Female B6D2F1 mice, 8–12 weeks old, were purchased from the Ontario Cancer Institute and kept under sterile conditions in a specific pathogen-free animal facility. 10–16 week old mice were used in all experiments. At the end of any experiment (typically 100 days), animals were terminated by CO₂ asphyxiation/cervical dislocation. All experimental procedures were approved by the Animal Care Committee of the Ontario Cancer Institute.

70Z/3-L leukemia cells, derived from B6D2F1 mice, were maintained in complete Opti-MEM: Opti-MEM (Gibco, USA) with 5% heat-inactivated fetal calf serum (FCS) (Gibco, USA), 1xpenicillin/streptomycin (Wisent, Canada), and 5.5 × 10⁻⁵M β-mercaptoethanol, in a humidified atmosphere at 37 °C and 5%CO² [3, 16].

We constructed two novel lentiviruses (LVs) (Fig. 1). LV15sol contains the signal sequence from tissue plasminogen activator, a commonly used signal sequence (s.s.) that designates the fused protein for secretion, fused to the cDNA of mouse IL-15. LV15Rc contains a partial cDNA of mouse IL-15Rα including its signal and pro-peptides as well as the sushi domain required for binding to IL-15 [17] linked to the mature processed mouse IL-15 cDNA. With human cDNA, this complex has been reported to display increased stability, secretion, and bioactivity in comparison to IL-15 alone [18].Fig1Fig. 1

Construction of lentivirus constructs LV15sol and LV15Rc. Diagram of the lentiviruses (LVs) pDy.tpa-mIL15 (LV15sol) and pDY.mIL-15Rc (LV15Rc). Both LVs contain LTR, HIV long terminal repeat; Ψ, human immunodeficiency virus packaging sequence; SD, 5′ splice donor; ΔGAG, truncated group antigen sequence; RRE, Rev. response element; SA, 3′ splice acceptor; cPPT, central polypurine tract; EF-1α, elongation factor-1 alpha promoter; WPRE, woodchuck hepatitis virus post-transcriptional regulatory element; SIN/LTR, self-inactivating HIV long terminal repeat. LV15sol: The signal sequence (s.s.) and pro-peptide of tissue plasminogen activator (tPA) (amino acids 1–35 as predicted by Uniprot bioinformatic analyses) replaced the endogenous signal sequence and pro-peptide (amino acids 1–48 as predicted by Uniprot bioinformatic analyses) of mouse IL-15. A DNA cassette comprising a Kozak consensus sequence and this IL-15sol cDNA was synthesized by Genscript (Piscataway, USA) and subcloned into the lentiviral backbone pDY.cPPT-EF1α.WPRE downstream of the EF-1a promoter. LV15Rc: The first 98 amino acids of mouse IL-15Rα, including its signal peptide and sushi domain as identified by Prosite bioinformatic analysis (SIB Swiss Institute of Bioinformatics), were fused to mouse IL-15 amino acids 49–162 (signal and pro-peptide removed by Uniprot bioinformatic analyses) by a linker (SGGSGGGGSGGGSGGGGSLQ). A DNA cassette comprising a Kozak consensus sequence upstream of this IL-15Rc cDNA was synthesized by Genscript (Piscataway, USA) and subcloned into the 3′SIN, HIV-1-based, lentiviral backbone pDY.cPPT-EF1α WPRE, downstream of the EF-1α promoter. Both vectors were verified by restriction enzyme digestion and DNA sequencing. Lentiviral particles were produced at the University Health Network Vector Production Facility

The parent leukemia cell line was transduced with an approximate multiplicity of infection of 10. After 3 washes cells were seeded into Terasaki plates at a density of 0.3 cells/well to ensure the presence of single-cell wells. Cells were expanded and clones were quantitated for the production of IL-15 at 10⁶ cells/ml/hour. The following ELISA kits were used: mouse IL-15 DuoSet ELISA for IL-15sol (DY447, R&D Systems, USA) and mouse IL-15/IL-15R complex ELISA for IL-15Rc (88–7215, Invitrogen/Thermo Fisher, USA).

70Z/3-L lines were expanded in complete Opti-MEM, washed twice and resuspended at a density of 5 × 10⁶ cells/ml. Each B6D2F1 mouse received 10⁶ cells in a volume of 200 μl PBS injected into the left abdominal cavity using a 1 ml syringe and 26-gauge needle. Mice were then monitored for onset of sickness. The control group usually showed first signs of sickness around day 10. The end point of most experiments was set to day 100.

To evaluate protective immunity, surviving mice were re-challenged with 10⁶ cells of the parent strain after 100 days of the initial challenge.

B6D2F1mice were depleted of T-cell subsets as well as NK-cells using specific depletion antibodies as previously described [3]. Depletion antibodies were injected ip at a dose of 0.5 mg antibody/mouse on days − 2, 3, 6, 10 and 13, and after that once weekly for another month. Cells were injected on “day 0”. Prior to depletion experiments, the efficacy of depletion antibodies was demonstrated in vivo by flow cytometry (data not shown).

70Z/3-L cells were co-cultured with the GFP retroviral packaging line GP + E (ATCC, USA) to render them GFP⁺. After 48 h of co-culture, suspension 70Z/3-L leukemia cells were removed from the adherent packaging line, passed multiple times to ensure no cells of the packaging line were carried over, and expanded. GFP⁺ 70Z/3-L cells were sorted and the top 10% of GFP⁺ expressing cells were collected. GFP expression was stable, as revealed by repeated flow analysis. 10⁶ GFP⁺ leukemia cells were injected ip into mice to monitor cell expansion in vivo. GFP⁺ clones yielded similar survival compared to their non-GFP counterparts (data not shown).

Cells were harvested from the peritoneum by peritoneal lavage with PBS containing 1%FCS. Peritoneal cells were washed, counted, and stained for 30 min with specific antibodies: CD4 (RM4–5; BioLegend), CD8 (53–6.7; BD Biosciences), NK1.1 (PK136; BioLegend), CD44 (IM7; BD Biosciences), CD62L (MEL-14; eBioscience), CD25 (PC61; BioLegend), FOXP3 (MF-14; BioLegend), PD-1 (29F.1A12; BioLegend). For GrzB staining cells were first stained with antibodies directed against antigens on the cell surface, then fixed overnight using the Foxp3 Fixation/Permeabilization Concentrate/Diluent kit (eBioscience, USA). The next morning, cells were permeabilized and stained with GzmB-FITC (GB11; BioLegend) for 45 min. Flow cytometry was performed using the FortessaX20 (Becton Dickinson). Analysis was done using FlowJo software (TreeStar).

Transduced 70Z/3-L clones were evaluated for their production of IL-15sol and IL-15Rc by ELISA. IL-15sol was expressed at levels ranging from approximately 40–500 pg/ml/hr./10⁶cells among various clones (Fig. 2a) which remained stable over time as determined by repeated testing of the clones (data not shown). The range of secretion of IL-15Rc was approximately 100–5000 pg/ml/hr./10⁶cells among clones (Fig. 2b) which also remained stable over time (data not shown).Fig2Fig. 2

Increasing levels of leukemia cell-mediated IL-15 positively correlates with improved survival. a IL-15 levels of LV15sol clones (for both (A) and (B) secretion levels are mean + SEM calculated from 2 to 4 individual ELISAs with duplicate wells). b IL-15 levels of LV15Rc clones; Correlation of survival with IL-15 output: a LV15sol clones; (p <  0.005, LV15sol.1–.7 vs 70Z/3-L, Log-rank, Mantel-Cox test); b LV15Rc clones; (p <  0.0001 for LV15Rc.1–.4, p <  0.003 for LV15Rc.5–.7 vs 70Z/3-L, Log-rank, Mantel-Cox test); Mice were injected ip with 10⁶ cells of either the parent line, or one of the transduced clones and monitored for onset of disease. Based on the secretion levels of IL-15 a theoretical threshold was established (arrow indicates the threshold), below which the protective effect of IL-15 was not observed. Multiple experiments were pooled for survival curves for some clones (n numbers are indicated in brackets). c Side by side comparison of the survival rates of the two IL-15 models using clones LV15sol.1 and LV15Rc.4 in an extended 250-day experiment. P-values: p <  0.001 for both IL-15 groups vs 70Z/3-L controls; p = 0.0003 LV-15Rc vs LV-15sol (Log-rank, Mantel-Cox test); d H³-Thymidin incorporation and e total live cell counts demonstrate that the representative LV15Rc and LV15sol clones used throughout this study grow at a similar rate as the 70Z/3-L parent strain. Results are mean + SEM calculated from 2 to 4 individual experiments with triplicate wells

In order to determine whether the secretion of either IL-15sol or IL-15Rc by the transduced leukemia cells would elicit a protective immune response in the host, a number of clones spanning a range of IL-15 secretion levels were injected into the peritoneal cavity of B6D2F1 mice (10⁶cells/mouse).

We saw a clear trend of improved survival with higher concentrations of IL-15. Clones producing less than 200 pg/ml/hr./10⁶cells of IL-15sol failed to elicit a protective immune response and mice perished around the same time as the control group (Fig. 2a). Clones secreting more than 200 pg/ml/hr./10⁶cells of IL-15sol elicited a protective immune response which led to long-term survival of almost 100% of mice on day 100 (Fig. 2a).

Similarly, clones producing less than 250 pg/ml/hr./10⁶cells of IL-15Rc failed to elicit a protective immune response (Fig. 2b). Clones secreting more than 250 pg/ml/hr./10⁶cells of IL-15Rc elicited an immune response leading to partial protection and a proportion of mice survived up to 100 days (Fig. 2b). However, mice started dying after 50 days post-injection, suggesting incomplete leukemia cell clearance.

To better understand the survival of IL-15Rc leukemia cell-bearing mice a larger experiment with 50 mice in the Rc-cohort was launched using the clones LV15sol.1 and LV15Rc.4, which secrete similar levels of IL-15 as determined by ELISA and shown in Fig. 2. This study demonstrated that over the course of 250 days almost all mice in the Rc cohort perished (Fig. 2c). The experiment was terminated with only 3 mice alive in the IL-15Rc group (3/50 = 6%). Conversely, all mice that had received cancer cells secreting IL-15sol survived the 250 days (Fig. 2c), suggesting a different mechanism of action for IL-15Rc and IL-15sol. It should be noted that clones secreting the highest levels of IL-15Rc fared no better than mid-range level clones, indicating that the IL-15 threshold has been reached by mid-level.

To test for the presence of residual tumor cells in surviving mice on day 100, splenocytes and peritoneal cells were cultured without growth factors, and monitored for the growth of 70Z/3-L cells. Cultures generated from mice injected with IL-15sol-secreting leukemia cells were free of leukemia cells (n = 10 mice). However, 100% of peritoneal cultures and 90% of splenic cultures harvested from mice injected with IL-15Rc-secreting cancer cells were confluent with 70Z/3-L cancer cells after 2–3 weeks (n = 10 mice). These IL-15Rc clones secreted similar levels of IL-15Rc as prior to injection (data not shown). The cells were further re-injected into naïve mice to test whether 100 days in vivo had changed their properties. However, they yielded similar survival patterns to the original clones (data not shown).

One hundred days after the initial ip injection of 10⁶ IL-15-secreting leukemia cells, surviving mice were re-challenged with 10⁶ cells of the parent cell line in order to test whether long-lasting, IL-15-independent, immunity had been established. To test for the efficacy of 70Z/3-L to induce leukemia, a naïve control group was injected with the parent line. All mice surviving the initial injection of cells secreting IL-15sol survived the re-challenge for an additional 100 days (Fig. 3a), suggesting that IL-15sol bestows long-lasting protective immunity. While cell-mediated IL-15sol therapy leads to immunity, IL-15Rc therapy does not (Fig. 3b).Fig3Fig. 3

Leukemia cell-mediated IL-15 therapy leads to long-term immunity and protection against 70Z/3-L for IL-15sol but not IL-15Rc. a, b Mice were initially injected with 10⁶ cells of either a LV15sol clones, or b LV15Rc clones. After 100 days surviving mice were re-challenged with 10⁶ 70Z/3-L cells to test whether immunity was established. A naïve control group received 10⁶ 70Z/3-L cells to control for the efficiency to cause leukemia. a p = 0.0062 for all 3 surviving groups vs 70Z/3-L controls; b p = 0.0246 (LV15Rc.3-primed), p = 0.9876 (LV15Rc.4-primed) vs 70Z/3-L controls; c-e Both T-cell subsets are required for leukemia cell-mediated IL-15 therapy. Mice were depleted of certain cell populations prior to challenging them with either c 10⁶ LV15sol.1 cells; or d 10⁶ LV15Rc.4 cells. In both instances, CD4⁺ and CD8⁺ T-cell subsets were required. e In order to establish whether the same T-cell populations are necessary in the secondary challenge, we immunized 55 mice with the IL-15sol clone LV15sol.1. After 100 days (all 55 mice survived) they were depleted of various cell populations and then re-challenged with the parent strain. Again, both T-cell subsets were required. In all instances, a control group (naïve mice) received 10⁶ 70Z/3-L cells to control for their efficiency to cause leukemia. In the re-challenge experiment one group of not re-challenged (NR) mice was included. Depletion of NK-cells did not have an effect in any of the experiments

Antibody-depletion is commonly used to eliminate various lymphocyte subsets in vivo. It is a useful tool to analyze the roles of different cellular subsets in immune responses and immunological diseases. Both CD4⁺ and CD8⁺ T-cells were essential to establish protective immunity in the IL-15sol (Fig. 3c) as well as the IL-15Rc (Fig. 3d) model, as depleting either T-cell subset ablated the protection. There were no fatalities in any of the isotype or PBS control groups in the IL-15sol model (Fig. 3c). Due to the poor long-term survival of mice receiving IL-15Rc-secreting leukemia cells we saw fatalities in the isotype and PBS groups (Fig. 3d), reflecting the results in Fig. 2c. Depleting NK-cells using the depletion antibody Anti-Asialo GM1 did not lead to deaths in either IL-15 model.

CD4⁺ and CD8⁺ T-cell populations were also necessary for survival after re-challenge and thus IL-15sol-induced long-term protective immunity (Fig. 3e). Due to the poor long-term survival of mice receiving IL-15Rc-secreting leukemia cells a secondary immune challenge was not performed.

The fact that IL-15sol and IL-15Rc-mediated cancer cell therapy revealed dramatically different survival rates prompted us to search for differences between IL-15sol- and IL-15Rc-mediated responses. Since cytokine treatment induces the secretion of other factors, we measured the levels of various cytokines/chemokines in the serum of mice that received IL-15sol vs IL-15Rc-secreting cancer cells. Blood was taken from a group of mice before as well as post-injection on days 8, 18, 30 and 40. 10 μl of serum was diluted with 40 μl of PBS and cytokine levels were measured using a 31-plex analysis (EveTechnologies, Clagary, Canada).

The analysis revealed different serum cytokine/chemokine profiles in host mice injected with leukemia cells secreting IL-15sol vs IL-15Rc. Serum of mice that had received cancer cells secreting IL-15sol showed significantly elevated levels for Eotaxin, G-CSF, IFN-γ, IL-1α, IL-5, IL-6, IP-10, KC, MCP-1 and MIG (Fig. 4a). Interestingly, serum from mice injected with IL-15Rc clones resembled serum cytokine levels of naïve mice, and serum taken from 70Z/3-L-injected mice had slightly elevated serum cytokine/chemokine levels (Fig. 4a). Fig. 4a shows the result of day-8 serum, which represented the peak of the response. To confirm this observation was not unique to the clones used in this experiment, additional clones were tested and showed similar results (data not shown).Fig4Fig. 4

Serum cytokine profile of mice injected with IL-15-secreting leukemia cells. a Serum was taken eight days post ip-administration of leukemia cells. A control group of naïve mice was included. Results shown are mean + SEM calculated from 4 individual mice/group from one representative experiment performed at least twice. Corresponding p values can be found next to the graph. b Only IL-15sol could be detected in mouse serum using the 31-plex analysis (EveTechnologies, Calgary). We repeated the analysis of IL-15 in mouse serum using our ELISA systems to detect IL-15sol c as well as IL-15Rc d. Both c and d show a time course where mice were bled prior to injection and then on days 5, 8, 16 and 30 post-injection of IL-15 secreting leukemia cells

The differences in survival together with our cytokine results suggest a different mechanism of action between the immune responses initiated by IL-15sol vs IL-15Rc. We next turned to flow cytometry to investigate the fate of the leukemia cells in the peritoneum. GFP experiments revealed that IL-15sol-secreting leukemia cells expanded in the host peritoneal cavity approximately 100-fold (from 10⁶ to about 10⁸), peaking around day 7, before they were eliminated by the immune system (Fig. 5a). Interestingly, cancer cells secreting IL-15Rc expanded only 2–5-fold in the host peritoneal cavity (Fig. 5a). Fig. 5a also depicts the parent cancer cell line which expanded to almost 10⁹ cells which ultimately led to the death of the host around day 10. The dramatic difference in cancer cell expansion and clearing between IL-15Rc and IL-15sol may be due to the activation of different cell types.Fig5Fig. 5

Kinetics of expansion of (A) leukemia, (B) NK1.1⁺, (C) CD4⁺, (D) CD8⁺, (E) NK1.1⁺GrzB⁺, (F) CD4⁺GrzB⁺, and (G) CD8⁺GrzB⁺ cells in the peritoneum of mice injected with IL-15 secreting leukemia cells. Every graph shows a time course, and every time point has been obtained by sacrificing 3–6 mice/group. Numbers shown are total cell numbers calculated based on the absolute cell numbers obtained by peritoneal lavage. Naïve mice were included to obtain baseline levels and were included on the graph on the day they were analyzed. Experiments have been performed three times with similar results, and were pooled. For statistical analysis peak values on day 8 were compared within groups using 1-way ANOVA (see Table 2 for statistical analysis). a GFP⁺70Z/3-L parent strain expands significantly more in vivo than LV15Rc or LV-15sol secreting line. b NK1.1⁺ cells expand significantly more in mice injected with LV15Rc.4 compared to all other groups. c CD4⁺ cells expand significantly more in mice injected with LV15sol.1 compared to all other groups. d CD8⁺ cells expand significantly more in mice injected with LV15sol.1 vs LV15Rc.4. e The increase of NK1.1⁺GrzB⁺ cells in the peritoneum of mice injected with LV15Rc.4 was significantly higher compared to LV15sol.1 (p = 0.0147, day 8); f The total numbers of CD4⁺GrzB⁺ cells in the peritoneum of mice injected with LV15sol.1 were significantly higher compared to LV15Rc.4 (p = 0.0070, day 8). g The total numbers of CD8⁺GrzB⁺ cells in the peritoneum of mice injected with LV15sol.1 were significantly higher compared to LV15Rc.4 (p = 0.0171, day 8)

Flow cytometry further revealed that IL-15sol lead to a heightened increase of both CD4⁺ (Fig. 5c) and CD8⁺ T-cell numbers (Fig. 5d) in the peritoneum compared to IL-15Rc, suggesting that IL-15sol is a better activator of T-cells. IL-15Rc on the other hand, is a potent activator of NK1.1⁺ cells, leading to a massive influx and/or expansion of NK1.1⁺ cells in the peritoneal cavity, while NK1.1⁺ cells only increased marginally in the presence of IL-15sol-secreting cells (Fig. 5b). We next tested whether these IL-15Rc-induced NK1.1⁺ cells, stained positive for Granzyme-B (GrzB). Figure 5e shows an approximately 10-fold increase of GrzB⁺NK1.1⁺ cell numbers in the peritoneum of mice injected with IL-15Rc-secreting cancer cells compared to IL-15sol, naïve mice, and mice injected with the parent strain. On the contrary, we saw significantly higher numbers of GrzB⁺CD4⁺ cells (Fig. 5f) and GrzB⁺CD8⁺ cells (Fig. 5g) in the peritoneum of mice injected with leukemia cells secreting IL-15sol compared to IL-15Rc.

Since mice injected with IL-15Rc-secreting tumor cells failed to develop long-term protective immunity, we wanted to test whether their T-cells were exhausted. We investigated whether PD-1 was upregulated on T-cells of IL-15Rc mice 50⁺ days post-injection. We found significantly more CD4⁺ and CD8⁺ T-cells expressing PD-1 in the peritoneal cavity of IL-15Rc mice compared to IL-15sol mice (Fig. 6a).Fig6Fig. 6

T-cells of mice injected with IL-15Rc secreting leukemia cells express significantly higher levels of exhaustion and activation markers 50+ days post-ip-injection. a PD-1; b CD4⁺/CD8⁺; c CD25⁺; d CD44⁺; e GrzB⁺ and f NK1.1⁺GrzB⁺ cells. Each data point represents 1 mouse. Numbers shown are total cell numbers calculated based on the absolute cell numbers obtained by peritoneal lavage. Naïve mice were included to obtain baseline levels and were included on the graph on the day they were analyzed. Experiments have been performed four times with 1–2 mice sacrificed per group, and results were pooled. For statistical analysis mean values were compared within groups using 2-way ANOVA with Tukey’s post-test. P-values are denoted with asterisks (* = p = < 0.05,** = p = < 0.01, *** = p = < 0.001, **** = p = < 0.0001)