From controversy to opportunity: experts weigh in on myeloid states as determinants in cancer immunotherapy

Across solid tumors, cells of the myeloid lineage—monocytes, macrophages, dendritic cells, and granulocytes—play major roles in cancer biology by regulating immune responses, angiogenesis, tumor cell invasion and metastasis, and treatment outcomes. These cells arise from multiple ontogenies, are highly adaptable to tissue cues, cancer-associated inflammation, obesity, and aging, with vast transcriptional diversity that reflects functional heterogeneity. Although the majority of myeloid subsets limit durable anticancer responses, populations with antitumor activity are also documented. This profound heterogeneity underscores the need for a deep understanding of myeloid functional states to define which populations should—and should not—be targeted therapeutically.

To address these challenges and the persistent gaps in mechanistic and translational understanding of myeloid cells in cancer, we convened this JITC Special Series, in collaboration with The Myeloid Network. We invited leading experts in myeloid biology to provide perspectives on the roles of myeloid cells in cancer immunotherapy¹ as well as on how their metabolic traits govern their functions.² We invited experts to tackle key outstanding questions surrounding the application and interpretation of next-generation multiomic approaches in myeloid biology; the promises and challenges of targeting granulocytes in cancer; the current and future potential of myeloid cells as clinical biomarkers, including what constitutes definitive evidence of predictive or prognostic value; and the strengths and limitations of existing experimental models used to study myeloid cell function in cancer immunotherapy. Their perspectives reflect both the remarkable progress enabled by new technologies and therapeutic strategies, as well as the conceptual, technical, and translational hurdles that continue to limit clinical impact. Collectively, these expert insights highlight areas of consensus, points of active debate, and critical gaps in knowledge, offering a roadmap for how the field can move beyond simplified frameworks toward more rigorous, mechanism-driven approaches that accelerate effective myeloid-targeted therapies.

As the field advances, it is essential to set clear definitions of the subtypes of myeloid cells that are being investigated. It is necessary to abandon the outdated and biologically inaccurate macrophage (M1/M2) and neutrophil (N1/N2) dichotomies. These frameworks represent an oversimplification, reflecting a conceptual inertia that obscures biological reality and actively impedes meaningful progress in interpreting and therapeutically targeting myeloid function in vivo. Myeloid cells exist in distinct, dynamic states that support context-dependent functions, including immune defense, tissue homeostasis, repair, and immunosuppression. Progress will require ontogeny-resolved, spatially and temporally defined, multiomic frameworks that capture myeloid states as dynamic programs, coupled with rigorous functional validation to distinguish pathogenic from protective subsets.

Significant progress has already come from multiomic analyses revealing regulation across RNA, protein, chromatin, metabolism, space, and time, defining associations between myeloid states and outcomes that are highly context-specific and patient-specific.³ A broad toolkit exists to identify myeloid cells in human and mouse tumors—tumor dissociation with immune enrichment, flow cytometry, CyTOF (cytometry by time of flight, or mass cytometry, which is a high-dimensional single-cell analysis technique that uses metal-tagged antibodies instead of fluorophores to detect over 40-50 protein markers simultaneously), single-cell and spatial transcriptomics, CITE-seq (Cellular Indexing of Transcriptomes and Epitopes by sequencing, which is a single-cell multiomics technique that simultaneously measures cell surface proteins and mRNA from the same individual cells), and immunohistochemistry—each with complementary strengths and distinct limitations. Modeling human myeloid biology in mice remains imperfect; species differences, artificial inflammatory states, and transplantable tumor systems can distort myeloid recruitment and activation relative to human disease. Genetically engineered mouse models, orthotopic tumors, and alignment to human multiomic datasets improve relevance but are costly and still fail to fully capture human myeloid complexity, particularly for granulocytes. An essential priority is to better align preclinical models, biomarkers, and inflammatory states with defined human disease contexts to enable predictive translation and rational development of myeloid-targeted therapies.

Favorable responses are associated with myeloid programs supporting antigen presentation, type I interferon signaling, and pro-inflammatory, immune-stimulating functions. Unfavorable responses are linked to immunosuppressive and metabolically rewired states, including arginase-1 (ARG1), folate receptor beta (FOLR2), interleukin-10 (IL-10), osteopontin (OPN/SPP1), and triggering receptor expressed on myeloid cells 2 (TREM2) programs as well as lipid metabolism, lactate accumulation, and suppressive metabolites.^{1 2 4 5} These states appear conserved across tumor types, suggesting shared immune-evasion mechanisms rather than purely tumor-specific phenomena. The field must address which molecular features are drivers versus correlates of immune responses, whether unfavorable states can be safely and durably reprogrammed, how temporal trajectories (state transitions over time) determine treatment response, and how myeloid heterogeneity can be incorporated into predictive biomarkers and therapy design.

Granulocytes—particularly neutrophils—are highly heterogeneous and cannot be reduced to simple protumor or antitumor categories. Their functions span direct tumor cell killing, immune activation, angiogenesis, immune suppression, and metastatic support in a context-dependent and stage-dependent manner. In most chronically inflamed human tumors, tumor-associated neutrophils predominantly support tumor growth through immunosuppression, neutrophil extracellular trap (NET) formation, angiogenesis, and facilitation of metastasis. Antitumor neutrophil activity is more often observed in early disease or following therapeutic interventions that reprogram the tumor microenvironment.6,9 Tumor-derived factors and microenvironmental conditions such as hypoxia and metabolic changes support accumulation of immature, immunosuppressive neutrophils, which inhibit T and natural killer cell function and correlate with poor clinical outcomes. NETs, while critical for host defense, are frequently co-opted in cancer to remodel extracellular matrix, prime pre-metastatic niches, and suppress antitumor immunity. Pathological neutrophil activation arises not from new identities but from chronic, dysregulated deployment of otherwise protective programs, initiated early during bone-marrow development and reinforced within blood and tumor niches. The term myeloid-derived suppressor cells (MDSCs) has been used to describe a heterogeneous population of pathologically activated immature myeloid cells—including neutrophil-like polymorphonuclear cells and monocytic cells—that arise in cancer and chronic inflammation and are defined by their immunosuppressive function rather than a stable lineage identity. Accordingly, the term MDSC is most informative when applied carefully and in appropriate biological context to avoid overlooking meaningful myeloid diversity.

A central unknown is how to distinguish and therapeutically manipulate transient, beneficial granulocyte activation versus sustained, tumor-educated pathological states to improve outcomes without inducing systemic immunopathology. Therapeutically, selective granulocyte targeting is likely to be most effective when tumor-promoting neutrophil states dominate in the chronically inflamed or pre-metastatic setting—and combination approaches—rather than indiscriminate neutrophil depletion—appear most promising.¹⁰ The unresolved challenge is how to selectively modulate pathological neutrophil while preserving antimicrobial defense and harnessing context-dependent antitumor neutrophil functions.

Eosinophils remain less understood, but emerging clinical and preclinical evidence links them to improved immunotherapy responses, potentially through recruitment and activation of CD8+T cells, although their low abundance, fragility, and functional heterogeneity remain major gaps. Similar challenges are faced for monocytes and macrophages. The central imperative is moving beyond descriptive atlases toward causal, spatially and temporally resolved biology that translates into precise, validated myeloid-targeted therapies and biomarkers. Promising targets and biomarkers increasingly focus on pathways driving pathological myelopoiesis and immune suppression—such as colony-stimulating factor 1 receptor (CSF1R/CD115), inflammasome/IL-1 signaling, IL-4 signaling, type I interferon, phosphoinositide 3-kinase gamma (PI3Kγ), spleen tyrosine kinase (SYK) / Bruton’s tyrosine kinase (BTK), and soluble factors like S100A8/A9—often with greatest efficacy in combination and sequential strategies.11,16 Redundancy and compensation are major barriers: high plasticity allows myeloid cells to rapidly compensate when individual cytokines or factors are targeted. Identifying mechanisms fundamental to the survival and activity of pathological myeloid states will be required. Another key unknown is how to optimally sequence and combine myeloid-modulating therapies with cytotoxic agents and immune checkpoint inhibitors to suppress deleterious programs while preserving or enhancing myeloid-driven antitumor immunity.

Despite its transformative power, multiomics is limited by static snapshots that obscure dynamic transitions and causality, technical and analytical challenges in cross-modal integration, and costs that impede scalability and clinical translation. Current approaches are largely correlative and constrained by insufficient incorporation of spatial organization and temporal context. Moving forward, the field must develop standardized, reproducible analytical frameworks, rigorously link multiomic states to in vivo function, and define which myeloid programs are truly actionable for therapy.