C57BL/6 mice were obtained from Charles River Laboratory and Envigo (Frederick, MD). IL-15 knock out (KO) mice were purchased from Taconic. IFN-γ KO were kindly provided by the Cancer Inflammation Program, National Cancer Institute.

Murine carcinoma-38 (MC38) colon carcinoma and Tissue culture-1 (TC-1) lung epithelial cell-derived carcinoma cells were maintained in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (Sigma), penicillin/streptomycin, essential amino acids and 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES). Animals aged 6–8 weeks received 1–2×10⁵ tumor cells by subcutaneous (SC) injection in the flank. Five days after inoculation of tumor cells, tumor-bearing mice were randomized into two groups, untreated and hetIL-15 treated. hetIL-15 was purified from HEK293 cells (Admune Therapeutic LLC/Novartis). For hetIL-15 treatment, mice received intraperitoneal (IP) injection of 3 µg (molar mass of IL-15) of hetIL-1541 51 three times/week for a total of 4–8 injections. Tumor area (length x width) was measured every 2–3 days in a blinded fashion. To block protein secretion from tumor cells in vivo, 100 µg brefeldin (Sigma) at a concentration of 10 mg/mL were delivered intratumorally (IT), 30 min before sacrifice and tumor excision, in some experiments.

Excised tumors were cut into small pieces and digested using the Tumor Dissociation Kit (130-096-730, Miltenyi Biotec), as per manufacturer’s instructions. In some experiments, CD8⁺ T cells were isolated from tumor cell suspensions using Dynabeads FlowComp Mouse CD8 Kit (ThermoFisher Scientific), as per manufacturer’s instructions. Spleens were dissociated using a 100 µm cell strainer and washed to remove any remaining organ stroma. Cells were washed and stained with a fixable viability dye (ThermoFisher Scientific) for 30 min at 4°C. Surface staining was performed using antibodies for the following markers: CD3 (145–2 C11), CD4 (RM4-5), CD8 (53–6.7), CD45 (30-F11), CD49b (DX5), CD64 (X54-5/7.1), CD19 (ID3) from BD Biosciences; CD11c (N418), CD11b (M1/70), MHC-II (M114.15.2), XCR-1 (ZET), CXCR3 (CXCR3-173) from Biolegend and CD103 (2E7) from ThermoFisher Scientific. For intracellular staining, cells were fixed and permeabilized using the Foxp3 staining buffer (ThermoFisher Scientific), following manufacturer’s instructions. Samples were stained with Ki67 (SOLA15) and B-cell lymphoma-2 (Bcl-2) (3F11) from BD Biosciences; Foxp3 (FJK-16s) and IRF8 (V3GYWCH) from ThermoFisher Scientific; GzmB (GB11) and CXCL9 (MIG-2F5.5) from Biolegend; and X-C Motif Chemokine Ligand 1 (XCL1) (BAF486) from R&D Systems followed by PE-Streptavidin (BD Biosciences). Cells were acquired in a Fortessa flow cytometer (BD Biosciences). Data analysis was performed using FlowJo software (Tree Star, Ashland, Oregon, USA). Analyses of tumor immune infiltrates were performed in two to three independent experiments for each type of tumor.

Tumor cell suspensions were seeded in 24-well plates in the presence of beads conjugated with α-CD3/CD28 antibodies (Dynabeads Mouse T Activator CD3/CD28, 11 452D, ThermoFisher Scientific) and incubated at 37°C in the presence of anti-CD107a antibody (1D4B, BD Biosciences) and monensin (GolgiStop, BD Biosciences). Similar cultures without antibody-stimulation were used as negative controls. After 6 hours of incubation, cells were harvested and stained for surface markers and intracellular IFN-γ (XMG1.2, BD Biosciences).

Immunohistochemistry (IHC) was performed as previously reported.52 After deparaffinization, 4.5 µm thick sections were incubated with rabbit anti-CD3 (SP7, Abcam, ab16669, 1:100), followed by treatment with tyramide-fluorophore reagent (PerkinElmer, NEL791001KT; Life Technologies, T20950) at 1:100 dilution in Amplification plus buffer (PerkinElmer, NEL791001KT) for 10 min at room temperature (RT) and washed in Tris-Buffered Saline, 0.1% Tween^®20 Detergent (TBS-T) and H₂0. Similar procedure was performed for the rat anti-CD8 (4SM15, ThermoFisher, 14-0808-82, 1:2000), rat anti-CD4 (4SM95, ThermoFisher, 14–9766, 1:200) and rat anti-Foxp3 antibodies (FJK-16s, ThermoFisher, 145 773–82, 1:100) using an anti-rat secondary HRP (Vector Labs, MP-7444–15). After washes in TBS-T, 4′,6-diamidino-2-phenylindole (DAPI) (Life Technologies, D1306, 1 mg/mL stock, 1:500 in Phosphate buffered saline (PBS)) was added to slides for 10 min at RT. Slides were imaged at both 4 x, and 20 x using Vectra imaging software (PerkinElmer), and the number of cells were enumerated from the top nine hotspots images, to provide equal weighting for each mouse, using inForm analysis software (PerkinElmer).

Excised tumors were resuspended in T-Per Tissue Protein Extraction Reagent (ThermoFisher Scientific) and homogenized using 2.8 mm ceramic (zirconium oxide) beads (Precellys Lysing kit) and Precellys Evolution Homogenizer. Tumor lysates were recovered after spin at 13 000 rpm for 15 min at 4°C and analyzed for cytokine levels using U-PLEX Biomarker group 1 mouse Assay (35 analytes; Meso Scale Diagnostics) and ELISAs specific for the mouse chemokines XCL1 (ThermoFisher Scientific) and CXCL9 (ThermoFisher Scientific), following manufacturers instructions.

Tumors were mechanically disrupted in RLT buffer (QIAGEN) using 2.8 mm ceramic beads (Precellys Lysing kit) and Precellys Evolution Homogenizer. RNA extraction was performed with RNeasy (QIAGEN) including on-column DNase I digestion, according to the manufacturer’s instructions. nCounter PanCancer Immune Profiling Panel (NanoString Technologies) was used to monitor the expression of a panel of 780 genes related to immuno-oncology. The mRNA molecules were counted with the NanoString nCounter at the Laboratory of Molecular Technology (Advanced Technology Program, Frederick National Laboratory). Analysis was performed with a workflow written in R and through a user interface developed on the Foundry Platform (Palantir Technologies). Raw data were filtered to remove low expressed genes, leaving 758 genes out of the original 780. To normalize data, log transformation and quantile normalization were applied. To define differentially expressed genes, we run limma53 and set thresholds at two-fold change and p<0.05 difference between groups (adjusted p value following Bonferroni multiple comparison testing). Gene enrichment analysis on the Gene Ontology (GO) database was subsequently performed on the top 150 genes (sorted by t-statistic) using the Fisher’s exact test (https://github.com/CCBR/l2p). Heatmaps were represented as Z-score centered and rescaled. To determine immune cell scores, the geometric mean of the cell marker genes were calculated from normalized gene expression values within each sample. Cell type gene markers for cytotoxic cells and DCs were determined as in Danaher et al. 54

Differences between untreated and hetIL-15 treated mice were evaluated by a two-tailed Mann-Whitney U test. Tumor areas were plotted as mean±standard error of the mean (SEM) for each data point, and tumor growth curves were compared using mixed-effects analysis of variance (ANOVA). The p values were corrected for multiple comparisons using Holm-Sidak test. Prism V.8.1 software package was used for analysis.

To address the role of endogenous IL-15 in tumor growth, MC38 colon carcinoma cells were administered SC to C57BL/6 wild type (wt) and age-matched IL-15 KO mice, and tumor development was monitored overtime. Tumors showed an accelerated growth in mice lacking IL-15 compared with wt mice (online supplementary figure 1), suggesting that IL-15-dependent immune mechanisms can limit the development of tumors. We next used human cell-produced hetIL-15 treatment as an anticancer immunotherapeutic drug in the MC38 colon carcinoma mouse model. hetIL-15 administration, following a 2-week schedule with three injections/week and a dose of 3 µg/injection by IP delivery, resulted in a significant delay of tumor growth (figure 1A). Analysis of the tumor immune infiltrate from untreated and hetIL-15 (day 8, 1 day after the fourth administration) treated mice was performed by both flow cytometry (figure 1B–G) and IHC (figure 2). hetIL-15 treatment increased the number of intratumoral CD8⁺ T cells by ~3 x (figure 1B). The proliferation rate of intratumoral CD8⁺ T cells in untreated mice was ~50%–60%, suggesting rapid lymphocyte turnover in tumor sites. On hetIL-15 treatment, the percentage of Ki67⁺ CD8⁺ T cells significantly increased to ~80% (figure 1C). Tumor-infiltrating CD8⁺ T cells were also characterized by a significant upregulation in the expression of Bcl-2 in comparison to control mice, indicative of increased survival (figure 1D). Similar results were also observed in the NK cells analyzed in MC38-bearing mice (figure 1E–F). hetIL-15 treatment promoted a ~3 x increased accumulation of NK cells in the tumor sites (figure 1E) and these NK cells showed significantly higher proliferation rate in comparison to control mice (figure 1F).

Favorable outcome in cancer immunotherapy treatments has also been linked to the ratio of CD8⁺ T cells to Tregs.8 For this reason, we determined the CD8⁺/Treg ratio within the tumors from untreated and hetIL-15 treated mice. As previously reported in the B16 melanoma model,47 MC38 tumors were characterized by a significant infiltration of Treg cells with a CD8⁺/Treg ratio of ~2. hetIL-15 promoted a significant accumulation of intratumoral CD8⁺ T cells resulting in a ~5 x increase in the CD8⁺/Treg ratio (figure 1G).

The lymphocyte infiltrate in MC38 tumors was also analyzed by IHC (figure 2). Staining of tumor sections with antibodies against CD3, CD4, CD8 and Foxp3 confirmed that the lymphocytes isolated on in vitro digestion were of intratumoral origin rather than peripherally associated with the excised tumors (figure 2A). Quantification of CD8⁺ T cells/mm² (figure 2B) by IHC showed a significant increase in the hetIL-15 group, with lymphocytes distributed throughout the tumor. In contrast, hetIL-15 treatment did not affect the number of tumor-infiltrating Tregs (figure 2C). As a consequence of CD8⁺ T cell expansion, a significant increase in the CD8⁺ T cells/Treg ratio was also observed by IHC on hetIL-15 administration (figure 2D).

The anticancer efficacy of hetIL-15 was verified in the mouse TC-1 carcinoma model using the same methods (online supplementary figure 2A). hetIL-15 treatment was associated with significantly increased intratumoral CD8⁺ T cells, unchanged frequency of tumor infiltrating Tregs and increased CD8⁺/Treg ratio (online supplementary figure 2B–F).

The systemic effects of hetIL-15 treatment were evaluated in MC38-bearing mice. hetIL-15 induced a significant increase in the absolute number of CD8⁺ T and NK cells in spleen (online supplementary figure 3A–D, respectively) with a significant upregulation in the expression of Ki67⁺ cells (online supplementary figure 3B–E, respectively) and the survival marker Bcl-2 (online supplementary figure 3C).

Taken together, these data show that hetIL-15 sustains proliferation and survival of lymphocytes both in tumor and spleen, and verify the efficacy of hetIL-15 as anticancer immunotherapeutic agent.

We performed gene expression analysis of MC38 tumors using a panel of 780 immune-oncology related gene probes (Nanostring Technology). Tumors were excised 3 hours after the fourth hetIL-15 administration and processed for mRNA isolation. Comparison of gene expression identified several significantly overexpressed genes in tumors recovered from hetIL-15-treated mice (figure 3A). Gzmb and Gzma were the most upregulated genes (~5 x, adjusted p<0.01). Xcl1, Prf1, Ctsw, CD69, Klrc1, Ncr1 and Fasl were also significantly overexpressed in hetIL-15-treated mice (figure 3A). These upregulated genes after hetIL-15 treatment represent an expression signature that corresponds to activated TILs with cytotoxic phenotype. Nanostring analysis also identified additional functional pathway signatures, including signal transducer and activator of transcription intracellular signaling, T-cell receptor (TCR) recognition of cognate antigen, IFNs signaling, increased metabolic rate and immune cell chemotaxis (online supplementary figure 4).

To confirm the transcriptomic data, we analyzed the GzmB content of TILs. Flow cytometric analysis demonstrated that provision of hetIL-15 resulted in higher proportion of tumor-infiltrating CD8⁺ and CD4⁺ T lymphocytes, as well as NK cells harboring GzmB in comparison to untreated mice (figure 3B–D, respectively). Thus, hetIL-15 treatment led to a significant accumulation of GzmB⁺ CD8⁺, CD4⁺ T and NK cells per tumor. Similar results were obtained in the TC-1 tumor model (online supplementary figure 2G). We also analyzed the GzmB content of splenic CD8⁺ T and NK cells. In untreated animals, both CD8⁺ T and NK cells were negative for GzmB, but on hetIL-15 administration, we observed a significant increase in the frequency of lymphocytes harboring GzmB (online supplementary figure 3F and G, respectively). Interestingly, we identified an intratumoral CD8⁺ T cell subset characterized by the expression of GzmB and lack of Ki67 (figure 3E, red) in hetIL-15 treated mice, but these cells were almost completely absent within the spleen. These GzmB⁺Ki67^- CD8⁺ T cells are indicative of a terminally differentiated cytotoxic phenotype and represented ~15% of the total CD8⁺ T cells within the tumors (figure 3E). These data indicate that hetIL-15 treatment preferentially induces intratumoral accumulation of CD8⁺ T cells committed to killing.

We also investigated the production of IFN-γ and the degranulation activity (CD107⁺) by MC38 tumor-derived T cells on ex vivo stimulation with beads coated with α-CD3/CD28 antibodies. We found that ~10% of CD8⁺ (figure 3F) and ~5% of CD4⁺ (figure 3G) tumor-infiltrating T cells were able to secrete IFN-γ in untreated control mice. Significant increase in the frequency of IFN-γ⁺ CD8⁺ (~20%) and CD4⁺ (~10%) T cells was found in tumors from the hetIL-15-treated group, suggesting that hetIL-15 increases the production of IFN-γ within the tumor. Similarly, we found an increased frequency of degranulating CD107a⁺ CD8⁺ T cells in hetIL-15-treated in comparison to control mice (figure 3F).

Overall, hetIL-15 administration shapes the tumor microenvironment towards a Th1 profile, promoting activation and cytotoxic functional commitment in tumor-resident T and NK cells.

Differential gene expression analysis showed that, in addition to pathways related to NK and T cell activation (online supplementary figure 4), hetIL-15 treatment upregulated several intratumoral chemokines stimulating pathways related to myeloid cell migration and chemotaxis (figure 4A, GO:0002548, p=0.039). Gene expression analysis of the whole tumor revealed a ~4 x increase in the expression of XCL1 mRNA on hetIL-15 administration (figure 4B). Analysis of XCL1 protein within the tumor was performed by ELISA at 6 hours after hetIL-15 IP administration. A ~3 x increase in XCL1 concentration was detected in tumor extracts from hetIL-15 treated mice compared with tumors from PBS treated mice (figure 4C). To identify the source of XCL1, cell suspensions from MC38 tumors were analyzed for the chemokine production in vivo. On day 8 after initiation of treatment, mice received an IT injection of brefeldin to block protein secretion in vivo. Tumors were excised and isolated cells were analyzed by flow cytometry using a panel of antibodies for the detection of NK and T cells as well as XCL1. We found that both tumor-infiltrating NK and CD8⁺ T cells produce XCL1 (figure 4D). hetIL-15 treatment resulted in a significant increase of XCL1 production by both cell subsets, as suggested by increased XCL1 mean fluorescence intensity.

XCL1 has been linked to the tumor recruitment of cDC1.11 16 55–57 These cells are known to support antitumor responses through the attraction, stimulation and expansion of tumor-specific CD8⁺ T cells.58 For this reason, we evaluated the myeloid cell infiltrate in MC38 tumors on hetIL-15 treatment. A previously established flow cytometry protocol allows the distinction between cDC1, defined as CD45⁺MHCII^highCD64^-CD11c⁺CD11b^dim, and other myeloid populations including CD64⁺ macrophages and CD11b^high conventional type 2 DC (cDC2).13 59 Additionally, expression of CD103, XCR1 and IRF8 on cDC1 has been linked to their ability to cross-present antigens and stimulate efficient CD8⁺ T cell responses.14–16 55 The strategy outlined in figure 5A was used to identify cDC1 expressing XCR1, CD103 and IRF8 within CD45⁺ immune cells infiltrating the tumors. cDC1 constitute approximately 0.1% of intratumoral CD45⁺ leucocytes in MC38 tumors, in agreement with a previous report.10 The expression of CD64 on these cells was ~4 x reduced in comparison to MHCII⁺CD11b⁺ myeloid cells (online supplementary figure 5A). Tumor-infiltrating B cells represented between 0.05% and 0.6% of CD45⁺MHCII⁺ and did not stain positive for CD11c, XCR1 and IRF8 (online supplementary figure 5B–D).

Treatment with hetIL-15 resulted in significantly increased accumulation of cDC1 expressing CD103⁺XCR1⁺IRF8⁺ in MC38 tumors (figure 5B, left panel). Expression of the XCL1 receptor (XCR1) on this cell subset suggests its ability to migrate toward the XCL1 gradient within the tumors stimulated by hetIL-15 treatment. Analysis of tumor-infiltrating cDC1 was also performed in control and hetIL-15-treated mice bearing TC-1 carcinoma. Similarly, hetIL-15 treatment was associated with ~2 x increase in the number of cDC1 expressing CD103, XCR1 and IRF8 localized within the tumors (figure 5B, right panel). These results are in agreement with the gene expression data revealed by transcriptomics (Nanostring Technology). Based on the differential expression of cell type specific genes, cell scores were calculated for different immune cell subsets as described in Material and Methods. hetIL-15 treatment induced changes in the composition of the tumor immune infiltrate (figure 5C). hetIL-15 therapy was associated with a significant upregulation of the gene expression profile of both cytotoxic cells (figure 5D) and DCs (figure 5E). In contrast to the results obtained within the tumors, we observed a reduction in the frequency of XCR1⁺CD103⁺IRF8⁺ cDC1 in the spleens of hetIL-15 treated animals compared with control mice (online supplementary figure 3H).

Taken together, these data indicate that hetIL-15 treatment promotes the production of the chemokine XCL1 by intratumoral lymphocytes leading to increased tumor infiltration by CD103⁺XCR1⁺IRF8⁺ cDC1.

The intratumoral accumulation of cDC1 and cytotoxic lymphocytes promoted by hetIL-15 was accompanied by increased IFN-γ production by T cells (figure 3) resulting in higher IFN-γ concentration as measured in tumor lysates by Meso Scale Discovery (MSD) quantitation (figure 6A). Differential gene expression analysis showed an increase in IFN-γ mRNA (figure 6B) and IFN-γ responsive genes and confirmed the enrichment of this pathway within the tumor in hetIL-15 treated animals (GO:0034341, p=0.049; figure 6B). The IFN-γ-dependent chemokines CXCL9 and CXCL10 were also assessed in tumor lysates obtained from MC38 bearing-wt and IFN-γ KO mice. Before hetIL-15 treatment, CXCL9 tumor concentration was slightly lower in IFN-γ KO mice compared with wt mice, although this difference did not reach statistical significance by one-way ANOVA (figure 6C, left panel). CXCL10 levels were similar in both strains of mice (figure 6C, right panel). hetIL-15 administration resulted in a ~3 x increase in the intratumoral concentration of both CXCL9 and CXCL10 only in wt mice, with no changes being observed in IFN-γ KO animals (figure 6C). To dissect the mechanism of intratumoral chemokine production in response to hetIL-15, MC38-bearing wt and IFN-γ KO mice were treated with brefeldin delivered IT to block protein secretion in vivo. CXCL9 production in the excised tumors was analyzed by flow cytometry using a mixture of antibodies targeting both T and myeloid cells. We found that all tumor-associated myeloid cell subsets analyzed (macrophages, monocyte-derived DC and cDC1) produce CXCL9 in vivo (figure 6D). hetIL-15 administration increased significantly the frequency of myeloid cells staining positive for CXCL9 in wt mice (figure 6D). cDC1 were one of the sources of CXCL9 (~8% and~2% in wt and IFN-γ KO mice, respectively). The frequency of cDC1 cells producing CXCL9 increased to ~25% in wt mice, but no changes were observed in IFN-γ KO mice (figure 6E), indicating that IL-15-stimulated CXCL9 production is IFN-γ-dependent. To confirm these data, single cell suspensions of MC38 tumors were cultured for 6 hours in the presence of either hetIL-15 or recombinant IFN-γ. Flow cytometry analysis showed that cDC1 produced CXCL9 in response to hetIL-15 only from wt mice(figure 6F). As expected, stimulation with recombinant IFN-γ restored the ability of cDC1 from IFN-γ KO mice to produce CXCL9 to similar levels observed in wt mice (figure 6F). These results are in agreement with the measurements of intratumoral chemokine levels and demonstrate that hetIL-15 stimulation of CXCL9 and CXCL10 production is, at least in part, IFN-γ dependent.

CXCL9 and CXCL10 are key regulators of leucocyte trafficking.60 Effector CD8⁺ T cells and a subset of NK cells expresses CXCR3, the receptor for CXCL9, CXCL10 and CXCL11, and they migrate following gradients of these chemokines.13 23 61 We observed an increased presence of CD8⁺ T lymphocytes and NK cells expressing CXCR3 in blood of hetIL-15 treated wt mice (figure 6G). Evaluation of CXCR3 expression in TILs was not possible, due to the use of collagenase necessary for tumor processing. Overall, these data support a mechanism of hetIL-15 action in which lymphocyte trafficking to the tumor is controlled by a gradient of the chemokines CXCL9 and CXCL10. Myeloid cells, including macrophages, monocyte-derived DC and cDC1 are among the sources of these chemoattractants in the tumor in an IFN-γ dependent fashion. Hence, the antitumor effects of hetIL-15 treatment are linked to the increase of intratumoral lymphocyte infiltration by an IFN-γ-dependent mechanism.