Female CD45.2 C57BL/6J mice were purchased from Janvier labs (France). Male CD45.1 C57BL/6 mice were obtained from laboratory animal facility. Ncr1 ^iCre mice were provided by Pr. E. Vivier (42) and were crossed with TGFβRII ^fl/fl mice provided by Dr. G. Guasch (43, 44) to produce Ncr1 ^Cre/WT TGFβRII ^fl/fl mice. All mice were bred and maintained in the Cancer Research Center of Marseille animal facility under specific pathogen-free conditions. All mice were used between 6 and 12 weeks of age. In some experiments, mice were treated intraperitoneally with 100µg Poly(I:C) (In vivogen) and sacrificed 16 hours later.

All animal protocols were performed in strict accordance with the European Directive 2010/63/EU on the use of animals for scientific purposes (agreement No. 2016041809182209) and were approved by an Institutional Animal Care and Use Committee.

Two AML cell lines were used in our experiments, FLB1 cells provided by Prof. O. Hérault (45) and C1498 cells purchased from ATCC. Both leukemia cell lines are syngeneic to C57BL/6 mice. Freshly thawed 50 000 FLB1 cells (CD45.1) were resuspended in PBS and injected intravenously in congenic CD45.2 mice. For C1498, cells were expanded in DMEM medium supplemented with 10% Fetal Calf Serum (FCS) (Life Technologies) for one week. 25 000 C1498 cells (CD45.2) were then resuspended in PBS and injected intravenously in CD45.1 congenic mice. In all cases, AML development was monitored by flow cytometry analysis of the percentage of CD45.2 or CD45.1 positive cells in the blood. Control mice consisted of PBS injected mice.

In addition to transplantable models, inducible MLL-AF9 (iMLL-AF9) leukemic mice were kindly provided by Prof. J. Schwaller (Basel, Switzerland). Briefly the leukemia was induced by addition of 0.4mg/mL doxycycline in the drinking water until the appearance of leukemia as measured by white blood cell (WBC) count and a May-Grünwald-Giemsa staining (46). Mice were sacrificed when WBC count was four- to six-fold that of normal control mice. Control mice consisted of iMLL-AF9 transgenic mice not treated by doxycycline or mice having the doxycycline promoter but without the MLL-AF9 transgene (46).

FLB1 cells were injected in mice and the percentage of leukemia cells in the blood was monitored by flow cytometry as previously described. Control mice were injected with PBS. Treatment of leukemic and control mice with rapamycin or with vehicle was started when the percentage of FLB1 cells in leukemic mice reached 0.5-2% of total CD45⁺ cells in the blood, which corresponds to 17 days post-injection. Rapamycin (R706203), dissolved in DMSO to a final concentration of 20 mg/mL and further diluted in 10% PEG-400/87,5% H2O, was daily injected intraperitoneally in mice to a final dosage of 3 mg/kg. Treatment of leukemic and control mice with rapamycin or with vehicle was stopped when the percentage of FLB1 cells in leukemic mice reached 10-20% of total CD45⁺ cells in the blood. Mice were then sacrificed immediately, or kept for 24h then treated intraperitoneally with 100µg Poly(I:C) (In vivogen) and sacrificed 16 hours later ( Figure 6A ).

All patient studies were carried out in accordance with the declaration of Helsinki with the agreement of the Institut Paoli-Calmettes institutional review board and all patients gave informed consent. All patients were diagnosed with AML and peripheral blood samples were taken and frozen by the institutional biobank (agreement N° LAM-NK2020-IPC 2020-019–20-001). Frozen samples were selected irrespective of AML subtypes. Healthy donor samples were provided by the local blood bank (Etablissement Français du sang) and frozen until use.

For murine cell phenotyping, Single-cell suspensions were prepared from BM, spleen and peripheral blood and were depleted of red blood cells using 1X RBC lysis buffer (eBioscience). For extracellular staining, cells were incubated with appropriate antibodies diluted in 2.4G2 supernatant, to block non-specific antibody binding, for 30 min at room temperature. Intracellular staining for IFN-γ was performed using BD Cytofix/Cytoperm™ Fixation/Permeabilization kit (BD Biosciences). Intracellular staining of Eomes, T-bet, Ki-67, Perforin, and Granzyme B was performed using FoxP3 Fixation/Permeabilization kit (eBioscience). For pSTAT5 and pS6 measurements, BM cells were prepared and cultured with murine IL-15 (mIL-15, eBioscience) at the indicated concentrations for one hour at 37°C. Intracellular staining of phosphorylated proteins was performed using the PerFix staining kit (Beckman Coulter). For absolute counting of cells, 10µL of CountBright absolute counting beads (ThermoFisher Scientific) were added to the stained cells just before acquisition on the flow cytometer.

For human NK cell phenotyping, frozen PBMCs were thawed and counted. Cells were then stained immediately or after 48h of culture at 37°C in RPMI medium (Gibco) supplemented with 10% FCS (Life Technologies), with or without 20ng/mL recombinant human IL-15 (Miltenyi Biotec). Extracellular staining was performed by incubating cells with antibodies for 30 min at RT. Intracellular staining of Eomes and T-bet was performed using FoxP3 Fixation/Permeabilization kit (eBioscience).

For human and murine NK cell phenotyping, acquisition was performed by a BD LSRII-SORP flow cytometer or a BD Fortessa flow cytometer (BD Biosciences). Data analysis was conducted with FlowJo software (BD). A list of all antibodies used in this study can be found in Supplemental Table 1 .

Control or leukemic mice were treated with two intraperitoneal injections of 100µg EdU (Invitrogen) 12 hours apart. Thirty-six hours after the second injection, mice were sacrificed and single cell suspensions were prepared from BM and labelled with antibodies against surface antigens. To measure EdU incorporation, cells were fixed and stained with Click-it Plus EdU flow cytometry kit (Invitrogen) according to manufacturer’s instructions.

Cells labelled with antibodies specific for surface proteins were incubated for 15 min at room temperature in the dark with Annexin-V (BioLegend) and 7-AAD (BD Biosciences) diluted in Annexin-V binding buffer (BioLegend). Cells were then immediately analyzed by flow cytometry. Apoptotic cells are defined as total Annexin-V positive cells.

NK cells were sorted from the spleen of leukemic or control mice using mouse NK cell isolation kit (StemCell, 19855) and FACS Aria II (BD Biosciences). Target YAC-1 cells or FLB1 cells were stained with 4µM of Cell Proliferation Dye eFluor™ 670 (Life Technologies) according to manufacturer’s instructions. Sorted NK cells were then incubated with target cells for four hours at 37°C at different effector to target (E:T) ratios. Target cell killing was measured using CellEvent™ Caspase-3/7 Green Detection Reagent (Life technologies) and analyzed by flow cytometry.

One million splenic cells of mice pre-stimulated or not with Poly(I:C) were cultured alone, or with 200 000 YAC-1 target cells, or FLB1 or C1498 leukemia cells, or stimulated with 200ng/mL PMA and 1µg/mL ionomycin (Sigma). After four hours of incubation at 37°C in the presence of anti-CD107a antibody, golgi stop and golgi plug (BD Biosciences), cells were stained and the percentages of NK cells positive for CD107a and IFN-γ were measured by flow cytometry. In some experiments, FLB1 leukemia cells were magnetically depleted from the spleen of leukemic mice by means of anti-CD45.1-biotin antibody (eBioscience) staining and EasySep™ Mouse Streptavidin RapidSpheres™ Isolation Kit (StemCell).

RNA was isolated from total BM cells of leukemic or control mice using the RNeasy mini kit (Qiagen). cDNA was then obtained using the SuperScript II Reverse Transcriptase (Invitrogen). Quantitative PCR was performed using TGF-β1 and Gapdh (internal control) TaqMan assays (Applied Biosystems) and a CFX96 real-time PCR detection system (Bio-Rad). Gene expression was normalized to internal control (ΔCt = Ct gene of interest–Ct internal control) and relative mRNA expression to GAPDH was calculated as 2^–ΔCt.

BM cells from five mice per group (control or FLB1 injected) were pooled per experiment prior to NK cell sorting, and three independent experiments were performed. NK cells were first sorted using EasySep™ Mouse NK Cell Isolation Kit (STEMCELL Technologies), then with Aria II on the basis of live/CD3^-CD19^-NK1.1⁺CD27⁺CD11b^— gating. RNA was isolated from sorted cells using the RNeasy micro plus kit (Qiagen) and mRNA quality was evaluated using an Agilent 2100 (Pico Chip). RNA Sequencing was performed by the GenomEast platform, a member of the ‘France Génomique’ consortium (ANR-10-INBS-0009). RNA-Seq analysis was done with a production pipeline, as follows. Quality control was done with FastQC and sequence alignment against mm10 genome with Subread 1.6. Quality control revealed an optimal quality level (Phix) and a high number of aligned reads. Alignment was done in pair-read mode with removal of multiple-aligned reads. Gene count was performed with featureCount (SubRead package). Further analysis of gene counts (Differential analysis of gene expression) was done using R with EdgeR package. Count normalization was done with TMM method (EdgeR). Analysis of differential expression was performed with Limma on the web-based application Phantasus (https://bioconductor.org/packages/release/bioc/html/phantasus.html or https://artyomovlab.wustl.edu/phantasus/). Analysis of enriched pathways was first performed using EnrichR (https://amp.pharm.mssm.edu/Enrichr/) (47). Finally, Gene set enrichment analysis (GSEA) was also performed, with the GSEA software from the Broad Institute, using the Hallmark Gene Sets from the molecular signature database v7.1 (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp). The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: GEO, GSE180409.

Freshly isolated BM or splenic cells were resuspended in glucose free RPMI (Gibco) supplemented with 100µM 2-NBDG (2-(N-(7-Nitrobenz-2-oxa-1,3-diazol-4-yl)Amino)-2-Deoxyglucose, Invitrogen) for 10 min et 37°C. Cells were then washed with PBS, stained with antibodies and analyzed by flow cytometry.

Total mitochondrial mass or mitochondrial superoxide were measured by incubating cells for 10 min at 37°C with 500nM MitoTracker™ Deep Red (Life Technologies) or 5µM MitoSOX™ Red (Life Technologies) respectively. Cells were washed twice with PBS, stained with antibodies and analyzed by flow cytometry.

Femurs and tibias, isolated from leukemic or control mice, were dissociated from adjacent muscles, cut in two, and centrifuged in RPMI medium. BM supernatants of 5 different leukemic or control mice were pooled and analyzed for 97 different mouse cytokines using mouse cytokine array C6 according to manufacturer’s instructions (Antibodies Online). The cytokine membranes were quantified by densitometric analysis using ImageJ.

For IL-15 measurement, BM supernatants of leukemic or control mice were analyzed using mouse IL-15 DuoSet ELISA (R&D systems) according to manufacturer’s recommendations.

Statistical analyses were performed using Prism software (GraphPad). For two data sets comparison, unpaired nonparametric Mann-Whitney t-test was used. For comparison of more than two data sets, unpaired nonparametric Kruskal-Wallis t-test was used. Levels of statistical significance are expressed as P values: *P< 0.05, ** P <0.01, *** P <0.001, **** P <0.0001. Error bars represent SEM.

To gain insight into the mechanisms underlying NK cell defects observed in AML patients, we used FLB1 cells as a syngeneic mouse model of AML. This model is based on the overexpression in hematopoietic stem cells of hoxa9 and meis1 genes that are commonly overexpressed in AML (48, 49). FLB1 cells are of myeloid origin and phenotype (data not shown) and almost recapitulate human AML (45). They are recognized and killed by syngeneic NK cells in vitro, although to a lesser extent than the classical murine NK target cell YAC-1 ( Supplemental Figure 1A ). Intravenous injection of FLB1 cells in syngeneic mice resulted in a progressive bone marrow (BM), blood and spleen invasion ( Supplemental Figure 1B ). Following AML development, NK cell frequency was increased in leukemic mice BM ( Supplemental Figure 1C ), albeit a significant decrease was observed in non-leukemic and NK cell numbers ( Supplemental Figures 1D, E ). There was no significant change in the percentage nor in the number of NK cells in the spleen of leukemic mice compared to control mice ( Supplemental Figures 1C–E ).

We next analyzed NK cell maturation by means of CD27 and CD11b expression. We observed an increase in the percentage of the immature CD27⁺CD11b^— NK cell subset and a sharp decrease in the percentage of mature CD27⁺CD11b⁺ NK cells in the BM and spleen of leukemic mice ( Figure 1A ), while no significant change could be observed for CD27^—CD11b⁺ NK cell frequency ( Figure 1A ). With respect to numbers, all subsets were reduced in leukemic mice BM and not changed in the spleen, except CD27⁺CD11b^— immature NK cells which were increased in the spleen of leukemic mice ( Supplemental Figure 1F ). Moreover, we noted that NK cell maturation imbalance in the BM paralleled leukemia progression in this organ ( Figure 1B ). Finally, we evaluated NK cell maturation in other mouse models of AML: the murine C1498 transplantable model and the genetically induced model of MLL-AF9 transgenic mice. Both tumors invade progressively the BM and induce AML-like diseases (46, 50). Similarly to the FLB1 model, we found a significant increase in the percentage of the immature CD27⁺CD11b^— NK cells in both leukemia models ( Supplemental Figure 1G ). These data suggest that the hypomaturation of NK cells is common in leukemic mice, and validate the use of our FLB1 injected mice as a surrogate model for the study of NK cell alterations in AML.

NK cell maturation is regulated by transcription factors, including T-bet and Eomesodermin (Eomes) (51). In particular, it has been shown that Eomes regulates NK cell maturation from CD27⁺CD11b^— stage to CD27⁺CD11b⁺ stage (52, 53). To assess whether Eomes and T-bet are dysregulated in NK cells from leukemic mice, we evaluated their expression in BM and spleen NK cells by flow cytometry. We observed a decrease in the percentage of Eomes-expressing NK cells in leukemic mice ( Figure 1C and Supplemental Figure 1I ). We also noted a decrease in the level of Eomes expression in Eomes⁺ NK cells, although this decrease is not statistically significant in the BM ( Supplemental Figure 1H, J ). Moreover, reduced expression of T-bet in NK cells from leukemic mice was found ( Figure 1D and Supplemental Figure 1K ), suggesting an altered expression of the transcription factors regulating NK cell maturation after leukemia progression in mice.

The hypomaturation of NK cells in leukemic mice could also originate from an increase in the proliferation rate of immature subsets and/or in the apoptosis of mature subsets. To address this, we first assessed in vivo homeostatic proliferation of NK cell subsets by means of EdU incorporation. We observed a decrease in the percentage of proliferating NK cells in the BM of leukemic mice compared to control mice ( Figure 1E ). This decrease was observed for all subsets, especially for CD27⁺CD11b^— NK cells ( Figure 1E ) and was confirmed by a measurement of Ki67 expression ( Figure 1F ). Moreover, ex vivo apoptosis analysis by measuring Annexin-V expression revealed a decreased apoptosis especially in mature CD11b⁺ subsets ( Figure 1G ). Altogether, these data suggest an altered homeostatic proliferation, apoptotic rate, and maturation in NK cells from leukemic mice.

Next, we compared the functions of splenic NK cells from leukemic and control mice. NK cells from leukemic mice tend to produce less IFN-γ and degranulate less than control NK cells in response to YAC-1 target cells ( Figure 2B ). Priming of NK cells in vivo with the TLR3 ligand poly(I:C), injected 16h prior to the functional analysis, improved their degranulation and cytokine production but these functions remained lower in NK cells from leukemic mice compared to controls ( Figures 2A, B ). Ex vivo stimulation with PMA and ionomycin, performed in order to measure the maximal functional potential of NK cells, also resulted in decreased degranulation and IFN-γ production in NK cells from leukemic mice ( Figure 2B ). A similar reduction of NK cell functions was observed when FLB1 cells were depleted from the spleen of leukemic mice before ex vivo stimulation, ruling out that the presence of leukemic cells during functional assays may be responsible for the observed effects ( Supplemental Figure 2A ). In line with the reduction of NK cell functions, NK cells from leukemic mice expressed lower levels of perforin and granzyme B ( Figure 2C ). Finally, we assessed the functions of NK cells from leukemic mice, primed with poly(I:C), in response to AML cell lines, FLB1 and C1498. Decreased degranulation and IFN-γ production by NK cells were observed as compared to controls ( Figure 2D ). Noteworthy, defective functions of NK cells in response to in vivo/ex vivo stimulation were independent of their maturation status ( Supplemental Figures 2B, C ). Altogether, these results show that NK cells in leukemic mice display a reduction of their effector functions.

NK cell maturation, homeostasis, as well as activation, are critically dependent on cytokine priming. We hypothesized that NK cell defects in leukemic mice might stem from changes in cytokine signaling. To address this point, we first measured the modifications in the cytokine profile of the BM supernatants of leukemic and control mice using a mouse cytokine array for the quantification of 97 different cytokines. Several cytokines and soluble factors were differentially expressed ( Supplemental Figure 3A ). We noted an increase in the levels of some pro-inflammatory cytokines, such as Leptin, IFN-γ, CD27, CD30, hepatocyte growth factor (HGF), but also of some anti-inflammatory cytokines such as IL-10 and CTLA-4 or some chemokines such as CXCL1 and CCL21 ( Supplemental Figure 3A ). Next, we performed RNA sequencing analysis on the immature CD27⁺CD11b^— NK cells from the BM of leukemic and control mice. We used a web-based software (EnrichR) which provides the 10 most significantly enriched pathways ( Figure 3A ). Enrichment analysis revealed a strong response to different proinflammatory cytokines including type-I IFNs and Interleukin/IL-2 signaling in NK cells from leukemic mice. A further analysis of the RNA sequencing data using GSEA confirmed that the IL-2/IL-15 pathway was engaged in NK cells isolated from leukemic mice ( Figures 3B, C ). Amongst the genes from this pathway, Cish, known to be induced by IL-2 and IL-15, was significantly up-regulated (FC=1.9, adjusted p value = 0.02) ( Figure 3C ) (24).

IL-15 is one of the most important cytokines for NK cells. In particular, IL-15/mTOR pathway regulates NK cell maturation from CD11b^— to CD11b⁺ stage and facilitates NK cell activation in periphery (15, 54), two checkpoints that are altered in our leukemic mice. We first measured by ELISA the levels of IL-15 in the BM supernatants. Interestingly, concentration of IL-15, measured by ELISA, in the BM of FLB1-injected mice was higher compared to controls an increase in the levels of IL-15 in the BM of FLB1-injected mice compared to controls ( Figure 3D ). We also noted a tendency towards an increase in the levels of IL-15 in the BM of MLL-AF9 leukemic mice ( Supplemental Figure 3B ). Next, we measured ex vivo levels of phosphorylation of STAT5 (pSTAT5) and of the mTOR associated ribosomal protein S6 (pS6) in NK cells. We noted an increase in the amount of phosphorylated STAT5 and S6 in NK cells from leukemic mice compared to controls ( Figure 3E ), suggesting an activation of the signaling pathways downstream of IL-15 receptor. In line with an activated profile, flow cytometry analysis showed an increase in the percentage of NK cells expressing CD69 and CD25, two markers that can be induced by IL-15 (25, 55) ( Figure 3F ). Finally, we measured the expression of IL-15 receptor subunits on BM-NK cells. The β chain of the IL-2/IL-15 receptor (CD122) was downregulated in FLB1-injected and MLL-AF9 leukemic mice ( Figure 3G and Supplemental Figure 3C ). We also noted a decrease in the surface expression of the common γ chain (CD132) in NK cells from FLB1-injected mice, independently of the stage of NK cell maturation ( Figure 3G ). A total (intracellular and extracellular) versus extracellular staining of CD122 in NK cells showed that the whole surface and intracellular levels for this receptor were decreased in leukemic mice NK cells ( Supplemental Figure 3D ). The increase in the levels of IL-15 in the BM and the activation of the IL-15 signaling pathway in NK cells from leukemic mice strongly suggest that the downregulation of IL-15 receptor subunits is due to ligation to IL-15 and increase in its degradation (24, 56), in line with cish mRNA upregulation. Altogether, these data strongly suggest that NK cells are chronically stimulated by cytokines in the leukemic microenvironment and in particular by the pro-inflammatory cytokine IL-15.

Our data indicate that NK cells from leukemic mice are exposed to IL-15 in vivo. We wanted to assess the consequences of this exposure to their responsiveness to this cytokine ex vivo. For this, we cultured BM cells, from leukemic or control mice, with increasing concentrations of IL-15 for 1 hour, and we measured the expression of phosphorylated STAT5 and S6 in NK cells by flow cytometry. IL-15-induced phosphorylation of STAT5 was reduced in total and CD27⁺CD11b^— NK cells from leukemic mice compared to control mice, at all IL-15 concentrations tested ( Figure 4A ). The S6 phosphorylation was induced in NK cells only at high concentrations of IL-15 known to activate the mTOR pathway (15). Nonetheless, S6 phosphorylation was lower in NK cells from leukemic mice compared to control mice, and this reduction was independent of the maturation stage of NK cells ( Figure 4B ). These data indicate that NK cells from leukemic mice are hyporesponsive to additional short stimulation with IL-15 in vitro. Of note, culturing NK cells for a longer period of time (24h) with IL-15 increased their responsiveness to target cells but NK cells from leukemic mice were still poor responders compared to controls (data not shown), thus confirming that the responsiveness of NK cells to IL-15 stimulation in vitro was reduced in leukemic mice.

Chronic stimulation of NK cells with IL-15 is known to generate deleterious effects, assimilated to cell exhaustion, on NK cell functions and metabolism (55, 57–59). So far, our results suggested that BM-NK cells in leukemic mice bath in a proinflammatory milieu and in particular they may be chronically stimulated by IL-15 which leads to deregulation in the mTOR pathway. Therefore, we next wondered what would be the consequences of this exposure on NK cell metabolism. First, we measured by flow cytometry the size and granularity of NK cells ex vivo as simple surrogate markers of NK cell activation and metabolic response (15, 60). In contrast to the activation status, the size and granularity of BM-NK cells, but not that of spleen NK cells, tended to be lower in leukemic mice compared to control mice ( Supplemental Figure 4A ). We next measured the expression of the heavy chain of L-amino acid transporter SLC3A2 (CD98) and the transferrin receptor (CD71) ex vivo. Indeed, these markers are known to be upregulated in metabolically active NK cells (15, 60). Interestingly, the expression of CD98 was reduced, irrespective of maturation stages, in leukemic mice BM-NK cells compared to control mice ( Supplemental Figure 4A ). In addition, we observed a tendency towards a decrease in the expression of CD71 ( Supplemental Figure 4A ). These data indicate that despite their activation, NK cells of leukemic mice seem to be less metabolically active.

In vivo injection of poly(I:C) is known to strongly enhance NK cell metabolism via IL-15 and the mTOR pathway (15, 60). Therefore, we next compared the metabolic responses of NK cells in leukemic and control mice after poly(I:C) injection 16h prior to sacrifice ( Figure 5A ). We observed a reduced NK cell size, granularity, and expression of nutrient transporters in BM-NK cells from leukemic mice ( Figure 5B and supplemental Figure 4B ). We also observed a tendency towards a decrease of these markers in spleen-NK cells from leukemic mice compared to controls ( Figure 5C and supplemental Figure 4B ). Moreover, we assessed the uptake of glucose by ex vivo NK cells primed with Poly(I:C) by measuring 2NBDG staining. We showed a reduced glucose uptake in BM-NK cells from leukemic mice compared to controls ( Figure 5D and Supplemental Figure 4C ). Finally, mitochondrial activity in NK cells, including mitochondrial mass and mitochondrial reactive oxygen species (ROS), measured by MitoTracker and MitoSOX staining, respectively, were also reduced in leukemic mice ( Figures 5E, F and Supplemental Figure 4D ). Altogether, these data suggest that NK cells from leukemic mice exhibit reduced metabolic response after activation of the mTOR pathway.

Our data suggest that IL-15/mTOR pathway is chronically stimulated in NK cells from leukemic mice which is associated with metabolic and functional defects. We hypothesized that the reduction of the activation of the mTOR pathway during AML progression might reduce these defects. Therefore, we treated mice with daily injections of rapamycin, an inhibitor of the mTOR complex 1 (mTORC1), during the exponential phase of the engraftment of AML cells in the organs ( Figure 6A and Supplemental Figure 1B ). First, we measured the levels of S6 phosphorylation in NK cells ex vivo and observed a decrease in leukemic mice treated with rapamycin compared to those treated with vehicle ( Figure 6B ). This increased S6 phosphorylation remained higher in NK cells of leukemic mice compared to control mice even in rapamycin treated group, indicating that the regimen of rapamycin treatment used in our experiments reduced, but did not completely abrogate, the activation of the mTOR pathway ( Figure 6B ). Next, we assessed NK cell metabolic markers such as the size, granularity and expression of nutrient transporters CD98 and CD71. We showed an increase in these markers in leukemic mice treated with rapamycin compared to mice treated with vehicle ( Figures 6C, D ). Finally, NK cell secretion of IFN-γ, but not their degranulation, tended to be higher in leukemic mice treated with rapamycin and activated with Poly(I:C) injection 16h prior to sacrifice compared to mice treated with vehicle ( Figure 6E ). Altogether, these data indicate that the reduction of the activation of mTOR during AML development partially rescued NK cell metabolic and functional defects.

TGF-β is a prominent immunosuppressive cytokine that can inhibit the metabolism and the functions of NK cells, in particular by repressing the mTOR pathway (61–63). TGF-β can also decrease the expression of the β subunit of IL-15 receptor leading to reduced IL-15 signaling (64). In our model we noted an increase in phosphorylated SMAD2/3 (pSMAD2/3) proteins ( Supplemental Figure 5A ) hinting at activation of TGF-β signaling as evidenced by an increase in phosphorylated SMAD2/3 (pSMAD2/3) proteins ( Supplemental Figure 5A ). Not surprisingly TGF-β mRNA levels also had a tendency to increase in BM cells of leukemic mice compared to control mice ( Supplemental Figure 5B ). Therefore, we wondered whether TGF-β could be involved in the alteration of NK cell IL-15 signaling, metabolism, and functions. To address this question, we used mice in which the TGF-βRII receptor was depleted in NK cells specifically (NK-TGFβRII-/- mice). When NK-TGFβRII^-/- mice were injected with FLB1 cells, we did not observe any restoration of NK cells defects, although phosphorylation of SMAD2/3 was completely abrogated ( Supplemental Figure 5C–G ). Hence, CD122 expression, STAT5 and S6 phosphorylation, degranulation, IFN-γ production, and expression of nutrient transporters remained reduced in NK cells from leukemic mice compared to controls ( Supplemental Figure 5D–G ), ruling out the implication of TGF-β in the alterations of NK cells in leukemic mice.

Thus far as our data suggest that NK cells in leukemic mice are chronically stimulated by IL-15 and exhibit defective homeostatic proliferation, signaling defects and reduced functional and metabolic responses, suggesting an exhaustion status. Indeed, the reduced expression of Eomes by BM-NK cells from leukemic mice ( Figure 1C ) is in line with this hypothesis (59, 65). To confirm this hypothesis, we measured the expression of another marker of exhaustion of NK cells, the co-inhibitory molecule TIGIT (66), and we observed a higher frequency of TIGIT positive NK cells in leukemic mice ( Figure 6F ). Altogether, these data indicate that NK cells of leukemic mice are exhausted and suggest an implication of the chronic pro-inflammatory stimulation, partially mediated by IL-15 in the BM.

We next sought to determine whether NK cells from AML patients also displayed IL-15 signaling defects and metabolic reduction. We analyzed peripheral blood NK cells from patients diagnosed with AML and compared with healthy donors’ NK cell expression of IL-15Rβ/CD122. We noted a lower surface expression of CD122 on AML patients’ NK cells ( Figure 7A ). in humans, the final steps of NK cell maturation are characterized by the expression of the marker CD56/NCAM defining 2 different maturation subsets: CD56bright representing immature NK cells, which give rise to more mature CD56dim NK cells. Although not all patients had sufficient numbers of CD56^bright NK cells, as previously shown (67), we found that both NK cell maturation subsets displayed a reduced CD122 expression in AML patients ( Figure 7A ). Next, we analyzed the expression of CD98 and CD71 as surrogate markers of cellular metabolism (68). In contrast to leukemic mice NK cells, AML patients’ NK cells had similar levels of expression of CD98 and higher levels of expression of CD71 as compared to that of healthy donors ( Supplemental Figure 6 ). However, NK cells from AML patients were less able to increase the expression of CD98 and CD71 in response to a stimulation with IL-15 in vitro for 48h ( Figure 7B ). Finally, we analyzed the expression of exhaustion markers on NK cells such as PD-1 and TIGIT. While TIGIT expression was slightly lower on AML patients’ NK cells, a fraction of patients’ NK cells expressed high levels of PD-1 ( Figure 7C ).