Myeloid drugs rarely prove immune conversion in solid tumors

A targeted review of myeloid-directed cancer drugs finds that most programs show pharmacological activity without proving that suppressive cells have been converted into a state that restores antitumor immunity. The authors introduce the Myeloid Immune Conversion Audit Trail, or MICAT, a four-level framework distinguishing intratumoral target engagement, conversion of the intended myeloid state, propagation through the tumor immune ecosystem, and randomized clinical attribution. Applying the rubric to representative studies, they conclude that no carcinoma trial reviewed has completed the full causal chain.

Across several strategies, compensation between myeloid lineages, selective resistance and damage to beneficial immune functions repeatedly interrupt otherwise plausible therapeutic effects. The review, published on October 20, 2026, in Frontiers in Pharmacology, examines macrophages, monocytes, neutrophils, myeloid-derived suppressor cells and dendritic cells that jointly regulate immune exclusion, T-cell dysfunction, angiogenesis, stromal remodeling and resistance to immune checkpoint blockade. These populations are not interchangeable drug targets.

Human tumor-associated macrophages can simultaneously express inflammatory, scavenging, angiogenic, lipid-handling, matrix-remodeling and antigen-presentation programs, making the familiar M1-versus-M2 classification unsuitable as a pharmacodynamic endpoint. Suppressive-cell phenotypes are also difficult to identify reliably because polymorphonuclear myeloid-derived suppressor cells overlap with neutrophils, while monocytic populations overlap with conventional monocytes. Studies should therefore predefine the population being targeted, report absolute cell abundance, distinguish depletion from redistribution and use functional or spatial evidence alongside marker expression.

MICAT extends an established pharmacological audit trail by inserting an ecosystem-propagation step between modification of a myeloid cell and a clinical outcome. Each level receives a score from zero to two: zero for absent or negative evidence, one for partial or indirect evidence, and two for a demonstrated criterion. Level 1 requires direct tumor exposure or receptor engagement linked to proximal pharmacodynamic effects.

Level 2 requires a prespecified lineage-specific change in cell state or function, with absolute abundance accounted for. Level 3 demands concordant evidence in at least two downstream domains, such as antigen-presenting-cell function, adaptive-effector activity, clonal change or spatial organization. Level 4 requires benefit attributable to the experimental agent in a component-controlled randomized comparison.

The eight-point total measures completeness of causal evidence, not drug potency, commercial promise or probability of future success. CSF1R inhibition illustrates the difference between target activity and immune conversion. In the randomized phase III ENLIVEN trial for advanced tenosynovial giant-cell tumor, pexidartinib produced a week-25 response in 39% of patients, compared with none receiving placebo, establishing randomized benefit in a strongly CSF1-dependent neoplasm.

That result does not prove immune conversion in carcinomas because the trial did not demonstrate conversion of a myeloid state and downstream ecosystem propagation. In the phase I MEDIPLEX trial in colorectal and pancreatic cancers, pexidartinib plus durvalumab produced systemic CSF1R pharmacodynamic effects, but only one of 47 patients achieved a partial response, in MSI-high colorectal cancer. Pexidartinib’s additional inhibition of FLT3 and c-KIT impaired dendritic-cell differentiation, reduced dendritic-cell subsets and weakened interferon-lambda-related function, showing how a nominally macrophage-directed drug can damage the antigen-presenting bridge required for checkpoint therapy.

Antibody programs targeting CSF1R have similarly separated target engagement from clinical benefit. Emactuzumab plus atezolizumab reduced CSF1R-dependent populations and produced objective response rates of 5.6% to 12.5% across cohorts, without evidence that checkpoint resistance had been reversed. AMG 820 plus pembrolizumab increased CSF1 and IL-34, reduced selected circulating monocytes and raised several intratumoral immune markers, yet immune-related partial responses occurred in only 3% of 116 patients.

Toxicities include edema, transaminase elevations, fatigue and rash, while pexidartinib also carries risks of mixed or cholestatic hepatotoxicity. The review does not invalidate CSF1R as a target, but narrows the conditions for success: carcinoma trials should select a defined suppressive macrophage state, document its disappearance in tumor tissue, verify preservation of conventional type 1 and type 2 dendritic cells, and compare continuous with pulsatile schedules. Recruitment blockade is vulnerable to biological redundancy.

CCR2 and CCR5 direct monocyte trafficking, while CXCR1 and CXCR2 recruit neutrophils and polymorphonuclear suppressor-like cells. In paired samples from pancreatic ductal adenocarcinoma, CCR2 inhibition reduced monocyte and macrophage influx but allowed CXCR2-positive neutrophils to substitute; mouse experiments showed that blocking either population could expand the alternative one. Clinical results have been inconsistent.

In first-line pancreatic cancer, BMS-813160 with chemotherapy and nivolumab produced a 37% response rate, compared with 26% for BMS-813160 plus chemotherapy and 28% for chemotherapy alone; in second-line colorectal cancer, neither experimental regimen clearly improved activity over FOLFIRI. Earlier CCR2 blockade produced encouraging signals in small, nonrandomized studies but not convincing added efficacy in metastatic disease, and it was associated with pulmonary toxicity in 24% of one trial’s patients, including grade 4 hypoxia. Navarixin with pembrolizumab showed little activity across prostate cancer, microsatellite-stable colorectal cancer and non-small-cell lung cancer.

PI3Kγ inhibition offers a different mechanism because it seeks to reprogram myeloid cells without necessarily depleting them. In the single-arm phase II MARIO-3 trial of eganelisib, atezolizumab and nab-paclitaxel in first-line metastatic triple-negative breast cancer, paired biopsies from 29 patients showed increased MHC class I and II and CXCL9/CXCL10 programs, reduced SPP1, metalloprotease and extracellular-matrix programs, and an increased CXCL9-to-SPP1 balance. The digital-spatial and multiplex-imaging analyses used smaller subsets of 12 and 11 patients, respectively.

These regimen-level changes were consistent with myeloid-state conversion and ecosystem propagation, but the trial did not demonstrate agent-specific intratumoral PI3Kγ engagement, and both chemotherapy and checkpoint blockade can alter the same readouts. The decisive test proposed by the review is a randomized comparison using the same chemotherapy and checkpoint backbone with or without eganelisib, direct tumor engagement and a conversion score locked before unblinding. CD40 agonism attempts to license dendritic cells, restore co-stimulation and cross-priming, and reshape macrophage and stromal interactions, but receptor clustering, epitope, valency, tumor anchoring, dose and schedule can make the agents behave differently.

In the randomized phase II PRINCE trial, neither sotigalimab-containing arm met its prespecified one-year overall-survival benchmark, although exploratory CD4, dendritic-cell and antigen-presenting-cell features were associated with benefit. Single-arm studies of mitazalimab reported encouraging response and survival estimates alongside dynamic circulating immune changes, but could not isolate its contribution. Tumor-localized agonists made the distinction between delivery and efficacy especially clear: MP0317 and RO7300490 achieved tumor localization and partial CD40 pathway activation, yet objective responses were absent or exceptional.

The review places high priority on sequence-aware studies in resectable pancreatic ductal adenocarcinoma, where surgery can provide tissue for measuring antigen-presenting-cell activation, stromal remodeling and T-cell effects. CD47–SIRPα blockade seeks to release macrophage inhibition of phagocytosis, but CD47 is widely expressed on normal cells, creating anemia, thrombocytopenia, hemagglutination, transfusion interference and infusion-reaction risks. In first-line head and neck squamous-cell carcinoma, adding magrolimab to pembrolizumab and chemotherapy did not improve progression-free survival and increased grade 3–4 adverse events.

ELEVATE CRC was also stopped early: median progression-free survival was 6.2 months with magrolimab plus bevacizumab-FOLFIRI and 6.7 months with bevacizumab-FOLFIRI alone, while grade 3 or higher treatment-emergent adverse events occurred in 72.7% and 52.4% of the respective groups. An exploratory conference report from ASPEN-06 suggested that fresh-biopsy confirmation of retained HER2 might enrich activity for evorpacept paired with HER2-directed therapy, but the historical comparison was nonsignificant and the subgroup analysis exploratory. Preservation of the tumor antigen used for opsonization is therefore a candidate biomarker requiring prospective confirmation.

Emerging targets and engineered platforms remain earlier in development. Early Clever-1, TREM2 and LILRB2 programs established dose feasibility and initial safety profiles, but peripheral occupancy, exploratory interferon signatures and small paired-tissue subsets did not establish intratumoral target conversion. In vivo lipid-nanoparticle macrophage programming, bispecific myeloid engagers and artificial-intelligence-assisted spatial scoring remain subject to the same evidentiary requirements as conventional drugs.

The review’s practical rule is strict: stop dose escalation when tumor engagement is saturated but myeloid conversion is absent, redesign a combination when conversion fails to propagate, and withhold claims of a validated immune switch until randomized benefit completes the causal chain.

Subject of Research: Evidence standards for pharmacological reprogramming of suppressive myeloid cells in solid tumors

Article Title: Pharmacological reprogramming of myeloid suppression in solid tumors: mechanisms, evidence standards for immune conversion, and biomarker-informed combinations

Article References: Maletin, N., Denda, N., Paut Kusturica, M., Rašković, A., Babović, S., & Golušin, Z. (2026). Pharmacological reprogramming of myeloid suppression in solid tumors: mechanisms, evidence standards for immune conversion, and biomarker-informed combinations. Frontiers in Pharmacology, 17, Article 1986842. https://doi.org/10.3389/fphar.2026.1986842

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1986842

Keywords: myeloid immune conversion, solid tumors, tumor-associated macrophages, immune checkpoint blockade, CSF1R inhibition, PI3Kγ inhibition, CD40 agonism, CD47 blockade, pharmacodynamic biomarkers, MICAT, Pharmacological, reprogramming