NK Dendritic Cells Expanded in IL-15 Exhibit Antitumor Responses In Vivo¹

Natural killer cells are innate immune cells involved in tumor surveillance, immunity against viruses, and IFN-γ secretion (1). Dendritic cells (DC)³ are rare, bone marrow-derived leukocytes that specialize in uptake, transport, processing, and presentation of Ag to T cells (2). NKDC are a novel immunologic cell type with properties of both NK cells and DC (3). These unique cells can directly lyse tumor cells, present Ag to naive T cells, and secrete copious amounts of IFN-γ upon activation. Understanding the immunobiology of NKDC has recently become an area of intense investigation (4, 5, 6, 7, 8). Initial studies of NKDC properties were hindered by their scarcity, because they represent only 1% of murine splenocytes. Previously, we have shown that in vivo overexpression of fms-like tyrosine kinase 3 ligand (Flt3L) expands NKDC (4). Flt3L-expanded NKDC retain their potent cytolytic activity and become more effective stimulators of T cells. Because in vitro expansion of NKDC would be desirable for human immunotherapeutic strategies, we sought a method to generate NKDC from freshly isolated cells.

IL-15 is a 14- to 15-kDa cytokine with pleiotropic functions and is secreted by placenta, muscle, kidney, lung, heart, monocytes, macrophages, and dendritic cells (9). It was first identified on the basis of its ability to mimic IL-2-induced T cell proliferation (9, 10). Functional studies using Abs that selectively blocked various IL-2R subunits determined that IL-15 utilizes the IL-2Rβ and common γ-chain (γ_c), but not IL-2Rα (11). Shortly after the discovery of IL-15, numerous studies established the obligate role of IL-15 in the development, homeostasis, and survival of NK cells. IL-15^−/−, IL-15Rα^−/−, IL-2Rβ^−/−, and γ_c^−/− mice all contain profoundly reduced numbers of mature NK cells (12, 13, 14, 15). Additionally, mice with defects in signaling components that operate downstream of the IL-15R complex, such as Jak3, STAT5a/b, or STAT5b, also demonstrate NK cell defects (16, 17, 18). After adoptive transfer into IL-15^−/− or IL-15Rα^−/− recipients, NK cells have limited survival (19, 20). IL-15 has also been implicated in the augmentation of NK cell cytotoxicity and IFN-γ secretion (9, 11). Additionally, IL-15-transgenic mice have been found to be more resistant to tumors (21). The effects of IL-15 on the development, homeostasis, survival, and activation of NKDC are unknown.

Adult 6- to 8-wk-old male C57BL/6 (B6, H-2K^b), BALB/c (H-2K^d), and IL-15^−/− (B6, H-2K^b) mice were purchased from Taconic Farms. OT-I and OT-II OVA TCR-transgenic mice on the Rag2^−/− B6 background were also acquired from Taconic Farms. IFN-γ^−/− mice were obtained from The Jackson Laboratory. Animals were maintained in the pathogen-free animal housing facility at Memorial Sloan-Kettering Cancer Center, and all procedures were approved by the Institutional Animal Care and Use Committee.

Mice were administered daily i.p. injections of murine IL-15 (1 μg; BD Pharmingen) at various time points. Splenocytes were isolated as previously described (4). Briefly, animals were euthanized by CO₂ inhalation. Spleens were mechanically disrupted before being passed through a sterile 70-μm nylon mesh filter (BD Falcon). The resulting cell suspension was pelleted (300 × g for 7 min), RBC were lysed using a hypotonic solution, and the remaining cells were washed twice in complete medium (RPMI 1640, 10% FCS, 2 mM l-glutamine, 0.1% 2-ME, 100 U/ml penicillin, and 100 U/ml streptomycin). Splenocytes were separated into CD11c⁺CD11c⁻ or NK1.1⁺NK1.1⁻ fractions with immunomagnetic beads per the manufacturer’s protocol (Miltenyi Biotec). Before all immunomagnetic bead incubations, FcγIII/IIR were blocked with the mAb 2.4G2 (Fc block; 1 μg/million cells; Monoclonal Antibody Core Facility, Sloan-Kettering Institute, New York, NY). Enriched CD11c⁺ or NK1.1⁺ cells were stained with fluorescently conjugated Abs to B220, NK1.1, CD3, and CD11c (all BD Pharmingen) for further separation of DC subtypes using a MoFlo cell sorter (DakoCytomation). Dead cells were excluded with 4′,6-diamidino-2-phenylindole (Molecular Probes). Care was taken to exclude highly autofluorescent cells during FACS, and sorted cell populations were consistently >98% pure for the desired set of surface markers.

For in vitro expansion, FACS-purified NK cells, NKDC, and DC isolated from B6 or IFN-γ^−/− mice were cultured in complete medium with IL-15 (20 ng/ml) for various times. The medium was changed on days 3 and 7 of culture.

Five-color flow cytometry was performed on a FACScan flow cytometer (BD Biosciences) with modifications by Cytek. Voltages were determined using unstained cells. Single-stained positive controls for each fluorochrome were used to set compensation. Samples were incubated with Fc block before staining. Approximately 5 × 10⁵ cells were labeled with 0.1 μg of FITC, PE, peridinin chlorophyll-a protein (PerCP), allophycocyanin, allophycocyanin-Cy7, or biotin-conjugated Ab (all BD Pharmingen). Biotinylated Abs were secondarily stained with streptavidin-PerCP. Cells were stained for CD3ε (145-2C11), CD11c (HL-3), CD40 (3/23), CD45R/B220 (RA3-6B2), CD69 (H1.2F3), NK1.1 (PK136), Ly49C/I (5E6), Ly49D (4E5), and Ly49F (HBF-719). Appropriate Ig isotype controls were used for phenotype analysis. Flow cytometry data were analyzed with FlowJo software (Tree Star).

In vitro cytokine production was assessed by culturing 3 × 10⁴ purified NK cells and NKDC in a 96-well U-bottom tissue culture plate (BD Falcon) in 100 μl of medium for 72 h. Supernatant IFN-γ content was assayed using a cytometric bead array per the manufacturer’s protocol (BD Biosciences). IL-12 (20 ng/ml; BD Pharmingen) and IL-18 (20 ng/ml; BioSource International) were added to some wells.

Lysis assays were performed as before (3) using FACS-purified B6 splenic NK cells and NKDC expanded in IL-15 for 7 days. Briefly, cells were cocultured with 1 × 10³ [⁵¹Cr]sodium chromate (PerkinElmer, Life and Analytical Sciences) labeled Yac-1 cells or CHO cells (American Type Culture Collection) for 6 h in 96-well V-bottom plates (BD Falcon) in a total of 200 μl of medium. Concanamycin A (100 ng/ml; ICN Biomedical) was added to some wells containing effectors for 2 h of incubation before the addition of Yac-1 targets. [⁵¹Cr]sodium chromate release was measured with a TopCount NXT microplate scintillation and luminescence counter (PerkinElmer). Spontaneous release (no effectors) and maximum release (2% Triton X-100; Sigma-Aldrich) were also assayed. Percent specific lysis was calculated as (cpm experimental − cpm spontaneous release) × 100/(cpm maximum release − cpm spontaneous release). MLRs were performed by combining stimulator cells with allogeneic T cells. T cells were purified using Thy1.2 (CD90.2) immunomagnetic beads (Miltenyi Biotec) according to the manufacturer’s protocol. FACS-purified NK cells, NKDC, and DC from B6 mice were added in various numbers to 1 × 10⁵ BALB/c T lymphocytes in 96-well U-bottom plates (BD Falcon) in a total of 200 μl of medium. Ag-specific T cell activation was assayed in a similar fashion with OT-I CD8⁺ transgenic T cells specific for SIINFEKL peptide (OVA_257–264), or OT-II CD4⁺ transgenic T cells specific for KISQAVHAAHAEINEAG peptide (OVA_323–339). Stimulators were loaded with the appropriate OVA peptide (1 μg/ml; Peptide Synthesis Core, Sloan-Kettering Institute) and plated at various concentrations with OT-I or OT-II T cells (3 × 10⁴/well) in a 96-well U-bottom plate. On day 3, the cultures were pulsed with [³H]thymidine (1 μCi/well; PerkinElmer) and radioactive uptake was measured 20 h later with a TopCount NXT microplate scintillation and luminescence counter.

The B16F10 melanoma cell line (American Type Culture Collection) was cultured in complete medium and tested negative for known mouse pathogens. Cultured tumor cells were washed twice in PBS, and 1 × 10⁵ tumor cells in 300 μl of PBS were injected via the lateral tail vein into B6 or IFN-γ^−/− mice. Spleen NK cells and NKDC from B6 donors were FACS purified, expanded in IL-15 for 7 days, and cultured with either IL-12 (20 ng/ml) and IL-18 (20 ng/ml) or medium for 2 h. Eight hours after administration of tumor cells, 1 × 10⁶ of the IL-15 expanded NK cells or NKDC were adoptively transferred into the tumor recipients by i.v. injection in 300 μl of PBS. Animals were sacrificed on day 15 and the number of lung metastases was quantified with the aid of a dissecting microscope. Statistical analysis was performed using one-way ANOVA test and Student’s t test. All p < 0.05 were deemed statistically significant.

IL-15 is known to be an important regulator of NK cell survival and proliferation. Mice lacking IL-15 have a severe reduction or absence of cells with a NK phenotype (12). Because NKDC (CD3⁻NK1.1⁺CD11c⁺) share NK cell (CD3⁻NK1.1⁺CD11c⁻) features, we hypothesized that IL-15 may play an essential role in the proliferation and survival of NKDC as well. Although NKDC constitute ∼10–15% of all splenic CD11c⁺ cells in B6 mice, we found that IL-15^−/− mice had a marked decrease in the percentage and total number of NKDC (Fig. 1). Daily administration of exogenous murine IL-15 to IL-15^−/− mice, however, restored normal percentages and absolute numbers of NKDC and NK cells to that of wild-type B6 mice. Interestingly, NKDC were reconstituted faster than NK cells, as seen on day 4 (Fig. 1 B), whereas NK cells required up to 9 days of IL-15 administration before reaching values present in wild-type B6 mice. Discontinuation of IL-15 for as little as 2 days resulted in the loss of all NK1.1⁺ cells, confirming that IL-15 is essential for the development and homeostasis of NK cells as well as NKDC (data not shown).

Once we established the obligate role of IL-15 in the development and maintenance of NKDC, we tested whether administration of exogenous IL-15 would expand NKDC in wild-type B6 mice. After 7 daily i.p. injections of murine IL-15 to B6 mice, we observed a 2-fold increase in the percentage of CD11c⁺ splenocytes and the counts of splenic NK cells and NKDC increased 2- and 3-fold over baseline, respectively (Figs. 1,A and 2, A and B). Because expansion of cells in an in vitro setting is more pertinent for translation to human studies, we tested the effects of IL-15 treatment on freshly sorted splenic NK cells and NKDC from B6 mice. NKDC cultured in IL-15 for 7 days expanded 10-fold in number, while NK cells increased 5-fold (Fig. 2 C). The ability of both NK cells and NKDC to divide in response to IL-15 was finite, since longer culture periods did not further increase cell yields. Notably, DC (CD3⁻NK1.1⁻CD11c⁺) did not expand and were not able to survive past 3 days of culture in IL-15. Because IL-2 and IL-15 share common signaling components (IL-2/15Rβγ_c), we investigated the effect of IL-2 culture on NKDC. In contrast to IL-15, IL-2 culture resulted in <20% cell viability by day 3 and near complete cell loss on day 7 for NKDC as well as NK cells (data not shown).

Since 7 days of IL-15 culture led to the optimal expansion of both NK cells and NKDC, all remaining experiments were performed using cells prepared in this way. NKDC expanded in IL-15 were both larger in size and more granular than fresh splenic NKDC and fresh or expanded NK cells, as indicated by microscopy and forward and side scatter analyses (Fig. 2,D). Additionally, compared with IL-15-expanded NK cells, IL-15-expanded NKDC had a phenotype associated with a greater activation state, as indicated by a lower expression of the inhibitory NK cell marker Ly49C/I and increased expression of the activation markers Ly49D and CD69 (Fig. 3). Furthermore, expanded NKDC down-regulated the inhibitory marker Ly49F, while IL-15-expanded NK cell expression of Ly49F remained unchanged.

We next tested whether NKDC expanded in IL-15 had altered function. Using a chromium release assay, we found that NKDC expanded in IL-15 had similar cytolytic capacity as fresh NKDC (Fig. 4,A). In contrast, NK cells expanded in IL-15 gained lytic capacity and became equipotent to NKDC in this regard (Fig. 4,B). Addition of IL-2 to the IL-15 cultures on days 3 and 6 or culturing cells in the presence of anti-NKG2D did not alter the lytic capacity of NKDC or NK cells (data not shown). When cultured with concanamycin A, however, percent specific lysis of NKDC decreased substantially (Fig. 4 C).

IFN-γ is a pleiotropic cytokine that is mainly a product of NK⁺ cells and CD8 T cells. We have previously shown that freshly isolated splenic NKDC produce higher levels of IFN-γ than NK cells under a variety of conditions (3, 5). We next tested whether NKDC expanded in IL-15 had augmented IFN-γ production. After culture in IL-12 plus IL-18, NKDC expanded in IL-15 secreted 985 ± 20 ng/ml IFN-γ, nearly twice as much as IL-15-expanded NK cells and 10-fold higher than fresh NKDC (Fig. 4 D).

Although NKDC expanded in IL-15 had potent lytic ability and produced large amounts of IFN-γ, they had undetectable APC function. NKDC expanded in IL-15 lost their ability to induce allogeneic T cell proliferation in an MLR (Fig. 4,E) and they no longer stimulated division of TCR-transgenic OT-I (Fig. 4,F) and OT-II (Fig. 4 G) T cells after culture with their respective peptides. On average, fresh NKDC possess approximately one-third of the capacity of DC for stimulating naive T cells (3, 4). As expected, NK cells expanded in IL-15 did not induce any T cell proliferation. Flow cytometry analysis of MHC I, MHC II, and costimulatory molecules (CD40, CD80, and CD86) did not show any differences between fresh and IL-15-expanded NKDC, except that IL-15-expanded NKDC had a slightly higher level of CD86 expression (data not shown). In contrast, NK cells failed to gain CD86 expression during IL-15 culture. Addition of various classical DC maturation agents including TNF-α, IFN-γ, IL-12, IL-18, CpG, and LPS to the IL-15 cultures did not result in an up-regulation of maturation markers of NKDC (data not shown).

Because NKDC expanded in IL-15 secreted such high levels of IFN-γ in vitro, we investigated their antitumor potential in a melanoma lung metastasis model, which depends on IFN-γ. NKDC expanded in IL-15 were treated with IL-12 plus IL-18 in vitro for 2 h and injected via tail vein into wild-type B6 mice, which had been administered B16F10 melanoma cells 8 h earlier. Mice treated with NKDC expanded in IL-15 developed ∼50% fewer lung metastases than mice treated with IL-15-expanded NK cells (Fig. 5,A), although the findings were not statistically significant. The results were even more striking when IFN-γ^−/− recipients were used and NKDC provided a statistically significant benefit. Adoptive transfer of NKDC expanded in IL-15 resulted in 67% fewer lung metastases compared with PBS controls and 60% fewer tumors than mice injected with IL-15-expanded NK cells. These results correlated when lungs harvested from the recipient mice were weighed (Fig. 5,B). To determine whether the antitumor properties of expanded NKDC were strictly due to their IFN-γ secretion, we used IFN-γ^−/− NKDC expanded in IL-15, which demonstrated similar in vitro expansion kinetics as wild-type NKDC. IFN-γ^−/− NKDC expanded in IL-15 for 7 days were stimulated with IL-12 plus IL-18 for 2 h and subsequently adoptively transferred into IFN-γ^−/− recipients. Mice treated with IFN-γ^−/− NKDC expanded in IL-15 had a similar number of lung metastases as controls treated with PBS or IFN-γ^−/− NK cells expanded in IL-15 (Fig. 5 C). Thus, in this tumor prevention model, the antitumor effects were predominantly due to IFN-γ secretion. Additionally, treatment with β₂m^−/− NKDC expanded in IL-15 did not significantly alter the number of metastases, when compared with mice that had received wild-type NKDC (data not shown).

NKDC are now recognized as a novel immunologic cell type with pleiotropic functions (3, 4, 5, 6, 7, 8). Originally, Homann et al. (22) reported murine DX5⁺CD11c⁺ cells that had lytic and APC function and mediated tolerance in a transgenic model of diabetes. Subsequently, we detected NK1.1⁺CD11c⁺ cells in the liver, as well as the spleen, lymph nodes, and thymus of wild-type B6 mice. We labeled the cells as NKDC (3). Recently, cells with similar phenotype (DX5⁺CD11c⁺B220⁺) and function (cytolysis, APC function, and IFN-γ production) to that of NKDC were reported simultaneously by two groups and termed IFN-γ- producing killer DC (IKDC) (23, 24). In addition, IKDC were reported to secrete type I IFN, but more recent data suggest this may have been due to contamination by plasmacytoid DC (pDC) (25), which also express DX5 and B220, but not NK1.1. Notably, we have not found NKDC to produce IFN-α (our unpublished data). IKDC may represent a mature subtype of NKDC that expresses B220. However, the importance of B220 expression is unclear because upon activation, B220⁻ NKDC gain higher MHC II expression and APC function, distinguishing them from NK cells, but not IKDC (our unpublished data).

Our initial investigations of NKDC have been constrained by the difficulty in isolating them due to their scarcity (3). Because expansion of cells in an ex vivo environment would be desirable for translation to human studies, we sought to determine the ideal conditions for in vitro expansion of NKDC. Because of its essential role in the development, homeostasis, and survival of NK cells (9, 11, 13, 14, 15, 16, 17, 18), we tested the effects of IL-15 on NKDC. We found that IL-15 was also critical for the ontogeny of NKDC because they were markedly reduced in IL-15^−/− mice but could be restored by exogenous IL-15 (Fig. 1,A). The more rapid reconstitution of NKDC compared with NK cells in IL-15^−/− mice treated with IL-15 (Fig. 1,B) could not be explained by differences in expression of the IL-15R complex, because cell surface expression of IL-15Rα, IL-2Rβ, and γ_c was similar between NKDC and NK cells (data not shown). In the absence of immunologic stimuli, mature NK cells have been reported to persist in the periphery with a half-life of ∼8 days (20). However, we observed that as little as a 2-day interruption in the administration of IL-15 resulted in the disappearance of all NK1.1⁺ cells. Consistent with our findings in IL-15^−/− mice, IL-15 also expanded NKDC in wild-type B6 mice (Fig. 2, A and B). In vitro, IL-15 preferentially expanded freshly isolated splenic NKDC compared with NK cells (Fig. 2,C). IL-15-expanded NKDC were larger in size and more granular than freshly isolated NKDC (Fig. 2,D) and exhibited a more activated phenotype (Fig. 3). IL-15 expansion of NKDC did not alter their lytic capacity (Fig. 4,A). The lysis mechanism was not NKG2D mediated, because addition of an anti-NKG2D Ab to the cultures did not cause a decrease in lytic capacity (data not shown). Addition of concanamycin A to the cultures, however, led to a substantial decrease in percent lysis by both NKDC and NK cells (Fig. 4,C), showing that the majority of IL-15-expanded NK cell- and NKDC-mediated lysis is perforin dependent. NKDC cytotoxicity was not limited to Yac-1 cells, because they also lysed CHO cells. IL-15 expansion, however, did increase the ability of NKDC to produce IFN-γ by 10-fold (Fig. 4 D). Meanwhile, IL-15-expanded NK cells produced 2-fold less IFN-γ than IL-15-expanded NKDC.

Although IL-15-expanded NKDC retained their potent cytolytic capacity and secreted a larger quantity of IFN-γ than freshly isolated splenic NKDC, they surprisingly lost their ability to present Ag to allogeneic T cells (Fig. 4,E) and Ag-specific TCR-transgenic T cells (Fig. 4, F and G). Analysis of maturation markers on NKDC expanded with IL-15 in vivo or in vitro showed similar cell surface expression of CD40, CD80, CD86, MHC I, and MHC II as freshly isolated splenic NKDC. These results were in stark contrast to our previous work using in vivo overexpression of Flt3L by an adenovirus vector to expand NKDC, which preserved splenic NKDC lytic function but augmented their ability to stimulate naive T cells, both in vitro and in vivo (4). Our findings are consistent with the work of Chan et al. (23), who demonstrated that IKDC can differentially exhibit NK cell or APC properties. In the periphery, IKDC preferentially have NK cell function; however, upon systemic infection, IKDC traffic to lymph nodes, where they lose NK cell properties and primarily act as APC. Taken together, it is clear that the method of NKDC expansion may alter their functional properties.

Although NKDC may play a variety of roles under physiologic and pathologic conditions, they have already been found to be important in the innate immune response to infectious organisms and tumors. Homann et al. (22) demonstrated that DX5⁺CD11c⁺ cells play a role in viral infection. Recently, we (6) have shown that NKDC are the initial source of IFN-γ in mice during systemic infection by Listeria monocytogenes. Taieb et al. (24) have demonstrated that IKDC possess antitumor properties after activation with imatinib mesylate and IL-2. Similarly, using adoptive transfer studies, we have shown that freshly isolated NKDC activated with IL-18 and CpG were more potent than NK cells in preventing tumor metastases (5). In the current study, we show that NKDC expanded in IL-15 and activated with IL-12 plus IL-18 were more effective at preventing melanoma lung metastases than IL-15-expanded NK cells (Fig. 5,A). The protective effect was not due to a superior survival advantage of IL-15-expanded NKDC in vivo, since similar organ distribution and survival of CD45.2⁺ NK cells and NKDC were observed via flow cytometry analysis on day 3 after adoptive transfer into CD45.1⁺ mice (data not shown). The antitumor mechanism relied on IFN-γ production by NKDC as proven with IL-15-expanded IFN-γ^−/− NKDC and IL-15-expanded β₂m^−/− NKDC (Fig. 5 C and data not shown).

The developmental origin of NKDC remains to be precisely defined, although substantial evidence already exists that NK cells and DC share a common linage. Murine fetal thymic progenitor cells possess the capacity to differentiate into T cells, NK cells, or DC (26). Additionally, a bipotential NK cell and DC precursor has also been identified in human thymus (27) and human CD34⁺Lin⁻CD38⁺ progenitors can differentiate into NK cells, B-lineage cells, myeloid cells, and DC (28), presenting evidence that NKDC may arise from such a common NK cell/DC precursor. A recent study by Welner et al. (29) reported that IKDC arise from Lin⁻Sca-1⁺c-kit^highThy1.1⁻L-selection⁺ lymphoid progenitor cells and diverge early from the precursors also responsible for giving rise to NK cells, pDC, B, and T cells.

The human counterpart to murine NKDC has yet to be identified. Nevertheless, Hanna et al. (30) found that human NK cells have the potential to gain APC function upon activation during extended culture in IL-2. Recently, human pDC were shown to acquire TRAIL-mediated lytic activity when activated with influenza virus or CpG (31). Although, human NKDC remain to be identified, our results suggest that like human NK cells, NKDC may be expanded in vitro by IL-15. However, the method of expansion may considerably alter their functional properties.

In summary, our findings demonstrate the obligate role of IL-15 in the development and survival of NKDC. IL-15-expanded NKDC retain their potent lytic capacity, secrete an abundant amount of IFN-γ, and inhibit melanoma lung metastases in a tumor prevention model. While acquiring more effective NK cell properties, NKDC expanded in IL-15 have negligible APC function, suggesting that the method used for NKDC expansion may skew their pleiotropic properties. Nonetheless, IL-15-expanded NKDC are attractive candidates for cellular immunotherapy.

We thank the Swim Across America Foundation for their generous support.

The authors have no financial conflict of interest.