PB samples from healthy adult donors were provided by the Transfusion Center-IRCCS Ospedale Policlinico San Martino (Genoa, Italy). The donors gave their informed consent to participate in this study, approved by the Ethics Committee of the Liguria Region (Protocol no. 39/2012, number CER Liguria: DB ig 10125), in accordance with the Declaration of Helsinki.

CMV serology was evaluated on plasma samples obtained from donors PB, using an enzyme immunoassay for virus-specific IgG immunoglobulins (CMV IgG ELISA Kit, Technogenetics, Lodi, Italy).

PB mononuclear cells (PBMCs) were separated by density gradient centrifugation (Ficoll-Hypaque solution, Sigma-Aldrich, St. Louis, MO), analyzed by immunofluorescence and flow-cytometry to evaluate the donor eligibility (CMV serostatus, NK cell frequency in PBMC over 10%) and frozen. Subsequently, PBMC from selected donors (see section 3.1) were thawed and processed as indicated below.

The mAb used in this study recognizes the CD94/NKG2C and CD94/NKG2A heterodimeric complexes on human cells by interacting specifically with the CD94 molecule and was generated and characterized in our laboratory (38–40). The purified mAb used in cell cultures was obtained from supernatants collected after hybridoma culture with serum-free medium (CD Hybridoma, Gibco, ThermoFisher Scientific, Waltham, Massachusetts, USA). MAb purification was performed by affinity chromatography using HiTrap Protein G HP columns (GE Healthcare, Chicago, Illinois, USA).

The medium used in cultures and throughout the experiments was RPMI-1640, supplemented with 2mM L-glutamine, 1% penicillin-streptomycin, 10% heat-inactivated Fetal Bovine Serum (FBS, Gibco). Recombinant human (rh) IL-2 (Proleukin, Chiron, Emeryville, CA) or rhIL-15 (Peprotech, London, UK) was added to the medium where indicated (from now on, rh will be omitted).

The protocol developed after testing different experimental conditions, requires the following steps:

In particular, PBMCs obtained from selected donors were depleted of T lymphocytes by positive selection with anti-CD3 magnetic microbeads (MACSmicrobeads Miltenyi Biotec, Bergisch Gladbach, Germany). T cell-depleted PBMCs, enriched in NK cells, were cultured as depicted in Figure 1B . Briefly, T cell-depleted PBMCs were cultured in the presence of IL-2 or IL-15 at the final concentrations of 300 IU/ml and 20 ng/ml respectively, in the presence or in the absence of the anti-CD94 mAb, at a final concentration of 0.125 μg/ml, a dosage chosen after testing a range of concentrations from 0.0125 μg/ml to 5 μg/ml. The mAb was added to cultures only at day 0. Cells were seeded at a concentration of 5x10⁵/ml in round 96-well plates (10⁵ post-depletion cells/well). On day 5, fresh medium containing the correct cytokine was added to cultures, while on days 7 and 10 a 40%-50% change of medium was carried out. On days 5, 7, 10 and 14, cells were collected, counted (Trypan Blue, Sigma-Aldrich was used to exclude dead cells) and characterized by flow-cytometry analyses as described in the text.

In each experiment a fraction of T cell-depleted PBMCs was labeled with the intracellular fluorescent dye 5 (6)-carboxyfluorescein diacetate succinimidyl ester (CFSE, Molecular Probes, Life Technologies, Thermo Fisher Scientific, California, USA), as previously described (28), and cultured according to the protocol shown in Figure 1B . On the indicated days, the cells were subjected to flow cytofluorimetric investigations to evaluate their proliferative abilities indicated by CFSE dilution.

The different mAbs, as well as the isotype-specific secondary reagents used in flow-cytometry analyses, are shown in Supplementary Table S2 ( Supplementary Table S2 ). See Supplementary File for details on flow-cytometry procedures and analyses.

On day 14 of culture, the cells expanded in the presence of cytokine+mAb were collected, washed, and incubated with different cell lines at an effector NK cell:target cell ratio of 1:1 for 3 hours, in culture medium supplemented with anti-CD107a mAb-PE as previously described (41). In particular, the following cell lines were used: K562, NALM-16 (B-Cell Precursors Acute Lymphatic Leukemia cell line), HT-29 and LoVo (colorectal cancer cell lines), HL-60 (Acute Myeloid Leukemia cell line), Raji previously opsonized or not with the anti-CD20 mAb Rituximab (1 μg/ml), LCL 721.221 wild type (221wt) and LCL.721.221AEH (221E) (i.e., LCL 721.221 transfected to express HLA-A leader sequence), kind gift from Prof. Miguel Lopez-Botet and Dr. Aura Muntasell (Hospital del Mar Medical Research Institute IMIM and University Pompeu Fabra, Barcelona, Spain), originally developed by Geraghty D. and coworkers (42). Basal degranulation of NK cells was assessed in 3-hour cultures without tumor cell lines. Subsequently, cells were collected, stained and analyzed by flow-cytometry as described in Supplementary Methods .

To detect NK intracellular production of IFN-γ and TNF-α, on day 14 mAb-expanded NK cells were washed and incubated with different cell lines. i.e. K562, 221wt, 221E, Raji rituximab-coated or not, or medium alone, an effector NK cell:target cell ratio of 1:1 for 4 hours in the presence of GolgiStop (BD Biosciences Pharmingen). Thereafter, cells were washed, stained as described above for CD107a assays and then fixed and permeabilized with Foxp3 permeabilization buffer Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). IFN-γ and TNF-α production was detected by subsequent intracellular staining with PE-labeled mAbs and cytofluorimetric analysis.

Nonparametric Wilcoxon-Mann-Whitney tests were used. Statistical significance (p value) is indicated in the different graphs (*p<0.05; ** p<0.01; *** p<0.001; **** p<0.0001). Median Fluorescence Intensity values were normalized before calculating statistical significance. Graphs and statistical analyses were performed with GraphPad Prism 8 (GraphPad Software, www.graphpad.com).

To set up a method efficiently expanding NKG2C⁺ NK cells, we first collected and stored PBMC from several CMV⁺ healthy donors (HD); the selected donors were characterized by a percentage of CD56^dimNKG2C⁺ NK cells close to the median frequency observed (26%, Figure 1A ). The gating strategy is shown in Supplementary Figure 1A ( Supplementary Figure S1A ). Indeed, the percentage of the NKG2C⁺ NK cell population in PB of CMV⁺ HD ranged from 1 to 82% ( Figure 1A ): most NKG2C⁺ NK cells coexpressed the marker of terminal differentiation CD57, representing the adaptive NKG2C⁺CD57⁺ NK cell subset (median=23%, range =1-80%, Figure 1A ), while NKG2C⁺ NK cells lacking CD57 were present at lower frequencies in most CMV⁺ HD (median=6%, range=0-32%). In CMV-seronegative (CMV^–) HD, low amounts of NKG2C⁺ NK cells could be observed (median=6%, range=1-16%, Figure 1A ) (19, 28).

After testing different experimental conditions, we developed an efficient expansion method summarized in Figure 1B . The protocol requires three steps, i.e. PBMC isolation from selected CMV⁺ HD, T-cell depletion to obtain NK-enriched cell populations and 14-day culture in the presence of a specific anti-CD94 mAb, with IL-2 or IL-15.

We selected a mAb produced in our lab that recognizes CD94, the non-transducing monomer that associates with NKG2A or NKG2C to form the inhibitory or activating heterodimeric NK cell receptor, respectively (39, 40). The mAb-mediated engagement of CD94 induces inhibitory signals in CD94/NKG2A⁺ NK cells while triggering activating signals in CD94/NKG2C⁺ NK cells (43, 44). Notably, at variance with previously described methods, our culture protocol did not include stimuli like feeder cells or genetically manipulated HLA-E⁺ transfected cell lines (34, 35, 37). However, a regular addition or substitution of medium with fresh cytokines every 2-3 days was required ( Figure 1B ).

As shown in Figure 1C , upon T-cell depletion and before expansion (day 0), the cell population included NK cells (mean=32%; range 25-52%), B cells (mean=22%; range 10-34%), and monocytes (mean=41%; range 20-60%), while T cells were present at negligible frequencies (mean=0, range 0-1%) (see Supplementary Figures S1A, B for gating strategy). Following the mAb-mediated NK stimulation, the frequency of NK cells progressively increased, representing virtually the only cell type found in cultures on days 10 and 14; this occurred in the presence of either IL-2 ( Figure 1C left) or IL-15 ( Figure 1C right). In particular, the mAb addition on day 0 significantly increased the frequency of NKG2C⁺ NK cells in both cytokine conditions, driving the selective expansion of NKG2C⁺ NK cells that were predominant on day 14 ( Figure 2A left). A large fraction of NKG2C⁺ NK cells expanded with cytokine+mAb were represented by adaptive NKG2C⁺ CD57⁺ NK cells ( Figure 2A middle), while limited numbers of NKG2C⁺ NK cells lacking CD57 were detectable ( Figure 2A right). In parallel cultures, T-depleted PBMC were labeled with CFSE and stimulated with mAb+either of the two cytokines; the progressive expansion of NKG2C⁺ NK cells, predominantly expressing CD57 (right plots) was also clearly demonstrated, showing a tendency to proliferate more rapidly when IL-15+mAb was used, compared to IL-2+mAb (see representative donor in Figure 2B ).

Importantly, the NKG2C⁺ NK cell numbers observed upon 10-14 days of culture in the presence of anti-CD94 mAb were significantly higher as compared to cultures carried out without the mAb ( Figure 2C ). We observed a tendency to obtain higher numbers of NKG2C⁺ NK cells in the presence of IL-15+mAb ( Figure 2C right) compared to IL-2+mAb ( Figure 2C left) as also suggested by the fold expansion rate reported in Figure 2D for the different culture conditions. We did not observe any further cell number increase after day 14 of culture, although NKG2C⁺ NK cells were able to survive longer. Finally, when we cultured T-cell depleted PBMCs from CMV^– HD using the same protocol, we did not succeed in expanding high numbers of NKG2C⁺ NK cells (not shown). Interestingly, we observed that the mAb could be eliminated from the culture 5 days after cell seeding, without affecting the expansion of adaptive NK cells (not shown).

The prolonged culture of NK cells in the presence of activating cytokines such as IL-2, IL-12 and IL-15, can induce upregulation of NKG2A and other inhibitory molecules that could decrease the therapeutic potential of NKG2C⁺ NK cells (36, 45). We assessed the NKG2A expression in our culture system (mAb+cytokine) and we did not detect any significant increase of CD94/NKG2A⁺ NK cells, which were even significantly lower if comparing NK cells on day 0 and day 14 ( Figure 3A left). In addition, we did not observe any expansion/induction of NK cells coexpressing CD94/NKG2C and CD94/NKG2A ( Figure 3A center). It is thus reasonable to speculate that the CD94 engagement by triggering the NKG2A-mediated inhibitory signal prevented the expansion of both NKG2A⁺ NKG2C^– and NKG2A⁺NKG2C⁺ NK cells. Conversely, it efficiently promoted the expansion of NKG2C⁺ cells lacking NKG2A ( Figure 3A right).

Besides NKG2A, the cytotoxic activity of NK cells can be dampened by other ICs whose expression can be induced or increased following NK cell activation. We focused on the expression of PD-1 and other relevant ICs (i.e., TIM-3, LAG-3 and TIGIT) on expanded NKG2C⁺ NK cells. At variance with other methods, neither the cytokine nor the mAb+cytokine stimulation induced PD-1 expression in NKG2C⁺CD57⁺ NK cells ( Figure 3B ). On the other hand, both TIM-3 and LAG-3, virtually absent on day 0, were significantly upregulated in a relevant fraction of NK cells by exposure to IL-2 or IL-15. However, neither the expression of these ICs nor the frequency of TIM-3⁺ and LAG-3⁺ cells increased significantly upon the addition of the mAb ( Figure 3B ). Finally, TIGIT expression could be detected already at day 0 on most adaptive NK cells and its expression level slightly increased at day 14 after expansion independently from mAb addition, although reaching statistical significance only with IL-2 ( Figure 3B right).

As mentioned previously, CMV-driven adaptive NK cells are characterized by the expression of self iKIRs, i.e., iKIRs that recognize HLA-I molecules expressed by the individual’s healthy cells. In a subgroup of donors, the KIR and KIR-ligand genotypes were analyzed and the expansion of single self iKIR⁺NKG2A^–NKG2C⁺ NK cells upon cytokine+mAb stimulation was assessed. The gating strategy is shown in Supplementary Figure S1C . As shown in Figure 3C (left), the frequency of NKG2A^– NK cells expressing a single self iKIR significantly increased (KIR2DL1 in 1 donor, HD A; KIR2DL3 in 3 donors, HD B, C, and F; KIR2DL2/L3 in 2 donors, HD D and E, Supplementary Figure S2 ). On day 0, virtually all NKG2C⁺CD57⁺ NK cells were NKG2A^– single self iKIR⁺, a profile that was maintained upon mAb-mediated expansion ( Figure 3C center). The frequency of NKG2A^– single self iKIR⁺ NK cells significantly increased on day 14 also in the NKG2C⁺ CD57^– NK cell subset, whose profile was more heterogeneous on day 0 ( Figure 3C right). In line with previous studies, we could not find single self KIR3DL1⁺ NKG2C⁺ NK cells either before or after expansion (not shown) (22, 25, 26). Overall, these observations show that our mAb+cytokine protocol selectively expands NKG2C⁺ NKG2A^– single self iKIR⁺ NK cells without inducing de novo PD-1 expression or further upregulating cytokine-induced TIM-3 and LAG-3 expression.

A variable fraction of adaptive NKG2C⁺ NK cells lacks the expression of the adaptor proteins FcεRγ and Syk due to epigenetic modifications induced by CMV infection ( Figure 4A day 0) (22, 25, 26). Following mAb-mediated expansion, this peculiar characteristic was maintained, as indicated by the frequency of NKG2C⁺ CD57⁺ FcεRγ^– and Syk^– NK cells at day 14 ( Figure 4A day 0 vs 14 and representative experiment in Supplementary Figure S3A ). The expanded NKG2C⁺ CD57⁺ NK cells normally expressed the signaling molecule CD3ζ and the activating receptor CD16, although CD16 MFI was reduced in cultured cells as reported also in other expansion methods (34, 35, 37) ( Figure 4A ; Supplementary Figure S3 ), and were characterized by high levels of perforin and granzyme B, as expected for highly differentiated NK cells (not shown) (2). Importantly, the main activating NK receptors (i.e., NKp46, NKp30, NKp44, NKG2D, and DNAM-1) expression was maintained or upregulated compared to day 0 ( Figure 4B ; Supplementary Figure S3B ). Notably, mAb-expanded adaptive NKG2C⁺CD57⁺ NK cells were highly functional in degranulation assays against tumor cell lines of different hystotypes, i.e. against K562, a classic NK cell susceptible HLA-I negative target, Lovo (HLA-I^-), but also against HLA-I⁺ tumor cells as NALM-16 and HT-29 ( Figure 4C IL-2+mAb and 4D IL-15+mAb).

As described previously, the lower FcεRγ expression in adaptive NK lymphocytes is associated with strong ADCC capabilities. Indeed, adaptive NKG2C⁺CD57⁺NKG2A^– self iKIR⁺ FcεRγ^–/low NK cells showed high degranulation levels in response to IgG-opsonized tumor cells (Rituximab-coated Raji), after expansion with mAb and IL-2 or IL-15 ( Figures 4C, D and representative experiment in Supplementary Figure S4 ). Furthermore, such expanded NK cells were found to be highly functional also against a HLA-E⁺ transfected cell line (221E, Supplementary Figure S5 ; Figures 4C, D , Supplementary Figure S4 ). The high response against the HLA-E⁺ target indicates that the expression and function of the activating receptor CD94/NKG2C are well preserved during the CD94 mAb-driven expansion. The NKG2C⁺ CD57^– NK cell subset that moderately expanded in our experimental system ( Figure 2A ), displayed similar functional abilities (not shown). A modest degranulation was observed also when mAb+IL-15-induced adaptive NK cell expansions were tested against primary myeloid leukemic blasts ( Supplementary Figure S6 ). Notably, degranulation levels were increased upon mAb-mediated masking of HLA-I molecules, by interrupting inhibitory KIR-HLA-I interactions between expanded adaptive NK cells and primary AML. This observation confirms the relevance of choosing a suitable donor mismatched with the HLA-I haplotype of the patient to achieve the highest killing potential of KIR⁺ adaptive NK cells.

Interestingly, mAb-expanded adaptive NKG2C⁺CD57⁺ NK cells were also capable of producing IFN-γ and TNF-α in response to K562, 221E and Rtx-Raji, with higher efficiency when NK cells were expanded in the presence of IL-15 ( Figure 4E ).