IL-21 is a recently identified cytokine that stimulates mouse NK cell effector functions in vitro. In this study we demonstrate that IL-21 achieves its stimulatory effect by inducing the development of mature NK cells into a large granular lymphocyte phenotype with heightened effector function. IL-21 treatment results in increased cell size and granularity and a corresponding decrease in cell viability and proliferative potential. These cells up-regulate the expression of the inhibitory CD94-NKG2A receptor complex and the activation markers CD154 and killer cell, lectin-like-receptor G1. Surprisingly, IL-21 treatment also results in down-regulation of the pan-NK marker, NK1.1. Coinciding with these cellular changes IL-21 enhances cytolytic capacity across a spectrum of target sensitivities and induces IL-10 and IFN-γ production. In vivo treatment with IL-21 results in a very similar activation and phenotypic maturation of NK cells as well as a potent increase in NK cell-mediated anti-tumor immunity that is perforin dependent. These developmental changes suggested that IL-21 functions to induce the terminal differentiation of mouse NK cells, resulting in heightened NK cell-mediated cytotoxicity and immune surveillance.
Natural killer cells are components of the innate immune system that play a protective role against some viral infections and tumors (1, 2). These functions are achieved by the ability to recognize and lyse target cells via effector molecules such as perforin and the provision of cytokines such as IFN-γ to activate further immune responses. Considerable advances have been made in understanding the receptors that activate and inhibit functionally mature NK cells; however, much less is known about the regulation of NK cell differentiation.
The emergence of functional NK cells in vivo is known to require signaling through the IL-15R (reviewed in Ref.3). The first committed NK cell progenitor expresses IL-2/15R β-chain (CD122) in the absence of NK1.1 (also known as CD161c or NKR-P1C) (4). Thereafter NK1.1 expression is maintained throughout the lineage and is followed by the rapid acquisition of the NKG2-CD94 and Ly49 receptor families (5). The factors that subsequently control mature NK cell differentiation and homeostasis remain unclear: however, NK1.1+ mature NK cells are known to continually require IL-15 for survival (6, 7, 8). In vitro, NK cells can be stimulated by IL-15 or IL-2; once activated, cellular differentiation can be further augmented by addition of IL-12, an initiator of IFN-γ production (9).
IL-21 is structurally related to the lymphoid cytokines IL-2, IL-4, and IL-15 and has been demonstrated to be expressed by activated CD4+ T lymphocytes (10, 11). The IL-21R is expressed in lymphoid tissues, shows homology to the β-chain of IL-2/IL-15Rs, and forms a complex with the common γ-chain (12). IL-21 has pleiotropic roles in the lymphoid lineages, including the promotion of CD8+ T cell function (13) and the inhibition of IgE production in B cells (14, 15). Within the NK cell lineage, IL-21 enhances maturation from human multipotent bone marrow progenitors and activates peripheral NK cells in the absence of other stimuli (10). In contrast, IL-21 has been reported to have an inhibitory effect on the IL-15-promoted expansion of mouse NK cells and to have no function in the absence of activating signals (13). Once mouse NK cells are stimulated, IL-21 has been demonstrated to increase cytotoxicity and IFN-γ production over that observed for IL-15 alone. This late function of IL-21 in the murine NK cell lineage is compatible with the phenotype of IL-21R−/− mice, whose in vivo resting NK cell compartment is normal (13).
We have analyzed the differentiation and functional characteristics of IL-21-treated mouse NK cells. In agreement with previous studies, IL-21 application leads to a reduction in NK cell expansion capacity in response to IL-15. This effect was accompanied by a dramatic increase in cell size and granularity, loss of expression of NK1.1, and up-regulation of the NKG2-CD94 complex. These observations correlated with a significant increase in cytotoxicity and a massive induction of cytokine secretion. Finally, we show that this activity is mirrored upon in vivo treatment with IL-21, whereby a proportion of NK cells again display a remarkably similar NK1.1low/CD94high phenotype and are activated in their cytotoxic activity. Significantly, IL-21-treated mice displayed a marked increase in their capacity to inhibit both liver and lung tumor metastases that was dependent upon both NK cells and perforin-mediated cytotoxicity. We propose a model in which IL-21 functions to induce the terminal differentiation of mouse NK cells.
Materials and Methods
Monoclonal Abs and flow cytometry
Anti-NK1.1 (PK136), anti-CD154 (MR1), anti-Ly49A (YE132), anti-Ly49C/I (SW5E6), anti-Ly49D (4E5/E1), anti-Ly49G2 (4D11), anti-TCRβ (H57-5921), anti-IFN-γ (HB170 and XMG1.2), and anti-IL-10 (JES5-2A5.1) were purified from hybridoma supernatants on protein G-Sepharose columns (Amersham Pharmacia Biotech, Arlington Heights, IL) and conjugated to biotin (Pierce, Rockford, IL) and FITC (Molecular Probes, Eugene, OR) as recommended by the suppliers. Biotinylated anti-CD94 (18d3), anti-DX5, anti-NKG2A/C/E (20d5), anti-Ly49F (HBF-719), anti-killer cell, lectin-like-receptor G1 (anti-KLRG1; 4 2F1), anti-CD122 (TM-β1), anti-IL-10 (SXC-1), and PE-labeled anti-CD49b (HMα2) were purchased from BD PharMingen (San Diego, CA). For flow cytometry, single-cell suspensions were stained with the appropriate mAb in PBS containing 2% FCS. Biotinylated mAb were revealed by PE-streptavidin (Southern Biotechnology Associates, Birmingham, AL). Cells were analyzed on a FACScan (BD Biosciences, Mountain View, CA), and cell sorting was conducted on high speed flow cytometers (Moflo; Cytomation, Fort Collins, CO; and BD Biosciences). Dead cells were excluded by propidium iodide (PI) staining.
NK cell isolation, propagation, and Western blotting analysis
Bone marrow- and spleen-derived RAG1−/− mature NK cells were enriched using DX5-conjugated microbeads as recommended by the manufacturer (Miltenyi Biotec, Auburn, CA). The enrichment step typically resulted in >70% CD49b-positive NK cells. Recombinant cytokines (R&D Systems, Minneapolis, MN) were used at the following concentrations unless specified: mouse IL-2, 50 ng/ml; mouse IL-21, 100 ng/ml; and human IL-15, 50 ng/ml. NK cells were cultured for 3–7 days ex vivo in IMDM supplemented with 50 μM 2-ME, 10% heat-inactivated FCS, and the appropriate cytokine. Membrane labeling of DX5+ cells was conducted using 10 mmol/liter CFSE (Molecular Probes) at 37°C for 10 min before culture in appropriate cytokines. All cells were CD49b+/CD122+ after 3-day culture with any cytokine combination. After 7 days cell morphology was examined by cytospin and staining with May-Grunwald-Giesma solutions. Analysis of perforin protein expression was performed on equal numbers of cells exactly as previously described (21). Unless otherwise noted in the text, all experiments were performed with bone marrow and spleen NK cells with identical consequences. Similar results were obtained in all experiments with either IL-2 or IL-15.
Cell proliferation and viability assays
DX5+ purified cell suspensions were cultured for 3 or 7 days in IL-15 (or IL-2) with or without IL-21, and viable cell number was determined by trypan blue exclusion or by flow cytometry using calibrite beads (BD PharMingen). Cell proliferation was determined after culture for 3 days with IL-15, followed by plating at 1 × 104 cells/well in 96-well, flat-bottom plates with or without IL-21 and wide range of IL-15 concentrations (0.1–100 ng/ml) for 24 h. Plates were then pulsed with 1μCi of [3H]thymidine (NEN, Boston, MA) and harvested after 4 h onto glass-fiber mats, and incorporation was determined by scintillation counting. Identical IL-15-supplemented cultures were propagated for 3 days, washed twice in PBS, and seeded with or without cytokine for 24 or 48 h. Cell viability/apoptosis of these cultures was determined using annexin V/PI staining; viable cells were annexin V−/PI−, and apoptotic cells were annexin V+/PI−. All assays were performed in triplicate.
Total RNA from NK cells was subjected to RT-PCR as previously described (22). PCR was performed for 20–27 cycles in the presence of [α-32P]dATP, and the appropriate linear amplification range was determined. The primers were: GM-CSF, 5′-ACCCGCTCACCCATCACTGTC-3′ and 5′-CTGGCCTGGGCTTCCTCATTT-3′; TNF-α, 5′-CAGGGGCCACCACGCTCTTC-3′ and 5′-CTTGGGGCAGGGGCTCTTGAC-3′; IL-10, 5′-TGCTATGCTGCCTGCTCTTACTGA-3′ and 5′-CCTGCTCCACTGCCTTGCTCTTAT-3′; and NK1.1, 5′-GCTGTGCTGGGCTCATCCT-3′ and 5′-GCAATGTGTACCTTAGTCCAATCC-3′. Hypoxanthine phosphoribosyltransferase primers were previously described (23). PCR products were resolved on 6% polyacrylamide gels by autoradiography. Primer pairs span introns.
Cytokine production assays
DX5+ cell suspensions were cultured in appropriate cytokines for 6 days, then washed, and 1 × 105 viable cells were seeded in fresh cytokine for 24 h, and supernatants harvested. IFN-γ levels were assayed by sandwich ELISA using HB170 as the capture mAb, XMG1.2 biotin as the detection mAb, and standard protocols. The IL-10 ELISA used JES5-2A5.1 as a capture mAb and SXC-1 biotin for detection. IL-21 was detected using purified and biotinylated goat anti-IL-21 (R&D Systems). Cytokine production was determined using recombinant mouse standards (R&D Systems). IFN-γ and IL-10 assays were performed three times in triplicate. IL-21 ELISAs were performed on three to eight mice per group. The lower signal intensity and the higher background staining resulting from the single available anti-IL-21 polyclonal Ab precluded an absolute determination of the serum IL-21 concentrations. Relative signal intensity between IL-21, control pORF, and naive serum was calculated and is shown in arbitrary units per milliliter.
Hydrodynamic gene transfer
The expression plasmid IL-21 pORF and the control vector pORF (provided by Dr. P. Hwu, National Cancer Institute, Bethesda, MD) were injected i.v. into mice as previously described (25, 26). Briefly, mice were injected with 20 μg of plasmid in a volume of 2 ml of saline over a 5- to 7-s period. The volume of saline was based on the age and weight of the mouse and did not exceed 10% of body weight. Liver and splenic mononuclear cells were analyzed on day 7. Serum was collected on days 1 and 3 after plasmid injection and was analyzed for IL-21 concentration by ELISA.
Tumor metastases models
Groups of five to 10 C57BL/6 wild-type (WT) or gene-targeted mice were pretreated with IL-21 or control pORF by hydrodynamic gene transfer 2 days before i.v. or intrasplenic (i.s.) inoculation with B16F10 cells at a dose of 5 × 105 cells, which has previously been shown to result in similar numbers of lung or liver metastases, respectively (27).
Renca renal cell carcinoma and DA3 mammary carcinoma.
BALB/c WT or gene-targeted mice were treated with IL-21 or control pORF 1 day after i.v. inoculation with Renca (2.5 × 105) (28) or DA3 (1 × 105) (27) cells. Some groups were additionally injected with Abs specific for NK cells (using 100 μg of rabbit anti-aGM1 Ab (Wako Chemicals) or anti-NK1.1, PK136 i.p.), TRAIL (N2B2; 250 μg), or Fas ligand (FasL; 250 μg; MFL-1) on days 0, 1, and 7 after tumor inoculation. Mice were then euthanized 14 days after tumor inoculation, the lungs or livers were removed, and surface metastases were counted with the aid of a dissecting microscope. In all metastasis models, the data were recorded as the mean number of metastases ± SE. Significance was determined using a Mann-Whitney rank-sum U test.
IL-21 limits NK cell proliferative signals mediated by IL-2 or IL-15
To investigate the function of IL-21 in mouse NK cell biology, we purified mature DX5+ cells from RAG1−/− bone marrow and spleen. DX5+ bone marrow cells were cultured in the presence of IL-15 or IL-2 (with or without IL-21) and displayed the NK cell surface phenotype CD49b+/NK1.1+/CD122+/TCRβ− 3 days ex vivo (data not shown). Cultures were then continued for an additional 4 days, and cell yield was determined. The addition of IL-21 resulted in a 5- to 10-fold reduction in total NK cell number, whereas IL-21 alone did not support proliferation or survival (Fig. 1 A). Similar effects of IL-21 have been observed for splenic NK cells (13) (data not shown).
To attempt to explain the decrease in cell yield in response to IL-21, mature DX5+ cells were cultured for 3 days in IL-15 alone. The addition of IL-21 to these cultures resulted in a decreased proliferation rate and reduced cell number across a broad titration of IL-15 concentrations (Fig. 1, B and C). These results suggested that IL-21 exerted its effect uniformly on activated NK cells. To test a direct effect of IL-21 on cell viability, DX5+ cells were grown for 5 days in the presence of IL-2 (or IL-15), then reseeded with or without cytokine for 24–48 h. IL-21 had very little survival-promoting ability on its own, and in the presence of IL-2 or IL-15 it had an antagonistic effect on cell viability (Fig. 1, D and E). IL-21 treatment resulted in an increased steady state apoptosis rate after 24 h (4–7%) compared with IL-15 alone (2–3%) as assessed by annexin V staining (n = 3 experiments). The vav-Bcl2 transgenic NK cells that expressed Bcl2 under the control of the pan-hemopoietic vav regulatory elements (20) were completely resistant to IL-21-induced cell death at both time points, indicating that the canonical apoptotic pathway was being induced by IL-21 treatment (Fig. 1 D).
These data confirm and extend previous studies (13) by identifying a role for IL-21 in controlling NK cell expansion in response to activating cytokines. This effect resulted from a combination of subtle decreases in cell division rate and viability that occurred throughout the culture period as opposed to direct effects on the initial resting NK cells.
IL-21 stimulated NK cells acquire a large granular lymphocyte phenotype
NK cells cultures grown in the presence of IL-15 or IL-2 proliferated extensively and displayed a characteristic NK cell morphology with invagination of the cytoplasm and some granularity (Fig. 2,A). In contrast, cultures propagated in the presence of IL-21 displayed a large, round, and highly granular phenotype after 7 days ex vivo (Fig. 2,A). This increased size and granularity were also readily apparent by flow cytometric analysis and were accompanied by a large increase in perforin expression (Fig. 2, B and C).
The pan-NK surface Ag NK1.1 is down-regulated by IL-21
The antiproliferative and cell morphological effects of IL-21 outlined above suggested that IL-21 is inducing a differentiation or maturation process in NK cells. We then examined the cell surface phenotype of bone marrow and splenic NK cultures after 7 days with or without IL-21. Strikingly, whereas NK cells propagated in the presence of IL-2 or IL-15 abundantly expressed the pan-NK cell marker NK1.1, IL-21 induced a down-regulation of surface NK1.1 levels (Fig. 3, A and B). This effect occurred in bone marrow- and spleen-derived NK cells and in the presence of IL-2 or IL-15. The reduction in NK1.1 was dependent on the IL-21 concentration, as the mean fluorescence index was proportional to a titration of IL-21 levels (Fig. 3 D).
Costaining with CFSE was used to examine whether NK1.1 down-regulation was associated with cell division number. Examination of the CFSE profiles indicated extensive proliferation of the NK cells in all conditions, with the majority of cells having undergone at least four divisions (Fig. 3, A–C). An overlay of IL-15 with or without IL-21 bone marrow culture histograms indicated very similar cell division profiles between the samples, with a relatively minor, but consistent, increase in cell number in the earlier three divisions in the presence of IL-21 (Fig. 3,C). The identical levels of NK1.1 down-regulation within individual division cohorts indicated a lack of division association to this differentiation event (Fig. 3, A and B). Finally, to test whether the NK1.1 down-regulation occurred at a transcriptional level, NK cells were sorted from IL-2 with or without IL-21 bone marrow cultures and subjected to quantitative RT-PCR analysis (Fig. 3, E and F). These data clearly indicate that the loss of NK1.1 on the surface of IL-21-treated cells is a result of decreased transcriptional output from the NK1.1 (NKR-P1C) gene.
IL-21 modulates NK cell surface marker expression
Having observed the down-regulation of NK1.1 by IL-21, we examined a large panel of NK cell surface molecules to begin to functionally and phenotypically describe this differentiation event. The activating and inhibiting receptors of the Ly49 family are stochastically expressed on individual NK cells, such that individual receptors are expressed on only a subset of NK cells (29). Therefore, alterations in the frequency of individual receptor specificities by IL-21 would be indicative of the selective expansion or depletion of particular NK cell subsets by the cytokine. The Ly49 family members Ly49A, C/I, and G2 were equivalently expressed on NK cells regardless of whether IL-21 was added to the cultures (Fig. 4,A). In contrast, IL-21 induced a consistent decrease in the number of Ly49D+ cells and a near total loss of Ly49F+ cells (Fig. 4,A). In addition to the Ly49 family, we examined a panel of other NK-associated or activation-induced receptors. The majority of proteins, including CD49b (Fig. 4,B), MacI (CD11b), IL-2/15Rβ (CD122), FcRγ (CD16), B220, CD28, CD69, and 2B4 (CD244), were unchanged as a result of IL-21 exposure (data not shown). In contrast, four Ags, CD94, NKG2A/C/E, KLRG1, and CD154 (CD40 ligand (CD40L)), were modulated by IL-21 (Fig. 4,B). KLRG1 is induced in vivo upon activation of NK cells via pathogenic and nonpathogenic stimuli (30), whereas human CD40L is regulated by IL-2 activation (31). The addition of IL-21 induced both these molecules, although in distinct manners. KLRG1 was expressed on only a minority of treated NK cells (20–30%), whereas the low CD40L levels in IL-2 were uniformly up-regulated by IL-21 (Fig. 4,B). Most interesting was the regulation of the CD94 molecule, which forms a heterodimer with members of the NKG2 family (Fig. 4, A, C, and E) (32). IL-2 (or IL-15)-stimulated cultures contained two distinct CD94+ populations (termed low and high), whereas the NKG2A/C/E mAb recognized 25% positive and 75% negative cells. Examination of the CD94-NKG2 profiles after IL-21 treatment revealed a shift in the populations, such that both Ags were uniformly positive (Fig. 4,B). To determine whether this phenotype is the result of up-regulation of CD94-NKG2 expression or the selective outgrowth of the CD94-NKG2high population, we isolated NK1.1+/CD94low and NK1.1+/CD94high cells from RAG1−/− spleen and bone marrow (Fig. 5,A). These populations were then propagated in IL-2 with or without IL-21 for 7 days and re-examined for CD94-NKG2 expression. These data clearly indicated that CD94-NKG2high cells continued to express high levels of the Ags with or without IL-21, whereas CD94-NKG2low cells induced high levels of the receptor complex and concurrent down-regulation of NK1.1 upon IL-21 exposure (Fig. 5,C). CD94-NKG2low cells exposed to only IL-2 retained low levels of CD94-NKG2 expression (Fig. 5 C). The NKG2 locus consists of three genes (A, C, and E) that are all recognized by the 20d5 mAb; however, RT-PCR analysis has indicated that the NKG2A message represents the majority of the total NKG2 transcripts (33). To investigate whether IL-21-induced up-regulation of the NKG2-CD94 dimer resulted in altered NKG2 gene usage, we repeated RT-PCR analysis and gene-selective restriction enzyme digestion as previously reported (33). In accordance with the literature, the majority (95%) of transcripts were NKG2A, a finding that did not change upon the addition of IL-21 to the cultures (data not shown).
Functional maturation of NK cell cytotoxicity upon IL-21 exposure
The phenotypic changes, including increased perforin expression, we have observed as a consequence of IL-21 stimulation led us to hypothesize that these alterations would correlate to changes in NK effector functions. IL-21 has previously been reported to enhance the cytotoxic activity of IL-15 or poly(I:C)-activated splenocyte cultures against the canonical NK cell-sensitive YAC-1 target cells (13). To extend these findings we performed lysis assays using the RMA lymphoma cell line (MHC class I+) as well as the TAP-deficient RMA-S (MHC class I−) and a variant of RMA-S expressing Rae-1β (RMA-S-Rae-1β) (24). Rae-1β is one of a family of inducible ligands for the activating receptor NKG2D that represents one of the initial mechanisms for detecting abnormal or stressed cells in vivo (34). Bone marrow and splenic NK cells propagated for 7 days in IL-2 (or IL-15) showed cytolytic activity against all three targets, with RMA-S-Rae-1β having much greater sensitivity than RMA-S while, as expected, the RMA parental line was the most resistant (Fig. 6). Interestingly, although IL-2 and IL-15 were equivalent in all other aspects examined, IL-15 was more potent at inducing target cell killing than IL-2 (Fig. 6). The addition of IL-21 to the cultures resulted in marked enhancement of target cell lysis across all three lymphomas, including enhancement of the already cytolytic NKG2D pathway (Fig. 6). This functional enhancement provides further evidence that IL-21 has induced the cellular differentiation of activated NK cells.
IL-21 stimulates the maturation of cytokine production and secretion in NK cells
In addition to cytolytic activity, the predominant biological activity of NK cells is the production of immunoregulatory cytokines, especially IFN-γ. We have examined the cytokine production profiles of IL-2 (or IL-15)-treated NK cell cultures with or without IL-21. RT-PCR analysis of bone marrow- and spleen-derived NK cells indicated that the transcription of some cytokines, such as TNF-α, did not change in response to IL-21 (Fig. 7,A). In contrast, IL-10 was induced, and GM-CSF repressed in an IL-21-dependent manner in the same cell populations (Fig. 7,A). Previous reports have indicated that IL-21 has a stimulatory effect on IFN-γ production (13). We have confirmed these findings with purified bone marrow and spleen NK cells (Fig. 7,B). IL-21 induced massive IFN-γ secretion in cultures derived from either tissue. The induction was apparent at 3 days ex vivo and was further enhanced by 6 days (data not shown) and 7 days (Fig. 7,B). A similar synergistic induction profile of IL-10 was observed for both tissues in the presence of IL-2 and IL-21 (Fig. 7 C). In contrast, IL-21 induced little IL-10 secretion (<2 ng/ml) in the absence of IL-2 stimulation.
IL-21 promotes the phenotypic and functional maturation of NK cells in vivo
To date, no studies have been reported on the in vivo effects of IL-21. To determine whether the phenotypic and functional changes we observed in vitro also occur in vivo, we used a hydrodynamics-based gene delivery technique to express mouse IL-21. This procedure has been demonstrated to result in transient high level expression of injected plasmids in hepatocytes (25, 26). A single injection of 20 μg of mouse IL-21 pORF plasmid DNA resulted in significantly elevated concentrations of circulating IL-21 in vivo as detected by ELISA (Fig. 8,C). Serum levels of IL-21 were highest 24 h after injection and were decreased, but still detectable, on day 3. IL-21 levels were below the ELISA detection level in the sera of the majority of mice injected with the same amount of control pORF plasmid (six of seven mice; Fig. 8 C). Western blot analysis of IL-21 and control pORF-treated livers showed similar IL-21 expression on days 1 and 3 postinjection (data not shown).
Liver and spleen NK cells from IL-21 or control pORF-injected RAG1−/− mice were analyzed by flow cytometry on day 7 after injection. Cellularity was increased 2- to 3-fold in the spleen, liver, and bone marrow on day 7 after IL-21 pORF injection (data not shown). NK cells were identified by staining with CD122 (IL-2/15Rβ) and were examined for NK1.1, DX5, and CD94 expression. Whereas all CD122+ NK cells were DX5 (CD49b)+, a proportion (∼19%, representative of 10 mice) had down-regulated NK1.1, and the majority had increased CD94 expression. These phenotypic changes were highly reminiscent of the effects of IL-21 in vitro and provide the first evidence of NK1.1− mature NK cells in vivo (Fig. 7 A).
To determine whether the NK cell effector function was also enhanced in IL-21-treated mice, we examined the cytotoxicity of their freshly isolated spleen and liver cells against RMA-S target cells. In agreement with our in vitro data, NK cells from IL-21 pORF-injected mice were highly cytolytic compared with those from pORF control mice (Fig. 8 B). Therefore, in vivo-stimulated NK cells responded with striking similarity to those treated with IL-21 in vitro.
IL-21 stimulates NK cell- and perforin-mediated antimetastatic activity
The enhanced effector function observed in the IL-21-treated mice raised the possibility that IL-21 may also enhance the natural antitumor activity of NK cells. We have used a number of well-established metastases models to assess the antitumor effects of systemic IL-21 (27, 28, 35). In these models tumor cells can be injected i.s. (B16F10 melanoma cells) to obtain liver metastases and i.v. to obtain lung metastases (B16F10, Renca renal cell adenocarcinoma, and DA3 mammary carcinoma cells).
In the B16F10 model, IL-21 or control pORF plasmid was injected on day 0, and cells were injected i.v. or i.s. 2 days later. The effect of IL-21 expression on lung and liver metastases was measured 14 days after B16F10 injection. IL-21 treatment resulted in a >80% reduction in tumor colonies in the lung and the virtual absence of liver tumor colonies in C57BL/6 WT mice (Fig. 9,A). By contrast, pORF plasmid treatment was without effect. The heightened response in liver was presumably a result of higher local levels of IL-21 produced after plasmid integration into hepatocytes (25). This effect was independent of T and B cells, but was dependent on NK cells, as RAG1−/− mice were normally responsive and IL-21 was ineffective in NK cell-depleted (anti-aGM1 or anti-NK1.1 mAb) mice (Fig. 9). Importantly, the antimetastatic activity of IL-21 required perforin, but not IFN-γ, production (Fig. 9 A). Similarly, IL-21 was effective in both IL-12−/− and IL-18−/− mice (data not shown), further supporting a role for IL-21 in enhancing the cytolytic granule function independently of inflammatory cytokines.
To assess the role of FasL and TRAIL in IL-21-mediated antitumor activity, we used two previously characterized BALB/c models: the FasL-resistant, TRAIL-sensitive Renca renal cell adenocarcinoma (28) and the FasL-sensitive, TRAIL-sensitive DA3 mammary carcinoma (36). Both models produced liver metastases after i.v. injection into pORF-treated WT and gene-targeted mice (Fig. 9, B and C). Similarly to the B16F10 model, IL-21 treatment resulted in a >80% reduction of liver metastases that was NK cell and perforin dependent and, in the case of DA3, insensitive to Ab-mediated depletion of FasL-expressing cells (Fig. 9, B and C). Interestingly, both models showed only a marginal dependence on the TRAIL pathway (Fig. 9, B and C). In summary, in four independent experimental tumor models IL-21 exerted a pronounced antimetastatic effect, primarily through the up-regulation of the perforin pathway.
In this study we have attempted to define the biological role played by IL-21 in NK cell differentiation. The addition of IL-21 to NK cell cultures led to a number of striking phenotypic changes, including reductions in proliferation, viability, and NK1.1 expression; dramatic increases in cell size and granularity; and changes in cell surface marker expression. In addition, IL-21 treatment led to a maturation of NK cell functions, including cell lysis and cytokine secretion. These effects were also apparent upon in vivo IL-21 treatment and resulted in markedly enhanced natural antimetastatic activity of NK cells.
The functionality of NK cells is regulated by the combination of activating and inhibitory signals (29, 37). The changes in cell surface expression of a number of NK receptors in the presence of IL-21 and the increased cytolytic capability and cytokine production of the same cells raised the possibility that these events were causally related. Two families of receptors, the Ly49 and CD94-NKG2 families, are important regulators of cellular responses. IL-21 stimulation did not alter the expression of most tested Ly49 molecules, with the exception of a reduction of Ly49D and Ly49F levels (Fig. 5,A). In contrast, IL-21 induced CD94high-NKG2+expression in the entire CD94low-NKG2− population (Fig. 6). These phenomena may be related, as it has been reported that Ly49D expression negatively correlates with NKG2A expression (38). NK1.1, a commonly used pan-NK marker, was expressed uniformly by NK cells in the presence of IL-2 or IL-15, but was significantly down-regulated by in vitro and in vivo IL-21 stimulation. This down-regulation occurred at the transcriptional level and provides the first evidence for an NK1.1low/− NK cell population in C57BL/6 mice. It is possible that IL-21 has a specific function that requires NK1.1 repression; alternatively, as NK1.1 is often used as a defining marker for the NK cell lineage, it is possible that NK1.1− NK cells arise in a variety of physiological settings and have not been recognized as members of the NK cell lineage. The ligand for NK1.1 is not known; however, cross-linking of NK1.1 with an mAb is a potent inducer of IFN-γ (39, 40). Moreover, the production of IFN-γ is known to inversely correlate with NK1.1 expression levels (39). Therefore, it is possible that NK1.1 down-regulation is directly related to the NK cell activation state induced by IL-21. As NK1.1 and Ly49D are activating receptors, and the CD94-NKG2A dimer is inhibitory, it could be argued that the expected consequence of IL-21 stimulation would be to inhibit NK cell-mediated cytotoxicity. In contrast, IL-21 treatment resulted in enhanced killing activity against a selection of target sensitivities. The presumption that IL-21 induced a high state of NK cell activation is supported by the increase in expression of CD154 and KLRG1, two markers associated with activated NK cells (30, 31). Intriguingly, KLRG1 is also associated with the loss of proliferative capacity of human effector T lymphocytes and thus may be indicative of the IL-21 antiproliferative effect (41).
Terminal differentiation is the term used to describe the attainment by a cell of its final cellular functional characteristics. The data presented in this study have led us to propose that IL-21 induced the terminal differentiation of mature NK cells. We have used the following criterion to support this hypothesis: 1) decreased cell proliferation, 2) decreased cell half-life, 3) increased cell size and granularity, 4) maturation of cell surface marker expression, 5) enhancement of cytotoxic effector function, and 6) maturation and enhancement of cytokine secretion profile. This differentiation involved a synergistic interaction between IL-15 (or IL-2) and IL-21. Whereas IL-21 alone was not able to support the proliferation or survival of NK cells, IL-15 was sufficient to activate the cells to proliferate and undergo a degree of differentiation into functional NK cells. This already considerable ability to lyse target cells as well as to secrete IFN-γ, however, was not associated with the other maturation indicators outlined above. The addition of IL-21 was required to fully engage this differentiation program. Interestingly, the effect was similar when IL-21 was added initially or after 3 or 5 days of culture in IL-15, indicating that the combination of IL-15/IL-21 was not selectively amplifying a distinct NK cell precursor, but was acting on the entire population (Fig. 1). Moreover, no significant differences were observed between the isolation of DX5+ cells from bone marrow and spleen, suggesting the resting mature NK cells were functionally similar between these tissues.
The development of mouse NK cells in vivo has been proposed to involve five stages to the mature NK cell (5). The mature stage V cell has a cell surface phenotype very similar to that of the IL-15 (or IL-2)-stimulated NK cells described in this study (CD122+/NK1.1+/DX5+/MacI+), with mosaic expression of CD94-NKG2 and Ly49, and a degree of cytotoxicity and IFN-γ production (5). IL-21 treatment therefore generates a cell with a distinct cellular phenotype from that currently identified for mouse NK cells in vivo (CD122+/NK1.1−/CD94high). Previous studies have suggested that the final differentiation of NK cells relies on, in addition to IL-15, the in vivo activation of the IFN response by IFN-αβ (42, 43) or in vitro stimulation by IL-12 (44). Our in vitro and corroborating in vivo data suggest that IL-21 is a key player in NK cell differentiation. Moreover, preliminary data demonstrate distinct differences between IL-12 and IL-21 when acting in concert with IL-15. Similarly to IL-21, IL-12 acts in synergy with IL-15 to induce IFN-γ production; however, in contrast to IL-21, IL-12 does not enhance cytotoxicity levels over those seen for IL-15 alone (data not shown).
Human NK cells have been proposed to differentiate in vitro along a pathway where IL-2-responsive cells proceed from IL-13+ to IFN-γ+ mature cells through a double-positive intermediate (45). The comparison of mouse and human cell differentiation stages is complicated by the fact that human NK cells are characterized by CD56bright and CD56dim populations, an Ag not expressed by mouse NK cells (46). By a number of criteria, IL-21-treated mouse NK cells appear similar to CD56bright cells; that is, they have high levels of CD94-NKG2A, IL-10, and IFN-γ production (47). However, in contrast to CD56bright cells, IL-21 treatment does not decrease mouse CD16 levels. Moreover, using ELISA and RT-PCR analysis, we detected only minimal transcription and no IL-13 secretion in mouse NK cell cultures (data not shown). NK cells have also been demonstrated to express a number of cytokines, including TNF-α, IL-5, IL-10, and GM-CSF (48, 49, 50, 51, 52). Our RT-PCR and ELISAs indicate that IL-15-treated mouse NK cells transcribe IFN-γ, TNF-α, GM-CSF, and IL-5, but little or no IL-4, IL-10, or IL-21 (Fig. 7 and data not shown). Upon IL-21 treatment there is further induction of IFN-γ, loss of GM-CSF, and initiation of IL-10 production (Fig. 7). These results support a role for cytokine expression maturation in mouse NK cells; however, the existence of the IL-13+-IFN-γ+ pathway proposed for human cells remains to be determined. Human CD94 has been shown to correlate with IL-13 loss and IFN-γ acquisition, a finding compatible with our observation that IL-21 up-regulates IFN-γ and CD94 expression (Figs. 5 and 7) (45).
In summary, the studies presented in this study show that IL-21 functions in vitro and in vivo to mediate the functional maturation of mouse NK cells. As the only known source of IL-21 is activated CD4+ Th2 cells (10, 53), it has been proposed that IL-21 functions as a mediator of the transition from innate to adaptive immunity (13). In this model NK cells are induced to proliferate in response to a viral infection by local stimuli such as IFN-αβ and IL-15. Upon the recruitment of Ag-specific T cells to the immune reaction, CD4+ Th2 cell-produced IL-21 would limit the NK cell expansion and enhance CD8+ T cell effector functions. However, we propose a more extensive role for IL-21 in NK cell biology. Rather than simply reducing the half-life of activated NK cells, IL-21 induces full differentiation of the lineage. The consequences are enhanced NK cell production of cytokines such as IFN-γ and IL-10, potentiated cytotoxic function, and limited NK cell proliferation. This model is also compatible with the normal numbers of mature NK cells reported in IL-21R−/− mice (13), as our results extend the differentiation pathway of NK cells beyond the stage examined previously. The recent postulated NK cell regulation of CD4+ T cell responses via IL-10 production (54) raises the possibility of an important network between CD4+ T cells, APC, and NK cells that enables both the functional maturation of NK cell responses and the subsequent development of adaptive immunity.
This study also provides key insights into the in vivo role of IL-21 in NK cell function and differentiation. Importantly, systemic IL-21 generated a NK cell population very similar in terms of cell surface phenotype and cytotoxic function to those NK cells generated by IL-21 in vitro (Fig. 8). Our data are also the first to illustrate the antimetastatic activity of IL-21. This antitumor effect was clearly mediated by NK cells and perforin-dependent cytotoxicity. Surprisingly, IL-21 treatment did not require host expression of IL-12, IL-18, or IFN-γ (Fig. 9 and data not shown), suggesting that the increased cytolytic potential of the IL-21-activated NK cells was sufficient to suppress tumor metastases. Although perforin has been shown to be a major contributor to NK cell-mediated control of metastases (36, 55), many cytokines and other biological response modifiers that activate NK cells invariably display some or complete IFN-γ dependence in their action (27, 56). This IFN-γ independence was mirrored in vitro, as IFN-γ−/− NK cells responded to IL-21 in an identical manner as WT NK cells (data not shown). Similarly, although NK cells have been shown to be able to use the TRAIL and Fas death receptors to induce cytotoxicity and immune surveillance (57), IL-21 functioned largely independently of these pathways (Fig. 9, B and C). After submitting this manuscript, similar tumor rejection properties were reported using a system in which melanoma cells were engineered to ectopically express IL-21 and elicit a local antitumor response (58). The observations of Ma et al. (58) support the systemic antimetastatic effects of IL-21 reported in this study. These new data, indicating the ability of IL-21 to fully differentiate the cytolytic program of NK cells, suggest that IL-21 may be used effectively in combination with other immunomodulators that stimulate NK cell proliferation and maturation.
We thank M.-C. Gouel for technical assistance; D. Foster (ZymoGenetics) for the initial supply of IL-21; J. Adams for the vav-Bcl2 mice; P. Hodgkin and J. Hasbold for assistance with CFSE labeling; K. Shortman, D. Raulet, A. Jamieson. and S. Anderson for the mAb; and M. O’Keeffe for help with the ELISA.
This work was supported by The Walter and Eliza Hall Institute of Medical Research Metcalf fellowship (to S.N.), a National Health and Medical Research Council Fellowship (to M.S.), a Cancer Research Institute Postdoctoral Fellowship (to Y.H.), and National Health and Medical Research Council and Human Frontiers in Science Program Grants (to M.S. and Y.H.).
Abbreviations used in this paper: KLRG1, killer cell, lectin-like-receptor G1; CD40L, CD40 ligand; FasL, Fas ligand; i.s., intrasplenically; PI, propidium iodide; WT, wild type.