Increased Antitumor Effects Using IL-2 with Anti–TGF-β Reveals Competition between Mouse NK and CD8 T Cells

Natural killer cell–based immunotherapy is a promising treatment against multiple cancers as a result of the ability of NK cells to eliminate tumor cells without prior immunization (1). IL-2 is used widely to activate NK cells both in vivo and in vitro, and it is approved for treatment in metastatic melanoma and renal cell carcinoma (1, 2). However, as a cancer therapeutic, benefits in survival have been hampered (1, 3), in part because of limitations in systemic IL-2 administration and associated toxicities (4, 5), as well as the potential expansion of regulatory T cells (Tregs) by engaging the high-affinity IL-2R (CD25) (6).

Secretion of immunosuppressive cytokines, such as TGF-β, by Tregs and/or tumor cells results in NK cell suppression. TGF-β inhibits IFN-γ production, impairs degranulation, and decreases expression of activating receptors, such as NKG2D and/or NKp30 on NK cells, resulting in diminished tumor lysis (7, 8) and allogeneic bone marrow (BM) rejection (9). Furthermore, NK cell homeostasis (8) and ontogeny (10) are negatively controlled by TGF-β. Therapeutically, neutralization of TGF-β using mAbs, TGF-β receptor kinase inhibitors, or antisense TGF-β oligonucleotides led to promising results in several cancers by preventing tumor-sensitized Treg expansion, augmenting antitumor responses in an NK cell and/or CD8 T cell–dependent manner, and suppressing tumor progression and metastasis (6, 11–18). TGF-β blockade also restored NKG2D expression and IFN-γ secretion by NK cells (7).

Despite these promising results, immunotherapeutic strategies that favor NK cells by promoting immune activation and preventing immune suppression could lead to greater antitumor efficacy. We showed previously that the combination of anti-CD25 and IL-2 improved NK cell antitumor responses by eliminating Tregs (19). Additionally, the development of nanolipogels that allow sustained delivery of IL-2 combined with TGF-β receptor inhibitor resulted in delayed tumor growth because of the increased presence of NK cells and effector CD8 T cells at the tumor site (20). In this study, we investigated the efficacy of using anti–TGF-β (clone 1D11), which neutralizes the three isoforms of TGF-β, combined with low-dose (LD) IL-2 in NK cell and T cell expansion and function. We report that combination immunotherapy allows for greater expansion and activation of NK cells and CD8 T cells, increased antitumor effects, and diminished toxicities. Furthermore, mechanistic assessment revealed a dual regulatory role between NK cells and T cells, limiting each other’s expansion and effects that can account for the immunotherapeutic success of NK cell– and CD8 T cell–based cancer therapies.

The University of California, Davis Institutional Animal Care and Use Committee approved all studies and protocols. Female C57BL/6 mice were purchased from the Animal Production Area, National Cancer Institute (Frederick, MD). Perforin-deficient (C57BL/6-Prf1tm1sdz), B6Smn.C3-Faslgld (FasL^−/−), and wild-type (WT) counterparts were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were used at 8–12 wk of age and were housed under specific pathogen–free conditions.

Mice were treated i.p. with 240 μg anti–TGF-β clone 1D11 (National Cell Culture Center) every other day and/or with 0.2–1 million IU recombinant human IL-2 (National Cancer Institute repository, Frederick, MD) daily for 7 d. Rat IgG (rIgG; Jackson ImmunoResearch) and/or PBS were used as controls. Some mice received 200 μg anti-NK1.1 (clone PK136; National Cell Culture Center) or anti-CD8 (clone YTS169.4; Taconic) i.p. 2 d prior to anti–TGF-β and IL-2 administration. Organs were collected 1 d (24 h) or 3 d (72 h) after 7 d of treatment with IL-2.

Ab staining of single-cell suspensions was performed as previously described (21). A Foxp3 Intracellular Kit (eBioscience) was used, following the manufacturer’s instructions (19). PE anti–granzyme B or isotype control (Invitrogen, Grand Island, NY) was used for intracellular staining. Stained cells were analyzed with an LSRFortessa cytometer (Becton Dickinson, San Jose, CA), and FlowJo software (TreeStar) was used for data analysis.

NK cell cytotoxic function was determined by a standard 4-h [⁵¹Cr]-release assay against the NK cell–sensitive tumor cell line Yac-1 (American Type Culture Collection, Manassas, VA) (22) using treated splenocytes or purified NK cells (NK negative selection kit; STEMCELL Technologies, Vancouver, BC, Canada) as effector cells. CD8 T cell cytotoxic function was determined by a redirected assay, as previously described (23).

Two million splenocytes from C57BL/6 mice were cultured with 1000 IU/ml recombinant human IL-2 and 20–80 μg/ml anti–TGF-β in six-well plates by triplicate at 37°C and 5% CO₂. rIgG was used as control (80 μg/ml). At day 7, cells were collected, and viability was determined by trypan blue staining. Flow cytometry was used to determine the percentage of NK cells (CD45⁺CD3⁻NK1.1⁺). After T cell depletion using anti-Thy1.2 and rabbit complement (24), two million cells were also cultured in the presence of IL-2 and anti–TGF-β. At day 7, adherent lymphokine-activated killer cells were collected, and viability was measured by trypan blue.

IL-6 cytokine bead array and liver enzyme alanine transaminase [ALT; IDTox Alanine Transaminase (ALT) Enzyme Assay Kit, ID Labs, London, ON, Canada] levels were quantified in serum samples collected at 24 h posttreatment, according to the manufacturer’s instructions and as previously described (25). Absorbance of each well was determined at 340 nm on a plate reader (VERSAmax turntable plate reader). Each sample was run in triplicate. Histology analysis also was performed for liver, lung, and gut collected 24 h after the end of treatment, and pathology was assessed. Lungs, livers, and guts were flushed and fixed in 10% neutral-buffered formalin. Samples were embedded in paraffin, cut into 5-μm-thick sections, and stained with H&E at Histology Consultation Services (Everson, WA). All slides were coded and read in a blinded fashion. The scoring was determined by a board-certified pathologist using previously described criteria (25)

A total of 2 × 10⁵ cells of the 3LL Lewis lung carcinoma cell line (American Type Culture Collection) were injected i.v. into mice 4 d prior to initiation of combination therapy (CT). Some mice also received 200 μg anti-CD8 and/or anti-NK1.1 or rIgG on the same day of tumor injection and once a week until the end of the experiment. Mice were monitored for survival and euthanized when moribund.

Each experiment was performed at least two times (with three to eight mice/group), with the exception of the FasL^−/− experiment, which was done one time. The Student two-tailed t test, one-way ANOVA (Tukey posttest analysis), two-way ANOVA (Bonferroni posttest analysis), or log-rank test was used when appropriate to assess the statistical significance (Prism 4; GraphPad, La Jolla, CA). The p values < 0.05 were considered statistically significant.

We first investigated the effects of administration of LD IL-2 and anti–TGF-β (1D11 clone) CT on NK cell responses in vivo at 24 or 72 h after cessation of therapy. As expected, both NK cell numbers and percentages were significantly increased after CT compared with IL-2 alone (Fig. 1A, data not shown), whereas anti–TGF-β treatment alone did not have any effect.

More importantly, after CT, NK cell activation was significantly augmented, as evidenced by the higher expression of the activation marker Thy1.2 and enhanced tumor cell lytic capability compared with mice receiving IL-2 alone (Fig. 1B, 1C). These results correlated with significantly higher levels of granzyme B (data not shown). Additionally, slight upregulation of the NK cell receptors NKG2D, NKG2A, and Ly49G2 was detected, which also can be correlated with activation (data not shown) (21).

However, the effect of CT on the NK cell population was not observed 72 h after treatment cessation (Fig. 1A). Analysis of NK cells 48 h after the end of treatment revealed an intermediate number between the results obtained at 24 and 72 h posttreatment, suggesting an NK cell contraction (data not shown). This significant reduction in the total number of NK cells after cessation of treatment is likely due to the short in vivo half-life of IL-2 and 1D11, as well as NK cells’ need for exogenous cytokine stimulation due to the NK cell “cytokine addiction” phenomenon, indicating that the continued presence of the therapy is likely needed for sustained effects (26–28).

Given the reported role of TGF-β in NK cell ontogeny (10), we hypothesized that the benefit of CT is a consequence of the expansion from the mature NK (mNK) cell compartment, as well as increased NK cell development and maturation. In fact, an increase in NK cell progenitors was observed after NK cell stimulation. CT resulted in a significant increase in the numbers of precursor NK (CD3⁻CD122⁺NKG2D⁺NK1.1⁻DX5⁻), immature NK (CD3⁻CD122⁺NKG2D⁺NK1.1⁺DX5⁻), and mNK (CD3⁻CD122⁺NKG2D⁺NK1.1⁺DX5⁺) cell subsets in both spleen and BM (Fig. 1D, 1E). Expression of CD27 and CD11b also has been used to further differentiate mNK cells (29). In the spleen, an increase in both the immature-like phenotype (CD27^highCD11b^low, CD27^highCD11b^high) and the mature-like phenotype (CD27^lowCD11b^high) was observed (Fig. 1F), whereas in the liver and BM CT caused primarily increase in the immature-like phenotype (data not shown). These data suggest a possible de novo generation of NK cells in the BM and liver due to CT that could repopulate other organs, such as the spleen and lymph nodes (LNs). They also revealed a higher NK cell developmental rate that may contribute to the expansion of mNK cells after CT administration.

The effect of anti–TGF-β on NK cells was indirect, because in vitro coincubation of splenocytes with IL-2 and anti–TGF-β induced a higher percentage and number of NK cells in a dose-dependent manner, whereas no effect was observed when splenocytes were T cell depleted, indicating that T cells may directly regulate and inhibit NK cell expansion via TGF-β (Supplemental Fig. 1). These data indicate that CT results in the expansion of mNK cells, as well as improves NK cell cytotoxic capabilities and accelerates NK cell ontogeny.

Previous studies reported a positive impact of IL-2 and TGF-β neutralization on the adaptive immune cell compartment (15, 20, 30). Thus, as expected, CT also significantly enhanced CD4 and CD8 T cell numbers, whereas Tregs were not significantly affected compared with IL-2 alone (Fig. 2A–C). The impact of CT on T cell expansion can be the consequence of neutralizing the levels of TGF-β produced by IL-2–stimulated Tregs in the presence of anti–TGF-β (Fig. 2D). Similar to NK cells, the effect of CT on the T cell compartment was not observed 72 h posttreatment, suggesting a T cell contraction (Fig. 2B). The loss of T cells after stimulation can be a consequence of the role of IL-2 in activation-induced cell death (31).

Further analysis of the CD8 T cell population, a cell of interest that contributes to the antitumor benefits reported for CT, revealed a significant expansion of total effector memory (CD44⁺CD62L⁻) CD8 T cells and, in particular, cells with the phenotype of non–Ag-specific bystander-activated memory (CD44⁺CD25⁻NKG2D⁺) CD8 T cells after CT in the spleen (Fig. 2E–G), which correlated with a markedly increased lytic ability that was indicative of bystander T cell expansion in response to cytokine-alone stimulation (23) (Fig. 2H). Therefore, CT also affected the CD8 T cell compartment, resulting in bystander and memory cell expansion and enhanced lytic capabilities.

We next determined the safety and efficacy of CT using LD IL-2 (0.2 × 10⁶ IU) and compared it with treatment with high-dose (HD) IL-2 (1 million IU) monotherapy that induces significant toxicities in mice. The treatment with CT or HD IL-2 resulted in comparable expansion of Tregs, NK cells, and CD4 and CD8 T cells (Fig. 3A–C) and similar NK cell and CD8 T cell activity, as determined by granzyme B expression and cytolytic capability (Supplemental Fig. 2A, 2B). The NK cells expanded after CT or HD IL-2 displayed similar phenotypic profiles (Supplemental Fig. 2). However, CT did not lead to systemic toxicities, whereas HD IL-2 did, as evidenced by high serum levels of IL-6 and the liver enzyme ALT, as well as significant liver, lung, and gut damage (Fig. 3D–G). Histological examination showed that HD IL-2 resulted in higher periportal lymphocytic aggregates in the liver and significant peribronchial and perivascular infiltrates and interstitial lymphocytic pneumonitis in the lung (Fig. 3F). More importantly, ∼50% of the mice that received HD IL-2 did not survive the duration of the treatment compared with 100% of the CT group (Fig. 3H). Additionally, no differences were observed when anti–TGF-β was combined with HD IL-2 (Supplemental Fig. 2D), indicating that the additive effect of anti–TGF-β and IL-2 treatment is IL-2 dose dependent, because high levels of IL-2 are sufficient to induce maximal immune responses. These data suggest that the use of anti–TGF-β and LD IL-2 can achieve NK cell and T cell augmentation that is comparable to HD IL-2, without the marked toxicities.

We then evaluated the antitumor efficacy of systemic CT in a murine lung metastatic carcinoma model (3LL). Similar to what was described for other mouse tumor models (20, 30), CT administration significantly increased the survival of metastatic tumor–bearing mice compared with IL-2 alone (Fig. 4A). The antitumor responses of CT were both CD8 cell– and NK cell–dependent, because only the depletion of both populations abrogated the survival effects mediated by CT (Fig. 4B). These data indicate that CT results in enhanced antitumor responses by improving both NK cell and CD8 T cell compartments, similar to what was observed by Park et al. (20).

The minimal effect of either single NK cell or CD8 T cell depletion on the survival of tumor-bearing mice was initially surprising, particularly in the case of NK cell depletion because of the importance of NK cells in the initial stages of antitumor responses (Fig. 4B). When NK cells or CD8 T cells were depleted in tumor-bearing rIgG-treated mice without prior stimulation, depletion of NK cells resulted in slightly accelerated tumor progression and mouse death compared with control or CD8-depleted groups (data not shown), which demonstrated the need for NK cells in the initial phase of the antitumor responses. These data also indicated that after CT, NK cells and CD8 T cells could compensate for each other to mount strong antitumor effects.

In multiple infection models, it was postulated that regulation between NK cells and CD8 T cells exists indirectly through elimination of APCs, CD4 T cells or by direct CD8 T cell elimination (32–34). Furthermore, both populations respond to the same cytokines (35), such as IL-2, which could result in their direct competition. To investigate the possible existence of dual regulation between NK cells and CD8 T cells after immunotherapy, we depleted mice of either cell type 2 d prior to CT (Supplemental Fig. 3) and determined the impact of these depletions on the cell expansion of the remaining population. CD8 depletion resulted in a significantly greater expansion of NK cells in the spleen that was characterized by having both immature-like and mature phenotypes and similar receptor expression patterns as the nondepleted group (Fig. 5, data not shown). NK cell cytotoxic function was comparable (Fig. 5D), suggesting that the primary effects of CD8 T cells were related to NK cell expansion. Interestingly, prior NK cell depletion induced a greater expansion of CD8 T cells, whereas CD4 T cells and Tregs were not affected following CT (Fig. 6A). Particularly higher numbers of effector and bystander memory–activated CD8 T cells were observed (Fig. 6B–D), which also correlated with elevated cytotoxic function (Fig. 6E). Similar effects were observed in LNs (data not shown). A stimulatory environment seems to be required, because depletion of NK cells or CD8 T cells in resting mice did not increase CD8 T cells or NK cells, respectively (Supplemental Fig. 3B, 3C). Moreover, this regulation was independent of TGF-β, because similar effects were observed in IL-2–treated mice (Supplemental Fig. 3D–H). These results demonstrate a dual regulation between NK cells and CD8 T cells that could be a consequence of cytokine competition for IL-2, homeostatic proliferation, or direct elimination of CD8 T cells by NK cells.

Because a role for NK cells in preventing CD8 T cell–dependent toxicity during infection was suggested previously (32, 33), we also determined the impact of NK cell depletion on the induction of toxicity. Higher levels of ALT, but not IL-6, were detected in serum in the NK cell–depleted group compared with the non–NK cell–depleted groups after CT (Fig. 6F, 6G). These data point to a role for NK cells in restraining T cell numbers and activity, as well as limiting pathology.

There are multiple mechanisms that can be involved in this dual NK cell–CD8 T cell regulation. Some studies from infection mouse models showed a direct or indirect impact on CD8 T cells by NK cells in a perforin-dependent manner (32, 34) or CD4 T cell–dependent manner (33), respectively. However, despite the impaired NK cell function, NK cell depletion of IL-2–treated perforin-deficient mice still resulted in CD8 T cell expansion comparable to that in WT mice (Supplemental Fig. 4A–E). To exclude the CD8⁺ T cell regulation via CD4⁺ T cell elimination that was suggested by Waggoner et al. (33), anti-CD4 was used with or without anti-NK1.1 prior to CT. Anti–TGF-β was used in these experiments to minimize the effect of Treg depletion on CD4-depleted mice. Anti-CD4 administration did not alter CD8⁺ T cell expansion after CT unless NK cells also were depleted (Supplemental Fig. 4F, 4G), suggesting a direct regulation of CD8⁺ T cells by NK cells.

Interestingly, upregulation of Fas expression also was observed in CT-treated CD8 T cells after NK cell depletion, particularly in the effector and bystander memory–activated CD8 T cell subsets, whereas no differences were found in naive CD8 T cells (Fig. 6H) and in effector CD4 T cells (Supplemental Fig. 4H). High levels of Fas also were induced by IL-2 treatment alone (Supplemental Fig. 3I), which was reported previously (36). In contrast, no differences were found for PD-1 or Rae1δ (data not shown), which could indicate that CD8 T cell elimination takes place through PD-1–PD-L1 and NKG2DL-NKG2D pathways, respectively. Therefore, these results suggest the Fas-FasL pathway as a regulatory mechanism between NK cells and CD8 T cells after exogenous stimulation with IL-2 and/or anti–TGF-β. To determine the impact of Fas-FasL on the regulation of CD8 T cells by NK cells after immunotherapeutic stimulation, we administered CT to mice lacking FasL expression. CT resulted in a comparable NK cell expansion (Fig. 6I) while negatively affecting NK cell function (data not shown). However, a greater expansion of CD8 T cells was accomplished by CT in FasL-deficient mice compared with WT mice, with no differences if mice were depleted of NK cells prior to immunotherapy (Fig. 6J). Additionally, Fas expression was not altered by NK cell depletion prior to CT in FasL-deficient mice (Fig. 6K), whereas it was increased in WT mice (Fig. 6H, 6K). Altogether, these data suggest a direct elimination of CD8 T cells by NK cells in a Fas-FasL–dependent manner.

In conclusion, this study provides evidence for the existence of dual NK cell–CD8 T cell regulation and suggests a regulatory role for NK cells to prevent exacerbated CD8⁺ T cell responses during non–Ag-specific stimulatory models, similar to what was suggested previously for virus or Ag-specific stimulatory models.

The concurrent blockade of inhibitory signals while providing positive stimuli has ushered in a new era in cancer immunotherapy. We demonstrated that the combination of LD IL-2 (stimulatory) with anti–TGF-β (suppressive) resulted in significant increases in both activated NK cells and effector CD8 T cells correlating with longer survival in tumor-bearing mice. More importantly, CT did not induce appreciable toxicities, yet it resulted in comparable effects on NK cells and CD8 T cells in comparison with HD IL-2. The use of this CT (IL-2 and anti–TGF-β) also resulted in the generation of de novo NK cells, confirming a role for TGF-β in their maturation (10). In contrast to the study by Marcoe et al. (10), anti–TGF-β alone did not improve NK cell maturation or numbers. The accumulation of mNK cells observed in adult CD11cdnR mice is likely the consequence of the lack of response to TGF-β over time as a result of deficiency in TGF-β receptor signaling (14). Additionally, in our study, the use of anti–TGF-β with a short half-life may not be sufficient or complete enough to observe the effects on NK cell development without prior stimulation or unless myelosuppressive therapies, which promote NK cell ontogeny, are applied. Nevertheless, the impact of TGF-β neutralization on NK cell development during IL-2–dependent stimulation of NK cells provides another mechanism of action for this immunotherapeutic strategy that makes it more attractive than others because of its multilayered effect. Blockade of TGF-β prevents the immunosuppressive effects of IL-2–induced Tregs, allowing for greater expansion and improved functionality of NK cells, and it generates new mNK cells that will benefit from the therapy that might result in enhanced antitumor responses compared with other therapies that only target existing NK cells without generation of de novo NK cells.

A dramatic reduction in immune cell expansion was observed shortly after immunotherapy cessation. Previously, our group described a CD4 conventional T cell loss following immunotherapy with agonistic CD40 and IL-2 that was associated with IFN-γ–dependent upregulation of PD-L1 (37, 38); IFN-γ also was shown to negatively regulate influenza virus–specific CD8 T cell responses (39). It is possible that IL-2–dependent IFN-γ production can lead to immune cell loss following treatment cessation. Furthermore, cytokine deprivation is also important for Ag-stimulated T cell contraction (40). Because of the loss of effect after immunotherapy, it is clear that approaches similar to the one developed by Park et al. (20) might be necessary to achieve prolonged antitumor effects. However, it is important to take into account the induction of immune cell exhaustion because NK cell anergy after sustained stimulation with IL-15 (41) and CD8 T cell exhaustion were described after chronic virus infection (42).

In this study, we also demonstrated for the first time, to our knowledge, that a dual regulation between NK and CD8 T cells might exist after immunotherapy, which seems to have a stronger regulation toward CD8 T cells by NK cells. NK cell inhibition of CD8 T cells was shown previously in various infection models in which the regulation can be done by directly eliminating CD8 T cells in a perforin-, IL-10–, and/or NKG2D-dependent manner or indirectly by affecting the levels of mature dendritic cells and CD4 T cells (32, 33, 43–46). More importantly, this regulatory effect of the adaptive immune response by NK cells is not limited to infection models, and it was suggested for autoimmune disorders by limiting the proliferation of autoreactive T cells (47). For example, depletion of NK cells resulted in an increase in encephalitis and demyelization after Theiler’s murine encephalitis virus (48) and in various experimental autoimmune encephalomyelitis models (47). Additionally, Barber et al. (49) suggested a negative regulation of tumor-infiltrated CD8 T cell by NK cells in a low-immunogenic lymphoma mouse model. Our study, together with these other studies, suggests a general role for NK cells in the regulation and inhibition of the adaptive immune response that seems to occur as a mechanism to most likely prevent a strong immune response that otherwise could lead to autoimmunity.

Similarly, and in agreement with our study, CD8 depletion prior to mouse CMV infection also was shown to induce enhanced NK cell numbers and activity (50). Because NK cells and CD8⁺ T cells use the same cytokines for development, maintenance, and activation, such as IL-2 and IL-15 (35), it is possible that this dual regulation is due, in part, to competition for cytokines by both immune cells, limiting the ability of each population to proliferate after stimulation. Homeostatic proliferation also could partially explain our results. However, the lack of CD8⁺ T cell expansion after CD4⁺ T cell depletion, which creates a bigger niche for their proliferation, suggests otherwise. Because NK cells are able to mount a faster immune response compared with CD8⁺ T cells, the use of cytokines by NK cells could restrict, to a greater extent, the expansion of CD8⁺ T cells. Additionally, we observed a greater expansion of both effector and bystander memory–activated CD8⁺ T cell subsets when NK cells were not present in both spleen and LNs. We previously demonstrated that, after systemic administration of cytokines, non–Ag-specific bystander CD8 T cell expansion occurs primarily in the memory (CD44^high) population, which highly expresses CD122, the low-affinity IL-2R, compared with naive cells, highlighting the need for heightened cytokine levels in their expansion and maintenance of their activated state (23). This requirement for IL-2 would explain the differences observed in this population when NK cells are absent, indicating a cytokine competition between NK cells and CD8 T cells. Moreover, the effect of immunotherapy on the CD4 T cell compartment was minimal, because ∼30% of this compartment is Tregs after treatment. This lack of effect on the CD4 T cell compartment could be explained by induction of apoptosis via the PD-1–PD-L1 pathway after immunotherapy, as suggested previously (37, 38).

NK cell–CD8 T cell regulation can also be explained by direct elimination of activated CD8 T cells by NK cells. In a mouse CMV infection model, perforin- and IL-10–dependent regulation of CD8 T cell responses by NK cells prevented animal death (32). In our model, perforin did not seem to have a role in the NK cell–CD8 T cell regulation because no differences were observed in perforin-deficient mice. NKG2D could also play a role in CD8 T cell regulation, because expression of NKG2D ligands is observed 24–48 h after activation of CD8 T cells and was suggested in a CD8 T cell–priming model (34); however, we did not observe upregulation of Rae1-δ expression on CD8 T cells after treatment, suggesting that other ligands may be involved. Upregulation of polivirus receptor also was observed on Ag-specific T cells after Staphylococcus aureus enterotoxin B treatment that render elimination of T cells by NK cells in a NKG2D-DNAM1 dependent manner (51); again, no differences in polivirus receptor expression were detected on effector CD8 T cells. In contrast, our model suggests that the FasL-Fas pathway is involved in the direct regulation of CD8 T cells by NK cells. Upregulation of Fas is observed after IL-2 treatment and has been involved in the activation-induced death of T cells (31, 36). Additionally, Fas-FasL–induced cell death is important in the elimination of alloreactive CD8 T cells after hematopoietic stem cell transplantation (52, 53).

There are significant differences between mouse and human NK cells, including their tissue location. In contrast to humans, whose NK cells can be found in the LNs in steady-state conditions, mouse NK cells rapidly accumulate in the LNs only during inflammation, immunization, or infections (29, 54). It is likely that this dual regulatory ability of NK cells and T cells may occur at earlier stages of activation in human LNs compared with mouse models.

In conclusion, this study suggests the existence of a dual regulation between NK cells and CD8 T cells in a stimulatory non–Ag-specific model that can have tremendous impact on the efficacy of the immunotherapeutic strategy chosen. It also proposes critical protective role of NK cells in regulating T cell function during infection, as well as in other types of stimulation to prevent exacerbated responses that would result in detrimental toxicities. Finally, combination immunotherapy targeting positive and negative regulators may lead to greater antitumor efficacy.

We thank Weihong Ma and Monja Metcalf for technical assistance.

The authors have no financial conflicts of interest.