Research ArticleImmunometabolism

TGF-β inhibits the activation and functions of NK cells by repressing the mTOR pathway

See allHide authors and affiliations

Science Signaling  16 Feb 2016:
Vol. 9, Issue 415, pp. ra19
DOI: 10.1126/scisignal.aad1884

Relieving NK cell suppression

The immunosuppressive cytokine transforming growth factor–β (TGF-β) has beneficial effects when it resolves inflammation and prevents autoimmunity, but not when it inhibits antitumor immune responses. Viel et al. found that TGF-β signaling in mouse and human natural killer (NK) cells, cytotoxic cells that target tumor cells, inhibited the activation of the kinase mTOR, a central regulator of cellular metabolism and cytotoxic function. NK cells deficient in a TGF-β receptor subunit showed enhanced antitumor activity in mouse models of metastasis, suggesting that enhancing metabolism in NK cells may provide a therapeutic strategy to kill cancer cells.

Abstract

Transforming growth factor–β (TGF-β) is a major immunosuppressive cytokine that maintains immune homeostasis and prevents autoimmunity through its antiproliferative and anti-inflammatory properties in various immune cell types. We provide genetic, pharmacologic, and biochemical evidence that a critical target of TGF-β signaling in mouse and human natural killer (NK) cells is the serine and threonine kinase mTOR (mammalian target of rapamycin). Treatment of mouse or human NK cells with TGF-β in vitro blocked interleukin-15 (IL-15)–induced activation of mTOR. TGF-β and the mTOR inhibitor rapamycin both reduced the metabolic activity and proliferation of NK cells and reduced the abundances of various NK cell receptors and the cytotoxic activity of NK cells. In vivo, constitutive TGF-β signaling or depletion of mTOR arrested NK cell development, whereas deletion of the TGF-β receptor subunit TGF-βRII enhanced mTOR activity and the cytotoxic activity of the NK cells in response to IL-15. Suppression of TGF-β signaling in NK cells did not affect either NK cell development or homeostasis; however, it enhanced the ability of NK cells to limit metastases in two different tumor models in mice. Together, these results suggest that the kinase mTOR is a crucial signaling integrator of pro- and anti-inflammatory cytokines in NK cells. Moreover, we propose that boosting the metabolic activity of antitumor lymphocytes could be an effective strategy to promote immune-mediated tumor suppression.

INTRODUCTION

Natural killer (NK) cells are innate lymphoid cells that have an important role in the defense against intracellular pathogens and tumors. NK cells have the ability to kill other cells that are recognized as targets through an arsenal of receptors that recognize major histocompatibility complex (MHC) class I molecules or various surface ligands associated with cellular stress (1). NK cells also secrete large amounts of interferon-γ (IFN-γ) and other cytokines in response to stimulation through NK cell or cytokine receptors. NK cells develop in the bone marrow and in other organs such as the liver (2, 3) in response to the cytokine interleukin-15 (IL-15). IL-15 is pivotal to maintain NK cell survival in the periphery by stimulating the phosphorylation (and activation) of the transcription factor signal transducer and activator of transcription 5 (STAT5) (4). After they have become committed to the NK cell lineage, NK cells undergo a phase of intense proliferation before reaching the blood circulation. This is followed by a process of maturation that includes at least three stages defined by the cell surface expression of CD11b and CD27 (5, 6). CD11bCD27+ cells (also called CD11b cells) are the most immature cells and are found mostly in the bone marrow and lymph nodes. CD11b+CD27+ (or double-positive cells) represent the intermediate stage, whereas CD11b+CD27 (or CD27 cells) are the most mature cells, having the full set of NK cell receptors and the greatest amount of S1P5, a sphingosine 1-phosphate receptor that enables the egress of NK cells from the bone marrow and promotes their circulation in the blood (7).

The cytotoxic potential of NK cells is enhanced upon inflammation mediated by Toll-like receptor (TLR) ligands. Mechanistically, this phenomenon involves trans-presentation of IL-15 and IL-15 receptor α (IL-15Rα) complexes by myeloid cells, such as dendritic cells, and the cell surface amounts of these complexes are increased during infections (8). We previously showed that increased concentrations of IL-15 stimulate activation of the serine and threonine kinase mammalian target of rapamycin (mTOR) in NK cells, enhancing both their metabolism and their cytotoxic arsenal. Stimulating mTOR activity is essential for the activation of peripheral NK cells, which demonstrates that it is a key checkpoint to control the effector potential of these innate effectors (9). Similarly, the control of mTOR and cellular metabolism is central to the regulation of effector functions of other immune subtypes, such as T cells and dendritic cells (10). Hence, pharmacological agents such as rapamycin or its derivatives are powerful immunosuppressant molecules and are used in the clinic to inhibit graft rejection in solid organ transplantation (11).

Transforming growth factor–β (TGF-β) is a cytokine of the bone morphogenetic protein (BMP)–activin family that mediates a wide range of actions in the immune system (12). Many cell types produce TGF-β, and virtually all cells of the immune system express the TGF-β receptor (TGF-βR). TGF-β is secreted as an inactive dimer that requires processing through different mechanisms to become active. Active TGF-β binds to a tetrameric receptor that is composed of two TGF-βRI chains and two TGF-βRII chains. On binding to TGF-β, the type II receptors phosphorylate the type I receptors, which then propagate the signal by phosphorylating the transcription factors Smad2 and Smad3 (collectively known as Smad2/3). This complex then shuttles to the nucleus and binds to Smad4 and additional cofactors to repress or activate the expression of target genes. In addition to this Smad-dependent pathway, which is often referred to as the canonical signaling pathway, TGF-βRs also activate various other signaling events that involve p38 mitogen-activated protein kinase (MAPK), phosphoinositide 3-kinase (PI3K)–Akt, tumor necrosis factor receptor (TNFR)–activated factor 6 (TRAF6), among others. The importance of this Smad-independent pathway is variable depending on cell types, as well as other parameters, such as proliferation and cellular environment (13).

TGF-β is commonly viewed as the most powerful immunosuppressive cytokine (14). Deletion of TGF-β or its receptor induces a massive T cell–mediated autoimmune syndrome that causes rapid death in mice. This occurs because of the essential role of TGF-β in the control of T cell tolerance in part by promoting the development of the natural regulatory T (Treg) cell lineage and inducing the differentiation of peripherally induced Tregs, and also by directly restraining the activation and differentiation of CD4+ and CD8+ T cells (14). Furthermore, TGF-β plays an important role in the suppression of innate immune cells, such as NK cells (15, 16). The addition of TGF-β to NK cell cultures containing IL-2 inhibits NK cell proliferation, cytotoxicity, and IFN-γ secretion (17, 18). In vivo, the transgenic expression of a dominant-negative (DN) form of TGF-βR in CD11c+ cells (which is expressed at least by dendritic cells and NK cells) in CD11c-dnTGFβRII mice stimulates substantial NK cell proliferation and promotes their maturation but does not induce adverse autoimmune reactions (19). The production of TGF-β by tumors or in the context of chronic infections inhibits NK cell function, in particular through the decreased abundance of NK cell receptors (20, 21). Reciprocally, deletion of TGF-βRs in T cells leads to their activation and NK cell–like differentiation, which suggests that TGF-β is a powerful repressor of NK cell receptor expression (22). The molecular mechanism by which TGF-β inhibits NK cell differentiation is unclear. SMADs repress expression of the gene encoding the transcription factor T-bet (Tbx21) by directly binding to the Tbx21 promoter, thereby limiting IFN-γ production by NK cells (23). Whether this mechanism is sufficient to explain the range of inhibitory activities of TGF-β on NK cell proliferation and cytotoxicity and the expression of NK cell receptors is unknown.

Here, we generated several mouse models that enabled deletion of the TGF-βR or constitutive activation of TGF-β signaling in NK cells to revisit the mechanisms of action of TGF-β in NK cells. We found that TGF-β caused the very early inhibition of mTOR activity in NK cells stimulated with IL-15. The effect of TGF-β on mTOR was comparable to that of rapamycin, a specific mTOR complex 1 (mTORC1) inhibitor, in both intensity and kinetics. Moreover, TGF-β and rapamycin had very similar effects on the activation of NK cells in vitro, and mTOR deletion or constitutive TGF-β signaling in NK cells had comparable deleterious effects on NK cell development and differentiation in vivo. Together, our data establish that a major target of early TGF-β signaling in NK cells is the kinase mTOR.

RESULTS

TGF-β is not essential for conventional NK cell development

To study the role of TGF-β signaling in mouse NK cells, we measured the abundance of TGF-βRII in NK cell subsets defined by the cell surface expression of CD11b and CD27. TGF-βRII was found in all subsets but was maximal in immature CD11bCD27+ NK cells (Fig. 1A). Accordingly, upon ex vivo treatment with TGF-β, CD11b NK cells had greater amounts of phosphorylated SMAD2 (pSMAD2) and pSMAD3 (pSMAD2/3) than did mature NK cells (Fig. 1B), whereas the total amounts of SMAD2 and SMAD3 were similar between subsets (Fig. 1B). Because immature NK cells actively proliferate to generate the pool of peripheral NK cells, these data suggested a possible role for TGF-β in the regulation of the size of this pool. Previous studies that showed that the expression of a DN form of TGF-βRII in CD11c+ cells (CD11c-dnTGFβRII mice) results in increased numbers of peripheral NK cells further support this hypothesis (19, 24). To directly test this, we deleted Tgfbr2 specifically in NK cells by crossing Ncr1Cre mice (25) with Tgfbr2fl/fl mice (26) to obtain Ncr1Cre/+xTgfbr2fl/fl mice (hereafter called NK-Tgfbr2−/− mice). Tgfbr2−/− NK cells were as unresponsive to TGF-β as were NK cells from transgenic CD11c-dnTGFβRII mice (Fig. 1C). However, and in contradiction with results obtained from the CD11c-dnTGFβRII model, the distribution and the maturation of NK cells were normal in NK-Tgfbr2−/− mice (Fig. 1, D to F). The only evidence of an inhibitory role of TGF-β on NK cell development was the increased proliferation measured for bone marrow NK cells in NK-Tgfbr2−/− mice compared to that in bone marrow NK cells from control mice (Fig. 1G). Thus, TGF-β had very limited activity on conventional NK cell development and homeostasis under steady-state conditions. The difference in NK cell phenotypes between CD11c-dnTGFβRII mice and NK-Tgfbr2−/− mice could be due to the different approaches used to abrogate TGF-β signaling.

Fig. 1 TGF-βR is functional on NK cells but does not regulate their homeostasis under steady-state conditions.

(A) Flow cytometric analysis of the cell surface expression of TGF-βRII on the indicated gated NK cell subsets from mouse spleen. Left: Histogram plot is representative of three experiments. Right: Averaged mean fluorescence intensity (MFI) of the TGF-βRII staining calculated from a total of three mice in two experiments. DP, double-positive. (B and C) Splenocytes from the indicated mouse strains were cultured for 1 hour with TGF-β before the extent of SMAD2/3 phosphorylation was measured by flow cytometry. (B) Left: Histogram plot is representative of three experiments. Middle: Averaged MFI of the pSMAD2/3 staining was calculated from six wild-type (WT) mice from two experiments. Right: Averaged MFI of total SMAD2/3 staining was calculated from seven WT mice from two experiments. (C) Averaged MFI of the pSMAD2/3 staining calculated from three mice of each strain. ns, not significant. (D to F) Flow cytometric analysis of NK cells in the indicated organs of WT and NK-Tgfbr2−/− mice. The percentages (D) and numbers (E) of NK cells were calculated from eight mice from each group. (F) Density plot of the cell surface expression of CD27 and CD11b in NK cells. Data are representative of three experiments for each group. (G) Flow cytometric analyses of bromodeoxyuridine (BrdU) incorporation by bone marrow (BM) NK cells from WT and NK-Tgfbr2−/− mice. Data are the average percentages of BrdU-positive cells in six mice from three independent experiments. *P < 0.05, **P < 0.01, and ***P < 0.001 by Student’s t test.

TGF-β inhibits the antitumor function of NK cells

To identify in vivo situations of NK cell exposure to TGF-β, we measured the extent of SMAD2/3 phosphorylation in isolated cells from mice that were left unchallenged or were challenged with different agents ex vivo. Very little pSMAD2/3 was measured in splenic NK cells from unchallenged mice or from mice injected with the classical NK cell activator polyinosinic:polycytidylic acid [poly(I:C)] or with IL-15–IL-15Rα complexes (Fig. 2A). Similar results were obtained when mice were infected with influenza virus or Listeria monocytogenes; however, substantial SMAD2/3 phosphorylation was measured in NK cells extracted from solid tumors, such as NEU15 mammary tumors (Fig. 2A), which is suggestive of the exposure of NK cells to TGF-β within the tumor environment. To examine the role of TGF-β in NK cells in the context of a tumor, we challenged NK-Tgfbr2−/− mice and littermate controls with intravenous injection of B16-F10 melanoma cells or RM-1 cells and counted lung metastases 2 weeks after injection. The NK-Tgfbr2−/− mice exhibited better suppression of metastases than did the control mice (Fig. 2, B and C). Depletion of NK cells in both mouse strains with anti–asialo-GM1 antibody led to an equivalently large increase in the number of metastases (Fig. 2, B and C). Thus, we conclude that TGF-β impairs the antitumor function of NK cells.

Fig. 2 TGF-β inhibits the antitumor function of NK cells.

(A) Flow cytometric analysis of Smad2/3 phosphorylation in gated NK cells. NK cells from the spleens of WT C57BL/6 mice that were left unchallenged, were injected with poly(I:C) 18 hours earlier, were injected twice with IL-15–IL-15Rα complexes, or were infected with influenza or Listeria 7 days earlier were analyzed. NK cells from the tumors of NEU15 cell–injected mice were also analyzed. Results are averaged from at least four mice for each condition. MFIs were normalized to the amount of pSMAD2/3 in control NK cells. (B and C) Control and NK-Tgfbr2−/− mice were injected intravenously with B16-F10 cells or RM-1 cells (2 × 105). Groups of mice were treated with 50 μg of asialo-GM1 antibody or control immunoglobulin (Ig) on the day before, the day of, and 7 days after tumor inoculation. Mouse lungs were harvested and fixed on day 14, and the numbers of metastases were counted under a dissecting microscope. Each symbol represents an individual mouse. *P < 0.05 and **P < 0.01 by Mann-Whitney test.

Loss of TGF-β signaling derepresses NK cell activation in vitro

To investigate how TGF-β inhibited the activation of NK cells, we compared the effect of stimulation of NK-Tgfbr2−/− mice and wild-type control mice with IL-15 on parameters of NK cell activation. Tgfbr2−/− NK cells had more granzyme B, T-bet, and KLRG1 than did control NK cells upon stimulation with IL-15 (Fig. 3A). Furthermore, IL-15–stimulated Tgfbr2−/− NK cells also exhibited increased size and granularity as measured by forward scatter (FSC) and side scatter (SSC), respectively, and increased CD98 and CD71 abundances than did similarly treated control NK cells (Fig. 3A). The latter parameters correlate with metabolic activity, as we previously showed (9). For this reason, we hypothesized that Tgfbr2−/− NK cells might have increased mTOR activity upon stimulation with IL-15. To test this hypothesis, we measured the extent of phosphorylation of the ribosomal S6 protein, which is a substrate of S6 kinase (S6K), a major target of mTORC1 in response to IL-15 signaling (9). We detected substantially greater S6 phosphorylation in Tgfbr2−/− NK cells than in control NK cells but saw no difference in the extent of STAT5 phosphorylation upon IL-15 stimulation (Fig. 3B), suggesting that endogenous TGF-β may inhibit NK cell activation by limiting mTOR activity.

Fig. 3 Loss of TGF-β signaling derepresses NK cell activation induced by IL-15.

(A and B) Splenocytes from WT and NK-Tgfbr2−/− mice were cultured with IL-15 (100 ng/ml) for 3 days (A) or 1 hour (B). The relative abundances of the indicated intracellular and cell surface proteins were then measured by flow cytometric analysis (A). Bar graphs show averaged MFI ratios calculated from a total of five mice from three experiments and normalized to those of control NK cells. Histograms show FSC and SSC from a representative experiment. (B) The abundance of pS6 was measured in splenocytes from CD45.1 WT and NK-Tgfbr2−/− mice that were mixed in a 1:1 ratio at the beginning of the culture. This procedure was used to minimize variations between experiments. Bar graphs show averaged MFIs from four mice from two experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test.

TGF-β inhibits mTOR and metabolic activity in mouse and human NK cells in response to IL-15

Next, we directly measured the influence of exogenous TGF-β on mTOR activity induced by IL-15 in splenic NK cells in vitro. TGF-β inhibited the induction of S6 phosphorylation (Fig. 4A). This effect occurred with the same kinetics as did the phosphorylation of SMAD2/3, suggesting that mTOR inhibition occurred just downstream of TGF-βR activation. We also observed SMAD2/3 phosphorylation in the absence of exogenous TGF-β, suggesting that splenic cells secreted active TGF-β under these conditions. Accordingly, the addition of a TGF-β–blocking antibody enhanced the mTOR activity induced by IL-15 in NK cells (Fig. 4A). Similar results were obtained from experiments with SB-431542, an inhibitor of the kinase activity of TGF-βRI (Fig. 4B). Neither TGF-β nor the anti–TGF-β antibody altered the extent of STAT5 phosphorylation stimulated by IL-15, thus showing a very specific inhibition of mTOR activity by TGF-β and no general impairment of IL-15 signaling (Fig. 4A). The inhibitory effect of TGF-β was not restricted to S6 phosphorylation but was also observed on other mTOR substrates, including 4EBP1 and Akt (Fig. 4C). This effect was comparable to that of the specific mTORC1 inhibitor rapamycin, with a notable difference regarding the phosphorylation of Akt, which was insensitive to rapamycin (27) but was inhibited by TGF-β (Fig. 4C). Moreover, TGF-β also substantially inhibited mTOR activity in human NK cells in response to a 1-hour treatment with IL-15 (Fig. 4D).

Fig. 4 TGF-β counters the IL-15–stimulated activation of mTOR and bioenergetic metabolism.

(A to C) Splenocytes from WT mice were stimulated for the indicated times with IL-15 alone or in combination with either anti–TGF-β or TGF-β. (A) The relative abundances of pS6, pSTAT5, and pSMAD2/3 were measured by flow cytometric analysis. Graphs show averaged MFI ratios in gated NK cells calculated from a total of three mice from three experiments, which were normalized to those of control (Ctrl), unstimulated cells. (B) The extent of phosphorylation of S6 and SMAD2/3 in WT NK cells was measured in the presence of the indicated concentrations of the TGF-βRI kinase inhibitor SB-431542. Data are means ± SD of three experiments. (C) Analysis of the phosphorylation of Akt, 4EBP1, and S6. Left: Histograms show the phosphorylation of the different proteins in NK cells under the indicated conditions and are representative of three experiments. Right: Bar graphs show averaged phosphoprotein MFIs in NK cells from four mice from two independent experiments. (D) Phosphorylation of S6 in human NK cells (CD56brightCD3). Human peripheral blood mononuclear cells (PBMCs) were cultured under the indicated conditions for 1 hour before they were analyzed by flow cytometry. Data are averaged pS6 MFIs in NK cells from seven donors from three independent experiments. (E) Splenocytes from WT mice were treated with the indicated reagents for 24 hours. The relative abundances of the indicated intracellular and cell surface proteins and the extent of 2-NBDG incorporation were measured by flow cytometry. Bar graphs show averaged MFIs calculated from three mice in two experiments. (F) Primary NK cells were cultured in low-dose IL-15 for 5 days, as described previously (51), and then were stimulated with IL-15 (100 ng/ml) for 20 hours, and NK cell metabolism was analyzed under the indicated conditions. ADG, OxPhos (OCR for O2 consumption rate), and GC were analyzed in real time. Bar graphs show the average of three independent mice each analyzed from three independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test. Rapa, rapamycin.

Because we previously demonstrated that mTOR regulates NK cell bioenergetic metabolism, we compared the effect of TGF-β and rapamycin on the following metabolic parameters: abundances of the amino acid transporters CD71 and CD98, FSC flow cytometry parameters proportional to cell size, 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose (2-NBDG) glucose incorporation, glycolysis [glycolytic capacity (GC) and acidification due to glycolysis (ADG)], and oxidative phosphorylation (OxPhos). TGF-β was as efficient as rapamycin in inhibiting the metabolic activity of NK cells irrespective of the parameter analyzed and in inhibiting the expression of the associated cell surface markers CD71 and CD98 (Fig. 4, E and F). Thus, TGF-β inhibits mTOR-dependent metabolic activity in NK cells stimulated by IL-15.

The inhibition of mTOR is not a result of decreased IL-15R abundance or inhibition of the cell cycle

Although the effect of TGF-β on mTOR activity in NK cells was rapid, we considered the possibility that it might be a consequence of the well-described inhibitory effect of TGF-β on cell cycle activity. To address this point, we first stimulated freshly isolated NK cells with IL-15 in the presence of TGF-β or a wide range of chemotherapeutic compounds that inhibit proliferation through various mechanisms. Doxorubicin, methotrexate, cyclophosphamide, and 5-fluorouracil (5-FU) all inhibited the IL-15–stimulated proliferation of NK cells to varying degrees (Fig. 5A); however, unlike TGF-β or rapamycin, none of these compounds inhibited the early mTOR activity stimulated by IL-15 (Fig. 5A). Second, we measured the effect of TGF-β on the incorporation of BrdU by proliferating NK cells in vitro in response to IL-15. We did not detect a substantial effect of TGF-β on NK cell proliferation before 48 hours of culture (Fig. 5B), well after the observed inhibitory effect of this cytokine on mTOR activity. These experiments excluded the possibility that mTOR inhibition was a consequence of cell cycle inhibition in TGF-β–treated NK cells.

Fig. 5 Early inhibition of mTOR activity by IL-15 is not a consequence of cell cycle inhibition but is independent of the amounts of T-bet and IL-15R.

(A to C) Splenocytes from WT mice were cultured in vitro under the indicated conditions. (A) S6 phosphorylation was measured after 1 hour (right), and cell proliferation was measured after 3 days (left). Data are means ± SD of four mice from two experiments. Doxo, doxorubicin; MTX, methotrexate; Cyclo, cyclophosphamide. (B) WT splenocytes were cultured in the presence of BrdU for the indicated times. The percentages of NK cells that were BrdU+ were determined by flow cytometric analysis. Data are means ± SD of a total of three mice from two experiments. (C) The relative abundances of CD122 (left) and CD132 (right) in the treated NK cells were measured at the indicated times. Data are means ± SD from three experiments. (D) Splenocytes from WT (control) and Tbx21−/− mice were cultured with IL-15 alone or with TGF-β in the presence of BrdU. The extent of S6 phosphorylation was measured at 1 hour (left), whereas the percentages of BrdU+ NK cells (middle) and the MFI of granzyme B were measured at 3 days. For pS6, results are shown as MFI ratios relative to those in cells treated with IL-15 alone. Data are means ± SD of five mice from two experiments. *P < 0.05, **P < 0.01 by Student’s t test.

Next, we considered the possibility that the inhibition of mTOR by TGF-β was indirectly a result of the inhibition of Tbx21 expression. Indeed, T-bet is required for the expression of the gene encoding CD122, the β chain of the IL-15R (28), and TGF-β inhibits Tbx21 expression (23). However, the cell surface abundance of CD122 (or of CD132, the γ chain of IL-15R) was not altered by TGF-β or the anti–TGF-β antibody, at least not in the first 4 hours after in vitro treatment (Fig. 5C). Moreover, TGF-β still had an inhibitory effect on NK cells isolated from Tbx21−/− mice, which lack T-bet, regardless of the parameter analyzed, that is, mTOR activity or cell proliferation, as assessed by BrdU incorporation (Fig. 5D). Together, these results suggest that mTOR inhibition is a proximal signaling event downstream of the TGF-βR in NK cells and not a distal effect that is indirectly a result of the inhibition of other biological processes.

TGF-β and rapamycin have similar effects on NK cell activation in vitro

To evaluate the contribution of mTOR inhibition to the inhibitory effect of TGF-β on NK cells, we compared the effects of TGF-β and rapamycin on NK cell activation parameters in vitro upon stimulation with IL-15 for 3 days. We assessed NK cell proliferation by measuring BrdU incorporation and the abundances of a large panel of cytotoxic or lymphocyte markers. TGF-β and rapamycin had equivalent negative effects on NK cell proliferation (Fig. 6, A and B). Flow cytometric analysis revealed that most of the markers that we analyzed were similarly regulated by TGF-β and rapamycin, which resulted in the coclustering of the TGF-β and rapamycin conditions by global clustering analysis (Fig. 6C). Many of the markers that were analyzed were decreased in abundance in the presence of TGF-β or rapamycin. For some of these, TGF-β was the more potent inhibitor (see, for example, granzyme B, CD24, and CD223), whereas the opposite was true for some other markers, including CD71 and CD98 (Fig. 6C). Note that the abundance of CD122, the IL-15Rβ chain, was substantially decreased in abundance by both rapamycin and TGF-β at this late time point. Thus, except for TNF-related apoptosis-inducing ligand (TRAIL), whose abundance was increased by TGF-β but was left unchanged by rapamycin, and CD62L, which displayed the reciprocal pattern, there was an overall strong similarity in the effects of both compounds. A similar conclusion was reached when we compared the effects of TGF-β and rapamycin on the cytotoxicity of NK cells toward YAC1 cells (Fig. 6D). Finally, TGF-β and rapamycin had analogous influences on the amounts of granzyme B and perforin produced by human NK cells, as well as on the extent of degranulation and IFN-γ and MIP-1β (macrophage inflammatory protein-1β) secretion in response to stimulation by K562 cells in the presence of IL-2 (Fig. 6, E and F). Thus, the effect of TGF-β on IL-2– and IL-15–mediated NK cell activation was recapitulated by inhibiting mTOR with rapamycin.

Fig. 6 TGF-β and rapamycin have similar effects on IL-15–induced NK cell activation and cytotoxicity in vitro.

(A and B) WT NK cells that had been allowed to proliferate in IL-15 were restimulated under the indicated conditions for 12 hours in the presence of BrdU. BrdU incorporation was measured by flow cytometry. (A) Representative flow cytometry density plots of NK1.1 and BrdU in gated NK cells. (B) Data are means ± SD of the percentage of BrdU+ NK cells from a total of four mice from two experiments. (C) Splenic NK cells from WT C57BL/6 mice were cultured for 3 days under the indicated conditions. The cells were then analyzed by flow cytometry to determine the relative abundances of the indicated proteins. Results are presented as a hierarchical cluster heat map. Each row denotes a parameter, and each column denotes a cell culture condition, as indicated. The color scale indicates relative protein abundance as determined by MFI. Dendrograms denote the Euclidean distances between clustered conditions. Results are representative of four independent experiments. (D) WT NK cells treated overnight under the indicated conditions were then cocultured for 4 hours with YAC1 target cells at the indicated effector-to-target (E:T) ratios. Data are means ± SD of cytotoxicity as determined by calculation of the percentages of propidium iodide–positive (PI+) target cells from four mice in two independent experiments. (E) Human PBMCs were cultured for 36 hours with medium alone or with IL-2 alone or in the presence of either TGF-β or rapamycin. The MFIs of granzyme B (left) and perforin (right) in gated NK cells were then analyzed by flow cytometry. Data are averaged MFI ratios in the treated NK cells relative to those in cells treated with medium alone. Data are from nine healthy donors in three independent experiments. (F) NK cells treated with IL-2 alone or in the presence of TGF-β, anti–TGF-β, or rapamycin were cocultured for 4 hours with K562 cells at a 1:1 ratio. Cells were then analyzed by flow cytometry to determine the percentages of cells positive for CD107a (left), IFN-γ (middle), and MIP-1β (right). Data are means ± SD of six donors. *P < 0.05, **P < 0.01 by Student’s t test.

Deletion of mTOR and constitutive TGF-β signaling have similar effects on NK cell development

Next, we sought to compare the effects of TGF-β signaling and mTOR inhibition on NK cells in vivo. We took advantage of mTorfl/fl mice (29) and Tgf-βRICA mice (30). The former mice enable mTOR deletion, whereas the latter mice enable constitutive TGF-β signaling in Cre-expressing cells. These mice were crossed with Ncr1Cre mice to generate NK-mTor−/− and NK-TgfβRICA mice. The percentage of peripheral NK cells was very low in both strains (Fig. 7A). Moreover, splenic NK cells in both mouse strains had similar phenotypes, with a predominance of immature CD27+CD11b NK cells (Fig. 7B). In the bone marrow, the frequencies and numbers of NK cells were similar in wild-type, NK-mTor−/−, and NK-TgfβRICA mice, whereas NK cells tended to be more immature in the NK-mTor−/− mice and NK-TgfβRICA mice than in the wild-type control mice.

Fig. 7 Constitutive TGF-β signaling and mTOR deletion have comparable effects on the development and maturation of NK cells.

(A and B) Ex vivo analysis of the percentages (top) and numbers (bottom) of NK cells (A) and NK cell maturation as determined by flow cytometric analysis of the relative abundances of CD11b and CD27 (B) in cells from the spleens and bone marrow of WT, NK-Tgfbr1CA, and NK-mTor−/− mice. Data are means ± SD of 10 experiments. (C) Flow cytometric analysis of the relative abundances of various cell surface or intracellular proteins in gated splenic CD11bCD27+ NK cells from WT, NK-Tgfbr1CA, and NK-mTor−/− mice ex vivo. Data are averaged MFIs ± SD of the indicated markers calculated from a total of six mice of each group from three experiments. Data were normalized to the MFIs of cells under control conditions. Histograms show FSC and SSC from a representative experiment. (D) Homeostatic NK cell proliferation, as measured by BrdU incorporation in the spleens and bone marrow of WT and NK-Tgfbr1CA mice. Bar graphs show the average percentages of BrdU+ cells from four mice in two experiments. (E) Splenocytes from the indicated mice were cultured for 1 hour in the presence or absence of IL-15. Gated CD11bCD27+ NK cells were analyzed by flow cytometry to determine the relative abundances of pS6 (left) and pSTAT5 (right). Data are means ± SD of three mice from each group. (F) Splenocytes from WT, NK-Tgfbr1CA, and NK-mTor−/− mice were cocultured for 4 hours with target YAC1 cells at a 1:1 ratio. NK cell degranulation was then measured by flow cytometry. Data are the mean percentages ± SD of CD107a+ cells among gated CD11b NK cells for the different mouse strains. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 by Student’s t test.

We performed an in-depth, flow cytometric analysis of NK cell phenotypes in both strains of mice, focusing our analysis on the splenic CD27+CD11b NK cells that were present in all strains to avoid a bias caused by the skewed maturation of cells in the NK-mTor−/− and NK-TgfβRICA strains. We then selected the flow cytometric parameters that were statistically significantly different between the control mice and at least one of the mutant strains of mice (Fig. 7C). Three groups of markers were identified: those that showed co-regulation in mTor−/− and TgfβRICA NK cells, which were statistically significantly different to those of wild-type NK cells (Fig. 7C, top row). This category included most of the markers analyzed and notably KLRG1, CD146, granzyme B, Ly49H, T-bet, and CD122. The other categories correspond to proteins whose abundances were not correlated between NK-mTor−/− mice and NK-TgfβRICA mice, which included the NK cell receptors 2B4 and NKG2D, whose abundances were decreased in mTor−/− NK cells, but not TgfβRICA NK cells (Fig. 7C, bottom row).

Next, we compared homeostatic proliferation and early phosphorylation events in response to IL-15 in NK cells from TgfβRICA mice and those from control mice. Similarly to mTor−/− NK cells, TgfβRICA NK cells exhibited defective proliferation in the bone marrow, but not the spleen (Fig. 7D). In comparison with control NK cells, TgfβRICA NK cells had a decreased amount of phosphorylated S6 (pS6), but not phosphorylated STAT5 (pSTAT5), when stimulated with IL-15 (Fig. 7E). Together, these data suggest that NK cells deficient in mTOR behave similarly to those expressing a constitutively active form of TGF-βRI, which further suggests the contribution of mTOR inhibition to the overall effect of TGF-β on NK cells. This inhibition is relevant to the antitumor function of NK cells, because mTOR-deficient NK cells are poorly responsive to IL-15 in vivo and have reduced effector functions upon engagement of activating NK cell receptors (9). Moreover, both mTor−/− NK cells and TgfβRICA NK cells exhibited a reduced ability to degranulate in response to YAC1 tumor targets (Fig. 7F).

DISCUSSION

The molecular mechanisms underlying the inhibitory activity of TGF-β on NK cells remain mostly unknown. Here, we provide evidence that a major target of TGF-β in NK cells is the serine and threonine kinase mTOR. As direct evidence, we showed that TGF-β signaling opposed the phosphorylation of the mTOR substrates S6, 4EBP1, and Akt in response to IL-15. This effect was very rapid, concomitant with SMAD phosphorylation, and thereby consistent with a proximal event downstream of the TGF-βR. Indirect evidence included the observation that mTOR deletion and constitutive TGF-β signaling in NK cells had comparable effects on NK cell development and maturation, as well as cytotoxic potential. Moreover, TGF-β and rapamycin, a highly specific mTORC1 inhibitor, had almost identical effects on IL-15–mediated NK cell activation in vitro, in terms of bioenergetics metabolism, proliferation, cytotoxic activity, and the expression of cytotoxicity-associated receptors and transcription factors. Similarly, previous reports described that TGF-β can be substituted for rapamycin to induce the differentiation of naïve T cells into Tregs (31) or follicular T helper cells (32) and that TGF-β and rapamycin have similar effects on anti-CD3– and anti-CD28–mediated T cell proliferation after in vitro culture for several days (33).

The control of metabolic activity emerges as a key event in immune cell regulation (10). Rapamycin and its derivatives are currently used for the prevention of kidney transplant rejection, the treatment of graft-versus-host disease, and the chemotherapy of some cancers (11). The mechanisms by which rapamycin suppresses immune responses have been extended from inhibition of T cell proliferation to suppression of dendritic cell maturation and sustenance of Tregs (34). Our own studies previously highlighted the crucial role of mTOR in NK cell activation. Rapamycin dampened in vivo NK cell cytotoxicity against MHC-I–negative target cells, demonstrating the central role of mTOR and bioenergetics metabolism in the control of NK cell function. Mechanistically, we found that mTOR was essential for controlling the production of several cytotoxic mediators, such as granzyme B and perforin, as well as increasing NK cell reactivity against target cells through positive feedback on signaling by activating NK cell receptors (9). The mTORC2-dependent regulation of the actin cytoskeleton may also contribute to the establishment of immunological synapses (35). The similarity of the effects of rapamycin and TGF-β thus suggests that a major effect of TGF-β in cytotoxic lymphocytes is to restrain bioenergetics metabolism by inhibiting mTOR activity to antagonize the effects of the proinflammatory cytokines IL-2 and IL-15.

The Smad-induced inhibition of Tbx21 expression was proposed to explain the inhibition of IFN-γ secretion by NK and T cells exposed to TGF-β (23, 36); however, as shown here, the amounts of T-bet and of its downstream target CD122 were not substantially altered by TGF-β before at least 4 hours of treatment. Moreover, we showed that the well-known antiproliferative effect of TGF-β occurred 2 days after the inhibition of mTOR activity. These data thus suggest that mTOR inhibition is a very early event during TGF-β signaling in NK cells and that it contributes to the inhibition of proliferation rather than acts as a surrogate marker of this inhibition.

How does TGF-β inhibit mTOR? A previous study of epithelial cells found many interactions between TGF-β signaling and other signaling pathways, which raises several possibilities (37). We found that TGF-β–mediated inhibition required the kinase activity of TGF-βRI, which controls Smad-dependent and Smad-independent pathways. Previous studies established that TGF-β activates Akt and mTOR in a Smad-independent way in epithelial cell types during the epithelial-to-mesenchymal transition, highlighting the context-dependent effects of TGF-β (38, 39). Indeed, the type 1 TGF-βR constitutively binds to FK506-binding protein-12 (FKBP12) (40), and this signaling molecule is released upon TGF-β signaling. Rapamycin also binds with high affinity to FKBP12 (41), and the FKBP12-rapamycin complex inhibits mTOR activity. Endogenous partners of FKBP12 that have a similar activity to that of rapamycin are yet to be identified, but FKBP12 represents a potential link between TGF-β and mTOR. Note that the development of Tregs, which depends on both TGF-β signaling and the inhibition of mTOR, is reduced in Fkbp12−/− mice (42). Moreover, in transplanted patients, treatment with tacrolimus (also known as FK506), an inhibitor of calcineurin and FKBP12, decreases the percentage of Tregs that are generated (43).

The early effect of TGF-β on mTOR activity is likely to be reinforced in an indirect way at later stages. Indeed, our results confirm previous findings that the amount of CD122, the IL-15RB subunit, is decreased upon TGF-β signaling in vitro or in vivo, which may contribute to the overall decrease in metabolic activity observed in NK cells at late stages of TGF-β signaling. Overall, mTOR inhibition by TGF-β is therefore likely to be due to direct and indirect effects that operate with different kinetics. Our results thus identify mTOR as a crucial integrator of cytokine signaling, which is capable of translating antagonistic signals into adapted cellular responses. Because mTOR activity is also modulated by antigen receptors, it is tempting to speculate that mTOR activity is a key molecular switch in immune cells such as NK cells, integrating signals from surface receptors for various metabolic, immunologic, or hormonal signals. In this context, it will be interesting to study the effects of inhibitory cues, such as adenosine or corticosteroids, on mTOR activity in NK cells.

We noted a few differences between the effects of TGF-β and rapamycin on NK cell activation in vitro. For example, TGF-β stimulated the production of TRAIL in NK cells, whereas rapamycin had the opposite effect. Similarly, constitutive TGF-β signaling in NK cells led to the cell surface expression of CD24, CD29, and BTLA (B and T lymphocyte–associated) in vivo, whereas mTOR deletion did not influence these markers. Reciprocally, mTOR deletion had an inhibitory effect on the generation of a series of markers, such as 2B4, NKG2D, the IL-12 receptor (CD212), and CD11c, whereas constitutive TGF-β signaling had no such effect. Therefore, TGF-β cannot be seen only as an “endogenous rapamycin”; it probably has activities beyond the control of mTOR that may be linked to the transcriptional activity of Smads on target genes. Future studies will have to identify all of the Smad target genes in NK cells to precisely address this point.

We found very little consequences of Tgfbr2 deficiency in NK cells at steady state. This result is in clear contrast with previous findings that transgenic expression of human TGF-βRII truncated from its kinase domain in CD11c+ cells induced substantial expansion in the size of the NK cell population (19). The reason for this discrepancy is unclear, because we found that NK cells from both Tgfbr2−/− mice and CD11c-dnTGFβRII mice were unresponsive to TGF-β in terms of Smad phosphorylation. A similar discrepancy has been reported in CD8+ T cells when comparing the effects of both genetic approaches (44). One can speculate that the Itgax promoter may be expressed at earlier stages of NK cell development than is the Ncr1 promoter that drives Tgfbr2 deletion in our system. Another possibility could be an effect of the DN receptor both in cis and in trans on myeloid cells interacting with NK cells and trans-presenting IL-15. Our work demonstrates the lack of a role of TGF-β on the homeostasis of NKp46+ NK cells, which includes the most conventional NK cells, suggesting that under physiological conditions, NK cells are not exposed to active TGF-β. Similarly, we found very little Smad2/3 phosphorylation in NK cells exposed to various inflammatory stimuli, such as TLR ligands, IL-15, or intracellular bacteria or viruses, which suggests that TGF-β plays a limited role in controlling NK cell activation in these contexts. By contrast, very strong Smad2/3 phosphorylation was observed in NK cells extracted from tumors, confirming the relevance of the TGF-β pathway in immunosuppression induced by the tumor microenvironment. This may be especially true in the context of breast cancer, because several studies have highlighted a role for TGF-β in suppressing NK cell activity in models of breast cancer in vitro (45) and in vivo (46, 47).

In conclusion, we demonstrated the existence of an evolutionarily conserved molecular pathway whereby TGF-β inhibits the metabolic activity of NK cells by opposing the induction of mTOR activity mediated by IL-2 or IL-15. These data specify a molecular mechanism for the immunosuppressive effects of TGF-β and point to mTOR as a key integrator of cytokine signals in NK cells and T cells. Our data provide a rationale for developing therapeutics aimed at increasing mTOR activity in tumor-infiltrating lymphocytes, such as NK cells, to restore their cytolytic activity by countering the effect of TGF-β.

MATERIALS AND METHODS

Mice

This study was performed in strict accordance with French recommendations for the ethical evaluation of experiments using laboratory animals, the European guidelines 86/609/CEE, and the QIMR Berghofer Medical Research Institute animal ethics committee. C57BL/6 and CD45.1 C57BL/6 mice were purchased from Charles River Laboratories. Ncr1iCre mice were crossed with mTorfl/fl (29) mice or LSL-TgfβRICA (30) or TGFβRIIfl/fl (26) mice. Tbx21−/− mice were previously described (48). All strains were bred at the Plateau de Biologie Expérimentale de la Souris (PBES) or the QIMR Berghofer Medical Research Institute. Litters of 8- to 24-week-old mice were used for NK cell analysis. In some experiments, mice were treated twice intraperitoneally with 5 μg of IL-15–IL-15Rα complex (eBioscience) at 24-hour intervals and then were sacrificed 24 hours later or were treated with 150 μg of poly(I:C) (Invivogen) and sacrificed 18 hours later. For in vivo BrdU incorporation, mice were injected twice at 24-hour intervals with 2 mg of BrdU in saline. BrdU incorporation was measured the next day. In some experiments, mice were infected with 2 × 105 TCID50 (median tissue culture infectious dose) influenza virus (H1N1 WSN strain) intranasally or 2 × 103 L. monocytogenes intravenously. Mice were sacrificed on day 6 after infection.

Flow cytometry

Single-cell suspensions of bone marrow, blood, spleen, and liver from mice were obtained and incubated with the appropriate antibodies (table S1). Whole-blood samples from healthy human donors were collected by venous puncture in heparin-containing vials. PBMCs were isolated by Ficoll gradient centrifugation. Intracellular staining of transcription factors or intracellular cytotoxic mediators was performed with Foxp3 Fixation/Permeabilization Concentrate and Diluent (eBioscience). Intracellular staining of phosphorylated proteins was performed with Lyse/Fix and Perm III buffers (BD Biosciences). BrdU incorporation was measured with the BrdU Flow Kit (BD Biosciences). Flow cytometric analysis was performed on FACSCanto, FACS LSR II, FACS Fortessa (Becton-Dickinson), or Navios flow cytometers (Beckman Coulter). Data were analyzed with FlowJo software (Tree Star).

Cell culture and stimulation

Splenic lymphocytes were prepared and cultured with the cytokines recombinant murine IL-15 (rmIL-15, 100 ng/ml) and recombinant human TGF-β1 (rhTGF-β1, 10 ng/ml), both from R&D Systems. In some culture experiments, we used anti–TGF-β blocking antibody (1 μg/ml, 1D11, Bio X Cell), 25 nM rapamycin (Sigma-Aldrich), the TGF-βRI kinase inhibitor SB-431542 (1 to 10 nM, Sigma-Aldrich), 5 μM doxorubicin (Healthcare), 50 μM methotrexate (Mylan), 50 μM 5-FU (Pfizer), 50 μM cyclophosphamide (Baxter), or 10 μM BrdU (Sigma-Aldrich). In experiments comparing mTOR activity between wild-type and NK-Tgfbr2−/− NK cells, splenocytes were mixed at a 1:1 ratio before stimulation to minimize variability linked to experimental conditions. Staining of cell surface and intracellular targets was then performed. PBMCs were isolated by Ficoll gradient centrifugation and stimulated with rhIL-2 (1000 U/ml, corresponding to ~15 ng/ml) or hIL-15 (100 ng/ml, PeproTech) in the presence or absence of TGF-β or rapamycin.

Assessment of glucose uptake

Glucose uptake was measured with 2-NBDG (Invitrogen). Freshly isolated cells were resuspended in RPMI 1640 medium (Life Technologies) in the presence of 100 μM 2-NBDG and were cultured for 10 min at 37°C before cell surface markers were stained.

Killing assays

Splenic cells from Rag2−/− mice were stimulated overnight with IL-15 alone or together with anti–TGF-β, TGF-β, or rapamycin and then were cocultured for 4 hours at different E:T ratios with YAC1 target cells that were previously labeled with carboxyfluorescein diacetate succinimidyl ester (CFSE) (Invitrogen). The percentage of dead cells within CFSE-positive YAC1 cells was measured by flow cytometry after staining with PI.

Degranulation assays

Human PBMCs were stimulated overnight with IL-2 alone or with anti–TGF-β, TGF-β, or rapamycin and then were cocultured for 4 hours with K562 cells at a 1:1 ratio. The percentages of NK cells that were positive for CD107a, IFN-γ, or MIP-1β were then measured by flow cytometry.

Tumor models

B16-F10 melanoma cells (2 × 105) or RM-1 prostate adenocarcinoma cells (2 × 105) were resuspended in phosphate-buffered saline and injected into the tail veins of wild-type control or NK-Tgfbr2−/− mice. The numbers of lung nodules were counted under a dissecting microscope 14 days after injection, as previously described (49). The NEU15 cell line was established from a spontaneous mammary tumor harvested from a mouse mammary tumor virus (MMTV)–neu transgenic female mouse (50). FVB/N mice were injected with 5 × 106 NEU15 cells into the fourth mammary fat pad, and tumors were harvested 7 weeks after injection. Tumor volume was between 124 and 628 mm3.

Seahorse analysis

OCR and extracellular acidification rate (ECAR) were measured in XF medium [nonbuffered Dulbecco’s modified Eagle’s medium containing 2 mM glutamine and 10 mM glucose (pH 7.4)] with the XFe24 Extracellular Flux Analyzer (Seahorse Bioscience). ECAR and OCR were measured under basal conditions and after the sequential addition of 2 μM oligomycin, 4 μM antimycin A and 0.1 μM rotenone, and 30 mM 2-deoxyglucose (all from Sigma-Aldrich), which enabled the accurate quantitation of oxygen consumption as a result of OxPhos and of ADG. Cultured NK cells were purified, activated with IL-15 (100 ng/ml) for 20 hours, and then plated (at 750,000 cells per well) in a Seahorse plate coated with Cell-Tak (BD Biosciences) for analysis.

Statistical analysis

Error bars represent the SD. Statistical analyses were performed with two-tailed t tests or nonparametric tests where appropriate. These tests were performed with Prism software (GraphPad). Levels of statistical significance are expressed as P values: *P < 0.05, **P < 0.01, ***P < 0.001.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/9/415/ra19/DC1

Table S1. Antibodies used in this study.

REFERENCES AND NOTES

Acknowledgments: We thank the core facilities of the SFR Biosciences, Gerland, J. Sutton (QIMR Berghofer), and K. Elder (QIMR Berghofer) for mouse breeding, maintenance, and genotyping. We thank D. Mittal (QIMR Berghofer) for technical assistance. Funding: The T.W. laboratory is supported by the Agence Nationale de la Recherche, European Research Council (ERC-Stg 281025), INSERM, CNRS, Université de Lyon, and ENS de Lyon. M.J.S. is supported by a National Health and Medical Research Council (NHMRC) of Senior Principal Research Fellowship (1078671) and Program grant (1013667). F.S.-F.G. is supported by a National Breast Cancer Foundation, an NHMRC Early Career Fellowship, and a Cure Cancer Australia Priority-Driven Young Investigator Project Grant. Author contributions: S.V., A.M., F.S.-F.G., R.L., J.R., M.G., S. Degouve, S. Djebali, A.S., E.C., and L.T. performed experiments; S.V., A.M., F.S.-F.G., J.B., J.C.M., C.C., J.M., N.D.H., L.B., D.F., M.J.S., and T.W. analyzed data; and M.J.S. and T.W. designed the study and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article