Murine N-809 and N-803 (previously ALT-803) were generously provided by ImmunityBio under a Cooperative Research and Development Agreement with the National Cancer Institute. Murine N-809 consists of N-803, an IL-15 N72D/IL-15RαSushi superagonist complex containing the Fc-domain of murine IgG2a fused to two anti-PD-L1 ScFv domains. αPD-L1 (10F.9G2) and CD8 (2.43) antibodies were from BioXcell. The NK depletion antibody (anti-asialo-GM1) was from Wako Chemicals. FTY720 was from Sigma.

Female Balb/c and C57BL/6 mice aged 6–10 weeks were obtained from NCI Frederick Cancer Research Facility and maintained under specific pathogen-free conditions in accordance with Association for Assessment and Accreditation of Laboratory Animal Care guidelines.

4T1 triple negative (TN) breast carcinoma and Yac-1 lymphoma cells were obtained from American Type Culture Collection. MC38 colon carcinoma cells are previously described.22 All cell lines were cultured according to providers’ instructions, determined as mycoplasma free by MycoAlert Mycoplasma Detection Kit (Lonza) and used at low passage number.

For antitumor studies, 4T1 tumor cells (5×10⁴, subcutaneously) were orthotopically implanted into the mammary fat pad of female Balb/c mice. MC38 (3×10⁵, subcutaneously) tumor cells were implanted into the right flank of female C57BL/6 mice. Tumors were measured biweekly using digital calipers, and volumes were determined as (length²×width)/2. Mice were randomized based on tumor size and treatment initiated when tumors reached 50–100 mm³. Unless otherwise stated, mice received two doses of N-809 (subcutaneously) or two doses of N-803 (0.3 µg, subcutaneously) plus αPD-L1 (200 µg, intraperitoneally). Quantification of 4T1 lung metastasis was performed as previously described.23

For all N-809 studies, unless otherwise stated, immune cells in the lymph nodes, spleen and tumors were isolated 2 days after the final treatment as previously described.16 Cell counts were performed using 123count eBeads (Thermo Fisher Scientific).

Antibody labeling of cells for flow cytometry (1–10×10⁶ immune cells) was performed using the BD Cytofix/Cytoperm Kit (BD Biosciences) according to the manufacturer’s instructions. Antibodies (online supplementary table S1) and matched isotypes were obtained from the listed manufacturers. Live/Dead Fixable Dead Cell Stain was from Invitrogen. Flow cytometry (≥1×10⁵ events) was performed on a BD LSRFortessa flow cytometer (Beckton Dickinson) and analyzed with FlowJo FACS Analysis Software V.9.9.6 (Treestar). Cell populations were identified as listed (online supplementary table S2). Expression of phenotypic proteins was determined by subtracting the respective isotype, set between 1% and 5% of the population.

Isolated immune cells were stimulated with αCD3 (2C11, BD Biosciences) +α CD28 (37.51, BD Biosciences) as previously described.16 Frequency of interferon gamma (IFNγ)⁺ and/or tumor necrosis factor alpha (TNFα)⁺ cells were calculated by subtracting the non-stimulated controls.

NK cell killing of Yac-1 targets was determined as previously described.16 24

Tumor fragments were preserved in RNAlater (Thermo Fisher Scientific) and stored at −80°C. RNA was extracted using the RNeasy Mini Plus Kit (Qiagen) following the manufacturer’s protocol. RNA purity was assessed on the Nanodrop One Spectrophotometer (Thermo Fisher Scientific) and Agilent Bioanalyzer (Agilent). RNA analysis was performed using the PanCancer Mouse IO 360 Panel and data analyzed using the nSolver Software and nCounter Advanced Analysis Software (NanoString). Heatmaps were generated using the Morpheus Software (Broad Institute) for the fold change of a given treatment over phosphate-buffered saline (PBS) calculated by NanoString analyses.

Statistical analyses were performed in Prism V.7.0a or V.8.2 (GraphPad Software). Unless otherwise stated, data presented in bar graphs or scatter plots were analyzed using one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons. Two-way ordinary ANOVA was used to analyze tumor growth curves. Survival was analyzed using log-rank (Mantel-Cox) test. Outliers were identified using Robust regression and Outlier removal (ROUT test). Statistical significance was set at p<0.05.

Additional materials and methods are described in online supplementary additional files.

First, we validated PD-L1 binding ability and IL-15 activity of N-809. In vitro, N-809 bound to murine PD-L1 in a cell-free system (figure 1A) and on murine MC38 colon tumor cells (figure 1B) with similar affinity to a commercially available αPD-L1. In vivo administration of N-809 reduced detectable surface expression of PD-L1 on CD45⁺ cells in MC38 primary tumor, draining and non-draining lymph nodes (LN) and spleen (figure 1C).

IL-15 is a NK cell prosurvival and proliferation-inducing factor.25 To determine IL-15 activity of N-809 in vitro, splenic NK cells were incubated with N-809 or N-803 and viability/proliferation was determined. Untreated NK cells did not survive (figure 1D). N-809 significantly enhanced NK cell viability/proliferation, although 41-fold less than N-803 (figure 1D). When bound to murine PD-L1, N-809 retained 75% of its IL-15 activity (figure 1E). N-809 also demonstrated IL-15-related activity in vivo, as shown by minor changes to CD4⁺ T-cell number (figure 1F) and IL-15Rβ expression (figure 1G) and major dose-dependent alterations to NK and CD8⁺ T cells. CD4⁺ T cells are known to be less responsive than NK and CD8⁺ T cells to IL-15.25 N-809 greatly enhanced NK cell number (figure 1H) and expression of IL-15Rβ (figure 1I) and Ki67 (figure 1J). Similar effects were observed on CD8⁺ T cells (figure 1K–M). Together, these data demonstrate that the αPD-L1 and IL-15 components of N-809 are functional in vitro and in vivo.

Clinically, αPD-L1 is administered intravenously9 while experimentally given intravenously or intraperitoneally.15 16 N-803 is given subcutaneously both in the clinic and experimentally.14–16 26 Since N-809 consists of αPD-L1 and N-803, we determined whether route of administration impacted antitumor efficacy. Subcutaneous and intraperitoneal administration of N-809 to 4T1 tumor-bearing mice significantly reduced lung metastasis burden by 36% and 35%, respectively (online supplementary figure S1a-c). As route of administration did not affect N-809 efficacy and N-803 subcutaneously is better tolerated clinically,14 26 N-809 was administered subcutaneously.

We also determined the optimal dose of N-809 for antitumor efficacy in the highly aggressive 4T1 TN breast and MC38 colon carcinoma models. N-809 did not greatly affect primary 4T1 tumor growth at any dose (online supplementary figure S1a,d), but 50 µg N-809 induced the greatest reduction in 4T1 lung tumor burden compared with 25 and 100 µg N-809 (online supplementary figure S1b,c,e,f). In contrast, 100 µg N-809 best decreased tumor burden and increased median overall survival (mOS) in MC38 tumor-bearing mice (online supplementary figure S1g-i). Thus, N-809 was used at 50 µg for 4T1 tumor studies and 100 µg for MC38 tumor studies.

Next, we compared the antitumor efficacy of N-809 versus N-803+αPD-L1 in murine 4T1 and MC38 tumor models, which are resistant to either αPD-L1 or N-803 monotherapy, respectively.16 Due to significant treatment-related mortality of mice with high-dose N-803 (online supplementary figure S1j,k), we were unable to use equimolar doses of N-809 and N-803+αPD-L1 (online supplementary table S3). Instead, N-809 was compared with 0.3 µg (15 µg/kg) N-803 and 200 µg (10 mg/kg) αPD-L1. These doses are similar to those used in the clinic9 14 15 and have not been previously evaluated in murine tumor models.

N-809 and N-803+αPD-L1 were well tolerated in both tumor models, as neither 4T1 nor MC38 tumor-bearing mice significantly lost weight after treatment (online supplementary figure S2a,c). After N-809 treatment, 4T1 and MC38 tumor-bearing mice had increased inflammation in the lung (online supplementary figure S2b) or kidney, liver and lung (online supplementary figure S2d), respectively. However, these effects were transient and displayed no evidence of morbidity (online supplementary figure S2d).

We have previously demonstrated that N-803 increases systemic PD-L1 expression on CD45⁺ cells, in particular monocytic (M-MDSC) and granulocytic (G-MDSC) myeloid-derived suppressor cells.16 Furthermore, N-803 greatly enhances macrophage PD-L1 expression (online supplementary figure S3a). N-809 reduced detectable surface PD-L1 expression less than N-803+αPD-L1 on CD45⁻ and CD45⁺ cells in the primary tumor and periphery of 4T1 (figure 2A) and MC38 (figure 2B) tumor-bearing mice, suggesting lower PD-L1 blockade by N-809. However, N-809 induced greater antitumor efficacy versus N-803+αPD-L1. N-803+αPD-L1 significantly reduced 4T1 lung tumor burden by 36% vs PBS, whereas N-809 decreased lung metastasis by 60% (figure 2C–E). In addition, N-809 significantly reduced MC38 tumor burden (figure 2F,G) and increased mOS from 24 to 29 days (figure 2H) versus N-803+αPD-L1. N-809 cured 8% of MC38 tumor-bearing mice (figure 2G), and mice did not develop palpable tumors after rechallenge (figure 2I,J). Together, these data demonstrate that N-809 is well-tolerated and has increased antitumor efficacy versus N-803+αPD-L1.

In order to dissect the mechanism of action of N-809, we first examined serum cytokine levels in MC38 tumor-bearing mice. Immunostimulatory cytokines IFNγ (figure 3A) and TNFα (figure 3B) were significantly elevated by N-809 compared with PBS and N-803+αPD-L1. N-809 also elevated IL-10 (figure 3C), an immunosuppressive cytokine which counteracts high levels of IFNγ and TNFα,27 but did not increase serum IL-6 (figure 3D), a known driver of IL-15-associated toxicity.11 26 N-803+αPD-L1 did not increase serum cytokine levels compared with PBS (figure 3A–D).

Intratumoral cytokine signaling was examined by NanoString. Overall, N-809 induced greater changes to genes involved in interferon (figure 3E) and cytokine (figure 3F) signaling than N-803+αPD-L1. Notably, N-809 increased expression of IFNγ targets (MHC class I genes H2-Aa and H2-Q10) and genes associated with positive regulation of IFNγ signaling (Jak1 (JAK1), Ifngr1 (IFNγRI)). Many downregulated genes were type I IFN-related, including Irf7, Ifit1-3, Isg15, Rsad2, Oasl1, Oas1a, Oas2-3 (figure 3E). N-809 also increased gene expression related to immunostimulatory cytokines IL-2 (Il2ra (IL-2Rα/CD25)), IL-12 (Il12ra (IL-12Rα), Il12rb2 (IL-12RβII)) and IL-18 (Il18r1 (IL-18RI)) (figure 3F). Il2rb, a component of the IL-15 receptor (CD122/IL-15Rβ), had the greatest upregulation of all cytokine-associated genes after N-809 treatment (figure 3F). Together, these results suggest that N-809 induces an immunostimulatory environment both systemically and in the TME.

Changes to cytokine levels and signaling suggested major alterations to the immune compartment by N-809, especially compared with N803+αPD-L1. Supporting this, NanoString analysis of the tumor (figure 4A) showed an enrichment of genes associated with the lymphoid compartment (figure 4B), cytotoxicity (figure 4C), costimulation (figure 4D) and the myeloid compartment (figure 4E) by N-809, not N-803+αPD-L1. The most upregulated genes by N-809 were related to NK and CD8⁺ T-cell cytotoxicity and infiltration, including Grzma (granzyme A), Prf1 (perforin) and Grzme (granzyme E), Cxcr3 (CXCR3), Tbx21 (T-bet), Lck (LCK) and Cd3e (CD3ε) (figure 4A).

In order to confirm and expand these findings, we performed a comprehensive analysis of immune subsets in the TME by flow cytometry. Corroborating the gene expression results, NK (figure 4F) and CD8⁺ T-cell (figure 4G) numbers were significantly enhanced after N-809 treatment while the number of T_reg (figure 4H) and M-MDSC were decreased (figure 4J). Reduction in total macrophage numbers (figure 4K) was due to decreased immunosuppressive M2-like (figure 4M) but not pro-inflammatory M1-like (figure 4L) macrophages. N-809 did not alter CD4⁺ T-cell (online supplementary figure S3b) and G-MDSC (online supplementary figure S3c) numbers. Overall, N-809 promoted an inflamed TME by increasing the proportion of immune effector to suppressive cells, as shown by increased CD8⁺ T cell:T_reg (figure 4I) and M1-like:M2-like macrophage (figure 4N) ratios. Other than reducing total macrophage numbers (figure 4K), although lesser than N-809, N-803+αPD-L1 did not affect TME immune cell numbers (figure 4F–N).

We also evaluated immune cell populations in the draining (d) LN. N-809 significantly enhanced NK cell (figure 4F), CD8⁺ T-cell (figure 4G) and CD4⁺ T-cell (online supplementary figure S3b) numbers. In contrast to the TME, T_reg (figure 4H), M-MDSC (figure 4J) and total macrophage numbers (figure 4K) increased in the dLN with N-809. The expanded total macrophage number (figure 4K) was due to increased M1-like (figure 4L), not M2-like (figure 4M), macrophages. G-MDSC were not affected by treatment (online supplementary figure S3c). N-809 increased the ratio of immune effector to suppressive cells in the dLN, as shown by enhanced CD8⁺ T cell:T_reg (figure 4I) and M1-like:M2-like macrophage (figure 4N) ratios. N-803+αPD-L1 treatment did not affect immune cell numbers in the dLN (figure 4F–N). These data demonstrate that N-809, not N-803+αPD-L1, enhances the ratio of immune effector to suppressive cells within the TME and periphery.

Given the increased number of NK and CD8⁺ T cells with N-809 treatment (figure 4F,G), we next performed depletion studies to determine whether these cell types were required for N-809 efficacy. Depletion of either CD8⁺ T cells (figure 5A) or NK cells (figure 5B) completely abrogated the antitumor efficacy of N-809 (figure 5C–E). Depletion of both CD8⁺ T cells and NK cells (figure 5A,B) did not further enhance this effect (figure 5C–E). Thus, CD8⁺ T cells and NK cells are required for the antitumor efficacy of N-809.

Next, we performed a comprehensive analysis of NK and CD8⁺ T-cell phenotype and function to further interrogate the mechanism of action of N-809.

In the TME, N-809 significantly increased NK cell expression of activation-induced NKG2D (figure 6A), proliferation-associated Ki67 (figure 6B) and cytolytic molecule granzyme B (figure 6C) versus PBS and N-803+αPD-L1. Similar effects were observed in the dLN (figure 6A–C). In contrast to the TME, where N-803+αPD-L1 did not affect NKG2D, Ki67 or granzyme B levels, the combination significantly enhanced NKG2D (figure 6A) and granzyme B (figure 6C) in the dLN, although less than N-809 (figure 6A–C). The function of NK cells markedly improved with N-809 versus PBS or N-803+αPD-L1, as shown by increased NK killing of Yac-1 target cells at all effector:target (E:T) ratios (figure 6D). N-803+αPD-L1 also enhanced NK cell killing, but it was significantly reduced versus N-809 (figure 6D). Similar effects on NK-mediated lysis were observed in 4T1 tumor-bearing mice (online supplementary figure S3d). Thus, N-809 promotes greater activation and function of NK cells than N-803+αPD-L1.

Similar to the effect on NK cells, N-809 enhanced expression of Ki67 (figure 6E) and granzyme B (figure 6F) on activated intratumoral CD8⁺ T cells as compared with PBS and N-803+αPD-L1. Additionally, central memory (T_CM) (figure 6G), but not effector memory (T_EM) (figure 6H) or effector (T_eff) (figure 6I) CD8⁺ T cells, were expanded. N-809 did not affect CD8⁺ T-cell clonality in the tumor (online supplementary figure S3e,f). Similar to the TME, N-809 enhanced Ki67 (figure 6E) and granzyme B (figure 6F) expression on activated CD8⁺ T cells in the dLN. In contrast to the tumor, N-809 increased the proportion of T_CM (figure 6G), T_EM (figure 6H) and T_eff (figure 6I) CD8⁺ T cells. N-803+αPD-L1 did not significantly alter CD8⁺ T-cell phenotype in the tumor or dLN (figure 6E–I).

Correlating with the development of a more active CD8⁺ T-cell phenotype, N-809 increased the proportion of activated CD8⁺ T cells in the TME and dLN that produced IFNγ (figure 6J), TNFα (figure 6K) and both IFNγ and TNFα (figure 6L) versus PBS and N-803+αPD-L1. IFNγ production in the tumor did not change with treatment (online supplementary figure S3g), but N-809 significantly increased TNFα production on a per cell basis in the dLN (online supplementary figure S3h). N-803+αPD-L1 only increased the proportion of CD8⁺ T cells producing IFNγ in the dLN, but not to the level induced by N-809 (figure 6J). N-809, but not N-803+αPD-L1, also enhanced CD8⁺ T-cell function in the lungs of 4T1 tumor-bearing mice, as shown by the increased proportion of IFNγ-producing (online supplementary figure S3i), TNFα-producing (online supplementary figure S3j) and IFNγ/TNFα-producing (online supplementary figure S3k) CD8⁺ T cells. Together, these data suggest that N-809 induces activation and increases function of CD8⁺ T cells greater than N-803+αPD-L1.

N-809 promotes the accumulation of a large number of active, functional CD8⁺ T cells and NK cells within the TME (figures 4F,G and 5). This occurs through proliferation within the tumor or migration of de novo CD8⁺ T cells and NK cells from the periphery. In order to distinguish between these possibilities, we administered FTY720 to mice treated with N-809. FTY720 is a sphingosine-1 phosphate receptor inhibitor that sequesters lymphocytes within secondary lymphoid organs and prevents migration in the blood.28 CD8⁺ T cells are more sensitive to FTY720 than NK cells,29 30 as demonstrated by a significant reduction in CD8⁺ T cells (figure 7B), but not NK cells (figure 7A), in the blood after FTY720 treatment. Thus, only CD8⁺ T-cell expansion versus migration could be evaluated. N-809-induced increase of CD8⁺ T-cell number in the TME was completely abrogated by FTY720 administration (figure 7C), suggesting that N-809 promotes lymphocyte trafficking of CD8⁺ T cells into the tumor, not in situ expansion.

To investigate the effect of N-809 on CD8⁺ T-cell and NK cell trafficking, intratumoral gene expression associated with chemokine signaling was evaluated by NanoString. N-809 had a greater overall effect on genes related to chemokine signaling than N-803+αPD-L1 (figure 7D). N-809 significantly enhanced the expression of genes associated with migration of CD8⁺ T cells and NK cells, including Cxcr3 (CXCR3), Ccr5 (CCR5) and Ccl5 (CCL5), with Cxcr3 being the most upregulated gene in the TME (figure 7D).

To confirm these findings, we determined chemokine levels in the TME and chemokine receptor expression on NK and CD8⁺ T cells. Particular focus was placed on NK and CD8⁺ T-cell-expressed chemokine receptors CXCR3 and CCR5 and their respective ligands CXCL9/10 and CCL4/5, as they are the main drivers of NK and CD8⁺ T-cell migration to inflamed tissues and tumors.31–33 In the TME, N-809 significantly increased the level of CXCR3 ligand CXCL9 (figure 7E) but not CXCL10 (figure 7F). N-809 also increased CCR5 ligands CCL4 (figure 7G) and CCL5 (figure 7H). N-809 greatly upregulated expression of CXCR3 and CCR5 on NK (figure 7I,J) and CD8⁺ T cells (figure 7K,L) in the dLN and spleen. There was no expression of CXCR3 or CCR5 on NK or CD8⁺ T cells in the tumor (online supplementary figure S4). Correlating with its inability to increase immune cell numbers in the tumor, N-803+αPD-L1 did not affect chemokine levels in the TME (figure 7E–H) or chemokine receptor expression on NK or CD8⁺ T cells (figure 7I–L). These data suggest that N-809 increases effector cell numbers in the tumor through chemokine-induced CD8⁺ T-cell and NK cell trafficking to the TME.