Healthy donor (male and female, age range 22-60 years) peripheral blood mononuclear cells (PBMCs) were obtained with full ethical approval from the National Research Ethics Committee (reference 06/Q1701/120). HCC patients ( Supplementary Table 1 ) were recruited from the outpatient clinic at Southampton General Hospital and provided informed consent and approved by NRES Committee South Central-Southampton-B 12/SC/0521. CLL samples ( Supplementary Table 2 ) were obtained from patients recruited in the “real world” observational study at the University of Southampton (NIHR/UKCRN ID: 31076, CI F.Forconi) following written informed consent in accordance with Ethics Committee approvals (UK National Research Ethics Service number 19/WM/0262) and the Declaration of Helsinki. Diagnosis of CLL was according to the 2008 International Workshop on CLL (IWCLL2008)/National Cancer Institute (NCI) criteria (26). Diagnosis was confirmed by a flow cytometry “Matutes score” >3 in all cases. Phenotypic and immunogenetic characteristics (tumour IGHV usage and mutational status) were determined as previously described (27). CLL patient, HCC patient and healthy donor PBMCs were cryopreserved in liquid nitrogen.

Human liver cancer cell lines (HepG2 and PLC/PRF5) were maintained in DMEM (Gibco) supplemented with 1% penicillin-streptomycin (Life Technologies) and 10% FBS (Sigma). PBMC, Raji cells and the HLA-null 721.221 B cell lymphoblastoid cell line were cultured in RPMI 1640 (Gibco) supplemented with 1% penicillin-streptomycin and 10% FBS (R10). Status of HLA-C expression of these cells lines has previously been reported (25).

NK cells were isolated from healthy donor PBMCs using the Miltenyi NK isolation kit (130-092-657) with ~95% purity ( Supplementary Figure 1 ). NK cells were cultured in expansion media consisting of NK MACS medium (Miltenyi Biotech) supplemented with 1% NK supplement (Miltenyi Biotech), 5% human AB serum (Sigma-Aldrich) and 500 IU/mL IL-2 (Miltenyi Biotech) for 14-21 days at 37°C before use in functional assays as indicated in the figure legends. Alternatively, NK cells were cultured in 10 ng/mL IL-12, 20 ng/mL IL-15 and 50 ng/mL IL-18 (R&D Systems) for 16 hours before the media was replaced with expansion media as above. After the first 5 days, expansion media was replenished every 2-3 days to maintain cells at 0.4-0.5 x 10⁶ cells/mL.

Healthy donor PBMCs stimulated overnight in 1 ng/mL IL-15 (R&D Systems), or NK cells expanded for 14-21 days were used as effector cells as indicated in the figure legends. The tumour-targeting antibodies cetuximab (human IgG1, in-house), avelumab (human IgG1, Southampton General Hospital (SGH) pharmacy), rituximab (human IgG1, in-house), obinutuzumab (human IgG1, SGH pharmacy) or isotype control (human IgG1, ChiLob7-4, in-house) were added to the target cells as indicated for 20 minutes before being washed twice with media to remove unbound antibody. Effector cells were then co-incubated with target cells at the indicated E:T ratios for 4 hours at 37°C/5% CO₂ in a 96 well round bottomed plate. Samples were then stained with 0.17 μg/mL anti-CD107a-e660 (eBioscience) and after 1 hour of co-culture, Golgistop (BD Biosciences) was added for a further 3 hours. Samples were then stained for surface markers with the following antibodies in FACS buffer (PBS, 1% BSA, 0.05% Sodium Azide) for 30 minutes at 4°C: CD3-PerCP (UCHT1, Biolegend), CD56-PE/Cy7 (HCD56, Biolegend), KIR2DL3/L2/S2-PE (CH-L, BD Biosiences) and KIR2DL3/L2-FITC (REA147, Miltenyi Biotech). For assessment of IFNγ expression, cells were permeabilised with BD Cytofix/Cytoperm (BD Biosciences) and stained with IFNγ-BV421 (Biolegend) following the manufacturer’s protocol and analysed using a BD FACS Aria II and FlowJo V10.8.1 software. Data shown in graphs were calculated by subtracting CD107a or IFNγ readings of the no target control from data with targets.

HCC and CLL patient PBMCs were used in assays on the day of thaw and were not cultured in IL-15 due to poor long-term sample viability. For experiments using KIR2DS2+ CLL donors, samples with ≥6% CD5-CD19- lymphocytes were selected and CLL donor PBMCs (250,000 cells/100 μL) were labelled with 0.17 μg/mL anti-CD107a-AF660 and incubated with indicated tumour-targeting antibodies for one hour prior to addition of Golgistop for a further 3 hours. PBMC were stained with 1 μg/mL DAPI and DAPI+ (dead) cells were removed from the analysis. This assay set-up lacked a no target control, therefore data presented in graphs are the raw CD107a and IFNγ expression values.

To assess direct cytotoxicity of the different NK cell subsets, KIR2DS2^high, KIR2DL3/L2^high, and KIR2DL3/L2/S2− CD56^dim CD3− healthy human cells were sorted using a BD FACS Aria II using the surface staining protocol described above. NK cells were then cultured overnight in R10 supplemented with 1ng/ml IL-15. 721.221 cells were stained with Cell Trace™ Violet Cell Proliferation Kit (Invitrogen) and were then co-cultured with the sorted NK cells at an effector:target (E:T) ratio of either 1:1 or 3:1 for 4 h at 37°C. After co-culture, cells were stained with propidium iodide (Invitrogen) and lysis of 721.221 cells was assessed on a BD FACS Aria II (BD Biosciences) using FACSDiva software (BD Biosciences) and analysed with FlowJo v10.7.1 (BD Biosciences).

To assess CD16 expression on patient samples, cells were stained with CD16-APC (3G8, Biolegend) in FACS buffer for 30 minutes at 4°C. To assess expression of activating receptors on the surface of NK cells before and after expansion, cells were stained with NKp30-PerCP (9E2, Biolegend), NKp46-APC (P30-15, Biolegend), NKG2D-APC/Cy7 (1D11, Biolegend) and CD57-APC (HNK-1, Biolegend) in FACS buffer for 30 minutes at 4°C. All samples were analysed using a BD FACS Aria II and FlowJo V10.8.1 software.

Full-length PCR amplification of HLA-C was performed with VeriFi DNA polymerase (PCR Biosystems, UK) using in-house primers and protocols at the Anthony Nolan Research Institute (ANRI). Amplicon size and concentration were determined using a Fragment Analyzer (Agilent). Sequencing libraries were generated using PacBio TPK3 (Pacific Biosciences) as per manufacturer’s instructions and then sequenced by a Pacific Biosciences Sequel machine (28). Analysis was performed using an in-house HLA genotyping pipeline at ANRI as described in (29, 30). KIR genotype was confirmed by polymerase chain reaction sequence-specific primer (PCR-SSP) reactions detailed in (31).

Statistical significance was determined using GraphPad PRISM software (version 10.0.2). One-way ANOVA was used to compare more than 2 groups with one independent variable and two-way ANOVA was used to compare more than 2 groups with 2 independent variables. Geisser-Greenhouse correction for unequal variability and Dunnett’s correction for multiple comparisons were applied to two-way ANOVA analyses. Data was considered statistically significant at p<0.05.

HCC has an extremely poor survival rate (32) and NK cell therapies represent a promising novel treatment approach that are currently under clinical evaluation (NCT04162158, NCT02008929, NCT05040438). KIR2DS2^high NK cells isolated from healthy donors have previously been shown to possess enhanced activation against lymphoma cells in the presence of anti-CD20 antibodies (25) in comparison to KIR2DS2- NK cells, however, it is not known whether this enhanced activity is also present for KIR2DS2^high NK cells in response to antibodies directed against solid tumour targets. Both PD-L1 and EGFR can be upregulated on the surface of HCC tumour cells (33) and antibodies targeting the PD-1: PD-L1 axis are approved in HCC (34). We therefore assessed NK activation against HCC cell lines in the presence of the anti-EGFR antibody cetuximab and the anti-PD-L1 antibody avelumab, both of which are known to induce ADCC against target cells (35, 36). HepG2 and PLC/PRF/5 HCC tumour cells were incubated with the indicated concentrations of avelumab or cetuximab as indicated prior to co-culture with healthy donor NK cells that had been primed overnight with 1 ng/mL IL-15. All healthy donors used in this study had at least one copy of HLA-C1 ( Table 1 ) and their NK cells were therefore educated through KIR2DL3/L2/S2. Using a previously published flow cytometry panel to separate NK cells based on KIR2DL3/L2/S2 expression ( Supplementary Figure 2 ) (37), KIR2DS2^high NK cells demonstrated significantly higher degranulation against HCC cells compared with KIR2DL3/L2^high and KIR2DL3/L2/S2- NK cells alone and in combination with all concentrations of cetuximab tested (p<0.01 or p<0.05) ( Figures 1A, B ). Induction of IFNγ was also higher in the KIR2DS2^high population relative to the other NK cell subsets ( Figure 1C ). Importantly, CD107a and IFNγ expression were comparable between the NK cell subsets in the absence of target cells ( Supplementary Figure 2 ). Furthermore, KIR2DS2^high NK cells demonstrated enhanced degranulation and IFNγ expression against HCC cells alone and in combination with avelumab compared to KIR2DL3/L2^high and KIR2DL3/L2/S2- NK cells (p<0.01 or p<0.05, respectively) ( Figures 1D-F ). At the higher concentrations of avelumab used, which were based on the achievable serum concentrations in patients (38), degranulation was saturated ( Figure 1E ), potentially masking the differences in ADCC between the NK cell subsets. Furthermore, in accordance with their enhanced activation status, KIR2DS2^high NK cells induced greater lysis of target cells compared to KIR2DL3/L2^high and KIR2DL3/L2/S2- NK cells ( Supplementary Figure 3 ).

Because therapeutic strategies including tumour-targeting mAbs, cytokines and NK cell engagers (NKCEs) rely on the effector functions of patient NK cells, we sought to evaluate whether the enhanced function associated with KIR2DS2 was also evident in NK cells isolated from patients with cancer. PBMCs isolated from HCC patients were co-cultured with target cell lines, in the presence or absence of the indicated ADCC-inducing antibodies. Against HepG2 cells, KIR2DS2^high NK cells derived from the peripheral blood of HCC patients (n=5) demonstrated enhanced activation compared to KIR2DL3/L2/S2- NK cells (p<0.01) in both the presence and absence of avelumab ( Figures 1G, H ). Significantly higher activation was also observed compared to KIR2DL3/L2^high NK cells (p<0.05) in the presence of avelumab ( Figure 1H ). In addition, KIR2DS2^high NK cells from HCC patients had enhanced activation against HLA-deficient 721.221 target cells both in the absence and presence of the type II anti-CD20 mAb obinutuzumab (p<0.01) compared to KIR2DL3/L2^high and KIR2DL3/L2/S2- NK cells ( Figures 1I, J ). This indicates that the enhanced activation associated with KIR2DS2 is not solely due to less inhibitory signalling compared to KIR2DL3/L2^high cells or binding of KIR2DS2 to HLA-C. This is in accordance with data previously observed using healthy donor derived NK cells (25). Of the three HCC patients tested, CD16 expression was not significantly higher on KIR2DS2^high NK cells compared to KIR2DL3/L2^high and KIR2DL3/L2/S2- NK cells ( Figures 1K, L ), indicating that the enhanced ADCC of the KIR2DS2^high population was not simply due to greater expression of CD16.

To determine the antibody-dependent activation of KIR2DS2^high NK cells against primary tumour samples, we used peripheral blood derived chronic lymphocytic leukaemia (CLL) cells as targets in combination with the clinically relevant anti-CD20 antibodies rituximab and obinutuzumab. Obinutuzumab is known to induce ADCC more potently than rituximab due to its enhanced affinity for CD16 (38). In the presence of these antibodies, KIR2DS2^high NK cells from healthy donors primed overnight with 1 ng/mL IL-15 showed significantly higher CD107a (p<0.0001) ( Figures 2A, B ) and IFNγ (p<0.0001, p<0.001 or p<0.01) ( Figure 2C ) expression compared to the other NK cell subsets in accordance with previous reports against lymphoma cell lines (25). Furthermore, as the lymph nodes are a key tissue site for CLL proliferation and drug resistance in patients (39), we tested the activation of the NK cells in the presence of signals mimicking the lymph node microenvironment (CD40L and IL-4). KIR2DS2^high NK cells retained enhanced activation in the presence of CD40L and IL-4 (p<0.01 or p<0.05) ( Figure 2D ).

Subsequently, we assessed the function of KIR2DS2^high NK cells from PBMC isolated from CLL patients against their own autologous tumour cells ( Figure 2E ). Six CLL patients positive for KIR2DS2 expression with sufficient (≥6%) healthy CD5-CD19- lymphocytes were identified by flow cytometry (37). Patient PBMC were incubated with rituximab or obinutuzumab as indicated before assessment of NK cell degranulation by flow cytometry in the CD3-CD56+ lymphocyte population. KIR2DS2^high NK cells from CLL patients showed significantly enhanced activation in the absence of anti-CD20 mAb compared to KIR2DL3/L2/S2- (p<0.05) but not KIR2DL3/L2^high NK cells ( Figures 2F, G ). In the presence of rituximab, enhanced activation of the KIR2DS2^high population was evident compared with both KIR2DL3/L2/S2- (p<0.01 for 0.01 and 0.1 μg/mL, p<0.05 for 1 μg/mL) and KIR2DL3/L2^high NK cells (p<0.01 for 0.01 μg/mL, p<0.05 for 0.1 μg/mL) ( Figure 2G ). In the presence of obinutuzumab, KIR2DS2^high NK cells showed enhanced activation compared to KIR2DL3/L2/S2- cells (p<0.05 for 0.01 μg/mL, p<0.01 for 0.1 ug/mL), but did not reach statistical significance compared to KIR2DL3/L2^high NK cells ( Figure 2G ). Finally, similar to HCC patient NK cells, levels of CD16 expression were comparable between KIR2DS2^high and KIR2DL2/L3^high NK cells, however CD16 was significantly higher on KIR2DS2^high cells compared to KIR2DL3/L2/S2- cells (p<0.05) ( Figures 2H, I ). These data indicate that KIR2DS2 expression is associated with a population of NK cells in cancer patients that have superior activation against target cells both in the presence and absence of tumour-targeting antibodies.

Ex vivo expansion is required for adoptive NK cell therapies. We therefore assessed whether the enhanced activation of KIR2DS2+ NK cells was retained following expansion ex vivo and could therefore potentially be used to select a population of NK cells with enhanced anti-cancer activity. NK cells were isolated from the peripheral blood of healthy volunteers, mimicking the source of NK cells in allogeneic adoptive transfer strategies (7). The NK cells were then expanded in Miltenyi Biotech NK expansion medium with human IL-2 for 14-21 days ( Supplementary Figure 4 ). The Miltenyi Biotech NK expansion method has recently been shown to have potential for the clinical production of expanded NK cells (40). Expanded NK cells were then co-cultured with HepG2 cells for 4 hours in combination with avelumab ( Figure 3A ). After expansion, KIR2DS2^high NK cells lost their enhanced activation relative to the KIR2DL3/L2/S2- subset but not the KIR2DL3/L2^high subset as measured by both degranulation (p<0.05 KIR2DS2^high vs KIR2DL3/L2^high) and IFNγ (p<0.05 KIR2DS2^high vs KIR2DL3/L2/S2-) expression ( Figures 3A-C ). In accordance with the data against HepG2 target cells, KIR2DS2^high NK cells also lost their enhanced activation over the KIR2DL3/L2/S2- but not the KIR2DL3/L2^high (p<0.01 for control, p<0.05 for obinutuzumab) subset of NK cells against CD20+ malignant B cells (Raji) following ex vivo expansion ( Figures 3D-F ). Expression of the activating receptors NKp30, NKp46 and NKG2D was significantly increased on the surface of each subset of NK cells after ex vivo expansion in IL-2 for 14 days compared to day 0 (p<0.0001 for NKp30 and NKG2D, p<0.001 for NKp46), although there were no significant differences in expression of these receptors between NK cell subsets on day 0 or day 14 ( Supplementary Figures 5A-C ). In addition, CD57 was consistently higher on KIR2DS2^high NK cells compared to KIR2DL3/L2/S2- NK cells and decreased on all three NK cell subsets after expansion in IL-2 ( Supplementary Figure 5D ). To determine if the loss of enhanced activation in the KIR2DS2^high NK cell subset also occurred with other methods of NK expansion, we tested the effect of an expansion method utilizing IL-12/15/18. The IL-12/15/18 cytokine cocktail has been shown to generate cytokine-induced memory-like NK cells and is currently in clinical trials for patients with cancer (41, 42). In accordance with the results from the previous experiments, KIR2DS2^high NK cells expanded in IL-12/15/18 did not have higher activation compared to the KIR2DL3/L2/S2- subset in combination with rituximab and obinutuzumab ( Figures 3G-I ).

To determine whether this effect was due to inhibitory KIR: HLA interactions, we tested ex vivo expanded NK cell activation against the HLA-deficient B cell lymphoblastoid cell line 721.221. In this setting, KIR2DL3/L2^high NK cells showed a trend for higher activation compared to KIR2DS2^high NK cells and KIR2DL3/L2/S2- NK cells at both a 10:1 and 20:1 E:T ratio ( Figures 3J-L ). Similar results were obtained for KIR2DS2^high NK cells expanded in IL-12/15/18 ( Figures 3M, N ). Overall, these results demonstrate that the superior effector function associated with KIR2DS2+ NK cells may not be applicable in an adoptive transfer setting that requires ex vivo expansion of NK cells.