IL-15/IL-15Rα/CD80-expressing AML cell vaccines eradicate minimal residual disease in leukemic mice

Older individuals with acute myelogenous leukemia (AML) have poor outcomes because of more frequent high-risk features and comorbidities.¹  The improved survival achieved with allogeneic hematopoietic stem cell transplantation because of curative graft-versus-leukemia responses conferred by donor T cells provides evidence for the potential efficacy of immunotherapies (reviewed by Dombret and Gardin¹ ). Because many older individuals are ineligible for transplants, there is an unmet need for novel therapeutic approaches.

Although immunotherapies for AML have been explored, to date none have reliably reduced relapse rates.^2-5  In this context, autologous cell vaccines may have advantages for stimulating antileukemic immunity because responses are directed to multiple leukemia-associated antigens, some of which are patient specific. Immune responses generated against an autologous AML vaccine would obviate the problem of lack of a priori knowledge of the dominant antigens present in each patient’s leukemia. Previous trials with autologous cell vaccines have induced antileukemic immunity, but responses were variable.⁶  This is in part because AML blasts are ineffective in T-cell stimulation, because of their downregulation or absent expression of specific costimulators such as CD80,^7,8  and because of their immune evasive effects including upregulation of checkpoint molecules and stimulation of inhibitory immune effectors (reviewed by Teague and Kline⁹ ). In older patients, immunotherapeutic efficacy may be further limited by a decline in T-cell responsiveness.^10-13 

Because engineering patient AML cells to express the missing costimulatory protein CD80 has shown promise,^7,8,14,15  we engineered AML vaccines to express a novel combination of CD80 and the heterodimeric complex interleukin-15 (IL-15) and IL-15 receptor α (IL-15Rα) to improve the induction of antileukemic cytolytic responses. The IL-15/IL-15Rα heterodimer is the naturally occurring form of the cytokine and is a member of the γc cytokine family that engages a heterodimeric receptor comprising the IL-2Rβ/IL-15Rβ subunit (shared with the IL-2 receptor) and γc.¹⁶  Efficient IL-15 production requires coexpression of IL-15 and IL-15Rα in the same cell,^17-22  which substantially increases IL-15 half-life and activity through the IL-2/IL-15R γc receptors.^23-25 

IL-15 has major advantages as an immune-stimulatory cytokine because in contrast to the effects of IL-2, previously used in immunotherapy, IL-15 reverses CD8⁺ T-cell unresponsiveness to tumor-associated antigens, renders T effector cells resistant to suppressive regulatory T cells (Treg’s), and participates in antiapoptotic signaling to effector T cells (reviewed by Waldmann ²³ ). IL-15 also stimulates more effective induction of antigen-specific cytotoxic lymphocytes and more durable immunity through actions on memory T cells, and it has important roles in natural killer (NK) and NK T-cell activation, proliferation, and survival.²³  Although systemic IL-15 administration has less toxicity than does high-dose IL-2 infusion, it does cause neutropenia, fever, and other side effects.²⁶  Thus, local expression of IL-15 by genetically modified AML vaccines has the potential advantages of IL-15 immune stimulation, with reduced risk of systemic toxicities. Finally, local expression of membrane-bound and secreted heterodimeric IL-15/IL-15Rα together with costimulation by CD80 may mimic the interactions of professional antigen-presenting cells with lymphocytes, required for triggering effective cell-mediated immune responses.

In these studies, the leukemia-specific cytolytic activity stimulated by coexpression of IL-15/IL-15Rα and CD80 in lentivirally transduced irradiated AML cell vaccines was evaluated in 32Dp210 myeloid leukemia-bearing mice.^27-30  We also developed a novel model of postremission minimal residual disease (MRD) in leukemic mice that in part recapitulates the clinical setting in which AML cell vaccines would be administered. Our studies show that in mice with prior high leukemic burdens, postremission therapy with 32Dp210-IL-15/IL-15Rα/CD80 vaccines can produce long-term survival and eradication of leukemia.

32Dp210 leukemia cells transformed by the p210 bcr abl transcript³¹  were provided by R. Arlinghaus (University of Texas MD Anderson Cancer Center). For longitudinal in vivo bioluminescence studies, 32Dp210 leukemia cells were transfected with a luciferase vector (Cat.VSL-0074; Capital Bioscience, Gaithersburg, MD). Other 32Dp210 cells were transduced with enhanced green fluorescent protein for flow cytometric analysis, or with the herpes simplex thymidine kinase (HSV-TK) suicide gene, to confer sensitivity to ganciclovir (GCV) and tumor-specific lethality after drug administration. For each experiment, 32Dp210 aliquots were thawed; expanded in RPMI, 10% fetal bovine serum (Gibco, Life Technology), and 1% glutamine; and then washed in phosphate-buffered saline and injected IV via lateral tail vein. 32Dp210-Luc⁺, 32Dp210-GFP⁺, and 32Dp210-Luc-HSV-TK⁺ cells all showed comparable growth in vivo.

Antibodies to H-2D^K (15-5-5), H-2K^k (36-7-5), H-2Kb (AF6-88.5), CD3 (2C11), CD4 (RM4-5), CD8 (53-6.7), CD25 (PC61), CD335 (29A-1.4), CD11b (M1/70), Ly-6G and Ly-6C (Gr-1), CD80 (16-10A1), IL-15 (16-7154-85), IL-15Rα (DNT15Ra), CD279 (J43), CD44 (IM7), CD62L (MEL-14), FoxP3 (FM23), interferon γ (IFN-γ; 14-4-4s) and Fc Block (2.4G2), PD-L1 (BV711 [Rat anti-mouse CD274, clone MIH5], matched BV711-conjugated isotype control [mAb; rat IgG 2a]), and PD-1 (BV421 [Hamster anti-mouse CD279, clone J43]) were purchased from BD Biosciences (San Jose, CA). Cultured and tissue-derived cells were suspended in phosphate-buffered saline, labeled with Fc block (BD Biosciences), and stained with the indicated antibodies. Intracellular staining for FoxP3 and IFN-γ was performed using Cytofix/Cytoperm-Fixation/Permeabilization Kit (Becton Dickinson), and data were analyzed using FlowJo software (Treestar, Ashland, OR).

Lentiviral vectors were constructed by homologous recombination in the pMLV-mIL-2-Furin-CD80 backbone, containing the murine leukemia virus promoter, murine IL-2, and murine CD80.³²  The murine IL-2-Furin-CD80 cassette in this vector was replaced either with mouse CD80 alone or with codon-optimized mouse IL-15 with the leader sequence from granulocyte-macrophage colony-stimulating factor to increase expression, and a RNA/codon-optimized, membrane associated, complete IL-15Rα.^17,33  The IL-15 cassette was linked by a P2A sequence, to generate a self-cleaving peptide³⁴  producing mouse IL-15-P2A-IL-15Rα. A tri-cistronic lentiviral vector was constructed by linking the murine IL-15-P2A-IL-15Rα to murine CD80 with an F2A sequence (coding for a second self-cleaving peptide)³⁵  3′ to the IL-15/IL-15Rα cassette generating IL-15-P2A-IL-15Rα-F2A-CD80 (supplemental Figure 1). Lentivirus production and transduction was performed as described.³⁶ 

32Dp210 cells were transduced with 8 µg/mL polybrene in 6-well plates. To achieve comparable levels of expression of mouse IL-15, IL-15Rα, and CD80 in 32Dp210 vaccines, transduced cells were first purified based on high-level and comparable CD80 expression. Then, purified 32Dp210-IL-15/IL-15Rα⁺ and 32Dp210-IL-15/IL-15Rα/CD80 cells with comparable expression of IL-15Rα were isolated and expanded.

Female C3H mice (8-10 weeks; Charles River, San Diego, CA) were housed under pathogen-free conditions at University of California, San Francisco. Experiments were conducted with supervision of University of California, San Francisco Institutional Animal Care and Use Committee according to approved protocol (AN108913-02A).

Lentivirally transduced or untransduced 32Dp210-luc parental cells were irradiated with 40 Gy prior to subcutaneous administration. At lower doses of irradiation, occasional viable tumors were observed. Tumor progression and responses to vaccinations were analyzed in a Xenogen in vivo imaging system (IVIS), after intraperitoneal injection with 150 mg/kg luciferin and anesthesia with 3% isofluorane. Images were analyzed using Living Image software (Xenogen, Alameda, CA).

Ten days after IV injection of 1 × 10⁵ 32Dp210-luc, all mice had detectable leukemia by IVIS and began GCV treatment (50 mg/kg/day × 14 days) (APP Pharmaceuticals LLC, Schaumburg, IL). To confirm that GCV had no effects on peripheral blood counts, counts were assessed after administration of 100 mg/kg per day × 14 days (supplemental Table 1).

In vivo cell depletion by intraperitoneal injection of antibodies was carried out on day 0 after inoculation of 1 × 10⁴ 32Dp210 cells. Antibodies (100 µL diluted rabbit anti-mouse anti-asialo-GM1, or 500 µg rat-anti-mouse CD4, or 500 µg rat anti-mouse CD8 [BioXCell, West Lebanon, NH]), produced >90% depletion of target cells 24 hours after injection, verified by staining for CD3, CD4, CD8, and NKp46 in blood and spleen.

The 96-well plates (MAIP S4510; Millipore) were coated with anti-IFN-γ antibody, washed, and blocked per manufacturers’ instructions (Mabtech). Splenocytes (3 × 10⁵) from vaccinated mice were plated in duplicate with equal numbers of irradiated (100 Gy) 32Dp210 cells and incubated at 37°C and 5% CO₂ for 20 hours. Plates were developed by incubation with biotinylated anti-IFN-γ antibody (Mabtech), and activity was detected using a colorimetric AP kit (Bio-Rad, CA) and counted (ELISpot reader, Cellular Technology Ltd.).

Splenocytes from naïve mice were labeled with CellTrace Violet or carboxyfluorescein diacetate succinimidyl ester (CFSE); 2 × 10⁶ cells were cocultured with an equal number of irradiated (100 Gy) 32Dp210-Luc, 32Dp210-Luc–IL-15/IL-15Rα, 32Dp210-Luc-CD80, or 32Dp210-Luc–IL-15/IL-15Rα/CD80 cells for 12 days. Splenocytes were then stained with anti-CD3, anti-CD4, and anti-CD8 antibodies, and proliferation was quantified by flow cytometric analyses.³⁷ 

After 4 weekly vaccinations with 32Dp210, 32Dp210-IL-15/IL-15Rα, 32Dp210-CD80, or 32Dp210-IL-15/IL-15Rα/CD80, harvested splenocytes were restimulated in vitro with irradiated 32Dp210 parental cells for 5 days. These splenocytes were purified over Ficoll-Hypaque gradients and used as effectors. Target 32Dp210 cells were labeled with CellTrace 7-hydroxy-9H(1,3-dichloro-9,9-dimethylacridin-2-one) succinimidyl ester (DDAO-SE). Splenic effector cells were cocultured at the indicated ratios for 24 or 48 hours with (1 × 10⁵) 32Dp210 target cells. In studies to assess whether responses to the human p210 breakpoint cluster region gene–Abelson murine leukemia viral oncogene homolog 1 (BCR-ABL) peptide were dominant, unmodified 32Dp210 cells, or BCR-ABL (GFKQSSKAL) peptide-loaded, or control (HYLSTQSALSK) peptide-loaded C3H splenocytes were used as targets (effector-to-target ratio = 10:1). The fractions of apoptotic 32Dp210 cells were assessed using the Active Caspase-3 apoptosis kit (BD Bioscience) and flow cytometry.

C3H mice were vaccinated weekly 4 times with 32Dp210-derived vaccines as described previously. Splenocytes (1 × 10⁶) from immunized C3H mice were stimulated ex vivo for 20 hours with either (1) irradiated unmodified 32Dp210 (5 × 10⁵), (2) PMA + ionomycin, or (3) C3H splenocytes from naïve mice loaded with 5 μg/mL BCR-ABL–derived peptide or with 5 μg/mL control peptide. Intracellular IFN-γ expression was quantified after staining and fluorescence-activated cell sorter analysis.

Secreted mouse IL-15 levels were measured by enzyme-linked immunosorbent assay (ELISA) using recombinant mouse IL-15 as a standard and polyclonal rabbit anti-mouse IL-15 antibody (H114; Santa Cruz Biotechnology Inc., Santa Cruz, CA).

Experiments were repeated at least 2 times unless indicated otherwise. Statistical analyses were performed using Prism 6 (GraphPad Software Inc., La Jolla, CA). Results are reported as the mean ± standard error of the mean (SEM) in independent experiments. The significance of differences was determined using Student paired t test.

Dose-finding studies with 32Dp210 leukemia demonstrated that the minimal cell dose that produced reliable leukemic engraftment, and adequate survival to allow induction of immunity (∼30-40 days^38-40 ), was 10⁴ cells administered IV (Figure 1A, upper panel). Because leukemia-associated stimulation of inhibitory immune responses poses challenges in the clinical setting,⁹  we assessed whether 32Dp210 leukemia has immunosuppressive effects comparable to those observed in patients. Fourteen days after leukemia inoculation, T cells increased the expression of the negative regulatory receptor PD-1, and 32Dp210 cells upregulated expression of the PD-1 ligand, PD-L1, on engrafted leukemia cells (Figure 1A lower panels). The numbers of CD3⁺CD4⁺ T cells, CD3⁺CD8⁺ T cells, and the frequency of CD3⁺CD8⁺CD44^hi T cells were all decreased in leukemic mice, whereas the frequency of inhibitory CD4⁺FoxP3⁺ Treg’s and CD11b⁺Ly6G⁺ myeloid derived suppressor cells increased, demonstrating 32Dp210-mediated immunosuppression (supplemental Figures 2 and 3). 32Dp210 leukemia also reproduced early homing of blasts to the bone marrow compartment seen clinically (0.2% GFP⁺ cells at day 3) (Figure 1B).

Because coexpression of IL-15 and IL-15Rα markedly increases the secretion, half-life, and bioactivity of IL-15, codon-optimized murine IL-15 and murine IL-15Rα constructs were both incorporated in vectors, with or without CD80 (supplemental Figure 1).^17,33  As expected, IL-15 was barely detected in transduced 293T cell supernatants in the absence of IL-15Rα expression; however, coexpression of IL-15 and IL-15Rα resulted in high levels of IL-15 secretion (Figure 2A). Transduction with the IL-15/IL-15Rα/CD80 vector produced somewhat lower levels of IL-15 secretion both in 293T cells (Figure 2A) and in 32Dp210 cells, an effect of insert size (Figure 2B). Irradiation, required for vaccine preparation, did not adversely affect IL-15 secretion (Figure 2B) or cell surface expression of IL-15Rα, IL-15, or CD80 (Figure 2C-D).

Primary stimulation of splenocytes from naïve mice with irradiated 32Dp210-IL-15/IL-15Rα or 32Dp210-IL-15/IL-15Rα/CD80 vaccines produced a fivefold increase in CD3⁺CD8⁺ T-cell proliferation compared with coculture with untransduced 32Dp210 cells (Figure 3A). Lower levels of proliferation were observed after stimulation with 32Dp210-CD80. In the case of CD3⁺CD4⁺ T cells, proliferation levels stimulated by coculture with either parent 32Dp210 cell vaccine, 32Dp210-CD80, or 32Dp210-IL-15/IL-15Rα were similar; however, the greatest responses were seen after 32Dp210-IL-15/IL-15Rα/CD80 vaccination (Figure 3B). To assess whether the combination of IL-15 and CD80 expression is synergistic in stimulating T cells in the absence of antigen presenting cells, and other coexisting populations in spleen, CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells were purified and then stimulated with irradiated 32Dp210-derived cells. Consistent with our hypothesis, coexpression of IL-15/IL-15Rα and CD80 showed striking synergy in stimulating purified CD3⁺CD8⁺ (Figure 3C) and CD3⁺CD4⁺ T cells (Figure 3D).

The immune-stimulatory effects of 32Dp210-based vaccination were then evaluated in vivo in normal, non-tumor-bearing mice where leukemia-mediated immunosuppression would not confound assessment of maximal induced immune responses. C3H mice were vaccinated with irradiated 32Dp210-CD80, 32Dp210-IL-15/IL-15Rα, or 32Dp210-IL-15/IL-15Rα/CD80 weekly for a month. Controls included age-matched, untreated C3H mice, and mice vaccinated with the irradiated 32Dp210 parent line. Thereafter, splenocytes were harvested, restimulated with 32Dp210 parent cells, and stained for flow cytometric analysis. The highest levels of CD3⁺CD8⁺ T-cell proliferation were observed in splenocytes from mice vaccinated with 32Dp210-IL-15/IL-15Rα and 32Dp210-IL-15/IL-15Rα/CD80 (Figure 3E). Prior stimulation with the 32Dp210-IL-15/IL-15Rα/CD80 vaccine also resulted in enhanced proliferation of CD3⁺CD4⁺ T cells (Figure 3F).

In ex vivo assays, leukemia-specific cytolytic activity stimulated by 32Dp210-derived vaccines against 32Dp210 targets was greatest in splenic T cells from 32Dp210-IL-15/IL-15Rα/CD80 vaccinated mice at all effector to 32Dp210 target cell ratios (Figure 4A). Vaccination with 32Dp210-IL-15/IL-15Rα also stimulated significant responses. ELISpot assays demonstrated increased frequencies of IFN-γ-expressing cells after secondary stimulation of splenocytes, consistent with activation and expansion of cytotoxic T cells (Figure 4B). In these assays, it was necessary to restrict the number of peripheral blood mononuclear cell to 3 × 10⁵, as higher numbers of peripheral blood mononuclear cells produced too many spots for accurate quantification, because of the high magnitude of immune stimulation. Robust T-cell stimulation by vaccines was further supported by demonstration of increased intracellular IFN-γ expression in CD3⁺CD8⁺ and CD3⁺CD4⁺ T cells after 20 hours coculture with unmodified 32Dp210 cells. Again, highest levels of immune stimulation were induced by 32Dp210-IL-15/IL-15Rα/CD80 vaccination (Figure 4C-D).

To determine whether expression of the human BCR-ABL oncogene expressed by 32Dp210 is the dominant antigen driving antileukemic responses, intracellular IFN-γ expression in T cells from vaccinated animals was assayed after stimulation with either (1) 32Dp210 or (2) naïve C3H splenocytes loaded with a BCR-ABL fusion peptide GFKQSSKAL shown to be targeted in other studies,⁴¹  or an irrelevant peptide. Intracellular IFN-γ expression in splenocytes from vaccinated mice after in vitro stimulation with BCR-ABL peptide-loaded cells was comparable to background levels seen after coculture with media alone or control peptide (Figure 4C-D). The absence of a dominant response to the BCR-ABL peptide was confirmed by ELISpot assays where BCR-ABL-loaded cells provided no significant stimulation (Figure 4E). The lack of a major response to BCR-ABL expressed by 32Dp210 supports our hypothesis that the potent antileukemic responses observed are induced by multiple leukemia-associated epitopes on 32Dp210 vaccines. Clinically, the magnitude of the protective response, elicited by a single peptide, is likely to be substantially lower than the aggregate response to a larger panel of patient specific and leukemia associated antigens, which collectively provide greater protective immunity. To further examine the role of BCR-ABL expression in vaccine responses, the ex vivo cytolytic activity of splenocytes against either unmodified 32Dp210, BCR-ABL-loaded, or control peptide-loaded C3H targets was compared. As previously, splenocytes from 32Dp210-IL-15/IL-15Rα/CD80 treated animals showed highest levels of cytolytic activity to 32Dp210 targets, whereas stimulation with either BCR-ABL or control HYLSTQSALSK peptide did not exceed background levels (Figure 4F).

In parallel experiments, serially vaccinated non-tumor-bearing mice underwent 32Dp210 tumor challenge. 32Dp210-IL-15/IL-15Rα/CD80-vaccinated mice had higher survival rates after IV 32Dp210 inoculation than mice treated with 32Dp210-IL-15/IL-15Rα or 32Dp210-CD80 vaccines (Figure 4G). Mice surviving 100 days after vaccination and initial tumor challenge demonstrated long-lasting antileukemic immunity as they were able to reject 32Dp210 leukemia after rechallenge with 32Dp210 leukemia at day 150.

The therapeutic efficacy of 32Dp210-derived vaccines was then tested in mice with established leukemia. Mice received 3 weekly vaccinations, 3 days after injection of 1 × 10⁴ 32Dp210 cells, when leukemia has engrafted (Figure 1B). All unvaccinated leukemic controls died by day 70 (Figure 5A). Vaccination with parent 32Dp210 cells slightly prolonged survival, consistent with the previously described low-level immunogenicity of irradiated tumor cells.^42,43  Treatment with any of the lenti-engineered vaccines produced greater survival after tumor challenge than did treatment with unmodified 32Dp210 cells (Figure 5A), but highest survival rates (80%) were observed after vaccination with 32Dp210-IL-15/IL-15Rα/CD80. These mice also rejected leukemia on rechallenge. Thus, coexpression of CD80, and IL-15/IL-15Rα, induces highly effective antileukemic activity, even in the context of leukemia-associated immunosuppression.

To define immune effectors mediating leukemia-specific responses, in vivo antibody-mediated depletion studies were performed with either anti-CD4 or anti-CD8 or anti-asialo-GM1 (AS-GM1) antibodies. The AS-GM1 antibody, which binds to NK cells and also to some activated T cells^44,45  and to basophils,⁴⁶  was used instead of anti-NK1.1 antibody, as the NK1.1 antigen (CD161b/CD161c) is not expressed in C3H mice.^47,48  While depletion with anti-CD4⁺ or anti-AS-GM1 antibodies had minimal effects on survival outcomes, depletion of CD8⁺ cells completely abrogated the effects of 32Dp210-IL-15/IL-15Rα/CD80 vaccination (Figure 5B).

The clinical application of an autologous AML vaccine strategy would be as postremission immunotherapy. To recapitulate this clinical setting, we developed a murine model of MRD for testing our vaccine. Several chemotherapeutic regimens previously shown to induce transient responses in 32Dp210 leukemia were tested.⁴⁹  However, advanced 32Dp210 leukemia proved to be highly chemorefractory as dose-finding studies with cytosine arabinoside (AraC), and/or high dose dasatinib to target the BCR-ABL tyrosine kinase, were either ineffective or resulted in toxicity (data not shown). We therefore adopted an alternative strategy for achieving postremission MRD in mice with high leukemia burdens by engineering 32Dp210 cells to express the HSV-TK suicide gene.³⁶  Treatment with GCV could induce remission without the confounding immune and cytopenic effects of conventional chemotherapy.⁵⁰  Because effects of GCV are primarily restricted to suppression of marrow proliferation, but not inhibition of immune function, treatment should not alter leukemia-associated immune deviation, except as associated with reduction of tumor burden.⁵¹  Blood samples confirmed that there was no significant effect on counts after 2 weeks of GCV administration (supplemental Table 1).

To validate this model, mice were treated daily with GCV once a significant leukemia burden was evident by IVIS. Although expression of HSV-TK has sometimes been reported to be immunogenic,⁵²  HSV-TK expression did not affect leukemic engraftment/progression (Figure 6). Remission was arbitrarily defined as a level of bioluminescence comparable to background levels in luciferin-injected non-tumor-bearing mice. (Figure 6A). When GCV treatment began on day 14 after tumor inoculation, 7/10 mice showed remission at day 28 (Figure 6A). These mice had comparable leukemic burdens (Figure 6B), and GCV treatment decreased 32Dp210-GFP⁺/HSV-TK⁺ cells to background levels (Figure 6C). All GCV-treated leukemic mice exhibiting background levels of in vivo bioluminescence, and by definition clinical remission, also showed pathological remission (<5% blasts in bone marrow), but persistence of MRD (<0.5% GFP⁺ cells in BM, n = 7/7) (Figure 6C).

The efficacy of postremission vaccination with 32Dp210-IL-15/IL-15Rα/CD80 was then tested in mice with GCV-induced MRD. A 10-fold larger dose of 32Dp210-luc-HSV-TK⁺ cells (1 × 10⁵) was administered to ensure that all mice had evidence of leukemia on day 10 by IVIS, when GCV treatment started (Figure 7A, upper panel). By day 5 of GCV administration (day 15), most leukemic mice responded (Figure 7A, lower panel). All animals that received no further therapy after achieving remission with 14-day GCV, relapsed; however, postremission treatment with 32Dp210-IL-15/IL-15Rα/CD80 was curative, eliminating MRD in 50% of mice, despite the persistence of leukemia-mediated immune suppression (Figure 7B).

Prior AML vaccines have not reliably increased relapse-free survival, because of insufficient induction of antileukemic cytolytic activity; however, the promise of IL-15 in cancer immunotherapy provided a compelling rationale for combining IL-15 and IL-15Rα, with CD80-mediated costimulation to improve vaccine efficacy.

In ex vivo assays, treatment with the 32Dp210-IL-15/IL-15Rα/CD80 vaccine stimulated the highest levels of cytolytic activity when compared with effects of vaccination with 32Dp210-IL-15/IL-15Rα, 32Dp210-CD80, or unmodified 32Dp210 cells. The effects of CD80 costimulation and IL-15/IL-15Rα were additive in this context and correlated with induction of cytotoxic lymphocytes with increased IFN-γ expression. Major responses most likely reflect responses to multiple leukemia-associated antigens and were not targeted to the human BCR-ABL antigen as demonstrated by the lack of responses after secondary stimulation with the BCR-ABL peptide GFKQSSKAL.⁴¹ 

The 32Dp210 leukemia recapitulates many features of human AML including early homing to bone marrow and negative effects on host immunity. Although treatment with all of the engineered vaccines increased survival in leukemic mice, 32Dp210-IL-15/IL-15Rα/CD80 vaccination produced the highest survival rates, consistent with the effects of combined IL-15/IL-15Rα and CD80 expression in stimulating cytolytic activity, also shown ex vivo. Tumor-specific cytotoxic treatment with GCV in HSV-TK-expressing 32Dp210 leukemia induced remission in the majority of mice, as defined pathologically, and by IVIS. However, similar to the clinical setting, MRD persisted after remission induction as evidenced by (1) the presence of MRD (<5% leukemia in bone marrow) in all mice meeting remission criteria by IVIS and (2) the universal relapse in mice that achieved remission with GCV treatment but no further treatment. Postremission 32Dp210-IL-15/IL-15Rα/CD80 vaccination was curative in 50% of GCV-treated subjects, while depletion of CD8⁺ cells, but not CD4⁺ cells or asialo-GM1⁺ NK cells, abrogated vaccine efficacy. Although it is possible that differential levels of in vivo depletion could account for the modest effects of anti-CD4 or anti-AS-GM1 antibodies, dependence on CD8⁺ T cells for immunotherapeutic efficacy is consistent with recent studies with the IL-15/IL-15Rα/Fc superagonist ALT-803 in a myeloma model,⁵³  and with effects of IL-15 in a transgenic adenocarcinoma of the mouse prostate (TRAMP-C2) cancer model.⁵⁴ 

Other studies have also demonstrated the superiority of the IL-15/IL-15Rα heterodimer in stimulating cell-mediated antitumor responses.^53,55-57  Thus, a 10-fold decrease in breast cancer metastases was observed in tumor-bearing mice treated with IL-15/IL-15Rα, compared with outcomes in untreated mice.⁵⁸  Addition of IL-15 and IL-15Rα during ex vivo T-cell expansion for adoptive immunotherapy stimulated T cells responding to lower peptide concentrations that lysed targets at lower effector-to-target ratios, than did cells without IL-15 treatment.⁵⁹ 

Although IL-15 has been shown to play an important role in NK cell proliferation and homeostasis,²³  the role of NK cells in some studies of IL-15-mediated antitumor immunity has been less clear. For example, treatment with IL-15- and IL-12-expressing tumor cell vaccines resulted in similar effects on tumor rejection in normal vs NK-depleted mice.^60,61  In contrast, the efficacy of immunotherapy in murine lymphoma⁶²  and neuroblastoma⁶³  with IL-15 and IL-12 required the presence of both CD8⁺ T cells and NK cells. In our studies, NK depletion by in vivo administration of anti-As-GM antibodies during 32Dp210-derived vaccination did not have a major effect. This could in part be because of the persistence of some tissue-resident NK cells, recently shown to be depleted by administration of anti-NK1.1 antibodies in C57Bl/6 mice, but not by anti-asialo-GM antibodies.⁴⁸  As stated previously, NK1.1 is not expressed in the C3H strain and therefore could not be tested in our model.

The feasibility of using IL-15 to enhance stimulation of autologous antileukemic immunity has been demonstrated with T cells from leukemia patients, previously shown to have an exhausted phenotype.⁶⁴  Patient-derived T cells, collected at presentation, were expanded and then stimulated in vitro with IFN-γ-treated autologous AML cells weekly for 3 weeks, in combination with IL-15 and agonistic anti-CD28 antibodies. AML-reactive, leukemia-specific T cells were generated in 5/8 patients.⁶⁴  The availability of clinical grade IL-15/IL-15Rα, and of an Fc fusion to a mutated form of the heterodimer (ALT803), has allowed the ongoing clinical evaluation of the protein in cancer patients (www.clinicaltrials.gov; #NCT02452268, #NCT01670994). ALT 803 produced clinical responses with CD8⁺ and NK cell activation and expansion, without stimulatory effects on Treg’s, in patients relapsing after allo-transplant (#NCT01885897).⁶⁵  Although significant toxicity was observed with systemic administration of ALT803, these effects were mitigated by subcutaneous administration.

Our studies indicate that AML cell vaccines may enable induction of leukemia-specific immunity by amplification of responses to subdominant leukemia antigens, even when those against the immune dominant antigens are lost. There is now evidence from preclinical studies and recently published phase 1 trials with engineered autologous AML vaccines showing that they (1) are readily generated as cells can be reliably collected, cryopreserved, and engineered^7,66 ; (2) can overcome tumor-related immunosuppression; and (3) stimulate leukemia-specific cytolytic immune responses when administered either after remission/consolidation chemotherapy or in immunosuppressed patients early after hematopoietic stem cell transplantation.^66,67  The potent immune-stimulatory combination of heterodimeric IL-15/IL-15Rα and CD80 expression in autologous leukemia cells provides a promising and universally applicable approach to generation of personalized leukemia vaccine therapy that could improve progression-free survival.

The authors thank William J. Murphy (University of California, Davis), Lewis L. Lanier (University of California, San Francisco), Michael Rosenblum (University of California, San Francisco), Kenneth I. Weinberg (Stanford University), and Donald B. Kohn (University of California, Los Angeles) for helpful discussions, and Ralph B. Arlinghaus (University of Texas MD Anderson Cancer Center) for providing the 32Dp210 cells.

This work was supported by grants from the National Cancer Institute, National Institutes of Health (1 R21 CA177284-01A1) (K.M.L.G.); University of California, San Francisco–King’s Health Partners Faculty Fellowship Travel Grant (K.M.L.G.); University of California, San Francisco/Clinical and Translational Science Institute Pilot Project (K.M.L.G.); and a Leukemia and Lymphoma Society Translational Research Grant (6532-18) (K.M.L.G.).

Contribution: K.M.L.G., F.F., and Y.S. designed the experiments and discussed the results; Y.S., L.D.-V., J.P.F., and C.E.A. performed or participated in experimental work; K.M.L.G. wrote the manuscript; F.F. and G.N.P. edited the manuscript; G.N.P. and C.B. provided IL-15 and IL-15Rα plasmid constructs; and F.F. provided CD80 plasmids.

Conflict-of-interest statement: The authors declare no competing financial interests.

Correspondence: Karin M. L. Gaensler, Department of Medicine, Division of Hematology/Oncology, University of California San Francisco School of Medicine, 502 Parnassus Ave, M1286, Box 1270, San Francisco, CA 94143; e-mail: karin.gaensler@ucsf.edu.