Myeloid-driven immunosuppression as a context-dependent therapeutic bottleneck in pancreatic ductal adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) remains one of the most treatment-refractory solid malignancies despite improvements in perioperative care and multi-agent chemotherapy [1,2] . Immune checkpoint inhibitors (ICIs) provide little benefit outside the uncommon mismatch repair-deficient or microsatellite instability-high subgroup, and adding PD-1/PD-L1 or CTLA-4 blockade to chemotherapy has not produced a reproducible survival benefit in unselected PDAC [3-5] . The resulting description of PDAC as an "immune-cold" tumor is convenient but biologically imprecise. A microsatellite-stable subset has high tumor mutational burden and prominent immune infiltration [6] ; neoantigen quality is associated with immunoediting and long-term survival [7] ; and individualized RNA vaccination can induce high-magnitude, polyfunctional, and durable neoantigen-specific T cells after resection [8,9] . PDAC is therefore not uniformly incapable of generating antitumor immunity. The more consistent problem is the failure to convert immune recognition into sustained tumor control.We argue that suppressive myeloid programs should be treated as a context-dependent therapeutic bottleneck rather than a universal target. When such programs are demonstrably rate-limiting, durable benefit from T-cell-directed therapy is unlikely unless the barrier is directly reprogrammed or functionally bypassed. Here, functional bypass denotes restoration of antigen presentation, lymphocyte access, or T-cell effector function without direct pharmacological targeting of the myeloid population that creates the barrier. Conversely, a change in myeloid abundance or phenotype should not be regarded as therapeutic success unless it is followed by improved tumor-reactive T-cell function. This framework does not justify pan-myeloid depletion or assume that one target applies across PDAC. It calls for selective correction of the relevant defect, integrated with antigenic priming and T-cell support.The PDAC myeloid compartment includes tumor-associated macrophages (TAMs), monocytic and polymorphonuclear MDSCs, neutrophils, and conventional dendritic-cell subsets. These populations are heterogeneous and plastic; lineage markers and the binary M1/M2 macrophage model do not reliably predict function [10][11][12] . Tumor genotype, tissue injury, hypoxia, nutrient competition, microbial products, chemotherapy, and fibroblast-derived signals can alter recruitment and cell state. Immunosuppressive P2RX1-negative neutrophils, for example, accumulate in pancreatic cancer liver metastases and are associated with impaired T-cell activity [13] . Single-cell and spatial studies further show that myeloid, fibroblast, and lymphocyte states form multicellular ecosystems that vary between patients and after treatment [14,15] . Genetically and anatomically distinct PDACs can therefore use different combinations of immune suppression [16] .Two implications follow. First, cell density is an inadequate therapeutic classifier: abundant macrophages may reflect suppressive scavenging and fibrosis in one tumor but retain phagocytic or antigen-presenting functions in another. Second, a peripheral-blood change does not establish that the relevant intratumoral state has changed. Trial design therefore needs to distinguish abundance, location, and function rather than treat all CD68-positive or CD163-positive cells as one population.Myeloid dysfunction can limit immunity at three connected stages (Figure 1). Reduced abundance or function of type 1 conventional dendritic cells impairs cross-presentation and tumor-reactive CD8 + T-cell priming. TAMs, MDSCs, and neutrophils can then contribute to chemokine imbalance, abnormal vascular trafficking, extracellular-matrix remodeling, and fibroblast activation, separating lymphocytes from malignant cells. After T cells enter the tumor, arginase activity, reactive oxygen species, adenosine, IL-10, TGF-β, and checkpoint-ligand expression can suppress proliferation, cytokine production, and cytotoxicity. Productive immunity is consequently determined not only by cell counts but also by spatial cooperation among dendritic cells, helper T cells, cytotoxic T cells, and myeloid populations [10][11][12]16,17] .This framework also explains why pan-myeloid depletion is a poor default. Dendritic cells are indispensable for priming, and some macrophage states support phagocytosis, tissue repair, and antigen presentation. Broad depletion could remove beneficial cells or induce compensatory recruitment [10][11][12]16] . The therapeutic objective is selective inhibition of a defined suppressive function or reprogramming of a harmful state while preserving host defense and antigen presentation. Selected studies are summarized in Table 1. Randomized studies have not established an incremental benefit from empirically adding checkpoint blockade or CD40 agonism to chemotherapy in unselected PDAC, whereas single-arm studies repeatedly show that immune or stromal pharmacodynamic changes can occur without proven clinical benefit. In PA.7, adding durvalumab and tremelimumab to gemcitabine plus nab-paclitaxel did not improve overall survival [4] . The negative result is compatible with multiple unresolved immune barriers, but the trial did not determine which barrier was limiting.The randomized phase II PRINCE trial was informative but not definitive. Nivolumab plus chemotherapy met its prespecified 1-year overall-survival benchmark, whereas the sotigalimab-containing arms did not [5] . The study was not designed to establish superiority between arms. In exploratory analyses, outcomes within the nivolumab/chemotherapy arm were associated with activated, antigen-experienced T-cell features, whereas outcomes in sotigalimab-containing arms were associated with antigen-presenting-cell and CD4 + T-cell features. These arm-specific associations are hypothesis-generating rather than validated predictive biomarkers.Subsequent studies further illustrate the gap between pharmacodynamic activity and demonstrated clinical benefit. CD40 agonism produced intratumoral immune remodeling in the neoadjuvant setting and preliminary activity in an uncontrolled metastatic phase Ib study [18,19] . Entinostat plus nivolumab altered myeloid and dendritic-cell features without meeting its primary efficacy endpoint [20] . Toripalimab plus gemcitabine and nab-paclitaxel yielded single-arm activity, while exploratory spatial analyses suggested that dendritic, helper T, and cytotoxic T-cell cooperation may be more informative than isolated cell density [21] . The COMBAT regimen was associated with altered immune-cell trafficking, but its multicomponent, nonrandomized design precluded attribution of either pharmacodynamic or clinical effects to motixafortide alone [22] . OPTIMIZE-1 reported encouraging activity and coordinated myeloid/T-cell activation with mitazalimab plus mFOLFIRINOX, although the single-arm design cannot establish incremental CD40 benefit [23,24] . REACtiVe-2 detected vaccine-specific T-cell responses and, in a limited set of paired biopsies, increased intratumoral T cells and reduced collagen without new objective radiological responses [25] . Plerixafor plus cemiplimab increased T-cell recruitment but produced no objective responses and coincided with enrichment of suppressive macrophage states [26] . ARC-8 linked CD73 inhibition to reduced adenosine-associated NR4A signaling and T-cell activation in paired biopsies, but lacked a concurrent chemotherapy-only control [27] . Together, these findings favor mechanism-matched trials with prespecified downstream T-cell readouts over increasingly complex all-comer combinations. Setting and intervention Key clinical and translational findings Evidentiary interpretation PA.7 [4] Randomized phase II; n=180; untreated mPDAC Gemcitabine plus nab-paclitaxel (GnP) with or without durvalumab and tremelimumab Dual ICI did not improve overall survival (P=0.72).ctDNA KRAS-wild-type status was prognostic in both arms.Regimen-specific randomized negative result: adding durvalumab and tremelimumab to GnP did not improve outcomes in unselected PDAC. The trial did not identify which immune barrier remained limiting.PRINCE [5] Randomized APX005M phase Ib [18] Open-label; n=30 (24 DLT-evaluable); untreated mPDAC Sotigalimab plus GnP, with or without nivolumabThe recommended phase II dose was 0.3 mg/kg.Responses were reported in 14 of 24 DLT-evaluable patients.Established feasibility and preliminary activity, but not CD40-specific efficacy.In PRINCE, the sotigalimab-containing arms did not meet their prespecified historical survival benchmark; the trial did not establish comparative efficacy between treatment arms.Neoadjuvant selicrelumab [19] Selicrelumab alone or with GnP before Resected tumors showed T-cell enrichment and Direct evidence of intratumoral pharmacodynamic Pilot; n=16; resectable PDAC surgery activation, more mature dendritic cells, fewer M2-like TAMs, and reduced fibrosis versus external comparator specimens.activity. The small nonrandomized cohorts and noncontemporaneous comparator specimens preclude causal efficacy or survival inference.Entinostat phase II [20] Single-arm; 27 evaluable; previously treated advanced PDAC A 14-day entinostat lead-in followed by entinostat plus nivolumab ORR was 11% (three partial responses); the primary endpoint was not met. Dendritic-cell activation/maturation and inflammatory transcriptional changes were observed.Target perturbation can coexist with poor population-level efficacy. One possible interpretation is that the magnitude, location, or functional consequence of myeloid perturbation was insufficient; the study did not establish which factor limited efficacy.Toripalimab phase Ib/II [21] Single COMBAT [22] Open-label phase IIa; two OPTIMIZE-1 [23,24] Single-arm phase Ib/II; 70 Encouraging activity with coherent pharmacodynamics, but historical comparison cannot establish incremental CD40 benefit. Biomarker associations require independent validation.REACtiVe-2 [25] Single-center phase I; n=16; Plerixafor/cemiplimab [26] Single-arm phase II; n=21; Whether concurrent macrophage-directed intervention improves efficacy requires prospective testing.ARC-8 [27] Phase Ib dose escalation/expansion; n=138; Abbreviations: AUC, area under the curve; CTL, cytotoxic T lymphocyte; DCR, disease control rate; DC, dendritic cell; DLT, dose-limiting toxicity; GnP, gemcitabine plus nab-paclitaxel; ICI, immune checkpoint inhibitor; mPDAC, metastatic PDAC; MDSC, myeloid-derived suppressor cell; NR4A, nuclear receptor subfamily 4 group A; ORR, objective response rate; OS, overall survival; PFS, progression-free survival; TAM, tumor-associated macrophage.Mechanism matching does not require a drug labeled as myeloid-targeted; it requires evidence that the prespecified myeloid bottleneck is no longer rate-limiting. For trial design, it is useful to distinguish at least three conceptual scenarios: deficient dendritic-cell priming; monocyte/macrophage-dominant suppression with stromal exclusion; and granulocytic inflammation with systemic immunosuppression. These categories are not validated clinical subtypes, may coexist within the same tumor, and should be treated as working hypotheses rather than fixed classifications. CD40 or innate-sensing approaches are most relevant when antigen presentation is inadequate. Recruitment blockade is defensible only when the targeted chemokine axis is active, whereas epigenetic or signaling interventions require evidence that the suppressive state is reversible. Early studies of epigenetic therapy underscore that tissue pharmacodynamics-not drug exposure alone-must establish whether the intended immune state changed [20,28] .Proximal target engagement must be distinguished from downstream immune rescue. For a myeloid-directed agent, target engagement may include inhibition of the prespecified pathway, a prespecified and directionally appropriate change in the intended intratumoral myeloid state, or reduced recruitment of the targeted suppressive population. Downstream mechanistic evidence should then establish improved antigen presentation, lymphocyte access, or expansion and functional activity of T-cell clones with experimentally demonstrated tumor reactivity. A fall in circulating monocytes, by itself, is insufficient without evidence of a concordant intratumoral effect or a validated blood surrogate. When a functional bypass is proposed, the protocol must define the downstream readout that would show the bottleneck is no longer limiting.A complete immune strategy treats myeloid modulation as an enabling component rather than an endpoint. ICI resistance can arise from failures in antigenicity, priming, trafficking, or effector function, and correcting one stage may leave another rate-limiting [29] . Therapeutic vaccines can generate tumor-specific T cells [30] , neoantigen-specific T-cell receptor therapy has produced objective regression in an individual patient [31] , and the nonrandomized phase I/II TACTOPS study demonstrated manufacturing feasibility, cellular persistence, and preliminary disease-control signals in a heterogeneous, selected population receiving autologous multiantigen-targeted T cells [32] . These findings establish immune potential but do not determine whether suppressive myeloid programs were directly corrected, functionally bypassed, or remained limiting.A complete regimen links antigen generation or vaccination with correction or functional bypass of the relevant myeloid barrier and sustained T-cell activity. Treatment sequence is itself a biological variable and requires prospective testing. A short myeloid-modulating lead-in may be appropriate when target engagement must precede priming, whereas other regimens may require concurrent exposure. Timing should therefore follow an explicit biological hypothesis and serial pharmacodynamic measurement rather than operational convenience.The stromal compartment also requires context-specific interpretation; fibrosis is not uniformly harmful. Combined MEK and STAT3 inhibition can remodel cancer-associated fibroblast states and improve immune responsiveness in experimental PDAC, but stromal plasticity can also generate compensatory states [33] . Fibroblast subsets may either restrict or support antitumor immunity [34] . Myeloid and stromal interventions are therefore best developed as coordinated rather than independent strategies.No single biomarker captures myeloid-driven resistance. PD-L1, tumor mutational burden, macrophage density, and circulating MDSC frequency each measure only one dimension and have important technical limitations [35] . A practical strategy integrates tumor genotype, myeloid functional state, dendritic-cell competence, T-cell fitness, spatial architecture, and systemic inflammation. When clinically safe, paired pretreatment and on-treatment biopsies provide the most direct evidence of intratumoral change; blood assays complement rather than replace tissue unless prospectively validated as surrogates.Spatial analysis is particularly important because similar cellular compositions can produce different outcomes depending on organization. Conserved dendritic-cell regulatory programs can limit antitumor immunity across cancers [36] , and spatial studies indicate that local immune cooperation can be more informative than abundance alone [15,17,21] . Durable tumor-specific T-cell activity and clinical benefit without measurable resolution of the prespecified myeloid defect would argue that the defect was not rate-limiting, was functionally bypassed, or was inadequately measured. A downstream myeloid change after T-cell activation should not be misclassified as the cause of benefit.The failure of checkpoint blockade in most PDAC should not be interpreted as evidence that adaptive immunity is irrelevant. Neoantigen studies, RNA vaccination, adoptive T-cell therapy, and spatial profiling show that tumor-specific immunity can exist or be induced. A coherent interpretation is that these responses are often limited by defective antigen presentation, exclusionary tissue organization, and suppressive myeloid function.Myeloid-driven immunosuppression is therefore best viewed as a context-dependent therapeutic bottleneck, not a universal target or an obligatory drug class. Future trials need to identify the dominant immune defect, select an intervention that corrects or bypasses that defect, verify proximal intratumoral target engagement, and determine whether tumor-specific T-cell activity and clinical outcome improve. Biomarker-enriched randomized designs will be more informative than increasingly complex all-comer combinations.The practical goal is to replace the label "immune-cold PDAC" with a testable description of where the cancer-immunity cycle fails in an individual patient. The decisive test is whether correction-or functional bypass-of a defined myeloid defect restores tumor-reactive T-cell function and improves outcome. Trials that document myeloid pharmacodynamic change without this downstream effect should be regarded as mechanistically negative.