cIAP1/2 antagonism eliminates MHC class I negative tumors through T cell-dependent reprogramming of mononuclear phagocytes

Abstract

Loss of major histocompatibility complex (MHC) class I and interferon (IFN)-γ sensing are major causes of primary and acquired resistance to checkpoint blockade immunotherapy. Thus, additional treatment options are needed for tumors that lose expression of MHC class I. The cellular inhibitor of apoptosis proteins 1 and 2 (cIAP1/2) regulate classical and alternative nuclear factor kappa B (NF-κB) signaling. Induction of non-canonical NF-κB signaling with cIAP1/2 antagonists mimics costimulatory signaling, augmenting anti-tumor immunity. We show that induction of non-canonical NF-κB signaling induces T cell-dependent immune responses, even in beta-2-microglobulin (β2M)-deficient tumors, demonstrating that direct CD8 T cell recognition of tumor cell expressed MHC class I is not required. Instead, T cell-produced lymphotoxin reprograms both mouse and human macrophages to be tumoricidal. In wild type mice, but not mice incapable of antigen-specific T cell responses, cIAP1/2 antagonism reduces tumor burden by increasing phagocytosis of live tumor cells. Efficacy is augmented by combination with CD47 blockade. Thus, activation of non-canonical NF-κB stimulates a T cell-macrophage axis that curtails growth of tumors that are resistant to checkpoint blockade due to loss of MHC class I or IFN-γ sensing. These findings provide a potential mechanism for controlling checkpoint blockade refractory tumors.

One Sentence Summary:

T cell-dependent reprogramming of intratumoral macrophages by cIAP1/2 inhibition leads to control of MHC class I negative pancreatic cancer in mice.

Introduction

Tumor cell evasion of immune attack is an important component of tumor progression. Mutations in genes responsible for interferon (IFN)-γ signaling and major histocompatibility complex (MHC) class I antigen presentation, which render tumor cells invisible to CD8 T cells, are common means of acquired resistance to checkpoint blockade in melanoma and implicate IFN-γ and CD8 T cells as critical components of the immune response in this setting (1–4). Whereas MHC class I expression lost due to lack of cytokine responsiveness could potentially be restored, the solution to genetic loss of beta 2 microglobulin (β2M) or other components of MHC class I antigen presentation is not immediately obvious (5). Other types of immune cells could, in theory, be mobilized for tumor cell destruction. Natural killer (NK) cells can identify and destroy MHC-negative, virally-infected cells. Although NK cells are infrequent in most solid tumors, their activity can be augmented by stimulator of interferon genes (STING) agonists or pegylated interleukin (IL)-2 (6, 7). Tumor cells express the NK cell-activating ligand, MHC class I chain-related protein A (MICA), but proteolytically cleave this ligand to prevent NK cell-mediated cytotoxicity (8, 9). Tumor-targeting antibodies might also promote destruction of MHC class I deficient escape variants via engagement of Fc receptors on NK cells or on myeloid cells. Strategies targeting myeloid cells may prove beneficial, and macrophages in particular can be tumoricidal in some cases (10, 11). However, myeloid cells lack antigen specificity, are notoriously plastic, and become immunosuppressive in the setting of cancer. Chronic inflammation can lead to tumor progression through elaboration of cytokines such as IL-1 and IL-6, both of which promote malignant cell growth (12, 13).

T cells can recognize tumor antigens presented on MHC by cross-presenting dendritic cells or tumor cells. Direct cytolysis of tumor cells by CD8 T cells can occur and is likely a major pathway by which checkpoint blockade operates. In addition to cytotoxicity, both CD4 and CD8 T cells produce cytokines and chemokines to coordinate responses by multiple immune cell types, including mononuclear phagocytes. This coordination between tumor-specific T cells and innate immune cells in cancer is underexplored. The cellular inhibitor of apoptosis proteins (cIAPs) regulate nuclear factor kappa B (NF-κB) signaling and control cytokine output from activated T cells (14–18). The cIAPs can be antagonized pharmacologically by peptide mimetics of SMAC, the natural binding partner for cIAP1/2 (19–21). cIAP1/2 antagonists such as LCL161 have minimal effect on T cells in the absence of T cell receptor (TCR) stimulation, but can deliver a costimulatory signal to natural killer T (NKT), CD4 and CD8 T cells from both mice and humans (14, 18, 22), resulting in increased production of IL-2, IFN-γ, tumor necrosis factor (TNF)-α and granulocyte-macrophage colony-stimulating factor (GM-CSF). cIAP1/2 antagonism in mice augments anti-tumor responses in several tumor models (14, 18, 23–25). We previously showed that cIAP1/2 antagonism acts by promoting NF-κB-inducing kinase (NIK) accumulation in T cells, mimicking stimulation through co-stimulatory ligands of the tumor necrosis factor receptor (TNFR) superfamily, to enhance cytokine production and anti-tumor immunity in the context of a melanoma vaccine in mice (14). Subsequent reports showed that cIAP1/2 antagonism augments TNF-α and IFN-γ production from CD8 T cells, which leads to direct TNF-α-mediated tumor cell death in mouse models of glioblastoma and reprogramming of the tumor microenvironment in breast cancer models (23, 25). cIAP1/2 antagonism also leads to enhanced phagocytosis by macrophages in multiple myeloma patients; studies of cIAP1/2 antagonism in a murine model of multiple myeloma confirmed that enhanced phagocytosis led to short term control of tumors, although long term durable remissions still required adaptive immunity (24). How T cells interface with macrophages in tumors is unclear. We hypothesize that augmentation of T cell-produced cytokines with cIAP1/2 antagonists may reprogram innate immune cells to phagocytose tumor cells, thereby bypassing the need for direct recognition of MHC class I.

Pancreatic ductal adenocarcinoma (PDAC) is a poorly immunogenic tumor replete with myeloid cells, which has proven refractory to all T cell-based immunotherapies outside of the 1% of patients with microsatellite instability (26, 27). PDAC is sensitive to tumoricidal myeloid cells reprogrammed by IFN-γ-induced downstream of agonistic anti-CD40, although systemic delivery of anti-CD40 is complicated by toxicities (10, 28, 29). We therefore tested the effects of cIAP1/2 antagonism in both implantable and autochthonous murine models of pancreatic cancer and found that cIAP1/2 antagonism robustly elicited tumoricidal macrophages in a T cell-dependent fashion.

Results

IAP antagonists are effective in models of either primary or acquired resistance to checkpoint blockade.

To model either primary or acquired resistance to checkpoint blockade, we selected two cancer models, one that is highly infiltrated by CD8 T cells and responds well to programmed cell death protein-1 (PD-1) blockade and one that has few infiltrating T cells and is refractory to PD-1 blockade. We used CRISPR/Cas9 to delete B2m from the immunotherapy-sensitive colorectal tumor cell line MC38 and the immunotherapy-resistant pancreatic cancer cell line 6694c2. For both subcutaneous tumor models, the parental cell lines were controlled by treatment with the cIAP1/2 antagonist, LCL161 (Fig. 1A and B). A few of the 6694c2 tumors initially responded to LCL161 treatment, but then grew progressively. We generated cell lines from the resistant tumors. Reimplantation of these cell lines into naïve mice showed that the tumors were still capable of responding to cIAP1/2 antagonism, suggesting that the acquired resistance to cIAP1/2 antagonism in this model is tumor cell extrinsic (fig. S1). Of note, cIAP1/2 antagonism controlled growth of β2M^−/− tumors, indicating that cIAP1/2 antagonism is effective in the settings of primary and acquired checkpoint blockade resistance when tumors have lost expression of MHC class I (Fig. 1A-B).

Fig. 1. cIAP1/2 antagonism overcomes primary and acquired resistance to checkpoint blockade.

(A) MC38 or MC38 β2M^−/− tumor cells were implanted subcutaneously into the flanks of C57BL/6 mice (300,000 cells per mouse). Mice were treated with LCL161 (75mg/kg) or sodium acetate vehicle solution by oral gavage every 3 days starting on day 3, or were treated with anti-PD-1 (200μg/mouse) by intraperitoneal injection on days 6, 9, 12. Tumor size was measured over time. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (B) 6694c2 or 6694c2 β2M^−/− tumor cells were implanted subcutaneously into the flanks of C57BL/6 mice (300,000 cells per mouse). Mice were treated with LCL161 (75mg/kg) by oral gavage every 3 days starting on day 3, or were treated with anti-PD-1 (200μg per mouse) and anti-CTLA4 (200 μg per mouse) by intraperitoneal injection on days 6, 9, 12. Tumor size was measured over time. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (C) Poorly immunogenic pancreatic cell lines 6694c2 and 6499c4 were implanted orthotopically in C57BL/6 mice treated with vehicle (veh), LCL161 (LCL) alone (75mg/kg every 3 days starting on day 4) anti-PD-1 alone (200μg per mouse, twice weekly), or LCL161 plus anti-PD-1. Tumors were analyzed at day 21. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (D) LSL-Kras^G12D;p53+/flox;p48cre+ mice were monitored by ultrasound and enrolled into the study when tumors were greater than 4mm in diameter. Mice were treated with LCL161 or vehicle every 2 days and with or without anti-PD-1 (200μg per mouse) every 4 days. Survival was monitored. Mice were euthanized when primary tumor volume exceeded 2cm³ by ultrasound or for metastasis-associated morbidity (jaundice, ascites, labored breathing). Cause of death due to tumor was confirmed by necropsy. All data are shown and p values were calculated by Log-rank (Mantel-Cox) test.

Most patients with PDAC are unresponsive to checkpoint blockade, making this an area of urgent unmet need (30). We conducted subsequent in vivo experiments in mice using poorly immunogenic pancreatic tumor cells implanted orthotopically to better model the PDAC microenvironment. We used KPCY cell lines 6694c2 and 6499c4, which were previously reported to contain no detectable antigenic single nucleotide polymorphisms and which are unresponsive or partially unresponsive to combination checkpoint blockade (31). In vivo, we confirmed that orthotopically implanted 6694c2 and 6499c4 tumors were unresponsive to PD-1 blockade; nevertheless, we observed a 50% reduction in tumor size at 3 weeks with cIAP1/2 antagonism (Fig. 1C). Similar findings were observed in the lox-stop-lox-Kras^G12D;p53+/flox;p48cre+ genetically engineered mouse model (KPC GEMM) of spontaneous PDAC, in which treatment was initiated upon ultrasound confirmation of tumors greater than 4mm in diameter. cIAP1/2 antagonism more than doubled median survival (9 days versus 21 days, P = 0.02, Fig. 1D). Combination of cIAP1/2 antagonist with anti-PD-1 did not further augment survival above that seen with cIAP1/2 antagonist monotherapy (P = 0.739).

IAP antagonists were developed as cancer therapeutics based on their ability to sensitize tumor cells to TNF-α-mediated apoptosis (32). TNF-α sensitivity occurs in glioblastoma and is partially responsible for the efficacy of cIAP1/2 antagonism in this setting (23). However, pancreatic cancer cell lines 6694c2 and 6499c2 were highly resistant to direct killing by cIAP1/2 antagonism in vitro. Despite verified loss of cIAP1 protein by western blotting, cytotoxicity could not be enhanced through co-treatment with TNF-α (fig. S2).

cIAP1/2 antagonists elicit T cell-dependent responses to MHC class I negative tumors.

T cells recognize tumor antigens via peptide-MHC complexes presented on the surface of tumor cells. Pancreatic tumor cells express negligible MHC class I, but expression can be induced upon exposure to IFN-γ (Fig. 2A). We therefore used clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 to generate variants of 6694c2 cells lacking surface MHC class I via loss of β2M, transporter associated with antigen processing 1 (TAP1), or components of the IFN-γ receptor. Loss of MHC class I was verified by flow cytometry, and in the case of β2M^−/− 6694c2 cells, we additionally verified inability to present exogenously loaded peptide to cognate CD8 T cells (fig. S3). Nearly all of the MHC class I-deficient lines were sensitive to cIAP1/2 antagonism upon orthotopic implantation into wild-type mice (P = 0.0003 (β2M^−/−), 0.012 (TAP1−/−), 0.0015 (IFNγR1−/−), 0.49 (IFNγR2−/−), and 0.0006 (β2M^−/− survival), Fig. 2B and C), demonstrating again that direct T cell recognition of peptide-MHC on tumor cells is not required. This is in contrast to both endogenous and checkpoint blockade-induced immune-mediated control of tumor growth in melanoma and other cancers, where MHC class I and IFN-γ sensing are critical (1–3, 33–35). To ascertain whether NK cells might be responsible for recognizing MHC class I negative tumors, we implanted wild type or β2M^−/− 6694c2 cells into mice treated with depleting antibodies to NK1.1. NK cell depletion showed no impact on the effectiveness of cIAP1/2 antagonism in vivo (P = 0.0004); however, T cells were critical, as demonstrated by the failure mice depleted of both CD4 and CD8 T cells to control tumor burden (P = 0.79, Fig. 2D and fig. S4A and B).

Fig. 2. The immune response to MHC class I negative tumors is T cell-dependent.

(A) Flow cytometry of cultured wild-type (WT), β2M^−/−, or TAP1^−/− tumor cells confirms lack of IFN-γ-inducible MHC class I. (B) Parental or CRISPR/Cas9-edited 6694c2 cells were inoculated orthotopically into C57BL/6 mice that were treated with vehicle or LCL161 (75mg/kg) by oral gavage every 3 days starting 4 days post-inoculation. Tumors were harvested at day 21. Results are compiled from two separate experiments. Data are presented as weights normalized to the mean of the vehicle treatment for each tumor type ± SEM and p-values were calculated by ANOVA. (C) 6694c2 parental or β2M^−/− cells were implanted orthotopically into C57BL/6 mice. Mice were treated with vehicle or LCL161 (75mg/kg) by oral gavage every 3 days starting 4 days post-inoculation. Survival was monitored. Data are representative of 2 independent experiments with 5 mice per group. P-values were calculated by log-rank (Mantel-Cox) test. (D) 6694c2 parental or β2M^−/− cells were implanted orthotopically into C57BL/6 mice. Mice were treated with vehicle or LCL161 (75mg/kg) by oral gavage every 3 days starting 4 days post-inoculation and depleting antibodies to CD4 and CD8 or NK1.1 (100 μg each antibody per mouse) on day 0 and then concurrently with LCL161 dosing. Control animals received isotype treatment. Data are compiled from 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (E) 6694c2 cells were implanted orthotopically into C57BL/6 mice or mice of the indicated genotype, and mice were treated with vehicle or LCL161 (75mg/kg) by oral gavage every 3 days, starting on day 5 post-inoculation. Tumors were harvested at day 21 and weight was recorded. Results are compiled from three separate experiments. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (F) 6694c2 β2M^−/− tumors were implanted into C57BL/6, Prf1^−/− or Batf3^−/− mice and mice were treated with vehicle or LCL161 (75mg/kg) by oral gavage every 3 days starting 4 days post-inoculation. Tumors were harvested at day 21 and weight was recorded. Results are compiled from two separate experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA.

To further determine the cellular components required for the in vivo efficacy of cIAP1/2 antagonism in poorly immunogenic PDAC, 6694c2 cells were implanted orthotopically into mice lacking the following cell types: cross-presenting DCs (Batf3^−/−); CD8 and MHC class I-restricted T cells (β2M^−/−); αβ T cells (TCRα^−/−); CD4 T cells (I-A^b−/−); or B cells (μMT^−/−). Sensitivity to cIAP1/2 antagonism was ablated in Batf3^−/− (P = 0.99) and TCRα^−/− mice (P = 0.99) and reduced in β2M^−/− mice (P = 0.47), demonstrating that cIAP1/2 antagonism requires an antigen-specific T cell response (Fig. 2E). Mice lacking CD4 T cells were still responsive to cIAP1/2 antagonism indicating lack of a strict requirement for CD4 T cells. Nevertheless, the trending reduction in tumor size in cIAP1/2 antagonist treated β2M^−/− mice suggests that CD4 T cells may be partly redundant with CD8 T cells. Direct cytotoxicity was not required, however, given that cIAP1/2 antagonism was similarly effective at controlling β2M^−/− 6694c2 tumors in wild-type and perforin-deficient (Prf1^−/−) mice relative to vehicle-treated mice (P = 0.0062, Fig. 2F). These findings show that direct recognition of tumor-expressed MHC class I and perforin/granzyme mediated cytolysis are not required, suggesting that T cells are exerting their effector function through activation of alternative immune pathways. Importantly, MHC class I is expressed by host cells, including cross-presenting dendritic cells, which are critically important for activation of CD8 T cells. cIAP1/2 antagonism was unable to control β2M^−/− 6694c2 tumors in Batf3^−/− mice, underscoring the importance of CD8 T cell recognition of tumor antigens presented by dendritic cells (Fig. 2F).

cIAP1/2 antagonism induces accumulation of phagocytic macrophages in PDAC tumors.

Intratumoral CD8 T cell frequencies averaged 2% of total CD45⁺ cells in orthotopic 6694c2 and in KPC GEMM tumors and did not increase with cIAP1/2 antagonism when examining orthotopic tumors 18 days post-implantation or end stage KPC GEMM tumors (fig. S5A and B). This paucity of CD8 T cells, combined with a lack of requirement for direct recognition of MHC class I on tumor cells, suggests that, in poorly immunogenic tumors, another cell type may be involved in restricting tumor growth. We performed single cell transcriptional profiling of total CD45⁺ cells from vehicle or cIAP1/2 antagonist-treated tumors at mid-stage of growth when tumor size is relatively similar (Day 12), expecting that we would capture the early processes of anti-tumor immunity.

Among pancreatic tumor infiltrates, we found granulocytes, lymphocytes, dendritic cells, and 6 clusters of macrophages which were relatively evenly distributed between vehicle and LCL161 treatment conditions, with the notable exceptions that cIAP1/2 antagonism induced a slight increase in T cells at this time point and the appearance of a distinct macrophage cluster (Fig. 3A-B and fig. S5C). Four of the macrophage clusters clearly aligned with subsets recently defined in colorectal cancer (36). Half of pancreatic cancer macrophages have been reported to arise from tissue-resident embryonic macrophages through local proliferation (37), and indeed we found an actively cycling macrophage cluster (Macro-cycling, Fig. 3A and C). The final macrophage cluster, which is highly enriched with LCL161 treatment, is defined by co-expression of a variety of T and B cell lineage markers (Macro-phagocytic, Fig. 3A and C). RNA counts were consistent with these being single cells rather than doublets as they contained similar numbers of recovered transcripts per cell to other macrophage clusters (Fig. 3D), and we hypothesized that these macrophages likely phagocytosed one or more lymphocytes. These putative phagocytic macrophages were more abundant in cIAP1/2 antagonist-treated compared to control tumor infiltrates ( Fig. 3E). We looked for surface markers that might distinguish these phagocytic macrophages and found that the transcripts for Ly6c and the MHC class II molecule, H2-Aa, were co-expressed (Fig. 3F). Although these could have originated from T and B cells, respectively, we found that cell surface-expressed Ly6C and MHC class II protein could be detected by flow cytometry of F4/80⁺CD11b⁺ macrophages from cIAP1/2 antagonist-treated tumors from both orthotopic (P = 0.009) and spontaneous (P < 0.0001) tumor models (Fig. 3G to I). Similar populations were not found in vehicle treated tumors (Fig. 3G to I). These data indicate that cIAP1/2 antagonism induces a phagocytic macrophage subset, which can be identified in tumors by co-expression of the surface markers Ly6C and MHC class II.

Fig. 3. Transcriptional analysis of tumor infiltrates reveals a cIAP1/2 antagonism-induced shift toward phagocytic macrophages.

(A) CD45⁺ cells were sorted from orthotopic 6694c2 pancreatic tumors implanted in C57BL/6 mice treated with vehicle or LCL161 for 12 days. n=5 tumors pooled per group. ScRNAseq libraries were constructed using the 10X Genomics platform and visualized by tSNE analysis. Both vehicle and LCL161 treatment groups were combined to identify cell clusters. Cluster names were defined using top differentially expressed genes as indicated. Macrophage (macro), conventional dendritic cell (cDC), plasmacytoid dendritic cell (pDC). (B) Violin plot shows expression of cluster-defining genes for populations shown in (A). (C) Heatmap shows z-scores of expression of top cluster-defining genes for each of the 6 macrophage clusters identified in (A). T and B cell clusters are included for comparison. (D) Violin plot shows RNA read counts for each of the macrophage clusters defined in (A). All counts are consistent with originating from a single cell. (A to D) show results from combined vehicle and LCL161 treatment groups. (E) Fractional representation of each cluster based on treatment group. (F) tSNE plots showing cells highlighted in green that co-express H2-Aa and Ly6c2. (G) Tumor infiltrates from end stage KPC GEMM mice treated with vehicle (n=5) or LCL161 (n=4) were analyzed by flow cytometry. Ly6c⁺MHCII⁺ cells were quantified as a percentage of total macrophages (CD11b⁺F4/80⁺Gr1^l°^w). Data are presented as mean ± SEM and p-values were calculated by Mann-Whitney test. (H) Tumor infiltrates from mice with orthotopic 6694c2 tumors treated with vehicle or LCL161 were harvested at day 18 and analyzed by flow cytometry. Representative flow plots are gated on CD11b⁺F4/80⁺Gr1^l°^w macrophages. (I) Quantification of Ly6c⁺MHCII⁺ cells shown in H.n=10 per group. Data are compiled from 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by Mann-Whitney test.

Macrophages display enhanced phagocytosis of tumor cells in vivo following cIAP1/2 antagonism.

To determine whether macrophages in vivo are involved in the response to cIAP1/2 antagonism, we inoculated mice with orthotopic 6694c2 tumors and treated with antibody blockade of colony stimulating factor 1 (CSF-1), a critical survival factor for macrophages that has been previously reported to play a role in pancreatic cancer (38). CSF-1 blockade abrogated the effect of cIAP1/2 antagonism, suggesting that macrophages may be involved in tumor reduction (P = 0.42, Fig. 4A and fig. S6A). To measure phagocytosis in vivo, we implanted mice orthotopically with mCherry expressing 6694c2 cells to enhance detection of tumor cell fragments by flow cytometry after treatment with vehicle or cIAP1/2 antagonist. At mid-stage of tumor growth, tumors were harvested, and infiltrating macrophages were analyzed by flow cytometry (Fig. 4B). cIAP1/2 antagonist treated tumors showed increased frequencies of mCherry⁺ macrophages expressing high abundance of MHC class II, indicating that cIAP1/2 antagonism enhances phagocytic uptake of tumor cells in vivo (P = 0.016, Fig. 4B and C). This process requires tumor-specific T cells, as Batf3^−/− mice lacking the ability to prime T cell responses showed a blunting of intratumoral phagocytic capacity consistent with the lack of reduction in overall tumor size (P = 0.0088, C57BL/6 vs. Batf3^−/−, Fig. 4D and Fig. 2E). Imaging flow cytometry confirmed that tumor cell fragments were contained in vesicle-shaped structures on the interior of tumor-infiltrating mononuclear phagocytes (Fig. 4E). Phagocytosis could be directly visualized in excised tumor tissue in situ by immunofluorescent staining of 6694c2 tumors from cIAP1/2 antagonist treated mice using antibodies to F4/80 and yellow fluorescent protein (YFP), which 6694c2 tumors expressed (Fig. 4F). The fraction of tumor-containing macrophages was significantly higher in tumors isolated from mice treated with LCL161 (P = 0.0097), even though total F4/80⁺ cell numbers did not differ between control and treated tumors (P = 0.99, Fig. 4G). These results demonstrate by an orthogonal method that cIAP1/2 antagonism in immune-competent mice induces tumor cell uptake by mononuclear phagocytes.

Fig. 4. Intratumoral phagocytes engulf pancreatic tumor cells in vivo in mice capable of generating tumor-specific T cell responses.

(A) Mice bearing orthotopic 6694c2 tumors were treated with LCL161 and blocking antibodies to CSF-1 or isotype control. Tumors were harvested on day 21 and tumor weights were recorded. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (B) mCherry⁺ 6694c2 cells were inoculated orthotopically into C57BL/6 mice that were treated with LCL161 or vehicle every 3 days. On day 12, tumor infiltrates were analyzed by flow cytometry. Macrophages were defined as CD45⁺ CD11b⁺ cells. Representative plots showing tumor⁺ myeloid cells. Representative of three independent experiments. (C) Quantification of n=5 mice per group from B. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by Mann-Whitney test. (D) mCherry⁺ 6694c2 cells were inoculated orthotopically into C57BL/6 or Batf3^−/− mice that were treated with LCL161 or vehicle every 3 days. On day 12, tumor infiltrates were analyzed by flow cytometry and gated as shown in B. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA.(E) Imaging flow cytometry of individual cells shown in the gate in B. Cell identification number is in the upper left of each panel. Representative of 8 images. (F) 6694c2-YFP orthotopic tumors were implanted into C57BL/6 mice, treated with vehicle or LCL161 (75mg/kg), and harvested at mid-stage of growth on day 12. A representative image is shown. (G) Quantification was performed on one 20x image (high-powered field, hpf) per sample (n=5 per group). Macrophages were defined as F4/80⁺. (H) CD47^−/− 6694c2 cells were generated by CRISPR/Cas9 and implanted orthotopically into C57BL/6 mice. Mice were treated with LCL161 or vehicle. Tumor were harvested on day 21 and weighted. Data are compiled from two independent replicates and normalized to vehicle treated controls. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (I) Mice bearing orthotopic 6694c2 tumors were treated with LCL161 or vehicle and CD47nb (200μg per mouse, daily) or a control nanobody starting on day 4. Data are representative of 2 independent experiments with 5 mice per group. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (J) Survival curve of KPC GEMM mice treated with vehicle (n=9), LCL161 (n=14, 75mg/kg every 3 days), or LCL161 and CD47nb (n=5, 200μg per mouse daily). All data are shown and p values were calculated by Log-rank (Mantel-Cox) test.

Tumors express a combination of pro-phagocytic and anti-phagocytic signals, the most well-studied example being the “don’t eat me” ligand, CD47. Binding of CD47 to signal regulatory protein alpha (SIRPα) on phagocytes triggers an inhibitory signal that blocks phagocytosis (39–41). Orthotopic 6694c2 tumors with genetic loss of CD47 were more effectively cleared upon cIAP1/2 antagonism, suggesting that pharmacologic blockade of this pathway may augment efficacy (P = 0.0005, Fig. 4H and fig. S6B). Alpaca-derived antibody fragments, or nanobodies, are only 15kDa in size, are not glycosylated, and do not require pairing with light chains, making them ideal candidates for large-scale production and diffusion into dense tissues (42). We treated C57BL/6 mice bearing orthotopic 6694c2 tumors with a high affinity nanobody targeting mouse CD47 (43). Although the CD47nb achieves only partial blockade of CD47 in tumors (44), we found that daily dosing of CD47nb, as compared to an irrelevant nanobody (VHHcont) combined with cIAP1/2 antagonism was effective at reducing tumor weight (P = 0.0001, Fig. 4I). CD47nb and cIAP1/2 antagonist combination therapy in the autochthonous LSL-Kras^G12D;p53+/flox;p48cre+ (KPC GEMM) model extended overall survival significantly compared to cIAP1/2 antagonism alone (P = 0.0219, Fig. 4J). These results suggest that cIAP1/2 antagonism increases the frequency of tumoricidal phagocytes and can be combined with therapies that further enhance phagocytosis.

T cell production of lymphotoxin is a critical factor for efficacy of cIAP1/2 antagonism.

We hypothesized that T cells activated in the presence of cIAP1/2 antagonist produce one or more soluble factors capable of augmenting the phagocytic capacity of macrophages. CD8 T cells isolated from an OT-I mouse were activated in vitro with anti-CD3/CD28 beads and analyzed for production of 22 cytokines. Supernatants showed a significant, dose-dependent increase in production of lymphotoxin beta (LTβ, P = 0.0008), GM-CSF (P = 0.0081), IL-2 (P = 0.022), TNF-α (P = 0.025), and IL-17A (P = 0.0022), but not IFN-γ (P = 0.31), consistent with previous reports of cIAP1/2 antagonists skewing cytokine production (Fig. 5A) (18, 45). To determine which of these cytokines are produced by T cells in vivo, we sorted CD4 and CD8 T cells by fluorescence-activated cell sorting (FACS) from mid-stage orthotopic tumors of mice treated with vehicle or cIAP1/2 antagonist. Of our candidate cytokines, only Ltb, Il17a, and Csf2 were significantly upregulated in cIAP1/2 antagonist treated tumors (P = 0.002 for Il17a, 0.0031 for Ltb, and 0.0316 for Csf2), and only Ltb and Ifng were transcribed to an appreciably high degree (Fig. 5B).

Fig. 5. cIAP1/2 antagonist-induced lymphotoxin (LTα1β2) is required for anti-tumor efficacy in vivo.

(A) CD8 T cells were isolated by negative selection on magnetic beads from an OT-I mouse and activated in vitro with anti-CD3/CD28 beads in the presence of the indicated concentrations of LCL161. 48 hours later, supernatants were collected and analyzed by cytokine/chemokine bead array. The indicated cytokines were significantly increased in LCL161 conditions. Data are presented as mean ± SEM and p-values were calculated by two-way ANOVA. *p<0.05. (B) Orthotopic pancreatic tumors from mice treated with vehicle or LCL161 were harvested on day 12, digested, and CD4 and CD8 T cells were sorted by FACS for limited input bulk RNAseq (n=3 biological replicates per group). Read counts from CD4 and CD8 samples were combined and differential expression analysis was performed between vehicle and LCL samples (n=6 per group). Cytokines from (A) are shown in the heatmap along with Log2 fold change between LCL161 and vehicle samples. (C) 6694c2 cells were implanted orthotopically in C57BL/6 or IFN-γ^−/− mice. Mice were treated with LCL161 (75mg/kg) by oral gavage every 3 days starting on day 4. Recombinant LTβR-Fc (60μg per mouse) was used as a decoy receptor to block lymphotoxin signaling. Tumors were harvested at 20 days post-inoculation and weighed. Results are combined from 2 independent experiments. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (D) 6694c2 cells were implanted orthotopically in C57BL/6 mice. Mice were treated with LCL161 (75mg/kg) by oral gavage and were also treated with blocking antibodies to the indicated cytokines or isotype control antibodies (100μg per mouse) or with recombinant Ltbr-Fc (60μg per mouse) every 3 days starting on day 0. Tumors were harvested at 20 days post-inoculation and weighed. Results are combined from 3 independent experiments. Data are presented as mean ± SEM and p-values were calculated by ANOVA.

To address the role of T cell-derived cytokines in vivo, we inoculated wild type or IFN-γ^−/− mice with 6694c2 orthotopic tumors. Despite the known role of IFN-γ in both anti-tumor immunity and in macrophage polarization, IFN-γ^−/− mice showed reduced tumor burden when treated with cIAP1/2 antagonist (P = 0.042, Fig. 5C). This efficacy was abrogated upon treatment with a decoy version of the lymphotoxin receptor (LTβR-Fc) that prevents lymphotoxin signaling in vivo (P = 0.87, Fig. 5C). Genetic deficiency of particular cytokines may influence immune cell development; for example, lymphotoxin-deficient mice have abnormal T cell mediated immunity due to failure to develop secondary lymphoid organs. We therefore used cytokine blocking antibodies or the lymphotoxin receptor decoy to test whether transient cytokine blockade would affect pancreatic tumor growth and response to cIAP1/2 antagonism (Fig. 5D). Importantly, blockade of TNF-α had no loss of efficacy (P = 0.0001), consistent with PDAC tumor cells being resistant to TNF-α-mediated cell death even in the presence of cIAP1/2 antagonist (Fig. 5D and fig. S2). TNF-α RNA and protein abundance were uniformly low in pancreatic tumors, suggesting that TNF-α has only a minor role in this tumor model (fig. S7). Similar to the genetic loss of IFN-γ, antibody blockade of IFN-γ failed to prevent the reduction in tumor burden downstream of cIAP1/2 antagonism (P = 0.033), indicating that IFN-γ does not play a critical role. LTβR-Fc alone (P = 0.82) or in combination with IFN-γ blockade (P = 0.75) eliminated the efficacy of cIAP1/2 antagonism (Fig. 5D). We therefore concluded that lymphotoxin is induced by cIAP1/2 antagonism in vivo and that this cytokine is responsible for the anti-tumor efficacy.

Lymphotoxin and cIAP1/2 antagonism reprogram macrophages to enhance phagocytosis of live tumor cells.

To better model the effects of cytokine signaling combined with cIAP1/2 antagonism on macrophage phagocytosis, we developed an in vitro phagocytosis assay. Bone-marrow derived macrophages (BMDMs) treated with cIAP1/2 antagonist displayed increased phagocytosis of cocultured mCherry⁺ tumor cells (P = 0.095, Fig. 6A), suggesting a direct effect of noncanonical NF-κB signaling in macrophages, consistent with previous reports (24). However, phagocytosis was significantly augmented by addition of recombinant lymphotoxin (LTα1β2) (P = 0.0067), to a similar degree as seen with blockade of the known phagocytosis inhibitory ligand CD47 (P = 0.026, Fig. 6A and fig. S8).

Fig. 6. Macrophages treated with cIAP1/2 antagonist and lymphotoxin phagocytose live tumor cells.

(A) BMDMs were cultured with vehicle or 500nM LCL161 and LTα1β2 as indicated for 24 hours, washed and cocultured with mCherry-expressing 6694c2 cells. In some cases, a nanobody targeting CD47 (CD47nb) or a control nanobody (VHHcont) were included with the tumor cells. Phagocytosis was measured by flow cytometry. Data are representative of 3 independent experiments. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (B) In vitro phagocytosis assays were performed using healthy donor PBMCs and CFSE-labeled human PDAC cells (cell line 8988T). Total PBMCs were cultured for 6 days with 20ng/mL human rGM-CSF. Cells were washed and plated with CFSE-labeled tumor cells in media containing vehicle or 500nM LCL161 for 18 hours. Phagocytosis was measured by flow cytometry. Data are representative of 3 independent experiments. Data are presented as mean ± SEM and p-values were calculated by Mann-Whitney test. (C) 500nM LCL161 was added with BMDM or with 6694c2-mCherry cell culture media. 24 hours later, media was removed and 6694c2-mCherry cells were plated onto macrophages at a 1:5 ratio. Media was supplemented with 500nM LCL161 or LTα1β2 as indicated. Phagocytosis was measured by flow cytometry after 18 hours of coculture. Data are representative of 3 independent experiments. Data are presented as mean ± SEM and p-values were calculated by ANOVA. (D) BMDMs were plated onto glass bottom tissue culture dishes and cultured with 500nM LCL161, IFN-γ, and LTα1β2 for 24 hours. Cells were washed and cocultured with YFP-expressing 6694c2 cells and Sytox blue. Cultures were imaged for 18 hours in a CO₂ controlled humidified incubator. Still frames with white arrow show a live tumor cell being engulfed by a macrophage. (E) BMDMs and 6694c2-zsGreen cells were plated onto glass bottom tissue culture dishes and treated with 500nM LCL161 and LTα1β2 for 18 hours. Cells were fixed and stained with anti-CD11b and DAPI. Representative images of zsGreen⁺ macrophages with multiple green puncta showing phagocytosis (top panel) or single large zsGreen⁺ inclusions with DAPI⁺ nuclei showing phagocytosis with intact nuclei (bottom panel, arrowhead indicates tumor cell nucleus) are shown. Scale bars = Scale bar=10μm. (F) Quantification of the proportion of phagocytic zsGreen⁺ macrophages at the indicated time points. N > 100 total macrophages per time point. Two independent experimental replicates were combined. Cells were counted by a blinded investigator. (G) Of the total zsGreen⁺ macrophages, the percentage of zsGreen⁺ macrophages with visibly intact tumor cell nuclei was calculated. (H) Confocal z-stack rendering of a tumor cell from (G) being phagocytosed by a macrophage. Scale bar=10μm.

To determine whether a similar outcome could be observed in human cells, healthy donor peripheral blood mononuclear cells (PBMCs) were cultured with recombinant human GM-CSF to induce differentiation of monocyte-derived macrophages while maintaining T cells in the culture mix. Human PDAC 8988T tumor cells were labeled with carboxyfluorescein succinimidyl ester (CFSE) and mixed with monocyte-derived macrophages in the presence of vehicle or cIAP1/2 antagonism and human LTα1β2. cIAP1/2 antagonism plus lymphotoxin significantly increased phagocytosis of human tumor cells, indicating that enhancement of phagocytosis can occur in human monocyte-derived macrophages as well (P = 0.0001, Fig. 6B). Lymphotoxin alone was inadequate to induce phagocytosis (P = 0.99), indicating a combination benefit of lymphotoxin signaling and cIAP1/2 antagonism (P = 0.0001, Fig. 6C). Both wild type and β2M^−/− 6694c2 cells were phagocytosed equally well by tumoricidal macrophages (P = 0.0001, fig. S9). Macrophages pre-treated with cIAP1/2 antagonist were more effective phagocytes than macrophages exposed to tumor cells that had been pretreated with cIAP1/2 antagonist, suggesting that the pro-phagocytic effect of cIAP1/2 antagonism is due to macrophage reprogramming rather than to enhanced tumor cell sensitivity (P = 0.0001, Fig. 6C).

Macrophage uptake of tumor cell fragments is increased upon cIAP1/2 antagonism; however, this could be either because macrophages engulf live tumor cells and kill them in the phagolysosome, or because the tumor cell dies by some other means and macrophages phagocytose the apoptotic remnants. We therefore conducted live cell imaging of tumor and macrophage co-cultures with cIAP1/2 antagonist and T cell-derived cytokines. We pretreated BMDMs with cIAP1/2 antagonist, lymphotoxin, and IFN-γ for 24 hours, washed the cells and added YFP-expressing 6694c2 cells and the dead cell marker Sytox blue. We observed macrophages engulfing and destroying live tumor cells (Fig. 6D and Movie S1-S2). Importantly, no such engulfment was seen when non-pretreated macrophages were used, and tumor cells remained viable and proliferated when cultured with cIAP1/2 antagonist, lymphotoxin, and IFN-γ in the absence of macrophages (Movie S1). Although we do not exclude a role for tumor cell apoptosis, we have demonstrated that engulfment of live tumor cells can occur, and this process requires only cIAP1/2 antagonist, lymphotoxin, and IFN-γ pretreated macrophages and cancer cells.

To quantify phagocytic events, we cocultured BMDMs with zsGreen-expressing 6694c2 cells, and treated with cIAP1/2 antagonist and lymphotoxin. zsGreen is highly resistant to lysosomal degradation and photobleaching (46). Upon coculture with tumor cells, macrophages were clearly observed with punctate zsGreen⁺ staining, indicating phagocytosis, although whether from live or dead tumor cells is not possible to distinguish (Fig. 6E). In some cases, the zsGreen signal was confined to a single phagosome and contained an intact, non-apoptotic nucleus, as identified by 4′,6-diamidino-2-phenylindole (DAPI) staining (Fig. 6E, white arrowhead). These rarer events were counted as live cell engulfment and may underestimate of the true percentage of live cell engulfment since they only capture early events while the tumor is still in a single phagosome. Both total phagocytosis and live cell engulfment rates were higher in cultures with cIAP1/2 antagonist and LTα1β2 combination treated macrophages (Fig. 6F and G). Confocal images taken through a z-stack and rendered into three dimensions show that macrophages clearly engulfed live tumor cells with intact nuclei (Fig. 6H).

Discussion

We have found that induction of non-canonical NF-κB signaling promotes anti-tumor immunity across a range of tumor models and induces T-cell dependent immune responses in settings where checkpoint blockade fails. Unlike current T cell-based immunotherapies, direct CD8 T cell recognition of tumor cell expressed MHC class I is not required. Rather, we have found that MHC class I negative tumors can be controlled in a T cell-dependent fashion via activation of tumoricidal macrophages (fig. S10. This mechanism of action inspired combination therapy with CD47 blockade, a strategy which induced curative responses in otherwise refractory preclinical models. This therapeutic strategy retains the benefit of tumor specificity while enabling responses to tumors that have evaded host immunity through loss of MHC class I.

Other strategies to overcome resistance to checkpoint blockade fall into three major approaches: 1) further augmentation of T cell priming, activation, or reversal of exhaustion; 2) activation of NK cells; and 3) targeting of immunosuppressive myeloid cells (4). For tumors such as pancreatic cancer that are poorly infiltrated by T cells and NK cells, neither of the first two approaches have proven fruitful (4, 27). Elimination or reprogramming of myeloid cells has some benefit, with inhibition of the CC- and CXC-chemokine receptors CCR2 and CXCR2 showing the most promise in mouse models and early phase clinical trials (47–50). However, targeting of myeloid cells alone is of short-term benefit, and is best used clinically in patients with locally advanced disease as a bridge to surgery (48). Long-term, durable responses require the antigen specificity of adaptive immunity, and indeed the rare long-term survivors of PDAC were shown to have tumor-specific memory CD8 T cells that persisted for life (51).

T cells have pleiotropic functions, and although CD8 T cells are best known for perforin/granzyme-mediated cytolysis, they are also capable of cytokine and chemokine secretion, functional properties which are not necessarily linked (52). For tumors that are well-infiltrated by CD8 T cells, direct cytotoxicity may be key to elimination of tumor cells. Indeed this model is supported by the multiple studies showing endogenous and checkpoint blockade-induced T cell responses that are dependent on MHC class I antigen presentation (1–3, 33, 34). However, direct cytotoxicity is not the only means by which CD8 T cells can influence tumor outcomes. Cytokine-mediated killing of tumor cells is the major pathway by which antigen loss variants can be eliminated (53, 54), and in tumors that are poorly infiltrated by CD8 T cells, cytokine secretion and repolarization of innate cells in the microenvironment may be critical. Neoantigen-specific CD4 effector T cells can induce potent tumor regression in mice and humans, suggesting that cytokine elaboration is an important property of tumor-specific T cells (55–58). IFN-γ is directly cytostatic, and the TNF family ligands TNF-α and TNF-related apoptosis-inducing ligand (TRAIL) can directly induce tumor cell apoptosis in several cancer types (59–61). We propose a model whereby T cells produce the cytokine lymphotoxin, which is spatially restricted to a small diffusion radius within the tumor microenvironment where it robustly induces phagocytic macrophages capable of engulfing and destroying live tumor cells. The antigen restriction is therefore maintained via spatial constraints that limit macrophage activity to a defined region. Similar spatial restraints have been proposed to regulate the bystander effects of IFN-γ (62, 63). Although we did not find a role for IFN-γ in our pancreatic cancer models, IFN-γ production is augmented by cIAP1/2 antagonism and may be relevant in other cancer types (14, 18, 22). We found that CD8 T cells were primarily required for the responses to cIAP1/2 antagonism in pancreatic cancer; however, it is possible that CD4 T cells or other sources of cytokines, such as targeted delivery or oncolytic viruses encoding cytokines, would also suffice in other tumor settings.

The heterotrimeric lymphotoxin complex LTα1β2 augments the effect of cIAP1/2 antagonism in vitro and in vivo, despite signaling pathways that both converge on increasing nuclear p52 (64, 65). It is possible that cIAP1/2 antagonism provides an insufficient or inconsistent degree of NF-κB signaling that can be maintained above a critical threshold in macrophage by further exposure to lymphotoxin. However, our observation that lymphotoxin was unable to induce phagocytosis on its own suggests a non-redundant role of cIAP1/2 antagonism. LCL161 induces transient hyperactivation of the E3 ubiquitin ligase function of cIAP1/2, leading to degradation of cIAP1/2 protein but simultaneously increasing activity of Myc (66). Myc activity has been linked to efferocytosis in macrophages, but whether it is necessary for engulfment of live tumor cells is unknown (67). cIAP1/2 antagonists may also act through canonical NF-κB signaling, thereby potentiating tumoricidal functions of macrophages as has been observed with agents that induce nuclear translocation of p65 (64, 68).

Recombinant cytokines, including IFN-γ and TNF-α, have been used in patients with cancer, but are limited by severe systemic toxicities (69). Local delivery of cytokines has shown some promise, including in models of pancreatic cancer (70), suggesting that T cells may be replaced by pharmacologic substitutes. However, cytokine-based therapies are limited by multiple factors including proper localization, the range of cytokines produced, and appropriate regulation of cytokine secretion. Induction of tumor-specific T cells to act as local sources of cytokine production is a superior alternative. cIAP1/2 antagonism in T cells lowers the threshold for TCR affinity at priming by mimicking co-stimulatory signaling pathways (14, 22). These newly primed T cells also show augmented production of certain cytokines, with both CD4 and CD8 T cells skewing toward increased IL-2 and TNF-α production, and CD4 T cells skewing toward increased IFN-γ production (18). Here we detected increased production of lymphotoxin from tumor-infiltrating T cells, showed that lymphotoxin is required for reduction of tumor burden. Further, we showed mechanistically that combination of lymphotoxin with cIAP1/2 antagonism converted quiescent macrophages into tumoricidal macrophages.

We found that co-expression of MHC class II and Ly6C identified phagocytic macrophages, suggesting they originated from infiltrating monocytes. In pancreatic tumors, Ly6C⁺ inflammatory monocyte-derived macrophages can be beneficial through depletion of extracellular matrix after treatment with agonistic anti-CD40 (28), but have also been implicated in immunosuppression following radiation treatment (71). Ly6C⁺ cells could be converted to an anti-tumor phenotype by exposure to the toll-like receptor 9 ligand, CpG (72). These results suggest that Ly6C⁺ monocyte-derived macrophages have functional plasticity, although no reports have yet linked this plasticity to antigen-specific T cells. Here, we showed that anti-tumor T cells augment phagocytosis of tumor cells by reprogramming Ly6C⁺ macrophages toward a tumor destructive phenotype. Importantly, this phagocytic repolarization occurs in human monocyte-derived macrophages as well, indicating conservation of the mechanism across both species.

Given a mechanism of action centered on phagocytosis, we added CD47 blockade as a potential combinatorial strategy. Although combination of cIAP1/2 antagonism with checkpoint blockade did not increase efficacy above cIAP1/2 antagonist monotherapy in poorly immunogenic tumors, combination of cIAP1/2 antagonism with CD47nb induced profound responses in a T cell-low model of pancreatic cancer. CD47 blockade is currently in clinical trials as a monotherapy and in combination with agents such as tumor-specific antibodies that can augment phagocytosis through engagement of Fc-receptors on macrophages (40, 73). CD47 blockade in tumors requires very high coverage of CD47, a strategy that is limited by the expression of CD47 on red blood cells (44). Combination of CD47 blockade with prophagocytic reprogramming, such as that induced by cIAP1/2 antagonism, could be a means of lowering the threshold of CD47 blockade required. Other strategies to augment phagocytosis or block immunoreceptor tyrosine-based inhibitory motif (ITIM) signaling in macrophages (74) may also combine with cIAP1/2 antagonists to improve antitumor responses. Given that β2M in humans binds to leukocyte immunoglobulin-like receptor B1 (LILRB1), an inhibitory receptor on macrophages, prophagocytic reprogramming may be even more effective in MHC class I negative tumors (75).

Our study has several limitations. Although we could show that human macrophages can be reprogrammed by cIAP1/2 antagonism and lymphotoxin to phagocytose human pancreatic cancer cells in vitro, most of our experiments were conducted in mice. Whether human T cells make lymphotoxin in response to cIAP1/2 antagonism is unclear, as is the reliance of human macrophages on T cell-derived cytokines in human cancer patients. Furthermore, our work clearly implicates pro-phagocytic receptor-ligand interactions between macrophages and tumor cells, although the identify of these pro-phagocytic signals remains unknown. A mechanistic understanding of how live tumor cells are recognized by tumoricidal macrophages will be important to identifying susceptible cancer types and predicting tumor cell resistance.

Loss of MHC class I is a central problem in cancer immunology and contributes to both primary and acquired resistance to checkpoint blockade therapies. All current strategies to overcome MHC class I loss do so while sacrificing antigen-specificity. Here we have shown that T cells, primed by cross-presenting dendritic cells, can induce tumor regression through local cytokine production and activation of tumoricidal phagocytes. Although we used cIAP1/2 antagonists to induce these properties, cIAP1/2 antagonists are likely not the only means by which T cell reprograming of phagocytes can be achieved, and we propose this as a potentially more general mechanism for controlling checkpoint blockade refractory tumors.

Materials and Methods

Study design

Previous studies from our lab and others have identified cIAP1/2 antagonism as a means of augmenting immune responses to tumors. However, their precise mechanism of action and whether these agents would effectively clear checkpoint blockade resistant tumors was unknown. Therefore, our primary objective was to rigorously evaluate the efficacy and mechanism of action of the cIAP1/2 antagonist LCL161 across several mouse models of poorly immunogenic cancer, including β2M^−/− tumor models. Efficacy of LCL161 was tested in subcutaneous MC38, MC38 β2M^−/−, 6694c2, and 6694c2 β2M^−/− models, as well as resistant cell lines derived from LCL161-treated 6694c2 tumors. We also evaluated efficacy of LCL161 in orthotopic 6694c2 and 6499c4 tumors as well as the lox-stop-lox-KrasG12D;p53+/flox;p48cre+ genetically engineered mouse model (KPC GEMM) of PDAC. All efficacy studies in subcutaneous models enrolled 5–10 mice per group and were repeated at least one. Efficacy studies in spontaneous KPC GEMM enrolled 8–14 mice per group and all data points from 2 experiments are shown. Both male and female KPC GEMM mice were enrolled and were randomized to treatment groups according to a pre-assigned order. Tumor ultrasound measurements were performed by a blinded investigator. Orthotopic experiments were performed using 5–10 mice per group and were sometimes compiled from multiple experiments as indicated in the figure legends. For orthotopic and subcutaneous tumor experiments, mice were randomized to treatment groups after tumor inoculation. Quantification of the tissue staining in Figure 4 was done manually by a blinded investigator.

Cell lines

KPCY cell lines 6694C2 and 6499C4 derived from a LSL-Kras^G12D;p53+/floxed, Pdx-cre, YFP-floxed mouse were previously described (31). Human PDAC cell line 8988T was a gift from Dr. Andrew Aguirre. MC38 cells were obtained from the American Type Culture Collection (ATCC) and modified in house using CRISPR/Cas9. Cells were cultured at 37°C in a humidified incubator with 5% CO₂. RPMI-1640 media was supplemented with 10% fetal bovine serum (FBS), 2 mmol/L L-glutamine, 1% penicillin/streptomycin, 1% minimal essential media (MEM) non-essential amino acids, 1 mmol/L sodium pyruvate, and 0.1 mmol/L β-mercaptoethanol. Cells used for in vitro experiments were cultured with 500nM LCL161 or 0.1% dimethyl sulfoxide (vehicle). Cytokines IFN-γ, TNF-α, LTα1β2 were purchased from Peprotech and used at the indicated concentrations. Cells used for in vivo experiments had been passaged for less than 2 months, were negative for known mouse pathogens, and were implanted at >95% viability. CRISPR modifications were performed using pSpCas9(BB)-2A-Puro (PX459) V2.0 vector that was a gift from Feng Zhang (Addgene 62988). Single guide RNAs (sgRNAs) were cloned into the vector by restriction enzyme digest (BbsI) and ligation with T4 DNA ligase. Cells were transfected with Lipofectamine Stem reagent and used according to the manufacturer’s protocol. Cells were selected with puromycin for successful transfection, FACS sorted for purity, and absence of protein was confirmed by flow cytometry.

Mice

All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the Dana-Farber Cancer Institute (DFCI) (protocol #14–019, 14–037, 10–055) and are in compliance with the NIH/NCI ethical guidelines for tumor-bearing animals. The following mouse strains were purchased from Jackson labs: C57BL/6 (000664), μMT^−/− (002249), Batf3^−/− (013755), TCRα^−/− (002116), β2M^−/− (002087), I-Ab^−/− (005589), RAG2^−/− (008449), Perforin^−/− (002407), OT-I (003831). TRP1^high mice were bred in house (76). LSL-Kras^G12D;p53+/flox,p48-cre mice (77) were bred in house as part of the Dana-Farber Cancer Institute Hale Center for Pancreatic Cancer Research. LSL-Kras^G12D;p53+/flox,p48-cre mice monitored weekly by palpation until tumor development was suspected as previously described (78). Mice were then shaved and received ultrasound monitoring to determine the size of the pancreatic tumor. Mice with ultrasound confirmed tumors greater than 4mm in diameter, defined as the longest dimension in any orientation, were enrolled into experiments according to a prespecified enrollment schedule.

Orthotopic pancreatic tumors

Orthotopic surgeries were performed as described (79). Briefly, C57BL/6 or RAG2^−/− mice were anesthetized with a ketamine/xylazine cocktail, shaved on the left flank, and the surgical site cleaned with ethanol and betadine. An incision was made in the skin and peritoneum, and the pancreas externalized with forceps. Panc02 or KPC cells were resuspended in phosphate buffered saline (PBS) and mixed 1:1 by volume with matrigel (Corning) for a total of 100,000 cells per 30 μL. The cell suspension was kept on ice and drawn into a chilled insulin syringe. Cells were then injected into the tail of the pancreas, and a bubble was observed. Mice that showed signs of leakage were removed from the experiment. The pancreas was left external to the body cavity for 1 minute with the mice on a warming pad to solidify the matrigel. The pancreas was then reinserted, peritoneum sutured with one stitch of absorbable suture, and the skin stapled with a sterile wound clip. Mice were given analgesia (Buprenex) and monitored post-surgery according to protocols approved by the Dana-Farber IACUC. Mice were euthanized 21 days post-surgery unless otherwise indicated. Tumors were weighed at the time of euthanasia.

Tumor infiltrate analysis

Pancreatic tumors were excised, weighed, minced and incubated in RPMI-1640 containing collagenase type IV (Sigma) and anti-trypsin at 37°C for 1 hour. Tumors were filtered through a 40 micron cell strainer, washed with PBS, and centrifuged. The resulting cell pellet containing tumor debris and infiltrating immune cells was resuspended in FACS buffer (PBS with 2% fetal calf serum) and stained with a master mix of antibodies. Cells were incubated with staining mix for 30 minutes at 4°C, washed once in PBS, and resuspended in 1% formalin prior to analysis on either a spectral flow cytometer (Sony SP6800) or a Fortessa cytofluorimeter (BD). Flow cytometry antibodies used in this study were purchased from BioLegend and used at 1:200 dilution (αCD45[30-F11], αCD4[RM4–5], αCD8[53–6.7], αNK1.1[PK136], αCD103[2E7], Ly6C[1A8], I-A/I-E[M5/114.5.2], F4/80[BM8], SigF[E50–2440], αCD11b[M1170] αCD11c[N418], αGR1[RB6–8C5], H2-Kb[AF6–88.5]). For one experiment, we used an ImageStream^x MkII Imaging flow cytometer (IFC) (Amnis Corporation) and IDEAS™ersion 6.2.64.0 software was used for compensation and data analysis.

In vivo treatments

LCL161 was provided by Novartis Pharmaceuticals and solubilized in 0.1 N HCL and diluted in sodium acetate buffer to a final concentration of 10 mg/ml and pH 4.3. Vehicle solution was made using HCL in sodium acetate buffer and adjusted to pH 4.3. LCL161 was administered by oral gavage at 75mg/kg in a total volume of 150 μL every 3 days starting at day 4 post tumor-inoculation unless otherwise indicated.

Mice received depleting antibodies on the day of tumor inoculation and then every 3 days until euthanasia. Depleting antibodies were given intraperitoneally at a dose of 100μg/mouse. Anti-PD-L1 [clone 10F.9G2] and/or anti-CTLA4 [clone 9D9] were administered intraperitoneally at a dose of 200 μg/mouse starting at day 4 post inoculation every 3 days. All antibodies were purchased from Bioxcell: αCD4[GK1.5], αCD8[2.43], αNK1.1[PK136], αCSF-1[5A1].

Expression of CD47nb and VHHcont

CD47nb (A4 clone) has been previously reported (43, 44). The A4 or VHHcont coding sequence was sub-cloned into the E.coli periplasmic expression vector pET22b, with the inclusion of a C-terminal sortase motif and His₆ tag. BL21(DE3) E. coli containing the plasmid was grown to mid-log phase at 37°C in terrific broth (TB) plus ampicillin, and induced with 0.5 mM Isopropyl β-d-1-thiogalactopyranoside (IPTG) overnight at 25°C. Cells were harvested by centrifugation at 5000xg for 15 minutes at 4°C, then resuspended in 25 mL 1x TES buffer (200 mM Tris, pH 8, 0.65 mM EDTA, 0.5 M sucrose) per liter culture, and incubated for 1 hour at 4°C with agitation. Resuspended cells were then submitted to osmotic shock by diluting 1:4 in 0.25x TES, and incubating overnight at 4°C. The periplasmic fraction was isolated by centrifugation at 8000rpm for 30 min at 4°C, and then loaded onto Ni-NTA (Qiagen) in 50 mM Tris, pH 8, 150 mM NaCl and 10 mM imidazole. Protein was eluted in 50 mM Tris, pH 8, 150 mM NaCl, 500 mM imidazole and 10% glycerol, then loaded onto a Superdex 75 16/600 column (GE Healthcare) in 50 mM Tris, pH 8, 150 mM NaCl, 10% glycerol. Recombinant VHH purity was assessed by SDS-PAGE, and peak fractions were recovered and concentrated with an Amicon 10,000 KDa MWCO filtration unit (Millipore), and stored at −80°C. To remove lipopolysaccharide (LPS), VHHs were immobilized on HisTrap HP 1 mL columns (GE Healthcare) in PBS, washed with 40 column volumes PBS + 0.1% TritonX-114, and eluted in 2.5 column volumes Endotoxin-free PBS (Teknova) with 500 mM imidazole. Imidazole was removed by PD10 column (GE Healthcare), eluting in LPS-free PBS. LPS content was tested using the LAL Chromogenic Endotoxin Quantitation Kit (Pierce) according to the manufacturer’s instructions.

Immunofluorescence

Tumors were resected from tumor-bearing mice and cut in half. The tissue was incubated in fixation solution [4% paraformaldehyde/20% sucrose] for 1 hour at room temperature and 5 hours at 4ºC. Tissue was frozen on dry ice in optimal cutting temperature (OCT) medium and stored at −80ºC. To obtain tissue slides, the tissue was sectioned to 0.08 μm with a Leica 3000 cryostat. After sectioning, slides were dried at 37ºC and permeabilized for 20 minutes at room temperature with 1 mM Triton-x100 in TBST before being blocked for 1 hour at room temperature with 50% SEA Block in PBS. Tissue was stained with antibodies to F4/80–AF647 at 1:100 dilution (BioLegend), GFP-AF488 at 1:200 dilution (Abcam) and DAPI at 1:10,000 dilution in 5% SEA Block. Images were taken using a fluorescent microscope. Three-dimensional rendering was performed using images from a confocal microscope at the Harvard Medical School MicRoN Imaging Core Facility.

Generation of Bone-Marrow Derived Macrophages

Bone marrow was flushed from C57BL/6 mouse femurs and erythrocytes lysed in hypotonic buffer. 50,000 cells were plated per well of a 48-well tissue-culture treated plate in 500 μL RPMI-1640 medium supplemented with 20 ng/mL M-CSF. After 3 and 6 days of culture, additional M-CSF containing media was added. After day 7, macrophages were confirmed by flow cytometry performed as described above to be >80% CD11b⁺F4/80⁺ and used for in vitro phagocytosis assays. Cells were dissociated by treatment with 1x TrypLE (Gibco) for 10–30 min at 37°C and vigorous pipetting.

CD8 T cell isolation and in vitro stimulation

CD8⁺ T cells were isolated from spleen and lymph nodes of OT-I mice using EasySep™ Mouse CD8+ T Cell Isolation Kit (StemCell catalogue # 19853) according to the manufacturer’s instructions. CD8 T cells were cultured in RPMI-1640 medium with anti-CD3/CD28 Dynabeads (ThermoFisher) for 48 hours. Supernatants were harvested. Cytokines and chemokines were analyzed by cytokine bead array (Discovery Assay 31-plex mouse cytokine and chemokine panel, Eve Technologies).

Time-lapse microscopy

BMDMs were plated onto glass-bottomed imaging dishes (MatTek Corporation) and cultured for 24 hours as described above. Cells were washed with Hank’s buffered salt solution (HBSS) and cocultured with YFP-expressing 6694c2 cells and Sytox blue in RPMI-1640 complete medium (ThermoFisher Scientific). The cells were then imaged using a Nikon Ti inverted microscope equipped with a Nikon linear encoded motorized stage, an Andor Zyla 4.2 sCMOS monochromatic camera, a Lumencor Spectra X LED light engine and a Plan Apo 20x/0.8 DIC objective lens. Imaging was performed at 37°C and 5% pre-mixed CO2 using a custom-built incubator enclosure and an OkoLab gas chamber. Fluorescence and transmitted light images were acquired sequentially every 2.5 min for 288 frames (12 hour elapsed time) using NIS Elements 4.30 software. Fluorescence from YFP and Sytox Blue was collected using Chroma filter cubes 49003 and 49000, respectively. A neutral density filter (ND4) was inserted in the lightpath to minimize photobleaching and phototoxicity. Movies were converted with Image J software and edited with Adobe Premiere Pro 2020.

For still immunofluorescence images, tumor-macrophage cocultures were incubated for the indicated number of hours, then fixed and stained with anti-CD11b Alexa Fluor594 and DAPI. Phagocytosis rates and intact nuclei were counted by a blinded investigator using >100 macrophages per condition.

Tumor confluency

A Celigo Image Cytometer (Nexcelom 200-BFFL-5c) was used to monitor cell growth of cell lines in conditioned media over 5 days. Cells were seeded in 96 well plates and treated with TNF-α (Peprotech) in combination with 500 nM LCL161. Confluence was measured.

Single-Cell RNA sequencing and data analysis

CD45⁺ cells were sorted by FACS from orthotopic 6694c2 pancreatic tumors implanted in C57BL/6 mice treated with vehicle or LCL161 for 12 days. N=5 mice per group were pooled prior to sorting. One library from each condition was then constructed using the Chromium Single Cell 5’ Kit (10x Genomics PN-1000006). Libraries were sequenced on an Illumina HiSeq 2500 system at the DFCI Center for Cancer Genome Discovery. The 10x CellRanger software (v3.0.2) was used to demultiplex base call files, align reads to the mm10 reference genome, and generate a single-cell feature count matrix for each library using default parameters. The count matrices were imported for downstream analysis into R using the “Seurat” package (v3.1.4) (80). Genes expressed in fewer than 3 cells were discarded from further analysis. Barcodes were classified as cells if they satisfied the following criteria: reads detected in greater than 100 distinct genes, percentage of mitochondrial reads less than 2 standard deviations (SDs) from the mean, and total reads within 3 SDs of the mean. Then, counts were log-normalized, scaled, and subject to dimensionality reduction using Principal Component Analysis (PCA) based on the 2,500 most variable genes. To visualize results in two dimensions, the top 19 PCs were used to generate a t-distributed Stochastic Neighbor Embedding (t-SNE) plot (perplexity = 60). Unsupervised clusters were identified first by constructing a Shared Nearest Neighbor (SNN) graph based on each cell’s 20-nearest neighbors and then applying modularity refinement with the Louvain algorithm. Another round of unsupervised clustering restricted to cells within the B cell and T cell clusters identified distinct sub-clusters that expressed macrophage markers and were manually labeled as such. Markers for each cluster were identified by comparing gene expression using Model-based Analysis of Single-cell Transcriptomic (MAST) (81).

Bulk RNA sequencing and data analysis

CD45⁺ CD4⁺ or CD8⁺ cells were sorted by FACS from orthotopic pancreatic tumors implanted in C57BL/6 mice treated with vehicle or LCL161 for 12 days. Total RNA was prepared using Qiagen RNeasy kit according to the manufacturer’s protocol. Library construction and Illumina sequencing were performed by the DFCI Molecular Genomics Core Facility reads were trimmed for adapter sequences using the tool “cutadapt” (v2.9) then aligned to the GRCm38 genome with the corresponding ENSEMBL 97 annotation using “STAR” (v2.7.0f). Feature counting was performed using the R package “Rsubread” (v1.32.4), allowing for multi-mapping and multi-overlapping reads.

Statistical analysis

All tumor weight and tumor infiltrates data are presented as mean with SEM error bars unless otherwise noted. When two groups were being compared, significance was determined using a two-sided Mann-Whitney test to compare ranks, without assuming Gaussian distribution. For three or more groups, ordinary ANOVA with Sidak’s multiple comparison test was performed. GraphPad Prism software was used to analyze data. Exact p values are reported in each figure panel. The number of individual data points shown in each graph represent independent biological replicates. For some figures, data from multiple experiments were combined for presentation in the same graph. In these cases, tumor weights were normalized to the average of the control (vehicle) group within each experiment and data are presented as “normalized tumor weights” on the y-axis.