Where does the immune pressure go? New framework explains immunotherapy resistance

Antibodies that block the PD-1 or PD-L1 checkpoint have reshaped modern oncology, delivering durable tumor control and, in some patients, years of treatment-free survival across melanoma, lung cancer, renal-cell carcinoma and other malignancies. Yet for every patient whose tumor melts away, there are many more in whom the drugs never work at all, work only briefly, or stop working after months or years of apparent control. A major new review published in Molecular Cancer argues that this frustrating patchwork of outcomes is not a random collection of failures, but the predictable consequence of a single biological phenomenon the authors call immune-pressure redistribution.

The review, led by Xiaodong Wang and Yeqian Feng together with an international team, rejects the conventional habit of cataloguing resistance by cell type — tumor cells here, T cells there, myeloid cells somewhere else. Instead, it asks a treatment-oriented question that classical cancer immunoediting theory does not address directly: once PD-1/PD-L1 blockade releases the brakes on antitumor immunity, where does the resulting immune pressure actually go? The answer, the authors propose, falls into three interconnected routes. Tumors can transfer the pressure into cell-intrinsic escape, weaken the pressure by crippling T-cell function, or unload it entirely into stromal, vascular, myeloid, microbial and systemic host compartments.

Each route has rich mechanistic underpinnings. Transfer, the first route, is exemplified by loss of antigen presentation. Under relentless cytotoxic T-cell selection, tumors can inactivate B2M, drop allele-specific HLA class I expression, or mutate components of the peptide-processing machinery such as TAP and the immunoproteasome, rendering themselves effectively invisible to CD8 T cells. Parallel defects in JAK1/2 abolish interferon-gamma responsiveness, blunting both inducible antigen presentation and interferon-dependent growth inhibition. Beyond genetics, oncogenic rewiring through WNT/beta-catenin activation, PTEN loss, and STK11/LKB1 alterations can generate T-cell-poor, non-inflamed tumor states, while lineage plasticity allows melanoma cells to suppress differentiation antigens and adopt NGFR-high or mesenchymal phenotypes that resist immune attack.

The second route, weakening, centers on the differentiation state of the tumor-reactive T-cell pool itself. PD-1 expression alone does not mark irreversible exhaustion: the proliferative response to PD-1 blockade is driven predominantly by TCF1-positive progenitor-exhausted cells that self-renew and feed downstream effector populations. When the infiltrate is dominated by TCF1-low, TOX-high terminal states, reinvigoration fails regardless of checkpoint intensity. Compensatory checkpoints such as LAG-3, TIM-3 and TIGIT can further constrain residual function — a principle validated clinically by the survival benefit of dual nivolumab–relatlimab blockade in untreated advanced melanoma. Metabolic warfare compounds the problem: highly glycolytic tumors deplete glucose and flood the microenvironment with lactate, which suppresses effector T and natural killer cells while paradoxically fueling intratumoral regulatory T cells, which import lactate to preserve their fitness and suppressive activity.

Perhaps the most conceptually striking mechanism in the review is the temporal duality of interferon signaling. Acute type I and type II interferon signaling is essential for initiating antitumor immunity, driving dendritic-cell cross-priming, MHC-I induction and CXCL9/CXCL10-mediated T-cell recruitment. Sustained signaling, however, flips into a liability: prolonged interferon exposure establishes multigenic inhibitory programs, imprints epigenetic inflammatory memory in cancer cells, and pushes CD8 T cells toward lipid-peroxidation-associated terminal exhaustion. Two 2024 proof-of-concept studies captured the therapeutic corollary — adding the JAK1 inhibitor itacitinib after an anti-PD-1 lead-in in metastatic non-small-cell lung cancer yielded a 67 percent response rate, while ruxolitinib plus nivolumab achieved a 53 percent response rate in checkpoint-refractory Hodgkin lymphoma. Crucially, the authors distinguish this sequenced, transient JAK inhibition from tumor-cell JAK1/2 loss, which abolishes interferon sensing altogether and cannot be corrected by the same strategy.

Unloading, the third route, extends resistance beyond the tumor itself. Cancer-associated fibroblasts and TGF-beta-driven stromal programs build collagen-rich matrices and immune-excluded borders that physically separate CD8 T cells from tumor nests. VEGF-driven angiogenesis impairs dendritic-cell maturation and erects endothelial barriers to effector entry, a mechanism whose clinical relevance is supported by randomized phase III successes of PD-1/PD-L1 blockade combined with VEGF-pathway inhibition in renal-cell and hepatocellular carcinoma. Suppressive myeloid states — SPP1-high macrophages, myeloid-derived suppressor cells, emergency myelopoiesis driven by tumor-derived G-CSF and IL-6 — and regulatory T cells that intercept B7 costimulation through CTLA-4 form coordinated immunosuppressive ecosystems. Even the tumor-draining lymph node emerges as an active resistance node: PD-1/PD-L1 interactions between tumor-reactive T cells and PD-L1-positive dendritic cells within nodes can restrain the generation of the very progenitor-exhausted cells that checkpoint therapy depends on, while preclinical work suggests elective nodal irradiation can blunt combined radiotherapy–immunotherapy efficacy.

A central theme of the review is sober honesty about clinical translation. The framework explicitly separates mechanisms with prospective clinical validation from those resting on retrospective association or preclinical models. The TIGIT story is instructive: an encouraging randomized phase II signal with tiragolumab gave way to divergent phase III outcomes across tumor types, demonstrating that receptor expression is not evidence of pathway dependence, particularly given the requirement for an intact CD226 costimulatory axis. Similarly, the ECHO-301/KEYNOTE-252 failure of epacadostat plus pembrolizumab showed that systemic kynurenine suppression cannot be assumed to control intratumoral tryptophan–kynurenine–aryl hydrocarbon receptor biology, while bintrafusp alfa’s defeat by pembrolizumab in PD-L1-high lung cancer and preclinical findings that broad fibroblast depletion can accelerate pancreatic cancer warn against indiscriminate stromal targeting.

Turning biology into bedside decisions, the authors propose a biomarker-guided strategy that reads the resistance topology across five dimensions: tumor visibility (B2M, HLA-I, JAK status, mutational and neoantigen fitness), immune-cell state (TCF1 versus terminal exhaustion, co-expression of inhibitory receptors), spatial architecture (infiltrated versus excluded patterns, tertiary lymphoid structures, cDC1 niches), systemic inflammation (neutrophil-to-lymphocyte ratio, lactate dehydrogenase, IL-8 — acknowledged as prognostic rather than treatment-specific), and early treatment dynamics. Serial circulating tumor DNA offers perhaps the most actionable dynamic signal, with early clearance predicting response and molecular relapse sometimes preceding imaging, while paired baseline and on-treatment biopsies can reveal whether an immune response failed to initiate or was rapidly counter-regulated. Microbiome biomarkers, despite associations with taxa such as Akkermansia muciniphila and promising fecal microbiota transplant trials in melanoma, remain investigational given poor cross-cohort reproducibility of species-level signatures.

The practical payoff is a shift from uniform escalation to topology-matched combinations and adaptive sequencing. Fixed biallelic B2M loss nominates MHC-independent effectors such as NK-cell and bispecific platforms rather than more checkpoint blockade; an inflamed, progenitor-rich, PD-1/LAG-3 co-expressing tumor rationalizes dual checkpoint inhibition; a TGF-beta-high, fibroblast-rich excluded lesion calls for stromal or vascular modulation; and persistent post-priming interferon signaling may justify time-limited JAK inhibition. Neoadjuvant checkpoint therapy in resectable melanoma, dose-optimized ipilimumab regimens, tumor-infiltrating lymphocyte therapy after PD-1 failure, and personalized neoantigen vaccines when antigen presentation remains intact all illustrate how sequencing and route-matching can outperform reflexive combination. The authors emphasize that the framework is a trial-design strategy, not a validated algorithm — no single assay captures the topology, and most candidate biomarkers still await prospective validation.

What the review ultimately offers is a conceptual reframing with immediate research consequences. Resistance to PD-1/PD-L1 blockade stops being an inert label applied after progression and becomes a measurable, evolving process: baseline profiling defines the initial resistance topology, early sampling shows how therapy reshapes it, and subsequent interventions target whichever compartment has become rate limiting. It also dissolves apparent contradictions that have long puzzled the field — interferons both ignite and extinguish antitumor immunity, lactate starves effectors while feeding suppressors, and inflamed tumors can still resist when inflammation is misdirected in space or time. The goal, the authors conclude, is not indiscriminate immune activation but sustained alignment among tumor visibility, effector competence and a permissive tissue ecosystem — an adaptive treatment system designed to keep productive immune pressure exactly where it belongs: on the tumor.

Subject of Research: Mechanisms, biomarkers, and therapeutic strategies underlying resistance to PD-1/PD-L1 immune checkpoint blockade in cancer

Article Title: Immune-pressure redistribution in resistance to PD-1/PD-L1 blockade: mechanisms, biomarkers, and therapeutic design

Article References: Immune-pressure redistribution in resistance to PD-1/PD-L1 blockade: mechanisms, biomarkers, and therapeutic design. (n.d.). https://doi.org/10.1186/s12943-026-02786-4

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02786-4

Keywords: PD-1 blockade, PD-L1, immune checkpoint resistance, immune-pressure redistribution, cancer immunotherapy, tumor microenvironment, T-cell exhaustion, antigen presentation loss, interferon signaling, tumor-draining lymph nodes, biomarkers, combination therapy