Healthy donors were recruited by the Blood Transfusion Centre of Policlinico Umberto I in anonymized form and gave written informed consent. All the procedures involving participants were in accordance with the declaration of Helsinki. The study was approved by the Ethics Committee of Sapienza University of Rome (approval number 639/16 RIF/CE 4179). Only donors with a percentage of CD56^dimFcεRIγ⁻ NK cells≥5%, evaluated by flow cytometry, were included in this study.

PBMC were obtained by lymphoprep (Ficoll-Hypaque, Cedarlane, Burlington, ON, Canada) density gradient centrifugation of heparinized peripheral blood samples. Then, PBMC were washed with phosphate buffered saline (PBS) and used for in vitro assays, immunostaining and cytofluorimetric analysis.

CD20⁺ lymphoblastoid Raji B cell line, provided by Dr. F. D. Batista (The Ragon Institute of Mass General, MIT and Harvard, Cambridge, MA, USA) was cultured in RPMI 1640 medium (Euroclone, Italy), supplemented with 10% heat-inactivated Fetal Calf Serum (FCS) (Gibco, Thermo Fisher Scientific, USA) and 2 mM L-glutamine Sigma (Sigma-Aldrich, Germany) (complete medium), at 37°C, 5% CO₂ and 100% humidity. Raji cell line was kept in culture for less than 2 consecutive months and routinely tested for positivity to mycoplasma.

One hundred thousand PBMC/well were seeded in round-bottomed 96-well plates in complete medium with 1% Penicillin/Streptomycin and co-cultured with γ-irradiated (3.5 Gy) Raji cells, opsonized or not with 1 µg/1x10⁶ rituximab (kindly provided by Dr. Christian Klein, Roche Innovation Centre Zurich, Schlieren, Switzerland) (Effector: Target Ratio = 3:1), for different time lengths. In some experiments, cells were pre-treated with 100 nM rapamycin (cat. R0395-1MG, Sigma) for 2 hrs, and the inhibitor was maintained throughout the culture.

For plate-bound antibody stimulation, 24-well plates were incubated with 10 µg/ml goat anti-mouse IgG F(ab’)₂ fragment (GaM) (cat. 55487, Cappel, Ohio, USA) for 18 hrs at 4°C in PBS; afterwards, plates were washed twice in PBS and coated with 5 µg/ml αCD16 mAb (clone: B73.1) for 3 hrs at 37°C. PBMC were added (1-1.5x10⁶ cell/well) either in the presence or absence of human recombinant IL-15 (5 ng/ml, Peprotech, UK) for different time lengths.

To analyze IFN-γ production, PBMC were cultured with γ-irradiated Raji cells, opsonized or not with rituximab (3:1 Effector: Target ratio), for 5 hrs at 37°C in the presence of 50 µM Monensin (Golgi-stop; cat.#: M5273, Merck), as previously described (27). Brefeldin A (10 µg/mL, cat.#: B7651, Merck) was added after the first hour of stimulation. In some experiments, PBMC were pre-incubated with 40 mM L-amino acid transport blocker 2-aminobicyclo[2.2.1]heptane-2-carboxylic acid (BCH) (cat.#: A7902-250MG, Sigma), or with 100 nM rapamycin (cat.#: R0395-1MG, Sigma) for 2 hrs. Metabolic inhibitors were present throughout stimulation. Percent of inhibition was calculated as: (1 – the ratio between the percentage of IFN-γ-producing cells in the presence of the inhibitor over the percentage of IFN-γ-producing cells in the presence of the vehicle) x 100. For BCH inhibition, stimulation was carried out in RPMI 1640 medium supplemented with 10% dyalized FCS (cat.#: A3382001, Thermofisher) and 1% L-glutamine.

PBMC were pretreated overnight at 37°C with 10 ng/ml of human recombinant IL-15 (Peprotech, #200-15) in complete medium. Cells were then washed and incubated with APC-Vio770-conjugated anti-CD56 (Clone: REA 196, cat.#: 130-114-548) and PerCP-Vio700-conjugated anti-CD3 (Clone: BW264/56, cat.#:130-113-132) antibodies (all from Miltenyi Biotech) for 25 minutes on ice. After washing, samples were incubated with the minimum saturating dose (1µg/1x10⁶ cells) of anti-CD16 (clone B7.31) mAb for 20 minutes on ice, washed and resuspended in pre-warmed RPMI1640 medium. Goat F(ab’)₂ fragment anti-mouse IgG (H+L) (GaM) (Jackson ImmunoResearch Laboratories, #115-006-003) at 0.6 µg/1x10⁶ cells, was added for different time lengths at 37°C. Stimulation was stopped by the addition of Intracellular Fixation & Permeabilization Buffer Set commercial kit (# 00-5523-00, eBioscience, Thermo Fisher Scientific), according to manufacturer’s instructions, centrifuged for 5 minutes at 1400rpm at 4°C and stained with PE-conjugated anti-pS6 ribosomal protein (S235/236) rabbit monoclonal antibody (clone: D57.2.2E, Cell Signaling Technology, #5316S) and FITC-conjugated anti-FcεRIγ subunit polyclonal antibody (Merck, # FCABS400F). Cell samples were acquired on a FACSCanto II (BD Bioscience) and analyzed with FlowJo v.9.3.2 (TreeStar) software.

Kynurenine (K8625, Sigma, Germany) was resuspended at 800 µM in Hank’s Balanced Salt Solution (HBSS, cat.#: H8264-500ML, Sigma, Germany) and pre-warmed at 37°C. After surface staining, PBMC were washed with HBSS and incubated with 200 µM Kynurenine for 5 min at 37°C in HBSS. Uptake was stopped by adding PFA (to a final concentration of 1%) for 30 min at RT. After fixation, samples were permeabilized with a 0.05% Triton X-100-containing PBS solution for 20 min at RT, and subjected to intracellular staining. In some experiments, samples were pre-incubated with 40 mM BCH in HBSS, for 2 hrs and then stained for surface markers. Afterwards, samples were washed and incubated with Kynurenine, in the presence of 40 mM BCH for 5 minutes. Kynurenine and BCH were replaced with an equal volume of HBSS in control samples. Kynurenine signal was detected with 405 nm laser, 450/50 BP filter and fluorescence minus one (FMO, w/o Kynurenine) sample was used as negative control. Percent of inhibition of Kynurenine uptake was calculated as: (1 – the ratio between net mFI of intracellular kynurenine in the presence of BCH over net mFI of intracellular kynurenine in the presence of the vehicle) x 100. Cell sample fluorescence signals were acquired with FACSCanto II (BD Biosciences, Franklin Lakes, NJ, USA) and analyzed with FlowJo v10.9.0 (Becton-Dickinson Biosciences, San Jose, CA, USA) software (modified from ref. 63).

Freshly-isolated or cultured PBMC underwent surface staining for 25 minutes at 4°C with saturating concentrations of the following fluorophore-conjugated antibodies: anti-CD3 PerCP-Vio700 (clone: BW264/56, cat.#:130-113-132), anti-CD56 APC-Vio770 (clone: REA 196, cat.#: 130-114-548), anti-CD16 PE-Vio770 (clone: REA 423, cat.#: 130-113-394), anti-NKG2C-PE (clone: REA 205, cat.#: 130-119-776), anti-CD57 APC (clone: REA 769, cat.#: 130-111-811) or anti-CD57 PE-Vio770 (clone REA 769, cat.#: 130-111-812), anti-CD98 APC (clone REA 387, cat.#: 130-127-296), isotype control mouse IgG1 APC (clone: REA 205, cat.#: 130-113-196), all from Miltenyi Biotec (Germany); anti-CD71 Alexa Fluor 647 (clone: OKT9, cat.#: 566724) and anti-GLUT1 Alexa Fluor 647 (clone: 202915, cat.#: 566580) were from BD Pharmingen (New Jersey, USA). After surface staining, cells were fixed with 2% PFA (Merck, Germany) for 15 min at RT, permeabilized with PBS with 0.05% Triton X-100 for 20 min at RT and subjected to intracellular staining for 30 min at 4°C with saturating concentrations of the following antibodies: FcεRIγ FITC (rabbit, polyclonal, cat.#: FCABS400F, Merck Millipore) and IFN-γ APC (clone: B27, cat.#: 554702, BD Pharmingen, New Jersey, USA). Cells were washed with a PBS solution containing 2% FCS and 2 mM EDTA disodium salt (washing buffer) after each step.

Live lymphocytes were selected according to FSC-A and SSC-A physical parameters, doublets were excluded based on FSC-H and FSC-A, and memory and conventional NK cell subsets were identified as depicted in Supplementary Figure S1 . Positivity marker was set according to isotypic control sample (<0.7% positive cells).

Statistical analysis was performed with Prism v9.5.0 (GraphPad, San Diego, CA, USA) software package. Wilcoxon signed rank, Mann-Whitney, and ANOVA for repeated measures tests were used, as appropriate. All p values were subjected to a significance threshold of 0.05.

Although some metabolic peculiarities have been described in memory NK cells (12, 19), the asset of nutrient transporters has not been defined yet. We have initially analyzed the expression of CD98 neutral amino acid antiporter and CD71 transferrin receptor on freshly isolated peripheral blood (PB) NK cell populations of healthy individuals with an expanded FcεRIγ^- NK cell pool (see Materials and Methods).

CD98 was expressed on all PB NK cells, without any significant differences between CD56^dim and CD56^bright NK cells ( Figure 1A ), in accordance with a previous report (59). CD56^dim memory NK cells displayed significantly higher levels of CD98, that were specifically attributable to the more differentiated (CD57⁺) fraction; indeed, the expression of the antiporter did not appreciably differ between memory and conventional (FcεRIγ⁺) CD57^- NK cells ( Figure 1B , right panel).

CD71 transferrin receptor was poorly expressed on mostly resting PB NK cells, with CD56^bright displaying higher levels than CD56^dim ( Figure 1C ), as previously reported (59). CD71 was significantly more expressed on memory NK cells, limited to the CD57^- fraction, when evaluated as either percentage of positive cells and intensity ( Figure 1D , right panel, and Figure 1E , respectively). Of note, CD57^- conventional cells displayed higher expression of CD71, as compared to their CD57⁺ fraction ( Figure 1D , right panel, and Figure 1E ), in accord with the previously reported higher proliferative competence of less mature population (60).

In marked contrast, the major glucose transporter GLUT1, present on a higher fraction of CD56^dim than CD56^bright PB NK cells ( Figure 1F ), was more abundantly expressed on conventional NK cells, independently of their functional maturation stage ( Figure 1G ). Taken together, these data indicate that circulating FcεRIγ^- memory NK cells display a distinct pattern of nutrient transporters, characterized by higher expression of CD98 and CD71 and decreased levels of GLUT1, with respect to conventional ones. The data also suggest the impact of functional maturation state on the in vivo distinctive profile of memory population; indeed, enhanced expression of neutral amino acid antiporter positively associates with a more advanced stage of differentiation, while more elevated expression of transferrin receptor characterizes less mature cells.

CD16 engagement by contact with IgG-coated targets triggers the full effector program of human CD56^dim NK cells (22–26), in the antibody-dependent killer activation (ADKA) modality. Moreover, hyperresponsiveness to CD16-initiated signals is the most prominent functional feature of HCMV-dependent FcεRIγ^- memory cells (1–3, 5–8, 14–20). However, the capability of CD16-dependent activation to induce a persistent metabolic reprogramming is still mostly unexplored.

Twenty-four hour co-culture with rituximab (RTX)-opsonized Raji B cell lymphoma targets induced robust upregulation of CD98 and CD71 nutrient transporters on memory and conventional NK pools ( Figures 2A, E ). The response was largely limited to CD57⁺ memory cells, while it occurred irrespectively of the maturation status in the conventional NK cell pool ( Figures 2B, F ). Interestingly, CD98 levels were higher on ADKA-stimulated CD57⁺ memory NK cells, as compared with the conventional counterpart ( Figures 2A, B ), thus reproducing the difference observed in ex vivo populations ( Figure 1B ). At variance, the percentage of CD71⁺ cells was comparably increased in memory and conventional NK cell pools by 24-hr stimulation with RTX-opsonized targets ( Figure 2E ). In the absence of opsonizing mAb, CD57^- memory NK cells exhibited the highest expression of CD71, that was unaffected by CD16 engagement ( Figure 2F ); in the same conditions, both CD57^- fractions of memory and conventional NK cells showed higher expression of CD71, as compared to their more terminally differentiated counterparts, similarly to what observed ex vivo ( Figure 2F , compared with Figure 1D ).

More prolonged (3 days) stimulation with mAb-opsonized targets sustained the upregulation of both nutrient transporters in FcεRIγ^- and FcεRIγ⁺ populations, as compared with co-culture in the absence of antibody ( Figures 2C, G ). At this time point, CD98 and CD71 reached significantly higher levels in conventional NK cells, as compared to memory counterpart, in contrast with what observed in a short-term stimulation (compare Figure 2A with Figure 2C , and Figure 2E with Figure 2G ). Similarly to what observed after 24-hr stimulation, CD16-dependent upregulation of nutrient transporters was largely confined to CD57⁺ memory fraction, while robustly occurring in all conventional cells ( Figures 2D, H ). Of note, CD57^- memory and conventional NK cells had higher levels of CD71 expression in control cultures, as compared to their CD57⁺ counterparts, thus maintaining the profiles observed in either freshly isolated and short-term stimulated populations. Overall, these data firstly show that human CD56^dim NK cells rapidly (24 hrs) and persistently (up to 3 days) upregulate CD98 and CD71 nutrient transporters in response to contact with mAb-opsonized targets. Additionally, the efficiency of ADKA stimulation to promote the regulation of nutrient transporters strictly depends on functional maturation stage, as marked by CD57 expression, in FcεRIγ^- memory, but not FcεRIγ⁺ conventional population; indeed, the upregulation of transporters on memory NK cells was stronger on more differentiated fraction, while more limited (CD98) or absent (CD71) in the CD57^- subset, at both time points. Kinetic analysis indicates the memory and conventional pools differently respond to the persistence of Ab-dependent stimulation, as short-term stimulation efficiently supported the higher, ex vivo, expression of CD98 on memory cells, but this advantage was lost upon chronic stimulation. On the other hand, the higher expression of CD71 observed on freshly isolated CD57^- memory NK cells persisted in the absence of opsonizing antibody, at shorter culture time point, while it was not supported by interaction with Ab-coated targets. Collectively, the data suggest that in vivo contact with IgG-coated cells may contribute to the enhanced expression of CD98, while other signals are responsible for sustaining the higher expression of transferrin receptor in freshly isolated memory NK cells.

Further stratification of CD56^dim NK cells revealed that the NKG2C/CD57 coexpressing subset tendentially displayed the highest level of both nutrient transporters, within either memory and conventional pools, at either time point of stimulation ( Supplementary Figures S2A–D ). This evidence supports the hypothesis that it represents a fully mature subset endowed with amplified sensitivity to CD16 engagement, as our group has previously shown for other functional responses (21).

Collectively taken, these data firstly show that effective interaction with antibody-coated targets promotes the upregulation of CD98 and CD71 transporters on human PB CD56^dim NK cells. The dynamics of the nutrient transporters in response to ADKA stimulation suggests that in vivo pattern of expression, functional maturation stage and persistence of the stimulus are all factors that contribute to the different behaviour of the two nutrient receptors in memory and conventional populations.

Several soluble and cell-associated signals may cooperate with CD16-initiated transduction pathways during NK cell interaction with antibody-opsonized targets. We thus assayed the regulation of CD98 and CD71 expression upon direct CD16 crosslinking. Stimulation with plastic-immobilized αCD16 mAb rapidly (24 hrs) and persistently (3 days) promoted the upregulation of CD98 on memory and conventional NK cells ( Figures 3A, C ). CD16-triggered upregulation of CD98 on memory cells at 24 hrs was restricted to the CD57⁺ fraction, and resulted in significantly higher expression of the transporter compared to conventional counterpart ( Figure 3B ). Both distinctive features were abrogated upon more sustained crosslinking of CD16 ( Figure 3D ), similarly to what observed in the ADKA setting ( Figures 2G, H ). Notably, unstimulated CD57⁺ memory cells displayed higher levels of CD98 than their conventional counterpart, in accordance with the profile of freshly isolated cells ( Figures 3B, D ). The pattern and kinetics of CD98 modulation by αCD16 mAb-mediated crosslinking reproduced what observed upon stimulation with mAb-opsonized targets, although to a more limited extent, suggesting that other factors in the co-culture with RTX-opsonized Raji cells may amplify CD16-dependent effect. Low dose IL-15 (5 ng/ml) enhanced CD98 expression on both memory and conventional NK cells ( Figures 3A, C ), irrespectively of the functional maturation stage ( Figures 3B, D ). Notably, CD16 crosslinking in the presence of IL-15 led to a further enhancement of CD98 expression in both memory and conventional populations ( Figures 3A, C ). At shorter stimulation timepoint, this cooperation occurred on conventional NK cell subsets independently of CD57 expression, while it was limited to CD57⁺ memory NK cells, in accordance with the lower sensitivity of the CD57^- subset to CD16 crosslinking ( Figures 3B, D ). Collectively, these data indicate that CD16 crosslinking is sufficient to promote CD98 upregulation during the interaction with IgG-coated target cells, and suggest that cytokine signals in the microenvironment amplify CD16-dependent regulation of the amino acid antiporter.

Short-term (24 hrs) CD16 stimulation, in combination or not with low-dose (5 ng/ml) IL-15, did not affect CD71 expression on PB memory and conventional NK cells (De Federicis et al, unpublished observations). However, three-day stimulation with plastic-bound αCD16 mAb induced significant upregulation of transferrin receptor on both CD57⁺ and CD57^- conventional NK cells, but not on the respective memory FcεRIγ^- counterparts ( Figures 4A, B ). Low-dose IL-15 augmented the percentage of CD71⁺ memory and conventional cells, and potentiated the effect of CD16 receptor crosslinking on all CD56^dim subsets, to a comparable extent. Of note, CD71 expression on unstimulated CD57^- memory NK cells was significantly higher than on their conventional counterpart, similarly to what observed in ex vivo populations ( Figure 4B ). Collectively, these data indicate that CD16 crosslinking is sufficient to induce upregulation of transferrin receptor on conventional NK cells, while memory cells require the presence of additional signals provided during the interaction with IgG-opsonized target cells.

Cytokine-induced upregulation of CD98 and CD71 in human NK cells was shown to involve mTORC1 activation (61). Interestingly, CD16 engagement acutely promoted stronger activation of mTORC1 in FcεRIγ^- memory NK cells with respect to FcεRIγ⁺ conventional population (19). Here we show that CD16-dependent activation of mTORC1, evaluated by phosphorylated S6 levels, was maintained at higher levels in memory NK cells in an extended kinetics ( Figure 5A ). Interestingly, Rapamycin, at a dose that specifically inhibits mTORC1 activation, reduced the extent of ADKA-induced upregulation of CD98 ( Figures 5B, C ) and CD71 ( Figures 5D, E ) on memory and conventional NK cell subsets. Of note, the presence of Rapamycin consistently reduced the levels of both transporters also in cultures with not-opsonized Raji cells, with the only exception of CD57⁺ memory NK cell subset ( Figures 5C, E ), suggesting that mTORC1 activation is also involved in regulating basal nutrient transporter levels.

These data demonstrate that the activation of the mTORC1 pathway is critically involved in the capability of CD16 engagement to upregulate two important nutrient transporters on human NK cells.

To evaluate whether higher expression of CD98 on PB memory NK cells correlates with enhanced neutral amino acid intake by its associated SLC7A5 light chain, we took advantage of Kynurenine, a violet-emitting tryptophan metabolite that is imported by SLC7A5/CD98 heterodimeric antiporter (58). In line with the higher expression of the antiporter, memory NK cells displayed a larger uptake of Kynurenine ( Figure 6A ); the presence of BCH, a selective inhibitor of the antiporter, efficiently abrogated Kynurenine uptake in both memory and conventional populations ( Figure 6B ).

We evaluated the impact of CD98/SLC7A5-dependent amino acid transport on CD16-triggered IFN-γ production upon stimulation with RTX-opsonized Raji cells. The presence of BCH led to a significant decrease of the percentage of IFN-γ-producing memory and conventional NK populations ( Figure 6C ), with a comparable effect on their CD57⁺ and CD57^- respective subsets ( Figure 6D ). BCH treatment also largely affected the low production of IFN-γ in the absence of opsonizing antibodies ( Figures 6C, D ). As a consequence, the amplified response of FcεRIγ^- memory cells (7, 8, 14, 15, 18, 19, 21), as well as the enhanced response of CD57⁺ cells to CD16 stimulation (28), persist in the presence of the inhibitor, when evaluated as the fold increase in the percentage of IFN-γ-producing cells stimulated by RTX-coated targets ( Figures 6E, F ).

Interestingly, memory NK cells exhibited lower sensitivity to BCH-mediated inhibition of IFN-γ expression ( Figure 6G ), in line with the higher neutral amino acid uptake ( Figure 6A ), suggesting that larger amino acid storage capability contributes to their enhanced responsiveness to CD16 signals.

Collectively, these data firstly show that CD98 amino acid transporter function is required for CD16-dependent IFN-γ production by CD56^dim memory and conventional NK cells.

CD98/SLC7A5-dependent import of neutral amino acids, such as leucine, sustains basal activation of mTORC1 and allows its further stimulation by growth factors or cytokines (39, 41–43, 62). Our previous work demonstrated the role of PI3K/mTORC1 axis in CD16-dependent priming of NK cell response to cytokine activation (54). However, the involvement of mTORC1 activation in the CD16-dependent production of IFN-γ is still unclear. Rapamycin treatment reduced CD16-triggered IFN-γ response of memory and conventional NK cells, to a similar and limited extent ( Figures 7A, C, D ), and independently of the functional differentiation stage ( Figure 7B ), suggesting that mTORC1 pathway is only partially involved in CD16-triggered IFN-γ secretion. Moreover, the significantly higher capability to produce IFN-γ was still apparent in Rapamycin-treated memory cell samples, when evaluated as the fold increase in the percentage of IFN-γ-producing cells stimulated by RTX-coated targets (data not shown). Notably, inhibition of CD98/SLC7A5 antiporter functionality by BCH more markedly impacted CD16-triggered IFN-γ production ( Figure 6G ) than Rapamycin effect ( Figure 7C ), in both memory (p=0.0002) and conventional (p<0.0001) NK populations. These data strongly suggest that neutral amino acid intake sustains CD16-triggered IFN-γ expression through both mTOR-dependent and -independent pathways. Of note, CD16-independent (co-cultures with not-opsonized Raji cells) production of IFN-γ by conventional, but not memory, CD57⁺ and CD57^- subsets was also impaired by rapamycin treatment ( Figures 7A, B, D ).

Our results demonstrate the contribution of mTORC1 activation in the CD16-triggered metabolic reprogramming and functional response, and suggest that the peculiar nutrient transporter profile contributes to memory NK cell hyperresponsiveness to CD16 engagement.