Analysis of the receptor-ligand interactions in the natural killer–mediated lysis of freshly isolated myeloid or lymphoblastic leukemias: evidence for the involvement of the Poliovirus receptor (CD155) and Nectin-2 (CD112)

Natural killer (NK) cells represent a lymphoid subpopulation characterized by the ability to raise a potent cytolytic activity against tumor or virally infected cells.^1,2  Their function is regulated by a series of surface receptors, capable of transducing either inhibitory or activating signals.^3,4  The inhibitory receptors that lead to NK cell inactivation represent the major fail-safe mechanism to prevent killing of normal major histocompatibility complex (MHC) class I⁺ autologous cells.⁵  An array of different MHC class I–specific inhibitory receptors has been identified.^6-8  In humans, killer immunoglobulin-like receptors (KIRs) are responsible for recognition of allotypic determinants that are shared by different HLA-class I alleles.⁹  Briefly, KIR2DL1 recognizes HLA-C alleles characterized by Lys at position 80 (HLA-C^Lys80), KIR2DL2/3 recognize HLA-C alleles characterized by Asn at position 80 (HLA-C^Asn80), KIR3DL1 is specific for HLA-B alleles sharing the Bw4 supertypic specificity (HLA-B^Bw4), and KIR3DL2 recognizes HLA-A3 and -A11 alleles. ILT2 receptor is more “promiscuous” since it recognizes many different HLA-class I alleles, although in most instances it displays a limited inhibitory capacity.¹⁰  CD94/NKG2A is specific for HLA-E, whose surface expression depends on the binding of peptides derived from the signal sequence of different HLA loci.¹¹  Thus, HLA-E levels of expression usually parallel HLA-class I surface density in cells with a complete haplotype and, consequently, CD94/NKG2A would sense levels of HLA class I expression.¹²  KIRs are clonally distributed and each type is expressed only by a subset of NK cells.⁶  Moreover, each NK cell expresses at least one receptor specific for self HLA class I molecules, thus preventing killing of normal autologous cells. This type of receptor distribution allows the whole NK cell pool to sense the loss of even single HLA class I alleles on self cells, a frequent event in tumor transformation.¹³  Moreover, in an allogeneic setting, KIR⁺ NK cells will lyse target cells in case of KIR/HLA-class I mismatch. Alloreactive NK cells were shown to play a key role in eradicating leukemias since recent studies demonstrated that KIR/HLA incompatibility in the graft-versus-host direction was associated with a better clinical outcome in patients receiving hematopoietic stem cell transplants from either haploidentical relatives or unrelated donors.^14-17 

In the absence of efficient inhibitory interactions, target cells are susceptible to NK-mediated killing. Different receptors and coreceptors are responsible for NK cell activation,¹⁸  including the natural cytotoxicity receptors (NCRs; ie, NKp46, NKp30, and NKp44),¹⁹  NKG2D,²⁰  2B4,²¹  NKp80,²²  NTBA,²³  CD59,²⁴  and DNAM-1.²⁵  Functional studies indicate that triggering receptors induce killing of target cells upon recognition of specific ligands.²⁶  So far, these ligands have been unveiled: MICA/B and ULBPs for NKG2D,^27-29  CD48 for 2B4,³⁰  Poliovirus receptor (PVR) and Nectin-2 for DNAM-1,³¹  whereas NTBA has a homophilic interaction.³²  NCRs are still orphan receptors, although they play a major role in the NK-mediated killing of most tumor cell lines, as revealed by monoclonal antibody (mAb)–mediated receptor masking experiments. In particular, lysis of acute myeloid leukemia (AML) has been described to be mainly NCR dependent; moreover, the lack of killing of leukemic blasts has been correlated either with an NCR^dull phenotype of the NK effector cells or with a lack of NCR ligands as suggested by indirect functional assays.³³ 

In this study we analyzed a large panel of myeloid and lymphoblastic leukemias for their susceptibility to NK-mediated lysis, with particular focus on the inhibitory or activating interactions occurring between NK and leukemic cells.

This study was approved by the Universitá di Genova institutional review board. Peripheral blood samples were obtained from patients after informed consent according to the Declaration of Helsinki. Diagnosis was established by cytologic and immunophenotypic criteria: AMLs were classified according to French-American-British (FAB) criteria, acute lymphoblastic leukemias (ALLs) by immunophenotyping for specific lineage markers. We analyzed 24 AMLs from M0 to M7 FAB subtypes, 1 chronic myeloid leukemia (CML), 11 ALLs, and 3 chronic lymphoblastic leukemias (CLLs). Leukemic cells, obtained by Ficoll-Hypaque density centrifugation of peripheral blood or bone marrow samples, were analyzed either freshly derived or after thawing viable cells, and cryopreserved in liquid nitrogen.

NK cells from healthy donors (some were relatives of the patients) were isolated using the RosetteSep method (StemCell Technologies, Vancouver, BC, Canada). NK cells were cultured for 5 days in the presence of 100 U/mL recombinant interleukin 2 (rIL-2; Proleukin; Chiron, Emeryville, CA) to obtain short-term activated polyclonal NK cell populations. NK cells were also cultured on irradiated feeder cells in the presence of 1.5 ng/mL phytohemagglutinin (Life Technologies, Paisley, Scotland) and rIL-2 to obtain proliferation of polyclonal or, after limiting dilution, clonal NK cells.²⁹ 

The following mAbs, produced in our laboratory, were used in this study: JT3A (IgG2a, anti-CD3), c127 (IgG1, anti-CD16), c218 (IgG1, anti-CD56), BAB281 and KL247 (IgG1 and IgM, respectively, anti-NKp46), Z231 and KS38 (IgG1 and IgM, respectively, anti-NKp44), A76 and F252 (IgG1 and IgM, respectively, anti-NKp30), BAT221 (IgG1, anti-NKG2D), GN18 and F5 (IgG3 and IgM, respectively, anti–DNAM-1), MA344 (IgM, anti-CD48), MA127 (IgG1, anti-NTBA), GL183 (IgG1, anti-KIR2DL2/3/S2, anti-CD158b1/b2/j), EB6b (IgG1, anti-KIR2DL1/S1, anti-CD158a/h), Z27 (IgG1, anti-KIR3DL1, anti-CD158e), AZ158 (IgG2a, anti-KIR3DL1, anti-CD158e), Z270 and Z199 (IgG1 and IgG2b, respectively, anti-NKG2A), F278 (IgG1, anti-ILT2), L95 (IgG1, anti-PVR), L14 (IgG2a, anti-Nectin-2), BAM195 (IgG1, anti-MICA) and A6-136 (IgM, anti-HLA class I).^{6,18,29,31,34}  M362 (IgG1, anti-MICB), M295 (IgG1, anti-ULBP1), M310 (IgG1, anti-ULBP2), M551 (IgG1, anti-ULBP3), and M475 (IgG1, anti-ULBP4) were kindly provided by Amgen, Seattle, WA. The anti-CD20 (Leu16, IgG1), anti-CD19 (Leu12, IgG1), anti-CD34 (8G12, IgG1), anti-CD33 (P67.6, IgG1), and anti-CD14 (MΦP9, IgG2b) were purchased from Becton Dickinson, Mountain View, CA. The anti-CD158b1/b2 (CH-L, IgG2b) was provided by BD Biosciences Pharmingen, San Diego, CA. The anti-CD56-PC5 (N901, IgG1) and anti-CD48 (J4-57, IgG1) were provided by Immunotech-Coulter, Marseille, France. To evaluate surface expression of HLA-class I molecules, W632 mAb was used.

To analyze the surface markers of leukemic cells, samples were first preincubated for 10 minutes at 4°C with hIgG (1 mg/mL) and then stained with the appropriate mAbs followed by phycoerythrin (PE)–conjugated AffiniPure F(ab')₂ goat antimouse IgG (Jackson Immunoresearch Laboratories, West Grove, PA). To compare the surface densities of various molecules among different leukemias, we calculated the MRFI, that is, the ratio between the mean fluorescence intensity of cells stained with the selected mAb and that of cells stained with the isotype-matched control mouse Ig.

To define the alloreactive NK cell subset, we performed either triple fluorescence analysis of peripheral blood mononuclear cells (PBMCs) gating on CD56⁺ lymphocytes (using anti-CD56–PC5 mAb) or double fluorescence analysis of purified IL2-activated NK cell populations. The alloreactive NK cell subset was evaluated as percentage of cells expressing only KIRs that don't recognize HLA-class I alleles of the patient's haplotype, identified with the same fluorochrome (fluorescein isothiocyanate [FITC] or PE), and cells not expressing KIRs that recognize HLA-class I alleles of the patient's haplotype, identified with the other fluorochrome. Briefly, cells were incubated with appropriate mAbs followed by FITC- or PE-conjugated isotype-specific goat antimouse second reagents (Southern Biotechnology Associated, Birmingham, AL).

Samples were analyzed by cytofluorimetric analysis on a FACSort and with the Cell Quest program (both from Becton Dickinson).³⁴ 

NK cells were tested for cytolytic activity in a 4-hour ⁵¹Cr-release assay as previously described.³⁴  When target cells were represented by the freshly isolated leukemic cells, we labeled them using 300 μg (instead of the standard 100 μg) ⁵¹Cr/10⁶ cells. In case of low efficiency of labeling, 10 000 (instead of the standard 5000) cells/well were plated. The effector-to-target (E/T) ratio was calculated accordingly to the target cell numbers. In all instances, leukemic blasts to be used as target cells were purified in order to obtain homogeneous (ie, more than 90%) tumor cell preparations. If needed, a negative selection was operated using mAb to markers selectively expressed in the normal counterpart and Dynabeads (Dynal, Oslo, Norway).³⁴  To correlate the susceptibility to lysis of leukemic cells to KIR/HLA ligand interactions, we also tested, as target cells, Epstein-Barr virus transformed B (B-EBV) cell lines derived from HLA-typed donors, characterized by the combination of informative groups of HLA-B and HLA-C alleles. In particular, in this study we used the following B-EBV cell lines: BM15 expressing the homozygous haplotype A1, B49 (Bw4), Cw7 (C^Asn80), kindly provided by professor G.B. Ferrara (Istituto Scientifico Tumori, Genova, Italy) and the cell line derived from donor AC expressing A2, B8,35 (Bw6), Cw2,7 (C^Lys80/C^Asn80). HLA-class I haplotypes of patients and potential NK donors were also analyzed. In Table 1 the 2 representative cases regarding ML2 and ML11 are reported.

For masking experiments, NK cells were preincubated with mAbs specific to the various receptors (of IgM isotype, if available) at the concentration of 10 μg/mL, and, after washing, used in the cytolytic assay.

For redirected killing assays, P815 cells were used as target cells in the presence of mAbs of IgG isotype at a concentration of 0.5 μg/mL.

The susceptibility of AML to NK-mediated lysis in haploidentical bone marrow transplantation has been shown to correlate with the existence of a mismatch between KIRs expressed by donor NK cells and HLA-class I alleles expressed by the patient.¹⁵  However, substantial differences may exist among possible donors in the proportions of NK cells expressing KIR that do not recognize the patient's HLA-class I alleles (referred as “alloreactive NK cells”).^6,16,35  The surface levels of HLA-class I molecules expressed in different leukemias may also play a role in the degree of NK-mediated lysis of leukemic cells. To address these points, we first analyzed a large panel of AMLs, belonging to different FAB subtypes, for the surface density of HLA-class I molecules. To this end, we utilized W632 mAb, a pan anti–HLA-class I antibody. In Table 2, the MRFI is indicated for each leukemia. It is evident that a great variability exists in the HLA-class I expression among different leukemias. In addition, no correlation appears to exist between the FAB subtypes and the surface levels of HLA-class I molecules.

In order to properly analyze the effect of alloreactive NK cells, present in NK populations derived from different donors, we used as target cells leukemias expressing high levels of HLA-class I molecules. In particular, as shown in Figure 1, we analyzed the leukemia ML2, belonging to M1 subtype (Figure 1A,C), and the leukemia ML11, belonging to M2 subtype (Figure 1B,D), for susceptibility to lysis by NK populations containing different proportions of potentially alloreactive NK cells (Table 1 illustrates HLA-class I haplotypes of patients and donors). Notably, the effector NK cells tested against ML2 were derived, respectively, from one relative of the patient (ie, a potential donor of haploidentical hematopoietic precursors, donor A) and an unrelated donor (donor B). The potentially alloreactive NK cells were identified on the basis of the surface expression of different KIRs and CD94/NKG2A. Thus, since ML2 expressed both HLA-C^Asn80 and HLA-C^Lys80 allotypes, no KIR/HLA-C mismatch was expected (Table 1). On the contrary, alloreactive NK cells could be represented by KIR3DL1⁺ cells since the HLA-B^Bw4 allotypes recognized by this receptor were absent in the patient's haplotype. Notably, the presence in the donor of cells expressing KIR3DL1 could be predicted by the presence in the HLA haplotype of both donors of HLA-B alleles belonging to the Bw4 supertypic specificity. Indeed, a population expressing KIR3DL1 but lacking KIR2DL1, KIR2DL2/3, and NKG2A existed in both donors, although in different proportions when freshly derived NK cells were analyzed (not shown). After a 5-day culture in the presence of IL-2, the cells maintained the same proportions of KIR and NKG2A expressing cells, that is, 4% in donor A and 35% in donor B (Figure 1A). These polyclonal NK cells were used as effector cells in a cytolytic assay against ML2 target cells. As shown in Figure 1C, NK cells derived from donor B, containing a higher proportion of potentially alloreactive cells, lysed more efficiently than those derived from donor A. This difference in the magnitude of the alloreactive effect was further verified in control experiments in which target cells were represented by the B-EBV cell line AC, selected according to the expression of HLA-B and HLA-C alleles belonging to the same groups, as defined by KIR-mediated recognition (ie, HLA-B^Bw6, HLA-C^Asn80/C^Lys80), expressed by ML2.

In the case of leukemia ML11, characterized by the expression of HLA-B^Bw4 and HLA-C^Asn80, the potentially alloreactive NK cells were expected to be confined to the subset expressing KIR2DL1 but lacking KIR2DL2, KIR3DL1, and NKG2A inhibitory receptors (Table 1). The presence of a KIR2DL1⁺ NK subset in these donors would be expected to be involved in keeping in check the expression of HLA-C^Lys80 alleles on autologous cells. Note that donor C was a relative of the patient, whereas donor D was an unrelated donor selected on the basis of a relatively high proportion of NK cells with a phenotype potentially alloreactive against ML11 (Table 1). NK cells from these 2 donors were cultured for 5 days in IL-2 and analyzed for expression of KIRs and NKG2A. As shown in Figure 1B, donor C and donor D had 2% and 22% potentially alloreactive NK cells, respectively. These polyclonal NK cells were tested against ML11 and the B-EBV cell line BM15, which expressed alleles belonging to HLA-B^Bw4 and HLA-C^Asn80, that is, corresponding to ML11. Again, both NK cell populations lysed target cells, although with different efficiency reflecting the different proportions of alloreactive NK cells contained in each population. It should be mentioned that donor D expressed both KIR2DL2 and KIR2DS2, as assessed by redirected killing assay using GL-183 mAb and confirmed by the analysis of KIR genotype (not shown). Furthermore, NK cell clones, derived from the same donor, coexpressing KIR2DL1 and KIR2DS2, displayed a potent alloreactivity against ML11 leukemia or different B-EBV cell lines characterized by the presence of HLA-C^Asn80 and the lack of HLA-C^Lys80. Thus, in this case, the proportion of alloreactive NK cells was even higher than that estimated by fluorescence-activated cell sorter (FACS) analysis, since available mAbs do not discriminate between activating or inhibitory KIR.

Although not shown, NK cell clones, characterized by the expression of CD94/NKG2A as the only inhibitory receptor, did not lyse both ML2 and ML11 AML, whereas killing was restored by mAb mediated–masking of either NKG2A or HLA-class I molecules. This suggests that, at least for leukemias expressing high levels of HLA-class I molecules, HLA-E can protect leukemic cells from NK cell lysis. Thus, the expression of NKG2A as an inhibitory receptor should be taken into account during the assessment of the proportion of alloreactive NK cells within a given NK population (Figure 1A-B). On the other hand, ILT2 expression was not considered in the double fuorescence analyses shown in Figure 1 since it was poorly represented (< 5% positive cells) in the NK cell populations tested.

Another central issue in the analysis of the relationship between NK cells and AML is the identification of triggering NK receptors that are actually involved in inducing the NK-mediated lysis of AMLs. To investigate this point, we analyzed the main receptors and coreceptors that are known to be involved in NK cell triggering in the process of tumor cell lysis. These include the NKp46, NKp30 and NKp44 (NCR), NKG2D, DNAM-1, and 2B4.¹⁸  mAb-mediated masking of a receptor that is involved in lysis would result in inhibition of cytolytic activity. Figure 2 shows one representative leukemia belonging to the M0 FAB subtype (ML1), one belonging to M1 (ML2), one belonging to M3 (ML14), and one to M5 (ML19). Effector NK cells were represented by polyclonal NK cells displaying alloreactivity against one or another of these leukemias. It is evident that mAb-mediated masking of NKp30 and NKp46 consistently resulted in a sharp inhibition of lysis. In contrast, NKG2D, another major receptor, does not appear to play any substantial role. This was true also for 2B4. On the other hand, mAb-mediated blocking of DNAM-1 consistently resulted in a significant inhibition (Figure 2). Although not shown, when an NCR^dull alloreactive NK cell clone was analyzed against ML1, the only receptor responsible for killing appeared to be DNAM-1. A similar involvement of NCR and DNAM-1 was observed in leukemias belonging to different FAB subtypes. In general, combined masking of both NCR and DNAM-1 receptors virtually abrogated cell lysis (not shown). It is of note that susceptibility of different AMLs to lysis by allogeneic NK cells was more readily detected in those leukemias expressing low surface density of HLA-class I molecules or, for those expressing high density, by the use of appropriate alloreactive NK cells. It is of note that only ML7 leukemia could not be lysed by NK cells even upon mAb-mediated masking of HLA-class I molecules. This might suggest the lack of ligands for relevant triggering receptors or the occurrence of still undefined inhibitory interactions involving non–HLA-specific receptors.

The fact that only some activating NK receptors are involved in lysis of AML blasts may suggest that the ligands for these receptors are expressed at the leukemic cell surface. While the cellular ligands of NCR have not been identified so far, ligands of NKG2D, DNAM-1, and 2B4 are known and could thus be investigated by specific mAb and FACS analysis. These include MICA/B and ULBPs (NKG2D ligands), Nectin-2 and PVR (DNAM-1 ligands), and CD48 (2B4 ligand). As shown in Table 2, all AMLs consistently expressed PVR and Nectin-2 (the only exception being represented by the ML7 leukemia). Note also that the levels of PVR and Nectin-2 were comparable in most AMLs, but we observed many cases, especially those expressing the CD34 molecule (not shown), in which Nectin-2 displayed a higher density as compared with PVR. These data are consistent with the inhibitory effect exerted by anti–DNAM-1 antibody in the NK-mediated lysis of AMLs (Figure 2). Regarding the NKG2D ligands, MICA and MICB were consistently negative in all AMLs analyzed; ULBPs were either negative or expressed in low amounts. Figure 3 shows the FACS profiles of representative markers in an M1 leukemia. In Figure 3A, gated leukemic cells of ML2 were analyzed for the expression of CD33, CD34, CD48, Nectin-2, PVR, and ULBP3 (normal monocytes were virtually absent, since CD14 was only 0.5%). It is of note that PVR and Nectin-2 were expressed at a higher density as compared with both ULBP3 (the only NKG2D ligand found in ML2, as shown in Table 2) and CD48. These data are in agreement with the finding that DNAM-1 but not NKG2D and 2B4 are involved in lysis of this leukemia (Figure 2B). Figure 3B shows FACS analysis of the whole ficoll-isolated cell fraction containing both leukemic and normal mononuclear cells. It is evident that staining with anti–Nectin-2 is restricted to the leukemic population (CD33⁺ and mostly CD34⁺), whereas normal lymphocytes are negative. Similar results were obtained with PVR (not shown). On the other hand, CD48 displays a high expression in normal lymphocytes, whereas it is poorly expressed in leukemic cells. Although not shown, similar patterns of expression were found in other cases in which normal lymphocytes were clearly distinguishable from leukemic blasts (ie, ML12 and ML14). Thus, the DNAM-1 ligands PVR and Nectin-2 were expressed on AML cells but not on normal lymphocytes, while CD48 appeared down-regulated in AMLs. In general, expression of PVR and Nectin-2 paralleled the susceptibility to NK-mediated lysis. Indeed, whereas all AMLs analyzed were PVR⁺ Nectin-2⁺ and were susceptible to NK-mediated lysis, ML7 did not express PVR or Nectin-2 and was resistant to lysis. In contrast, expression of NKG2D ligands in AML was only an occasional finding and, when present, was found at low surface density and, consequently, with a marginal functional relevance.

We analyzed a panel of lymphoblastic leukemias, both chronic and acute, of different subtypes for the expression of known ligands for activating NK receptors. The surface density of HLA-class I was also evaluated. These phenotypic analyses are relevant to predict whether a given leukemia will be susceptible to lysis by NK cells. As shown in Table 3, in general, lymphoblastic leukemias express high levels of HLA-class I molecules as compared with myeloid leukemias. This implies that lymphoblastic leukemias are more resistant to lysis by NK cells and that only KIR-mismatched NK cells may exert cytotoxicity against this tumor. Analysis of the expression of PVR and Nectin-2 or MICA and ULBPs revealed a relatively infrequent expression of the various ligands. These data suggest that lymphoblastic leukemias may be less susceptible to NK-mediated lysis, although the lack of information on NCR ligands does not allow any precise prediction.

In Figure 4 are shown selected leukemias that expressed Nectin-2. Surface expression of this DNAM-1 ligand was analyzed by double fluorescence against informative markers for staining of leukemic cells. We also evaluated the expression of NTBA, which is expressed on normal NK, T, and B lymphocytes and has been described as an activating coreceptor in NK cells. Since NTBA is characterized by a homophilic interaction, we evaluated its expression on lymphoblastic leukemias to analyze the possible expression of an additional ligand involved in NK cell activation. In Figure 4A, leukemic cells derived from patient LL10 expressed CD19 and CD34. Moreover, these markers were coexpressed with Nectin-2. On the other hand, leukemic cells did not express CD48 (or very weakly), thus suggesting that this surface antigen may be down-regulated not only in myeloid but also in lymphoblastic leukemias. Interestingly, similar to CD48, NTBA also was sharply down-regulated on leukemic cells. Therefore, on the basis of phenotypic analysis, this leukemia is expected to be susceptible to NK-mediated lysis by a DNAM-1–dependent mechanism. Leukemia LL8, shown in Figure 4B, homogeneously expressed CD19, whereas Nectin-2 was expressed in approximately one third of leukemic cells. In this case, CD34⁺ cells were contained within Nectin-2⁺ leukemic cells. Also in this case, CD48 and NTBA were down-regulated in leukemic cells, whereas the small fraction of normal lymphocytes was characterized by a bright CD48 and NTBA staining (although not shown, indeed CD48^bright cells were CD19^-). The leukemia LL2, shown in Figure 4C, is an undifferentiated leukemia characterized by the expression of CD34 and HLA-DR surface antigens, while lacking CD7, CD2, CD3, and CD19 molecules. In this case, Nectin-2 was expressed by virtually all CD34⁺ leukemic blasts. Note again the low expression of CD48 and the absence of NTBA on leukemic cells with respect to normal lymphocytes. Analysis of a large panel of leukemias indicated that the down-regulation of CD48 expression was generally more frequent than that of NTBA.

In order to evaluate the susceptibility of lymphoblastic leukemias to NK-mediated lysis, we used allogeneic NK cells displaying a KIR/HLA-class I mismatch. On the other hand, in cases in which this mismatch did not occur, we performed the cytolytic assay in the presence of anti–HLA-class I mAb of IgM isotype in order to prevent engagement of HLA-specific NK receptors. Figure 5 shows cytolytic assay with 2 freshly isolated leukemias (LL10, in Figure 5A, and LL3, in Figure 5B) that were efficiently lysed by alloreactive NK cells; this lysis was not substantially increased by the addition of anti–HLA-class I mAb. LL10 leukemic blasts (representing 65% of total peripheral blood lymphoid cells), to be used as target cells for cytotoxicity, were previously purified by negative selection using anti-CD48 mAb. Indeed, we had verified that this mAb exclusively stained normal mononuclear cells (Figure 4A). We analyzed the relative role of various activating receptors in the induction of lysis (Figure 5). In Figure 5A, a partial inhibition of lysis induced by mAbs directed to NKp30, NKp46, and DNAM-1, but not to NKG2D, could be observed. Moreover, a virtual abrogation of lysis was detected upon simultaneous masking of the 2 NCRs and DNAM-1, thus indicating that the cooperation between these receptors plays a crucial role in inducing lysis of LL10 leukemia. These data are consistent with the expression of Nectin-2 in LL10 leukemia, but also suggest that this leukemia expresses still-undefined ligands recognized by NCR.

LL3 leukemia expressed low levels of ULBP1 and PVR and no Nectin-2 (Table 3). In this case, maximal inhibition could be observed by the combined use of mAbs specific for NCR and NKG2D, although some involvement of DNAM-1 could be appreciated by the inhibitory effect exerted by mAbs simultaneously masking NCR and DNAM-1 (Figure 5B). Similar results could be obtained in leukemias expressing one or another DNAM-1 ligands. In most instances a cooperation between NCR and DNAM-1 could also be observed. Leukemias that were negative for NKG2D or DNAM-1 ligands were less susceptible to NK-mediated lysis as compared with the 2 cases reported in Figure 5. Moreover, in these cases the lysis appeared to be exclusively NCR dependent (not shown).

In this study we analyzed the susceptibility to NK-mediated lysis of a large panel of both myeloid and lymphoblastic leukemias. We show that, in most instances, cytolytic activity correlates with the surface expression of known ligands for activating receptors, such as the DNAM-1 and NKG2D ligands. Remarkably, almost all AMLs, independently on the FAB subtype, expressed both PVR and Nectin-2 (ie, DNAM-1 ligands); on the other hand, of the NKG2D ligands, MICA and MICB were consistently absent, while only few AMLs expressed low levels of one or another ULBP. Accordingly, lysis of most AMLs could be inhibited by mAb-mediated masking of DNAM-1 receptor. In these experiments a partial inhibition could be achieved also upon mAb-mediated blocking of NCR (mainly NKp30 and NKp46), whereas virtual abrogation of lysis required simultaneous blocking of NCR and DNAM-1, thus indicating that most AMLs also express the still-undefined NCR ligands. Our data are in agreement with a previous report showing that NKG2D plays only a marginal role in killing of AML.³³  Another report, however, described a dim expression of MICA in about half of the leukemias analyzed.³⁶ 

Remarkably, the surface expression of PVR and Nectin-2 in AML might even be necessary to promote an efficient lysis by NK cells as suggested by the observation that leukemia ML7, which lacks PVR and Nectin-2, was not lysed by alloreactive NK cells. Moreover, since DNAM-1/ligand interaction always played a role in leukemic cell killing, it is evident that PVR and Nectin-2 represent functionally relevant markers of AML.

Information regarding the susceptibility of AML to NK-mediated lysis is particularly relevant in view of the recent data showing that NK cells may play a crucial role in preventing leukemic relapse in patients undergoing haploidentical bone marrow transplantation.¹⁵  Indeed, in the case of KIR/HLA-class I mismatch in the donor versus recipient direction, the donor's NK cells arising from the grafted hematopoietic precursors were found to be responsible for a particularly favorable clinical outcome.¹⁶  In view of these data, it is important to define whether a given leukemia is susceptible to lysis by normal allogeneic NK cells. In particular, it is worth identifying the appropriate donor/patient combinations compatible with the presence of alloreactive NK cells. Although the presence of the donor's alloreactive NK cells might be predictable on the basis of the analysis of HLA-class I haplotypes in both donor and recipient, a more precise evaluation of the actual frequency of such alloreactive NK cells requires the study of the clonal repertoire¹⁵  or a combination of phenotypic and functional analyses at the population level.³⁵  As confirmed in this study, the presence of the donor's alloreactive NK cells correlates with the percentage of NK cells expressing the relevant KIR but lacking other inhibitory receptors engaged by HLA-class I alleles on the patient's target cells (eg, either KIR or CD94/NKG2A). Adequate information on the degree of “alloreactivity” of a given NK cell population can be obtained by assessing the proportion of NK cells that express only the KIR of interest as well as the magnitude of the NK-mediated cytolysis against the allogeneic target examined. Indeed, we show that potential NK donors displaying similar HLA-class I mismatch, as defined by KIR specificity, displayed different proportions of phenotypically defined alloreactive NK cells. Higher percentages of alloreactive NK cells resulted in higher levels of cytolytic activity against the patient's leukemic cells. Although the phenotypic analysis of expressed KIR is frequently informative on the degree of NK alloreactivity, it should be stressed that available anti-KIR mAbs do not discriminate between inhibitory or activating KIRs. This implies that in some donors the proportion of alloreactive NK cells cannot be precisely defined. Only recently, a novel reagent has been described that enables distinguishing KIR2DL3 from KIR2DL2 and KIR2DS2.³⁷  Indeed, the coexpression of an activating receptor with the KIR of interest may lead to an underestimation of the size of the functionally alloreactive NK subset. The presence of these activating KIRs can be further assessed by analysis of KIR genotype and by redirected killing assay using the FcγR⁺ P815 target cells in the presence of anti-KIR mAb. It is of note that allogeneic NK cells characterized by a KIR/KIR ligand incompatibility were also more cytotoxic against solid tumors, including melanoma and renal cell carcinoma.³⁸ 

Lymphoblastic leukemias were in general less susceptible to NK-mediated lysis. Again, a correlation existed between susceptibility to lysis and expression of PVR and Nectin-2. Moreover, also in lymphoid leukemias DNAM-1 appeared to act in a synergistic fashion with NCR. In contrast, NKG2D did not play a role in NK-mediated killing of most leukemias. An exception is represented by the leukemia LL3, which expressed significant levels of ULBP1. Our present data would suggest that the finding of a frequent resistance of lymphoblastic leukemias to NK-mediated lysis may merely reflect the lack of surface expression of ligands for major activating NK receptors. This should apply not only to the known ligands, but also to the cellular ligands of NCR, as suggested by our present study. The lack of expression of ligands for activating receptors, together with the higher levels of HLA-class I molecules as compared with AML, may provide an explanation as to why NK cells in a haploidentical bone marrow transplantation setting are not efficient in inducing a graft-versus-leukemia effect.

In conclusion, the DNAM-1 ligands, PVR and Nectin-2, play an important role as targets for NK-mediated killing and may also represent suitable markers for identifying leukemias that can be treated with NK-based immunotherapy.

We thank Dr S. Parolini for the generous gift of the valuable KL247, KS38, and AZ158 mAb, and the SC Laboratorio di Istocompatibilità/IBMDR, EO Ospedali Galliera Genova for HLA typing.