All reagents used in this paper and their source are listed in Supplementary Table 1 .

rSPs from different VOC derive from the following lineages: Ancestral Wuhan strain (lineage A), Alpha or G614 (lineage B.1.1.7), Indian Delta-Plus (lineages B.1.617.2.1 with mutation K417N), and Omicron (lineage B.1.1.529.2). For simplicity, Wuhan, G614, Delta-Plus, and Omicron were throughout the article.

PB was obtained from buffy coat of healthy donors (HD) admitted to the blood transfusion service of Bambino Gesù Children’s Hospital, Rome, Italy, after obtaining informed consent. The ethical committee approved the study that was conducted according to the Declaration of Helsinki tenets (2058_OPBG_2020). In this study, we also used peripheral blood mononuclear cells (PBMCs) from 11 informed subjects who showed recent (more than 3 months) mild infection/recovery (clinically and antigenically documented) from COVID-19 (Covid+) and from six informed individuals with no clinical and serological signs (absence of anti-nucleocapsid antibodies) of the disease (Covid−). At the sampling time, Covid+ subjects were antigen-negative. All selected individuals had received at least three doses of Comirnaty mRNA vaccine Pfizer/BioNTech (New York, USA), were negative for HBV, HCV, and HIV infections; and were cancer free.

Human NK cells (>98% CD3⁻CD56⁺) were isolated from PBMCs as previously described (25, 26). Briefly, following PBMC isolation by density Ficoll-Paque Plus, cells were resuspended in RosetteSep NK-cell enrichment cocktail (STEMCELL Technologies cat #15065) and NK cells were negatively selected by density centrifugation. Fresh NK cells were cultured in RPMI-1640 medium supplemented with 2 mM L-glutamine (Euroclone), 1% penicillin and streptomycin (Euroclone), and 10% fetal bovine serum (FBS; Thermo-Fisher Scientific, Massachusetts, USA). For masking experiments, NK cells were incubated with neutralizing antibodies (2 µg/mL, see Supplementary Table 1 ), in culture medium for 1 h followed by further stimulation with VOC-rSPs for 72 h at 37°C. At the end of culture, cell pellets were harvested and supernatants were collected for further analysis.

In order to exclude myeloid contaminants in NK cell preparations, evaluation of CXCL10, a typical cytokine produced by myeloid cells, was performed in the culture supernatants by ELISA.

The NALM-18 (B-cell childhood acute lymphoblastic leukemia) cell line was kindly provided by Dr. Pende D (IRCCS, Policlinico San Martino, Genoa, Italy), while HEK-293 (human embryonic kidney) Blue Null1, TLR2, and TLR4 reporter cell lines were purchased from InvivoGen (San Diego, California, USA). These cell lines were grown in RPMI-1640 (Euroclone, Milan, Italy) and DMEM, respectively, supplemented with 2 mM L-glutamine (Euroclone), 1% penicillin and streptomycin (Euroclone), and 10% FBS (Thermo-Fisher Scientific, Massachusetts, USA). HEK-293 Blue Null expression and TLR2 and TLR4 expression on reporter cell lines were maintained according to manufacturer’s specifications.

Freshly isolated NK cells were incubated with biotinylated spike protein (2.5 µg/mL Biotechne, Minneapolis, USA) and streptavidin PE [Becton Dickinson (BD), New Jersey, USA] in PBS at 4°C for 1 h. This mix was previously kept at 4°C for 1 h. Afterwards, cells were stained with fluorescent-labeled antibodies for flow cytometry detection as specified in Supplementary Table 1 (Reagent list).

Surface staining and intracellular staining of VOC-rSP-stimulated NK cells were performed with fluorescent-labeled antibodies to membrane molecules or cytokines, as previously described (25). Evaluation of apoptosis was carried out with FITC-conjugated Annexin V (BD, New Jersey, USA) and propidium iodide (PI; Sigma-Aldrich, Darmstadt, Germany) as previously described (26). Analysis of specific staining was performed by Cytoflex LX flow cytometer cytexpert 2.4 (Beckman Coulter, Brea, California.) and FlowJo Software (BD) and Kaluza software (Beckman Coulter).

NK cell cytotoxic activity was evaluated by a flow cytometric assay by using NALM-18 cells as targets, as previously described (25, 27). For CD107a expression, either untreated or Wuhan/G614-treated NK cells (72 h) were co-cultured for 4 h with NALM-18 target cells, the effector:target ratio being 1:1. Anti-CD107a eFluor 660 (Invitrogen) and GolgiStop (1:500) were added during stimulation. Afterwards, cells were stained with fluorescent-labeled antibodies for flow cytometry detection as specified in Supplementary Table 1 (Reagent list). For intracellular staining, cells were permeabilized using FOXP3 Staining buffer set (Miltenyi Biotec).

The presence of IFN-γ, TNF-α, and CXCL10 was investigated in the supernatants from NK cell cultures by DuoSet ELISA kits (R&D Systems Inc, Minneapolis, USA) according to the manufacturer’s instructions (Reagent list, Supplementary Table 1 ) The cutoff limits of the assays were as follows: 18.76 pg/mL for IFN-γ, 15.6 pg/mL for TNF-α, and 31.2 pg/mL for CXCL10.

Specific stimulation of human TLRs was carried out in HEK-293 Blue Null1, TLR2, and TLR4 reporter cells by measuring the activation of NF-κB (InvivoGen) (28), according to the manufacturer’s instructions.

TLR2 and TLR4 silencing experiments on freshly isolated NK cells were performed as previously described with minor changes (27). TLR2, TLR4, and negative control small interfering RNA (siRNA) are reported in Supplementary Table 1 . Following 48-h incubation, cells were activated with Wuhan-rSP (2.5 µg/mL) for 24 h in culture medium. Following this, cell pellets were processed for Western blot analysis whereas supernatants were collected for further analysis.

Total RNA extraction, reverse transcription, and gene expression studies on bulk, untreated, Wuhan- and G614-rSP-stimulated (6h) or transfected freshly purified NK cells were performed as previously described (25, 27). Total RNA extraction from NK cells was performed with an RNeasy Plus mini kit (Qiagen GmbH, cat #74134) following the manufacturer’s protocols. RNA was reverse-transcribed with oligo-dT primers using the Super Script IV first-strand synthesis (Invitrogen, cat #18091050) system following the manufacturer’s instructions. Reactions were carried out on a QuantStudio 7 Flex instrument using the recommended PCR cycling conditions. The following primers were used for TLR-2: forward 5′-CTT CAC TCA GGA GCA GCA AGC A-3′, reverse 5′-ACA CCA GTG CTG TCC TGT GAC A-3′; TLR-4: forward 5′-TGG ACC TGA GCT TTA ATC CC-3′, reverse 5′-GTC TGG ATT TCA CAC CTG GAT-3′, GAPDH: forward 5′-TCTTTTGCGTCGCCAGCCGA-3′, reverse 5′-ACCAGGCGCCCAATACGACC-3′.

PCR array was performed in custom 384-well TaqMan array microfluidic cards (Thermo Fisher Scientific, cat #4342259) to detect selected genes implicated in NK cell biology (ID reported in Supplementary Table 2 ). Analysis of PCR array data was performed by using relative threshold algorithm and normalized using the mean of β-actin expression, used as reference gene (TaqMan assay ID Hs99999903_m1, Applied Biosystems and Thermo-Fisher Scientific).

Whole-cell extracts from untreated, mock-, or TLR2/TLR4-silenced NK cells were examined to detect TLR2 and TLR4 and phosphorylated/total NF-kB, respectively, as previously described (29).

Data were analyzed by GraphPad Prism V.9.3.1 (San Diego, California). Reported values have been plotted as means ± SEM of independent experiments. The exact number of each experiment is reported in the figure legends. If not otherwise specified in the legend, statistical analysis was performed using paired Student’s t-test with Bonferroni adjustment and ANOVA Kruskal–Wallis multiple comparison test.

To establish whether rSP can directly interact with NK cells, we evaluated the binding of biotinylated rSP (Wuhan strain) on freshly purified NK cells. Even though a high variability was observed, biotinylated rSP bound the surface of a similar proportion of CD56^bright and CD56^dim NK cells ( Figure 1A ), mostly associated with NKG2A and NKG2C expression on the former and NKG2A and CD57 on the latter subset ( Figure 1B ).

To establish whether rSP could stimulate NK cell function and also to optimize the experimental conditions, we assessed the activation (CD25) and apoptosis (Annexin V) markers’ expression on NK cell-gated (CD3^-CD56⁺) rSP-stimulated PBMCs from HD, enrolled irrespective of their previous infection and/or vaccination. The time-course experiments indicated that, upon Wuhan-rSP stimulation, the proportion of activated NK cells was significantly increased, peaking at 72 h ( Supplementary Figure 1A ), with the optimal rSP dose established as 2.5 μg/mL ( Supplementary Figure 1A ). Such conditions were used in all experiments, unless otherwise specified. In addition, Annexin V expression was not changed upon the same experimental conditions ( Supplementary Figure 1B ), thereby suggesting that apoptosis was poorly affected by rSP ( Supplementary Figure 1B ).

In order to assess whether rSP exerted direct activity on the NK cell function, purified (>98%) NK cells obtained from PBMCs of HD were stimulated with rSP from different VOCs (Wuhan, G614, Delta-plus, and Omicron). Activated cells were then analyzed for phenotype, activation marker expression, cytokine release, and cytotoxic activity.

rSP (G614) stimulation did not modify the CD57 and NKG2A expression on NK cells, whereas it slightly, but significantly, increased that of NKG2C, without, however, changing the proportions of different NK cell subsets ( Figure 1C ). In the presence of all VOCs, both CD56^bright and CD56^dim NK cells displayed higher levels of surface CD69 than controls ( Supplementary Figure 1C ). Notably, the recombinant S1 subunit, more than the S2 subunit, was highly effective in inducing NK cell activation ( Supplementary Figure 2A ), whereas, among SARS-CoV-2 glycoproteins (gps), only E gp slightly increased CD25 expression on CD56^bright NK cells ( Supplementary Figure 2B ).

All VOC-rSPs induced a statistically significant increase in the frequency of cells expressing the activation markers CD25 and CD69 but not of HLA-DR (not shown) as compared to untreated NK cells. Of note, upon treatment with G614, Delta-plus, and Omicron, higher proportions of NK cells expressing activation markers were observed as compared to the Wuhan strain. This effect was more evident in the CD56^bright than in the CD56^dim NK cell subset ( Figure 1D ; Supplementary Figure 1C and data not shown). Of note, rSPs from VOC-G614 and Delta-plus induced activation of NK cells in the great majority of HD assayed ( Supplementary Table 3 ). In agreement with the above data, VOC-rSP-stimulated NK cells exerted significantly higher cytotoxic activity (left panels) and IFN-γ release (right panel) as compared to untreated NK cells upon co-culturing with an NK-sensitive cell line (NALM-18) ( Figure 2A ). Notably, rSP from both G614 and Delta-plus strains determined a significant increase of the IFN-γ release compared to the Wuhan strain ( Figure 2A , right panel). Among NK cell subsets, VOC-rSP-stimulated CD56^bright NK cells displayed higher CD107a and IFN-γ expression compared to CD56^dim NK cells ( Figures 2B, C ).

To establish which receptors are involved in rSP-NK cell binding and activation, we initially focused on ACE2 and the main NK cell activating receptors. NK cells did not express ACE2 at different time points of rSP stimulation ( Supplementary Figure 2C ). In addition, the masking experiments with neutralizing anti-ACE2 monoclonal antibody (mAb) did not affect CD25 and CD69 expression on NK cells induced by VOC (Wuhan, Delta-plus, G614)-rSP treatment ( Supplementary Figure 2D ). CD25/CD69 expression, as well as IFN-γ production on rSP-stimulated NK cells, was not inhibited by NKp46, NKp30, and NKp44 (NCR) and LFA1 neutralizing mAbs ( Supplementary Figures 3A, B ). Likewise, the expression on NK cell subsets of NKp46, NKp30, NKp44, LFA1, DNAM-1, and NKG2D was not modified by activation with VOC-rSPs (data not shown).

To further investigate the nature of the NK molecule(s) engaged by rSP, we evaluated the gene activation and the phosphorylation of some transcriptional factors upon rSP stimulation. The results of a customized CARD gene array performed on NK cells from PBMCs of three HD shortly (6 h) cultured with VOC (G614 and Wuhan)-rSPs indicated that several genes encoding for NK cell receptors, chemokines, signal transduction, transcription factors, and apoptosis were upregulated in stimulated compared to non-stimulated NK cells ( Figure 3A ). In particular, IRAK1, IRF3, and IKBKG genes (intermediate and terminal factors of the surface TLR signaling cascade) were highly upregulated ( Figure 3A ). Moreover, the same rSPs were able to induce the phosphorylation of NF-kB, strengthening the hypothesis of the possible role of surface TLR engagement ( Figure 3B ). Since in silico and structural studies indicated the direct interaction of rSP with TLR2 and TLR4 (30, 31), we checked the direct binding of all VOC-rSPs to TLR2 or TLR4 stable-transfected HEK-293 cell lines. rSPs from all VOCs bound both TLR2 and TLR4 with different binding capacity: Wuhan and Delta-plus rSPs bound TLR4 more than TLR2, while the inverse occurred for G614-rSP ( Figure 3C ).

On freshly isolated NK cells, TLR2 and TLR4 were expressed at the mRNA ( Figure 3D ) and protein (Western blotting) ( Figure 3E ) levels. However, the cytofluorimetric analysis indicated that both TLRs were poorly detectable on CD56^bright and CD56^dim NK cells and their expression was unchanged after rSP stimulation ( Supplementary Figures 4A, B ).

In agreement with these data, subsequent experiments demonstrated that neutralizing anti-TLR2 and -TLR4 antibodies inhibited the binding of biotinylated-SP (Wuhan strain) to freshly isolated NK cells ( Figure 4A ). More importantly, the combination of the two neutralizing antibodies resulted in the strongest inhibition of the CD25/CD69 expression compared to the partially and variably downregulation observed upon treatment with single neutralizing Abs ( Figure 4B ; Supplementary Figure 4C ). In line with these data, a reduction of IFN-γ and TNF-α production ( Supplementary Figure 4D ) was observed. Notably, the amount of IFN-γ released by rSP-stimulated NK cells was similar, often higher, to that produced in the culture with both TLR ligands [LPS, TLR4-ligand and Pam (3)csK(4), TLR2-ligand], suggesting similar affinity for the receptor(s) between rSP and the two agonists ( Supplementary Figure 4E ). Lastly, the silencing of TLR2 and TLR4 genes in NK cells from HD, by the transient transfection with specific siRNAs, inhibited rSP-induced phosphorylation of NF-kB and IFN-γ production by more than 50% ( Figures 5A–C ).

In order to evaluate the possible role of activated NK cells during SARS-CoV-2 infection, we examined the CD25 and CD69 expression and IFN-γ production of VOC (Wuhan and G614)-rSP-stimulated PBMCs from recovered patients (Covid+) and non-infected (Covid−) individuals who, at the sampling time, had received three doses of mRNA vaccine and were SARS-CoV-2 antigen-negative. We analyzed NK cells (gated as CD3^-CD56⁺) among VOC-rSP-stimulated PBMCs from Covid− subjects and observed higher proportions of CD25 and CD69 expression almost exclusively in the CD56^bright NK cell subset than in not stimulated cells ( Figure 6B ). By contrast, significantly higher proportions of both CD56^bright and CD56^dim NK cells expressing CD69 activation marker were found in Covid+ patients ( Figure 6A ). Accordingly, the IFN-γ release was significantly higher in culture supernatants of VOC (Wuhan and G614)-rSP-stimulated PBMCs from the Covid+ group as compared to the Covid− group ( Figure 6C ). Notably, also in this set of experiments, the highest activation of NK cells was observed upon G614-rSP stimulation compared to other rSPs (data not shown). Lastly, IFN-γ production was inhibited in vitro by anti-TLR2 plus anti-TLR4 neutralizing Abs ( Figure 6D ).

On the whole, the results on cytokine production and on the expression of activation markers show that the proportion of rSPs in vitro activable NK cells is highly increased during infection and after recovery, prevalently involving the CD56^dim subset.