Deficient expression of NCR in NK cells from acute myeloid leukemia: evolution during leukemia treatment and impact of leukemia cells in NCR^dull phenotype induction

Natural killer (NK) cells function as sentinels for the recognition of virus infection and cellular transformation by monitoring the expression of class I human leukocyte antigen (HLA) molecules at the cell surface.¹  Activation of NK cells is the result of a balance between inhibitory and activating signals.²  Inhibitory signals are provided by interactions with HLA class I molecules. Under physiologic conditions, normal cells are protected from NK-cell lysis because they express a sufficient amount of HLA class I molecules to inhibit NK cytotoxicity.³  In the absence of these inhibitory signals, activating receptors, if engaged by ligands on the target-cell surface, activate NK cytotoxicity. Many activating receptors have been described.⁴  NKG2D is an important activating receptor expressed by NK cells but also by virtually all cytolytic T lymphocytes.⁵  NKG2D plays a major role in NK-cell–mediated cytotoxicity against certain target cells. Ligands for NKG2D include the major histocompatibility complex class I chain-related (MIC)–A and –B stress-inducible molecules and the UL16-binding protein (ULBP) major histocompatibility complex class I–related molecules. FcγRIII receptor CD16 is responsible for the antibody-dependent cell cytotoxicity (ADCC) mediated by NK cells, but it also mediates cytotoxicity by yet unidentified ligands.⁶  Another important activating receptor expressed by NK cells is 2B4/CD244.⁷  This receptor belongs to the CD2 superfamily and is also expressed by all cytolytic lymphocytes.⁸  This receptor has an important role in the fight against Epstein-Barr virus (EBV) infection because its ligand, CD48, is up-regulated in EBV-infected cells and thus provides a coactivation signal for NK cells.^9,10  Finally, among activating receptors, the natural cytotoxicity receptors (NCRs) NKp30, NKp44, and NKp46 are NK-cell restricted.¹¹  Their engagement by still unidentified cellular ligands, or virus-derived components,^12,13  results in increased cytotoxicity and cytokine release by NK cells.¹⁴  NCR surface density, as measured by the brightness of immunofluorescence, varies. Thus, NK cells from most donors express a high density of NCR (NCR^bright), whereas in less than 20% of donors, NK cells display reduced quantities of NCR (NCR^dull). Importantly, a strict correlation exists between NCR density and NK-mediated cytolytic activity.^15,16 

We focused our attention on NK cells in patients with acute myeloid leukemia (AML) because several clues supported the potent antileukemia role of NK cells. Rapid disease progression and the high incidence of relapse after treatment with high-dose chemotherapy or after transplantation of allogeneic hematopoietic stem cells suggested that leukemic blasts can escape recognition by the immune system. It has been shown in patients with AML that impaired NK-cell function and cytokine production are associated with early relapse.¹⁷  Moreover, it as been observed that leukemia cells can express membrane-bound heat shock protein 70 (HSP70), which correlates with sensitivity to lysis by autologous NK cells.¹⁸  In addition, NK cells adhere to bone marrow fibroblasts and compete with myeloid leukemia cells for binding to the microenvironment, suggesting that they may inhibit the growth of leukemia cells.¹⁹  Finally, an important clue arguing for the major role of NK cells in AML clearance comes from allogeneic stem cell transplantation. When patients and donors were mismatched for HLA-specific inhibitory NK receptors (killer-cell–inhibitory receptors [KIRs]), the risk for relapse was virtually null compared with the 75% rate observed when perfect KIR matching was observed between donors and recipients.^20,21  Given that NK cells may have a potent antileukemia effect,²²  why do patients experience relapse and thus escape the innate immunity-mediated antitumor response? In agreement with this escape, it has recently been shown that low levels of NCR ligands were expressed by leukemia cells. In addition, the ligands might have been expressed by normal myeloid cells, thus disturbing NK-cell function in AML.²³  Another hypothesis is poor NCR expression by NK cells of AML patients (the so-called NCR^dull phenotype), a phenotypic abnormality observed in most (more than 80%) in vitro–expanded NK cells of leukemia patients.²⁴  This deficient NCR expression leads to poor antitumor cytotoxicity capacity because of the impossibility of NCR ligand engagement, though the cytotoxicity armamentarium of AML-NK cells is preserved as demonstrated by potent CD16-driven redirected killing.²⁴  Moreover, we have recently shown that this phenotypic defect is also correlated with poor killing of immature dendritic cells (iDCs), which could lead to iDC-mediated tolerance in AML-T cells.²⁵  Because phenotypic and functional analyses were performed after IL-2–driven expansion of AML-NK cells, we could not exclude in vitro–induced proliferation of a minor NCR^dull NK-cell subset because, in certain healthy donors, NCR^dull and NCR^bright subsets can coexist.¹⁵ 

In the present study, we first studied the AML-NK phenotype at diagnosis using 4-color flow cytometry analysis directly on total fresh uncultured peripheral-blood mononuclear cells (PBMCs) from 71 patients with AML. We then tested the in vivo reversibility of the NCR^dull phenotype and its dependence on leukemia cells by assessing the AML-NK phenotype after complete remission (CR) of leukemia and compared it with that of patients who did not attain CR or who experienced disease relapse. Then, to decipher mechanisms involved in that process, we verified whether NCR down-regulation could be obtained in culture. Finally, we compared NCR expression and clinical behavior of patients with leukemia.

Peripheral-blood samples were obtained from patients or healthy blood donors after informed consent was obtained in accordance with the Declaration of Helsinki. Patient samples were obtained before or after specific antileukemia therapy and were part of diagnostic procedures. Umbilical cord blood (UCB) samples from full-term newborns were collected at the Department of Gynaecology, Hôpital de la Conception (Marseille, France) after informed consent of the mothers. The entire research procedure was approved by the local ethics committee (Institut Paoli-Calmettes Marseille, France). To increase the statistical potency of our present analysis, data from 14 additional patients from our previous study²⁴  were analyzed regarding survival curves.

Cell surface analysis was performed through flow cytometry with the use of a FACScalibur cytometer and CellQuest software (Becton Dickinson, Mountain View, CA). NK cells were immunostained with fluorescein isothiocyanate (FITC)–conjugated anti-CD3, phycocyanin 5 (PC5)–conjugated anti-CD56, and allophycocyanin (APC)–conjugated anti-CD45 (to exclude myeloid blasts) monoclonal antibodies (mAbs) (Immunotech, Marseille, France). Indeed, blasts expressed lower levels of CD45 compared with nonleukemia cells. Thus, this marker is used for gating normal cells.²⁶  Triggering receptor expression (NKp44, NKp30, NKp46, CD16, 2B4/CD244, NKG2D) was measured with phycoerythrin (PE)–conjugated mAb (kind gift of Beckman-Coulter Immunotech, Marseille, France).

Freshly analyzed AML-NK cells were classified as NCR^bright or NCR^dull, depending on NKp30 and NKp46 MFI levels. Cut-off values for NCR^dull and NCR^bright were chosen on the basis of previously published studies (Costello et al,²⁴  Sivori et al¹⁵ ), in accordance with the subtypes described by Alessandro Moretta (Sivori et al¹⁵ ). Indeed, 80% of healthy donors have an NK-expressing NCR^bright phenotype.¹⁵  In our 23 samples from healthy donors, cut-off points to determine NCR^bright phenotypes were chosen with respect to this percentage. When NKp30 and NKp46 were expressed at low levels (MFI, respectively, less than 30 and less than 50), patients were classified with the NCR^dull phenotype. When NKp30 and NKp46 were both expressed at high levels (MFI, respectively, greater than 30 and greater than 50), patients were classified with the NCR^bright phenotype. Patients with the NKp30^bright/NKp46^dull phenotype or the NKp30^dull/NKp46^bright phenotype were further classified as NCR^discordant.

In all experiments, mature NK cells were derived from the peripheral blood of patients or healthy donors. NK cells were isolated by immunomagnetic negative selection according to the manufacturer's instructions (Miltenyi Biotec, Paris, France). Then NK cells were cultured for 2 to 3 weeks before use or were briefly cultured (1-5 days) with leukemia cells. To amplify NK cells, they were first isolated and then cultured for 3 weeks with 1000 IU/mL interleukin-2 (Proleukin; Chiron, Emeryville, CA), 1.5 ng/mL phytohemagglutinin A (Invitrogen, Cergy-Pontoise, France), and irradiated allogeneic normal PBMCs used as feeder cells in RPMI (Cambrex, Emerainville, France) and 10% fetal calf serum (FCS; Biowest, Paris, France). Then NK cells were tested for cytolytic activity or were harvested and cultured with leukemia cells for another 5 or 10 days.

Experiments were performed using a protocol modified from Sivori et al²⁷  and Carayol et al.²⁸  UCB mononuclear cells were obtained by Ficoll-Hypaque density gradient centrifugation. CD34⁺ cells were separated from mononuclear cells by use of the MACS system (Miltenyi Biotec, Auburn, CA). Cells obtained were 98% or more pure. CD34⁺ cord-blood precursors (2 × 10⁴ to 6 × 10⁴ cells) were incubated in 24-well plates (Falcon; Becton Dickinson) in 2 mL α-MEM (Cambrex) containing 10% human serum (ABCYS, Paris, France) and 5% FCS (Biowest). Human recombinant (hr) SCF (hrSCF; 20 ng/mL; ABCYS) and hrIL-15 (20 ng/mL; ABCYS) were added in all cultures. In all experiments, cells were also seeded on a confluent layer of murine MS-5 stromal cells. MS-5 cells were expanded in RPMI (Seromed, Biochrom KG, Berlin) with 10% FCS (Biowest) and were plated in 24-well plates at a concentration of 6 × 10⁴ cells/mL 24 hours before the initiation of cocultures. To test the effect of leukemia cells on differentiating NK cells, we added, on the first day of culture, 200 000 to 600 000 irradiated healthy donor PBMCs (control) or irradiated peripheral-blood cells (containing greater than 98% blasts) from patients with NCR^dull or NCR^bright AML. All cultures were maintained for 2 to 3 weeks, and half the medium was renewed twice a week.

Cytotoxic assay was performed with the use of a 4-hour chromium (Cr) 51 release assay. Effector cells were long-term cultured NK cells derived from PBMCs of patients at diagnosis or CR and from healthy donors. Target cells used to investigate the functional recovery of CR-AML-NK cells were the murine mastocytoma FcγRII+ P815 cell line (ATCC; LGC Promochem, Molsheim, France). Cytotoxicity was measured after redirected killing of P815 by NK cells using appropriate mAbs. Concentrations of the various mAbs were 5 μg/mL (KD1 anti–CD16 IgG1 mAb, BAB281 anti-NKp46, Z231 anti-NKp30; from the laboratory of Alessandro Moretta, Genova, Italy). All experiments were performed in triplicate in at least 3 independent experiments.

Resting or long-term activated NK cells were cultured in 96-well plates at 50 000 cells/mL, with 50 IU/mL rhIL-2 for 1 to 5 days for resting NK cells or 5 to 10 days for long-term activated NK cells (1000 IU/mL rhIL-2), in the presence of leukemia cells or healthy donors PBMCs (control) at 1 × 10⁶ cells/mL. NK cells were then analyzed for cell-surface–receptor expression by flow cytometry excluding blasts by CD45^high gating. We used low concentrations of rhIL-2 (50 IU/mL) to obtain resting NK cells (no up-regulation of cell-activation markers CD69 and NKp44), whereas high concentrations of rhIL-2 (1000 IU/mL) was used to obtain activated NK cells.

Survival analysis was performed with the use of SPSS software (SPSS, Chicago, IL). To analyze the repartition of age and sex, t tests were performed. To analysis survival we used the Kaplan-Meier method with the log-rank count, and for multivariate analysis we used the Cox model. For more homogeneous sample analyses, data from patients with biphenotypic (n = 2) and AML minimally differentiated (AML0) (n = 2) leukemia were not included. Eighty-one patients were included for survival studies. Results were considered significant when P values were less than or equal to .05.

AML-NK cells were analyzed at diagnosis and without previous in vitro amplification in 71 consecutive patients (40 males, 31 females; Table 1)Most (66 of 71) patients had AML, 2 had biphenotypic leukemia, and 3 had refractory anemia with excess blasts (RAEB). The median age was 61 years (range, 17-85 years). We chose to include in our study only patients with white blood cell counts lower than 50 × 10⁹/L (mean, 10.5 ± 22 × 10⁹/L) to facilitate direct analysis of NK cells, which, in all cases, represented a rare population. The absolute number of AML-NK cells was 0.248 × 10⁹/L ± 0.277 and, thus, was similar to values found in healthy donors under the same experimental conditions (ie, 0.264 × 10⁹/L ± 0.183). Seven of 71 (10%) patients had favorable karyotype abnormalities—3 with inv(16), 2 with t(8;21), and 2 with t(15;17).

We first analyzed, by flow cytometry, NCR expression on AML-NK cells and compared findings with those of 23 healthy control donors (Figure 1A). Because NKp44 expression is only observed after in vivo or in vitro stimulation, we failed to detect this molecule in healthy donors or unstimulated NK cells in patients with AML. Most (more than 70%) healthy donors had NK cells with an NCR^bright phenotype (NKp30 mean MFI, 72 ± 30; NKp46 mean MFI, 102 ± 56; sample of healthy donors in Figure 1A top row). One third of healthy donors had NK cells with the NCR^dull phenotype (NKp30 mean MFI, 20 ± 6; NKp46 mean MFI, 30 ± 10; data not shown). No healthy donor had NK cells displaying an NCR^discordant phenotype. In contrast, we observed that 47 (66.2%) of 71 patients with AML had NK cells with an NCR^dull phenotype (NKp30 mean MFI, 12 ± 8; NKp46 mean MFI, 23 ± 15; sample of AML-NK donor in Figure 1A, bottom row). Only 13 (18.3%) of 71 patients with AML had NK cells with a normal NCR^bright phenotype (NKp30 mean MFI, 41 ± 8; NKp46 mean MFI, 88 ± 31; sample of healthy donors in Figure 1A, middle row). We observed that 11 (15.5%) patients had an NCR^discordant phenotype; among them, 7 (10%) patients were NKp30^dull/NKp46^bright (NKp30 mean MFI, 23 ± 5; NKp46 mean MFI, 72 ± 15; data not shown), and 4 (5.5%) patients were NK30^bright/NKp46^dull (NKp30 mean MFI, 34 ± 7; NKp46 mean MFI, 46 ± 5; data not shown). In all cases, no difference in CD16 or CD56 expression was observed between normal and AML NK cells (Figure 1A, right column and data not shown, respectively). As shown in Figure 1B, NCR expression was decreased for NKp30 and NKp46 but was not associated with FAB classification subtypes. Patients with AML0 were NCR^bright, but this observation has no statistical significance because of the low number of patients with this subtype (n = 2; Figure 1B). In addition, no difference was observed in NCR expression between CD56^bright and CD56^dull NK subpopulations.

We performed a kinetic study comparing NCR expression and functions on AML-NK cells at diagnosis and after treatment, focusing first on patients who achieved CR. Mean expression of NKp46 was 43 ± 32 at diagnosis and reached 75 ± 50 after CR, corresponding to a statistically significant (P < .05) 1.75-fold increase (Figure 2A). NKp46 expression was further checked at different times after CR and remained stable at T2 (MFI, 88 ± 50) and T3 (MFI, 86 ± 37). Mean expression of NKp30 was 21 ± 14 at diagnosis and reached 35 ± 24 after CR, corresponding to a statistically significant (P < .05) 1.67-fold increase (Figure 2B). Expression of NKp30 was further checked at different times after CR and remained stable at T2 (MFI, 41 ± 26), with a moderate and nonstatistically significant decrease at T3 (MFI, 31 ± 15). As control, we verified that CD16 expression was comparable to that of healthy donor NK cells and did not change expression before (MFI, 242 ± 141) or after CR was achieved (MFI, 218 ± 133; data not shown). We then tested the putative cytolytic functions of AML-NK cells at diagnosis and after achievement of CR. AML-NK cells were tested in redirected killing with anti-CD16, anti-NKp30, and anti-NKp46 mAb (Figure 3). NCR^dull AML-NK cells derived from patients at diagnosis tested in redirected killing with anti-CD16 mAb displayed a cytolytic level comparable to that of healthy donor NCR^bright NK cells (43% ± 2% vs 60%  ± 1%). In contrast, anti-NKp30 mAbs and NKp46 mAbs failed to induce an increased cytotoxicity in NCR^dull AML-NK cells compared with nonredirected killing (20% ± 2% and 25% ± 4%, respectively, vs 15% ± 1%). As control, NKp30- and NKp46-redirected killing in healthy donor NK cells was as efficient as CD16-redirected killing (57% ± 2% and 58% ± 1%, respectively, vs 60% ± 1%). The CD16-driven redirected killing of AML-NK cells derived from patients in CR induced cytotoxicity that was comparable to that of healthy donor NK cells (44% ± 2%). AML-NK cells from patients in CR had restored NCR-driven cytotoxicity because NKp30- and NKp46-redirected killing induced a cytotoxicity that was comparable to that of control-redirected killing with anti-CD16 (42% ± 1% and 44% ± 1% respectively, vs 44% ± 2%). Restored NKp46 expression was also followed by restoration of NKp46-redirected killing (data not shown).

We then performed the same study for patients who failed to attain CR (Figure 4A) or who experienced disease relapse (Figure 4B). For 3 of the 6 patients who did not attain CR (Figure 4A, upper row), we did not observe the restoration of NKp30 or NKp46. The observed pattern was more complex for other patients. For unique patient number 87 (UPN87), we observed that NKp46 increased while NKp30 remained at a “dull” level. Patient UPN90 experienced NKp46 restoration to a normal level (represented by the plain line) whereas NKp30 remained lower than at the mean expression level observed for healthy donor NK cells (dashed line). For UPN25, we observed the complete restoration of NKp46 and the partial restoration of NKp30. In patients who experienced relapse (Figure 4B), we observed a homogeneous pattern regarding NKp46—a drastic decrease in NKp46 expression from CR to relapse (left panel). The same trend was observed regarding NKp30 (right panel). Moreover, for one patient in whom NCR expression was not restored after CR, no modification of NCR expression was observed after relapse (data not shown).

We sought to determine whether the NCR^dull phenotype could be induced in vitro by coculture of leukemia cells with developing or mature NK cells (Figure 5). NK cells at different stages of development were tested. These were NK cells undergoing differentiation from CD34 progenitors (Figure 5A), fresh mature NK cells from peripheral blood (Figure 5B), and long-term expanded and activated NK cells (data not shown).

To test the leukemic blast effect on NCR expression during NK ontogeny, NK cells were generated from CD34⁺ progenitors and analyzed after 15 days (greater than 98% CD3⁻CD56⁺ cells; data not shown) in the presence of normal allogeneic feeders (filled histograms), irradiated leukemia cells from NCR^dull patients (NKp30^dull/NKp46^dull; open histograms), or irradiated leukemia cells from NCR^bright patients (NKp30^bright/NKp46^bright; striped histograms). NKp30 expression obtained in the presence of leukemia cells from NCR^dull patients (MFI, 109 ± 37) was lower than expression in the presence of allogeneic normal feeders (MFI, 152 ± 57) or leukemia cells from NCR^bright patients (MFI, 147 ± 78). NKp46 expression obtained in the presence of leukemia cells from NCR^dull patients (MFI, 108 ± 40) was lower than expression in the presence of allogeneic normal feeders (MFI, 130 ± 23) or leukemia cells from NCR^bright patients (MFI, 171 ± 71). No significant difference in the activation-induced NKp44 expression and in NKG2D or 2B4/CD244 expression was observed (data not shown). As control, no significant difference in CD16 expression was observed between the different culture conditions (data not shown).

We then tested the effect of leukemia cells on mature NK cells. NCR^bright NK cells were isolated from healthy donors and cultured for 1 to 5 days in IL-2–supplemented medium and leukemia cells (Figure 5B). Rapidly, and at all times of culture, we observed a significant decrease (P < .05) in NKp30 expression when leukemia cells from patients with NCR^dull leukemia were used as feeders (open bars), in contrast to feeders from healthy donors (filled bars). In contrast, leukemia cells from patients with NCR^bright did not induce significant modification in NCR expression (data not shown). In addition, no difference in NKp46 expression was observed (data not shown). As control, CD16, 2B4/CD244, and NKG2D expression was checked but did not change when leukemia or normal feeder cells were used (data not shown).

We measured patient overall survival using the Kaplan-Meier model, determining different subgroups based on NKp30 and NKp46 expression. To obtain a sufficient number for potent statistical analysis, we included in our survival analyses patients phenotypically and functionally characterized in our previous study.²⁴  Among the 14 additional AML patients, 11 had NCR^dull and 3 had NCR^bright leukemia. Resultant duration of follow-up was 3.8 years. We first analyzed patient survival according to NKp30 expression (Figure 6A). Probability of survival at 5 years was 60% for patients with NKp30^bright AML and 33% in patients with NKp30^dull AML (P = .04). We then analyzed patient survival according to NKp46 expression (Figure 6B). We observed higher survival rates for patients with NKp46^bright than for those with NKp46^dull (72% vs 28% respectively; P = .003). We then compared the 3 following subgroups of patients: NCR^bright (NKp30^bright/NKp46^bright), NCR^dull (NKp30^dull/NKp46^dull), and NCR^discordant (Figure 6C). Probability of survival at 5 years was 64% for NCR^bright patients, 43% for NCR^discordant patients, and 31% for NCR^dull patients. These observations strongly suggest a correlation between poor survival and deficiency of NK-cell–triggering receptors. To analyze crucial variables that could affect overall survival, we performed multivariate analyses (Cox model; data not shown). We failed to reveal a statistical association between NCR phenotype and age, sex, white blood count, NK-cell count, karyotypic abnormality, relapse, and cause of death (equally represented by leukemia evolution and lethal infectious events). Along with karyotypic abnormalities, NCR expression was the only parameter that showed a significant association with survival.

Reduced NCR expression was observed in patients with long-term IL-2–expanded AML-NK cells²⁴  that could have been a consequence of the in vitro selection of NCR^dull NK cells. Against this hypothesis, we observed that the in vitro growth rate of NCR^dull AML-NK cells was comparable to that of their NCR^bright counterparts (data not shown). We then directly analyzed, without in vitro culture, AML-NK cells from 71 AML patients at diagnosis and observed an NCR^dull phenotype in most patients. Interestingly, we observed discordance between NKp30 and NKp46 expression—that is, cells were either NKp30^dull/NKp46^bright or NKp30^bright/NKp46^dull—in a significant fraction (15%) of patients. This phenotype has never been observed, to our knowledge, in healthy donors with parallel levels of expression for NKp30 and NKp46, with only NKp30^dull/NKp46^dull (NCR^dull) or NKp30^bright/NKp46^bright (NCR^bright) NK.¹⁵  This unusual differential expression of NKp30 and NKp46 sustains the hypothesis that the down-regulation of NKp30 and NKp46 observed in AML patients could rely on distinct mechanisms. An important issue raised by the abnormal NCR phenotype observed in AML patients was congenital origin (perhaps favoring leukemia occurrence) compared with acquired origin.²⁴  Because of the rarity of AML (incidence rate less than 1/100 000 per year), the possibility of a prospective study implicating extensive flow cytometry analysis of healthy subjects to determine NCR status before leukemia was excluded. As a surrogate, we initiated a kinetic study of NCR expression in AML patients. In most patients, we observed partial (NKp30) or complete (NKp46) recovery of NCR expression. Thus, NCR^dull was probably acquired as leukemia developed. This observation was completed by the detailed analysis of NCR recovery, depending on leukemia evolution during and after treatment. Interestingly, we observed that NCR recovery was mainly restricted to patients achieving CR. In addition, we observed a down-regulation of NCR expression in patients experiencing leukemia relapse. These data at least suggest a positive correlation between the NCR^dull phenotype and the presence of leukemia blasts. Given that few data are available regarding the regulation mechanism of NCR expression, we explored the unique clue available from the literature, which is the involvement of TGF-β1. TGF-β1 induces a drastic down-modulation of NKp30.²⁹  We thus measured TGF-β1 concentrations in peripheral-blood and bone marrow aspirate of AML patients (data not shown). As previously observed by another group, detected levels of TGF-β1 were lower in patients with AML than in healthy subjects.³⁰  Moreover, we failed to detect significant TGF-β1 concentrations in leukemia blast supernatants (data not shown). These experiments strongly suggest that TGF-β1 is probably not the mediator of NCR down-regulation in AML patients. In addition, TGF-β1 overexpression could explain NKp30, but not NKp46, down-regulation because the expression of the latter was not affected by TGF-β1,²⁹  suggesting at least the existence of another mechanism leading to NKp46 down-regulation. Nevertheless, we cannot exclude a paracrine effect of leukemia-secreted TGF-β1 undetectable in the sera of patients and not in our culture conditions. Because clinical data suggest that leukemia blasts induce the NCR^dull phenotype, we performed an in vitro analysis of the effects of neoplastic cells on NCR expression. Under NK-differentiating conditions, the coculture of CD34⁺ hematopoietic stem cells with leukemia cells of patients with an NCR^dull phenotype induced a decreased expression of NKp30 and, to a lesser degree, of NKp46. In those culture conditions, we did not observe the NCR^dull phenotype, as expected. This inability for blasts cells to induce the NCR^dull phenotype in differentiating NK cells might have had several origins: the process took place in mature NK cells, not during NK differentiation, or culture conditions were not favorable for AML blast effect (high cytokine concentration, leukemia cells, and CD34⁺ progenitor containment). Nevertheless, these data suggest leukemia cells could interfere with the initial steps of NK-cell differentiation. Moreover, culture of mature peripheral-blood normal NCR^bright NK cells with leukemia cells induced a significant down-regulation of NKp30, but not of NKp46. Because NKp30 and NKp46 are both down-regulated in AML-NK cells, our data suggest that leukemia cells may act during NK ontogeny and on mature NK cells to down-regulate NCR. We further completed our analysis by study of the in vitro reversibility of the NCR^dull phenotype. In most cases (there were some exceptions), the expansion of NCR^dull AML-NK cells with normal feeders, after the depletion of leukemia cells, failed to restore a normal NCR phenotype (data not shown). Thus, in vitro restoration of a normal NCR phenotype is likely to have been caused by novel NK populations and not by the restoration of normal NCR expression by primarily NCR^dull NK cells. We also performed experiments using leukemia blast culture supernatants (in resting or GM-CSF–stimulated conditions), but we failed to detect significant modification in NCR expression (data not shown). Altogether, these data argue for a mechanism of NCR down-regulation involving cell-to-cell contact between leukemia blasts and NK. Leukemia cells are able to down-regulate NCR expression. Therefore, what could have been the physiopathologic meaning of such a phenomenon? We can speculate that NCR down-regulation is another mechanism of tumor escape from innate immunity. Nonetheless, it could be argued that NCR expression, in most cases, was still present, though at reduced levels. We have shown that NCR down-regulation is sufficient to drastically reduce in vitro NK-cell cytotoxicity. Moreover, a recent study has shown that leukemia blasts express low amounts of the still unidentified NCR ligands.²³  The convergent effect of low NCR expression by AML-NK cells and low ligand expression by leukemia cells may suggest an evasion of leukemia from NK immune surveillance. From all these in vitro data, it could be hypothesized that the NCR^dull phenotype could have a consequence on leukemia clinical evolution. Kaplan-Meier survival curves markedly showed that patients with NKp30^bright or NKp46^bright leukemia survive longer than patients with NKp30^dull or NKp46^dull leukemia and that patients with NCR^bright leukemia survive longer than those with NCR^discordant leukemia, who in turn survive longer than those with NCR^dull leukemia. The NCR phenotype may influence patient outcome through an antileukemia effect or through anti-infection control because leukemia evolution and infectious complications equally contribute to patient outcomes. We can hypothesize that higher NCR expression at diagnosis and better recovery after CR enable better antiopportunistic infection and better antileukemia responses to AML-NK. Finally, our results showed that NCR expression can be considered a new prognostic marker for overall survival in AML patients. Aside from immunotherapy protocols with autologously restored NK cells, this new karyotype-independent prognostic marker deserves consideration in chemotherapy protocol design and analysis. Our findings represent a rationale for clinical immunotherapy studies aimed at restoring NK cytotoxic function through the restoration of NCR expression.

This work was supported by Groupement Entreprise Français Lutte Cancer, Fédération Nationale des Centres de Lutte Contre le Cancer, Fondation pour la Recherche Médicale, and Ligue Nationale contre le Cancer.

We thank Professor Alessandro Moretta for his valuable advice and commentary from the beginning of the study. We thank Florence Baratier and Dany Jayme for flow cytometry analysis. We thank Christine Fornelli of Beckman-Coulter-Immunotech for her help in mAb settings and for critical discussions.