, which are able to block activatory signals. These latter signals are mainly triggered by CD16, natural cytotoxicity receptors (NCRs), NKG2D, and the leukocyte adhesion molecule DNAM-1 (CD226) (1, 2). On the other hand, TNF ligand and receptor members are also involved both in NK cell-mediated cytotoxic function and in the control of NK cell response (3, reviewed in Ref.4).

Within the NK compartment, two phenotypically and functionally distinct NK subsets, CD56bright and CD56dim, have been described to represent either different stages of NK cell differentiation (Stage 4 and 5, respectively) or different subpopulations (similarly to CD4 and CD8 T cells, respectively) (2, 5). In particular, CD56dim/CD16bright/CD117neg NK cells are highly cytotoxic and preferentially express KIR inhibitory receptors, while the cytokine-producer CD56bright NK cells are characterized by CD117 (c-kit) cytokine receptor and CD94-CD159a inhibitory receptor, although they also express low percentages of KIR (5–10%) and CD16dim (20–30%) receptors (5–7). In vitro cultures of hematopoietic progenitors with addition of IL-21 favor the differentiation of hematopoietic progenitor cells toward CD56dim/CD16+ NK subset (8, 9). Differently, the classic cytokine combinations of FLT-3 ligand (FLT3-L) or stem cell factor plus IL-15 or IL-2 lead to the generation of CD56bright NK cells (2, 5, 7, 10–12). Moreover, IL-15 and IL-2 induce not only NK differentiation but also activation, suggesting that in vitro differentiated NK cells could be activated. As a matter of fact, cytokine activation of mature NK cells leads to an upregulation of intracellular lytic proteins (granzymes and perforin) and surface molecules such as adhesion, activatory, and TNF family proteins (3, reviewed in Ref.4), that render highly cytotoxic not only CD56dim but also CD56bright NK cells (13, 14). This in turn generates the so-called lymphokine-activated killer (LAK) cells able to perform spontaneous cytotoxicity via NKR/Ca++-dependent release of granzymes and perforins and/or via NKR/Ca++-independent apoptosis of sensitive target cells, mediated by TNF family members (reviewed in Ref.4). Interestingly, the same TNF receptor members exploited by LAK cells to kill their targets are able to drive activated NK cell apoptosis, finally controlling the LAK response but also limiting NK cell efficacy in adoptive transfer immunotherapies against tumor cells (4). Indeed, mature ex-vivo activated NK cells seems to have only a limited ability to counter cancer cells after their injection, instead, NK cells generated after CD34+ stem cells transplantation (particularly in KIR ligand mismatch regimens) can drive to a more efficient graft versus leukemia reaction (reviewed in Ref.4), suggesting that newly developed NK cells might be more suitable against tumor cells.

During differentiation and activation, NK cells sequentially acquire surface and intracellular molecules able to control both their cytotoxic functions and their susceptibility to apoptosis in a fashion that leads to self-tolerance, but the specific sequence through which this happens it has been only partially unraveled (10, 15). We had already described an immature CD161+/CD56 subset able to kill only via a TRAIL-dependent mechanism (11, 16). The phenotypical and functional characteristics of this subset strongly remind the immature NK Stage 3 classically depicted in lymph nodes by CD117 and CD94 markers (CD117bright/CD94) (17) and, more recently in cord and peripheral blood (PB) samples, by the low level of LFA-1 expression (18). The phenotype and the scatter characteristics (small agranular) of this immature LFA-1low subset of NK cells suggested that, although expressing NKp46 and NKG2D activatory receptors, it could not be still functional due to deficiencies in the cytotoxic machinery (18).

With the aim to investigate whether the characteristics of NK cells generated in vitro from CD34+ progenitors were suitable to counter cancer cells, we have evaluated the sequence through which developing NK cells acquire cytotoxic functions and susceptibility to apoptosis induced via TNF family members. A novel immature CD56bright NK subset, deficient in beta2-integrin and intragranular cytolytic proteins but already able to kill via TRAIL, has been identified. Finally, after 45 days of culture with IL-15 hematopoietic progenitor cells acquired phenotype, cytotoxic functions, and susceptibility to apoptosis similar to those of mature PB-activated CD56bright NK cells, indicating that a complete NK cytotoxic competence can be acquired in culture.


Reagents and Antibodies

MAbs for flow cytometry to CD3, CD11a, CD16, CD18, CD34, CD56, CD117, and granzyme-B were from Caltag Laboratories (Burlingame, CA), to CD94, CD161, CD158a,b (recognized by the HP-3E4 and CH-L clones), NKG2D, and Fas (CD95) were from BD Biosciences Pharmingen (San Jose, CA), to CD159a, NKp30, NKp44, and NKp46 were from Beckman Coulter (Fullerton, CA), to FasL (CD95L) and perforin were from Ancell Corporation (Bayport, MN), to TRAIL, TNF-Rs, and TRAIL-Rs were from Alexis Biochemicals (Gruenberg, Germany), to CD40L was from BenderMed Systems (Wien, Austria). Agonist anti-Fas IgM azide free antibody (CH11, Beckman Coulter, Fullerton, CA) was used for functional tests. TRAIL, TRAIL-Rs, and TNF-Rs staining was performed with a three-step procedure using first purified antibodies, then goat anti-mouse IgG-PE (Caltag Laboratories/Invitrogen) and, finally, directly conjugated mAbs. Information on monoclonal antibodies can be found in the Online Supplementary Table. Recombinant IL-15 and TNF-α were purchased from Peprotech EC (London, UK), recombinant FLT3-L, TRAIL, and IL-21 from R&D Systems (Minneapolis, MN). Fluorescent dyes DiOC18 (D-275) was from Invitrogen-Molecular Probes (Eugene, OR). Propidium iodide (PI) was from Sigma Chemicals (St Louis, MO). Actinomycin D was from USB Corporation (Cleveland, OH).

Cell Isolation and Culture

PB buffy coats were obtained from immobilized PB samples of healthy adults during the preparation of transfusion products. All samples were obtained, with informed consent, from donors at the Transfusion Centre of Urbino Hospital. PB mononuclear cells were isolated by density gradient centrifugation (Ficoll/Histopaque-1077, Sigma Chemical, St. Louis, MO) and allowed to adhere to plastic for 1 h at 37°C. Positive selection of CD34+ cells from mononuclear nonadherent cells was finally performed using the CD34 isolation kit and the Vario-MACS magnetic cell sorting program (Miltenyi Biotec, Auburn, CA). Purity was checked by flow cytometry on an aliquot of each selected population using anti-CD34-PE, (HPCA-2, BD Biosciences Pharmingen). Only samples with a purity >95% were used for in vitro NK cell differentiation. Purified CD34+ cells were cultured for 15–30 days in 1–2 mL ex-vivo 20 medium (BioWittaker, Walkersville, MA) supplemented with 10% human AB serum, FLT3-L (20 ng/mL) and IL-15 (20 ng/mL) with or without IL-21 (20 ng/mL). Cytokines were added at the beginning of the cultures and the medium was replaced every 4 days during culture period. To obtain higher percentages of CD56bright cells (>90%), FLT3-L administration was stopped after 15 days of culture.

CD56bright cells (> 95%), obtained after 30 days of primary cultures from CD34+ cells, were cultured in secondary culture for 15 days in 1–2 mL medium/well in the presence of IL-15, IL-21, or a combination of both (20 ng/mL).

In some experiments, after the positive selection of CD34+ cells, CD56+ NK cells were purified by negative selection from the eluted lymphocytes (purity > 95%) using the Vario-MACS and the NK isolation kit II (Miltenyi Biotec, Auburn, CA) and then cultured with IL-15 (20 ng/mL) for comparison with NK cells generated from CD34+ cells.

Flow Cytometric Surface and Intracellular Antigen Detection

Immunofluorescence analyses were performed with the indicated mAbs gating on the CD56 positive events within the lymphoblastic region of FSC vs. SSC plots (see Figure 1). Purified FITC-, PE-, biotin-, or PE-Cy5-conjugated antibodies were used. Quadrants were set based on background staining (<1–2%) with isotype control mAbs. Single stain controls with anti-CD45 were used to set instrument compensation.

To detect granzyme-B and perforin, cells were stained for surface NK markers and then treated with Fix/Perm cell permeabilization kit (Caltag Laboratories, Burlingame, CA) before intracellular protein detection.

Flow Cytometric Measurement of NK Cell-Mediated Cytotoxicity

NK cell-mediated cytotoxicity was tested against K562 (Fas/CD48) and Jurkat (Fas+/TRAIL-R2+/CD48+) cell lines (10, 11). Briefly, target cells were stained overnight with 5 μM DiOC18 (a green fluorescent probe, D-275, Molecular Probes). After washing, targets (5 × 104 cells) were incubated with effector cells for 2 h at different effector-target (E:T) ratios to detect granule release-mediated cytotoxicity. When indicated, 1 mM EGTA and 2 mM MgCl2 were added for 6-h to allow receptor-mediated Ca++-independent apoptosis (11). To detect both necrosis and apoptosis of target cells, a supravital PI staining was performed (19). Briefly, during the last 30 min of incubation, PI (50 μg/mL) was added to the samples, then cells were analyzed by flow cytometry without washing. This assay highlights differences in membrane permeability among living, apoptotic, and necrotic cells, staining both early apoptotic (PIdim) and (secondary) necrotic (PIbright) cells and allowing to discriminate them, as well as the annexin V/PI assay (19). The specific cell death percentage was calculated using the following formula: [percentage of DiOC18+/PI+ target cells after incubation with effector cells − percentage of DiOC18+/PI+ target cells without effector cells (spontaneous cell death)/100 – spontaneous cell death] × 100.

Flow Cytometric Evaluation of NK Apoptosis Mediated by TNF Family Members

To evaluate NK cell sensitivity to apoptosis, cells were pre-treated overnight with or without actinomycin D (100 ng/mL) and then incubated for 24 h at 37°C and 5% CO2 with TNF-α (200 ng/mL) or TRAIL (100 ng/mL) or agonistic anti-Fas antibody, CH11 (200 ng/mL). To detect apoptotic cell death induced by TNF family members, the previously described supravital PI staining was performed (19). Although this assay stains both early apoptosis (PIdim) and secondary necrosis (PIbright, 19), after the 24-h treatment, virtually all dead cells were already in secondary necrosis (not shown).

All samples for flow cytometry have been acquired on a FACScan or a FACScalibur flow cytofluorimeter (Becton-Dickinson BD Biosciences, San Jose, CA).

Statistical Analysis

During NK cell differentiation, percentages of expression of some NK antigens were calculated at different time points as mean values ± their standard deviations (mean ± SD).

To explore the existence of statistically significant differences, the two-tailed, two-sample Student t-test was used. P values <0.05 were considered statistically significant.


Expression of Cytotoxic Effector Molecules and TNF Family Members in CD56bright NK Cells Generated In Vitro from Purified CD34+ Hematopoietic Progenitors: Identification of an Immature CD56bright/granzyme-B/perforin/TRAIL+ NK Cell Stage

As already reported (9), after only 10–15 days of culture with FLT3-L and IL-15 plus IL-21, purified CD34+ hematopoietic progenitors differentiate into highly cytotoxic CD56dim/CD16+ NK cells that, as expected, express perforin, granzyme B, and LFA-1, but not CD117 (Fig. 1). As quickly as they appear, these cells die and disappear from cultures (not shown), indicating their predisposition to apoptosis. Differently from this subset, CD56bright NK cells, generated from CD34+ hematopoietic progenitors differentiated in the presence of FLT3-L and IL-15, appear later (usually after 20–30 days) and can be kept alive for at least a couple of months without a substantial appearance of apoptosis, suggesting their resistance to programmed cell death. For this reason, we have further investigated the functional characteristics of this second subset.

Figure 1.

Expression of cytotoxic effector molecules on CD56dim NK cells generated from CD34+ hematopoietic progenitors in the presence of FLT3-L and IL-15 with IL-21. Contour-plots of CD56+/CD3- NK cell population gated within the lymphoblastic region (R1). NK cells detected with FITC or PE-conjugated anti-CD56 and PE-Cy5-conjugated anti-CD3 mAbs were analyzed for intragranular lytic proteins (FITC-conjugated anti-perforin and PE-conjugated anti-granzyme-B mAb), LFA-1 (anti-CD11a-PE), and CD117 (anti-c-kit-PE) expression at day-20 of culture.

Developing CD56bright NK cells (typically characterized by CD117 antigen) express activatory molecules such as NCRs and NKG2D (but not CD16) before inhibitory CD94-NKG2A and KIR receptors (7, 18, 20, and not shown). Nevertheless the cytotoxic activity of in vitro developing CD56bright NK cells is usually low (9, 10, 16 and not shown). An inhibitory function of CD244, after the binding with its ligand (CD48), has been claimed to be responsible for their low cytotoxicity (10). However, this low cytotoxic activity appears to be even lower against CD48 K562 than against CD48+ Jurkat cell lines (16), suggesting that mechanisms other than CD48-CD244 binding, might account for the reduced cytolytic activity of immature CD56bright NK cells. For this reason, we evaluated the expression of NK cytotoxic effector molecules: granzyme-B, perforin, and TNF ligand members. At the 20th day of culture, a significative percentage of CD56bright/NCR+ NK cells co-expressed intragranular granzyme-B and perforin molecules (Fig. 2A) and their proportion progressively increased during the period of culture (Fig. 2B). Differently from intragranular proteins, TNF-related surface ligands (TRAIL, FasL, and CD40L) were hardly detectable on the surface of CD56bright NK cells during the first 20 days of culture (Fig. 2A and not shown). However, TRAIL was significatively expressed (although at low density) on the surface of most of CD56bright NK cells at day 30 of culture, while FasL and CD40L still remained undetectable (Fig. 2B and not shown). We have recently demonstrated that LFA-1 (detected equally by anti-CD11a and anti-CD18, Ref.18) as well as CD94 molecule (17) is a useful marker in the discrimination of immature NK cell stages, for this reason, we compared CD56bright/granzyme-B/perforin population with CD56bright/LFA-1 and CD56bright/CD94 ones. The simultaneous evaluation of surface and intracellular protein expression (Fig. 2B) revealed that, after 30 days of culture, the majority of the CD56bright NK cells co-expressed granzyme-B with LFA-1 and CD94 (whose distribution completely overlapped with CD159a antigen, Fig. 2B). Nevertheless, a minor proportion of LFA-1 negative cells expressing granzyme-B was usually present (Fig. 2B), suggesting that, during differentiation, intragranular cytotoxic molecules might precede LFA-1 expression. At this time point, within the CD56 positive NK cells, the percentages of CD56bright/granzyme-B/perforin cells (22 ± 14%, mean ± S.D., n = 8) were slightly but significantly lower than those of CD56bright/LFA-1 ones (34 ± 13%, mean ± S.D., n = 8, P < 0.01). On the contrary, TRAIL was similarly expressed at low density on both Stage 3 CD56bright/LFA-1 and Stage 4 CD56bright/LFA-1+ NK cells (Fig. 2B), indicating that TRAIL is a cytotoxic effector molecule expressible on the surface of immature NK cell stages (16). On the other hand, NK cells expressed Fas, TNF-R2, and, TNF-R1 at low density of expression, while TRAIL-R4 was present on a small subset and TRAIL-R1, -R2, and -R3 were undetectable (Fig. 2C and not shown).

Figure 2.

Expression of cytotoxic effector molecules: identification of an immature CD56bright/TRAIL+/granzyme-B/perforin NK cell stage. Contour-plots and histograms of CD56+/CD3- NK cell population generated from CD34+ hematopoietic progenitors in the presence of FLT3-L and IL-15. NK cells gated within the lymphoblastic region and detected with PE-Cy5 or FITC-conjugated anti-CD56 mAb were analyzed for intragranular lytic proteins (FITC-conjugated anti-perforin and PE-conjugated anti-granzyme-B mAb) and TRAIL (PE stained) expression at day-20 (A) and day-30 (B) of primary culture. (B) Correlate measurements of anti-perforin, anti-granzyme-B, anti-CD94, anti-CD159a, anti-TRAIL, or anti-LFA-1 (CD18) mAbs are also displayed. (C) Expression of TNF-receptor members at day-30 of primary cultures is shown. Contour-plot or histograms of NK cells detected with PE-Cy5-conjugated anti-CD56 are shown. Solid lines represent PE staining of the indicated surface antigens (shaded histograms are isotype controls). Percentage of positive cells is indicated. Experiment representative of at least five performed with similar results.

IL-15-Induced Terminal Differentiation of CD56bright NK Cells was Inhibited by IL-21

To induce further NK differentiation and the generation of highly cytotoxic CD56dim NK cells from CD117+/CD56bright ones, NK cells of a day-30 primary culture (containing both immature Stage 3 and mature Stage 4 CD56bright cells) were cultured for additional 15 days (secondary cultures) in the presence of IL-15 or IL-21 or a combination of both. As already described for mature mouse NK cells (21), IL-21 alone was unable to support NK cell survival and proliferation, and, when used in combination, it inhibited those induced by IL-15 (2.5 ± 0.2 vs. 3.7 ± 0.2, n = 5 mean fold expansion, P < 0,01). In accordance with other reports (14, 17, 22, 23), IL-15 induced further differentiation of CD117+/CD56bright NK cells, which reached a phenotype (Fig. 3A) and a cytotoxic function (Fig. 3B) similar to those of mature PB NK cells cultured in the presence of IL-15 (Fig. 3B and not shown). In particular, it increased the percentage of CD56bright cells expressing LFA-1 (87 ± 10%, mean ± S.D., n = 5), CD94-CD159a, granzyme-B, and perforin (Fig. 3A), as well as KIR and CD16 antigens (Fig. 3C), which were usually expressed at low to undetectable percentages on NK cells after 30 days of primary cultures (18 and not shown). In accordance with this phenotypic observation, in vitro generated CD56bright NK cells mediated a high cytotoxic activity against Jurkat and K562 cells (Fig. 3B). Interestingly, the presence of IL-21 did not induce the differentiation of CD56dim/CD16+ NK subset from CD56bright NK cells and, rather, it inhibited their differentiation since it induced a lower percentages of LFA-1+ (73 ± 17%, mean ± S.D., n = 5, P < 0,05), as well as CD16+ and KIR+ NK cells (Fig. 3C). Notably, CD16, CD158a, and CD158b molecules were confined to CD56bright/LFA-1+ NK cells (Fig. 3C), further confirming the immaturity of CD56bright/LFA-1 NK cells.

Figure 3.

The differentiation of CD56bright NK cells induced by secondary culture with IL-15 was inhibited by IL-21. (A) Immunophenotype of CD56bright NK cells after 15 days of secondary culture with IL-15. Correlate measurements of CD56 (PE-Cy5-conjugated, x axis) or LFA-1 (CD18 FITC-conjugated) and indicated markers (PE-conjugated, y axis) are displayed as contour-plots. Percentage of positive cells in each quadrant are indicated. (B) Cytotoxic activity against Jurkat (□, •) and K562 (▪, ○) target cells of CD56bright NK cells generated from CD34+ cells after a 15-day IL-15 secondary culture (□, ▪) and of PB NK cells after a 45-day IL-15 culture (•, ○ with a predominant, >70%, CD56bright phenotype) obtained from the same donor. Data shown are percentages of target cell death at different effector-target ratio in a 2-h assay. (C) Immunophenotype of CD56bright NK cells after 15 days of secondary culture with IL-15 in the presence (right panels) or absence (left panels) of IL-21. Correlate measurements of LFA-1 (CD18 FITC-conjugated, x axis) and indicated markers (PE-conjugated, y axis) are displayed as contour-plots. A representative experiment of at least four performed with similar results.

CD56bright NK Cells After Secondary Culture with IL-15: Upregulation of TNF Family Members and Relative Resistance to Apoptosis

Expression TNF family molecules have been evaluated to monitor the cytotoxic function and the potential apoptotic susceptibility of differentiating CD56bright cells.

After secondary cultures with IL-15, CD56bright NK cells increased their levels of TRAIL and FasL expression (Fig. 4A). However, surface FasL and CD40L molecules were usually low or hardly detectable (Fig. 4A), likely because they are preferentially secreted (14). As expected, CD56bright NK cells mediated a strong Ca++-independent cytotoxic activity against FasL- and TRAIL-sensitive Jurkat cells (in a 6 h cytotoxic assay with EGTA), but not against K562 cells (Fig. 4B), whose killing is completely Ca++-dependent (16).

Figure 4.

Expression and function of TNF ligand and receptor members on CD56bright NK cells after secondary culture with IL-15. (A) Flow cytometric analyses of CD56bright NK cells after secondary cultures. Histograms of NK cells detected with PE-Cy5-conjugated anti-CD56 mAb are shown. Solid lines represent PE staining of the indicated surface antigens (shaded histograms are isotype controls). (B) Ca++-independent cytotoxic activity against Jurkat (□, •) and K562 (▪, ○) target cells of CD56bright NK cells generated from CD34+ cells after a 15-day IL-15 secondary culture (□, ▪) and of PB NK cells after a 45-day IL-15 culture (•, ○) obtained from the same donor. Data shown are percentages of target cell death at different effector-target ratio in a 6-h assay with EGTA. Each experiment is representative of at least five performed with similar results. (C) Flow cytometric analyses of CD56bright NK cells after secondary cultures. Solid lines represent PE staining of the indicated surface antigens (shaded histograms are isotype controls). (D) Percentages of NK cell apoptosis after 18-h pre-treatment with (white bars) or without (gray bars) actinomycin D (100 ng/mL) and then for 24-h incubation with the indicated reagents. Data shown are means ± SD of three separate experiments. *P < 0.05, **P < 0.01 compared with controls.

On the other hand, IL-15 upregulated TRAIL-R2 and TRAIL-R4, but not TRAIL-R1 and TRAIL-R3, while Fas, TNF-R1, and TNF-R2 molecule expression remained constant on CD56bright NK cells (Fig. 4C and not shown). Since TNF receptor members are able to drive activated NK cell apoptosis, the functional activity of the TNF-related death receptors expressed on CD56bright NK cells was tested using a supravital PI staining assay (19). Similar to IL-15-activated mature NK cells (3), in vitro developing CD56bright NK cells are relatively resistant to apoptosis induced via TNF receptor members. In fact, a minority (<30%) of them are sensitive to recombinant TRAIL and anti-Fas agonistic antibody. Only a pre-treatment with actinomycin D (Fig. 4D), known to inhibit the apoptosis-protection mediated by c-FLIP (3), predisposes the majority (about 80%) of CD56bright NK cells to TRAIL-induced apoptosis.


To investigate the possibility to develop NK cells suitable to counter cancer cells in oncologic patients, we have evaluated the cytolytic function and susceptibility to apoptosis of human NK cells differentiated in vitro from CD34+ progenitors using cytokine combinations of FLT3-L plus IL-15 with IL-21 (known to generate the CD56dim subset) or without it (known to generate the CD56bright subset) (7–9, 18). First of all, it is important to underline that, upon cytokine stimulation, most of the NK cell markers used to distinguish CD56bright from CD56dim NK cells are either upregulated (CD56, CD16, KIR, CD25, granzymes, perforin, and cytotoxic function) or downregulated (CD62L, CCR7, CD127) (13, 14, 18, 23, 24, and not shown), rendering somehow difficult their reciprocal discrimination. To this regard, in our in vitro experiments, the use of a mAb conjugated with phycoerythrin (typically the brightest fluorochrome for flow cytometry applications) to CD117 (exclusive of CD56bright subset, 5) proved to be the best way to detect CD56bright NK cells and to distinguish them not only from CD56dim/CD117 NK cells but also from CD56+/LFA-1+/Granzyme monocytes (usually generated in the first 2 weeks of these cultures). Interestingly, while within 20–30 days of IL-21 containing cultures, the highly cytotoxic CD56dim subset died, cultures only with FLT3-L plus IL-15 generated CD117+/CD56bright NK cells that still showed several features of immaturity typical of Stage 3 NK cells (12, 17, 18). In particular, we observed a subset deficient in β2-integrins, CD94-CD159a, granzyme-B and perforin expression, suggesting that this could represent a stage still unable to perform the cytolytic activity. During normal in vivo differentiation, developing NK cells acquire activating and inhibitory receptors and cytotoxic functions in a fashion that prevents NK-mediated auto-aggression against normal cells. One possible explanation to this event is that the expression of functional inhibitory receptors, such as MHC-I receptors or 2B4 molecule on immature NK cells, would precede that of activatory ones (10). Nevertheless, our as well as other reports (7, 10, 17, 18) have already demonstrated both in vitro and in vivo that the expression of activatory molecules, such as NCRs and NKG2D, precede that of MHC-I inhibitory receptors and the inhibitory function of 2B4 on immature NK cells would not explain how CD48 (i.e., non-hematopoietic) autologous cells can be spared by these immature NK cells. NK self-tolerance might be guaranteed by the lack of ligands for NK activating receptors. However, immature myeloid cells, present invivo in the BM (site of NK and erythro-myeloid development) as well as in vitro during NK cell differentiation (10, 11), have shown to express NCRs ligands (25), thus the absence of these ligands could not be guaranteed. For this reason, we propose that, in addition to the inhibitory function of 2B4 molecule (10), immature NK cells would be non-lytic because of the deficiency of β2-integrin adhesion and intragranular cytolytic proteins, thus assuring a fine control of the fail-safe mechanism against all normal self cells. Moreover, MHC-I inhibitory receptors would be expressed as early as NK cells, by upregulating molecules of the cytolytic machinery, become potentially cytotoxic. The reason of an early expression of activatory molecules is not clear, anyway, some evidences support the idea that stimulatory signals are necessary to induce the expression of inhibitory receptors on developing NK cells (reviewed in Ref.15), finally leading to a functionally complete maturation. In our experiments, fully competent and highly cytotoxic NK cells, expressing MHC-I inhibitory receptors, have been obtained after a total of 45 days of IL-15 culture. However, this strong cytolytic function may also depend on a further process of activation induced by the cytokine itself, as suggested by the upregulation of TNF family ligands and receptors in the in vitro developing NK cells. Resembling the intracytoplasmatic TNF-α production (26), the membrane-bound form of TRAIL was observed on both immature (Stage 3) CD56bright/LFA-1 and more mature (Stage 4) CD56bright/LFA-1+ NK cells (18), confirming that TRAIL is a cytotoxic effector molecule expressible also by immature NK cell stages (16). Since immature NK cells are thought to be generated in BM, we had previously hypothesized and demonstrated TRAIL expression in that site (27). Of note, immature erythro-myeloid cells, generated in the BM as well, showed to be sensitive to recombinant TRAIL (27, 28). On the contrary, the highly cytotoxic LAK cells were resistant to apoptosis induced by TRAIL and became significantly susceptible to it only after the down-modulation of the cytokine-dependent c-FLIP (3). Therefore, depending on the cytokine microenvironmental milieu conditions, activated NK cells could develop resistance or sensitivity to negative feedback mechanisms induced by the TNF ligand family members, finally conditioning the length of NK cells response after their activation.

The expression not only of TNF members but also of MHC-I inhibitory receptors, adhesion and cytotoxic effector molecules as well as CD16 and KIR antigens increases with time suggesting that in vitro CD56bright NK cells continue to differentiate/activate from day 30 to day 45 of culture. CD56bright/CD16/KIR NK cells have been suggested to differentiate into either CD56dim/CD16+/KIR+ or into CD56bright/CD16+/KIR+ NK cells both in vitro after cytokine stimulation (14, 24, 29, 30) and in vivo in the humanized immune system (HIS) mice model (31) and after HLA-matched hematopoietic stem cell transplantation (32). Similarly, during our secondary culture (from day 30 to 45 of culture), some NK cells begin to express CD16 and KIR. However, the maintenance of CD117 molecule (even though at a lower density than Stage 3, 18) suggests that they would represent fully differentiated/activated CD56bright NK cells rather than CD56dim cells generated from CD56bright ones. Notably, an unusual CD56bright/CD117+/CD16low NK subset, overexpressing intracellular perforin and activating NKG2D and NKp46 receptors, emerges 3 months after hematopoietic stem cell transplantation and its appearance positively correlates with the increased level of IL-15 (32). Moreover, alike to our secondary cultures, Takahashi et al. described the in vitro development of potent effector CD56bright/CD16+ NK cells from CD56bright/CD16 NK cells upon cytokine administration (14).

Although the engagement of CD56 molecule with the FGF-R1 on fibroblasts has been described to induce the direct differentiation of CD56bright to CD56dim cells (33), it is still unclear which growth factors are involved and how they might produce this differentiation without a significant proliferation (29, 33), considering that these two subsets have significantly different telomerase length (30, 33). Moreover, it has been described that human CD56bright NK cells and their mouse counterpart, differently from CD56dim ones, are characterized by the expression of GATA-3 and CD127 (34), suggesting the presence of at least two independent pathways of NK cell differentiation: thymus/lymph node (CD56bright) and bone marrow/spleen (CD56dim). Consistent with this later hypothesis there are evidences of a substantially different activation pathways (CD56bright by cytokines, CD56dim by targets) and thus a different response not only to IL-2 and IL-15 but also to TGF-beta and IL-21 (35–37). In line with this, our data from cultures with IL-21 show a rapid development of highly cytotoxic CD16+/CD56dim/LFA-1+/CD117 NK cells from CD34+ hematopoietic progenitors (8, 9 and Fig. 1) but not from immature (Stage 3) CD56bright NK cells. This cytokine has rather an inhibitory effect on the development of CD56bright NK cells and it actually blocks their differentiation from Stage 3 to Stage 4, thus supporting a model where CD56bright and CD56dim NK cells would represent functionally distinct subsets of mature human NK cells that might diverge earlier than NK cell Stage 3 (12). Alternatively, it could be hypothesized that, starting from CD34+ cultures, IL-21 might induce the generation of FGF-R1 expressing cells able to induce the further differentiation of developing CD56bright cells into CD56dim ones. Standing to this latter hypothesis, however, it is still hard to explain how in IL-21 culture system, the generation of CD56dim NK cells is so short (15–20 days), considering that the generation of CD56bright cells (supposed to mature into CD56dim) takes 20–30 days without IL-21 (7, 9, 18) and, further, that this cytokine inhibits their maturation. Of interest, the CD56dim NK subset, but not the CD56bright one, is significantly decreased in the PB of patients affected by HIV-1, and this positively correlates with the number CD4 T cells. A possible explanation is that the development of CD56dim NK cells would be dependent on IL-21 (9), which is produced by CD4 T lymphocytes (38). In line with this hypothesis, there is the observation that after stem cell transplantation the recovery of CD56dim subset is markedly delayed and resembles the slow reconstitution of CD4+ T cells (39–41). These observations are suggestive of an influence of T cells, at least in the human species, in orienting NK cell progenitors toward CD56dim or CD56bright development.

A more detailed knowledge of human NK cell development can pave the way for designing new NK-based immunotherapies able to focus on the most suitable NK subset to counter tumor cells and other diseases. Of interest for cancer therapy, our data indicate that the majority (>70%) of these in vitro generated NK cells are resistant to apoptosis induced by TNF family members and that a complete cytotoxic competence of CD56bright NK cells, exploitable against tumor cells, can be acquired in culture starting from hematopoietic progenitor cells.


The authors thank Drs. Claudia Masoni, Federico Bastianelli, and Massimo Della Felice for their technical assistance and Dr. Mario D'Atri for critically reading the article.