Introduction
Multiple myeloma is caused by expansion of malignant plasma cells in the bone marrow (BM) and remains largely incurable with the current treatment options [
1]. Allogeneic hematopoietic stem cell transplantation (allo-HSCT) performed over HLA-C barriers can provide curative anti-myeloma responses [
2]. However, poor clinical outcomes and T cell-mediated graft-versus-host disease (GvHD) are major limitations of allo-HSCT for myeloma [
3‐
6]. Infusion of alloreactive natural killer cells (NK) cells could be an appealing strategy to improve clinical effectiveness since the risk of GvHD is low and there is no need for myeloablative conditioning [
7,
8].
NK cells are large granular lymphocytes with the potential to selectively kill malignant cells. The balance between tolerance to healthy cells and eradication of virally infected or tumor cells is established by a panel of activating and inhibitory receptors on the NK cell surface [
9]. Activating receptors, interacting with “stress-induced flags” on diseased cells, trigger NK cell killing. Inhibitory receptors bind to self-molecules such as HLA-class I and HLA-E that are expressed by virtually every normal cell and protect against unwanted NK cell reactivity. An important class of inhibitory receptors, the killer immunoglobulin-like receptors (KIRs), interacts with HLA-class I molecules on the target cell; KIR2DL1 ligates HLA-C group 2 (Lsy80) alleles, KIR2DL2/3 ligates HLA-C group 1 (Asn80) alleles and KIR3DL1 interacts with HLA-Bw4 motifs [
10,
11]. The lectin-like family member NKG2A interacts with HLA-E [
12].
Upon HLA-mismatched allo-HSCT, donor NK cells eliminate tumor cells when the patient lacks HLA epitopes from the donor, and donor NK cells express KIRs specific for these epitopes (KIR-ligand mismatch) [
13]. Landmark studies in the haplo-identical HSCT setting showed that KIR–ligand-mismatched NK cells exert anti-leukemic effects and improve survival of acute myeloid leukemia (AML) patients [
14,
15]. Also in myeloma, incompatibility between donor KIRs and patient HLA has been associated with a better clinical outcome [
2]. Moreover, a first clinical trial showed that isolated KIR–ligand-mismatched NK cells can be safely administrated to myeloma patients though the contribution of NK cells, in addition to the conditioning regime, could not be established in this trial [
16]. Beneficial effects of KIR–ligand-mismatched NK cell infusion have been observed in other types of cancer as well [
17‐
20]. Although these studies are encouraging, optimization of the clinical efficacy of transferred NK cells is still required.
While there are several reasons why NK cells therapy could be a major step forward in immunotherapy of cancer, various mechanisms have been described that could reduce clinical effectiveness, e.g., decreased expression of activating receptors [
21‐
23], hypoxia [
24], the expression of immunosuppressive soluble or surface molecules [
25‐
27] or infusion of insufficient numbers of (the right) NK cells and HLA-class I expression by myeloma cells [
28‐
30]. Genotypic selection of a KIR–ligand-mismatched donor seems to be a good strategy to reduce inhibitory effects by HLA-class I. However, because NK cells can express one or a combination of inhibitory receptors, not all donor NK cells will exclusively express KIRs that are mismatched with HLA-class I of the patient. Furthermore, of all peripheral blood NK cells, 20–80 % expresses NKG2A [
31,
32]. For clinical-grade expanded NK cells, from either hematopoietic progenitor cells [
33], blood-derived NK cells [
34], or pluripotent stem cells [
35], this percentage can be higher than 80 %. Because HLA-E is low polymorphic and all functional HLA-E alleles can inhibit NK cells through NKG2A [
36], NKG2A could abolish the overall NK cell response against HLA-E-expressing tumors, even in the KIR–ligand-mismatched setting. This makes HLA-E clinically highly relevant, but information on HLA-E expression and its inhibitory potential in myeloma is lacking.
In the present study, aimed to refine the selection process for NK cell therapy, we studied the relevance of HLA-class I and HLA-E for anti-myeloma reactivity of allogeneic NK cells. First, we determined the presence of HLA-class I and HLA-E on primary myeloma cells and on myeloma cell lines, revealing that primary myeloma cells express both HLA-class I and HLA-E. Also, we found major differences with myeloma cells cultured in vitro and the same cells grown in vivo. Second, we assessed the inhibitory potential of HLA-class I and HLA-E on different NK cell subsets in a co-culture system that allowed synchronized analysis of degranulation of individual NK cells subsets, demonstrating the potent inhibitory effects of HLA-class I and HLA-E on NK cell alloreactivity. To better predict the in vivo response, we performed these assays at ambient and clinically relevant low oxygen levels.
Materials and methods
Cell culture incubators
Experiments were performed in ambient 95 % humidified air, 21 % O2 incubator (SANYO) containing 5 % CO2. Hypoxia experiments were completely performed inside 0.6 % O2 incubators (Invivo2, 1000 Ruskinn Technology Ltd) at 37 °C with 5 % CO2.
Cell lines
Myeloma cell lines [
37] and HLA-class I-deficient K562 cell line (ATCC) were cultured in RPMI-1640 medium (Gibco) supplemented with 10 % fetal calf serum (Integro), 1000 U/mL penicillin (Gibco) and 100 µg/mL streptomycin (Gibco). The LME-1 cell line (received through Dr. R. van Oers, AMC, Amsterdam) was cultured in IMDM medium (Gibco), and XG-1 in IMDM medium with (500 pg/ml) IL-6 (Biosource).
Flow cytometric analysis of HLA-E and HLA-class I expression
Material from myeloma and plasma cell leukemia patients was obtained as a waste product from a diagnostic procedure that had been approved by the hospital’s medical ethical committee. Bone marrow cells were stained for CD38 (HB7, BD), CD138 (CLB-1D4, Sanquin), cytoplasmic kappa light chain (G20-193, BD), cytoplasmic lambda light chain (1–555–2, BD), HLA-class I (G46-2.6, BD) or HLA-E (3D12, eBioscience). HLA expression was determined on CD38high cells. All flow cytometry-based assays were performed on a BD FACS Canto II. Data were analyzed with BD FACSDiva v6.1.2 or FlowJo 7.6.
HLA genotyping and NK cell donor selection
Genotypic expression of HLA epitopes (HLA-C1, HLA-C2 and HLA-Bw4) in cell lines (supplemental table S1) and in healthy volunteers was determined by sequence-specific oligonucleotides (SSO) analysis and Luminex® according to the manufacturer’s guidelines (One Lambda). Healthy volunteers, with genotype HLA-C1+C2+Bw4+ and phenotypically expressing KIR2DL1, KIR2DL2/3 and KIR3DL1 were selected as NK cell donors. Donors signed informed consent forms.
NK cell isolation and CD107a degranulation assay
NK cells were isolated from blood by negative MACS selection (Miltenyi Biotec) according to manufacture's instructions. NK cells were activated with 1000 U/ml IL-2 (Proleukin) for 6 h at 21 % O2. Activated NK cells were co-cultured with target cells at 1:1 ratio in 96-well round-bottom plates in duplicate with anti-CD107a-Horizon-V450 (H4A3, BD). After 1 h, co-culture at 21 % O2 or 0.6 % O2, monensin (BD GolgiStop, Cat# 554724) was added. After another 8–9 h, co-cultures were stained on ice with anti-CD3-APC/H7 (SK7, BD), anti-CD56-PeCy7 (B159, BD), anti-KIR2DL1-APC (143211, R&D), anti-KIR2DL2/3/S2-PE (DX27, Miltenyi Biotec), anti-KIR3DL1-FITC (DX9, Miltenyi Biotec) and anti-NKG2A-PC5.5 (Z199, Beckman Coulter). For co-cultures with IL-2-activated NK cells, target cells were pre-incubated for 6 h at 21 % O2 or 0.6 % O2. For experiments without IL-2 activation, target cells cultured at 21 % O2 were used. Target cells were subsequently co-cultured with freshly isolated NK cells in a 10- to 12-h degranulation assay. For HLA-E blocking experiments, target cells were pre-incubated for 30 min at 37 °C with 10 µg/ml of anti-HLA-E (3D12; IgG1 isotype eBioscience) or IgG1 isotype control. Gating strategy is described in supplemental figure S2.
HLA-E induction by peptides
U266 cells were incubated with 500 µM of HLA-A1 (VMAPRTLLL), HLA-B7 (VMAPRTVLL) or a control peptide (RGPGRAFVTI) for 12 h at 37 °C. As additional negative controls, U266 cells were incubated without peptide or with DMSO, the solvent of the peptides. After 12 h, HLA-E expression was determined by flow cytometry. For degranulation assays, peptide incubation was done for 2 h after which the cells were co-cultured with IL-2-activated NK cells in a 12-h degranulation assay.
Mice and multiple myeloma disseminated growth model in RAG-2−/−γc−/− mice
RAG-2
−/−
γc
−/− mice were bred in the SPF unit Central Animal Facility of the Utrecht University using filter top cages. After a total body irradiation (3.0 Gy 200 kV X-rays), the mice received an injection with 5 × 10
6 cells intravenously from the green fluorescence protein (GFP)–luciferase-marked, human MM cell line U266. Tumor growth of the U266-luc cell line was monitored by bioluminescence imaging as described previously [
37]. Upon killing the mice, BM cells were stained for human-CD45, HLA-class I and HLA-E.
Statistical analyses
Statistical significances of differences were determined by paired t test or two-way repeated measures ANOVA with Bonferroni correction. A p value of <0.05 was considered significant. Analysis was performed with GraphPad Prism V (Graphpad Software Inc).
Discussion
Despite the introduction of novel drugs to treat multiple myeloma, this disease has remained largely incurable. NK cells can be novel immunotherapeutic candidates, and their safety has been documented in clinical trials [
17]. Exploitation of the full potential of NK cells requires selection of NK cell subsets with the highest possibility of triggering anti-myeloma and identification of the factors important for sustaining NK cell activity in vivo. In our quest of further refining NK cell therapy, we evaluated the clinical relevance of HLA-class I and HLA-E for the anti-myeloma response of alloreactive NK cells.
Expression of HLA-E in primary myeloma has not been reported, and the present study shows that HLA-E is indeed expressed by CD38
high cells of myeloma or PCL patients. Our data also demonstrate that primary myeloma cells remain positive for HLA-class I, which has been described before [
28‐
30]. HLA-class I was also strongly expressed by all cell lines in our panel. HLA-E expression, on the other hand, was lower on cell lines than on primary myeloma cells and ranged from completely absent, e.g., on OPM-1, to very low on most of the other myeloma lines. Low levels of HLA-E on solid tumor cell lines were also reported in three previous studies with 30–37 different in vitro cultured cell lines per study mainly of solid tumor origin [
42‐
44]. The most striking observation in our study was that the HLA-E level on U266 cells grown in the BM of RAG-2
−/−
γc
−/−mice was considerably higher than the level on in vitro cultured U266 and was relatively comparable to the expression levels in myeloma patients. Thus, there seems to be a difference between in vitro and in vivo expression levels of HLA, which might be the consequence of supportive factors present in vivo that are lacking in vitro. This observation is highly relevant for the interpretation of in vitro NK cell studies using HLA-E
low tumor cells in general [
45] because the in vitro situation might be very limiting in predicting the actual situation in cancer patients.
In a co-culture system allowing the synchronized analysis of degranulation of NK cell subsets, we identified HLA-E as a potent inhibitory factor for anti-myeloma reactivity of NK cells expressing the inhibitory NKG2A receptor (Fig.
5). The impact of HLA-E was most pronounced in the experiment where blocking of HLA-E enhanced degranulation of KIR-NKG2A+ NK cells and in the experiment where HLA-E binding peptides reduced KIR-NKG2A+ NK cell degranulation via the upregulation of HLA-E. This is consistent with the inhibitory role of HLA-E described in AML and ALL [
32]. A limitation of this study is lack of use of anti-HLA-E fragment associated with antigen binding (Fab). Therefore, we cannot fully exclude the possibility that the enhanced degranulation upon HLA-E antibody incubation was caused by triggering of ADCC. However, in our study, HLA-E blocking did not enhance degranulation of KIR+NKG2A− subsets which also express CD16 and can mediate ADCC, suggesting that the enhanced degranulation of KIR-NKG2A+ NK subsets did not trigger ADCC at global NK cell level. Lack of ADCC triggering by the antibody might be explained by the fact that the HLA-E antibody was mouse IgG1 isotype, which has been shown to poorly bind to human Fc receptors on monocytes [
46]. Another possible limitation of our study is that we did not confirm our results with MM cells from patient samples due to the very poor survival capacity of primary MM cells in culture. Nevertheless, because we showed that primary MM cells express relatively high levels of HLA-E, it will be important to take the inhibitory effects of HLA-E into account in the design of future (clinical) studies. Because a large fraction of circulating NK cells and the majority of NK cells in clinical products are NKG2A+, HLA-E can be expected to have major impact on overall NK cell alloreactivity against HLA-E-expressing tumors. HLA-E-mediated tumor resistance might at least partly explain the negative clinical trail results obtained so far. Our current data provide a strong argument to use NKG2A-negative NK cells in future clinical trials or develop strategies to abrogate inhibition by HLA-E.
Our findings consequently demonstrated a higher killing potential of KIR–ligand-mismatched NK cells as compared to their matched counterparts. We used NK cells from HLA-C1-, HLA-C2-, HLA-Bw4-positive donors to rule out hyporesponsiveness of non-licensed NK cells, and our data with K562 demonstrated that all subsets had more or less comparable intrinsic capability to kill MM cells. Therefore, it seems unlikely that differences between the subsets were caused by, for example, variation in the level of activating receptors. In addition, we have shown before that activating receptors (e.g., NKG2D and NCRs) are homogenously distributed on CD56+ cells [
24]. Previously, bortezomib-induced downregulation of HLA-class I [
47], KIR-blockade with IPH2101 [
48] and HLA-blockade with W6/32 [
16] have been shown to enhance killing of myeloma cells by NK cells. Our data are in agreement with these studies and confirm the relevance of HLA-class I in an alternative experimental setup. An advantage of our system is that it allows the simultaneous analysis of individual subsets, which additionally revealed that the anti-myeloma advantage of mismatched NK cells was reduced by co-expression of one or more matched KIR. Hence, selection of NK cell subsets with exclusively ligand-mismatched KIRs will be important to enhance clinical efficacy.
We recently showed that IL-2 activation is required for NK cell anti-myeloma responses in clinically relevant hypoxic conditions [
24]. In addition, we showed in that study that expression of HLA-class I, HLA-E and the NKG2D ligands MICA/B and ULPB1-2 by MM cells was not influenced by hypoxia [
24]. Here, we demonstrate that, even though IL-2 activation enhanced degranulation of all NK cell subsets, KIR–ligand-mismatched NK cells still mediated a more potent anti-myeloma response than matched NK cells. Importantly, we did not observe a difference between anti-myeloma responses at 21 and 0.6 % O
2, indicating that for all subsets, activation was sufficient to overcome the inhibitory effect of hypoxia. Recently, pre-activation of NK cells with a combination of IL-12/IL-15/IL-18 proved to be more efficient in enhancing NK cell effector function as compared to activation with IL-2 or IL-15 alone [
49,
50]. Thus, cytokines other than IL-2 might enhance the therapeutic potential of alloreactive NK cells even further.
We acknowledge that in our system, HLA-class I and HLA-E were not the only factors determining the magnitude of the anti-myeloma response, i.e., HLA-E blocking did not completely restore degranulation, and in response to some cell lines (e.g., UM-9), even fully KIR–ligand-mismatched NK cells did not vigorously degranulate (Fig.
4a). Previously, the activating receptors NKG2D, NCRs and DNAM-1 have shown to be important for NK cell anti-myeloma activity [
28,
29,
51]. Although our study was not designed to investigate the role of activating ligands, analysis of the NKG2D ligands MICA/B and ULPB2 showed that most myeloma cell lines in our panel expressed at least one of these ligands (supplemental table S2). UM-9, however, expressed very low levels of all three ligands, which might explain the low capacity to activate NK cells. To develop effective immunotherapeutic strategies for resistant myeloma, synergy between alloreactive NK cells and drugs such as lenalidomide and 17-AAG should be evaluated because these drugs have been shown to sensitize myeloma by enhancing the expression of activating ligands [
48,
52].
In summary, our results show that KIR–HLA-class I and NKG2A–HLA-E interactions are highly relevant for NK cell reactivity against myeloma. For accurate prediction of in vitro data to patient’s reality, two relevant in vivo realities have to be taken into account: (1) In the bone marrow myeloma, cells reside under hypoxic conditions and (2) in vivo, myeloma cells express both HLA-class I and HLA-E. Infusion of a high number of cytokine-activated alloreactive NKG2A-negative, KIR–ligand-mismatched NK cells or the use of KIR-blocking antibody (IPH2101) may help to improve the efficacy of alloreactive NK cell therapy.