Human ex vivo NK cells were enriched by using anti-CD56 MicroBeads with a MultiStand MACS (MACS, Miltenyi Biotec) from freshly isolated peripheral blood mononuclear cells (PBMCs).

Activation of human primary NK cells was pursued by culturing PBMCs for 5 days with Mitomycin C-treated R69- or 721.221-LCL at 10:1 PBMC:stimulator ratio. Subsequently, NK cells were enriched using anti-CD56 MicroBeads with a MultiStand MACS (MACS, Miltenyi Biotec).

PBMCs were obtained from blood from healthy donors by Ficoll gradient centrifugation (Blood and Tissue Bank of Aragon; approved by the CEICA, number: C.I.PI11/006). NK cell purity (CD56⁺/CD3⁻) was higher than 90% in all cases. Contamination with CD8⁺ CD3⁺ cells was less than 2%.

Blood samples from patients with B-CLL were obtained from Hospital Clinico Lozano Blesa (Zaragoza) and Hospital Puerta de Hierro-Majadahonda (Madrid). They were processed by Ficoll gradient centrifugation to obtain PBMCs and stored frozen in liquid nitrogen until their use. In all cases, the percentage of CD19⁺CD5⁺ cells was higher than 80% and no differences were observed between frozen and fresh B-CLL cells regarding their susceptibility to NK cells. This study was approved by Ethics Committee for Clinical Research of Aragon (CEICA), number: C.I.PI11/006; and by Ethics Committee for Clinical Research of Hospital Puerta de Hierro-Majadahonda, number: PI31_13.

IGHV gene mutational status was analyzed and classified according to ERIC recommendations (53). TP53 genetic abnormalities were detected by conventional cytogenetics, FISH, and Sanger sequencing analysis.

NK cells were labeled with 1 μM of CellTracker™ Green CMFDA (Life Technologies) and incubated with target cells at 9:1 e:t ratio for 4 h. Subsequently, phosphatidylserine (PS) translocation and membrane damage were analyzed in the green or violet fluorescence negative target cell population by flow cytometry using annexin V and 7AAD as previously described (51).

Killer-cell immunoglobulin-like receptor ligands were deduced from the cell HLA types, determined by standard methods approved by the European Federation for Immunogenetics, as in Ref. (54) or directly assessed by PCR with sequence-specific primers targeted to the critical polymorphic positions (oligonucleotide sequences and PCR conditions available upon request). Mismatch was defined, according to the missing-self model, as the absence in the target cell of a KIR ligand present in the effector cells.

The antibodies against HLA-E-PE (clone 3D12), HLA-ABC/E/F/G-FITC (clone W6/32), DR4-PE (clone DJR1), and DR5-PE (clone MD5-1) were from eBioscience. The antibody against ICAM-1-APC (clone HA58) was from BD. The antiCD19-APC (clone LT19) and antiCD5-FITC (clone UCHT2) were from Miltenyi Biotec. The NKR-Fc quimeras were from R&D Systems (NKG2D-Fc) or were provided by Miguel López-Botet (NKp30-Fc and NKp46-Fc). The anti-human IgG secondary antibody labeled with PE was from Jackson Immunoresearch. All antibodies were diluted in PBS with 5% FCS and 0.1% sodium azide for cell staining. To stain NKR ligands using NKR-Fc quimeras, cells were blocked with rabbit serum before staining with the respective chimeras.

Statistical analyses were performed using the GraphPad Prism v 4.0 software by one-way analysis of variance (ANOVA) as indicated in the figure/table legends. Regression-tree analyses (Decision tree learning) were used to classify B-CLL samples according to its susceptibility to NK cells. This is a predictive model that maps the values of cell death (percentage) to conclusions about categorical predictor variables.

As mentioned above, it is unclear whether NK cells activated with exogenous cytokines may kill B-CLL cells. Our group has previously described that NK cells activated with EBV-transformed R69 or 721.211 LCLs are able to eliminate cells from B-CLL patients without requiring supplementation with exogenous cytokines (51). Importantly, activated NK cells did not kill healthy non-transformed PBMCs. Here, we confirmed that the cytotoxic potential of NK cells activated with R69 LCLs was increased in comparison with freshly isolated NK cells from the same donors. As shown in Figure 1A, freshly isolated allogeneic NK cells from healthy donors did not kill B-CLL cells. In contrast, activated NK cells presented a significantly increased cytotoxic potential against B-CLL cells, even in cells from a patient (CLL2) with chemotherapy and Rituximab resistance (Table 1). Killing of Jurkat cells and R69 LCL (that expresses all known HLA ligands for inhibitory KIRs) was also significantly enhanced after NK cell activation. Cell death was almost completely inhibited by EGTA indicating the granule exocytosis was the main mechanism employed by NK cells to kill B-CLL cells (data not shown). As previously found in NK cells activated with R69 LCLs (51), the NK cell receptors NKG2A, NKp30, NKp44, and DNAM and the cytotoxic protease granzyme B were upregulated in activated NK cells (data not shown).

Next, we analyzed whether HLA-I expression in the LCLs used to activate NK cells might influence their anti-leukemic potential. To this aim we compared the cytotoxic activity of NK cells from 4 independent donors, activated either with HLA-I⁺ R69 LCLs or with HLA-I⁻ 721.221 cells, against B-CLL cells from four different patients. As shown in Figure 1, although with some individual variations, the cytotoxic potential of activated NK cell did not depend on HLA-I expression in stimulating LCLs (Figure 1B). Thus, we employed NK cells isolated from PBMC cultures activated for 5 days with R69 LCLs in the next experiments.

Our data confirm that LCL-activated NK cells are able to kill B-CLL cells. However, we have previously found that the cytotoxic potential of allogeneic NK cells against B-CLL cells varies depending on the NK cell donor/B-CLL patient pair (51).

Thus, experiments were set up to address the basis for the variable susceptibility of B-CLL cells to activated NK cells. Seven NK cell donors and six B-CLL patients were selected, and the level of cell death was evaluated in all combinations (42). Four and three different NK cells donors were blindly incubated with the six B-CLL patients on two different days. As shown in Figure 2A, the level of cell death varied depending on the pair of NK cell donors and B-CLL patients incubated in agreement with our previous results (51). Some donors like the one represented with circles presented a similar profile of killing against all CLL samples. However, this was not a general trend since others like the ones represented with triangles, diamonds, or inverted triangles presented a high variability depending on the CLL sample. Indeed, the profile of killing between the different NK cell donors was statistically similar (Supplementary figure). These killing profiles were not dependent on the level of NK cell activation since all of them similarly killed Jurkat cells and expressed comparable levels of granzyme B, NKG2D, and NCRs (NKp30, Nkp44, and Nkp46) (data not shown). Thus, a priori, this variability could be partly explained by (i) the degree of HLA–KIR mismatch between donor and patient as found in AML and ALL or (ii) the intrinsic resistance of cells from some B-CLL patients to activated NK cells.

As shown in the regression-tree analysis in Figure 2B, some B-CLL cells like CLL1 and CLL6 are more resistant than others to activated NK cells, irrespectively of the NK cell donor, supporting our second hypothesis. Thus, we decided to analyze the reason for the increased resistance of CLL1 and CLL6 to activated NK cells. First, we analyzed whether this could be related to the level of expression of ligands recognized by the NK cell receptors NKG2D, NKp30, NKp46, KIRs, and LFA-1. As shown in Figure 3A, the expression of NKG2D and NKp30 ligands were very low in most CLL samples, and only some CLLs expressed higher levels of NKp46 ligands. Jurkat cells expressed ligands for all these receptors, validating our approach. Next, we analyzed the expression of the adhesion molecule ICAM-1 and of HLA-I using both an antibody (clone W6/32) that recognizes all HLA-I molecules and another one specific for HLA-E. As shown in Figure 3B, all cells expressed HLA-I (ABC/E/F/G and E) and ICAM1, although the level of expression was quite variable. Of note, the level of expression of HLA-ABC and HLA-E in CLL1 and CLL6 was higher in comparison with the other B-CLL samples, suggesting that this expression could be related to the increased resistance of these samples. However, blocking of CD94 in NK cells, a protein required for the inhibitory signaling of the HLA-E ligand NKG2A, by using specific antibodies did not significantly enhance NK cell-mediated cytotoxicity (data not shown) suggesting that HLA-E expression in CLL1 and CLL6 was not the main responsible for NK cell resistance.

The adhesion molecule ICAM-1 was expressed by all B-CLL samples (Figure 3C), and the level of expression did not correlate with NK cell susceptibility. Although we did not test other adhesion molecules involved in NK cell recognition like LFA-3, we have previously shown that blocking ICAM-1/LFA-1 interaction almost completely eliminate NK cell cytotoxicity in hematological cancer cells (including B-CLL) [(55) and data not shown] suggesting a minor role for other adhesion molecules.

As mentioned above KIR–HLA mismatch has been shown to promote NK cell-mediated elimination of some hematological neoplasias, including acute monocytic and lymphocytic leukemias, and to a lesser extent multiple myeloma (56). However, this effect is not clear for B-CLL, and it could help to explain the different susceptibility of some B-CLL cells. Thus, we analyzed whether mismatches between NK cell donor KIRs and patient HLA ligands (HLA-A3/A11, -Bw4, -C1/2) would explain the differences observed. To this end, we determined the HLA epitopes for the seven NK cell donors and the six B-CLL patients. The results of these analyses are shown in Table 2 where the expression of known ligands for KIR3DL2 (A3/A11), KIR3DL1 (Bw4), KIR2DL1 (C2), and KIR2DL2/2DL3 (C1) are indicated.

Based on the degree of KIR ligand compatibility in donor to patient direction, we predicted matched (compatible) and mismatched (incompatible) combinations in the 42 NK-CLL pairs and analyzed whether the degree of mismatch significantly correlated with the susceptibility observed in the B-CLL samples. A summary of the percentage of matched (0) and mismatched (1) combinations is shown in Table 3. As expected, since the experimental design followed a double blind protocol, in which HLA genotype and mismatch was analyzed after testing NK cell-mediated cytotoxicity, a higher number of unmatched than matched combinations were analyzed. However, as shown in Figure 3, matched and unmatched NK-CLL combinations were tested at the same time, which partially compensate this potential limitation. As shown in Figure 4, except for one NK cell donor in CLL1, matched NK cell donors induced the lowest level of cell death in every B-CLL case, although in some instances differences were low. The median cell death in compatible B-CLL/NK cell combinations was 18%, and this value increased to 34% in mismatched combinations (Table 4). These differences became statistically significant using a one mixed-model analysis of variance (ANOVA). The model controlled for the within-subject nature of the seven NK cell donors by including random effects for B-CLL patient and B-CLL patient × NK cell interaction. Thus, although the statistical significance was close to 0.05, it should be noted that the model used for these analyses is very conservative taking into account that there is a non-controlled intrinsic variability in the components within each group (NK cell donors and B-CLL patients). Thus there was a good correlation between mismatching and NK cell cytotoxicity supporting that NK cell alloreactivity explains the variability observed in each B-CLL sample.

Although mismatch partly explained the variability observed in the different B-CLL/NK cell combinations, our results also suggest that in each B-CLL case there is/are intrinsic characteristic/s that regulate/s the susceptibility of B-CLL cells to NK cells which is independent on the expression of NK cell receptor ligands.

When we analyzed retrospectively the clinical evolution (response to treatment) of these patients, we realized that except for CLL3, paradoxically, samples from patients with a better response (CLL1) or in which no treatment was required (CLL6) were those more resistant to NK cell cytotoxicity (Table 1; Figure 2B). In contrast, the patient that did not respond to treatments (CLL2) was the most susceptible to NK cells. Based on this observation, we established a new experiment to analyze in more detail whether the prognosis of CLL patients correlated with the susceptibility of B-CLL cells to NK cells. Prognosis was established by determination of TP53 and IGHV mutational status as previously shown (16–19). Three different groups of B-CLL patients were established retrospectively according to the expression of wild-type (wt) or mutated/deleted (mut) TP53 and wt or mut IGHV: TP53^wtIGHV^mut (nine patients), TP53^wtIGHV^wt (eight patients), and TP53^mutIGHV^wt (five patients) as good, intermediate, and poor prognostic groups (Table 5). The samples of these groups were blindly incubated with NK cells from several donors randomly selected and the level of cell death was determined.

As shown in Figure 5, the susceptibility of B-CLL cells to NK cells was clearly enhanced in the worse prognostic group in comparison with good and intermediate prognosis. Expression of wt IGHV in the context of wt TP53 also increased the susceptibility of B-CLL cells to NK cells although did not reach statistic significance.