Configuration-dependent Presentation of Multivalent IL-15:IL-15Rα Enhances the Antigen-specific T Cell Response and Anti-tumor Immunity*

Here we report a “configuration-dependent” mechanism of action for IL-15:IL-15Rα (heterodimeric IL-15 or hetIL-15) where the manner by which IL-15:IL-15Rα molecules are presented to target cells significantly affects its function as a vaccine adjuvant. Although the cellular mechanism of IL-15 trans-presentation via IL-15Rα and its importance for IL-15 function have been described, the full effect of the IL-15:IL-15Rα configuration on responding cells is not yet known. We found that trans-presenting IL-15:IL-15Rα in a multivalent fashion on the surface of antigen-encapsulating nanoparticles enhanced the ability of nanoparticle-treated dendritic cells (DCs) to stimulate antigen-specific CD8+ T cell responses. Localization of multivalent IL-15:IL-15Rα and encapsulated antigen to the same DC led to maximal T cell responses. Strikingly, DCs incubated with IL-15:IL-15Rα-coated nanoparticles displayed higher levels of functional IL-15 on the cell surface, implicating a mechanism for nanoparticle-mediated transfer of IL-15 to the DC surface. Using artificial antigen-presenting cells to highlight the effect of IL-15 configuration on DCs, we showed that artificial antigen-presenting cells presenting IL-15:IL-15Rα increased the sensitivity and magnitude of the T cell response, whereas IL-2 enhanced the T cell response only when delivered in a paracrine fashion. Therefore, the mode of cytokine presentation (configuration) is important for optimal immune responses. We tested the effect of configuration dependence in an aggressive model of murine melanoma and demonstrated significantly delayed tumor progression induced by IL-15:IL-15Rα-coated nanoparticles in comparison with monovalent IL-15:IL-15Rα. The novel mechanism of IL-15 transfer to the surface of antigen-processing DCs may explain the enhanced potency of IL-15:IL-15Rα-coated nanoparticles for antigen delivery.

Here we report a "configuration-dependent" mechanism of action for IL-15:IL-15R␣ (heterodimeric IL-15 or hetIL-15) where the manner by which IL-15:IL-15R␣ molecules are presented to target cells significantly affects its function as a vaccine adjuvant. Although the cellular mechanism of IL-15 trans-presentation via IL-15R␣ and its importance for IL-15 function have been described, the full effect of the IL-15:IL-15R␣ configuration on responding cells is not yet known. We found that trans-presenting IL-15:IL-15R␣ in a multivalent fashion on the surface of antigen-encapsulating nanoparticles enhanced the ability of nanoparticle-treated dendritic cells (DCs) to stimulate antigen-specific CD8 ؉ T cell responses. Localization of multivalent IL-15:IL-15R␣ and encapsulated antigen to the same DC led to maximal T cell responses. Strikingly, DCs incubated with IL-15:IL-15R␣coated nanoparticles displayed higher levels of functional IL-15 on the cell surface, implicating a mechanism for nanoparticle-mediated transfer of IL-15 to the DC surface. Using artificial antigen-presenting cells to highlight the effect of IL-15 configuration on DCs, we showed that artificial antigen-presenting cells presenting IL-15:IL-15R␣ increased the sensitivity and magnitude of the T cell response, whereas IL-2 enhanced the T cell response only when delivered in a paracrine fashion. Therefore, the mode of cytokine presentation (configuration) is important for optimal immune responses. We tested the effect of configuration dependence in an aggressive model of murine melanoma and demonstrated significantly delayed tumor progression induced by IL-15:IL-15R␣-coated nanoparticles in comparison with monovalent IL-15:IL-15R␣. The novel mechanism of IL-15 transfer to the surface of antigen-processing DCs may explain the enhanced potency of IL-15:IL-15R␣-coated nanoparticles for antigen delivery.
It is well established that dendritic cells (DCs) 2 play a key role as professional antigen-presenting cells in initiating and coordinating antigen-specific adaptive immunity (1,2). By constitutively sampling the external microenvironment through receptor-mediated and non-receptor mediated endocytic pathways, DCs initiate, program, and regulate the effector response against a variety of antigens (3)(4)(5)(6)(7). Although much is known about the immunobiology of DC antigen processing and presentation and the effect of different cytokines on DC maturation and their subsequent effect on the immune response (8), little is known about how the configuration or geometry of cytokine transmission affects the DC response. For example, physiologically, although IL-2 and IL-15 belong to the same four a-helix family of cytokines, IL-2 is transmitted as a soluble paracrine or autocrine factor (9,10), whereas IL-15 is trans-presented on the surface of DCs and other cell types through a stable complex (11) with IL-15R␣ (12,13) or secreted as a soluble paracrine complex (14,15). The quality and manner by which cytokines are delivered to DCs can have a profound effect on the subsequent effector immune response (16,17).
IL-15 is a homeostatic cytokine for CD8 ϩ T cells and natural killer cells and an important regulator of the cytotoxic immune response (40). IL-15 signals by binding to cells expressing the IL-2/IL-15 ␤/␥ receptor and is found complexed to IL-15R␣ (12)(13)(14)(15). This IL-15:IL-15R␣ complex is termed heterodimeric . IL-15R␣ is expressed by several cell types in * C. B. and G. N. P. are inventors on United States Government-owned pat-the body, most notably by activated monocytes and DCs. Coexpression of IL-15 and IL-15R␣ in the same cell and intracellular complex formation is essential for the production and function of IL-15 (11,41,43). IL-15R␣ has a very strong affinity to IL-15 (K D ϳ10 Ϫ11 M). Indeed, the main bioactive form of IL-15 is found exclusively in a complex with IL-15R␣ in humans and mice (44) and is the main form of circulating IL-15 in the blood of lymphodepleted melanoma patients (45). This mechanism of IL-15 signaling has been shown to mediate the bulk of its immunological functions, including the differentiation, activation, and survival of CD8 ϩ memory T cells and natural killer cells (46 -49). IL-15 has already gained significant clinical interest as a highly promising immunotherapeutic agent (40, 50 -52), and the anti-tumor efficacy of the IL-15:IL-15R␣ complex has been demonstrated in a variety of preclinical tumor models (53)(54)(55)(56).
IL-15 has also been investigated for its ability to augment antigen-specific responses stimulated by vaccines. Multiple studies using DNA vaccination models have reported that the inclusion of an IL-15 expression plasmid enhanced functional and protective immunity (57)(58)(59). The administration of the IL-15 protein as an adjuvant enhanced the antigen-specific T cell response to peptide-pulsed DCs (60) and, in a separate study, was able to increase antibody titers against a recombinant bacterial superantigen by a DC-specific mechanism (61). Some of these effects may be due to the action of IL-15 on DCs. Indeed, IL-15-transduced DCs are more activated than normal DCs and show improved survival and function (62). When given during DC differentiation, IL-15 potentiated stronger DC-mediated CD8 stimulation (63)(64)(65) and mediated reversal of the suppressive effects of the tumor microenvironment by restoring the antigen-processing machinery inhibited by tumor gangliosides (66). Therefore, IL-15 may interact with DCs and other antigen-presenting cells (APCs) to enhance their stimulation of other immune cells.
Because of the substantial ability of IL-15 to promote the activation and survival of CD8 effector and memory T cells and as a vaccine adjuvant in general, we reasoned that the inclusion of IL-15 on the surface of NPs would potentiate the response of APCs to encapsulated antigen. In previous work, our group and others have shown that the quality of antigen-specific memory CD8 ϩ T cells after immunization is increased with the sustained delivery and increased availability of antigen afforded by PLGA-mediated prolonged antigen release (33,34,67,68). The humoral and cellular immune responses can be potentiated further by the introduction of rationally selected ligands to the PLGA formulation. The addition of pathogen-associated molecular patterns to antigen delivery systems, for example, generated protection against West Nile strains (33,34), lethal avian and flu strains in mice, and H1N1 in rhesus macaques (67). The use of IL-15 as an adjuvant on nanoparticulate PLGA delivery platforms for sustained release of antigen has not yet been explored.
Here we test this potential adjuvant by binding stable complexes of IL-15:IL-15R␣ to the surface of PLGA nanoparticles encapsulating the model antigen ovalbumin (OVA). IL-15:IL-15R␣ presented on nanoparticles acts as an adjuvant for the DC-mediated activation of OVA-specific immune responses, and this adjuvant activity (in terms of the cellular immune response) requires the delivery of antigen and heterodimeric IL-15 to the same DCs. The adjuvant activity is mediated through transfer of IL-15 from the nanoparticles to the surface of recipient DCs through a mechanism that involved internalization of the NP and export of IL-15 to the DC membrane. Our findings reveal new strategies for the enhancement of IL-15 function in vivo as an immunotherapeutic adjuvant.

Experimental Procedures
Nanoparticle Formulation-Endotoxin-free OVA (Invivo-Gen) was encapsulated in avidin-coated PLGA nanoparticles (NPs) using a water/oil/water double emulsion technique. 2 mg of ovalbumin protein was dissolved in 200 l of sterile PBS and added dropwise to 50 mg of PLGA (50:50 monomer ratio, Durect Corp.) dissolved in 3 ml of chloroform while vortexing. The emulsion was then added dropwise to a mixture of 3 ml of 5% poly(vinyl alcohol) and 1 ml of an avidin-palmitic acid conjugate at 5 mg/ml in 2% sodium deoxycholate while vortexing. The avidin-palmitate partitions to the chloroform-water interface, immobilizing avidin at the NP surface upon solvent evaporation (69). The double emulsion was sonicated three times for 10 s each on ice and then transferred into a 200-ml volume of 0.2% PVA in distilled H 2 O to evaporate the solvent by magnetic stirring at room temperature for 90 min. Particles were then washed three times with distilled H 2 O and lyophilized for long-term storage at Ϫ20°C. The encapsulation of IL-2 or class I peptides was performed in a similar manner. 200 g of SIINFEKL or 1 mg of MART-1 15-40 peptide (KGHGHSYT-TAEEAAGIGILTVILGVL) were dissolved in 200 l of sterile PBS and added dropwise, while vortexing, to the PLGA/chloroform solution before emulsification in PVA as described above. For interleukin 2 (Chiron Corp.), 200 l of a 1 mg/ml solution in PBS was used. The remaining nanoparticle formulation steps were followed after loading of the encapsulant. Nanoparticle size was analyzed using scanning electron microscopy as well as diffusion measurements in aqueous solution using a NanoSight NS500 instrument (NanoSight).
Cytokine Preparation-Optimized dual promoter plasmids expressing the two chains of the human heterodimeric cytokine IL-15:IL-15R␣Fc have been described previously. IL-15:IL-15R␣Fc is a fusion protein comprising the human soluble heterodimeric IL-15, in which the sIL-15R␣ chain is fused to the Fc region of human IgG1 (70). Human IL-15:IL-15R␣Fc was synthesized and secreted by a stable, clonal HEK293 cell line as reported previously (15) and purified by protein A affinity chromatography. The dose of heterodimeric IL-15 in these experiments was calculated and expressed as the equivalent amount of single-chain IL-15 found within the heterodimer, as determined by amino acid analysis. The heterodimer was reacted at a 1:10 molar ratio with NHS-LC-LC-biotin (Thermo Fisher Scientific) for 1 h at 37°C and then dialyzed for 48 h in PBS in a 3000 molecular weight cutoff dialysis bag (Thermo Fisher Scientific) to remove excess unreacted biotin.
Transmission Electron Microscopy (TEM)-IL-15-bearing NPs were treated with anti-human IL-15 (R&D Systems) and then treated with anti-mouse IgG labeled with 3-nm gold nanoparticles. The resulting NPs were mounted on copper grids, washed, and imaged on a FEI Tecnai Biotwin TEM instrument. NP-treated cells were fixed overnight in paraformaldehyde and then sectioned under liquid nitrogen before mounting and imaging by TEM.
Preparation of Murine Antigen-specific T Cells and Bone Marrow-derived Dendritic Cells-All mice used in these studies were purchased from Taconic Biosciences. Primary splenocytes were obtained from homogenized mouse spleens. CD8 ϩ T cells were purified from the splenic homogenate of 8-16-week-old female Rag2/OT-I transgenic mice by negative immunoselection according to the instructions of the manufacturer (StemCell Technologies). T cells were cultured in complete RPMI-10 medium consisting of RPMI 1640 medium, 10% FBS, 1% L-glutamine, 50 M ␤-mercaptoethanol, 1% penicillin, 1% streptomycin, 0.05% minimum essential medium vitamin solution, 1% sodium pyruvate, and 1% non-essential amino acids (all cell culture supplements were purchased from Invitrogen). When necessary, the splenic homogenate was labeled with CFSE before purification.
DCs were generated from murine bone marrow according to established protocols (72). Bone marrow was flushed out from the femora and tibiae of 8-16-week-old C57Bl/6 mice, and erythrocytes were depleted by hypotonic lysis (Lonza). The remaining bone marrow cells were cultured in RPMI 1640 medium supplemented with 10% FBS, 1% HEPES buffer, 1% penicillin-streptomycin-glutamine, 100 M ␤-mercaptoethanol, and 20 ng/ml murine GM-CSF. 5 days after the beginning of culture, non-adherent and semiadherent cells were collected and cultured in fresh medium for 2 more days before being used in experiments.
Preparation of Human Antigen-specific T Cells and HLA-A2 ϩ DCs-The human CD8 ϩ T cell receptor transgenic cell line DMF5, which is reactive to the MART-1 27-35 epitope, was provided by John R. Wunderlich (Surgery Branch, NCI/National Institutes of Health). Protocols for generating mature monocyte-derived DCs over a 48-h period (rather than a 7-to 10-day period) have been described previously (73)(74)(75). Peripheral blood was isolated from healthy human donors (HLA-A*0201 ϩ and HLA-A*0201 Ϫ ) after informed consent was obtained, and mononuclear cells were purified by Ficoll-Hypaque gradient centrifugation. Purified cells were suspended in serum-free DC medium (Cell-Genix) with 1% human serum and plated at 15 ϫ 10 6 cells/well in a 6-well tissue culture plate. Non-adherent cells and medium were removed and replaced with fresh DC medium 90 and 150 min after the start of incubation. The media was then replaced with DC medium supplemented with 800 IU/ml GM-CSF and 1000 IU/ml IL-4 for 24 h at 37°C to generate immature DCs capable of internalizing and cross-presenting antigen.
For stimulation with DMF-5 T cells, human DCs were first incubated with nanoparticles or other treatments overnight. A maturation mixture of 10 ng/ml TNF␣, 10 ng/ml IL-1␤, 1000 IU/ml IL-6, and 1 g/ml prostaglandin E2 was added to each sample for the final 4 h of culture. DCs were then harvested, washed, and co-cultured with DMF-5 T cells at a ratio of 1:2 DCs:T cells for 72 h at 37°C. The efficiency of antigen-specific T cell activation was assessed using a human IFN-␥ ELISA kit according to the instructions of the manufacturer.
T Cell and DC Co-culture Studies-DCs were seeded in 96-well U-bottomed plates at 25,000 cells/well. Nanoparticle constructs were added to DC cultures at 0.1 mg/ml and allowed to incubate for 8 h. When necessary, avidin-coated nanoparticles were allowed to bind with biotinylated IL-15:IL-15R␣ (prepared as described previously) by incubating at room temperature for 15 min. DCs were then washed twice by centrifugation at 1500 rpm for 5 min and aspirated carefully, and then the culture medium was replaced. T cells were added at 50,000 cells/well and allowed to co-incubate for 72 h. At the end of the incubation, all cells were centrifuged, and supernatants were recovered for analysis by ELISA.
Flow Cytometric Analysis-CFSE-labeled CD8 ϩ T cells were fixed in 2% paraformaldehyde and analyzed directly by flow cytometry. Before analysis, 5 ϫ 10 4 6-m polystyrene beads were added to each sample as an internal standard to determine absolute cell counts. For IL-15 retention studies, nanoparticletreated bone marrow-derived dendritic cells were stained with phycoerythrin-or allophycocyanin-labeled anti-human IL-15 (R&D Systems) diluted 1:20 in 1% FBS in PBS. Staining was performed for 30 min at 4°C, and then cells were washed once in 1ϫ PBS and fixed in 2% paraformaldehyde. For detection of phosphorylated signaling molecules, cells were stimulated with IL-15 for 1 h and then fixed with 4% paraformaldehyde for 20 min. Cells were washed and then permeabilized overnight in 100% methanol at Ϫ20°C. Finally, cells were washed and stained with Alexa Fluor 488-conjugated anti-phospho-STAT5 for 30 min at room temperature in the dark and then resuspended in 2% paraformaldehyde. All samples were run on an Accuri C6 flow cytometer (BD Biosciences), and data were analyzed using FlowJo (TreeStar) or Accuri C6 software.
Exosome Preparation-Exosomes were purified from DC culture supernatants by a sequential centrifugation protocol. DCs were incubated with NP treatments in complete medium for 24 h. DCs were then washed twice in serum-free media and replated in serum-free exosome production media containing 1% BSA for a further 24 h. Exosomes were purified from the resulting conditioned media by sequential centrifugation at 400 rcf, 2000 rcf, 10,000 rcf, and 100,000 rcf. Exosome preparations were adsorbed to 5 m of aldehyde-sulfate beads (Life Technologies) that were then stained with antibodies and analyzed by flow cytometry.
Transwell Experiments-DCs were incubated overnight with NP constructs and then washed three times in media. 2.5 ϫ 10 5 DCs/well were plated in a 24-well plate. CD8 ϩ T cells were added at a 1:1 ratio either in cell-cell contact or separated by a 400-m pore size transwell (Costar). Cells were stimulated for 3 days and then assayed for proliferation and cytokine secretion.
Artificial APC Studies-Artificial APCs (aAPCs) were constructed by incubating avidin-coated NPs with biotin-H2-Kb MHC dimers loaded with SIINFEKL (SIINFEKL:MHC) and equimolar anti-mouse CD28 antibody (BD Biosciences). When present, biotin-IL-15:IL-15R␣ was added at a concentration of 1.5 g of IL-15/mg of NP. To compare between NP-loaded DCs and aAPCs, 0.1 mg/ml of each NP construct was used to stimulate OT-I CD8 ϩ T cells. DCs were incubated with NPs overnight and washed before adding T cells. aAPCs were added directly to T cells. T cells were stimulated for 3 days and then analyzed for cell counts and cytokine secretion.
In Vivo Subcutaneous Tumor Studies-OVA-expressing murine B16 melanoma cells (B16-OVA) were a gift from the laboratory of Dr. Lieping Chen. B16-OVA cells were cultured in DMEM (Gibco) with 10% FBS and suspended in 1ϫ PBS on ice prior to injection. For metastatic studies, 8-week-old C57Bl/6 mice were challenged intravenously with 1 ϫ 10 5 cells injected in 50 l of PBS. Treatments began the day after B16-OVA challenge, and mice were euthanized by CO 2 asphyxiation when symptoms of pulmonary distress were observed. Each treatment consisted of 1 g of IL-15:IL-15R␣ and 0.4 mg of PEGmodified PLGA nanoparticles.
For subcutaneous studies, 8-week-old C57Bl/6 mice were sedated with isoflurane (Baxter), and the right hind flank was shaved prior to a subcutaneous injection of 0.5 ϫ 10 6 cells in 50 l of PBS. The mice were monitored, and treatment began when the largest tumor reached 4 mm in diameter, 7-10 days after B16-OVA injection. Tumor sizes were normalized between groups after initial measurement. Systemic nanoparticle injections were performed intraperitoneally. When each component was given, each dose consisted of 1 g of IL-15:IL-15R␣, 0.4 mg of PEG-modified PLGA nanoparticles, and 4 g of endotoxin-free ovalbumin. Tumor size was monitored every other day by caliper measurements. Mice were euthanized by CO 2 asphyxiation when tumors reached 15 mm in greatest dimension or when the mice exhibited signs of sickness.

Characterization of Antigen-loaded Nanoparticles Presenting
Multivalent IL-15:IL-15R␣-Biotinylated human IL-15:IL-15R␣ was mixed with avidin-coated nanoparticles at a ratio of 2 g/ml IL-15:1 mg/ml nanoparticles. Assuming maximal binding, this ratio yields ϳ800 molecules of IL-15:IL-15R␣ on the surface of each NP. To demonstrate complex exposure on the NP, we used phycoerythrin-labeled anti-human IL-15 and analyzed NP binding by flow cytometry. Bare nanoparticles were used as a negative control, and the dependence of IL-15:IL-15R␣ on the avidin-biotin interaction was tested using nanoparticles treated with excess free biotin before binding with IL-15:IL-15R␣. Nanoparticles treated with biotinylated IL-15: IL-15R␣ were stained uniformly with anti-IL-15, whereas pretreatment of nanoparticles with biotin completely inhibited IL-15 binding (Fig. 1a). The amount of avidin loaded onto NPs was determined by a total protein content assay to be 40 g/mg of NP.
To qualitatively demonstrate the presence of IL-15:IL-15R␣ on nanoparticles, we examined the modified NPs using TEM. Anti-human IL-15 was labeled with gold nanoparticle-linked anti-IgG1. IL-15:IL-15R␣-bound nanoparticles were labeled with gold nanoparticles, whereas gold labeling was not observed on bare nanoparticles at similar staining concentrations (Fig. 1b).
To show that IL-15:IL-15R␣-coated NPs were internalized by DCs, we prepared sections of DCs incubated with either IL-15:IL-15R␣-coated NPs or uncoated NPs and stained the sections for IL-15. On TEM analysis, NPs could be seen in the endosomal compartments of both DC preparations, but only the IL-15:IL-15R␣-treated DCs showed IL-15 staining in the endosomal compartments (Fig. 1c). This confirmed that the modified NPs were internalized by DCs and that intracellular IL-15 can be detected after NP internalization.
IL-15:IL-15R␣ on Nanoparticles Increases DC-mediated Stimulation of CD8 ϩ T Cells-To assess the effect of IL-15:IL-15R␣ NP (without encapsulated antigen) on DC-mediated stimulation of CD8 ϩ T cells, NPs with and without the IL-15 complex were first incubated with murine bone marrow-derived DCs. Because human IL-15:IL-15R␣ is cross-reactive with both mouse and human cells (15,44), it is suitable for use in our murine experimental system. CD8 responses were monitored using purified splenic CD8 ϩ T cells from OT-I transgenic mice specific for the K b -restricted peptide SIINFEKL (OVA 257-264) (76). DCs were incubated with 50 g/ml exogenous OVA in the presence of various IL-15:IL-15R␣-bearing nanoparticle constructs (Fig. 2a). Additionally, we compared the effect of free IL-15:IL-15R␣ together with free antigen. To control for effects induced by the nanoparticle vehicle itself, two additional controls were included: nanoparticles alone and nanoparticles preincubated with biotin to block surface binding sites and then mixed with IL-15:IL-15R␣ (Fig. 2b).
When DCs were incubated with soluble OVA and nanoparticles bearing IL-15:IL-15R␣, a greater production of IFN-␥ was observed compared with free IL-15:IL-15R␣ at identical concentrations (Fig. 2c). In the absence of CD8 ϩ T cells, little IFN-␥ was observed (data not shown), implicating CD8 ϩ T cells as the major producer of observed IFN-␥ upon DC stimulation. The activation of CD8 T cells was antigen-specific because OT-I T cells failed to proliferate when an equal concentration of exogenous BSA was used in place of OVA (Fig. 2d). Interestingly, increasing the density of surface-bound IL-15:IL-15R␣ correlated with increased levels of IFN-␥ produced by CD8 ϩ T cells, suggesting a dose-response effect on IFN-␥ because of IL-15:IL-15R␣ treatment (Fig. 2e). These results indicate that nanopar- ticles presenting IL-15:IL-15R␣ act to enhance the antigen-specific cross-priming of CD8 ϩ T cells by DCs.
Previous work has shown that the rate of antigen delivery (persistence of antigen) affects the long-term memory immune response and vaccine efficacy in pathogen recall (67,68). We therefore hypothesized that multivalent presentation of IL-15: IL-15R␣ may synergize with antigen encapsulation in PLGA to enhance CD8 ϩ T cell stimulation (Fig. 3a). PLGA nanoparticles encapsulating ovalbumin (NP/OVA) were prepared by a double emulsion method and used as the antigen source in place of exogenous OVA (Fig. 3b). NP/OVA particles were determined to have loaded 40 g of OVA/mg of NP. When IL-15:IL-15R␣ was trans-presented on nanoparticles that also encapsulated antigen, CD8 ϩ T cell activation by DCs that internalized these particles was similarly enhanced above that of free IL-15:IL-15R␣ alone, NP/OVA alone, or NP/OVA with IL-15: IL-15R␣ separated from the particle surface (Fig. 3c). As before, when T cells were removed, cytokine production was abrogated (data not shown). The antigen specificity of the CD8 ϩ response was confirmed when particles without antigen loading failed to elicit T cell proliferation (Fig. 3d). In addition, the IFN-␥ response was dependent on IL-15:IL-15R␣ surface density on  Fig. 1. c, the adjuvant effect of these formulations was evaluated by IFN-␥ production from OT-I CD8 ϩ T cells. Data were analyzed by a one-way ANOVA and a Tukey post-hoc test. *** indicates a significance level of p Յ 0.001. d, the antigen-specific nature of the T cell response was tested with nanoparticles encapsulating OVA (NP/OVA) and unloaded nanoparticles (NP/Ϫ) using CFSE-labeled OT-I CD8 ϩ T cells as responder cells. e, IL-15 was titrated on a fixed concentration of OVA-loaded nanoparticles to determine the dose dependence of IFN-␥ production on IL-15 concentration.
the antigen-loaded nanoparticles (Fig. 3e), similar to what was observed with exogenous antigen. The resulting maximal IFN-␥ responses in this experiment were much higher than those shown in Fig. 2 (6 ng/ml in Fig. 3c versus 2 ng/ml in Fig.  2c), indicating a stronger immune response elicited by the encapsulated antigen (38,39,77,78). We confirmed this by comparing free OVA and NP/OVA particles at the same dose (data not shown).
Taken together, these results demonstrate that the surface presentation of IL-15:IL-15R␣ on nanoparticles enhances the capacity of DCs to cross-prime CD8 ϩ T cells in an antigenspecific manner. This effect was confirmed in cases where antigen was either internalized exogenously ( Fig. 2) or present within the signaling nanoparticles (Fig. 3). We also verified that the DC-T cell interaction is critical for the cytokine response because the exposure of either cell type alone to IL-15:IL-15R␣ nanoparticle constructs failed to produce significant levels of cytokines (data not shown). We therefore conclude that IL-15: IL-15R␣ is able to act within the context of the DC:T cell interface to enhance the DC-mediated T cell response.

Synergy between Heterodimeric IL-15 and Encapsulated Antigen Requires Coordinated Delivery to DCs and T Cells-
The antigen delivery system employed here has two components: IL-15:IL-15R␣ trans-presented on NPs, which we have shown to be more effective than free IL-15:IL-15R␣ in inducing CD8 ϩ T cell responses, and encapsulated antigen (OVA), which is also more effective than free antigen for cross-presentation and induction of both CD4 ϩ and CD8 ϩ T cell responses (33,39,77). Given their separate functions, we asked whether the co-localization of both multivalent IL-15 and encapsulated antigen offer clear advantages compared with separate NPs delivering both elements. It is currently unknown whether the co-localized delivery of multivalent IL-15 and antigen to antigen-presenting cells has an effect on the effector response. We tested this hypothesis in three different ways, comparing our results each time to NPs presenting the IL-15 heterodimer and encapsulating OVA antigen (IL-15/NP/OVA, black columns to the right of each plot in Fig. 4).
First, we hypothesized that the delivery of multivalent IL-15 (IL-15-NP) and encapsulated OVA (NP/OVA) to the same cells would be more effective than if they were delivered to separate cells. DCs were incubated with empty IL-15:IL-15R␣-coated NPs (IL-15/NP-DCs). Separately, another population of DCs from the same batch were incubated with OVA-encapsulating NPs (NP/OVA-DCs). These cells were mixed together at a ratio equivalent to the ratio of IL-15:IL-15R␣ and antigen associated with a single particle (IL-15/NP-DC ϩ NP/OVA-DC). DCs incubated with separate or single particles (maintaining the same ratio of IL-15 to antigen) were then used to stimulate OT-I CD8 ϩ T cells. We found that mixed cells (separately receiving either IL-15 or antigen) were less effective at inducing CD8 T cell proliferation than the combined particle system (IL-15/NP/OVA) (Fig. 4a). Notably, the mixed DCs produced IFN-␥ responses on par with IL-15/NP-DCs and NP/OVA-DCs, suggesting that enhanced CD8 ϩ T cell responses are stimulated by the localization and potential synergy of IL-15 and antigen at the same DC.
IL-15 can powerfully affect CD8 ϩ T cell proliferation and activation and also enhance their persistence in vivo (17,48,79,80). Indeed, tumor or virus-specific T cells cultivated with IL-15 have been shown to exhibit a high activity against their respective antigenic targets (81). Therefore, we hypothesized that NPpresented IL-15:IL-15R␣ may also prime CD8 ϩ T cells against NP/OVA-loaded DCs. To address this question, we incubated OT-I CD8 ϩ T cells overnight with buffered saline as a control (T), IL-15:IL-15R␣ alone (15-T), or nanoparticle-bound IL-15: IL-15R␣ (15NP-T). We separately stimulated DCs with antigen by loading them with NP/OVA particles (Fig. 4b, left panel). After 12 h, T cells and DCs were washed, counted, plated together at the appropriate cell concentrations, and allowed to incubate for a further 72 h. The results are shown in Fig. 4b. We observed that CD8 ϩ T cell proliferation in response to antigenloaded DC was greater when T cells were preincubated with IL-15:IL-15R␣. Interestingly, NP-bound IL-15:IL-15R␣ was a stronger priming stimulus compared with free IL-15:IL-15R␣ (Fig. 4b, right panel). In addition, the proliferation of CD8 ϩ T cells preactivated with IL-15:IL-15R␣/NP and incubated with NP/OVA-DCs (Fig. 4b, NP/OVA-DC ϩ 15NP-T) was greater compared with the one obtained with DCs using the combined nanoparticle platform (Fig. 4b, IL-15/NP/OVA, black column). This result suggested that IL-15 trans-presented on NP and antigen-loaded NP may function on the separate axes of effector cells and antigen-presenting cells. IL-15 may act to prime and activate CD8s to respond to antigen, whereas nanoparticleencapsulated antigen is internalized by DCs and cross-presented to CD8s.
Having tested how our nanoparticle components work when delivered to different cells (separate delivery to different DCs and separate delivery to CD8 ϩ cells and DCs), we then tested a case where all nanoparticle components were delivered to the same DC. We asked whether the IL-15 and antigen components had to be presented on the same NP vehicle or whether trans-presented IL-15 and encapsulated antigen could still synergize when delivered on separate NP vehicles. We made separate nanoparticle constructs with trans-presented IL-15:IL-15R␣ and encapsulated antigen, respectively, and prevented transfer of any biotinylated IL-15:IL-15R␣ by blocking the avidin binding sites on NP/OVA using free biotin. We then tested these separated constructs (IL-15:IL-15R␣/NP ϩ NP/OVA) alongside the combined construct (IL-15:IL-15R␣/NP/OVA) and along with the separate components in soluble form. We found that the combined construct was the most effective at stimulating CD8 ϩ T cell proliferation (Fig. 4c), indicating that delivery on the same vehicle is needed for maximal stimulation. Again, both the separate and combined constructs performed significantly better than the soluble configurations where IL-15 and OVA were delivered apart from their optimal NP configurations ( Figs. 2 and 3). Overall, these results demonstrate a need for the localized delivery of trans-presented IL-15:IL-15R␣ and encapsulated antigen for synergy to occur.
Multivalent IL-15:IL-15R␣ on NP Functions in the Same Manner with Human Cells-The translational potential of these findings depends on whether the same findings are observed with human effector and antigen-presenting cells. To validate our findings using a clinically relevant human antigen, NPs were loaded with a 26-mer peptide derived from the Melan-A/MART-1 melanoma-specific antigen (82). IL-15:IL-15R␣ NPs loaded with the MART-1 antigen were incubated with human HLA-A2 ϩ DCs and then used to stimulate DMF-5 responder cells (Fig. 5a). DMF-5 cells carry TCR receptors specific for the MART-1 [27][28][29][30][31][32][33][34][35] sequence and respond to antigenspecific stimulation by producing IFN-␥ (83). Consistent with observed findings in the mouse system with the encapsulated OVA antigen, the MART-1 antigen encapsulated in NPs transpresenting multivalent IL-15:IL-15R␣ led to a greater IFN-␥ response from DMF5 cells compared with soluble IL-15:IL-15R␣ (Fig. 5b). This result highlights the clinical potential of this mode of IL-15 delivery and possibly points to the promise of enhancing a cytokine response by changes in the contextual delivery aspect compared with, for example, genetic manipulation of the IL-15 protein sequence to affect the desired response (84,85).  15R␣ and OVA delivery. a, the requirement for both particle types to go to the same cell was tested by mixing DCs that were separately given IL-15/NP or NP/OVA and comparing them with DCs given IL-15/NP/OVA. b, the effect of priming CD8ϩ T cells with IL-15 constructs was tested by priming CD8 ϩ T cells overnight with free or NP-bound IL-15:IL-15R␣ and stimulated with NP/OVA-DCs. c, to test the requirement for OVA or IL-15 to be delivered on the same particle, DCs were incubated with different particle constructs overnight and then used to activate OT-I CD8 ϩ T cells. In all cases, the T cell response was compared against untreated DCs and IL-15/NP/OVA-treated DCs. OT-1 cell counts were determined as a measure of total stimulatory capacity of each group. Data were analyzed by a one-way ANOVA and a Tukey post-hoc test. * indicates a significance level of p Յ 0.05, ** p Յ 0.01, and *** indicates p Յ 0.001.

The Mechanism of Action of Multivalent NP-presented IL-15: IL-15R␣ Involves Particle-mediated Transfer of IL-15 to the DC
Surface-We set out to determine the molecular mechanism by which IL-15:IL-15R␣ augments the stimulatory capacity of antigen-presenting DCs. Because IL-15 trans-presentation on DCs affects the activation of CD8 T cells, we asked whether delivered IL-15 may be retained on the surface of DCs for stimulation during T cell encounters. To address this question, we treated mouse DCs with free human IL-15:IL-15R␣, nanoparticle-bound IL-15:IL-15R␣, or IL-15:IL-15R␣ mixed with biotin-blocked nanoparticles and stained DCs for human IL-15 2 h later. Staining for human IL-15 ensured that we would detect only the delivered molecules and not endogenous murine IL-15. We found that incubating DCs with nanoparticle-bound IL-15:IL-15R␣ resulted in increased levels of detectable human IL-15 on the DC surface in a manner commensurate with the density of IL-15:IL-15R␣ on the nanoparticle surface. In contrast, free human IL-15:IL-15R␣ did not result in accumulation of human IL-15 on the DC surface (Fig. 6a).
We next sought to characterize the persistence of this IL-15: IL-15R␣ retention on the DC surface. DCs were treated with IL-15:IL-15R␣ constructs for 2 h, washed, and cultured over time before staining with phycoerythrin-labeled anti-human IL-15. DCs treated with NP-bound IL-15:IL-15R␣ showed an initial 200% increase in fluorescence because of IL-15 labeling compared with untreated controls (Fig. 6b). Although the fluorescence decayed steadily over 8 h, a strikingly high level of IL-15 was still retained on the surface of DCs, leading to a Ͼ100% increase in IL-15:IL-15R␣-associated fluorescence compared with untreated cells. This amount was steady after 8 h and remained despite wash steps. In contrast, when IL-15: IL-15R␣ was given in soluble form, very little IL-15 accumulated at the DC surface, and what little did accumulate was cleared quickly from the membrane (Fig. 6b).
Next we asked whether the surface-retained IL-15:IL-15R␣ was able to functionally stimulate CD8 ϩ T cells. DCs were treated with different concentrations of nanoparticle-bound IL-15:IL-15R␣ for 2 h, resulting in different levels of IL-15 retention on the DC surface, as revealed by flow cytometry. DCs were then washed and used to stimulate purified CD8 ϩ T cells at a 1:1 DC:T cell ratio for 1 h. T cells were then stained for phosphorylated STAT5 to evaluate IL-15-specific signaling. We found that increasing levels of IL-15 on the DC surface correspondingly increased the level of phosphorylated STAT5 in CD8 ϩ T cells (Fig. 6c), demonstrating that the IL-15 delivered to the surface of DCs remains functional and capable of stimulating CD8 ϩ T cells.
Because IL-15 is present in high concentrations at the DC surface within a short time after NP delivery, it is possible that non-internalized NP nonspecifically associated with the cell membrane and displayed IL-15, leading to the observed high IL-15 levels at steady state. If this is the case, then the treatment of DCs with avidin-coated NPs (the base scaffold on which IL-15 is displayed) would result in the display of avidin binding sites on the DC surface, allowing for the binding of biotinylated ligands such as biotin-IL-15:IL-15R␣. To test this hypothesis, DCs were incubated with avidin-coated NPs in either media or PBS for 1 h at 37°C. Biotin-IL-15:IL-15R␣ was then added to the DCs for 15 min ( Fig. 6d; NP, then IL-15:IL-15R␣; red line). All cells were washed and stained for IL-15 and analyzed by flow cytometry.
When the experiment was performed in media, avidin binding sites could not be detected via IL-15-associated fluorescence (Fig. 6d, left panel). However, when the experiment was performed in PBS, the "NP, then IL-15:IL-15R␣" group resulted in fluorescence similar to our positive control (Fig. 6d, IL-15:IL-15R␣/NP, blue line), suggesting the binding and detection of IL-15 at the DC surface (Fig. 6d, right panel). Biotin blockade prior to IL-15 binding ( Fig. 6d; NP, biotin block, then IL-15:IL-15R␣, green line) abrogated the IL-15 signal, showing that the effect is due to biotin binding sites on the cell surface. IL-15 labeling was not detected after treating the cells with free IL-15: IL-15R␣ alone, indicating that this effect is not due to any biotin binding proteins on the cell surface (Fig. 6d).
If IL-15-bearing NPs associate nonspecifically with the DC surface, then DCs may be able to retain IL-15 at their membranes in the absence of active cell processes. We therefore investigated the ability of fixed DCs to present IL-15 to CD8 ϩ T cells. DCs were fixed with paraformaldehyde and then incubated with IL-15:IL-15R␣/NP for 2 h. We found that these fixed, IL-15-treated DCs stained positive for IL-15, indicating that fixed DCs can accumulate IL-15 on their surface (Fig. 6e). The IL-15 staining was at a lower level than live DCs (Fig. 6e, green line) exposed to IL-15:IL-15R␣/NP for the same duration, indicating that some active cellular mechanisms may aid in (but are not required for) IL-15 accumulation on the cell surface.
Finally, we asked whether the IL-15 presented on the surface of fixed DCs was bioactive. Live and fixed DCs were co-incubated with IL-15:IL-15R␣/NP for 2 h and then washed extensively and incubated with CD8 ϩ T cells for 1 h. Phospho-STAT5 staining was used to assess IL-15 signaling in CD8 ϩ cells, and live or fixed DCs unexposed to IL-15 were used as controls. We found that the IL-15 present on the surface of both live and fixed DCs was able to stimulate STAT5 phosphorylation in CD8 ϩ T cells (Fig. 6f), demonstrating that fixed DCs are able to present bioactive IL-15.
Our data suggest that the incubation of DCs with NPs results in some of the NPs embedding in the cell surface, exposing their surface ligands (IL-15 or avidin) for interactions with neighboring cells. Such behavior could account for the retention of IL-15 on the DC surface. The fact that IL-15 binding (in the NP, then IL-15:IL-15R␣ group) could not be detected in the presence of media suggests that media components were able to shield or block the interaction of biotin-IL-15 with embedded avidin-NPs. Indeed, under the right circumstances, the avidin-biotin interaction on a surface can be blocked or inhibited (86,87). Furthermore, active cellular mechanisms are not required for the retention of bioactive IL-15 at the DC surface. Although fixed DCs cannot process and present antigen, they are able to act as physical scaffolds for NP-delivered multivalent IL-15. Overall, our findings demonstrate that nanoparticle-bound IL-15:IL-15R␣ delivers functional IL-15:IL-15R␣ to the surface of DCs in a manner dependent on the density of IL-15:IL-15R␣ bound to the nanoparticles and persists on the DC surface for at least 24 h.
NP-delivered IL-15 Is Not Observed on DC-derived Exosomes-We investigated the possibility that IL-15 might be exported from the DC membrane via exosomes. Indeed, IL-15R␣ has been detected previously on human DC-derived exosomes (88). We therefore purified exosomes from DCs by sequential ultracentrifugation (89). DCs were allowed to take up NPs overnight (untreated control, free IL-15:IL-15R␣, NP only, and IL-15:IL-15R␣/NP), and then exosomes were purified from the DC-conditioned media. Exosome preparations were examined by TEM, revealing circular and cup-shaped membrane vesicles about 100 nm in diameter (Fig. 7a). Exosome size distribution was also determined by particle tracking analysis (NanoSight) and found to be between 100 and 300 nm (Fig. 7b).
We then examined exosome-associated proteins in our preparations by flow cytometry. All exosome preparations stained positive for class II MHC and slightly positive for CD86 (Fig.  7c), both of which are known to be present in DC-derived exosomes (90). However, none of the exosome preparations contained detectable levels of IL-15 (Fig. 7c), suggesting that IL-15 is not associated with DC-derived exosomes.
If IL-15 and antigen were being rereleased by exosomes, then DC-derived exosomes may be able to directly stimulate OT-I T cells in the absence of cell contact. We therefore tested the ability of our NP-treated DCs to stimulate T cell activation in a transwell experiment. DCs were incubated overnight with the following nanoparticle constructs: IL-15:IL-15R␣/NP, NP/ OVA, and IL-15:IL-15R␣/NP/OVA. After extensive washing to remove excess NPs, DCs were used to stimulate CD8 ϩ T cells either in cell-cell contact or separated by a 400-m pore size transwell. The addition of the transwell completely abolished the proliferative (Fig. 7d, left panel) and cytokine response (Fig.  7d, right panel) of the CD8 ϩ T cells for all NP groups, suggesting that exosomes do not carry IL-15 or antigen to T cells across the transwell membrane. We therefore conclude that IL-15 is not observed on DC-derived exosomes and that it does not contribute to T cell stimulation via an exosome-mediated mechanism.
IL-15 Enhancement of the Antigen-specific CD8 ϩ Response Mediated by NP Multivalent Presentation Is Not due to Enhancement of Antigen Processing, NP Uptake, or DC-specific IL-15 Signaling-Given that NP delivered an encapsulated form of the whole OVA protein, we next asked whether IL-15:IL-15R␣ was acting on the antigen-processing machinery in DCs.
To answer this question, we encapsulated the SIINFEKL peptide within IL-15:IL-15R␣-coated nanoparticles. We observed that a stronger IFN-␥ response was obtained than when either IL-15:IL-15R␣ was not added or detached from the biotinblocked nanoparticle surface (Fig. 8a). APCs incubated with IL-15:IL-15R␣ NP therefore potentiate CD8 ϩ T cell stimulation in a manner independent of antigen.
Next we asked whether NP-presented IL-15:IL-15R␣ affected the rate of NP internalization and, therefore, antigen uptake. Avidin-coated NPs were surface-labeled with Alexa Fluor 488 and then incubated with DCs in the presence or absence of surface-bound IL-15:IL-15R␣. DCs were fixed and analyzed for Alexa Fluor 488 fluorescence by flow cytometry. We found that IL-15:IL-15R␣ had no discernable effect on the rate of fluorescence accumulation in DCs (Fig. 8b), indicating that the trans-presentation of IL-15:IL-15R␣ does not affect the rate of NP internalization in our system. Finally, we asked whether IL-15:IL-15R␣ induces its effects by activating signaling pathways within DCs during NP uptake. 2 h prior to NP uptake, DCs were incubated with either 10 g/ml anti-human IL-15, 50 M STAT5 inhibitor (NЈ-((4-oxo-4H-chromen-3-yl)methyl-ene)nicotinohydrazide), 2 M PI3K inhibitor (LY294002), or 2 M MEK inhibitor (U0126). We assessed the effect of these inhibitors by evaluating the -fold increase in CD8-mediated IFN-␥ secretion when surfacebound IL-15:IL-15R␣ was added to NP/OVA during particle uptake by DCs. Although the addition of anti-human IL-15 did partially decrease IFN-␥ production, none of the other inhibitors appreciably affected the activation of CD8 ϩ T cells by the NP-loaded DCs (Fig. 8c). In addition, DCs did not show elevated levels of STAT5 phosphorylation upon stimulation with IL-15:IL-15R␣, lending support to our hypothesis that the primary effect of delivered IL-15:IL-15R␣ to DCs is to stimulate neighboring T cells.
Artificial Antigen Presentation Together with IL-15:IL-15R␣ Improves the Sensitivity and Magnitude of the Antigen-specific Response-APCs provide at least three essential signals to activate T cells: antigen-specific stimulation by peptide-MHC complexes; costimulatory signals, most commonly via the CD28/B7 interaction; and the release of cytokines to modulate T cell expansion, survival, and function. Our data indicated that IL-15:IL-15R␣-bearing, antigen-loaded NPs are able to enhance specific T cell responses, transferring IL-15:IL-15R␣ molecules to the DC surface in a manner dependent on the NP surface density of IL-15:IL-15R␣ but independent of NP internalization, antigen processing, and DC signaling. This effect echoes the physiological mechanism of action of IL-15, which is trans-presented on the surface of activated monocytes and DCs via binding to membrane-bound IL-15R␣ (13). Therefore, we wanted to understand what happens when IL-15 is presented in a physiological fashion in a fully functioning system with specified parameters. To do so, we designed a nanoparticle-based antigen-presenting cell that provides all of the essential signals for T cell stimulation: MHC-complexed SIINFEKL for antigen presentation, anti-CD28 antibodies for costimulation, and the presence or absence of trans-presented IL-15:IL-15R␣ as the cytokine signal (Fig. 9a). We have demonstrated previously that an aAPC system consisting of antigen presentation, costimulation, and cytokine secretion is able to efficiently expand target T cell populations (16,91).
To assess the difference in levels of T cell stimulation by natural versus artificial APCs, we compared the ability of both systems to stimulate 1 ϫ 10 5 OT-I CD8 ϩ T cells under conditions of equivalent antigen presentation. To equalize antigen presentation between NP-loaded DCs and aAPCs, we loaded DCs with 0.1 mg/ml NP/OVA. Because DCs produce one SIINFEKL-presentingMHCforeveryϳ3000OVAproteinsprocessed (92), this level of NP/OVA loading produces peptide: MHC on the order of 1 nM. We therefore compared the following T cell stimulation systems: DCs that internalized hetIL-15coated, antigen-loaded NPs and aAPCs coated with hetIL-15 and 1 nM each of SIINFEKL:MHC and anti-CD28 (Fig. 9b). The FIGURE 7. IL-15 is not associated with DC-derived exosomes. a, exosome preparations were visualized by TEM. Scale bar ϭ 200 nm. A representative image is shown. b, representative size distribution plot of exosomes sized using nanoparticle tracking analysis. Black line, mean; red, error bars. c, exosome-associated proteins were analyzed by flow cytometry. Exosomes were adsorbed to aldehyde/sulfate beads and stained for MHC class II, CD86, and IL-15. d, NP-loaded DCs were assayed for their ability to stimulate antigen-specific T cell activation via exosomes. DCs were used to stimulate CFSE-labeled T cells either separated by a transwell (Transwell, black columns) or in direct cell contact (Co-culture, white columns). T cells were assayed for CFSE dilution (left panel) and IFN-␥ secretion. The differences in mean values between the co-culture and Transwell experiments for each condition were compared using an unpaired t-test. *** indicates a significance level of p Յ 0.001, ** indicates p Յ 0.01, and * indicates p Յ 0.05. same mass of NP and amount of IL-15 were used for each system.
Both systems stimulated CD8 ϩ T cell proliferation and led to IFN-␥ secretion. However, aAPCs stimulated greater cell proliferation and lower IFN-␥ production compared with NP-loaded DCs (Fig. 9c). These differences may reflect changes in the molecular composition and physical aspects of the natural and artificial APCs.
We then examined the effect of IL-15 presentation on our aAPC system. Equimolar concentrations of SIINFEKL-loaded MHC dimers and anti-CD28 were titrated onto a fixed concentration of nanoparticles that were then used to stimulate OT-I CD8 ϩ T cells in the presence or absence of a fixed concentration of IL-15:IL-15R␣ added to the nanoparticle surface. In the absence of IL-15:IL-15R␣, maximal T cell proliferation of 1 ϫ 10 4 cells occurred at about 10 nM peptide:MHC. In the presence of IL-15:IL-15R␣, a maximal proliferation of 1.5 ϫ 10 4 cells occurred at 1 nM peptide:MHC, representing a 10-fold increase in T cell sensitivity to antigen and a 50% increase in the magnitude of response (Fig. 9d, left panel). Similarly, the antigen concentration required for maximal IFN-␥ secretion was reduced from 30 to 10 nM peptide:MHC with the addition of IL-15:IL-15R␣, although the highest amount of IFN-␥ produced remained the same (Fig. 9d, right panel). This demonstrated the ability of trans-presented IL-15:IL-15R␣ to improve the sensitivity and magnitude of the antigen-specific CD8 ϩ T cell response.
Besides IL-15, another critical cytokine provided by APCs to T cells is IL-2. IL-2 and IL-15 share two of three receptor chains and identical signaling pathways but effect different responses in immunity through differential receptor distribution and modes of action (79,(93)(94)(95). IL-15 is trans-presented on the surface of activated monocytes and DCs, whereas IL-2 is secreted in a paracrine fashion from DCs or helper T cells. We hypothesized that the physiological presentation of both IL-15 and IL-2, through the trans-presentation of IL-15 and paracrine delivery of IL-2, can combine the effects of both cytokines to improve the T cell response in our aAPC system. We tested this by adding two different components to our system: exogenous IL-2 at a concentration of 0.5 ng/ml and encapsulating IL-2 within our PLGA nanoparticles at a concentration of 5 ng/mg PLGA. Without IL-15, exogenous IL-2 improves the CD8 response to approximately the same level as when IL-15:IL-15R␣ was added, with a maximum proliferation of 1.5 ϫ 10 4 cells at 1 nM peptide:MHC. Addition of IL-15:IL-15R␣ on the surface of the aAPCs further improves this to 2.1 ϫ 10 4 cells at 0.45 nM (Fig. 9e, left panel). A slight increase in the sensitivity of IFN-␥ production (from 3.6 to 3.1 nM peptide:MHC) was also observed (Fig. 9e, right panel). Encapsulated IL-2 is a powerful stimulant for T cell proliferation (16,96). Without IL-15:IL-15R␣, responding T cells peaked at 0.7 nM peptide:MHC for 2.6 ϫ 10 4 cells, whereas the addition of IL-15:IL-15R␣ increased the response still further to 2.9 ϫ 10 4 cells at 0.35 nM peptide:MHC, effectively doubling the sensitivity of T cells to antigen compared with encapsulated IL-2 alone (Fig. 9f, left  panel). The addition of encapsulated IL-2 appeared to saturate the IFN-␥ response, resulting in greatly increased maximal cytokine production but no change in antigen sensitivity of IFN-␥ production (Fig. 9f, right panel). Our results demonstrate that IL-15:IL-15R␣ is able to synergize with paracrine delivered IL-2 to enhance T cell activation.  15R␣/NP/OVA. Only the third configuration co-delivered IL-15 together with antigen. Control groups were given either PBS or blank nanoparticles alone. Although survival was not improved significantly between free and NP-bound IL-15:IL-15R␣, the addition of encapsulated OVA led to a significant increase in survival (Fig. 10a). IL-15:IL-15R␣ can therefore synergize with antigen delivery in the anti-tumor immune response.
We next tested the specific configuration of IL-15 and OVA necessary for anti-tumor responses. C57Bl/6 mice were inoculated subcutaneously with 0.  10b). Notably, configurations where either IL-15:IL-15R␣ or OVA was detached from the NPs did not enhance animal sur- vival beyond the control group. Our findings align with recent literature showing that OVA vaccination requires an adjuvant to potentiate antigen-specific protective immunity (97)(98)(99)(100) and further highlight the importance of adjuvant configuration in the case of IL-15. Overall, we show that nanoparticles transpresenting IL-15:IL-15R␣ and encapsulating antigen demonstrate in vivo efficacy in the treatment of an aggressive model of murine melanoma.

Discussion
We highlight in this work the effect of the mode of cytokine delivery on the immune response specifically with respect to IL-15 signaling. IL-15:IL-15R␣ on nanoparticle scaffolds enhances the sensitivity and magnitude of the antigen-specific stimulation of CD8 ϩ T cells upon stimulation by nanoparticletreated DCs. Our experiments highlight, we believe for the first time, the configuration-dependent delivery of IL-15:IL-15R␣ as a key factor for its biological action: that its presentation on the surface of a nanoparticle or cell leads to enhancement of its adjuvant effect. Several groups have previously compared the effects of soluble and trans-presented IL-15, but most compared the action of IL-15 in the presence or absence of cells or scaffolds presenting IL-15R␣ (17,48,101). However, because IL-15R␣ enhances activity when bound to IL-15 (11,44,55,56), these studies could not differentiate between the topological effects of IL-15 from the agonist behavior of IL-15R␣. Mortier et al. (47) have compared the activation of natural killer cells by DC supernatants containing IL-15:IL-15R␣ and DCs trans-presenting the same, although the membrane-bound concentration of IL-15:IL-15R␣ was unknown. Giron-Michel et al. (102) have demonstrated differential signaling and functions between soluble and membrane-bound IL-15:IL-15R␣ in human hematopoietic progenitor cells, although IL-15 was not controlled for concentration and the membrane-bound form of IL-15 investigated was a novel form complexed with the ␤␥ receptors. By showing that the same quantity of IL-15:IL-15R␣ delivered to DCs can differentially affect responding T cells, we show that the action of IL-15:IL-15R␣ can be regulated by its configuration during delivery or presentation to cells. Because IL-15:IL-15R␣ is physiologically present in both membranebound and soluble forms, our work suggests the increased efficacy of surface-presented IL-15:IL-15R␣ as an explanation for why the major effects of IL-15 are dependent on its trans-presentation on cell membranes.
We observed that the retention of functional IL-15:IL-15R␣ on the surface of DCs is a mechanism for the nano-adjuvant behavior seen with our IL-15:IL-15R␣ constructs. The nanoparticle-bound IL-15:IL-15R␣ is transferred to the DC cell surface, where it stimulates CD8 ϩ T cells alongside peptide:MHC engagement (Fig. 11a). IL-15R␣ and IL-2R␣ have been observed previously to co-localize with class I and class II MHC in supramolecular receptor clusters, supporting our observation that the transfer of IL-15:IL-15R␣ to the DC membrane could enhance the sensitivity and magnitude of T cell stimulation (103). Three mechanisms could account for this IL-15mediated adjuvant behavior: the scaffolding of IL-15:IL-15R␣ to the DC membrane by nanoparticles, the recycling of IL-15: IL-15R␣ to the DC membrane after nanoparticle internaliza- FIGURE 10. In vivo effect of trans-presented IL-15 and encapsulated antigen in a melanoma model. To test the adjuvant effects of surface-presented IL-15:IL-15R␣ in vivo, we used a B16 melanoma model. a, C57Bl/6 mice were injected intravenously with 1 ϫ 10 5 OVA-expressing B16 melanoma cells (B16-OVA) to establish pulmonary tumor nodules (n ϭ 5 mice/group). 1 day, 2 days, and 10 days after B16-OVA inoculation, mice were injected intraperitoneally with different IL-15:IL-15R␣ constructs. Mice were sacrificed upon signs of pulmonary distress. Survival curves were compared against the negative control (PBS) using the Mantel-Cox test. ** indicates a significance level of p Յ 0.01, and * indicates p Յ 0.05. b, to test the effect of IL-15 configuration, mice were inoculated subcutaneously with 0.5 ϫ 10 6 OVA-expressing B16 melanoma cells. Doses of nanoparticle or IL-15 constructs were injected intraperitoneally on days 1, 3, 5, 7, and 14 after tumors reached a minimum diameter of 4 mm. Arrows indicate treatment times (n ϭ 6 mice/group). Tumors were measured every other day, and mice were sacrificed when tumors exceeded 15 mm in greatest diameter. Survival curves were compared by Mantel-Cox and Gehan-Breslow-Wilcoxon analysis. Survival curves were compared against the negative control (NP only) using the Mantel-Cox test. ** indicates a significance level of p Յ 0.01 for the IL-15:IL-15Ra/NP/OVA group. tion, or export of IL-15 to T cells via DC-derived exosomes. The mechanism of NP-mediated scaffolding is supported by the observations that NPs accumulate at the cell membrane (104,105). Indeed, avidin binding sites could be detected on the DC surface shortly after addition of untreated NPs (Fig. 6d), and the nonspecific interaction of NPs with the DC membrane is supported by our observation that fixed DCs can present bioactive IL-15 (Fig. 6, e and f). Alternatively, studies by Waldmann and co-workers (13,48) have suggested that IL-15 and IL-15R␣ are recycled and stored intracellularly after internalization, indicating that IL-15 may be transported to the cell membrane after NP uptake, which may explain the longevity of surface-retained IL-15 on DCs in our experiments. We showed that intraendosomally delivered IL-15 could be detected after internalization, where it may survive to be recycled to the cell membrane. Finally, previous work has shown that IL-15R␣ associated with DC-derived exosomes can stimulate natural killer cells, leading to proliferation and IFN-␥ production (88). However, in our FIGURE 11. Proposed mechanism of adjuvant IL-15:IL-15R␣ action. a, the IL-15:IL-15R␣ on nanoparticles synergizes with encapsulated antigen to stimulate antigen-specific CD8 ϩ T cells. As nanoparticles are internalized, they embed in the cell membrane, presenting IL-15:IL-15R␣ on the cell surface. Alternatively, IL-15:IL-15R␣ could also be recycled to the DC membrane after it enters the nedosome. The PLGA nanoparticles aid in endosomal escape and cross-presentation of antigen on MHC class I. At the DC surface, IL-15:IL-15R␣ directly stimulates CD8 ϩ T cells and enhances their sensitivity to DC-presented antigen. b, IL-15:IL-15R␣ trans-presentation at the cell-NP interface may enhance interfacial IL-2 accumulation. The close apposition of an interacting cell and IL-2releasing nanoparticle normally leads to accumulation of IL-2 at the cell-NP interface. However, the addition of trans-presented IL-15:IL-15R␣ engages a portion of the IL-2R␤␥ receptors at the interface, reducing the number of receptors available for IL-2 binding and forcing IL-2 to accumulate over a wider area. This effectively increases the total number of activated IL-2R␤␥ receptors in the cell-NP interaction because the signaling induced by IL-15:IL-15R␣ transpresentation and IL-2 binding is identical. system, we did not observe proliferation or IFN-␥ being produced from CD8 ϩ transwell cultures, nor did we observe exosome-associated IL-15 (Fig. 7).
In our use of the artificial APC system to elucidate the mechanism of action of our NP-loaded DCs, we noted that the two systems generated differential responses. Although we observed increased proliferation with aAPCs compared with NPloaded DCs, that trend is reversed when we examined IFN-␥ levels. Antigen-specific proliferation and IFN-␥ release can be dissociated from each other in clonotypic activation of T cells in murine lines and in immune responses from splenic T cells (106,107). In addition, the presence of low levels of IL-2 can enhance IFN-␥ secretion independent of proliferation (107). We detected the presence of IL-2 produced in the case of NPloaded DCs but not from aAPCs (data not shown), which may contribute to the higher levels of IFN-␥ seen with NP-loaded DCs. aAPCs provide a very strong stimulus for T cell proliferation (16,96,108) and, indeed, outperform NP-loaded DCs in expanding T cells. We note that artificial APCs are by no means equivalent to natural DCs in terms of the breadth and intensity of immune responses triggered. However, by providing a basic approximation of antigen recognition, co-stimulation, and cytokine signaling, the utility of aAPCs for elucidating immune mechanisms is demonstrated in this work.
Our work affirms that the physiological presentation of IL-15 and IL-2 is important for their action on T cells. Use of the IL-15:IL-15R␣ complex and paracrine delivery of IL-2 were shown to better activate T cells than if either cytokine were delivered alone and could stimulate even stronger responses if delivered in combination in our aAPC system. We have previously described this phenomenon for IL-2 (16,96) and modeled the interaction of paracrine IL-2 with a receiving cell under a variety of conditions (71, 109 -110), which provides a basis for understanding how these two cytokines would interact with one another. When an aAPC releases a cytokine such as IL-2, mathematical modeling indicates that IL-2 accumulates significantly at the particle:cell interface. Locally concentrated IL-2 would bind readily to available IL-2/15␤␥ receptors, thereby enhancing T cell activation. If trans-presented IL-15:IL-15R␣ were present at the particle:cell interface, it would occupy a portion of the IL-2/15␤␥ receptors, allowing the local concentration of IL-2 to increase further. Such receptor occupancy would not hinder cytokine-induced T cell activation because the signaling ends of IL-2/15␤␥ are the same whether bound to IL-2 or IL-15:IL-15R␣ (93); nor would the number of IL-2/15␤␥ receptors decrease because engagement by trans-presented IL-15:IL-15R␣ does not trigger significant down-regulation of IL-2/15␤␥ (42). The combined action of physiologically presented IL-2 and IL-15 would therefore continue to increase T cell activation until available IL-2/15␤␥ receptors are saturated (Fig. 11b). These results shed more light on the interaction between IL-2 and IL-15 and show that the cytokines can be combined to stimulate robust T cell responses at levels below receptor saturation. These findings could have implications for the design of IL-2-and IL-15-based immunotherapies.
Multiple clinical trials for single-chain IL-15 and heterodimeric IL-15 are currently recruiting or underway. Initial results from the first-in-human phase I trial of single-chain IL-15 indi-cate a modest form of capillary leak that was less severe than that induced by IL-2 but also moderate toxicities, including hypotension and thrombocytopenia (52). In light of the significant clinical interest behind IL-15, our findings provide new strategies for optimizing its therapeutic delivery, aimed at maximizing its effectiveness while limiting its associated toxicities. Our results have particular significance in the application of IL-15 to therapeutic cancer vaccines where the nanoparticlemediated combination delivery of IL-15 and antigen to target DCs may prove to be a useful strategy.