Interleukin-15/Interleukin-15Rα Complexes Promote Destruction of Established Tumors by Reviving Tumor-Resident CD8⁺ T Cells

Cancer immunosurveillance is the process whereby innate and adaptive immune mechanisms suppress the growth of tumors (1, 2). CD8⁺ T cells and natural killer (NK) cells play important roles in this process by directly killing malignant cells (1, 3–5). Cancer immunosurveillance is regulated not only by the immune system but also by elements of the tumor microenvironment, including malignant cells, tumor stroma, and the vasculature (6). Indeed, tumors can escape immunosurveillance by disabling the function of cytolytic lymphocytes and antigen-presenting cells, by preventing blood-borne lymphocytes from infiltrating malignant tissue or by inducing tolerance (1, 7, 8).

Various immunotherapeutic strategies have been developed for the treatment of human cancers. Cancer vaccines strive to incite robust antitumor immunity by immunizing the cancer patient with different forms of tumor antigens; however, their effect on tumor burden has been modest (9–11). In adoptive cell therapy (ACT), patients are infused with autologous, tumor-specific T cells that can be derived from tumor-infiltrating lymphocytes (TIL) or from peripheral blood lymphocytes engineered to express a tumor-specific T-cell receptor (12, 13). Although ACT has been successful in inducing objective responses in select cancers such as metastatic melanoma, most patients still fail to respond despite increased frequencies of circulating, tumor-specific lymphocytes (14, 15). It is becoming increasingly clear that clinical efficacy in cancer immunotherapy may be more dependent on the ability of immune effector cells to access the tumor and to exert their tumoricidal functions therein rather than on the numbers of circulating, tumor-specific lymphocytes (16–19). Unfortunately, fewer efforts have focused on designing therapies that target tumor-resident T cells and boost their effector function in situ. Thus, new approaches are needed that either facilitate the infiltration of circulating leukocytes into solid tumors or that effectuate the tumoricidal function of TILs that persist in a functionally suppressed state in the malignant lesion.

The administration of cytokines to augment immunosurveillance has proven efficacious in the treatment of select cancers (20). For example, IL-2 is Food and Drug Administration approved for the treatment of renal cell carcinoma and metastatic melanoma. However, IL-2 therapy is limited by systemic toxicity, poor biological activity, and an inability to induce antitumor activity in most cancer patients (21) due to selective promotion of T-cell activation–induced cell death (AICD) and expansion of T regulatory cells (Treg; refs. 22–26). Highly related to IL-2 is the cytokine IL-15 (27, 28), which lacks these adverse effects. In addition to sharing the use of two receptor subunits (IL-2Rβ/CD122 and IL-2Rγ/CD132) and inducing similar intracellular signaling events, both IL-15 and IL-2 induce the mild expansion of memory CD8⁺ T cells, NK cells, and NKT cells (22). IL-15 has shown antitumor efficacy and enhances the effects of chemotherapy and ACT (29–32). However, like IL-2, IL-15 has a short half-life and high doses are needed to achieve biological responses in vivo (33, 34). Recently, it was shown that the biological activity of IL-15 could be increased ∼50-fold by administering preformed complexes of IL-15 and its soluble receptor, IL-15Rα (35, 36). This increase in activity is likely due to an increased half-life of the complex compared with IL-15 alone and that IL-15 is being presented by IL-15Rα to CD122⁺ cells similarly to how it is thought to be presented by dendritic cells in vivo. Compared with IL-15, IL-15/IL-15Rα complexes induce a dramatic expansion of CD122^hi cells, including antigen-experienced CD44^hi CD8⁺ memory and memory phenotype (MP) T cells, NK cells, and NKT cells (35, 36). Given these promising data, we hypothesized that IL-15/IL-15Rα complex–driven expansion of such lymphocytes may promote immune-mediated destruction of established solid tumors that would otherwise escape immunosurveillance.

Here, we report that systemic administration of IL-15/IL-15Rα complexes relieves tumor-resident CD8⁺ T cells from control by the tumor microenvironment, allowing these cells to expand and destroy advanced solid tumors without the need for vaccination or ACT. Unexpectedly, we found that tumor destruction occurred independently of circulating CD122⁺ lymphocytes or tumor-resident NK cells despite their marked expansion in lymphoid organs and peripheral blood. Our results have significant implications for immunotherapeutic intervention of human cancers and reveal a novel mechanism for reinvigorating the cytotoxic potential of TILs that persist within solid tumors.

Mice. The RIP1-Tag2 transgenic mouse line was previously described (37). RIP1-Tag2 mice were obtained from the Mouse Model of Human Cancer Consortium (National Cancer Institute) and were maintained on a C57BL/6 background. Perforin knockout mice, strain C57BL/6-Prf1^tm1Sdz/J, were obtained from The Jackson Laboratory, were maintained on a C57BL/6 background, and were crossed to the RIP-Tag2 mice. C57BL/6 mice were obtained from The Jackson Laboratory. All mice were bred and maintained under barrier conditions in the Dana-Farber Cancer Institute Animal Facility in accordance with NIH guidelines.

Tumor cell lines. B16F10 melanoma cells were maintained in DMEM with high glucose (Invitrogen Life Technologies) supplemented with 10% fetal bovine serum (FBS), 1× penicillin/streptomycin. Cells were harvested using 0.25% trypsin/EDTA (Invitrogen Life Technologies) when 50% to 90% confluent. MIN6 insulinoma cells were maintained in DMEM with high glucose (Invitrogen Life Technologies) supplemented with 10% FBS, 1× penicillin/streptomycin, and 0.1% β-mercaptoethanol (38). Cells were harvested using 0.25% trypsin/EDTA (Invitrogen Life Technologies) when 50% confluent and were labeled with 10 μmol/L of 5,6-carboxyfluorescein diacetate succinimidyl diester (CFSE; Molecular Probes) for 10 min at 37°C. Cells were then cultured for 4 d with graded doses of IL-15/IL-15Rα complexes (per 1 mL): 0 = medium only, 0.1× = (0.1 μg of IL-15 + 0.466 μg of IL-15Rα), 1× = (1 μg of IL-15 + 4.66 μg of IL-15Rα), 10× = (10 μg of IL-15 + 46.6 μg of IL-15Rα), and 100× = (100 μg of IL-15 + 466 μg of IL-15Rα). Cells were harvested and stained with 7-amino-actinomycin D (BD PharMingen) to evaluate their viability. Cells were analyzed immediately, without fixation, using a FACSCalibur (BD Biosciences). Data were analyzed using FlowJo Software (Tree Star).

Treatment with IL-15/IL15-Rα complexes and assessment of tumor burden. Recombinant murine IL-15 (eBioscience) and recombinant soluble murine IL-15Rα-Fc (R&D Systems) were suspended in 0.1% bovine serum albumin (BSA)/PBS, mixed, and incubated for 30 min at 37°C before injection. Unless specifically noted, RIP1-Tag2 mice at 10 to 11 wk of age received one injection of IL-15/IL-15Rα complexes, IL-15 alone, or PBS per day on 2 consecutive days. For the IL-15/IL-15Rα complexes, 2 μg of IL-15 were precomplexed with 12 μg of IL-15Rα-Fc in 300 μL PBS and injected i.v. For IL-15 alone, 2 μg of IL-15 were injected i.v. in 300 μL PBS. Mice were sacrificed 4 d after the first injection of IL-15/IL-15Rα (schematized in Fig. 1). For long-term treatments, RIP1-Tag2 mice at 10 to 11 wk of age received one injection of the IL-15/IL-15Rα complexes per day for 2 consecutive days i.v. followed by a total of 13 i.p. injections every 3 d. To measure pancreatic tumor burden, pancreata were excised from euthanized subjects and solid tumors were dissected away from surrounding exocrine tissue. Tumor diameters (d) were then measured and tumors were classified in four categories: A, d < 1 mm; B, 1 ≤ d ≤ 2 mm; C, d = 2 mm; and D, d > 2 mm. Pancreatic tumor burden was calculated as [(A × 1) + (B × 2) + (C × 3) + (D × 4)]. For B16 transplantation, 5 × 10⁵ cells were injected s.c. in 200 μL of PBS into the abdomen of 6- to 12-wk-old C57BL/6 mice. Ten days later, or when the tumors reached 2 to 8 mm in diameter, mice were treated with one injection of the IL-15/IL-15Rα complex per day on 2 consecutive days. Each injection was performed i.v. and consisted of 2 μg of IL-15 precomplexed with 12 μg of IL-15Rα in 300 μL PBS. Mice were sacrificed 4 d after the first injection of IL-15/IL-15Rα. B16 tumor burden was assessed by measuring the diameter (d) of each tumor using a caliper and percent tumor growth was calculated as (d_day4 − d_day0) / d_day0.

Phenotypic characterization of host leukocytes by cytofluorimetry. Spleens were processed into single-cell suspensions using glass slide disruption followed by RBC lysis. Solid tumors were removed from the pancreas and processed into single-cell suspension by gentle mechanical disruption with forceps followed by enzymatic digestion using the following digestion medium: 0.2 mg/mL collagenase P (Roche), 0.8 mg/mL dispase (Invitrogen), and 0.1 mg/mL DNase I (Sigma) in RPMI. Tumor suspensions were incubated in digestion medium at 37°C for 25 min. Released cells were collected and the remaining tumor tissue was subjected to further processing by incubation in fresh digestion medium for an additional 25 min at 37°C. After digestion, single-cell suspensions were filtered with 80-μm nylon mesh; RBCs were lysed; and suspensions were filtered again. Blood was collected into tubes containing 10 mmol/L EDTA PBS to prevent coagulation and RBCs were subsequently lysed. For staining, cells were suspended in PBS/5% FBS/2 mmol/L EDTA (FACS buffer) and Fc receptors were saturated using FcR-block antibody (2.4G2). Single-cell suspensions from spleen, blood, and tumors were stained with fluorescently labeled monoclonal antibodies specific for NK1.1 (PK136), CD3ε (145-2C11), CD45 (30-F11), CD44 (1M7), CD69 (H.1.2F3; Biolegend), CD8α (53-6.7; BD PharMingen), and CD122 (IL-2/IL-15Rβ; TM-B1; Caltag) at 4°C for 30 min; washed in FACS buffer; and analyzed immediately, without fixation, using a FACSCalibur (BD Biosciences). Data were analyzed using FlowJo Software (Tree Star).

For intracellular IFNγ analysis, total spleen or tumor cells were plated at 2 × 10⁶/mL in complete medium [DMEM supplemented with l-glutamine (2 mmol/L), 1× penicillin/streptomycin, and 10% FBS] and activated with phorbol 12-myristate 13-acetate (PMA; 5 ng/mL) and ionomycin (500 ng/mL) for 5 h, and with monensin in the last 3 h. After cell surface staining with anti–CD8-PerCP and anti–CD45-FITC antibodies, cells were fixed with 4% paraformaldehyde. Cells were then permeabilized with 0.1% saponin and stained with anti–IFNγ-APC or isotype control antibody. Cells were washed again and analyzed by flow cytometry. For granzyme B detection, intracellular staining was done the same way using anti-granzyme B–phycoerythrin antibody but without PMA and ionomycin stimulation.

Fluorescence microscopy. Pancreas or pancreatic tumors were fixed in 4% PFA/10% sucrose/PBS for 2 h and then incubated in 30% sucrose/PBS for 1 h. Tissues were snap-frozen in optimum cutting temperature (OCT) medium (Sakura Finetek). Sections (6 μm) were fixed in cold acetone for 10 min and blocked with 10% goat serum in PBS for 1 h. To visualize tumor-associated leukocytes, tissue sections were incubated with biotinylated anti-CD45 (Biolegend) or anti-CD8 (Caltag) for 30 min at room temperature, washed, and then incubated with streptavidin-Cy3 (Invitrogen) for 30 min at room temperature. All dilutions were made in 2% goat serum/PBS and sections were washed thrice with 2% goat serum/PBS after incubation with primary and secondary reagents. To visualize apoptotic cells within tumors, we used the In Situ Cell Death Detection Kit, Fluorescein [terminal deoxyribonucleotide transferase–mediated nick-end labeling (TUNEL); Roche] to label DNA strand breaks. The permeabilization step was performed at room temperature instead of 4°C with a modified permeabilization solution of 0.25% Triton-X 100/0.1% sodium citrate. Sections were then dried, mounted on glass coverslips with anti-fade mounting medium containing 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes), and examined on a fluorescence microscope. Images were captured using the Spot RT-Slider Digital Imaging System (Diagnostic Instruments).

Leukocyte depletion. Cells expressing CD8⁺ and NK1.1⁺ were depleted using monoclonal antibodies specific for CD8α (53-6.7) or NK1.1 (PK136), respectively. For depleting NK1.1⁺ cells, RIP1-Tag2 mice received one i.v. injection of anti-NK1.1 (500 μg) 24 h before treatment with IL-15/IL-15Rα complexes and two subsequent antibody injections of 75 μg each on days 0 and 1 of the regimen schematized in Fig. 1. For depletion of CD8⁺ cells, RIP1-Tag2 mice received one i.v. injection of anti-CD8 (100 μg) 24 h before treatment with IL-15/IL-15Rα complexes and two subsequent antibody injections of 50 μg each on days 0 and 1 of the regimen. At the end of the experiment, leukocyte depletion was verified by cytofluorimetry.

Adoptive transfer of RIP1-Tag2 leukocytes. Spleen and lymph nodes were removed from 10.5-wk-old CD45.1⁺ or CD45.2⁺ RIP1-Tag2 mice and processed into single-cell suspension using glass slide disruption followed by RBC lysis. Cells were then incubated in RPMI containing 0.1% BSA for 10 min at 37°C with 10 μmol/L of CFSE. The labeling reaction was stopped with RPMI/20% FBS on ice and the cells were washed twice in PBS/5% FBS/2 mmol/L EDTA. Ten million cells were injected i.v. in 200 μL of PBS into 10.5-wk-old CD45.2⁺ RIP1-Tag2 mice.

Assessment of host leukocyte proliferation. Cellular proliferation of host leukocytes was measured using (+)-5-bromo-2′-deoxyuridine (BrdUrd) incorporation. Mice were maintained on BrdUrd (Sigma) in drinking water (0.8 mg/mL) with 1% glucose throughout the experiment starting 1 d before IL-15/IL-15Rα complex treatment. Mice were injected on days 0 and 1 with 2 μg of IL-15 precomplexed with 12 μg of IL-15Rα in 300 μL PBS i.v. and sacrificed between 1 and 3 d later. Spleen, lymph nodes, and tumors were harvested and processed into single-cell suspension to evaluate BrdUrd incorporation within total leukocytes (CD45⁺) and CD8⁺ populations by cytofluorimetry using phycoerythrin-conjugated anti-BrdUrd antibody (BD PharMingen).

In vitro cytotoxicity assay. Cytolytic activity was determined by 5-h ⁵¹Cr release assay, as previously described (39). Briefly, target cells, RMA, and RMA pulsed with 1 μmol/L SIINFEKL (OVA peptide), were labeled for 1 h with ⁵¹Cr. Effector cells were harvested on day 4 from spleens of mice injected with IL-15/IL-15Rα complexes on days 0 and 1. For OT-I effector cells, spleens from OT-I^tg/RAG1^−/− TCR transgenic mice (40) were harvested and the percentage of OT-I cells was identified based on cytofluorimetric analysis of CD8 and Vα2 expression. Target cells were incubated with effector cells in 96-well round-bottomed plates for 5 h at 37°C with 5% CO₂ at the indicated ratios. The percentage of specific ⁵¹Cr release was calculated as follows: percentage of specific lysis = [(experimental release − spontaneous release) / (maximum release − spontaneous release)] × 100.

Histologic assessment. Tissue samples were collected, directly fixed in Bouin's solution, paraffin-embedded, and processed for H&E staining. Histologic analysis was performed in a blinded fashion.

Statistical analysis. All data are presented as mean ± SE. Statistical analysis used the two-tailed unpaired Student's t test for comparison of two experimental groups, and log-rank test for comparison of survival in Kaplan-Meier survival plot. P values <0.05 were considered significant.

In vivo delivery of IL-15/IL-15Rα complexes rapidly induces a significant reduction in solid tumor burden. The biological activity of IL-15 can be significantly increased when administered as a complex with soluble IL-15Rα. The remarkable potency of IL-15/IL-15Rα complexes could be beneficial in numerous clinical settings in which CD122⁺ cells (including NK cells, NKT cells, and CD8⁺ T cells) serve as critical effectors in immune surveillance of solid tumors and chronic infections; however, this has not been formally tested to date. Stoklasek and colleagues recently reported that systemic administration of IL-15/IL-15Rα complexes could prevent transplanted B16 melanoma cells from forming tumors, whereas IL-15 alone had no effect on tumor engraftment (36), providing evidence that prophylactic administration of IL-15/IL-15Rα complexes can prevent the generation of tumors. Whether this agent can affect immunosurveillance of established solid tumors had not been previously addressed. Thus, we tested the effect of IL-15/IL-15Rα complexes on established solid tumors in two different models: in one case, where tumors arise either from transplanted melanoma cells or in another where tumors arise spontaneously in the endocrine pancreas of transgenic animals. To this end, mice bearing palpable s.c. B16 tumors were treated with one i.v. injection of IL-15/IL-15Rα complexes per day for 2 days and tumor progression was evaluated 3 days later (Fig. 1A). Systemic administration of IL-15/IL-15Rα complexes had a marked effect on the overall tumor burden, impairing the growth of established B16 tumors in C57BL/6 mice by 75% compared with controls (30% versus 140%, P < 0.004; Fig. 1A).

Having confirmed that in vivo delivery of IL-15/IL-15Rα complexes could affect B16 tumors in a therapeutic setting, we next sought to determine whether this regimen would have any effect on solid tumors that arise spontaneously in a vital organ. To this end, we used the RIP1-Tag2 transgenic mouse model in which the SV40 T antigen (Tag) is expressed under the control of the rat insulin promoter (RIP), causing oncogenic transformation of the majority of pancreatic β cells (37). Tumor development in these mice unfolds in several distinct stages (Fig. 1B). At 3 to 4 weeks of age, β cells start proliferating in response to Tag expression (hyperplastic stage); at 7 to 8 weeks of age, new blood vessels form along with alterations in the microvasculature (angiogenic switch); and by 10 weeks of age, solid tumors will have developed in 100% of the mice. Evidence for tumor-specific T-cell responses in pancreatic lymph nodes at early stages of tumorigenesis suggest that hyperplastic islets are subject to immunosurveillance; however, such mechanisms fail to prevent tumor growth (41) for reasons that remain unclear. To examine the therapeutic efficacy of IL-15/IL-15Rα complexes on established solid tumors of the pancreas, we treated 10- to 11-week-old RIP1-Tag2 mice with the complexes or IL-15 alone once per day for 2 days and analyzed 3 days later (Fig. 1B). In vivo delivery of IL-15/IL-15Rα complexes caused a rapid and significant reduction in the number and size of tumors (Fig. 1B,, bottom left), diminishing pancreatic tumor burden by >50% (P < 0.0001) in RIP1-Tag2 mice (Fig. 1B,, bottom right). In contrast, however, IL-15 at the same dose used in the complexes had no significant effect on tumor burden in RIP1-Tag2 mice (Fig. 1B , bottom right). Importantly, the body weight of treated subjects was identical to control littermates throughout the regimen (Supplementary Fig. S1A) and no observable toxicity or autoimmunity to normal tissues was found in mice treated with IL-15 alone (data not shown) or IL-15/IL-15Rα complexes (Supplementary Fig. S1B and data not shown). Together, these data indicate that systemic administration of IL-15/IL-15Rα complexes inhibits the growth of established B16 tumors and causes significant regression of naturally arising solid tumors of the pancreas.

Having shown the rapid therapeutic effect of IL-15/IL-15Rα complexes on tumor burden, we then tested the effect of this agent on the survival of RIP1-Tag2 mice. Subjecting RIP1-Tag2 mice harboring solid tumors to long-term treatment with IL-15/IL-15Rα complexes resulted in a significant prolongation of survival (P < 0.05) compared with control-treated animals (Fig. 1C). Taken together, these results show that treatment with IL-15/IL-15Rα complexes seems to be well tolerated and can have a significant clinical benefit for animals bearing advanced solid tumors.

We then wanted to determine whether the reduction in pancreatic tumor burden was due to specific destruction of malignant β cells or of other cells in the tumor stroma. To evaluate this, tumors were harvested at various time points after injection of IL-15/IL-15Rα complexes and processed for TUNEL analysis by immunofluorescence microscopy. Within 48 hours of administering IL-15/IL-15Rα complexes, the number of TUNEL⁺ cells increased ≥7-fold (Fig. 1D), indicating that apoptosis was occurring in the tumors as a consequence of the treatment. To pinpoint the identity of the dying cells, we stained cryosections of the β-cell tumors with an insulin-specific antibody. The TUNEL⁺ cells were clearly costained with an insulin-specific monoclonal antibody (Fig. 1D , bottom right), whereas insulin⁻ cells were consistently TUNEL⁻, indicating that the apoptotic cells were malignant β cells rather than tumor stromal cells. Given how rapidly this novel regimen causes tumor cell apoptosis, we tested the direct effect of IL-15/IL-15Rα complexes on the viability and proliferation of malignant β cells in vitro, and neither of these were directly affected by IL-15/IL-15Rα complexes, even at high concentrations (Supplementary Fig. S2A). In support of these functional data, CD122 surface expression was undetectable on all tumor cells tested, whereas tumor-associated CD8⁺ T cells were CD122⁺ (Supplementary Fig. S2B). Thus, we conclude that tumor destruction following systemic administration of IL-15/IL-15Rα complexes does not result from a direct effect on tumor cells.

IL-15/IL-15Rα complexes expand CD8⁺ T cells and NK1.1⁺ cells in lymphoid organs and tumors of RIP1-Tag2 mice. Treatment of tumor-free C57BL/6 mice with IL-15/IL-15Rα complexes is known to expand CD122⁺ cells in the spleen, including NK cells, NKT cells, and MP CD8⁺ T cells (35, 36). To ascertain whether CD122⁺ cells in mice harboring solid pancreatic tumors respond in a similar manner, we treated 10- to 11-week-old RIP1-Tag2 mice with IL-15/IL-15Rα complexes and then analyzed the abundance and activation state of CD8⁺ T cells and NK1.1⁺ cells in various tissues. In agreement with previous reports (35), we found that the size and weight of the spleen increased significantly as a result of the treatment (Supplementary Fig. S3). Cytofluorimetric analysis indicated that the relative abundance of NK1.1⁺ splenocytes increased from 3% in PBS-treated controls to 11% in IL-15/IL-15Rα complex–treated mice (Fig. 2A,, left) with equal proportions of NK (CD3⁻) and NKT (CD3⁺) cells (Supplementary Fig. S4A). An expansion of the CD8⁺ T-cell compartment was also observed, increasing from 11% in controls to 18% in IL-15/IL-15Rα complex–treated mice (Fig. 2A , left). A comparable expansion of NK1.1⁺ cells and CD8⁺ T cells was also observed in peripheral blood (data not shown) and tumor-draining lymph nodes (data not shown), whereas surface expression of prototypical activation markers, such as CD44 and CD69, by circulating NK1.1⁺ cells and CD8⁺ T cells was only minimally affected (data not shown).

Next, we evaluated the effect of systemically administered IL-15/IL-15Rα complexes on tumor-associated leukocytes. For these studies, we established a sequential digestion method that allows for maximal release of single cells from large solid tumors that are surgically excised from the surrounding pancreatic tissue. Applying this method in combination with cytofluorimetry, we detected a population of CD45⁺ cells that comprise ∼20% of the total tumor cell suspension in untreated mice (Fig. 2B,, left). This observation was somewhat surprising because it had been previously reported that the advanced-stage tumors in RIP1-Tag2 mice were largely devoid of leukocytes (41). To confirm that the leukocytes detected by cytofluorimetry were bona fide intratumoral cells rather than contaminating peritumoral or extratumoral leukocytes that were physically associated with the excised tumor, we examined the localization of CD45⁺ cells in pancreata of RIP-Tag2 mice using immunofluorescence microscopy. This analysis showed that the CD45⁺ cells detected by flow cytometry were indeed derived from an intratumoral leukocyte population that was evenly distributed throughout the tumor parenchyma (Fig. 2B,, right). Interestingly, systemic administration of IL-15/IL-15Rα complexes caused the intratumoral leukocytes to expand to 30% (Fig. 2B,, left) of the tumor, and CD45⁺ cells were localized within the tumor bed of treated mice similar to control tumors (Fig. 2B , right).

Detailed analysis of the intratumoral leukocyte composition indicated that a variety of leukocyte subsets are represented in the tumor under steady-state conditions, including CD8⁺ T cells, NK cells, and NKT cells expressing CD122 (data not shown), as well as CD4⁺ T cells and Tregs. Treatment with IL-15/IL-15Rα complexes caused a 7- to 8-fold increase in intratumoral NK1.1⁺ cells (Fig. 2A,, right), which are largely NK cells (88% NK1.1⁺CD3⁻; Supplementary Fig. S4B). Similarly, intratumoral CD8⁺ T cells increased 4- to 5-fold upon treatment with IL-15/IL-15Rα complexes (Fig. 2A , right). Surface expression of CD44 and CD69 by tumor-associated leukocytes was only minimally affected (data not shown) by the administration of IL-15/IL-15Rα complexes. The percentage of tumor-associated CD4⁺ T cells and Tregs was similar between PBS and complex-treated RIP1-Tag2 mice (data not shown). In sum, these results show that systemic delivery of IL-15/IL-15Rα complexes increases the frequency of CD8⁺ T cells and NK1.1⁺ cells in lymphoid tissues and solid tumors of RIP1-Tag2 mice, but does not markedly alter their activation state.

Tumor destruction upon exposure to IL-15/IL-15Rα complexes is mediated by CD8⁺ T cells, not by NK or NKT cells. Having shown that NK1.1⁺ cells and CD8⁺ T cells were both responsive to IL-15/IL-15Rα complexes in RIP1-Tag2 mice bearing advanced-stage solid tumors, we asked whether one or both of these cell populations was responsible for the tumor cell destruction. To address this, RIP-Tag2 mice were injected with monoclonal antibodies that specifically deplete NK1.1⁺ cells or CD8⁺ cells or with isotype controls before the injection of IL-15/IL-15Rα complexes. Systemic depletion of NK1.1⁺ cells had no effect on the efficacy of IL-15/IL-15Rα complex therapy (Fig. 2C), whereas depletion of CD8⁺ T cells completely abrogated the therapeutic efficacy of IL-15/IL-15Rα complexes (Fig. 2D). We conclude that CD8⁺ T cells, but not NK or NKT cells, are responsible for the diminished tumor burden in IL-15/IL-15Rα–treated RIP1-Tag2 mice.

Tumor-resident, not circulating, CD8⁺ T cells mediate tumor destruction. Next, we sought to define the steps leading to tumor cell destruction by CD8⁺ T cells following in vivo delivery of IL-15/IL-15Rα complexes. We reasoned that these lymphocytes could promote tumor regression by two different mechanisms that were not mutually exclusive. One possibility was that CD8⁺ T cells within secondary lymphoid tissues or blood would undergo expansion and activation upon exposure to IL-15/IL-15Rα complexes, traffic via blood to tumors, infiltrate the tumor parenchyma, and finally kill malignant cells. A second possible mechanism was that tumor-resident CD8⁺ T cells would expand in the tumor itself upon signaling by IL-15/IL-15Rα complexes and then destroy neighboring tumor cells. To explore these possibilities, we first determined whether CD8⁺ T cells within solid tumors of RIP1-Tag2 mice divide upon exposure to IL-15/IL-15Rα complexes using BrdUrd incorporation. Intratumoral CD8⁺ T cells incorporated BrdUrd within 24 hours of treatment, whereas CD8⁺ T cells in lymphoid tissues did not respond until day 2 (Fig. 3A). Furthermore, the amount of BrdUrd incorporated by the intratumoral CD8⁺ T cells was higher than CD8⁺ T cells within lymphoid organs (Fig. 3B). These observations are consistent with the fact that 80% of the tumor-resident CD8⁺ T cells express CD122⁺, whereas only 24% of spleen CD8⁺ T cells are CD122⁺ in 10-week-old RIP1-Tag2 mice (Fig. 3C). Thus, tumor-resident CD8⁺ T cells rapidly proliferate in response to systemically delivered IL-15/IL-15Rα complexes.

Previous reports established that solid pancreatic tumors in RIP1-Tag5 mice become impenetrable by circulating leukocytes after the angiogenic switch because of alterations in the local vasculature (6, 42–44). However, it has remained unclear whether altered tumor vessels (45) prevent circulating leukocytes from infiltrating advanced solid tumors in the RIP1-Tag2 line. If such barriers indeed develop during tumorigenesis in the RIP1-Tag2 mouse model, then it would be imperative to evaluate whether such barriers persist after treatment with IL-15/IL-15Rα complexes so as to determine whether newly infiltrating or tumor-resident lymphocytes mediate the tumor destruction. In agreement with our cytofluorimetric enumeration of leukocyte expansion in IL-15/IL-15Rα complex–treated RIP1-Tag2 mice, histologic analysis of liver, kidney, and exocrine pancreas clearly showed an increase in leukocyte frequency in blood vessels of these tissues (Fig. 4A). In striking contrast, tumor vessels seemed to lack leukocytes under both conditions (Fig. 4B), suggesting that circulating leukocytes are excluded from solid tumors in RIP1-Tag2 mice, even after treatment with IL-15/IL-15Rα complexes. These data are further corroborated by our observation that CD8⁺ T cells are present in the tumor parenchyma (Fig. 4C) of control and IL-15/IL-15Rα complex–treated animals rather than in tumor blood vessels as previously reported for adoptively transferred cells (46).

To test this notion more stringently, we directly monitored the trafficking of circulating leukocytes in control and IL-15/IL-15Rα complex–treated RIP1-Tag2 mice. We transferred CFSE-labeled splenocytes from CD45.1⁺ congenic RIP1-Tag2 donors into CD45.2⁺ RIP1-Tag2 mice. At various times after transfer, lymphoid organs, liver, and tumors were harvested from control and IL-15/IL-15Rα complex–treated recipients and processed for cytofluorimetric enumeration of the donor cells. Donor cells were readily detected in spleen, lymph nodes, and liver of both IL-15/IL-15Rα complex–treated RIP1-Tag2 mice and PBS-treated control littermates (Fig. 5A). In contrast, the tumors in both conditions contained few if any CD45.1⁺ donor cells (Fig. 5A) within the first 72 hours after transfer. Consistent with our BrdUrd studies, we found that the donor cells started dividing, as measured by CFSE dilution, 2 to 3 days after transfer in peripheral lymphoid organs and liver of IL-15/IL-15Rα complex–treated mice (Fig. 5B). Tumor-infiltrating leukocytes were not detected until 3 days posttransfer and seemed to be progeny of donor cells that had divided elsewhere. Interestingly, these cells arrive after tumor burden is already significantly reduced in IL-15/IL-15Rα complexes (Fig. 5C), thus implicating tumor-resident rather than circulating lymphocytes in this process. These results establish two new points. First, advanced solid tumors in RIP1-Tag2 mice are not readily accessible to any circulating leukocytes. Second, systemic treatment of 10- to 11-week-old RIP1-Tag2 mice with IL-15/IL-15Rα complexes does not markedly increase leukocyte infiltration of solid tumors. These results suggest that tumor-resident CD8⁺ T cells play a major role in the tumor destruction that ensues after in vivo delivery of IL-15/IL-15Rα complexes.

IL-15/IL-15Rα complexes endow tumor-resident CD8⁺ T cells with tumoricidal potential. Despite the fact that endogenous CD8⁺ T cells with a memory phenotype are present within advanced RIP1-Tag2 tumors and that tumor-specific memory CD8⁺ T cells have been observed in these animals, such lymphocytes could be functionally impaired, unable to recognize antigen or ignorant under steady-state conditions because tumorigenesis is unchanged in T cell–deficient RIP1-Tag2 mice (47). However, systemic administration of IL-15/IL-15Rα complexes leads to the rapid expansion of tumor-resident CD8⁺ T cells that in turn destroy the surrounding tumor. To gain a better understanding of how this treatment leads to tumor cell destruction by CD8⁺ T cells, we initially examined the localization of CD8⁺ T cells within the tumors at early time points. Within 48 hours of treatment, we found that the apoptotic tumor cells were often in close proximity to or even in direct contact with tumor-resident CD8⁺ T cells, suggesting that the lymphocytes were directly lysing the malignant β cells (Fig. 6A). Consistent with this notion, we found that the cytolytic activity of splenic CD8⁺ OT-I T cells toward OVA peptide-pulsed targets was potentiated by IL-15/IL-15Rα complexes whereas control-treated CD8⁺ T cells exhibited no cytolytic potential (Fig. 6B). Thus, signaling by the IL-15/IL-15Rα complexes seemed to activate the killing potential of CD8⁺ T cells.

Next, we examined tumor-resident CD8⁺ T cells for their expression of molecules that have been implicated in tumor cell killing, including IFNγ, granzyme B, and perforin. We found that treatment with IL-15/IL-15Rα complexes lead to an increase in IFNγ and granzyme B expression by tumor-resident and splenic CD8⁺ T cells (Fig. 6C); however, tumor-resident CD8⁺ T cells expressed higher levels of both effector molecules under steady-state conditions and in treated animals (Fig. 6C). Because our attempts to examine perforin expression by intracellular flow cytometry were met with technical difficulties, we were unable to quantify changes in perforin protein after treatment. Given the effect of the IL-15/IL-15Rα complexes on expression of granzyme B, we generated perforin-deficient RIP1-Tag2 mice to evaluate the contribution of the granzyme B/perforin pathway in the IL-15/IL-15Rα complex–stimulated tumor destruction. As shown in Fig. 6D, the reduction in solid tumor burden in IL-15/IL-15Rα complex–treated mice was abrogated in the knockout animals, indicating that tumor destruction is dependent on the granzyme B/perforin pathway. In sum, we conclude that in vivo delivery of IL-15/IL-15Rα complexes endows tumor-resident CD8⁺ T cells with the capacity to rapidly expand and directly kill their tumor cell neighbors.

The potential use of IL-15 as a cancer immunotherapeutic agent has been investigated in several mouse models of cancer (29–31). Transfer of B16 cells into IL-15 transgenic mice showed that ubiquitous cytokine expression can act prophylactically to stimulate potent antitumor immunity and prevent tumor engraftment (36). Other studies investigating the curative potential of this agent showed that IL-15 could enhance the effect of chemotherapeutic agents and ACT on transplanted tumors, but had little or no effect when administered alone (29–32). IL-15 complexed with IL-15Rα exhibits remarkable potency compared with free IL-15 (35, 36) and has been shown to prevent engraftment of B16 tumors (36). We report here that two injections of IL-15/IL-15Rα complexes leads to CD8⁺ T cell–mediated regression of autochthonous solid tumors of the pancreas and also severely retards growth of established B16 tumors without significant toxicities. Remarkably, in vivo delivery of IL-15/IL-15Rα complexes promotes immune-mediated destruction of established tumors by endogenous, tumor-resident CD8⁺ T cells. These results are particularly striking as the therapeutic effect of IL-15/IL-15Rα occurs at a relatively low dose without the addition of chemotherapeutic agents, vaccination, ACT, or other cytokines. Although IL-15 lacks the adverse effects of IL-2 such as Treg expansion and AICD, it is of great interest to compare the effect of IL-15/IL-15Rα complexes to that of IL-2/anti–IL-2 complexes in vivo (48).

The superagonist activity of IL-15/IL-15Rα complexes is most pronounced for CD8⁺ memory-phenotype cells (35). Our results show that IL-15/IL-15Rα complexes stimulate the expansion of CD122⁺ cells and enhanced the effector functions of NK and CD8⁺ T cells. Of particular interest was the ability of these complexes to arm a small population of naturally occurring intratumoral MP CD8⁺ T cells with the capacity to proliferate and kill neighboring cancer cells in an environment that otherwise seems to be immunosuppressive. The existence of such cells in solid tumors of the pancreas in RIP1-Tag2 transgenic mice was unexpected as insulinomas are often described as lacking leukocytic infiltrates (6, 41, 42, 45, 49). In retrospect, it is understandable how a population of long-lived CD8⁺ T cells might have been overlooked because these cells are sparsely distributed throughout the tumor. Furthermore, processing of tumors by multiple rounds of enzymatic digestion allowed us to systematically analyze the cellular composition of these malignant lesions and the response of a population of MP CD8⁺ T cells to IL-15/IL-15Rα complexes within the tumor. Speiser et al. (41) also provided functional evidence for a small pool of tumor-specific, memory CD8⁺ T cells in RIP-Tag2 mice that rapidly expanded upon vaccination with tumor antigen and acquired sufficient cytolytic potential to kill tumor cells. Thus, treatment with IL-15/IL-15Rα complexes enables MP CD8⁺ T cells that persist in tumors to rapidly divide and kill neighboring tumor cells.

We and others have shown that circulating leukocytes cannot access solid tumors of the endocrine pancreas because of an atypical vascular network (45, 49). Like RIP1-Tag2 tumors, solid tumors in RIP1-Tag5 mice develop vascular alterations that mitigate leukocyte adhesion (6, 42–45). These observations suggest that immunotherapies that aim to target established tumors with circulating leukocytes, including vaccination and ACT, will not be fully effective in the face of such vascular barriers (50). Identifying agents that promote extravasation of blood leukocytes into tumors is of considerable interest. For example, treatment with either CpG-ODN or irradiation were recently found to promote infiltration of RIP1-Tag5 tumors by adoptively transferred tumor-specific T lymphocytes (6, 42). These vascular barriers may also be the cause of incomplete tumor destruction in both short- and long-term treatments. At any given time, RIP1-Tag2 mice contain tumors of various sizes and composition. Although IL-15/IL-15Rα complexes will lead to expansion and activation of CD122⁺ CD8⁺ T cells in large tumors, smaller tumors that contain fewer or lack CD122⁺ CD8⁺ T cells may be refractory to this treatment. On the other hand, tumor-resident CD8⁺ T cells may become functionally exhausted after treatment because they cannot be reinforced by circulating CD8⁺ T cells due to vascular barriers. Although IL-15/IL-15Rα complex treatment prolonged the survival of both RIP1-Tag2 and B16 tumors (51) in a CD8⁺ T cell– and NK cell–dependent manner, respectively, reversal of vascular barriers by other biologics may be required for efficient CD8⁺ T-cell infiltration and complete cure. Alternatively, as we show here, treatment with agents that sidestep vascular barriers by endowing intratumoral lymphocytes with tumoricidal function may prove to be highly efficacious in reducing tumor burden while minimizing the risk of systemic inflammation or toxicity. According to a recent study addressing the mechanism by which adoptively transferred CD8⁺ T cells penetrate and destroy transplanted EL4 tumors, circulating CD8⁺ T cells entered tumors from peritumoral vessels and began killing malignant cells shortly after extravasation (46). Migration of cytotoxic lymphocytes from the tumor margin to the tumor core occurred progressively as a result of tumor cell elimination (46). In contrast, we found that MP CD8⁺ T cells of host origin are distributed evenly throughout solid tumors of the pancreas and after exposure to IL-15/IL-15Rα complexes begin killing malignant cells from their long-term posts within the tumor core. Thus, the strategy by which adoptively transferred T cells infiltrate and destroy tumors seems to differ rather significantly from that used by naturally occurring effector cells that persist within solid tumors to kill malignant cells following IL-15/IL-15Rα treatment. Despite the marked increase in NK cell abundance in IL-15/IL-15Rα–treated RIP1-Tag2 mice, NK1.1⁺ cells do not contribute to the tumor destruction observed in our study. The cellular mechanism of immune-mediated tumor destruction is dictated in part by the status of MHC class I expression on the tumor cell targets (38). RIP1-Tag2 tumors are known to maintain expression of MHC class I molecules (41); therefore, they are likely to inactivate local NK cells by triggering inhibitory receptors and be targeted by CD8⁺ T cells.

Complexing free IL-15 with its soluble receptor IL-15Rα allows a novel mechanism for boosting the biological activity of this cytokine in various clinical settings. Here, we report that systemic administration of IL-15/IL-15Rα complexes can circumvent tumor immune evasion by endowing tumor-resident CD8⁺ T cells with the capacity to act as a “trojan horse” and destroy tumors from within. This is a particularly important mechanism as many tumors like the RIP1-Tag2 tumors develop barriers for new immune cell invasion. Given that immune effector cells have been detected in various types of solid tumors (16), it will be of considerable interest to test the efficacy of IL-15/IL-15Rα complexes in other tumor models and in human cancers.

Grant support: American Cancer Society (S.J. Turley), Richard and Susan Smith Family Foundation (S.J. Turley), V Foundation (A.W. Goldrath), Cancer Research Institute (A.W. Goldrath), Arthritis Foundation (J.A. Hamerman), Institut National de la Recherche Agronomique (M. Epardaud), and Prevent Cancer Foundation (M.P. Rubinstein).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Competing interests statement: The authors declare that they have no competing financial interests.

We thank K. Wucherpfennig, G. Dranoff, R. Johnson, and A. Doedens for helpful discussions of this work and L. Lanier (University of California San Francisco) and M. Kronenberg (La Jolla Institute of Allergy and Immunology) for the gifts of PK136.