The molecule at the center of the study is CXCL9, a chemokine that acts as a beacon for effector CD8-positive T cells. Higher CXCL9 expression has been linked to better responses to immunotherapy across multiple cancer types, making it an attractive lever for converting cold tumors into inflamed ones. The problem is that the signaling circuitry governing CXCL9 production inside tumor cells is poorly understood. The two best-known inducers, the inflammatory cytokines interferon-gamma and TNF-alpha, activate the JAK-STAT and NF-kappaB pathways respectively, but these pathways crosstalk extensively with PI3K-AKT, MAPK and p53 signaling, all of which are frequently rewired in cancer. Untangling which of these interactions actually matter in a given tumor cell is a combinatorial nightmare.

The Eindhoven team’s solution was to build interpretable mechanistic models of the signaling network and then let an active learning algorithm decide which experiments to run next. They started by curating a prior knowledge network specific to CXCL9 regulation, drawing on literature and the DoRothEA database of transcription factors. The network spans five major pathways: JAK-STAT, NF-kappaB, PI3K-AKT, MAPK and p53, connected to upstream cytokines including IFN-gamma, TNF-alpha, IFN-alpha and EGF. This scaffold was converted into a set of logic-based ordinary differential equations, a formalism that turns qualitative wiring diagrams into continuous dynamical systems without requiring the detailed kinetic parameters that are usually unknown in cancer signaling. Each edge in the network carries an adjustable strength parameter, which makes the fitted models biologically interpretable rather than black boxes.

To train these models, the researchers worked with two pancreatic cancer cell lines, AsPC1 and BxPC3, chosen because reanalysis of the Genomics of Drug Sensitivity in Cancer database showed they respond very differently to drugs. They exposed the cells to the two cytokines alone and in combination, alongside five clinically relevant inhibitors targeting JAK, IKK, PI3K, MEK and RAS, and measured secreted CXCL9 protein using a bead-based immunoassay with flow cytometry readout. The results confirmed the central role of JAK-STAT signaling: the JAK inhibitor momelotinib strongly suppressed CXCL9 in both lines, while IFN-gamma drove robust induction. More intriguingly, the PI3K inhibitor taselisib and the MEK inhibitor trametinib boosted CXCL9 expression, especially when combined with dual cytokine stimulation, pointing to previously underappreciated regulatory routes.

The fitted models, ensembles of ten optimizations per cell line, reproduced the experimental data with striking accuracy, achieving Pearson correlations of 0.998 for AsPC1 and 0.995 for BxPC3. In silico knockout experiments, in which individual regulatory edges were systematically removed from the models, then revealed context-specific control points. Deleting the ERK-AR interaction reduced CXCL9 in BxPC3 but not AsPC1, while JAK-STAT1 and STAT1-CXCL9 knockouts affected only AsPC1. The NF-kappaB pathway emerged as the key mediator of synergy between IFN-gamma and TNF-alpha in both cell lines, whereas JAK-STAT interactions contributed to synergy specifically in BxPC3. Bootstrapped parameter comparisons quantified these differences, showing that eight pathway parameters were significantly stronger in BxPC3 while two, including IFNGR-JAK, were stronger in AsPC1, providing a mechanistic explanation for the cell lines’ divergent drug responses.

The truly novel element, however, was the active learning pipeline coupled directly to these mechanistic models. Active learning is well established in drug discovery, where it helps algorithms pick the most informative compounds to test next, but it had never before been integrated with mechanistic biological models of this kind. The workflow is elegantly cyclical: the model ensemble predicts CXCL9 responses for all untested perturbation conditions, an acquisition function selects a small batch of the most valuable candidates, those are measured in the wet lab, and the models are retrained on the expanded dataset. The researchers benchmarked four acquisition strategies on synthetic data: greedy sampling, which chases conditions predicted to produce the highest CXCL9; uncertainty sampling, which targets conditions where the model ensemble disagrees most; a hybrid of the two; and random selection as a baseline.

The benchmarking produced a clear and practically useful picture. Greedy and the hybrid strategy discovered 1.4 to 1.9 times more CXCL9-inducing conditions than random sampling after five rounds, but they also generated more false positives when too many conditions were added per round. Uncertainty sampling was less aggressive at finding hits but delivered the best model generalization, reaching a mean R-squared of 0.93 across all conditions, including unseen ones, significantly outperforming every other strategy. The choice of initial training set also mattered: a carefully hand-picked set of ten conditions yielded seventeen final hits on average compared with nine for the worst random set, though the pipeline proved capable of recovering from suboptimal starts. These findings offer concrete design guidance for anyone attempting similar iterative experiments under real resource constraints.

Crucially, the team then took the pipeline back into the laboratory, running two rounds of active learning with real measurements in both cell lines. The qualitative differences between strategies seen in silico reproduced experimentally. Greedy and hybrid selections produced the strongest CXCL9 induction, while uncertainty-guided choices explored a broader response range and most consistently shrank the model’s prediction uncertainty. One complication surfaced: some greedy-selected drug combinations, such as PI3K plus ERK inhibition, yielded lower CXCL9 than expected because the high cumulative drug concentration triggered apoptosis. A Caspase-3 assay confirmed strong negative correlations between cell death and chemokine secretion, and after correcting for apoptosis, the expected hierarchy of acquisition strategies re-emerged. This observation may also help explain conflicting reports in the literature linking CXCL9 to both favorable and unfavorable prognosis in pancreatic cancer, since cytotoxicity can mask genuine immunostimulatory effects.

The study’s broader significance lies in demonstrating that mechanistic modeling and active learning, usually pursued on separate tracks, can be fused into a data-efficient engine for biological discovery. The interpretable logic-ODE framework kept the experimental design grounded in prior biological knowledge, while the learning loop squeezed maximum information from minimal measurements. Among the most tantalizing findings were the frequent selections of AKT and p53 inhibitors by the hybrid strategy, both validated as CXCL9 inducers despite sitting outside the canonical JAK-STAT and NF-kappaB regulatory axes, suggesting that less-characterized signaling mechanisms may hold untapped potential for immunomodulation. The authors caution that predictions remain constrained by the structure of the prior knowledge network, and that future work could expand it with transcriptomic data, add multiplexed readouts such as PD-L1 or TGF-beta, and employ Bayesian parameter inference for better-calibrated uncertainty. But the proof of principle stands: rational, mechanism-driven design of combination therapies aimed at warming up cold tumors is no longer a distant aspiration, but an iterative workflow that a small lab can start running today.

Subject of Research: Active learning-guided mechanistic modeling of CXCL9 chemokine regulation in pancreatic cancer cells

Article Title: Active learning-guided mechanistic modeling reveals context-specific regulators of CXCL9 expression in pancreatic cancer cells

Article References: Wang, B.-R., Yelleswarapu, M., Descamps, L., & Eduati, F. (2026). Active learning-guided mechanistic modeling reveals context-specific regulators of CXCL9 expression in pancreatic cancer cells. Molecular Systems Biology, 22(8), 1360-1375. https://doi.org/10.1038/s44320-026-00221-w

Image Credits: AI Generated

DOI: 10.1038/s44320-026-00221-w

Keywords: pancreatic cancer, CXCL9, active learning, mechanistic modeling, logic-ODE, immunotherapy, JAK-STAT, NF-kappaB, chemokines, cold tumors, drug combinations, systems biology

The enzyme at the center of the study, adenosine deaminase acting on RNA 1, or ADAR1, is one of the principal architects of the human transcriptome. It catalyzes the conversion of adenosine to inosine, a process known as A-to-I editing, in double-stranded RNA regions. Because cellular machinery reads inosine as guanosine, this editing can recode messenger RNA transcripts, alter RNA structures, and influence how RNAs are processed, translated or degraded. ADAR1 also plays a well-documented role in innate immunity, where it marks endogenous double-stranded RNA as self and prevents spurious activation of antiviral sensors. In cancer, elevated ADAR1 activity has been linked to tumor progression and immune evasion in several contexts, but its precise contribution to how bladder cancer evades immunotherapy had remained unclear until now.

Using a combination of whole transcriptome sequencing, whole genome sequencing, immunohistochemistry, and patient-derived materials, the team, led by corresponding authors Haiwei Yang, Qiang Lu and Xiao Yang, established that targeting ADAR1 in bladder cancer cells fundamentally reshapes the immune landscape of the tumor microenvironment. When ADAR1 was disabled in experimental models, two critical changes occurred. First, tumors became infiltrated by significantly more CD8-positive T cells, the cytotoxic soldiers of the adaptive immune system that are responsible for directly killing cancer cells. Second, tumor cells increased their expression of PD-L1, the checkpoint ligand through which cancers normally suppress T cell activity. That second change might at first seem counterproductive, but it is precisely what makes the tumors vulnerable to PD-1 blockade: a tumor expressing high levels of PD-L1 but crowded with CD8-positive T cells becomes an ideal target for checkpoint inhibitors, which release the molecular brakes on the waiting immune cells.

To dissect the mechanism, the researchers engineered a mutant form of ADAR1, designated ADAR1-E912A, that is defective in its RNA editing catalytic activity. Comparing cells carrying this editing-defective mutant with cells carrying wild-type ADAR1 allowed them to separate the consequences of the enzyme’s editing function from its other activities. The editing-defective mutant preserved the ability to promote CD8-positive T cell infiltration, and the team traced this effect to the chemokine CCL5, a signaling molecule known to recruit T cells to sites of inflammation and malignancy. Cells expressing the editing-defective mutant produced and secreted more CCL5, creating a chemotactic gradient that drew cytotoxic T lymphocytes into the tumor.

The molecular explanation for the increased CCL5 proved to be a matter of RNA stability. ADAR1 normally edits the messenger RNA encoding CCL5, and these edited transcripts are recognized by endonuclease V, an enzyme that cleaves inosine-containing RNA and thereby destines it for degradation. When ADAR1’s editing function was lost, CCL5 mRNA escaped this editing-dependent degradation pathway, remained stable for longer, and accumulated to higher levels within the cell. The result was a larger pool of CCL5 available for translation and secretion. In effect, ADAR1 acts as a post-transcriptional throttle on one of the most important T cell recruitment signals in the tumor, and switching off that throttle allows the tumor to summon the very immune cells it had been keeping at bay.

The second mechanism the team uncovered concerns PD-L1, the protein through which tumors engage the PD-1 checkpoint on T cells. The researchers found that ADAR1 cooperates with DICER, the cytoplasmic enzyme that processes microRNA precursors into their mature forms, to facilitate the maturation of a specific microRNA, miR-377-3p. Intriguingly, this cooperation did not require ADAR1’s editing catalytic activity, indicating that the protein performs a scaffolding or partner role in the microRNA biogenesis pathway independent of its enzymatic function. Mature miR-377-3p, in turn, represses the expression of PD-L1. In tumors where ADAR1 is abundant, the resulting PD-L1 suppression helps malignant cells keep a low immunological profile even as they hold cytotoxic T cells outside. When ADAR1 is removed, this repression is lifted, PD-L1 rises to the tumor cell surface, and the tumor becomes, paradoxically, an excellent candidate for PD-1 blockade because it now displays the molecular handle that checkpoint drugs are designed to disengage.

Perhaps the most clinically significant element of the study is the identification of an existing drug that can be repurposed as an ADAR1 inhibitor. Fludarabine, a nucleoside analogue long used as a chemotherapeutic agent in hematologic malignancies, was found to suppress ADAR1 on two fronts simultaneously: it inhibited the enzyme’s editing activity and reduced its expression. This dual suppression reproduced the effects seen with genetic targeting of ADAR1. In experimental systems, fludarabine treatment promoted CD8-positive T cell infiltration into tumors, increased PD-L1 expression on tumor cells, and sensitized bladder cancer to PD-1 blockade therapy in both in vitro and in vivo models. The finding suggests that a drug already approved and widely characterized in the clinic could be combined with checkpoint inhibitors to extend their benefits to bladder cancer patients who currently derive little benefit from immunotherapy.

The strength of the study lies partly in the breadth of its experimental evidence. The team worked across multiple platforms, including engineered bladder cancer cell lines with distinct ADAR1 statuses, co-immunoprecipitation and RNA immunoprecipitation assays to map protein and RNA interactions, immunofluorescence combined with fluorescence in situ hybridization to localize molecular events within cells, and quantitative PCR to quantify transcript changes. In vivo, they employed the N-butyl-N-(4-hydroxybutyl) nitrosamine model, a well-established chemical carcinogenesis system for bladder cancer, alongside co-culture experiments with peripheral blood mononuclear cells and experiments using patient-derived organoids that retain features of the original tumors. This triangulation across cell culture, animal models and human-derived material lends considerable weight to the conclusion that ADAR1 targeting genuinely reprograms the tumor microenvironment rather than merely altering isolated molecular readouts.

The implications extend beyond bladder cancer itself. Cold tumors across many cancer types share the fundamental problem of T cell exclusion, and any mechanism that reliably converts cold tumors into hot ones is of broad interest to the oncology community. If ADAR1 functions as a similar gatekeeper in other malignancies, the combination of an ADAR1 inhibitor with PD-1 blockade could represent a generalizable strategy. At the same time, the work highlights the dual and sometimes opposing roles that a single RNA binding protein can play: one activity, the editing of CCL5 mRNA, suppresses immune recruitment, while another, editing-independent cooperation with DICER, suppresses PD-L1 expression. Disentangling these functions will be important as inhibitors are developed and deployed. Fludarabine itself carries a known toxicity profile, and its dosing and safety when combined with checkpoint inhibitors in solid tumors will require careful clinical evaluation. Nevertheless, by identifying a druggable molecular switch that simultaneously pulls immune cells into the tumor and exposes a checkpoint target on the tumor cell surface, the Nanjing team has provided a persuasive mechanistic blueprint for turning one of immunotherapy’s most resistant solid tumors into a far more vulnerable one.

Subject of Research: Targeting the RNA editing enzyme ADAR1 to reprogram the cold tumor microenvironment of bladder cancer and enhance PD-1 blockade immunotherapy

Article Title: Targeting ADAR1 reprograms cold tumors to hot and enhances immunotherapy in bladder cancer

Article References: Bai, K., Zhuang, J., Yu, H., Lv, J., Chen, Y., Jiang, L., Li, K., Yang, H., Lu, Q., & Yang, X. (2026). Targeting ADAR1 reprograms cold tumors to hot and enhances immunotherapy in bladder cancer. Molecular Cancer. https://doi.org/10.1186/s12943-026-02778-4

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02778-4

Keywords: ADAR1, bladder cancer, RNA editing, PD-1 blockade, fludarabine, CD8-positive T cells, PD-L1, CCL5, miR-377-3p, cold tumors, tumor microenvironment, cancer immunotherapy

The central concept guiding this research is the tumor microenvironment, or TME, the scaffold of extracellular matrix within which tumor cells and host immune cells communicate and compete. Many cancers are described as immunologically cold, meaning they contain few functional cytotoxic immune cells and resist the effects of immune checkpoint inhibitors, the landmark therapies that block inhibitory receptors such as PD-1 and PD-L1 to unleash CD8-positive T cells. Checkpoint inhibitors have transformed survival for many patients since emerging in the 1990s, but their efficacy is throttled by features of the TME, including low T cell numbers, T cell exhaustion, oxygen-starved tissue, and dense populations of immunosuppressive cells. The review’s authors, led by Campbell M. Johnston and Hongji Zhang of the University of Virginia’s Department of Surgery, argue that exercise directly counters many of these barriers, effectively warming cold tumors and sensitizing them to drugs that were previously powerless against them.

One of the most detailed lines of evidence concerns tumor-associated macrophages, or TAMs, immune cells that are abundant within cold tumors and frequently promote malignancy. TAMs exist along a spectrum from the M1 phenotype, which fights tumors, to the M2 phenotype, which secretes immunosuppressive molecules such as interleukin-4, interleukin-10, PD-L1, and transforming growth factor-beta that blunt effector T cell function and empower regulatory T cells. Worse, TAMs can push CD8-positive T cells into a terminally exhausted state from which they cannot recover, and hypoxia accelerates this process. In mouse models of glioblastoma, depleting TAMs increased the proportion of stem-like progenitor-exhausted T cells and improved responsiveness to anti-PD-1 therapy, underscoring how central these cells are to treatment failure. Preclinical studies now show that aerobic exercise can tip the balance, repolarizing macrophages from the pro-tumor M2 state toward the anti-tumor M1 state and reducing total M2 numbers within tumors.

The macrophage data are remarkably consistent across exercise modalities. Breast cancer-bearing mice that ran on treadmills before and after tumor inoculation showed fewer M2 macrophages within their tumors, while medium-intensity treadmill running increased the M1-to-M2 ratio in similar models. In melanoma-inoculated mice, swimming prevented M2 polarization and reduced interleukin-6 production, inhibiting tumor glycolysis and lowering lactic acid accumulation. Exercise also increased the production of major histocompatibility complex class II molecules on macrophages, sharpening their ability to activate T cells. These findings matter clinically because pharmacological strategies targeting macrophage biology, including inhibitors of transforming growth factor-beta, have struggled in human trials, failing to show clear benefit or producing severe adverse events. Exercise, by contrast, achieves a similar biological reprogramming without toxicity, offering a route around a therapeutic bottleneck that has frustrated drug developers.

Myeloid-derived suppressor cells, or MDSCs, represent a second immunosuppressive population that exercise appears to tame. These cells promote immune evasion by impairing chemokine secretion, recruiting regulatory T cells, and increasing PD-1 expression on T cells. Multiple preclinical studies show that physical activity delays MDSC accumulation and reduces their numbers within tumors. Mice exercised before and after breast carcinoma inoculation had significantly lower intratumoral MDSC levels than sedentary controls, and treadmill running started after tumor inoculation reduced splenic MDSCs, slowed tumor progression, and increased immune cell infiltration in mammary carcinoma models, with an inverse relationship between MDSC abundance and CD8-positive T cell presence. Crucially, these findings extend to humans. In newly diagnosed breast cancer patients, a single session of acute exercise increased natural killer and CD8-positive T cell levels while reducing MDSCs. In esophageal cancer patients who exercised during neoadjuvant chemotherapy, CD8-positive T cell counts rose while inflammatory biomarkers associated with MDSCs and TAMs fell significantly.

Hypoxia and disordered blood vessel growth form a third pillar of the exercise-immunotherapy connection. Tumors grow so erratically that their vasculature becomes tangled and inefficient, starving tissue of oxygen and stabilizing hypoxia-inducible factors that drive further abnormal angiogenesis. The resulting hypoxic environment excludes natural killer and CD8-positive T cells, inhibits dendritic cells and antigen presentation, recruits immunosuppressive cells, and pushes macrophages toward the pro-tumor M2 phenotype through a CXCL8-interleukin-10 signaling axis, all of which correlate with poor prognosis and resistance to checkpoint blockade. Exercise directly counters this vicious cycle. Melanoma-bearing mice that swam at low or moderate intensity showed significantly reduced expression of hypoxia and glycolysis genes, along with greater CD8-positive T cell infiltration and cytotoxicity. Daily high-intensity exercise lowered intratumoral hypoxic fractions in breast carcinoma models, and in a landmark clinical observation, pancreatic cancer patients who exercised during preoperative therapy showed increased tumor vascularity, while exercised mice bearing patient-derived pancreatic tumors displayed vascular remodeling, accelerated regression, and delayed regrowth.

Natural killer cells, the innate immune system’s front-line tumor killers, emerge as perhaps the cells most responsive to exercise. NK cells mobilize more readily into tumors in exercised animals, and work by Cho and colleagues showed that NK cells from exercised individuals kill target cells more efficiently, with their cytotoxicity actually enhanced under hypoxic conditions, a striking advantage given the oxygen-poor nature of tumors. This resilience stems from exercise-induced metabolic reprogramming that reduces mitochondrial oxidative stress and, through interleukin-15 signaling, lessens sensitivity to reactive oxygen species such as hydrogen peroxide within the tumor microenvironment. Clinical translation is already visible: men with localized prostate cancer who adhered strictly to high-intensity interval training showed significantly increased NK cell infiltration into their tumors. Perhaps most dramatic, Pedersen and colleagues found that voluntary wheel running reduced tumor volume by 66 percent in mice lacking functional T cells, an effect abolished when NK cell production was blocked, proving that NK cells alone can mediate exercise-driven tumor suppression.

CD8-positive T cells, the primary targets of checkpoint inhibitors, are recruited into exercised tumors through well-defined molecular routes. The chemokine receptor CXCR3, which binds CXCL9, CXCL10, and CXCL11, guides these cells into tumor tissue, and CXCR3 knockout mice show reduced T cell infiltration and blunted responses to PD-1 blockade. Recent work demonstrated that four weeks of preoperative treadmill running increased CXCL9 release and CXCR3-positive T cell recruitment in colorectal liver metastases, an effect lost in CXCL9-deficient mice. Exercise also suppresses CCL5, a chemokine that recruits regulatory T cells, TAMs, and MDSCs and correlates with poor prognosis, while boosting interleukin-15, a cytokine essential for T cell survival that appears to actively shift CD8-positive cells from circulation into tumors rather than merely raising their blood counts.

The combination studies provide the most compelling case. Melanoma-bearing mice treated with exercise plus anti-PD-1 therapy developed smaller tumors with more apoptotic cells, more cytotoxic T cells, and fewer regulatory T cells than mice receiving the drug alone. Similar synergy appeared in triple-negative breast cancer, where treadmill running improved therapeutic response, boosted T cell and NK cell activation, and cut MDSC numbers alongside anti-PD-1 treatment. In transgenic breast cancer models, adding running to anti-PD-1 therapy delayed tumor growth and improved control, even when both interventions began only after tumors reached a clinically relevant size. Pancreatic ductal adenocarcinoma, notoriously resistant to checkpoint inhibitors, responded to low-intensity treadmill running combined with anti-PD-1 when neither approach worked alone, an especially promising result for patients too ill for vigorous activity. Clinically, hepatocellular carcinoma patients who exercised regularly had significantly better overall and progression-free survival on combined lenvatinib and anti-PD-1 therapy, with matching results in mouse models. Randomized trials such as HI AIM and ERICA are now testing supervised exercise before and during immunotherapy infusions in lung cancer patients.

The authors are careful to note that clinical evidence remains limited and that major questions persist regarding optimal exercise modality, intensity, frequency, timing, and patient selection across cancer types, ages, sexes, disease stages, and body mass indices. Yet the biological coherence of the evidence is difficult to dismiss: exercise relieves hypoxia, normalizes vasculature, repolarizes macrophages, suppresses MDSCs and regulatory T cells, mobilizes NK cells, and drives cytotoxic T cells into tumors through defined chemokine axes, collectively converting cold tumors into inflamed, drug-responsive ones. If ongoing adequately powered trials with longitudinal immune profiling confirm these mechanisms in patients, structured physical activity could become one of the first universally accessible adjuncts to cancer immunotherapy, a prescription written not on a pharmacy pad but into the daily routines of patients fighting some of medicine’s most treatment-resistant cancers.

Subject of Research: How physical exercise modulates the tumor microenvironment to enhance cancer immunotherapy efficacy

Article Title: Physical Exercise in Immunotherapy

Article References: Johnston, C. M., Kim, S. J., Zhang, Y., Tsung, C., Kent, E., May, A., & Zhang, H. (2026). Physical Exercise in Immunotherapy. Sports Medicine – Open, 12(1), Article 130. https://doi.org/10.1186/s40798-026-01101-1

Image Credits: AI Generated

DOI: 10.1186/s40798-026-01101-1

Keywords: physical exercise, cancer immunotherapy, immune checkpoint inhibitors, tumor microenvironment, tumor-associated macrophages, myeloid-derived suppressor cells, natural killer cells, CD8 T cells, PD-1, tumor hypoxia, interleukin-15, cold tumors