Molecular Interactions Between NK Cells and Acute Leukemic Cells: KIR2DL5 Drastically Limits NK Cell Responses

Abstract

Natural Killer (NK) cells naturally recognize and eliminate leukemic cells. However, the molecular interactions that govern these responses are diverse due to the large number of activating and inhibitory NK receptors that modulate NK functions and the diversity of corresponding ligands that are differentially expressed in acute lymphoblastic and myeloblastic leukemias. We identified resting NKG2A⁺ NK cells and NKG2A⁺KIR⁺ NK cell subsets as the most effective in eliminating lymphoid and myeloid leukemic cells respectively. The NKG2A⁺KIR^±CD57⁻ cell subsets show high expression of activating receptors and a functional transcriptomic profile, but differ in KIR2DL5 expression. The frequency of KIR2DL5⁺ NK cells increases with the number of expressed KIR. Furthermore, KIR2DL5 is preferentially co-expressed with KIR2DL1 and is negatively regulated by NKG2A. Of note, CD57 expression, regardless of the NK cell subset considered, is associated with reduced receptor expression, consistent with its reduced cytotoxic potential. Furthermore, molecular interactions between NK cells and leukemic cells influence NK cell responses, particularly the inhibitory KIR2DL5-PVR axis. The integration of these data is of importance for the optimization of NK cell-based immunotherapies, as the selection of NK cell donors represents a key parameter for the improvement of these therapies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10875-025-01913-y.

Introduction

Natural killer (NK) cells play a critical role in early defense against tumor cells due to their ability to recognize and eliminate abnormal cells without prior sensitization [1, 2]. Acute leukemia is a rapidly progressing cancer of the blood and bone marrow that results in the overproduction of immature blood cells, known as blasts. There are two main types of acute leukemia: acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML). NK cells express inhibitory receptors such as killer cell immunoglobulin-like receptor (KIR) and CD94/NKG2A that bind to HLA class I molecules (HLA-A, -B and -C) and the non-classical HLA-E molecule, respectively. HLA molecules are expressed on autologous cells, allowing NK cells to recognize them as self. However, the expression of self HLA molecules can be drastically reduced on tumor cells and may be absent in allogenic contexts. In these configurations, NK cells cannot recognize cells as self and their cytotoxicity is drastically activated by the absence of engagement of their inhibitory receptors with their cognate ligands.

Although NK cells belong to the innate immunity, during their maturation they undergo an education or licensing process mediated by these inhibitory receptors and autologous HLA class I molecules. This education is acquired by the engagement of KIR2DL1, KIR2DL2/3 and KIR3DL1 with their respective ligands (HLA-C C2 group, HLA-C C1 group, HLA-A and -B with Bw4 motifs) [3, 4]. NK cells also express activating receptors, such as DNAM-1, NKG2D, 2B4, and NCR (NKp30, NKp44 and NKp46) that bind to upregulated ligands on leukemic cells [5]. Overall, the response of NK cells is defined by the sum of signals received by the expressed receptors. This complexity in receptor expression leads to significant heterogeneity among NK cells. Horowitz and colleagues described more than 100,000 NK phenotypes in a single donor using 28 NK cell surface markers [6]. Furthermore, KIRs are stochastically expressed and the KIR-HLA pair is the most polymorphic in the human genome and influences the intra- and inter-donor diversity of the NK cell repertoire [7, 8]. Recently, we have shown that KIR and HLA genetics, as well as age, sex and cytomegalovirus (CMV) status, influence the diversity of the NK cell repertoire [9].

Phenotypic and functional diversity are intertwined, as evidenced by the identification of good and bad responders to leukemic cells [10]. Furthermore, NK cell functionality depends on the heterogeneous expression of ligands by AML and ALL cells, leading to the formation of multiple molecular interactions engaging NK cell receptors. Even though the role of NK cells against leukemia is well established, the identification of NK cell subsets most effective in targeting leukemic cells based on phenotypic markers remains an unresolved issue. Björkström et al. proposed a model of NK cell differentiation [11] that allows the discrimination of immature NK cells that express CD94/NKG2A, then KIR [12] and, finally acquire the maturation marker CD57 [13]. Additionally, CMV has been observed to influence NK cell maturation with expansion of the memory-like cell subset expressing CD94/NKG2C [14, 15]. Recent advances in single cell analysis have allowed a deeper understanding of the differentiation markers of NK cells, including numerous markers (cytokine production, cytotoxic molecules, transcription factors) [16, 17].

In the present work, NK cell subsets with the highest efficacy against lymphoid and myeloid leukemic cells were identified in considering the NK cell phenotype in the context of KIR and HLA genotypes, age, sex, and CMV status. Moreover, the expression of leukemic ligands expressed at the immunological synapse between NK cells and ALL or AML cells were determined to account for the heterogeneity of ligands expressed by leukemic cells. The ultimate goal of this research is to identify the most effective NK cell subsets that can be targeted in NK cell-based immunotherapy and to optimize clinical application in the context of acute leukemia and Hematopoietic Stem Cell Transplantation (HSCT).

Methods

Peripheral Blood Mononuclear Cells (PBMCs), Cell Lines and Primary Leukemic Cells

PBMCs were isolated as previously described [18]. All blood donors (n = 200) were recruited at the Etablissement Français du Sang (EFS, Nantes, France), and informed consent was given by all individuals. Lymphoid cell lines (n = 6): Raji, Daudi, H9, Molt4, Jurkat and HP Ball and myeloid cell lines (n = 4): KG1, NB4, HL60 and OCI-AML3, were used to investigate NK cell degranulation. Cells were cultured in RPMI 1640 medium (Gibco, Paisley, Scotland, UK) containing glutamine (Gibco) and penicillin–streptomycin (Gibco) and supplemented with 10% fetal bovine serum (FBS) (Gibco). Mycoplasma tests performed by PCR were negative for all cell lines. Blood and bone marrow samples from 33 patients were collected and leukemic blasts were isolated as previously described [18]. All patients gave their informed consent to physicians from the clinical hematology department of the Nantes University Hospital (Pr. Patrice Chevallier), in accordance with the Declaration of Helsinki. Patients’ characteristics are summarized in Supplemental Table 1. A declaration of preparation and conservation of these biocollections (AC-2021–4397) has been provided to French Research Minister and has received approval from the IRB (2015-DC-1).

HLA and KIR Genotyping

All blood donors and leukemic cells (lines and blasts) were HLA-A, -B, -C typed as previously described [19]. By using primers provided by Dr Ketevan Gendzekhadze (Duarte, CA, USA), KIR generic typing was performed on all individuals by KIR multiplex PCR-SSP method [20].

Cell Phenotyping

PBMC, cell lines and primary leukemic cells were phenotyped by eight-color multiparameter flow cytometry (MFC) by using mouse anti-human mAbs. Healthy cells were defined as CD45^high, enabling the discrimination from CD45^low leukemic cells [21]. All PBMCs were stained for 30 min at room temperature, and NK cells were phenotyped as CD3⁻ CD56⁺ or NKp46⁺. The deep analysis of NK cell and primary leukemic cells phenotypes were done by using the mAbs listed in Supplemental Table 2. Antibodies were used as suggested by the supplier. Spillover and coexpression were taken into consideration. Experimental controls included unstained and single stained controls.

CRISPR Cas9 Procedure

TrueCut™ HiFi Cas9 Protein (A50579, 5 µg/µL) and TrueGuide™ gRNA (1.5 nmol) were purchased from Invitrogen™. TrueGuide™ gRNA were designed using the Integrated DNA Technologies (IDT) pre-design CRISPR-Cas9 guide RNA tool (https://eu.idtdna.com/pages). We ensured that a PAM of NGG is located immediately downstream of the target sequence for Cas9 recognition. The PVR-targeting TrueGuide™ gRNA sequence was as follows: CCTGCGGGAACGTGACGAAC. TrueGuide™ gRNA were reconstituted in 1X TE buffer, aliquoted and stored at −20 °C. TrueCut™ HiFi Cas9 Protein and TrueGuide™ gRNA were mixed together at a 1:1 molar ratio in 1X PBS. The mixed complex does not exceed 1/10th of the total reaction volume. The mixed complex was incubated 15 min at room temperature. Cultured KG1 cells were diluted one day before their transfection established according manufacture’s protocol. Cells were resuspended, mixed with the TrueCut Hifi Cas9 protein + TrueGuide™ gRNA mix and transferred to a 0.2 cm-gap electroporation cuvette. Electroporation was performed on a Gene Pulser Xcell System (Biorad, France) (Mode: exponential; Voltage: 150 V; Capacitance: 700 F; Resistance: 50Ω.). After electroporation, cells were placed in one well of 24-well plate with culture medium. Transfection and stable knockout of PVR (PVR^KO KG1) was confirmed by MFC. Prior to cell sorting, the KG1 cell line containing PVR^KO KG1 cells was amplified and stained with labelled mAb specific for PVR. The PVR^KO KG1 cells sorted were then amplified in culture medium.

Differential Gene Expression Analysis

PBMCs were thawed on the day prior to cell sorting and maintained in culture overnight in RPMI 1640 medium containing glutamine and penicillin–streptomycin and supplemented with 10% FBS (Gibco). NK cells were stained with labelled mAbs specific for NKp46, KIR2DL1/S1, KIR2DL3, KIR3DL1, NKG2A, NKG2C, and CD57. All cells that expressed KIR3DL1⁺, NKG2C⁺, and/or CD57⁺ were excluded from the analysis. The stained cells were sorted into four alive NK cell subsets: KIR2DL1/S1⁻ KIR2DL3⁻ NKG2A⁻, KIR2DL1/S1⁻ KIR2DL3⁻ NKG2A⁺, KIR2DL1/S1⁻ KIR2DL3⁺ NKG2A⁺, and KIR2DL1/S1⁺ KIR2DL3⁻ NKG2A⁺ using a FACSAria™ III cell sorter (BD Biosciences). Total RNA was extracted directly, quantified and 3’ digital gene expression profiling was carried out as previously described [22]. Raw sequencing data were deposited in GEO under accession number (GEO GSE281383). The sequencing output were analysed using the 3’ Sequencing RNA Profiling (SRP) pipeline [23]. This methodology aligns with previous DGEseq analysis in various studies [22, 24]. Conditions were compared using the DESeq2 R package. The complete material and method for the bioinformatics part is available in Supporting Information Text.

CD107a Degranulation Assay

PBMCs from healthy blood donors were mixed with target cell lines in the presence of anti-CD107a-BV421 (H4A3, Sony) or anti-CD107a-BV510 (H4A3, Sony) mAbs, and NK-cell (CD3⁻ CD56⁺) degranulation was assessed after incubation for 5 h with different target cells (E:T ratio = 1:1) or medium (negative control) with brefeldin A (Sigma) at 10 μg/mL for the last 4 h. Then, NK cells were phenotyped as previously indicated. For blockage experiments of KIR2DL5-PVR interaction, PVR⁺ leukemic cells were pre-incubated with the PVR-specific mAb (SKII.4, Biolegend) at a concentration of 1 µg/ml for 20 min at room temperature. The cells were washed twice before incubation with healthy PBMCs for the degranulation assay.

Flow Cytometry Acquisition and Data Analysis

Every flow cytometry acquisition was done on a unique FACSCanto II cytometer in our ISO 9001 certified research laboratory. Systematically before acquisition, BD FACSDiva CS&T Research Beads (BD Biosciences) were used. The flow cytometry files were analysed using Flowjo™ 10.6 software (LLC, Ashland, OR).

Statistical Analyses

Comparisons between two groups were performed using Student’s t-test or Mann–Whitney test (n < 30). Comparisons between multiple groups were performed using one-way ANOVA or Kruskal–Wallis test (n < 30). Clustering analysis was performed by utilizing the k-mean method. Statistical analyses were performed using the GraphPad Prism version 9.3.1 software (San Diego, CA) or R software, version 4.2.2 (The R Foundation for StatisticalComputing). Statistical significance was set at p < 0.05.

Results

Markers Determining the Efficacy of NK Cell Subsets Against Leukemic Cell Lines

We evaluated the degranulation of ex vivo NK cells from 200 healthy blood donors with 10 myeloid and lymphoid cell lines. Lymphoid cells are better recognized by NK cells than myeloid cells (Fig. 1A, Supplemental Table 3). For the same target cell line, a large diversity of NK cell responses is observed. The best responders against one cell line are not systematically the best responders against other cell lines suggesting a broad functional diversity of the NK cell repertoire and different molecular interactions between NK cells and leukemic cells. We further investigated the degranulation of 24 NK cell subsets defined on the basis of NKG2A, KIR2DL1, KIR2DL3, KIR3DL1, CD57 and NKG2C markers (Figure sup 1) against myeloid versus lymphoid cell lines taking into account the nature of HLA-C ligands (C1C1, C1C2 and C2C2) of the blood donors and subset representativeness (Fig. 1B). Interestingly, NK cell subsets expressing NKG2A without any KIR, CD57 or NKG2C constitute the pool of NK cells that is the most activated by lymphoid cell lines (Figure sup 2 A). In contrast, NK cell subsets expressing NKG2A and KIR (mainly KIR2DL) constitute the most efficient NK cell subsets against myeloid cell lines (Fig. 1B, Figure sup 2B). In addition, the expression of CD57 on all NK cell subsets drastically limits their potential of degranulation compared to their CD57⁻ counterparts against lymphoid and myeloid cell lines (Fig. 1B, Figure sup 2A-B). NK cell education is illustrated by the better degranulation of KIR2DL1⁺ and KIR2DL2/3⁺ NK cell subsets from C2⁺ and C1⁺ blood donors, respectively. The education process is particularly clear on KIR⁺ NK cell subsets that do not express NKG2A. The adaptive NK cell subsets marked by NKG2C, CD57 and KIR2DL harbor limited degranulation against lymphoid and myeloid cell lines (Fig. 1B). The nature of KIR genotype preferentially affects the degranulation of KIR2DL1⁺ and KIR3DL1⁺ NK cell subsets against lymphoid targets although only the degranulation of KIR3DL1⁺ NK cells against myeloid targets was affected by KIR genotypes (Fig. 1C). NK cell subsets from haplotype B + individuals were less effective than AA counterparts in accordance with the lower potential of education of KIR B allotypes as previously described [25]. NKG2C⁺KIR^± NK cells were the unique cell subsets to present a better degranulation in CMV⁺ compared to CMV⁻ blood donors although NKG2C⁺ KIR3DL1⁺ cells were unresponsive (Figure sup 3A-B). However, the expression of CD57 on these subsets drastically limits their function. Sex and age did not impact drastically the NK cell degranulation (Figure sup 3C-D). Altogether, our results show a broad diversity of degranulation of NK cell subsets that suggests that NK cell subsets are differently engaged with myeloid and lymphoid leukemic cells. We further deciphered the deep phenotype of NK cell subsets defined on the basis of NKG2A, KIR2DL1, KIR2DL2/3 and CD57 expression. We focused our study on the most efficient NK cell subsets against leukemic cells, termed NK^2A+ (NKG2A⁺KIR⁻CD57⁻) and termed NK^2A+KIR+ (NKG2A⁺KIR2DL⁺CD57⁻). In comparison, we used a reference cell subset that does not express the NKG2A, KIR2DL1, KIR2DL2/3 and CD57 markers that presents a low degranulation against all leukemic cells, termed NK^neg (NKG2A⁻KIR⁻CD57⁻). We excluded KIR3DL1⁺ NK cell subsets for following investigations as the degranulation of these subsets was lower against AML cell lines compared to KIR2DL1⁺ and KIR2DL2/3⁺ NK cell counterparts and KIR3DL1 gene exhibits a broad allelic polymorphism associated with no or low membrane KIR3DL1 expression in numerous blood donors [26]. The immature NK^neg cell subset exhibits low expression of activating receptors (NKp46, DNAM-1) and inhibitory CD96, but a high expression of inhibitory receptors (2B4 and ILT-2) (Fig. 1D). In contrast, the NK^2A+ cell subset exhibited the highest expression of the activating receptor NKG2D and NKp46, while NK^2A+KIR+ showed the highest expression of DNAM-1. Compared to NK^neg, both NK2A⁺ and NK2A⁺KIR⁺ cell subsets exhibited higher expression of NKp46 and DNAM-1 and somewhat higher expression of NKG2D and NKp30. The inhibitory receptor TIGIT and the activating receptor CD16 were homogeneously expressed on the surface of the three cell subsets. Additionally, KIR2DL5 and KIR3DL2 were significantly more frequently expressed on NK^2A+KIR+ than on NK^2A+ NK cells (Fig. 1E). Furthermore, KIR2DL5 expression is strongly associated with KIR2DL1 expression. The expression of CD57, regardless of the cell subset considered, is linked to a reduced expression of the activating receptors, consistent with their lower cytotoxic potential (Fig sup 4). Overall, three distinct NK cell subsets were identified on the basis of their expression of NKG2A, KIR2DL1, KIR2DL2/3 and CD57. The NK^2A+ and NK^2A+KIR+ subsets were found to demonstrate the greatest efficacy in their ability to target leukemic cells in contrast to the reference cell NK^neg subset, which showed low degranulation.

Fig. 1.

Functional and phenotypic characterization of NK cell subsets in response to lymphoid and myeloid leukemic cells. Schematic representation of the method used to characterize the function and phenotype of NK cell subsets (A). Scatter plots illustrating the impact of HLA-C environment (C1C1, C1C2, C2C2) (B) and KIR genotypes (AA, B +) (C) on the degranulation of 24 NK cell subsets, defined by NKG2A, KIR2DL1, KIRDL2/L3, KIR3DL1, CD57 and NKG2C expression, against lymphoid and myeloid cell lines, taking into account the representativeness of NK cell subsets. Bar graphs comparing the relative MFI of activating and inhibitory receptors on NK cell subsets defined on the basis of NKG2A, KIR2DL1 and KIR2DL2/3 (KIR) expression (n = 40) (D) and the frequency of KIR3DL2 (n = 40) and KIR2DL5 (n = 20) on NK cell subsets defined on the basis of NKG2A, KIR2DL1 and KIR2DL2/3 expression (E). Relative MFI represents the ratio of the MFI of each receptor on the MFI of isotype control. Statistical significance is indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Comparisons were made by Mann–Whitney and Kruskall-Wallis tests. Bar graphs represent mean ± SD

Functional Transcriptomic Signature of the NK^2A+ and NK^2A+KIR+ Cell Subsets

To complement the phenotypic study, we analyzed the transcriptional profiles of the three cell subsets NK^neg, NK^2A+ and NK^2A+KIR+. For this analysis, CD57 and KIR3DL1 were excluded from the gating strategy, in line with the characterisation of efficient NK cell subsets against leukemic cells. To design the cohort, we included HLA-C typing (n = 12 C1C1, n = 12 C2C2), KIR genotypes (n = 12 AA, n = 12 Bx) and CMV status (n = 17 CMV⁻, n = 7 CMV⁺) as parameters susceptible to shape the transcriptional profile of NK cell subsets (Fig. 2A). Using RNAseq on sorted NK cells from 24 donors, analysis with DESeq2 revealed the transcriptional signature of these cell subsets. The NK^neg cell subset expressed the fewest unique upregulated genes (29 DEG), while the NK^2A+KIR+ cell subset displayed the most unique upregulated genes (5361 DEG) (Fig. 2B). Comparisons between cell subsets showed distinct expression patterns: the NK^2A+ cell subset had 266 upregulated and 108 downregulated genes compared to the NK^neg cell subset (Fig. 2C), while the NK^2A+KIR+ cell subset had 404 upregulated and 147 downregulated genes compared to the NK^neg cell subset (Fig. 2D). The NK^2A+ cell subset also exhibited 1848 upregulated and 5431 downregulated genes compared to the NK^2A+KIR+ cell subset (Fig. 2E). Comparison between NK^2A+ and NK^neg cell subsets revealed that NK^2A+ cell subset showed increased expression of cytotoxic (GZMK, GNLY), maturation CD56 (NCAM-1), and activating receptor transcripts (NCR2), alongside inhibitory receptors (KLRC1, CD96), transcription factor (TCF7), and adhesion molecule transcripts (CD2, GPR183, SELL, CD44) (Fig. 2F). On the other hand, the comparison between NK^2A+KIR+ and NK^neg cell subsets revealed that the NK^2A+KIR+ cell subset had higher expression of the cytotoxic transcripts PRF1, CD56 and DNAM-1 (CD226), along with various inhibitory receptor transcripts (KLRC1, KIR2DL1, KIR2DL2, KIR2DL3) (Fig. 2G). In contrast, both NK^2A+ and NK^2A+KIR+ cell subsets showed downregulation of CD3 chains, inhibitory receptor LAG3, and cytokine/chemokine-related transcripts (LIF, CCL1). The NK^2A+ cell subset specifically downregulated the adhesion molecule SPON2. Moreover, the NK^2A+KIR+ cell subset downregulated NKG2D (KLRK1). The comparative analysis between NK^2A+ and NK^2A+KIR+ cell subsets further revealed that the NK^2A+ cell subset had higher expression of multiple transcripts coding for cytotoxic molecules, activating receptors, transcription factors, and adhesion molecules, but lower levels of DNAM-1, PRF1, and cytokines/chemokines (CLL1, IL2RA, IL10RB) (Fig. 2H). The lists of all DEGs for each comparison are presented in Datasets S1. Overall, the efficiency of the NK^2A+ and NK^2A+KIR+ cell subsets is reflected in a transcriptomic signature characterized by increased cytotoxicity, enhanced cytokine production and greater adhesion capacity compared to the NK^neg cell subset.

Fig. 2.

Transcriptomic signature of NK^neg, NK^2A+ and NK^2A+KIR+ cell subsets. Schematic representation of the method used for the transcriptomic study (A). Three-way Venn diagram showing the number of shared and uniquely expressed transcripts between NK^neg, NK^2A+ and NK^2A+KIR+ (B). Volcano plots showing the differential gene expression analysis of the comparisons of NK^2A+ vs NK^neg (C), NK^2A+KIR+ vs NK^neg (D), and NK^2A+ and NK^2A+KIR+ (E). Representation of upregulated (green) and downregulated (red) key transcripts for the comparisons of NK^2A+ vs NK^neg (F), NK^2A+KIR+ vs NK^neg (G), and NK^2A+ and NK^2A+KIR+ (H). The size of the circle corresponds to the value of log2FoldChange

ALL and AML Cells Exhibit Differential Expression Patterns of Ligands for Key NK Receptors

We further conducted an analysis of commonly altered expressed ligands in patients with acute leukemia (Fig. 3A). For this purpose, 33 patients were enrolled between 2012 and 2023, 14 ALL patients and 19 AML patients (Figs. 3B and Supplemental Table 1). In all patients, the expression of activating and inhibitory NK receptors was evaluated by MFC. ULBPs and MICA/B act as ligands for the activating receptor NKG2D. PVR and Nectin-2 are implicated in a multitude of interactions by binding to the inhibitory receptors TIGIT, CD96, KIR2DL5 and the activating receptor DNAM-1. AML blasts exhibited significantly higher expression of ULBP-2/5/6, MICA/B and PVR than ALL blasts (Fig. 3C). Analysis revealed a subtle increase in the expression of ULBP-3 and Nectin-2 in AML blasts. In contrast, ALL blasts were characterized by significantly higher expression of PD-L1 and HLA-E. HLA class I molecules tended to be more highly expressed in ALL than in AML blasts. After k-mean clustering of the blasts, it was observed that a higher proportion of ALL blasts were present in cluster 1, whereas clusters 2 and 3 were characterized by a majority of AML blasts (Fig. 3D). PD-L1 characterized ALL blasts from cluster 1 (Fig. 3E). In contrast, AML blasts showed strong expression of PVR, as well as ULBPs and MICA/B. Cluster 2 AML blasts expressed higher levels of NKG2D ligands (ULBPs and MICA/B) compared to clusters 1 and 3. In contrast, cluster 3 AML blasts have drastic decreased expression of HLA-C and HLA-E compared to cluster 1. This dichotomy in ligand expression results in different receptor-ligand interactions between NK cells and acute leukemic cells. To confirm our functional results, we evaluated the degranulation of NK cells against primary leukemic cells from the different clusters. Our findings revealed that educated KIR2DL⁺ NK cell subsets were the most efficient against primary leukemic blasts representative of each cluster (Fig sup 5). Of note, the expression of CD57 on NK cell subsets did not enhance their efficacy. Overall, NK^2A+ and NK^2A+KIR+ constituted the most efficient NK cell subsets against primary leukemic blasts.

Fig. 3.

ALL and AML cells exhibit differential expression patterns of ligands for key NK receptors. Schematic representation of the method used to characterize the phenotype of leukemic cell patients (A). Swimmer’s plot of patient outcomes (n = 33) (B). Bar graphs comparing the MFI of various ligands expressed at the cell surface of leukemic cells from ALL or AML patients (C). K-mean clustering of ligands expressed by leukemic cells from ALL and AML patients (D). Bar graph comparing the ligand expression profiles of the three clusters identified by K-mean clustering (E). Statistical significance is indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Comparisons were made by multiple Mann–Whitney (Bonferroni-Dunn method) and Kruskall-Wallis tests. Bar graphs represent mean ± SD

Expression Profile of KIR2DL5 on Resting and Feeder or Cytokine Stimulated NK Cells

Our findings demonstrate that myeloid leukemic cells are better recognized by NK^2A+KIR+ NK cell subset that preferentially co-express KIR2DL5. It is important to note that the KIR2DL5 gene is not universally present and is associated with KIR B + haplotypes, which have been shown to confer a reduced potential for NK cell education in comparison to KIR A counterparts, as previously described [25]. Even when the KIR2DL5 gene is present, the expression of KIR2DL5A*001 and KIR2DL5A*005 allotypes is predominantly observed on membrane NK cells, with the KIR2DL5A*001 allotype being recognized by UP-R1 mAb [27]. In our cohort of 200 healthy blood donors, 96 were found to possess the KIR2DL5 gene, while 53 expressed KIR2DL5 on their NK cells (53 expressed KIR2DL5A*001 allele and 23 expressed the KIR2DL5A*005 allele). The expression of KIR2DL5 is observed on a limited proportion of NK cells (median of 6.88) (Fig. 4A). The expression levels of KIR2DL5 remained unaffected by the CMV status. Of note, CD57 negatively impacts the frequency of KIR2DL5 but not KIR3DL2 (Fig. 4B). The frequency of KIR2DL5 increased with the number of inhibitory KIRs (KIR2DL1, KIR2DL2/3, KIR3DL1) expressed whatever NKG2A expression (Fig. 4C). Nevertheless, the frequency of KIR2DL5 is also negatively impacted by NKG2A in NK cell subsets expressing 0, 1 or 2 KIRs (Fig. 4C). We further investigated the KIR2DL5 expression on NK cells expanded using a feeder cell system or through cytokine stimulation, as NK cell adoptive therapies frequently employ an NK cell product that has been expanded. The expansion of NK cells is drastically more efficient with EBV-B cell line stimulation in comparison to cytokine cocktail (IL-12, IL-15 and IL-18) (Fig. 4D-E). The expression of PVR specific inhibitory receptors as TIGIT and CD96 in contrast to activating DNAM-1 was increased on NK cells stimulated with feeder system (Fig. 4F). The KIR2DL5 expression is globally maintained on NK cells in both system of expansion (Fig. 4G). It is coexpressed with TIGIT and DNAM-1 on resting NK cells as previously shown and this co-expression is maintained on feeder or cytokine stimulated NK cells (Fig. 4H).

Fig. 4.

Expression profile of KIR2DL5 on resting and activated NK cells. Mean frequencies of KIR2DL5⁺ NK cells from 29 healthy blood donors expressing KIR2DL5A*001 allotype. Eighty individuals are CMV seronegative and 11 are CMV seropositive (A). Bar graphs comparing the mean frequency of CD57⁻ vs CD57⁺ NK cells that express KIR3DL2 and KIR2DL5 (n = 40) (B). Bar graphs comparing the mean frequency of NK cells expressing KIR2DL5 following the number of expressed KIRs (0 to 3) and NKG2A expression (n = 20) (C). Mean frequency of CD3⁻ CD56⁺ cells, described as NK cells, determined for 6 KIR2DL5A*001⁺ healthy blood donors by flow cytometry using specific mAbs described in methods at days 0, 7, 14 and 17 following stimulation with the irradiated (HLA-A*03:01, A*26:01, B*07:02, C*07:02) EBV-B cell line (EBV-01/10) or cytokines (CK) including IL-12 (10 ng/ml), IL-15 (10 ng/ml) and IL-18 (100 ng/ml) (D). Representative density plots showing the NK cell expansion with feeder or cytokine stimulation at day 0 and day 14 (E). Geometric mean fluorescent intensity (MFI) of marker (TIGIT, CD96 and DNAM-1) expressed on NK cells at day 0, 7, 14 and 17 following stimulation with the irradiated EBV-01–10 cell line or CK (F). Mean frequencies of KIR2DL5⁺ NK cells at day 0, 7, 14 and 17 following stimulation with the irradiated EBV-01–10 cell line or CK (G). Representative density plots showing the coexpression of KIR2DL5 with TIGIT and DNAM-1 on NK cells expanded after feeder or cytokine stimulation at day 0 and day 14 (H). Means are indicated with error at all kinetic points. Statistical significance is indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. Comparisons were made by Mann–Whitney and Kruskall-Wallis tests. Bar graphs represent mean ± SD

KIR2DL5-PVR Axis Drastically Limits NK Cell Responses Against Acute Leukemic Cells

Our findings demonstrate that myeloid leukemic cells are better recognized by KIR⁺ NK cells that preferentially co-express KIR2DL5. We postulated that the KIR2DL5-PVR interaction might influence the response of KIR⁺ NK cells against myeloid leukemic cells that express PVR. To gain a deeper understanding of the impact of this interaction, we conducted CD107a degranulation assays of ten NK cell subsets co-expressing or not NKG2A, KIR2DL1, KIR2DL3, and KIR2DL5 (Fig. 5A) against primary leukemic cells expressing PVR (Fig. 5B, Fig sup 6). Primary leukemic cells are resistant to NK lysis and the degranulation level was low in contrast to those observed against leukemic cell lines. The expression of KIR2DL5 on NK cell subsets limits the degranulation of NK cell subsets (Fig. 5C, Fig sup 6). Blockage of the KIR2DL5-PVR interaction with a PVR specific mAb restores the degranulation of KIR2DL5⁺ NK cell subsets (Fig. 5D). To further investigate the impact of KIR2DL5-PVR axis on NK cell degranulation against PVR⁺ target cells, experiments have been done against six leukemic cell lines either expressing or lacking PVR (Fig. 5E). The healthy blood donors from which NK cell function was studied were selected on the basis of their KIR2DL5A*001 allelic typing associated with KIR2DL5 membrane expression on NK cells. NK cells expressing KIR3DL1, NKG2C, and CD57 were excluded from the analysis.

Fig. 5.

KIR2DL5-PVR axis inhibits NK cell degranulation against PVR + leukemic cells. Schematic representation of the degranulation assay performed on 10 NK cell subsets co-expressing or not NKG2A, KIR2DL1, KIR2DL3 and KIR2DL5 against primary leukemic cells and leukemic cell lines expressing or lacking PVR (A). Histograms showing the PVR and Nectin-2 expression on primary leukemic AML6 cells (B). Bar graphs comparing the degranulation of KIR2DL5⁻ vs KIR2DL5⁺ NK cells against primary leukemic AML6 cells (C). Bar graphs comparing the degranulation of KIR2DL5⁻ vs KIR2DL5⁺ NK cells against primary leukemic AML6 cells preincubated with or without an anti-PVR mAb (D). Bar graphs comparing the degranulation of KIR2DL5⁻ vs KIR2DL5⁺ NK cells against PVR- leukemic cells (Raji, Daudi) and PVR + cell lines (NB4, KG1, Molt4, H9) (E). Histograms showing the PVR and Nectin-2 expression on PVR^KO KG1 cell line generated by CRISPR-Cas9 technique (F). Bar graph comparing the degranulation of total NK cells against the PVR^WT vs PVR^KO KG1 cell lines (G). Bar graph comparing the degranulation of KIR2DL5⁻ and KIR2DL5⁺ NK cell subsets against the PVR^WT vs PVR^KO KG1 cell lines (H). Statistical significance is indicated as follows: *p < 0.05. Comparisons were made by multiple Mann–Whitney and Kruskall-Wallis tests. Bar graphs represent mean ± SD

The presence of KIR2DL5 at the NK cell surface did not affect the degranulation of NK cells against the Raji and Daudi cell lines that do not express PVR (Fig. 5E). However, degranulation of KIR2DL5⁺ NK cells is drastically inhibited against target cells expressing PVR as NB4, KG1, Molt4 and H9 in contrast to KIR2DL5⁻ NK cell subsets. This suggests that the presence of KIR2DL5 markedly suppressed NK cytotoxic activity against PVR⁺ leukemic cell lines. To go further, a PVR knockout KG1 cell line, PVR^KO KG1 was generated using the CRISPR Cas9 approach (Fig. 5F). Degranulation of total NK cells was higher against the PVR^KO KG1 cell line, suggesting a functional role of the ligand (Fig. 5G). Next, we investigated the degranulation of the 10 NK cell subsets against the PVR^WT KG1 and PVR^KO KG1 cell lines. We confirmed that, regardless of the NK cell subset, degranulation against the PVR^WT KG1 cell line is lower than against the PVR^KO KG1 cell line (Fig. 5H). Furthermore, the degranulation of KIR2DL5⁺ NK cells is significantly reduced against the PVR^WT KG1 cell line while no reduction is observed against PVR^KO KG1 cell line. Overall, we demonstrate that the KIR2DL5-PVR interaction induces strong inhibition of NK cell degranulation.

Discussion

In the present study, we have provided a comprehensive illustration of the diverse range of NK cell responses, offering valuable insights into the NK cell subsets that are most effective against lymphoid and myeloid leukemia. Our findings indicate that lymphoid leukemia cell lines are generally more susceptible to NK cell recognition compared to those of myeloid origin in accordance with previous studies [10]. The NK^2A+ cell subset which is the predominant cell subset, was identified as one of the most efficient against lymphoid cell lines. In contrast, the NKG2A⁺KIR2DL⁺CD57⁻ NK cell subsets (NK^2A+KIR+) were effective against myeloid cell lines, as previously demonstrated [10, 28, 29]. Indeed, adoptive transfer of allogeneic NK cells is more effective against myeloid leukemias than against lymphoid leukemias. Furthermore, the NKG2A⁻KIR⁻CD57⁻ cell subset (NK^neg) displays a low cytotoxic potential against all leukemia targets, which aligns with the notion that in the absence of inhibitory KIR and NKG2A, the education of NK cells is constrained. It is noteworthy that educated NK cells were highly effective against primary leukemic blasts. The functional heterogeneity is corroborated by the transcriptomic signatures of these cell subsets, which substantiate the cytotoxic potential of NK^2A+ and NK^2A+KIR+ cell subsets in contrast to NK^neg cell subset in accordance with the findings of Vivier’s studies [16, 17]. Expression of CD57, a marker of near-senescent cells with short telomeres [30], further reduced the natural cytotoxicity of all NK cell subsets. Different studies have previously underlined the hyporesponsiveness of these adaptive NK cell subsets. [10, 31–33].

AML blasts tend to downregulate HLA class I molecules, which prevents the engagement of inhibitory KIRs and facilitates NK-mediated killing. Conversely, ALL blasts retain HLA class I expression, which can inhibit KIR⁺ NK cell subsets. Additionally, ALL blasts express HLA-E molecules. Thus, the inhibitory signal mediated by NKG2A engagement with HLA-E ligand should be overridden by activating signals due to engagement of activating NK receptors with cognate ligands expressed on ALL. Among all the molecular interactions studied, no particular activating receptor susceptible to engage a ligand induced on ALL was identified. The efficacy of NKG2A⁺ cell subset against lymphoid cells may be partly due by the interaction of 2B4 with its ligand CD48, specifically expressed on ALL cells (data not shown). Moreover, adhesion molecules, such as LFA-1 and CD2 on NK cells respectively binding to ICAM-1 and LFA-3 that can be downregulated on leukemic cells [34–37], reducing NK cell conjugation and inhibiting cytotoxicity. CD2 also enhances NK cell activation by synergizing with NKp46 and DNAM-1 [38]. Deciphering expression of these molecules should bring insight on the efficiency of the NK^2A+ cell subset against ALL.

Activating receptors, particularly NKG2D and DNAM-1, seems to be crucial for targeting AML blasts as their respective ligands MICA/MICB/ULBP (NKG2DL) and PVR/Nectin-2 (DNAM-1L), are preferentially expressed by those cells. AML cells were classified into two clusters based on the expression of ligands. Both clusters expressed DNAM-1L, but cluster 2 additionally co-expressed NKG2DL. AML cells have been observed to downregulate NKG2D ligands in order to escape cell NK cell killing [39] suggesting that the NKG2DL⁻ group may be eliminated with less efficiency than the NKG2DL⁺ group. However, activation of resting NK cells occurs after the co-engagement of activating receptors [38]. As DNAM-1 and NKG2D do not synergize to induce NK cell activation, alternative co-receptors may be implicated in the elimination of AML blasts. Interestingly, we documented a preferential expression of KIR2DL5 on KIR2DL⁺ NK cells that are more efficient against AML that express preferentially PVR in contrast to ALL. Of note, KIR2DL5 was preferentially expressed on KIR2DL1⁺ NK cells, which may explain why the degranulation of NKG2A⁺ KIR2DL1⁺ NK cells was lower than that observed for the NKG2A⁺KIR2DL3⁺ NK cell subset. Moreover, we showed a drastic inhibition of degranulation of all KIR2DL5⁺ NK cell subsets by KIR2DL5-PVR axis. This inhibition mechanism has also been observed in other pathological conditions, such as various solid and hematological tumors [40] as well as HIV [41]. Importantly, PVR serves as a ligand for a number of receptors, including the inhibitory receptors TIGIT and CD96, as well as the activating receptor DNAM-1. While KIR2DL5 has its own binding site on PVR [40], TIGIT, CD96 and DNAM-1 share a binding domain [42]. NK cell-based immunotherapies are currently of great interest and require further refinement to optimize them. One approach to optimize these therapies could be to improve the selection of NK cell donors. For AML patients, it would be advantageous to select a donor with the AA KIR genotype, who lacks the expression of KIR2DL5, in order to circumvent the inhibitory KIR2DL5-PVR interaction.

In conclusion, our findings allow us to identify NK cell subsets that are most effective in eliminating lymphoid and myeloid leukemic cells. Furthermore, we have shown that molecular interactions between NK cells and leukemic cells influence NK cell responses, particularly the inhibitory KIR2DL5-PVR axis. However, the findings of this study are applicable to expanded NK/AML interactions, though they have yet to be validated in AML patients. This represents a promising area for future investigation. Moreover, in order to propose an immunotherapy strategy that is capable of effectively targeting leukemia cells which express PVR, it is first necessary to define potential physiological interactions involving KIR2DL5. This is an area in which research is lacking. The integration of these data is of importance for the optimization of NK cell-based immunotherapies, as the selection of NK cell donors represents a key parameter for the enhancement of those therapies. Further studies should focus on characterizing how the interplay between activating and inhibitory receptors impacts NK cell activation, particularly in the context of AML where therapeutic options are limited and patients continue to experience relapse.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

Study design: acquisition and analysis of flow cytometry data: GD and CW; KIR genotyping: NL, PS and KG; donor sample collection and biological documentation: GD, EF, CW, NL, AW and KG; patient sample collection and biological documentation: CW, GD, EF, MB, MJ, PC; Data interpretation: EF, GD, PS, MJ, MB, KG, and CR; Figure and table design: EF, CR; Manuscript writing: EF, CR. All authors read and approved the final version of the manuscript.

Funding

This work was financially supported by the Etablissement Français du Sang (EFS)/Centre Pays de la Loire and by grants from la Ligue contre le Cancer (Loire-Atlantique and Ile et Vilaine), Leucémie Espoir Atlantique Famille (LEAF) and l’Agence de Biomédecine (ABM). PS and MJ are PhD students supported by the Région Pays de la Loire/EFS Centre Pays de la Loire (N^o2022-09690) and INSERM, respectively.

Data Availability

No datasets were generated or analysed during the current study.

Declarations

Competing interest

The authors declare no competing interests.