T cell immunotherapies engage neutrophils to eliminate tumor antigen escape variants

Summary

Cancer immunotherapies, including adoptive T cell transfer, can be ineffective because tumors evolve to display antigen loss variant clones. Therapies that activate multiple branches of the immune system may eliminate escape variants. Here we show that melanoma-specific CD4+ T cell therapy in combination with OX40 co-stimulation or CTLA-4 blockade can eradicate melanomas containing antigen escape variants. As expected, early on-target recognition of melanoma antigens by tumor-specific CD4+ T cells was required. Surprisingly, complete tumor eradication was dependent on neutrophils and partly dependent on inducible nitric oxide synthase. In support of these findings, extensive neutrophil activation was observed in mouse tumors and in biopsies of melanoma patients treated with immune checkpoint blockade. Transcriptomic and flow cytometry analyses revealed a distinct anti-tumorigenic neutrophil subset present in treated mice. Our findings uncover an interplay between T cells mediating the initial anti-tumor immune response, and neutrophils mediating the destruction of tumor antigen loss variants.

Graphical Abstract

Introduction

Immune surveillance eliminates cancers in early stages of malignant transformation. Selective pressure imposed by the immune system on tumor cells can “edit” tumors, yielding immune escape variants^1,2. Many tumors contain clones that cease to express antigenic proteins. Antigenic heterogeneity is therefore a mechanism by which tumors become refractory to immunotherapeutic interventions like adoptive cell therapies³.

Differentiation antigens are well-characterized tumor immunotherapeutic targets⁴. These are proteins shared with healthy tissues important for specialized biological processes. Both cancer patients and healthy individuals contain T cells that can recognize the melanoma differentiation antigens, associated with melanin synthesis^5,6. As such, immune destruction of melanocytes in melanoma patients undergoing immunotherapy can be associated with a favorable prognosis^7–9.

Current immunotherapies consist of strategies that mobilize the adaptive immune compartment. These include immune checkpoint blockade (ICB)^10–14, infusion of ex vivo educated or genetically modified tumor-specific T cells^15–20, and anti-tumor immunization²¹. Moreover, agents that target tumor necrosis factor receptor (TNFR) superfamily members expressed on the surface of activated T cells, such as glucocorticoid induced TNF receptor (GITR), 4–1BB and OX40 are being investigated²².

OX40 engagement leads to increased effector T cell function and reduction and/or deactivation of regulatory T cells^23–25. We and others have shown that agonistic anti-OX40 antibodies, in combination or as a monotherapy, can promote potent anti-tumor immunity pre-clinically^26–30. Consequently, OX40 agonists are being tested in clinical settings^31–34.

The complex role of neutrophils in tumor promotion versus tumor elimination is not well understood. The pro-tumorigenic role for neutrophils in chronic inflammation has been well described in multiple studies^35–39. Higher neutrophil-to-lymphocyte ratios are associated with deleterious outcomes in patients receiving ICB^40,41. Moreover, neutrophil effector mechanisms including the formation of extracellular traps (NETosis) can promote metastases and shield tumors from effector T cell elimination^42–44. In contrasting models, neutrophils and NETs were demonstrated to directly kill cancer cells^45–49.The role of neutrophils as potential effectors in the context of immunotherapy remains incompletely defined. This is particularly true for the role of neutrophils as effectors in interventions that solely target T cells.

We reported that combining cyclophosphamide (CTX), CD4+ T cells specific for the melanoma antigen Trp1 (Trp1 T cells)⁵⁰, and an anti-OX40 agonist antibody⁵¹ leads to elimination of advanced mouse melanoma tumors²⁶. Moreover, the Trp1 antigen-specific therapy eradicated antigen-loss variant clones in melanomas expressing Trp1 heterogenously. Here we show that tumors of treated mice have extensive infiltration with activated neutrophils, and similar infiltrates were observed in human patients after ICB therapy. Importantly, neutrophil depletion abrogated therapeutic efficacy in our mouse model of antigenic heterogeneity. We found that inducible nitric oxide synthase is important for neutrophil-mediated elimination of escape variants. Moreover, neutrophils from tumors in treated mice have a distinct transcriptomic and protein surface marker phenotype. These findings uncover a fundamental anti-tumor role for neutrophils eliminating antigen loss variants in the context of T cell immunomodulation.

Results

Potent T cell immunotherapies can eradicate tumor antigen loss variants

Tumors are genetically and phenotypically heterogeneous. We analyzed publicly available single cell sequencing data from melanoma patients⁵². The t-distributed stochastic neighbor embedding (t-SNE) plots in Figure 1A, depicting the transcriptional profiles from 15 pooled patients, show that malignant cells express different combinations of melanoma differentiation antigens. While some transcripts such as PMEL, TYR, and MLANA were present in most cells, transcripts for other antigens, such as GAGE1, ERBB2, and CTAG1B were rare. We next investigated whether intratumoral antigenic heterogeneity exists within a given individual tumor. We therefore analyzed cells from tumors of individual patients and visualized expression of the prototypical melanosomal differentiation antigen TYRP1 by t-SNE plots (Figure 1B).

Figure 1. Potent immunotherapies can eliminate antigen escape variants.

A) tSNE plots of transcripts of human melanoma antigens from 15 patients described in Tirosh et al⁵². B) tSNE plots of melanoma cells of 3 individual patients showing TYRP1 levels. C) Melanoma biopsies (n=10) stained for TYRP1, S100, and DAPI. A Z-score was calculated using the fraction population of double positive TYRP1 and S100. Bar =100μm. D) Summary of a previously published experiment of a model of antigenically heterogeneous tumor²⁶. Green check marks are long term survival of the applied therapy while red X marks are cohorts that succumbed to tumor burden. E) Mice were implanted with B16:B78H1 mixture in the same flank and treated as described in the methods. Growth curves of individual mice over time Left and Kaplan-Meier Survival plots Right are shown. Inset numbers indicate the number of mice that succumbed to tumor burden. ***P < 0.001, ****P < 0.0001. Mouse experiments were repeated at least twice with equivalent results.

To further characterize antigenic heterogeneity, we performed immunofluorescence analysis of melanoma biopsies. We stained 10 specimens for TYRP1, a late melanocyte differentiation marker⁵³ and S100, an early melanocyte development marker, widely expressed and used for the identification of melanoma. To determine the degree of heterogeneity in each tumor, we calculated the percentage of TYRP1 signal relative to S100 for each biopsy. As shown in Figure 1C, tumors were highly heterogeneous, expressing a wide range of TYRP1 antigen quantities.

Tumors expressing low levels of potentially immunogenic antigens can be predicted to be associated with reduced immunotherapeutic efficacy. In line with this reasoning, we analyzed TYRP1 mRNA gene expression in a cohort of 40 patients treated with anti-CTLA-4⁵⁴. The patients were stratified based on high or low TYRP1 mRNA levels. A trend for decreased overall survival was observed in patients with lower TYRP1 mRNA expression (Supplemental Figure 1A). Interestingly, untreated melanoma patients whose tumors exhibit high TYRP1 expression showed decreased survival (Supplemental Figure 1B, El Hajj et al.⁵⁵, and Journe et al.⁵⁶). We thus hypothesized that antigen density or variegated antigen expression may impact the efficacy of immunotherapeutic interventions.

Since many cancers exhibit antigenic heterogeneity, we sought to investigate if a potent immunotherapy can eliminate antigen escape variants using our previously established mouse model of tumor heterogeneity²⁶. Here, B16 melanomas (expressing Trp1, the mouse homolog of TYRP1) were co-implanted with B78H1 melanomas that do not express Trp1, MHC Class I, or MHC Class II (Supplemental Figure 1B and Graft et at.⁵⁷). Mice bearing large heterogeneous tumors (B16 + B78H1) were preconditioned with cyclophosphamide (CTX) followed the next day by adoptive transfer of Trp1 T cells. Mice were then treated with anti-OX40 antibody (Figure 1D). B16 alone, B78H1 alone, or B16 and B78H1 separately implanted on opposite flanks of the same mouse served as controls. The triple combination therapy (CTX + Trp1T cells + anti-OX40 antibody) effectively eliminated B16 tumors but had no effect on the Trp1-negative B78H1 tumors. When both tumor lines were implanted on opposite flanks, B16 tumors regressed but B78H1 continued to grow. Interestingly, heterogeneous B16:B78H1 tumors were eliminated by the triple combination therapy. These results show that potent immunotherapies can reject heterogeneous tumors even when not all cancer cells express the target antigen. The substitution of anti-OX40 with anti-CTLA-4 in the triple combination also showed efficacy eradicating heterogenous B16:B78H1 tumors, albeit to a lesser degree compared to anti-OX40 treatment (Figure 1E). These results further demonstrate the translational potential of the triple combination therapy in eliminating heterogeneous tumors containing antigen escape variants.

Elimination of tumor antigen escape variants does not require the endogenous adaptive immune system

We next sought to understand the mechanism underlying the elimination of the antigen loss variant B78H1 cells in the heterogeneous B16:B78H1 tumors. Using luciferase-based bioluminescence imaging, we tracked Trp1 T cells (Trp1-Luc T cell) and observed that their peak infiltration over time did not correlate with maximal tumor regression (Figure 2A). We used homogenous B16 tumors to maximize antigen expression, optimize signal and avoid any confounding effects during T cell tracking experiments. Tumors continued regressing after the T cell infiltration peak had subsided.

Figure 2: Elimination of antigen escape variants does not require the adaptive immune system.

A) Mice were implanted with B16 cells. After 3 weeks, mice were injected with CTX. The next day mice were injected with Trp1-Luciferase (Trp1-Luc) cells and anti-OX40. Photon flux was calculated (p/s) on a selected region at the tumor site. Black line represents tumor size over time (mm²). Red line represents Trp1 infiltration over time. Symbols represent mean ± s.e.m. Representative images are shown. B) WT or MHC Class II KO mice were implanted with B16:B78H1 mixture. After 3 weeks, mice with injected with CTX, the next day, the mice were treated with Trp1 cells and anti-OX40. Cohorts of mice were injected 3X weekly with anti-MHC class II antibody. Growth curves of individual mice are depicted. C) Mice were treated as described. Three or 13 days after treatment, cohorts of mice received anti-CD4 antibody 3X weekly. Growth curves of tumors of individual mice are depicted. D) WT or Rag^−/− mice were treated as described. Growth curves of individual mice are shown. Inset numbers indicate the number of mice that succumbed to tumor burden. Experiments were repeated at least twice with equivalent results.

To test the contribution of Trp1 T cells in mediating tumor elimination, we administered the triple combination therapy (CTX, Trp1 cells, anti-OX40) to tumor-bearing MHC class II deficient mice, which are unable to adequately prime CD4+ T cells in vivo. Additionally, we injected tumor-bearing MHC class II deficient mice with anti-MHC class II blocking antibody to fully prevent in situ Trp1 priming that could be elicited by expression of MHC class II on B16 tumor cells. The triple combination therapy was ineffective in promoting anti-tumor immunity in the absence of MHC class II on tumor or host cells (Figure 2B). These results suggest that adequate priming of adoptively transferred Trp1 T cells is necessary to induce anti-tumor immunity in the triple combination therapy protocol.

The infiltration kinetics suggested that Trp1 T cells are required at an early stage of tumor eradication (Figure 2A). Depleting Trp1 T cells at an early (day 3) time point, significantly diminished the anti-tumor efficacy of the triple combination therapy. However, later (day 13) depletion did not affect therapeutic efficacy (Figure 2C). To examine the contribution of the endogenous adaptive immune system to antitumor efficacy, we administered the treatment to RAG^−/− recipients, which are deficient in mature lymphocytes. The triple combination therapy was effective in eliminating antigenically heterogenous tumors in RAG^−/− mice (Figure 2D). Collectively, these findings suggest that additional immune cell populations of innate origin are responsible for complete heterogeneous tumor elimination.

Innate immunity is responsible for elimination of tumor antigen escape variants

To further understand mechanisms responsible for the late-phase tumor regression, we quantified cytokines from tumor extracts by multiplex analysis, and surprisingly observed a predominant innate signature (high levels of GM-CSF, IL-3, MIP-1α, and MIP-1β) in tumors from mice treated with our triple combination therapy (Figure 3A), despite the therapy’s primary targeting of T cells. Hematoxylin & eosin-stained tumor sections from treated mice revealed the presence of neutrophils; confirmed with positive Ly6G and CD11b staining (Figure 3B). Substituting anti-OX40 with anti-CTLA4 in our triple combination protocol led to an increased neutrophilic infiltration, albeit at lower levels than anti-OX40 (Figure 3B). This result was in line with observed tumor control results as shown in Figure 1E. Increased neutrophil infiltration was also confirmed by flow cytometry (Figure 3C). Neutrophil activation can lead to multiple effector functions, including the release of neutrophil extracellular traps (NETs) at the later stages of activation⁵⁸, marked by myeloperoxidase (MPO) and citrullinated histone-3 (CitH3). Fluorescence microscopy revealed increased neutrophils undergoing NETosis in tumors of treated mice, in line with the efficacy of each treatment regimen (Figure 3D). Overall, we found that immunotherapies that specifically target T cells also recruit and activate neutrophils.

Figure 3: Combination therapy increased activated neutrophils in the tumor.

Mice (5/group) were implanted with B16 cells and treated as described. After 10 days, tumor extracts were prepared and subjected to cytokine analysis. A) Volcano plot showing log fold change (FC) in cytokines concentration relative to the control vs. p-values (P). Dotted line represents P =0.05. B) Mice (5/group) were implanted with B16 cells. After 3 weeks, mice were treated as described. After 10 days, tumors were dissected and analyzed by Immunohistochemistry. Representative slides showing H&E, CD11b, and Ly6G staining. Bar= 80μm. Box-and-whisker plots are shown. 4–5 representative high-power fields from 5 tumors per condition were analyzed. Plots represent %CD11b area and Ly6G+ cells/mm². C) Mice (3–4/group) were treated as described. Single cell suspension was prepared from tumors at day 10 and analyzed by flow cytometry. Representative flow plots are shown for events gated on live CD45⁺CD11b⁺Ly6G^highLy6C^mid. Percentage of neutrophils/CD45+ cells ± s.e.m are quantified per treatment. Each symbol represents a mouse. D) Mice were treated as described. Left, representative micrographs of IgG and anti-OX40 treated mice groups where NETs are indicated by yellow arrows and MPO+ neutrophils are indicated with white arrows. Bar= 50μm Right, NETs were quantified by integrated intensity of Cit-H3 and MPO and the ratio of CitH3 to MPO was calculated. Box-and-whisker plots are shown. Two-five representative high-power fields from 5 different tumors per condition were analyzed. **P < 0.01, ***P < 0.001, ****P < 0.0001. Experiments were repeated at least twice with equivalent results.

Neutrophils are responsible for bystander killing of tumor antigen escape variants

We then asked if the tumor microenvironment in mice treated with the triple combination therapy is favorable to neutrophil activation (assessed by NET formation) by exposing naïve neutrophils to tumor extracts from mice receiving the triple combination therapy. Tumor extracts from mice treated with the triple combination therapy induced extensive NETosis in vitro, comparable to phorbol myristate acetate (PMA) stimulation, a classical inducer of NETosis (Supplemental Figure 2A). We then asked if the same tumor extracts can program neutrophils to kill escape variants. Naïve neutrophils exposed to tumor extracts from treated mice lead to increased killing of target B78H1 (Trp1 negative) cells in a clonogenic killing assay⁵⁹. Tumor extracts from mice treated with the combination therapy could program neutrophils to kill B78H1 cells ex vivo (Figure 4A). Close examination of Trp1 T cells and neutrophils in tumors of treated mice by immunohistochemistry showed no co-localization of Trp1 T cell and neutrophils, suggesting independent anti-tumorigenic mechanisms (Supplemental Figure 2B).

Figure 4: Neutrophils kill antigen escape variants.

A) Mice (5/group) were implanted with B16:B78H1 mix and treated as described. Neutrophils were cocultured with B78H1 cells in the presence of tumor extracts for 18h. B78H1 killing was measured by clonogenic assay at 1:1 target:effector ratio Left and flow cytometry measuring active caspase 3/7 at 1:10 target:effector ratio Right. Average ± s.e.m. was calculated from 4–6 different wells per condition. B) Trp1 cells were activated in culture with peptide and irradiated APCs. The next day, anti-OX40 or IgG was added to the cultures. After 6 days, the Trp1 cells were cocultured with B16 cells in a transwell assay at 1:10 target:effector ratio. On the top chamber, naïve neutrophils were cocultured with B78H1 for 16h. B78H1 killing was analyzed by flow cytometry. A diagram of the experimental setup is shown on the Left. Graphs represent the average ± s.e.m. % of active caspase 3/7 of 4 different wells per condition Right. C) Mice were treated as described. Cohorts of mice were injected with anti-Ly6G 3X weekly starting at day 3. Growth curves of individual mice Left, and Kaplan-Meier survival curves pooled from 2 independent experiments are depicted Right. Inset indicate the number of mice that succumbed to tumor burden. **P < 0.01, ***P < 0.001, ****P < 0.0001. Experiments were repeated at least twice with equivalent results.

Our results so far suggested that activated Trp1 T cells infiltrate tumors, eliminate antigen expressing clones (B16), and reprogram neutrophils to eliminate antigen loss variants (B78H1). To test this, in vitro activated Trp1 cells co-cultured with B16 cells were able to educate neutrophils to kill B78H1 cells in a contact-dependent manner (Figure 4B). Notably, when neutrophils were exposed to conditioned media (supernatant from in vitro activated Trp1 T cells), they promoted B78H1 killing, supporting the observation that secreted factors educate neutrophils to kill antigen escape variants (Supplemental Figure 2C).

Finally, we asked if neutrophils were essential in eliminating antigen escape variants in vivo. Mice bearing B16:B78H1 heterogenous tumors were treated with the triple combination therapy and depleted of neutrophils by anti-Ly6G treatment. Depletion was confirmed at day 10 after anti-Ly6G administration (Supplemental Figure 3A and B). Neutrophil depletion led to initial tumor regression, but at later time-points almost all mice succumbed to tumors (Figure 4C). The cytokine and chemokine profiles in Figure 3A are not specific for neutrophil recruitment and activation, and additional innate populations could be responsible for heterogeneous tumor elimination. We therefore depleted monocytes, macrophages, or NK cells in triple combination treated RAG^−/− mice bearing heterogeneous tumors. Only neutrophil depletion abrogated the anti-tumor responses of the triple combination treatment (Supplemental Figure 3C). Overall, these results suggest that neutrophils are necessary for eliminating tumor escape variants in T cell immunotherapies.

Transcriptomics and flow cytometry reveal a distinct anti-tumor neutrophil signature

Once we established that neutrophils are necessary for the elimination of antigen loss variants with the triple combination therapy, we asked if they also acquire a distinctive anti-tumorigenic phenotype. To address this, we analyzed sorted CD45+ Ly6G+ neutrophils from mixed B16:B78H1 tumors of treated mice by bulk RNA-seq. Principal component analysis and volcano plots of differentially expressed genes showed that neutrophils from tumors of treated mice adopt a distinct transcriptional signature (Figure 5A). To further characterize expression signatures and understand the neutrophil heterogeneity with higher resolution, we performed single cell RNA sequencing (scRNA-seq) coupled with cellular indexing of transcriptomes and epitopes by sequencing technique (CITE-seq) on tumor-derived myeloid cells of treated or control mice. We FACS-sorted myeloid CD45+ CD11b+ cells from tumors of individual mice treated with the triple combination therapy (CTX + Trp1+ anti-OX40), control therapy (CTX + Trp1 + IgG), or untreated mice. After rigorous quality control (Supplemental Figures 4A–D and methods), we projected the data into two dimensions using Uniform Manifold Approximation and Projection (UMAPs)⁶⁰. To address outlier event biases, we used hashtag oligomers to identify cellular components of individual mice (Supplemental Figure 4E). Unsupervised clustering revealed 18 clusters of cells (Figure 5B and C, Supplemental Figure 4F, and Supplemental Table 1). CITE-seq can detect surface proteins on clusters derived from transcriptomic data that is useful for identifying subsets cell populations (Figure 5D). We found that tumors from untreated mice contained mostly monocytes (Ly6C), macrophages (F4/80), and dendritic cells (CD11c), while tumors from mice treated with the CTX + Trp1 + anti-OX40 or IgG combination contained mostly neutrophils (Ly6G) (Figure 5D). This was validated by Immunological Genome (ImmGen) analysis of known signatures⁶¹ (Supplemental Figure 4G).

Figure 5: Tumor neutrophils from the triple combination therapy show a distinct transcriptional phenotype.

A) Mice were implanted with B16:B78H1 mix and treated as described. After 10 days, single-cell suspensions were prepared from tumors. Live⁺CD45⁺Ly6G⁺ cells were FACS-sorted, bulk RNA was prepared, sequenced, and analyzed. Left, PCA analysis of neutrophils from each mouse per treatment. Right, volcano plots depicting up and down-regulated genes per treatment. Lines are FC (Fold change) =2 and P (P-value) =0.05). B-D) Mice were implanted with B16:B78H1 mix and treated as described. A separate cohort of mice was implanted with B16:B78H1 mix a week later as untreated controls. After 10 days, single-cell suspensions were prepared from tumors. Live CD11b⁺CD45⁺ were sorted and analyzed by scRNA-seq as described in the methods. B) UMAP of each group showing different clusters. C) Heat map showing top genes expressed by each cluster (Refer to Supplemental Table 1). D) Detection of selected proteins derived from CITE-seq analysis on different regions of the UMAP.

Given that so few neutrophils were present in tumors from untreated mice compared to treatment groups and to further define differences in neutrophils important for tumor elimination, we reanalyzed the data derived from either CTX + Trp1 T cells + IgG or CTX + Trp1 T cells + anti-OX40 groups (thereafter called IgG or anti-OX40). After analysis and re-clustering, we obtained 6 clusters. Approximately 50% of the cells in the IgG group were enriched in Cluster 0 while about the same proportion of cells in the anti-OX40 group were enriched in Cluster 1 (Figure 6A). Given that the anti-OX40 treatment yielded anti-tumorigenic neutrophils, further characterization of the enriched clusters was warranted. In Cluster 1 we observed distinct transcriptional signatures (Figure 6B and Supplemental Table 2) with genes consistent with increased neutrophil function, activation, and chemotaxis including Il1b, Osm, Plaur, Socs3, and Ccl4. Interestingly, the most upregulated gene, G0s2, is expressed in activated mature neutrophils⁶². Cluster 1 also showed an enrichment in gene ontology (GO) pathways including chemotaxis, activation, NO biosynthetic process, and response to Il1β (Figure 6C and Supplemental Table 3). The transcriptional levels of Cxcr2 versus Cxcr4 in regions of the UMAP (Figure 6D), suggest that Cluster 1 is more closely associated with mature differentiated neutrophils. To further characterize differences in these 2 clusters, we searched for gene signatures of pathways important for neutrophil function. The regions of the UMAP of Cluster 1 was enriched with gene signatures that include chemotaxis, neutrophil killing, apoptosis, and NETosis, consistent with more mature and activated phenotypes (Figure 6E). Regions in the UMAP belonging to Cluster 0 were enriched for genes encoding for early granules proteins such as azurophil and gelatinase, substantiating a more immature phenotype^63,64. In addition, Cluster 1 neutrophils showed an increased apoptotic and NETosis signature. It should be noted, however, that an increase in RNA and protein expression may not necessarily reflect elevated functions and may reflect lower cell maturity given that azurophilic granule expression peaks early during differentiation.

Figure 6: Tumor neutrophils from treated mice acquired a more mature anti-tumorigenic phenotype.

A-F) Analysis of scRNA-seq comparing IgG and anti-OX40 groups. A) UMAP from IgG and anti-OX40 groups are depicted Left. Pie charts showing the proportion of each cluster per treatment and the number of cells analyzed Right. B) Heat map showing top genes expressed by each cluster (Refer to Supplemental Table 2). C) Dot plots of selected enriched terms from GO overrepresentation analysis. The circle represents the counts of enriched genes in the signatures. P value is shown in the legend (refer to Supplemental Table 3). D) Expression of Cxcr2 and Cxcr4 transcripts projected into the described UMAP. E) Expression of selected GO or published gene signatures¹⁰¹ projected into the described UMAP. F) Cell trajectory and pseudotime inference for the neutrophil populations. G) Heatmaps of pseudo-temporally expressed genes in each curve showing the top 50 most variable genes in each lineage. H-I) Mice were treated as described and single cell suspensions were analyzed by flow cytometry (refer to Supplemental Figure 5B). H) Log fold change of median fluorescence intensity relative to the average for each marker of gated neutrophils pooled from 3 independent experiments. Differences between markers were P<0.05 or lower. I) Flow plots and quantification of selected markers gated on neutrophils. Bars represent averages ± s.e.m. Symbols represent individual mice. J) WT or iNOS KO mice were implanted with B16:B78H1 mix and treated as described. Growth curves of individual mice Left, and Kaplan-Meier survival curves pooled from 2 independent experiments. Right. Inset numbers indicate the number of mice that succumbed to tumor burden. K) WT or iNOS KO mice (5 mice/group) were implanted with B16 tumors and treated as described. At day 10 after treatment, Ly6G+ neutrophils were FACS-sorted from pooled tumors. Neutrophils were co-cultured with B78H1 cell for 18h. Killing of B78H1 cells was measured by clonogenic assay. Average ±s.e.m. was calculated from 4–6 different wells per condition. *P < 0.05, P < 0.01, *** P < 0.001, **** P < 0.00001. Flow cytometry and mouse experiments were repeated at least twice with equivalent results.

To understand the evolution of both clusters and to test if neutrophils from Cluster 0 differentiated into Cluster 1, we performed trajectory analysis. Interestingly, Cluster 0 and 1 diverge into different trajectories from a common progenitor cell population that corresponded to Trajectory 1 and 2 (Figure 6F). Multiple genes including Lcn2, Lyz2, Ngp, Camp, S100a8, and S100a9, are shared at the beginning of both trajectories (Figure 6G). However, genes including chemokines important for chemotaxis (Cxcl3, Ccl4, Ccl3, Cxcl2), transcription factors involved in neutrophil maturation (Cebpb, Jun) responses to interferon (Isg15, Rsad2) and effector functions (Il1b, Fth1, Hilpda) were highly expressed at the later stages of Trajectory 2^65–67. The trajectory analysis revealed divergence of clusters and confirmed an enhanced effector function of Cluster 1 enriched in the anti-OX40 group.

To assess the clinical relevance of the observed gene signatures derived from the triple combination therapy, we stratified melanoma patients from the SKCM TCGA dataset based on Cluster 0 and Cluster 1 signatures. Patients with tumors showing an elevated Cluster 1 signature had favorable survival compared to patients with lower Cluster 1 signature expression. These results suggest clinical applicability of the signatures obtained by the scRNA-seq experiment (Supplemental Figure 5A).

Lastly, to validate some of these observations and further characterize tumor neutrophils, we performed flow cytometry on tumor infiltrates from treated mice. We selected markers of neutrophil function based on our transcriptomic analysis and published reports^64,68,69. In addition to established lineage markers (Supplemental Figure 5B), cell suspensions were stained for markers important for neutrophil function. Neutrophils from tumors exposed to the anti-OX40 combination treatment expressed higher levels of activation markers including CD14, CD117, IL1Rβ, CD18, CD40, CD177, SiglecF, CD49d, and CD16 (Figure 6H). Additionally, neutrophils from tumors in the anti-OX40 combination treatment group downregulated CD62L and CD206, consistent with a more activated neutrophil state. Interestingly, CD11b and CD24 were also downregulated. In agreement with the scRNA-seq data, we found a decrease of CXCR2+ CXCR4- in the anti-OX40 combination treatment group (Figure 6I). PD-L1 expression on neutrophils has been associated with a pro-tumorigenic and suppressive phenotype⁷⁰. We observed that the neutrophils derived from tumors treated with the combination therapy express lower levels of PD-L1 (Figure 6I). Lastly, mature neutrophils (SiglecF+) derived from tumors of treated mice show higher levels of inducible nitric oxide synthase (iNOS), which can enhance cytolytic potential (Figure 6I). Overall, the results from the transcriptomic and flow cytometry experiments suggest that the combination therapy promotes a distinct neutrophil signature characterized by a functionally mature and activated phenotype with anti-tumorigenic potential.

Neutrophils kill tumor antigen escape variants via an iNOS-dependent mechanism

The production of reactive oxygen species is a common mechanism by which neutrophils eliminate pathogens and tumors^71,72. Given that neutrophils from tumors of treated mice expressed elevated levels of iNOS, we hypothesized that reactive oxygen species eliminate tumor escape variants. When we treated iNOS deficient mice (iNOS KO) bearing heterogenous B16:B78H1 tumors with our triple combination therapy, we observed a substantial reduction in survival (Figure 6J). Clonogenic killing assays were performed using a co-culture of Ly6G+ neutrophils, FACS-sorted from B16 tumors of WT or iNOS KO mice treated with triple combination therapy, and B78H1 target cells. Neutrophils from iNOS KO mice were not as effective in lysing tumors as WT neutrophils (Figure 6K). Collectively, these results demonstrate that neutrophils kill heterogeneous tumors, in part, via iNOS. NETs derived from neutrophils have been demonstrated to lyse melanoma cells in vitro⁴⁹. Given the elevated numbers of NETs observed in tumors of treated mice, we asked if NETs were also responsible for eliminating tumor escape variants. PAD4−/− and NE−/− mice are deficient in enzymes crucial for NET formation. The triple combination therapy was equally effective in eliminating heterogeneous tumors in PAD4−/− and NE−/− as in wild type mice (Supplemental Figure 6A and B). These results suggest that although NETs represent a surrogate for neutrophil activation, key enzymes important for NETosis are not directly involved in killing escape variants in vivo.

Neutrophils are important for immunotherapies

Using monoclonal T cells (Trp1 T cells) against a target antigen (Trp1) allowed us to study preclinically the contribution of neutrophils in eliminating antigen escape variants. We next studied neutrophil activation in a clinical setting. We analyzed NET formation in pre- and post-treatment tissue biopsies from 14 cancer patients who received an anti-OX40 antibody in a phase I neoadjuvant trial³¹. Post-treatment biopsies from 6 patients showed a greater than 2-fold increase, 6 had a less than 2-fold increase, and only 1 patient’s specimen showed decreased NETs (Supplemental Figure 7) when compared to pre-treatment biopsies. Thus, anti-OX40 can activate neutrophils in a small cohort of cancer patients.

To further study the contribution of neutrophils in a system that more closely models the clinical precedent, we used a MC38 colon carcinoma murine model which is responsive to anti-OX40 monotherapy. Cohorts of mice received anti-Ly6G depleting antibody. The efficacy of anti-OX40 monotherapy was decreased in the absence of neutrophils, demonstrating their importance in anti-OX40 monotherapy in absence of monoclonal adoptive T cell transfer (Figure 7A).

Figure 7: Anti-tumor neutrophils are important for response to immunotherapies.

A) Cohorts of mice were implanted with MC38. At day 7, the mice were treated as described. Cohorts of mice received anti-Ly6G. Left, Schema of the experiment. Middle, Tumor curves of individual mice over time. Inset numbers indicate the number of mice with regressed tumors. Right, Survival curves pooled from 2 independent experiments. B) Biopsies from melanoma patients that were treated or not with ICB were stained for NETosis. Left, Schema of the experiment. Middle, Representative images. Yellow arrows indicate NETs and white arrows indicates MPO+ neutrophils. Bars = 50μm. Right, Semiquantitative supervised analysis of NETs per biopsy. Each dot represents one biopsy. NET score (Y axis): - = no NETs, −/+ = 5–10 NETs in the entire biopsy, + = few, ++ = intermediate, +++ = abundant number of NETs. C) Cohorts of mice were implanted with MC38 tumors and treated as described. Cohorts of mice received antiLy6G. Left, Schema of the experiment. Middle, Tumor curves of individual mice over time. Inset numbers indicate the number of mice that completely regressed their tumors. Right, Survival curves pooled from 2 independent experiments. Mouse experiments were repeated twice. D) Bulk RNAseq data from melanoma tumor biopsies generated from patients treated with anti-PD-1 alone (CTLA-4 naive) or anti-PD-1 in combination with anti-CTLA-4 (CTLA-4 experienced)⁷³. Using the 10 top expressed genes for each signature in Cluster 0 and Cluster 1 normalized z-scores of gene expressions were generated and unpaired two-sided T-tests was performed. Box plots and violin plots are shown for patients with disease progression or not. * P<0.05.

Numerous patients have already been effectively treated with ICB and most of the therapeutic effect is ascribed to the effector function of T cells. The relative contribution of innate immunity, particularly neutrophils, to treatment efficacy is unknown. Therefore, we procured biopsies from melanoma patients that were untreated or received anti-CTLA-4 and/or anti-PD-1 therapy. Tumors from ICB treated patients showed increased neutrophil activation as detected by NETosis staining (Figure 7B). The presence of activated neutrophils in biopsies from patients treated with ICB is consistent with our preclinical results. Therefore, we tested the role of neutrophils in ICB in the MC38 mouse model which is responsive to combined blockade with anti-CTLA-4 and anti-PD1. Depleting neutrophils in mice bearing MC38 tumors significantly reduced the efficacy of anti-CTLA-4 + anti-PD1 therapy (Figure 7C).

To further demonstrate the role of anti-tumorigenic neutrophils in conventional T cell-targeted therapies, we applied the neutrophil gene signature from our single cell sequencing studies to a cohort of melanoma patients. Transcriptomic and immune profiling was performed on tumor biopsies of patients who had been treated with anti-PD-1 monotherapy or combined anti-PD-1 and anti-CTLA-4⁷³. In agreement with our preclinical observations in mice, upregulation of our anti-OX40-derived gene signature (enriched in anti-tumorigenic neutrophils/Cluster 1) was associated with improved outcome in anti-CTLA-4-pretreated patients (Figure 7D). Additionally, the gene signature derived from IgG treatment (not enriched in anti-tumorigenic neutrophils/Cluster 0) showed no difference in expression between groups. In sum, these results attest to the clinical relevance of neutrophils in tumor elimination while using immune based therapies that mobilize T cells.

Discussion

Silencing expression of antigenic proteins allows tumors to evade elimination by antigen-specific immune cells. Loss of target antigen is a particularly problematic resistance strategy in response to adoptive T cell therapy and additional targeted immunotherapies. Indeed, adoptive transfer of melanoma-specific T cells promotes dedifferentiation with emergence of melanoma antigen loss variant clones^74,75. Similarly, the outgrowth of CD19^null tumor cells in part accounts for tumor progression in patients treated with chimeric antigen receptor T cells against CD19^3,76,77. Consistent with these observations, mouse models show that immune elimination requires that a considerable number of tumor cells express neoantigens^78,79. In humans, immune checkpoint blockade shows greatest efficacy against neoantigen high tumors, perhaps due a lower likelihood that complete antigen-null variants emerge⁸⁰. It follows then that strategies which engage antigen-specific and non-specific arms of the immune system would achieve superior elimination of antigenically heterogenous tumors.

The role of neutrophils in tumor immunity is context and subset dependent. Pro-tumorigenic functions of neutrophils have been described in the setting of chronic inflammation (so-called tumor-associated neutrophils type 2) and metastasis^{42,43,81–83}. Tumor-associated neutrophils can promote oncogenesis, tumor growth, angiogenesis, extravasation and metastasis, and direct suppression of the adaptive immune system^38,84. An elevated neutrophil to lymphocyte ratio is a robust predictor of negative prognosis in multiple cancer types and is associated with reduced effectiveness of immunotherapies^40,41,85.

Conversely, neutrophils show anti-tumorigenic properties in the context of acute inflammation or with therapeutic intervention. Indeed, neutrophils may initiate an immune response by increasing antigen presentation and T cell co-stimulation^86,87. Relieving hypoxia has been shown to augment neutrophil infiltration into murine uterine adenocarcinoma with subsequent tumor cell-killing via NADPH oxidase, reactive oxygen species, and MMP-9⁴⁸. Other reports show that B16 melanoma elimination via combinatorial immunotherapy was partially neutrophil dependent⁸⁸,that neutrophils kill Lewis Lung carcinoma cells via nitric oxide release⁷¹, and that neutrophils exposed to β-glucan suppress tumor growth in a ROS dependent mechanism⁸⁹. Most recently, it was shown that neutrophil elastase kills tumor cells by cleaving the death domain of CD95⁹⁰. Therefore, the role of neutrophils in cancer is complex and most likely context dependent.

In combination with anti-melanoma T cells, anti-OX40 therapy increased the number of tumor-infiltrating neutrophils, and these neutrophils displayed a distinct phenotype. Upon UMAP projection and clustering of scRNA-seq data of tumors from treated mice, we found the neutrophils contained in a unique cluster (“Cluster 1”) that was enriched in activation markers including IL1b, Osm, Plaur, Soc3, and Ccl4. Gene ontology analysis of Cluster 1 showed increased chemotaxis, neutrophil killing, apoptosis, and NETosis, suggesting that highly differentiated neutrophils adopt an anti-tumor phenotype. However, while the identified gene signature enabled us to cluster cell subsets, some of the markers which help classify them cannot be attributed to an effector function. As an example, several genes upregulated in Cluster 1 (both by transcriptomics and by flow cytometry), are shared by pro-tumorigenic “activated PMN-MDSC”^67,91. In contrast, genes including Lcn2, Lyz2, Ngp, Camp, S100a8, S100a9, expressed at the beginning of the trajectory, are shared with PMN-MDSCs and cells that exhibit a more immature phenotype. Confusingly, the anti-tumorigenic signature derived from Cluster 1 shares many similarities with a prior report describing pro-tumorigenic “trained” neutrophils in the setting of chronic inflammation⁸⁹. We wonder then, if activated neutrophils suppress the generation of anti-tumor immunity in the setting of chronic inflammation, while in the context of acute inflammation (instigated by T cell immunotherapy, for example), neutrophils kill tumor cells indiscriminately. Consistently, neutrophils that are chronically as opposed to acutely activated, are metabolically deficient and lack adequate killing capabilities⁹².

Importantly, we found that neutrophil iNOS was partly required for heterogeneous tumor elimination. Although nitric oxide produced by neutrophils has been shown to kill effector tumor-specific T cells^93,94, at the timepoints where neutrophils infiltrate the tumors in our system, very few T cells remain. Thus, neutrophil-mediated killing of tumor cells predominates. As a metric for neutrophil activation, we employed quantification of NETs given that NETosis is triggered by many stimuli and processes⁹⁵. Importantly, and consistent with our observations, nitric oxide is a potent promoter of NETosis⁹⁶. The role of NETs in tumorigenesis is controversial. Recent reports implicate NETs in metastasis and “shielding” tumors from activated T cells^42,44,97,98. However, NETs might also contribute to the elimination of tumors under acute inflammatory conditions since in vitro, NETs have been shown to inhibit melanoma cell migration and promote tumor lysis⁴⁹.

In our combination treatment, anti-OX40 showed superior activity to anti-CTLA-4, which is somewhat surprising because as a single agent, anti-OX40 is ineffective in poorly immunogenic established murine tumors. But, in combination with additional interventions, as we have shown, anti-OX40 can be extremely efficacious. Care must be taken at the pre-clinical level when choosing combinations, since hyperstimulation and activation induced death of anti-tumor T cells results from combination anti-OX40 and PD-1 blockade^99,100. To date, there has been only limited clinical data on OX40 and our pre-clinical mouse study highlights the need for further work along these lines.

In summary, the data presented here support a model in which tumor regression and tissue-specific inflammation start with T cell activation and T cell-mediated elimination of some tumor cells, followed by a wave of neutrophils that eliminate antigen-negative tumor cells. This is distinct from prior considerations of the role of neutrophils as either potentiating tumor growth or participating in antigen presentation. These observations warrant more careful analysis of innate immunity in clinical settings when immune checkpoint blockade and other adaptive immune mediators are used.

Limitations of the Study

This study was designed to investigate mechanisms by which neutrophils eliminate tumor antigen escape variants in the context of T cell-targeted therapies. We performed most of the experiments in mice treated with an experimental anti-OX40 antibody rather than FDA approved immunotherapies (e.g., anti-CTLA-4 and anti-PD-1). We used this model system because, in comparison with treatment currently in standard clinical use, anti-OX40 demonstrated superior anti-tumor potency and increased neutrophil activation in our mouse models, which made it a more adequate and comprehensive system to study neutrophils. An additional limitation of our report is the relatively low number of patients in which neutrophil activation after T cell immunotherapy was studied. These assessments need to be validated in larger prospective cohorts. Despite these limitations, our work strongly suggests that future trials should be designed to address in detail the contribution of neutrophils in the context of therapies that improve T cell immunity to tumors and/or adoptive cell therapy.

STAR methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Taha Merghoub (tmerghoub@med.cornell.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All data supporting the finding of this study are available within the paper (and its Supplemental information files). The bulk RNA-seq and scRNA-Seq raw data have been deposited in the GEO database: GSE224675. All sequencing data are publicly available as of publication. Accession number is listed in the key resources table. This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Anti-OX40 (clone OX86)	Bioxcell	AB_1107592
Control IgG (anti-HRPN)	Bioxcell	AB_2687731
Anti-CTLA-4 (clone 9D9)	Bioxcell	AB_10949609
Anti-PD1 (clone RMP1–14)	Bioxcell	AB_10949053
Anti-CD4 (clone GK1.5)	Bioxcell	AB_10950382
Anti-MHC class II (clone M5/114)	Bioxcell	AB_10949298
Anti-NK1.1 (clone PK136)	Bioxcell	AB_1107737
Anti-Ly6G (clone 1A8)	Bioxcell	AB_1107721
Anti-CSF1R (2–4A5–4)	Bioxcell	AB_2894766
Anti-CD45 Pacific Blue (clone 30-F11)	eBioscience	AB_10373710
Anti-CD11b PE-Texas Red (clone M1/70.15)	eBioscience	AB_10373548
Anti-Ly6G PercpCy5.5 (clone RB6–8C5)	eBioscience	AB_906247
Anti-Ly6C PE-Cy7 (clone HK1.4)	eBioscience	AB_2573503
Anti-Class I FITC (anti-H2Kd, clone SF1–1.1)	BD Biosciences	AB_394923
Anti-Class FITC (anti-I-A/I-E, clone 2G9)	BD Biosciences	AB_394958
Anti-CD62L BUV395 (clone MEL-14)	BD Biosciences	AB_2739966
Anti-CXCR2 BUV563 (clone V48–2310)	BD Biosciences	AB_2873085
Anti-CD40 BUV615 (clone 3/23)	BD Biosciences	AB_2875639
Anti-CXCR4 BUV805 (clone 2B11/CXCR4)	BD Biosciences	AB_2871282
Anti-IL-1Rβ BV421 (clone 4E2)	BD Biosciences	AB_2737896
Anti-CD49d SB436 (clone R1–2)	Invitrogen	AB_2744780
Anti-PD-L1 BV480 (clone 10F.9G2)	BD Biosciences	AB_2743606
Anti-Ly6C BV570 (clone HK1.4)	Biolegend	AB_10896061
Anti-F4/80 Pac Orange (clone BM8)	Invitrogen	AB_2539704
Anti-CD16 SB600 (clone 93)	Invitrogen	AB_2688072
Anti-CD11b BV650 (M1/70)	Biolegend	AB_2566568
Anti-CD18 BV711 (M18/2)	BD Biosciences	AB_2742350
Anti-CD24 BV750 (M1/69)	BD Biosciences	AB_2872059
Anti-CD45.1 BV785 (clone A20)	Biolegend	AB_2563379
Anti-CD206 FITC (clone MR5D3)	Invitrogen	AB_2538348
Anti-CD117 Alexa700 (clone ACK2)	eBioscience	AB_657583
Anti-MPO PE (clone MPO455–8E6)	Invitrogen	AB_10733265
Anti-Ly6G PerCP-Cy5.5 (clone 1A8)	BD Biosciences	AB_1727563
Anti-INOS APC (clone CXNFT)	eBioscience	AB_2573244
Anti-CD177 AF647 (clone Y127)	BD Biosciences	AB_2869790
Anti-SiglecF AF700 (clone 1RNM44N)	eBioscience	AB_2637127
Anti-CD45.2 APC-Cy7 (clone 104)	Biolegend	AB_830789
Anti-CD45 APC (Clone 30-F11)	BD Biosciences	AB_398672
Anti-CD11b PE (Clone M1/70)	BD Biosciences	AB_396680
Anti-Ly6G (Clone RB6-BC5)	AbCam	AB_881385)
Anti-CD11b (rabbit polyclonal)	AbCam	AB_1310048
Anti-TYRP1 (clone EPR13063)	AbCam	Cat# ab178676
Anti-S100a (Clone EPR5250)	AbCam	Cat# ab133519
Goat anti-mouse myeloperoxidase	R & D Systems	AB_2250866
Rabbit anti-citrullinated histone H3	R & D Systems	AB_2250866
Anti-Trp1 Alexa 488 (Clone TA99)	Monoclonal Antibody Facility at MSKCC	AB_304752
Anti-CD11b PE-Cy7 (Clone M1/70)	BD Biosciences	AB_394491
Anti-Ly6G PE (Clone 1A8)	BD Biosciences	AB_394208
Anti-CD45.2 PE (Clone 104)	BD Biosciences	AB_1727493
Bacterial and virus strains
Biological samples
Chemicals, peptides, and recombinant proteins
D-luciferin	Sigma-Aldrich	Cat# 122799
Leica Bond ER1 Buffer	Perkin Elmer	Cat# AR9961
Cyclophosphamide monohydrate	Sigma-Aldrich	Cat# 93813
Permount	Leica Biosystems	Cat# SP15
1x Antibody Diluent / Block	Fisher Scientific	Cat# ARD1001
AR9 buffer	Akoya Biosciences	Cat# AR900250ML
Hoechst 33342	Akoya Biosciences	Cat# H3570
ProLong Gold mounting medium	Invitrogen	Cat# P10144
Fc Receptor blocker	Invitrogen	Cat# NB309
Donkey serum	Innovex	Cat# D9663
DAPI	Sigma	Cat#D1306
Fixable Viability Dye eFluor 506	ThermoFisher	Cat# 65–0866-14
ACK lysing Buffer	ThermoFisher	Cat# 10–548E
Cellstripper	Lonza	Cat# 25–056-CI
Cytofix/Cytoperm™ Kit	Corning	Cat# 554714
Percoll	BD Bioscience	Cat# 17089101
cOmplete™ Protease Inhibitor Cocktail Tablets	Cytiva (GE Healthcare)	Cat# 11697498001
Mouse Cytokine/Chemokine 32 Plex kit	Roche	Cat# MCYTMAG70K-PX32
Trp1 peptide50	Millipore	N/A
Murine IL-2	Genemed Synthesis	Cat# 402-ML-100
Murine IFN-γ	Peprotech	Cat# 315–05
poly-D-lysine	PeproTech	Cat# P7280
Sytox™Green Nucleic Acid Stain	Millipore Sigma	Cat# S7020
TRIzol Reagent	Molecular Cytology core at MSKCC	Cat# 15596018
RNeasy Micro Kit	Invitrogen	Cat# 74004
KAPA Hyper Prep Kit	Qiagen	Cat# KK8504
CD4 beads (L3T4; Miltenyi Biotech)	Kapa Biosystems	Cat# 130–117-043
Zombie NIR	eBioscience	Cat# 423105
TotalSeq™-C0301 anti-mouse Hashtag 1 Antibody	Biolegend	Cat# 155861
TotalSeq™-C0301 anti-mouse Hashtag 2 Antibody	Biolegend	Cat# 155863
TotalSeq™-C0301 anti-mouse Hashtag 3 Antibody	Biolegend	Cat# 155865
TotalSeq™-C0301 anti-mouse Hashtag 4 Antibody	Biolegend	Cat# 155867
TotalSeq™-C0301 anti-mouse Hashtag 5 Antibody	Biolegend	Cat# 155809
CITE-Seq antibodies Mouse panel	Biolegend	part 900002538
Opal dye 520	Akoya	Cat# FP1487001KT
Opal dye 690	Akoya	Cat# FP1497001KT
Leica Bond ER2	Leica	Cat# AR9640
Critical commercial assays
Deposited data
Human scRNA-seq	Tirosh at al.52	https://www.ncbi.nlm.nih.gov/geo/; ID: GSE72056
Human RNAseq data	Gide et at.73	https://www.ebi.ac.uk/ena/browser/; ID: PRJEB23709
Bulk RNA	This paper	https://www.ncbi.nlm.nih.gov/geo/; ID: GSE224675
Single cell	This paper	https://www.ncbi.nlm.nih.gov/geo/; ID: GSE224675
Experimental models: Cell lines
B16F10	I. Fidler (MD Anderson)	N/A
B78H1	A. Albino (MSKCC)	N/A
MC38	Kerafast, Inc.	Cat# ENH204-FP
Experimental models: Organisms/strains
Mouse: C57BL/6J	Jackson Laboratory	Stock # 000664
Mouse: C57BL/6J MHC II−/−	Jackson Laboratory	Stock # 003584
Mouse: C57BL/6J Rag2−/−	Jackson Laboratory	Stock # 008449
Mouse: C57BL/6J Nos2−/−	Jackson Laboratory	Stock #002609
Mouse: C57BL/6J Pad4−/−	Jackson Laboratory	Stock # 030315
Mouse: C57BL/6J NE−/−	Jackson Laboratory	Stock # 006112
Mouse: Trp1 CD4+ TCR transgenic	N. Restifo (NIH)	N/A
Mouse: Trp1-luciferase transgenic	J.P. Allison (MSK)	N/A
Oligonucleotides
Recombinant DNA
Software and algorithms
Living Image®	Perkin Elmer	https://www.perkinelmer.com/category/in-vivo-imagingsoftware
Prism 7	GraphPad Software, LLC (online download)	https://www.graphpad.com/scientificsoftware/prism/
FlowJo v.10	FlowJo, LLC (online download)	https://www.flowjo.com/solutions/flowjo/downloads
HALO	Indica Labs	https://indicalab.com/halo/
CaseViewer 2.1	3D Histech	https://www.3dhistech.com/solutions/caseviewer/
LAS X software	Leica	https://www.leicamicrosystems.com
Partek Flow	Partek Incorporated	N/A
Inform 2.4 image analysis software	Akoya	N/A
ImageJ software	NIH	https://imagej.nih.gov/ij/download.html
Other
Cell culture inserts (0.4 μm pore size)	Fisher Scientific	Cat# 353095
VWR Oven Gr Con 2.3CF	VWR	Cat#89511–404
Countess II Automated Cell Counter	ThermoFisher	N/A
Chromium Chip A	10X Genomics	PN 230027
IVIS 200 in vivo imaging system	Perkin Elmer	N/A
Leica Bond RX	Leica Biosystems	N/A
Pannoramic Flash 250	SD Histech	N/A
Vectra Multispectral Imaging System version 3	Akoya	N/A

EXPERIMENTAL MODEL AND SUBJECT DETAILS Mouse Models

C57BL/6J, C57BL/6J MHC II^−/−, C57BL/6J Rag2^−/−, C57BL/6J Nos2^−/−, C57BL/6J Pad4^−/−, and C57BL/6J NE^−/− 8–10-wk-old males) were obtained from The Jackson Laboratory, and Trp1-specific CD4+ T TCR transgenic mice were obtained from the N. Restifo Laboratory (National Institutes of Health, Bethesda, MD). Trp1 CD4⁺ TCR transgenic mice were crossed to Rag1^−/− Trp1^−/− CD45.1^+/+ background. These mice were further crossed in-house to a C57BL/6J luciferase^+/+ background for imaging experiments (Trp1-luciferase cells; A gift from J.P. Alison Laboratory). All mouse procedures were performed in accordance with institutional protocol guidelines at Memorial Sloan-Kettering Cancer Center (MSKCC) under an approved protocol. All mice used were 6–8 week of age. Mouse experiments that included Trp1 T cell transfer used males since mice the TCR transgene was inserted on the Y chromosome⁵⁰. Female mice were used for the MC38 model experiments.

Mouse tumor cell lines

The B16-F10 mouse melanoma cell line was originally obtained from I. Fidler (M.D. Anderson Cancer Center, Houston, TX). B78H1 was obtained from A. Albino (Sloan Kettering Institute). Inoculation with 1×10⁵ B16-F10 or 1×10⁶ B78H1 cells was determined to be a lethal dose when injected intradermally in the flank (50 μl/injection). In experiments where B16-F10 and B78H1 were co-injected, the lethal dose for each tumor was mixed in the same volume. MC38 was obtained from Kerafast. 5×10⁵ cells were injected subcutaneously in the flank which was deemed a lethal dose.

Patient samples

All tissues were collected at MSKCC following study protocol approval by the MSKCC Institutional Review Board. Informed consent was obtained for all patients. The study was in strict compliance with all institutional ethical regulations. All tumor samples were surgically resected primary or metastatic melanoma. All treated samples are from patients that received immunotherapy either ipilimumab alone, nivolumab alone or both. Samples were collected between 30 to 90 days post treatment. Controls did not receive prior immunotherapy. All tumors were subjected to pathological review and histological confirmation by a dermatopathologist before analysis. Biopsies reflect the population demographics of the greater New York area.

METHOD DETAILS

Monoclonal antibodies, drug treatment, and adoptive transfer experiments

Anti-OX40 (OX86), anti-CTLA-4 (9D9), and control IgG (anti-HRPN), were purchased from Bio-X-Cell. For all treatments, a single dose of 0.5 mg anti-OX40. Starting day 1, 0.1 mg anti-CTLA-4 were i.p. administered every 3 days for 3 weeks. Cyclophosphamide monohydrate (Sigma-Aldrich) mixed in sterile PBS was i.p. Injected as a single dose at 250 mg/kg. Trp1 CD4+ T cells (Trp1 T cells) were purified from lymph nodes and spleen by positive selection magnetic cell sorting using CD4 beads (Miltenyi Biotech) according to the manufacturer’s instructions and 1 ×10⁵ purified Trp1 T cells were injected i.v. in the tail. For CD4+ T cell depletion, 0.25 mg anti-CD4 (Bio-X-Cell) was injected 3X weekly starting at day 3 or day 12 after treatment throughout the duration of the experiment. For MHC class II neutralization, 0.2 mg anti-MHC class II (Bio-X-cell) was i.p. injected 3X weekly starting at day 3. For neutrophil depletion, 0.5 mg of anti-Ly6G (Bio-X-cell) was i.p. injected at day 3 after treatment 3X weekly throughout the duration of the experiment. For NK cell depletion, 200 μg of anti-NK1.1 (Bio-X-Cell) was injected i.p. 2X weekly for 3 weeks. For monocyte/macrophage depletion, 300 μg of anti-CSF1R (Bio-X-Cell) was injected i.p. 2X weekly for 3 weeks. For MC38 experiments, 0.1 mg of anti-CTLA-4 (9D9) and 0.25 mg of anti-PD1 (Bio-X-Cell) i.p. every 3 days for 3 weeks. 0.5 mg of anti-OX40 was given 2X a week apart. At day 10, cohorts of mice received anti-Ly6G depleting antibody 3X weekly for 3 weeks.

In vivo Imaging

Mice were treated as described above with Trp1-specific, Luciferase expressing CD4 +T cells. For imaging, mice were anesthetized with isoflurane and i.p. injected with 100 μl of a 30 mg/ml D-luciferin solution (Perkin Elmer). After 10 minutes, the mice were imaged using an IVIS 200 in vivo imaging system (Perkin Elmer). Mice were imaged periodically (every 3–4 days) after treatment. Photon flux was measured with Living Image^® software (Perkin Elmer).

Immunohistochemistry

Tumor sections (5 μm) were prepared from formalin-fixed, paraffin-embedded material and collected on Superfrost^®/plus microscope slides (Fisher Scientific). After deparaffinization and rehydration, the slides were boiled in 50 mM citrate buffer (pH 6) for 30 minutes to retrieve the antigens. After the slides were allowed to cool to room temperature, immunohistochemistry (IHC) was performed on Ly6G (AbCam), CD11b (AbCAM). For CD45.1, sections were stained for immunohistochemistry on Leica Bond RX (Leica Biosystems) with 1ug/ml CD45.1 Mouse monoclonal (ThermoFisher Scientific) for 1 hour on Protocol J. The sections were pre-treated with Leica Bond ER1 Buffer (Leica Biosystems) for 10 min at 100C before antibody application. After staining the sections were air-dried and mounted with Permount for digital scanning. H&E staining was performed using standard procedures. Images were captured with Pannoramic Flash 250 (SD Histech) using 40X0.95 NA objective. For CD11b analysis representative regions were chosen and the area of positive staining was measured using HALO software. For Ly6G analysis representative regions of the tumor were selected and positive cells were manually counted CaseViewer 2.1 software (3D Histech).

Immunofluorescence

For Trp1/S100 immunofluorescent determination, three-color multiplex staining was performed after primary antibodies optimal concentration and antibody stripping condition were selected by MSK pathologist (TH) according to the results of diaminobenzidine IHC staining on the Leica Bond Rx Stainer with Leica Bond Polymer Refine Detection kit (DS9800). 4 μm thick FFPE tissue sections were baked for 3 hrs. at 62°C in an oven (VWR Oven Gr Con 2.3CF, 89511–404) upright with subsequent deparaffinization performed on Leica Bond RX, followed by 30 minutes of antigen retrieval with Leica Bond ER2 (Leica), followed by two sequential cycles of staining, each round of which included a 10-minute blocking (Akoya antibody diluent/block) and 30-minute primary antibody incubation. The primary antibody detection was performed using HRP conjugated species-specific secondary antibody polymer with 10-minute incubation. Each HRP-conjugated secondary induces the covalent binding of fluorophores to tissue using tyramide signal amplification with Opal dyes 520 (Akoya) and 690 (Akoya) accordingly. A heat induced stripping of the primary/secondary antibody complex using Akoya AR9 buffer and Leica Bond ER2 (90% AR9 and 10% ER2) at 100°C for 20 minutes after each staining cycle preceding the next staining cycle. After two sequential rounds of staining, sections were stained with Hoechst (Invitrogen 33342) to visualize nuclei and mounted with ProLong Gold antifade reagent mounting medium (Invitrogen). Two color multiplex-stained slides were imaged using the Vectra Multispectral Imaging System version 3 (Akoya). Scanning was performed at 20X (200X final magnification). Filter cubes used for multispectral imaging were DAPI, FITC, Cy3, Texas Red and Cy5. A spectral library containing the emitted spectral peaks of the fluorophores in this study was created using the Vectra image analysis software (Akoya). Using multispectral images from single-stained slides for each marker, the library was used to separate each multispectral cube into individual components (spectral unmixing) allowing for identification of two marker channels of interest using Inform 2.4 image analysis software. Images were exported to Indica Labs Halo image analysis platform and cell segmentation and signal thresholding was performed separately on each case using a supervised algorithm.

NET staining was conducted as follows: After deparaffinization and rehydration, sections underwent antigen retrieval in EDTA buffer (10 mM Tris Base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0). After blocking with Fc Receptor blocker (Innovex) followed by 1 x blocking buffer (5% donkey serum (Sigma), 2.5% BSA, 0.1% Triton X-100 in PBS), sections were incubated with primary antibodies: goat-anti-mouse myeloperoxidase (1:100, R&D) and rabbit-anti-citrullinated histone H3 (1:250, Abcam) in 0.5 x blocking buffer overnight at 4 °C. Fluorochrome-conjugated secondary antibodies (1:150, ThermoFisher) and DAPI (10 ug/ml, ThermoFisher) were incubated on sections the next day in 0.5 x blocking buffer for 45 min in the dark at room temperature. Images were captured using Leica confocal microscopy (SP8) and processed with LAS X software (Leica). For quantification of NETs, processed single channel images were analyzed using ImageJ software with the same threshold for all images. The level of NETs in each tumor was approximated by calculating integrated density of the fluorescence intensity in the citrullinated histone H3 channel as a fraction of the integrated density of the myeloperoxidase channel. Two-five representatively chosen high power fields from five tumors per condition were analyzed. In other experiments, NETs were either semi-quantified or quantified using HALO software.

Flow cytometry analysis

For neutrophil infiltration experiments, the following antibodies were used for flow cytometry analysis CD45 Pacific Blue, CD11b PE-Texas Red, Ly6G PercpCy5.5, Ly6C PE-Cy7, were obtained from eBioscience. Viability dye (LIVE/DEAD Fixable Aqua Dead Cell Stain kit) was obtained from Invitrogen. Anti-CD16/CD32 (“Fc block”) was produced and purified by the Monoclonal Antibody Core Facility at MSKCC. All antibodies were used according to the manufacturer’s instructions with the recommended buffers. Single cell suspensions from tumors were prepared passing the minced tissue through a 40μM filter, red blood cells were removed by incubation in ACK Lysing Buffer (Lonza) at room temperature. The single cell suspensions were resuspended in FACS buffer at approximately 10 million cells/ml and 100μl of sample was further stained. Before staining, cells were treated with Fc block in FACS buffer (2% bovine serum albumin and 2 mM EDTA in PBS) on ice for 15 min. Staining of surface antigens was performed in FACS buffer on ice for 40 min. Flow cytometry was performed on an LSRII (BD Bioscience). FlowJo software (version 10.6.1; FlowJo LLC) was used for all flow cytometry analysis.

For determination of MHC class I, class II, and Tyrp1 on B16 and B78H1, the cell lines were grown in vitro to confluency. The cells were detached for the tissue culture plate using Cellstirpper (Lonza) and resuspended in FACS buffer. A single cell suspension was stained with anti-Class I FITC (anti-H2K^d, BD Bioscience), anti-Class FITC (anti-I-A/I-E, BD Bioscience). For Trp1 determination, the cells were permeabilized with Fixation and Permeabilization Solution (BD Bioscience). After 20 min, the cells were washed with Perm/Wash Buffer (BD Bioscience) and stained in the same buffer with Anti-Trp1 Alexa488 (TA99 prepared and conjugated by Monoclonal Antibody Facility at MSKCC) and further analyzed by flow cytometry.

For neutrophil phenotyping, the following antibodies were used: CD62L BUV395 (BD Bioscience), CXCR2 BUV563 (BD Bioscience), CD40 BUV615 (BD Bioscience), CXCR4 BUV805 (BD Bioscience), IL-1Rβ BV421 (BD Bioscience), CD49d SB436 (Invitrogen), PD-L1 BV480 (BD Bioscience), Ly6C BV570 (BioLegend) , F4/80 Pac Orange (Invitrogen), CD16 SB600 (Invitrogen), CD11b BV650 (BioLegend), CD18 BV711 (BD Bioscience), CD24 BV750 (BD Bioscience), CD45.1 BV785 (BioLegend), CD206 FITC (Invitrogen), CD117 Alexa700 (eBioscience), MPO PE (Invitrogen), Ly6G PerCP-Cy5.5 (BD Bioscience), INOS APC (eBioscience), CD177 AF646 (BD Bioscience), SiglecF AF700 (eBiscience), CD45.2 APC-Cy7 (Biolegend). Cohorts of mice were treated as before. Single cell suspensions were prepared from tumors 10 days after treatments. Cells were resuspended in 3 ml of 40% Percoll (Fisher Scientific) and centrifuged for 10 min at 800 rcf. Red blood cells were removed by incubation in ACK Lysing Buffer (Lonza) at room temperature. The pellet was resuspended in PBS and 1:500 dilution of Zombie NIR^™ Fixable Viability dye was added for 15 min. Before staining, cells were treated with Fc block in FACS buffer on ice for 15 min. After wash, surface antibodies were added and incubated for 40 min Super Bright Complete Staining Buffer (Invitrogen). The cells were washed, fixed and permeabilized overnight with Cytofix/Cytoperm^™ Kit (BD Biosciences). The next day, the cells were washed in 1X Perm/Wash buffer^™ (BD Biosciences). Intracellular antibodies were incubated for 1 hour in Perm/Wash buffer^™ for 1 hour, washed and resuspended in FACS buffer. The cells were analyzed by an Aurora flow cytometer (Cytek^® Biosciences). FlowJo software (version 10.6.1; Tree Star) was used for all flow cytometry analysis. Heatmaps of neutrophil markers were generated using the log of the fold-change (log FC) of the mean fluorescence intensity (MFI) for each replicate divided by the average MFI of all the replicates in total (irrespective of treatment) for each marker.

Chemokine/Cytokine protein array

Mice were treated as described previously. On day 10 after treatment, cohorts of 5 mice/group were sacrificed and tumors were resected and snap-frozen in liquid nitrogen. The weight of frozen tumors was measured, and then extensively minced. Extraction buffer (PBS and 1X cOmplete^™ EDTA-free Protease Inhibitor Cocktail Tablets (Roche)) was added at 1 ml/100 mg of tissue. The samples were homogenized in a tissue homogenizer, centrifuged to remove debris, and supernatants were snap frozen in liquid nitrogen. The samples were thawed, and protein concentration was normalized to 1 mg/ml. Cytokine and chemokine were measured using MILLIPLEX MP Mouse Cytokine/Chemokine Magnetic Bead Panel-Premixed 32 Plex (Millipore) according to the manufacturer instructions. The samples were analyzed using a MAGPIX machine (Luminex).

In vitro killing assays

Target melanoma cells (B78H1), with or without neutrophils, were co-incubated with tumor homogenate from treated mice (10 μg/ml) in HTM media (RPMI 1640 supplemented with 10% FCS, 1× nonessential amino acids, 1 mM sodium pyruvate, 2mM l-glutamine, and 50 μM β-mercaptoethanol). After 48 hours, the cells were washed twice with cold PBS followed by trypsinization. The recovered melanoma cells were plated in 6-well plates for colony formation. After 7 days in culture, the plates were fixed with 3.7% formaldehyde, stained with 2% methylene blue, washed, and dried. Colonies were counted manually. The percentage of target cells killed was calculated using the following equation: 1- [melanoma + T cells]/[melanoma alone]. For flow cytometry, Target melanoma cells (B78H1), with or without neutrophils, were co-incubated with tumor homogenate from treated mice (10 μg/ml) in HTM media. After 1 hour, CellEvent^™ Caspase-3/7 Green Detection Reagent (Invitrogen) was added at 1:1000 dilution. After 16 hours, the plates were washed once with PBS and B78H1 cells were dissociated with CellStripper solution (Lonza). HTM media was added at 1:1 and the cells were collected, washed once with FACS buffer. DAPI (4′,6-diamidino-2-phenylindole) (ThermoFisher Scientific) was added at 1μg/ml. The samples were collected in an LSR II flow cytometer (BD Bioscience). For in vitro killing assays, purified Trp1 T cells were co-cultured with irradiated T cell depleted splenocytes as antigen presenting cells (APCs) pulsed with of 2.5 μg/ml Trp1 peptide⁵⁰ (synthetized by Genemed Synthesis, Inc.) with 20 U/ml of recombinant murine IL-2 (R&D Systems) in HTM. The next day, 10 μg/ml of Anti-OX40 or control IgG was added to the cultures. After 6 days, the cells were washed and counted. In the bottom of a 24 well plate (Corning), B16 (2 X10⁵) and activated Trp1 T cells (1 X10⁵) were added at 2:1 ratio. The B16 cells were previously treated overnight with 10 ng/ml recombinant murine IFN-γ (PeproTech) to upregulate MHC Class II. A cell culture inserts (Transparent PET membrane 0.4 μm pore size; Falcon) was placed in each well. In the top insert, 1 X10⁴ B78H1 cells and 2 X10⁵ naïve purified neutrophils were added. As control, equivalent numbers of neutrophils were added to the lower chamber. After 1 hour, CellEvent^™ Caspase-3/7 Green Detection Reagent (Invitrogen) was added at 1:1000 dilution. After 16 hours, the top inserts were washed once with PBS and B78H1 cells were dissociated with CellStripper solution (Lonza). HTM media was added at 1:1 and the cells were collected, washed once with FACS buffer. DAPI (4′,6-diamidino-2-phenylindole) (ThermoFisher Scientific) was added at 1μg/ml. The samples were collected in an LSR II flow cytometer (BD Bioscience). Supernatants from the original Trp1 culture (Conditioned medium) were added to a plate containing 1 X10⁴ B78H1 cells and 2 X10⁵ naïve purified neutrophils. Killing of B78H1 was performed as described.

Ex vivo killing assay

WT or Nos2^−/− mice was treated as described. After 10 days, single cell suspensions were prepared from tumors. Cells were resuspended in 3 ml of 40% Percoll (Fisher Scientific) and centrifuged for 10 min at 800 rcf. Red blood cells were removed by incubation in ACK Lysing Buffer (Lonza) at room temperature. Cells were incubated with Fc block in FACS buffer for 15 min. After wash, the cells were incubated with anti-CD45 APC (BD Biosciences), anti-CD11b PE-Cy7 (BD Biosciences), and anti-Ly6G PE (BD Bioscience). After wash, cells were treated with DAPI and neutrophils were sorted in an BD FACSAria^™ III Cell Sorter (BD Bioscience). Purified neutrophils were incubated with B78H1 in HTM media. After 24 hours, the cells were washed twice with cold PBS followed by trypsinization. The recovered melanoma cells were plated in 6-well plates for colony formation. After 7 days in culture, the plates were fixed with 3.7% formaldehyde, stained with 2% methylene blue, washed, and dried. Colonies were counted manually. The percentage of target cells killed was calculated using the following equation: 1- [melanoma + T cells]/ [melanoma alone].

In vitro NET formation

Neutrophils were purified from bone marrow of naïve mice by MACS following the manufacturer instructions (Miltenyi Biotec). Neutrophils (2 ×10⁵) were plated in triplicates in poly-D-lysine-coated glass plates (Milipore) in 200μl of HNB (PBS, 0.5% FBS (Sigma), + 0.5% D-(+)-glucose (Sigma) . To coat glass plates, 200μl of 1 mg/ml poly-D-lysine solution was added and incubated for 5 minutes. The plates were washed twice with ddH₂O and dried for 2 hours under aseptic conditions. Neutrophils were incubated overnight at 37°C in HNB. 0.5%. PBS, tumor protein extracts (10 μg/ml), or phorbol 12-myristate 13-acetate (Sigma) at 100 ng/ml was added. The next day, Sytox^™Green Nucleic Acid Stain (Termo Fisher Scientific) was added to each well for 15 minutes. The plates were carefully washed with PBS and mounted in Mowiol medium (Molecular Cytology core at MSKCC). For NET quantification, 14–18 random high-power fields (HPF) were captured with an Axio Observer Z1 inverted microscope (Zeiss). Data were analyzed using HALO software.

Single Cell RNA-seq Reanalysis

To understand the heterogeneity of transcription within melanoma tumors single-cell RNA-seq for cells profiled⁵² by Smart-seq2 were downloaded from the Gene Expression Omnibus under GEO accession GSE72056. The processed and filtered log-normalized counts were then loaded into Seurat version 2.3.4¹⁰² and the cells labeled as malignant melanocytes were selected as the matrix for further analysis. Following best practices in the package suggestions principal component analysis was performed on this matrix and after reviewing principal component heatmaps and jackstraw plots t-distributed stochastic neighbor embedding analysis (tSNE) visualization were performed on the top 10 components and clustering resolution was set at 0.6 for visualizations. Plots for specific antigen marker genes were made using Seurat functions FeaturePlot and VlnPlot.

Bulk RNA-seq sample collection, library preparation, sequencing, and analysis

Mice were treated as described with the combination therapy or control. At day 10, single cell suspensions were prepared from tumors of individual mice. CD45+ CD11b+ Ly6G+ neutrophils were sorted in a BD FACSAria™ III Cell Sorter (BD Bioscience). ~1 X10⁵ cells/sample was resuspended in Trizol (ThermoFisher) and frozen. Frozen tissue was homogenized in 1mL TRIzol Reagent (ThermoFisher catalog), phase separation was induced with 200 μL chloroform, and RNA was extracted from the aqueous phase using the RNeasy Micro Kit (Qiagen catalog) on the QIAcube Connect (Qiagen) according to the manufacturer’s protocol with 350 μL input. Samples were eluted in 13–15 μL RNase-free water. After RiboGreen quantification and quality control by Agilent BioAnalyzer, 720–800 pg total RNA with RNA integrity numbers ranging from 2.2 to 6.7 underwent amplification using the SMART-Seq v4 Ultra Low Input RNA Kit (Clonetech), with 12 cycles of amplification. Subsequently, 2 ng of amplified cDNA was used to prepare libraries with the KAPA Hyper Prep Kit (Kapa Biosystems) using 8 cycles of PCR. Samples were barcoded and run on a HiSeq 4000 in a PE50 run, using the HiSeq 3000/4000 SBS Kit (Illumina). An average of 87 million paired reads were generated per sample and the percent of mRNA bases per sample ranged from 84% to 90%. FASTQ files were imported into Partek Flow for analysis. Transcript reads were trimmed and short reads were aligned to STAR - 2.7.8a. Reads mapping to duplicate UMIs were removed. Transcripts were quantified using the Ensembl database (mm10 - Ensembl Transcripts release 95). Subsequent analysis was performed in Partek Flow. We used PCA to determine how individual samples relate to each other based on all gene expression profiles.

Single-cell transcriptome cell preparation

Mice were treated as described with the combination therapy or control. A group of untreated mice were included. At day 10, single cell suspensions were prepared from tumors of individual mice. After Percoll centrifugation, RBC lyse, and Fc block, the cells were stained form individual mice were stained with CD45.2 PE, CD11b APC (BD Bioscience), hashtag oligomers (TotalSeq^™ anti-mouse Hashtag 1–5 Antibodies; Biolgened) and CITE-Seq antibodies (TotalSeq-C Costume Mouse panel Biolegend). After wash, DAPI was added, and 20,000 cells were sorted in a BD FACSAria^™ III Cell Sorter (BD Bioscience). Cells from each group were pooled for further analysis.

Single-cell transcriptome sequencing

Cells were stained with Trypan blue and Countess II Automated Cell Counter (ThermoFisher) was used to assess both cell number and viability. Following QC, the single cell suspension was loaded onto Chromium Chip A (10X Genomics PN 230027) and GEM generation, cDNA synthesis, cDNA amplification, and library preparation of 1,700–6,400 cells proceeded using the Chromium Single Cell 5’ Reagent Kit (10X Genomics PN 1000006) according to the manufacturer’s protocol. cDNA amplification included 13–16 cycles and 11–50 ng of the material was used to prepare sequencing libraries with 14–16 cycles of PCR. Indexed libraries were pooled equimolar and sequenced on a NovaSeq 6000 in a PE28/91 run using the NovaSeq 6000 S1 Reagent Kit (100 cycles) (Illumina). An average of 160 million paired reads was generated per sample.

Cell surface protein feature barcode analysis

Amplification products generated using the methods described above included both cDNA and feature barcodes tagged with cell barcodes and unique molecular identifiers. Smaller feature barcode fragments were separated from longer amplified cDNA using a 0.6X cleanup using aMPure XP beads (Beckman Coulter). Libraries were constructed using the Chromium Single Cell 5’ Feature Barcode Library Kit (10X Genomics PN 1000080) according to the manufacturer’s protocol with 9 cycles of PCR. Indexed libraries were pooled equimolar and sequenced on a NovaSeq 6000 in a PE28/91 run using the NovaSeq 6000 S1 Reagent Kit (100 cycles) (Illumina). An average of 68 million paired reads was generated per sample.

Single cell RNA-Seq Analysis

Primary analysis was done using the CellRanger v6.1.1 software (10X Genomics). Alignment to the mouse reference transcriptome (mm10 ENSEMBL 98) for gene expression) was performed for RNA quantification. Hashtag Oligo identification (HTO) processing was performed also with Cellranger v6.1.1 and the demultiplexing was performed with Seurat v3 (HTODemux, and demultiplexing was done with custom Python scripts. We used the R statistics software version 4.0.5 to perform all further analysis using the Seurat package version 4.0.3¹⁰³. We considered all cells that were detected in the RNA-seq, cell hashing, and ADT libraries. We removed cells with < 300 and > 5500 detected genes, and > 40000 detected UMIs and mitochondrial RNA content of > 10%. We also removed cells with > 5000 ADT reads or < 30 detected ADT features particularly to avoid clumps of antibodies that can occasionally attach to cells.

Normalization of the single cell RNA-seq was performed using the SCTransform package¹⁰⁴ with the number of variable genes set at 10000. Low dimensional projection was performed using UMAP based on the top 40 principal components and for preliminary clustering of cells (resolution = 0.7). Differential gene expression analysis was performed using the FindMarkers function (method: wilcox). Normalization of the ADT data was performed (using functions NormalizeData and ScaleData) and low dimension projection was performed using UMAP at default settings. Subsequent analysis of the single cell RNA-seq subset containing the populations of IgG and anti-OX40 samples was performed with the same parameters used for normalization and clustering with resolution parameter adjusted to 0.25.

Automated cell-type labelling was performed on the single cell RNA-seq counts using R package SingleR¹⁰⁵ with reference purified microarray expression profiles from the Immunological Genome Project dataset¹⁰⁶. Cell trajectory and pseudotime inference for the neutrophil populations of IgG and anti-OX40 samples was performed using the package Slingshot¹⁰⁷ using default parameters on the UMAP reduced dimension and using start.clus=2. For the trajectories finishing in clusters 0 and 1 heatmaps of pseudo-temporally expressed genes in each curve were produced according to package vignette suggestions and the top 50 most variable genes in each lineage displayed.

Gene ontology biological processes

To perform the GO Biological Process Enrichment Analysis, we used the enrichGO function from the “clusterProfiler” R package (version 4.0.5) to compare the genes identified in cluster 1 and cluster 2 against the universe of genes that were present in our single-cell RNA-seq dataset. The genome wide annotation for mouse org.Mm.eg.db (version 3.13.0) was used and gene names were converted to the ENSEMBL format using the “biomaRt” R package (version 2.48.3). Adjusted P values were calculated using the Benjamin & Hochberg (BH) false discovery rate method. Values were plotted using the “ggplot2” R package (version 3.3.5). The following GO pathways were used: Neutrophil chemotaxis (GO:0030593), Neutrophil killing (GO:0070942), Azurophil (GO:0035578), Gelatinase (GO:0070820), Apoptosis (GO:0043065), and NETosis¹⁰¹.

Neutrophil gene signature on melanoma signatures

Bulk RNAseq data from melanoma tumor biopsies generated from patients treated with anti-PD-1 alone (anti-CTLA-4 naive) or anti-PD-1 in combination with anti-CTLA-4 (anti-CTLA-4 experienced)⁷³. After alignment and quality control (STAR assign rate > 75%) and gene counts were normalized in R package edgeR¹⁰⁸, we analyzed n=30 anti-CTLA-4 experienced and n=40 and anti-CTLA-4 naïve samples. Using the 10 top expressed genes list for each signature in Cluster 0 (enriched in IgG treatment) and Cluster 1 (enriched in anti-OX40 treatment) we generated normalized z-scores of gene expressions and performed unpaired two-sided T-tests. anti-CTLA-4 experienced progressed n=15 (no clinical benefit). anti-CTLA-4 experienced non-progressed n=7 (clinical benefit). Anti-CTLA-4 naïve progressed n=9 (no clinical benefit). anti-CTLA-4 naïve non-progressed n=21 (clinical benefit). Clinical benefit is defined in Gide et al⁷³

QUANTIFICATION AND STATISTICAL ANALYSIS

Kaplan-Meier survival curves

Overall survival and TYRP1 expression data of a cohort of patients with melanoma treated with anti-CTLA-4 antibodies ⁵⁴ was retrieved from original manuscript and from cBioportal,^109,110 respectively. Melanoma samples from TCGA (SKCM, TCGA Provisional) were filtered for patients that did not receive neo- or adjuvant therapy at the time of collection. Patients were stratified based on median TYRP1 mRNA levels (Z-score). Survival rates were estimated using Kaplan Meier method. The log-rank test was used to determine survival difference between groups. All analysis was performed in R version 3.6.3 (2020–02-29).

For survival of melanoma patients based in Cluster 1 vs. Cluster 2, the mRNA median Z-scores RNA-seq expression data and clinical data of 479 skin cutaneous melanoma patients were retrieved from the TCGA-SKCM firehose legacy dataset (https://cbioportal.org). Patients with missing survival information were excluded from the study. Signature scores were calculated for each patient by extracting the expression Z-score value for the top 10 most significantly enriched genes identified in each cluster and calculating the mean Z-score value. Stratification of patients was done by splitting at the median signature score value. Survival plots were graphed using the “survival” R package (version 3.2–13) and significance was calculated using the Log-Rank method using the “survminer” R package (version 0.4.9).

Statistics

P values comparing survival curves were calculated using the Log-rank (Mantel-Cox) test. Statistical differences between experimental groups were determined by the two-tailed Student’s t-test and Log-rank test or non-parametric Mann–Whitney U-test using Prism software (GraphPad Software).

Supplementary Material

1

Supplemental Figure 1: Melanoma antigen expression as prognostic factors upon ICB treatment (Related to Figure 1)

A) Kaplan-Meier survival curves of patients with melanoma stratified based on median tumor TYRP1 expression. B) Patients with metastatic melanoma treated with CTLA-4 blockade and available gene expression profile were selected from Van Allen ⁵⁴. Patients without any neo- or adjuvant therapy at tumor resection were selected from the TCGA melanoma database. C) B16 and B78H1 melanoma cell lines were cultured in vitro and treated with IFN-γ. After 24 hours, single cell suspensions were prepared and analyzed for the presence of MHC Class I, MHC Class II, or Trp1 by flow cytometry. A representative plot with its corresponding mean fluorescence intensity (MFI) is shown.

2

Supplemental Figure 2: T cell-dependent neutrophil activation is contact independent. (Related to Figure 3).

A) Mice (5/group) were implanted with B16 cells. Mice were treated as described. After 10 days, tumor extracts were prepared and pooled. The tumor extracts were cocultured overnight with naïve neutrophils in the presence of SytoxGreen to stain extracellular DNA. One representative image is shown per condition. NET quantification is depicted by SytoxGreen area. Bars =20 μm. Box-and-whisker plots are shown for 15 representative images per condition. ***P < 0.001. B) Mice were implanted with B16 cell and treated as described. After 10 days, tumors were analyzed by immunohistochemistry. Representative slides showing H&E, Ly6G, and CD45.1 staining of tumors. A representative area is shown where CD45.1 (Trp1) staining is prevalent and no observed Ly6G (neutrophils) staining is found Bottom Right. Conversely, a representative area is shown where Ly6G (neutrophils) staining is prevalent and no observed CD45.1 (Trp1) staining is found. Bottom Left. Green bar Top= 50 μm. White bar Bottom= 20μm. C) Trp1 cells were incubated with APCs in the presence of Trp1 peptide. The next day, anti-OX40 or IgG was added to the cultures. After 6 days, the conditioned media was added to a naïve neutrophil and B78H1 co-culture. PBS or PMA was added to the coculture as negative or positive controls respectively. Bars represent averages ± s.e.m of % Casp3/7+ B78H1 cells of 4–8 different wells per condition. Dot represent individual wells. **P < 0.01. Experiments were repeated at least twice with equivalent results.

3

Supplemental Figure 3: Depletion of macrophage/monocytes and NK cells do not affect the potency of the combination therapy. (Related to Figure 3)

Mice (7/group) were implanted with B16:B78H1 mixed cells and treated as described. After 3 days, cohorts of mice received anti-Ly6G antibody or isotype control. Ten days after treatment, single cell suspensions were prepared from tumors of treated mice and the levels of neutrophils were quantified by flow cytometry. A) Representative flow plots are shown. B) Bars represent averages ± s.e.m of Ly6G_high Ly6C_mid neutrophils and Ly6G_neg Ly6C_high monocytes gated on live CD11b+CD45.2+ cells. Symbols represent individual mice. *P < 0.05. or ns=not significant. Experiments were repeated at least twice with equivalent results. C) B16:B78H1 mixed cells and treated as described. On day 3 after treatment, cohorts of mice were injected with either anti-Ly6G antibody 3 times weekly, anti-NK1.1 twice weekly, or anti-CSF1R twice weekly. All depletion antibodies were injected for 3 weeks. Growth curves of individual tumors are shown. Inset numbers indicate the number of mice that succumbed to tumor burden/the total number of mice.

4

Supplemental Figure 4: Quality control of scRNA-seq for myeloid populations in treated mice. (Related to Figure 5)

A) Violin plots of the number of genes, number of UMIs, mitochondria count percentage of mRNA for different treatments. B) Number of total UMIs over % mitochondrial UMIs left or total number of genes right for different treatments. Number represents the correlation between total genes or %mitochondrial UMI vs UMIs. C) Violin plots of the number of proteins, number of UMIs for different treatments. D) Number of total UMIs over % mitochondrial UMIs left or total number of proteins right for different treatments. Number represents the correlation between total proteins or %mitochondrial UMI vs UMI. E) The contribution of cells from tumors of each mouse to the analysis segregated by different hashtag antibodies. Note that one mouse in the IgG group did not provide any cellular contribution. F) UMAP showing different treatment by color, Immgen analysis depicting different cell populations, and unsupervised clustering depicting 18 different clusters (refer to Supplemental Table 1). G) CITE-seq UMAPs (adtUMAP) showing different treatment by color, Immgen analysis depicting different populations on adtUMAPs, and unsupervised clustering depicting clusters in adtMAP.

5

Supplemental Figure 5: Increased survival in melanoma patients expressing gene signature from Cluster 1 and neutrophil gating strategy. (Related to Figure 6)

A) Kaplan-Meier survival curves of patients with melanoma stratified based on gene signature from Clusters 0 and 1. Patients were selected from the SKCM-TCGA melanoma database. Overall survival was defined from date of specimen procurement until deceased. P values were calculated using the Log-Rank test. B) Neutrophil gating strategy from tumors.

6

Supplemental Figure 6: Heterogenous tumors are rejected by the combination therapy in PAD4 KO and Neutrophil Elastase KO mice. (Related to Figure 6)

A) NE−/− or WT mice were implanted with B16:B78H1 mixed cells and treated as described. B), NE−/− or WT mice were implanted with B16:B78H1 mixed cells and treated as described. Tumor size of individual over time is shown.

7

Supplemental Figure 7: Increased neutrophil activation in cancer patients post anti-OX40 administration. (Related to Figure 7)

Biopsies obtained from a neoadjuvant anti-OX40 study³¹ were stained for NETs. A) Experimental diagram. B) Representative images pre and post anti-OX40 administration. C) Number of NETs before and after anti-OX40 administration is shown. Green depicts more than 2-fold increase in NETs, Yellow shows less than 2-fold increase in NETs and Red depicts a decrease in NETs. Bars= 100μm.

8

Supplemental Table 1. Complete gene list for 18 clusters (Related to Figure 5)

9

Supplemental Table 2. Complete gene list for 6 clusters (Related to Figure 6)

10

Supplemental Table 3. Complete GO pathways comparing Cluster 0 vs Cluster 1 (Related to Figure 6)

Highlights.

• T cell immunotherapies can eliminate antigenically heterogenous tumors

• Neutrophils are recruited and are responsible for killing antigen loss variants

• Anti-tumorigenic neutrophils display a distinct genotype and phenotype

• Neutrophil activation is observed in patients treated with T cell immunotherapies