Activating KIRs on Educated NK Cells Support Downregulation of CD226 and Inefficient Tumor Immunosurveillance

NK cells (NKc) play a central role in cancer immune surveillance (1), but the relevance of NKc education through inhibitory and activating receptors has just begun to be elucidated (2, 3). NKcs become educated and fully competent during a process known as “licensing” in which NKG2A or inhibitory killer-cell immunoglobulin-like receptors (iKIR) interact with their cognate human leukocyte antigen class-I (HLA-I) ligands (4–7). iKIR/ligand licensing interactions include KIR2DL1/HLA-C2 allotypes (Lys80), KIR2DL2-3/HLA-C1 allotypes (Asn80) (8), KIR3DL1/HLA-I alleles with Bw4 epitope (9), KIR3DL2/HLA-A*03,A*11 alleles (10), and NKG2A/HLA-E (6). Seminal research has shown that only NKcs expressing iKIRs for self-HLA-I ligands are functionally active against allogeneic-mismatched (“missing-self”) tumors (11, 12). NKc education is also involved in immune surveillance against autologous solid tumors (13). Apparently, educated NKcs are superior effectors against missing-self tumors, whereas uneducated/uninhibited NKcs are better against tumors that retain HLA-I expression (2). However, this statement may need reconsideration in view of the results reported for myeloma, which suggests the overexpression of HLA-I on target cells may make effective cancer immune surveillance highly dependent on multiple and diverse licensing interactions (14). In addition, transformed cells can specifically interfere with DNAM-1(CD226) upregulation induced by licensing interactions, as a specific mechanism to escape NKc immune surveillance (3).

Human NKcs can also express up to six different activating KIRs (aKIR): KIR2DS1-5 and KIR3DS1. Interactions have been reported between KIR2DS1/HLA-C2 allotypes (15, 16), KIR2DS2/C1 (17), KIR2DS2/HLA-A11 (18), KIR2DS2/non-HLA cancer cell–expressed ligand (19), KIR2DS4/HLA-A*11 and to a limited number of HLA-C1 and -C2 allotypes (20, 21), KIR3DS1/HLA-F (22), and KIR3DS1*014/HLA-Bw4 (23). However, the role of aKIRs in NKc education is less well known. In contrast to inhibitory receptors, aKIR signaling, specifically KIR2DS1 in HLA-C2 homozygous individuals, renders NKcs hyporesponsive, a mechanism that has possibly evolved to prevent autoreactivity (15, 16). Apparently, aKIR/self–HLA interactions have detrimental effects on the tumor immune surveillance of NKcs (24), with aKIR richer B-haplotypes showing higher susceptibility to cancer and/or worse outcomes (24–28). A higher frequency of aKIRs or Bx centromeric and telomeric genotypes has been associated with gastric cancer (26), non-Hodgkin lymphoma (27), and childhood acute lymphoblastic leukemia (28). Those genotypes are also associated with poorer prognosis in non-Hodgkin lymphoma (27), higher frequency of minimal residual disease after treatment in childhood acute lymphoblastic leukemia (25), or significantly adverse survival in patients with hematologic malignancies after hematopoietic stem cell transplantation from unrelated donors with NKs educated by aKIRs (24). However, these results contradict those reporting survival benefits for patients with acute myeloid leukemia (AML) who have received grafts from unrelated donors with 1 or 2 KIR B-haplotypes, thus providing evidence that donors with KIR B-haplotype should be used preferentially in HLA-unrelated donor transplantations (29). aKIR/self–HLA interactions also have detrimental roles in virus surveillance (30), except for the KIR3DS1/HLA–Bw4 interaction that slows down progression to AIDS (31) or telomeric aKIRs that protect against cytomegalovirus infection after kidney transplantation. In contrast, aKIRs seem to play decisive and favorable functions in human reproduction (31, 32).

The roles of inhibitory receptors for the functional education of NKcs was first described for mouse and human in 1998 and 2006, respectively (7, 33), but clear phenotypic and metabolic evidence of the involvement of iKIRs in NKc licensing was not described in humans until 2015 (34) and 2018 (35). NKc education through inhibitory receptors is strongly linked to a dynamic and coordinated expression of CD226 and to conformational changes in the lymphocyte function–associated antigen-1 (LFA-1) receptor. The expression of CD226 on NKcs positively correlates with the number and the affinity of iKIR2D/HLA-C licensing interactions as well as with the magnitude of NKc functional responses (3, 34). Furthermore, licensed NKcs show metabolic reprogramming toward glycolysis and mitochondrial-dependent glutaminolysis, whereas unlicensed NKcs solely depend on mitochondrial respiration for cytolytic function, suggesting that enhanced glycolysis in licensed NKcs supports their increased proliferative and cytotoxic capacities (35). Licensing interactions also induce lysosomal remodeling with accumulation of granzyme B in dense-core secretory lysosomes (36).

Although precise mechanisms have not yet been identified, education via aKIRs shares features with the hyporesponsiveness induced by chronic stimulation through the activating receptor NKG2D. Sustained stimulation with tumor cell–bound ligands reduces NKG2D association with its signaling adaptors DAP-10 and KARAP/DAP-12, and uncouples the NKG2D receptor from the intracellular mobilization of calcium and the induction of NKc-mediated cytolysis (37). In addition, it has been described that the overall expression of the non-KIR–inhibitory LILRB1 and NKG2A receptors are increased in donors with richer aKIR BB genotypes (38).

Although mechanisms involving the suppression of NKcs by tumors are widely known, the downmodulation or blockage of activating receptors (39) or the alteration of the stoichiometric ratio of activating and inhibiting molecules (3, 40) could be involved. This manuscript describes how aKIR can contribute to immune evasion of different types of solid cancers as a result of increased sensitivity of NKcs to tumor-induced downmodulation of the activating receptor CD226, and how this significantly impacts the survival of patients.

This prospective observational study included 621 healthy Caucasian volunteers (control group) and 249 consecutive Caucasian patients with melanoma (n = 80), bladder (n = 80), or ovarian (n = 89) tumors (Supplementary Table S1). These series of controls and patients with cancer coincide with the cohort used in a previous study aimed at describing the role of NKc education in tumor immune surveillance (3).

Peripheral blood samples anticoagulated with EDTA were obtained from controls and patients from Hospital Clínico Universitario Virgen de la Arrixaca and Hospital General Universitario Santa Lucía (Murcia, Spain). HLA-I (KIR ligands) and KIR genotypes, as well as the expression of activating (CD16, CD226 and aKIR) and inhibitory (iKIR and NKG2A) receptors on NKcs were analyzed. 10-year progression-free survival (PFS) and overall survival (OS) were analyzed according to the expression of CD226 and iKIR receptor on NKcs. Institutional review board (IRB-00005712) approved the study. Written informed consent was obtained from all patients and controls in accordance with the Declaration of Helsinki.

HLA-A, -B, -C, and KIR genotyping was performed on DNA samples extracted by QIAmp DNA Blood Mini Kit (Qiagen) using sequence-specific oligonucleotide PCR (PCR-SSO) and Luminex technology by Lifecodes HLA-SSO and KIR-SSO typing kits (Immucor Transplant Diagnostic, Inc.), as described previously (3, 14). HLA-C genotyping allowed distinction between HLA-C^Asn80 (group-C1) and HLA-C^Lys80 (group-C2) alleles (8). HLA-A and -B genotyping allowed identification of alleles bearing the Bw4 motif according to the amino acid sequences at positions 77 to 83 in the α1 domain of the HLA class-I heavy chain (9). KIR genotyping identified iKIRs (2DL1-L3/2DL5 and 3DL1-L3) and aKIRs (2DS1-S5 and 3DS1), as well as KIR2DL4, which has both inhibitory and activating potential (41). The method used could not distinguish between KIR2DL5A (telomeric) and KIR2DL5B (centromeric) forms. Different allotypes of KIR2DS4 were detected, including the expressed allotype KIR2DS4 full exon-5 (KIR2DS4full) and the nonexpressed KIR2DS4-deleted exon-5 (KIR2DS4dell).

KIR genotypes were grouped into AA if they contained only genes from the canonical A-haplotype (KIR3DL3, KIR2DL3, KIR2DL1, KIR2DL4, KIR3DL1, KIR2DS4, and KIR3DL2; ref. 42). KIR2DS4 is the only aKIR present in the telomeric A-haplotype. Any genotype containing additional KIR genes were considered Bx. KIR2DS2/KIR2DL2 and KIR3DS1/KIR2DS1 are localized centromerically and telomerically with respect to KIR2DL4 and are referred to as B-centromeric and B-telomeric genes, respectively. KIR2DS3 and KIR2DS5 can localize to both the centromeric or telomeric regions of B-haplotypes; centromeric KIR2DS3 was assigned in case of absence of KIR2DS1 and KIR3DS1 genes. Individual genotypes were classified as Bx-centromeric (cBx) or Bx-telomeric (tBx) if they carried at least one of the centromeric (KIR2DL2/KIR2DS2) or telomeric (KIR2DS1/KIR3DS1) group-B genes, whereas those remaining were classified as AA-centromeric (cAA) and AA-telomeric (tAA; Fig. 1).

The expression of CD226 (DNAM-1), NKG2A, and KIR receptors (KIR2DL1, 2DS1, 2DL2/S2, 2DL3 and 3DL1) on both CD3⁻CD16^−/+CD56⁺⁺ (CD56^bright) and CD3⁻CD16⁺CD56⁺ (CD56^dim) NKc subsets and CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells was evaluated as a percentage of positive cells, and as mean fluorescence intensity (MFI) using LSR-II and DIVA Software (BD Biosciences), as described previously (3). Photomultiplier (PMT) voltages were adjusted daily using rainbow calibration particles (BioLegend; Supplementary Fig. S1). Fluorescence compensation was finely adjusted using as reference negative events for each fluorochrome. The staining protocol included 11-color/12-mAb: CD3 AmCyam (clone SK7, BD Biosciences), CD4 PE-CF594 (RPA-T4, BD Biosciences), CD8 APCCy7 (SK1, BD Biosciences), CD16 PacBlue (3G8, BD Biosciences), CD56 BV711 (NCAM16.2, BD Biosciences), CD158a,h PECy7 (EB6, BD Biosciences, recognizes both KIR2DL1 and 2DS1), CD158b1/b2,j PE-Cy5 (GL183, Beckman-Coulter, recognizes KIR2DL2, 2DL3, and 2DS2), CD158a FITC (143211, R&D Systems, KIR2DL1), CD158b2 APC (180701, R&D Systems, KIR2DL3), CD158e APC (DX9, R&D Systems, KIR3DL1), CD226 PE (11A8, BioLegend), and NKG2A biotin (REA110, Miltenyi Biotec). Streptavidin AF700 (Life Technologies, Molecular Probes) was used to detect biotinylated NKG2A.

Supplementary Figure S1 shows the gating strategy used for the analysis. Briefly, a FSC/SSC dotplot was used to identify lymphocytes; a CD3/CD16 dotplot to select CD3⁺ T cells; a CD4/CD8 dotplot to identify principal T-cell subsets; a CD16/CD56 dotplot to identify CD56^bright and CD56^dim NKcs; a CD56/CD226 dotplot to select CD226⁺ cells; a CD3/NKG2A dotplot to identify NKG2A⁺ cells; a KIR2DL1S1/KIR2DL1 dotplot to distinguish KIR2DL1⁺, KIR2DS1⁺, double-positive KIR2DL1/S1, and double-negative KIR2DL1/S1 NKcs; and a KIR2DL2/S2/L3+ KIR3DL1 dotplot to distinguish KIR2DL2S2⁺, KIR2DL3⁺, KIR3DL1⁺, and triple-negative KIR2DL2S2/L3/3DL1 NKcs. Internal negative cells were used as isotype controls to set positive cutoffs. Gates were hierarchically and electronically combined to define the following NKc subsets: (i) CD56^bright; (ii) CD56^dim KIR-negative NKG2A⁺ NKcs; (iii) CD56^dim KIR-negative NKG2A⁻ (nonlicensed); (iv) CD56^dim NKcs single-positive for KIR2DL1, 2DS1, 2DL2/S2, 2DL3, or 3DL1; and (v) CD56^dim KIR double-positive KIR2DL1_L2/S2, KIR2DL1_L3, KIR2DL1_3DL1, and KIR2DL2/S2/L3_3DL1 CD56^dim NKcs.

All data were collected in a database (Excel 2003; Microsoft Corporation) and analyzed using SPSS-15.0 (SPSS Inc.). Pearson χ² and two-tailed Fisher exact tests and Mann–Whitney, ANOVA, and post hoc tests were used to analyze categorical (i.e., sex, tumor staging) or continuous (i.e., age, CD226 MFI) variables, respectively. Kaplan–Meier estimator and log-rank tests were used to analyze patient survival (PFS and OS). Time to events (progression or death) was estimated as months from the diagnosis date. ROC curve analysis was used to explore patient survival and to determine the optimal cut-off values for CD226-MFI on total NKcs as well as for CD226/KIR2DL1, CD226/KIR2DL2, CD226/KIR2DL3, and CD226/KIR3DL1 MFI ratios on NKc subsets. Sex, age, tumor type, tumor staging, and CD226-MFI on total NKcs were evaluated in a Cox regression analysis to confirm positive associations. The strength of association was estimated by OR and 95% confidence interval (95% CI). Data is expressed as mean ± SEM. P values < 0.05 were considered significant. The Bonferroni correction (Pc) was applied when needed.

Clinical and biological characteristics of patients and healthy controls included in this manuscript are described in a previous work aimed at analyzing the role of NKc education in tumor immune surveillance (3). Likewise the genomic landscape of iKIR and iKIR/HLA-I ligand interactions for this series of patients with cancer and healthy controls is described in that work. Detailed information on the frequency of aKIRs and putative aKIR/HLA-I ligand interactions is shown in Table 1. Patients from all cancer groups and healthy controls showed similar distribution of aKIRs and aKIR/HLA-I ligand interactions. Patients with melanoma (90.0% vs. 78.9%, P = 0.02, Pc > 0.05) and ovarian (88.6 vs. 78.9%, P = 0.03, Pc > 0.05) cancer had higher frequency of KIR2DS4del exon-5; patients with melanoma (41.3% vs. 29.1%, P = 0.014, Pc > 0.05) also showed higher frequency of KIR2DS5 than healthy controls. Nonetheless, all significant differences were lost after Bonferroni correction by the number of comparisons (n = 4).

There were no significant differences in the frequency of Bx centromeric or Bx telomeric KIR genotypes between control and cancer groups (Table 1). No differences were found either in the frequency of the 6 main centromeric/telomeric KIR combined genotypes between controls and patients in our series: cAA/tAA (26.3% vs. 23.2%), cAA/tBx-KIR2DS3⁺ (2.4% vs. 2.0%), cAA/tBx-KIR2DS5⁺ (11.2% vs. 14.8%), cBx-KIR2DS3⁻KIR2DS5⁻/tAA (20.3% vs. 19.2%), cBx-KIR2DS3⁺/tAA (10.6% vs. 10.4%), and cBx/tBx (29.2% vs. 30.3%; Fig. 1).

As described previously for this series of patients (3), solid cancer patients with low expression of CD226 in total NKcs (CD226^low, MFI < 7385) showed significantly shorter mean OS (60.7 vs. 98.0 months, P < 0.001) than patients with CD226^high (MFI ≥ 7385) (Fig. 2A). In this article, we explore the influence of individual iKIRs and aKIRs on the survival impact exerted by the low expression of CD226 in total NKcs (Fig. 2B). Differential mean OS for patients with the inhibitory KIR2DL1⁺, KIR2DL2/S2⁺, KIR2DL3⁺, KIR2DL5⁺, KIR3DL1⁺, or the activating centromeric-KIR2DS3⁺ genotypes compared with all patients with CD226^low ranged from −9.7 to +9.3 months. However, compared with the total number of patients, much shorter mean OS was observed for patients with CD226^low in the presence of B-telomeric activating KIR2DS1⁺ (−29.0 months, P < 0.05), KIR2DS3⁺ (−25.3 months, P < 0.05), KIR2DS5⁺ (−36.8 months, P < 0.05), and KIR3DS1⁺ (−29.0 months, P < 0.05), and the A-telomeric activating KIR2DS4full⁺ (−28.3 months, P < 0.05). In contrast, patients with CD226^high and different iKIR⁺ or aKIR⁺ genotypes showed minimal variations when compared with the total number of patients.

Centromeric genotype did not influence the survival impact exerted by the expression of CD226 on total NKcs (Fig. 2C, top plot). Thus, the mean OS for AA-centromeric or Bx-centromeric genotypes were 103 or 95.6 months for CD226^high patients and 22.0 and 34.4 months for CD226^low patients, respectively. However, the Bx-telomeric genotype in CD226^low patients showed significantly shorter mean OS (31.6 months, P < 0.001) than the Bx-telomeric with CD226^high (103 months) and AA-telomeric patients with CD226^high (95.2 months) or CD226^low (86.8 months; Fig. 2C, bottom plot).

The deleterious effect of the combined Bx-telomeric/CD226^low was observed in patients with melanoma (50%, P < 0.05), bladder (53.8%, P < 0.001), or ovarian (80%, P < 0.001) carcinomas, that showed higher rates of overall mortality than patients with melanoma, bladder, and ovarian cancer with AA-telomeric/CD226^high (8.1%), AA-telomeric/CD226^low (10.1%), and Bx-telomeric/CD226^high (21.1%) genotypes (Fig. 2D).

The A-telomeric activating KIR2DS4full also increased the deleterious effect of CD226^low, specifically, when patients were carriers of its cognate HLA-I ligands (HLA-A*11, C*02, C*04, C*16). In absence of KIR2DS4 ligands, comparable mean OS (104.5 ± 10.3 vs. 92.9 ± 8.2 months, P = 945) was observed for KIR2DS4full⁺ AA-telomeric patients with CD226^high and CD226^low. In contrast, and as expected, for KIR2DS4full⁺ Bx-telomeric individuals the mean OS (37.9 ± 9.6 vs. 104.4 ± 5.9 months, P < 0.001) was lower in CD226^low than in CD226^high patients. However, in presence of KIR2DS4 ligands, both AA-telomeric (25.7 ± 6.8 vs. 106.4 ± 7.4 months, P = 0.034) and Bx-telomeric (25.0 ± 11.0 vs. 101.6 ± 11.8 months, P = 0.034) KIR2DS4full⁺ patients showed shorter mean OS with CD226^low than those with CD226^high (Fig. 3A).

Altogether, B-telomeric (KIR2DS1, KIR2DS2, KIR2DS3, KIR2DS5, and KIR3DS1) and A-telomeric (KIR2DS4full) aKIRs led to a cumulative negative impact on the deleterious effect of CD226 downmodulation on total NKcs: 1 or 2 aKIRs (97.02 vs. 55.5 months mean OS, P = 0.055), 3 or 4 aKIRs (99.4 vs. 69.8 months mean OS, P = 0.015), or >4 aKIRs (104.7 vs. 33.4 months mean OS, P = 0.001) for CD226^high and CD226^low, respectively (Fig. 3B).

We next explored the impact of CD226/iKIR expression ratios from individual NKc subsets on the OS of patients with AA-telomeric or Bx-telomeric genotypes (Fig. 4). For AA-telomeric patients, differences in OS were clearly evident from 40 to 60 months onwards (long-term), with lower mean OS for low versus high CD226/KIR2DL1 (81.4 vs. 106.1 months, P = 0.05), CD226/KIR2DL2/S2 (87.2 vs. 105.9 months, P = 0.17), CD226/KIR2DL3 (86.7 vs. 97.2 months, P = 0.45), and CD226/KIR3DL1 (76.2 vs. 101.7 months, P = 0.05) expression ratios (Fig. 4A, top plots). However, differences in Bx-telomeric patients between low and high CD226/KIR2DL2/S2 (39.9 vs. 110.6 months, P < 0.001), CD226/KIR2DL3 (80.9 vs. 101.3 months, P = 0.02), and CD226/KIR3DL1 (72.3 vs. 108.7 months, P = 0.01) expression ratios were evident in the first 24 months (short-term) after diagnosis (Fig. 4A, bottom plots). No differences were observed for CD226/KIR2DL1 expression ratio in Bx-telomeric patients.

The presence of the cognate HLA ligand for each iKIR⁺ NKc subset on Bx-telomeric patients exacerbated the short-term detrimental effect of low versus high CD226/KIR2DL1 expression ratios with the C2-ligand (66.5 vs. 100.9 months, P = 0.05), CD226/KIR2DL2/S2 with the C1 ligand (24.8 vs. 109.5 months, P < 0.001), CD226/KIR2DL3 with the C1 ligand (80.6 vs. 100.6 months, P = 0.02), and CD226/KIR3DL1 with the Bw4 ligand (63.6 vs. 111.3 months, P = 0.003; Fig. 4B).

However, KIR ligands, by themselves, did not influence the deleterious effect of CD226^low (Supplementary Fig. S2). In fact, no differences were observed in the mean OS of patients with CD226^high (93.8 ± 7.0 vs. 97.6 ± 4.9 months) and CD226^low (50.5 ± 20.0 vs. 36.7 ± 4.70 months) in absence or presence of C1 ligands; with CD226^high (99.7 ± 5.5 vs. 100.0 ± 5.4 months) and CD226^low (35.0 ± .20.0 vs. 50.5 ± .10.9 months) in absence or presence of C2 ligands; with CD226^high (79.8 ± 11.7 vs. 104.0 ± 3.9 months) and CD226^low (21.0 ± .7.2 vs. 51.5 ± .10.8 months) in absence or presence of Bw4 ligands; or with CD226^high (101.5 ± 5.5 vs. 95.5 ± 7.2 months) and CD226^low (44.5 ± 5.3 vs. 54.1 ± .12.1 months) in absence or presence of KIR2DS4 ligands (HLA-A*11, C*2, C*4, and C*16).

In summary, low CD226/iKIR expression ratios exerted an earlier detrimental impact on the OS of Bx-telomeric patients, and a later effect on AA-telomeric patients. In Bx-telomeric patients, this effect was exacerbated by the presence of the cognate ligand of each iKIR⁺ NKc subset. KIR2DL2/S2⁺, followed by KIR3DL1⁺, KIR2DL1⁺, and KIR2DL3⁺ single positive NKc subsets had the most detrimental effects on Bx-telomeric patients with low CD226/iKIR receptor ratios.

Cox regression analysis confirmed that the type of tumor (P < 0.014; OR = 2.847), the tumor staging of each type of cancer (P < 0.042; OR = 2.602), and the expression of CD226 on total NKcs (P < 0.002; OR = 0.238) were independent variables that significantly impacted the PFS of patients with solid cancer. Furthermore, both the type of tumor (P < 0.05; OR = 10.435) and the expression of CD226 on total NKcs (P < 0.021; OR = 0.262) were independent variables associated with solid cancer OS (Supplementary Table S1).

We next analyzed the expression of CD226 on CD8⁺ T cells and total NKcs (Fig. 5). CD226 expression was specifically downmodulated on NKcs in patients with Bx-telomeric genotype who died during the follow-up period (7600 ± 676 vs. 9640 ± 880 MFI, P < 0.05) but not on CD8⁺ T cells (10,700 ± 1,160 vs. 10,800 ± 1,109 MFI; Fig. 5A).

As described previously (3), CD226/iKIR expression ratios increased proportionally relative to the number of cognate HLA-I ligands available for each NKc subset. For all patients with cancer, the expression ratios in the presence of 0, 1, or 2 cognate ligands were 4.8 ± 0.34, 9.3 ± 0.43, and 11.9 ± 0.92 (P < 0.001) for CD226/KIR2DL1; 0.33 ± 0.02, 0.41 ± 0.02, and 0.46 ± 0.03 (P = 0.042) for CD226/KIR2DL2/S2; 3.2 ± 0.30, 4.8 ± 0.53, and 6.1 ± 0.61 (P = 0.016) for CD226/KIR2DL3; and 0.78 ± 0.1, 0.9 ± 0.07, and 1.11 ± 0.09 (P = 0.032) for CD226/KIR3DL1 (Fig. 5B).

Patients with the Bx-telomeric genotype that had the highest tumor burden, and died during the follow-up period, when compared with patients who survived or bore the AA-telomeric genotype, showed decreased CD226/KIR2DL1 (−21.1%, P > 0.05), CD226/KIR2DL2/S2 (−50.7%, P < 0.001), CD226/KIR2DL3 (−11.1%, P > 0.05), and CD226/KIR3DL1 (−38.4%, P < 0.05) expression ratios on educated NKcs when double doses of ligands were present. However, the CD226/iKIR expression ratios remained unaltered on uneducated NKcs (Fig. 5B).

Compared with AA-telomeric patients, lower frequencies of CD56^dim NKcs were observed in patients with the Bx-telomeric genotype for all types of cancers, melanoma (12.6 ± 0.88% vs. 14.3 ± 1.2%, P < 0.05), bladder (13.3 ± 0.77% vs. 16.1 ± 1.3%, P < 0.05), and ovarian (10.96 ± 0.96% vs. 14.8 + 1.3%, P < 0.05) carcinomas, as well as in healthy individuals (11.3 ± 1.5% vs. 15.6 ± 1.5%, P < 0.05; Fig. 5C).

We next compared expression of the CD226 activating and the NKG2A inhibitory receptors on NKcs expressing or not expressing the KIR2DS1-activating receptor. The frequency of KIR2DS1⁺ NKcs increased proportionally to the number of copies of KIR2DS1 gene present in the genome. In AB-telomeric (KIR2DL1/S1 heterozygous) genotypes, approximately 50% of C2 ligand–specific NKcs expressed or coexpressed KIR2DS1, whereas 50% were KIR2DL1 single positive. In the case of BB-telomeric (KIR2DS1 homozygous) genotypes, approximately 75% of NKcs bearing specific receptors for C2 ligands expressed or coexpressed KIR2DS1, whereas 25% were KIR2DL1 single positive (Fig. 5D). Remarkably, KIR2DS1⁺ showed lower expression of CD226 (9,998 ± 203 vs. 11,127 ± 220 MFI, P < 0.001) and higher expression of NKG2A (725 ± 7 vs. 450 ± 21 MFI, P < 0.001), compared with KIR2DL1⁺S1⁻ NKcs (Fig. 5E).

We also investigated the impact of the total number of aKIRs that were available in the genomes of patients with cancer on the expression of CD226 on total NKcs. CD226 expression on total NKcs decreased proportionally to the number of aKIRs: 1–2 (9950 ± 0.3), 3–4 (9490 ± 0.31), and >4 (8790 ± 0.31, P = 0.01) aKIRs (Fig. 5F).

Finally, as described for colorectal cancer (43), CD16 expression was significantly downmodulated on peripheral blood NKcs in the three types of malignant cancers analyzed in our series, irrespective of their AA/Bx-telomeric genotype (Supplementary Fig. S3).

Immunotherapy using NKcs expanded/activated ex vivo or stimulated in vivo with new immune-stimulatory agents could offer effective alternative treatments against recurrent/refractory tumors (40, 44). For this purpose, it is useful to have relevant biomarkers that can help in the selection of patients who could benefit from these therapies, as well as for the selection of NKc donors that could offer the greatest possibilities for effective immune responses. However, there are issues associated with NKc biology that make these selections difficult, such as the role played by NKc education through activating and inhibitory receptors, whether acting separately or together. Although the favorable role of NKc education via iKIR in cancer immune surveillance has begun to be demonstrated (2, 3, 11–14), the role of aKIRs remains in the arena of scientific debate. The survival benefits in patients with AML that have received grafts with KIR B-haplotypes have guided the selection of aKIR-rich donors for unrelated haploidentical stem cell transplantation in the past decade (29); however, in general, the interaction of aKIR with their HLA ligands renders NKcs hyporesponsive (15, 16), and increases susceptibility to cancer and worsen patients’ outcomes (24, 25). In line with results observed in solid and hematologic malignances (25–28), data described in this article support a detrimental role for telomeric activating KIR2DS1, KIR2DS3, KIR2DS4full, KIR2DS5, and KIR3DS1 receptors in solid cancer immune surveillance, apparently by increasing the sensitivity of educated NKcs to tumor-induced CD226 downmodulation (3).

Results described by our group have demonstrated that NKc education through iKIR/HLA-I licensing interactions upregulates the expression of the activating CD226 receptor, reduces that of iKIR receptors, and shifts the CD226/iKIR expression ratio on NKc membranes. High expression of CD226 or increased CD226/iKIR ratios is associated with longer PFS and OS survival in solid cancers. In contrast, transformed cells can induce the downmodulation of these licensing-driven receptor rearrangements as a specific mechanism to evade NKc immune surveillance, and negatively impact the survival of patients (3). Complementary data shown in this article suggests that telomeric aKIRs may play dominant roles in the specific tumor-induced CD226 downmodulation on educated NKcs, and that this exerts a profound negative impact on the short-term survival observed in the case of Bx-telomeric patients, leaving a residual long-term detrimental effect as a consequence of reduced CD226/iKIR expression ratios in patients with AA-telomeric genotypes.

In this work, we have been able to demonstrate that reduced CD226/iKIR expression ratios from each iKIR⁺ NKc subset analyzed in peripheral blood, contributes to the short-term detrimental impact that manifests as a decreased OS in patients with the Bx-telomeric genotype, whereas in patients with the AA-telomeric genotype, long-term detrimental impact is the case. Nonetheless, a dominant role in this detrimental impact was revealed for KIR2DL2/S2⁺, followed by KIR3DL1⁺, KIR2DL1⁺, and KIR2DL3⁺ NKc subsets. Although the presence of putative HLA-I ligands per se (C1/C2, Bw4, or KIR2DS4 ligands) did not alter the survival pattern induced by the high or low expression of CD226, the short-term detrimental impact of reduced CD226/iKIR expression ratio observed for the Bx-telomeric genotypes from every iKIR⁺ NKc subset was enhanced in the presence of the corresponding cognate ligand, probably as a consequence of selective tumor-mediated CD226 downmodulation on educated NKcs, as is evident in Bx-telomeric genotypes, with this ratio left unaltered on uneducated NKcs in absence of specific ligands.

Although the underlying mechanisms are mostly unknown, they may be specific for NKcs, because no CD226 downmodulation was observed on CD8⁺ T cells from dying patients with the Bx-telomeric genotype. It can be hypothesized that the coexpression of aKIRs together with iKIR on the analyzed subsets of NKcs could have contributed to the suppression of NKcs through mechanisms similar to those described for the KIR2DS1/C2C2 interaction (15, 16) or for the continuous interaction of NKG2D with tumor cell–bound ligands. This interaction reduces NKG2D association with its signaling adaptors DAP-10 and KARAP/DAP-12 (37) and downregulates Zap70 and Syk (45), thus increases their activation thresholds, and renders NKcs hyporesponsive. The increase in the activation threshold, together with CD226 downmodulation, may make NKcs especially susceptible to tumor evasion. In support of this statement, the KIR2DL2/S2⁺ NKc subset showed the strongest impact on OS due to CD226/iKIR downmodulation in the Bx-telomeric patients. It should be taken into account that patients with the Bx-telomeric genotype and KIR2DL2/S2⁺ NKcs in peripheral blood must necessarily also be Bx-centromeric. Therefore, this analysis included patients bearing the highest number of aKIRs so that most likely, KIR2DL2/S2⁺ NKcs would also coexpress several aKIR simultaneously, thus turning them especially sensitive to the anergizing effect of permanent aKIR/ligand interactions that would occur with putative ligands expressed by the tumor (19). For a better understanding of these results, it should also be taken into consideration that aKIRs may provide Src kinases in trans that could phosphorylate the ITIMs of iKIRs, thus enhancing the inhibitory signaling of iKIR when they interact with their specific ligands (46).

Our analysis could not evaluate the coexpression of KIR2DL2 and KIR2DS2 receptors; nevertheless, we were able to evaluate the coexpression of KIR2DL1 and KIR2DS1. Up to 50% and 75% of C2-ligand specific NKcs (KIR2DL1/S1) expressed or coexpressed KIR2DS1 in the case of AB-telomeric (KIR2DL1/S1 heterozygous) and BB-telomeric (KIR2DS1 homozygous) patients, respectively. In addition, it has been reported that nearly two thirds of KIR2DL1⁺ circulating NKcs can coexpresses the activating KIR2DS4 receptor (32), and that a high proportion (over 50%) of circulating KIR2DL1⁺ and KIR3DL1⁺ NKcs can coexpress the activating KIR2DS2 receptor but not the inhibitory KIR2DL2 or KIR2DL3 receptors (47). Similar to our results with KIR2DS1, it has been described that a high proportion of NKs do express KIR3DS1, with nearly one third in heterozygous KIR3DL1/S1 and two thirds in KIR3DS1 homozygous individuals, expressing or coexpressing KIR3DS1 (48). Our data have brought to light a deleterious effect of telomeric aKIRs; however, we were not able to demonstrate this effect for centromeric aKIRs (KIR2DS2 and KIR2DS3/S5), most likely because these aKIRs are inherited in a block with other iKIRs (KIR2DL2, KIR2DL5, and KIR2DL4). Besides, our data showed that the negative impact exerted by the low expression of CD226 increases proportionally with the number of aKIRs, so that patients with more than 4 aKIRs presented the worst outcomes, which likely involves centromeric plus telomeric aKIRs, and therefore, indirectly indicates that both centromeric and telomeric aKIRs may have negative effects on tumor immuno-surveillance.

With the aim of generating NKc subsets tuned to recognize altered expression of specific HLA-I ligands, KIR promoters regulate gene transcription that lead to NKc repertoires with a variegated expression of iKIR receptors mostly specialized for responding exclusively to the loss of individual allotypes (49, 50). However, such mechanisms seem to be less restrictive for aKIRs, resulting in iKIR⁺ NKc subset that coexpress one or more aKIRs in direct proportion with the number of aKIR that are available in the genome. This would explain our results that demonstrated that the detrimental impact of low CD226 expression on patients’ survival was proportional to the number of aKIRs present in their genomes. In fact, our results show that aKIR expression, specifically KIR2DS1, resulted in reduced expression of CD226 when compared with NKc subsets exclusively expressing KIR2DL1. Therefore, it is reasonable to assume, and our data support, the notion that a larger number of aKIRs could progressively lead to reduced CD226 expression and to reduced survival of patients. These results are in concordance with those of Misra and colleagues (28) and Nowak and colleagues (24) that described higher incidences of hematologic malignances and shorter PFS after stem cell transplantation directly proportional to the number of aKIR/HLA–ligand interactions, suggesting that the coexpression of multiple aKIRs and their cognate HLA ligands on NKcs of donors (education system overlapping) can increase the tolerance of antitumor NKc subsets and worsen the clinical outcomes of patients in a quantifiable manner.

Although no experimental data are presented in this work to support the assertion that CD226 downregulation was tumor-induced, this could be assumed on the basis of the data reported by Rocca and colleagues describing that both tumor infiltrating and circulating NKcs displayed strong downregulation of CD226, CD16, and CD94 (43). In further support of this idea, our data showed that CD16 was also downmodulated in all patients with cancer from our series. Similarly, our data could not establish causality between CD226 downregulation and aKIR expression. Considering that the molecular pathways involved in both licensing-driven CD226 upregulation and cancer-induced CD226 downregulation are widely unknown, it is difficult to highlight potential mechanisms that are responsible for the drastic reduction in survival observed for aKIR^rich/CD226^low patients. However, it could be related to the signaling pathways described by Coodbridge and colleagues, in which a predominance of stimulatory signals through activating receptors lead to poorer functional responses (disarming), lower amounts of the granule matrix protein (serglycin) and the effector molecules (granzyme B and perforin), as well as a lack of dense-core secretory lysosomes. One putative pathway downstream of activation receptor signaling is PI3K/AKT, which stimulates the enzyme PIKfyve, which converts PI3P to PI(3,5)P2 and thereby, positively regulates the lysosome-specific Ca2+ channel, TRPML1. PIKfyve and TRPML1 are critically involved in lysosomal modulation in several cell types, including NKcs (36).

In line with the reported results showing an increased expression of LILRB1- and NKG2A-inhibitory receptors in donors with aKIR-rich BB genotypes (38), our results show that KIR2DS1⁺ NKcs had greater expression of NKG2A. In fact, the Bx-telomeric genotype has an even more global impact on NKc development, leading to significantly inferior numbers of total CD56^dim NKcs in Bx-telomeric patients with the three types of cancers and in healthy controls when compared with AA-telomeric individuals.

Altogether, (i) the lower numbers of CD56^dim NKcs systematically observed in all Bx-telomeric patients, (ii) the reduced expression of CD16 activating receptor in all patients with cancer, (iii) the higher expression or coexpression of aKIRs that may turn NKcs hyporesponsive and potentially contribute to the reduced expression of CD226, and (iv) the increased expression of NKG2A may contribute to the drastic reduction in the short-term survival observed in aKIR-rich Bx-telomeric patients with different types of solid cancers.

In summary, the results presented in this article provide relevant information on the role of aKIRs in solid cancer immune surveillance. Telomeric aKIRs, and possibly, centromeric aKIRs too, play an early detrimental role in solid cancer immune surveillance by increasing the sensitivity of educated NKcs to tumor-induced CD226 downmodulation or directly contributing to reduce CD226 expression on educated NKcs. The reduced expression of CD226 on NKcs with aKIR-rich genotypes may be a biomarker indicative of NKc hyporesponsiveness in patients that could benefit from new NKc immune-stimulatory therapies.

No potential conflicts of interest were disclosed.

Conception and design: B. Ferri, A. Minguela

Development of methodology: C.F. Guillamón, L. Gimeno, J.A. Campillo, B. Ferri, D.J. Abellán, I. Legaz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.V. Martínez-Sánchez, L. Gimeno, G. Server-Pastor, J. Martínez-García, J.A. Martínez-Escribano, A. Torroba, B. Ferri, D.J. Abellán, A. Minguela

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.F. Guillamón, L. Gimeno, J. Martínez-García, J.A. Martínez-Escribano, B. Ferri, M. Muro, A. Minguela

Writing, review, and/or revision of the manuscript: J.A. Campillo, J. Martínez-García, B. Ferri, M.R. Lápez-Álvarez, M.R. Moya-Quiles, A. Minguela

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B. Ferri, A. Minguela

Study supervision: G. Server-Pastor, B. Ferri, A. Minguela

We thank Dr. Valentine P. Iyemere of Bionodum Limited for critical review of the manuscript. The authors are also grateful to B. Rodriguez Martin-Gil for blood sample collection and patient management, J.M. Bolarin for statistical advice, as well as the oncologists (Drs. J.L. Alonso, P. Sanchez-Henarejos, A. Soto, A. Puertes, M.D. Gimenez, and M. Guirao), urologists (Drs. J.F. Escudero, P.A. López-González, and P. López-Cubillana), dermatologists (Drs. J. Frías, P. Sánchez-Pedreño, R. Corbalán, M. Lova, A.M. Victoria, J. Hernández-Gil, C. Brufau, A. Ramírez, M.E. Gimenez, E. Cutillas, C. Pereda, R. Rojo, P. Mercader, J.M. Ródenas, A. Peña, and J. Pardo), gynecologists (Drs. F. Barceló and B. Gomez-Monreal), and surgeons (Dr. P. Cascales-Campos and J. Gil, E. Gil) at the Clinic University Hospital Virgen de la Arrixaca (Murcia, Spain), and the oncologists (Dr. P. Cerezuela and E. Braun) at the University Hospital Santa Lucía (Cartagena, Spain) for their kind collaboration in enrolling and attending the patients. This work was funded by MINECO - Instituto de Salud Carlos III (PI1302297) and Fundación Séneca, Agencia de Ciencia y Tecnología, Región de Murcia (20812/PI/18). C.F. Guillamón was funded by the Fundación para el estudio y el desarrollo de la inmunogenética en Murcia (FEYDIM). M.V. Martínez-Sánchez was funded by the Asociación Pablo Ugarte (APU).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.