T-cell engagers, often called TCEs or bispecific T-cell redirecting molecules, are designed to do exactly what their name suggests: physically bridge a T cell to a tumor cell. One arm of the molecule binds a tumor-associated antigen on the cancer cell surface, while the other binds CD3, a component of the T-cell receptor complex. When both arms engage their targets, the T cell is forced into close contact with the tumor cell, triggering activation, release of cytotoxic granules, and killing of the malignant cell. The approach has produced striking clinical successes in blood cancers, where antibodies can reach malignant cells circulating in the bloodstream. Solid tumors, however, have proven far more stubborn, and gastric cancer is a textbook example of why.

The obstacle is not usually the lack of a good target. CLDN18.2, a member of the claudin family of tight-junction proteins, is expressed on the surface of gastric cancer cells and is largely absent from most adult tissues outside the stomach mucosa, making it one of the most attractive antigens in the field. The problem is that many gastric tumors are immunologically cold—poorly infiltrated by T cells, lacking the chemokine signals that recruit them, and dominated by immunosuppressive stromal features. A T-cell engager can only redirect T cells that are actually present in or near the tumor. If the tumor bed contains few CD8-positive cytotoxic T cells, even a perfectly designed bridging molecule has little to work with.

The design of the new molecule began in silico. Rather than starting from an existing antibody, the team used an RFdiffusion-assisted workflow—a computational approach that generates novel protein structures by a diffusion-based process, effectively imagining new binding surfaces that can grip a chosen target. Candidate CLDN18.2 binders were screened computationally, and the most promising designs were incorporated into both human and murine CD3-engaging T-cell engager formats, allowing the researchers to test the concept in parallel across human-relevant and mouse-model systems. This design philosophy is part of a broader shift in protein engineering: instead of repurposing natural scaffolds, scientists can now specify a desired binding interaction and let generative models propose molecular architectures that satisfy it.

The selectivity testing focused on a critical safety concern. Claudin proteins are close relatives, and a therapeutic that binds CLDN18.1 as well as CLDN18.2 could damage healthy lung epithelium, where CLDN18.1 is found. In cell-surface binding assays, the selected engager bound CLDN18.2-positive tumor cells and CD3-positive T cells while showing limited detectable binding to cells expressing CLDN18.1. Functional testing followed in increasingly realistic settings: engineered gastric cancer cell lines, patient-derived tumor cells, and patient-derived organoids—three-dimensional cultures grown from actual patient tissue that preserve much of the cellular complexity of the original tumor. Across all of these platforms, the molecule induced CLDN18.2-dependent tumor-cell killing and activated CD8-positive T cells, confirming that the computationally designed binder could perform its biological job, not just bind its target in a test tube.

Then came the in vivo experiments, and with them the study’s central twist. In both syngeneic mouse models—immunocompetent animals with intact immune systems—and humanized models created by reconstituting mice with human peripheral blood mononuclear cells and carrying patient-derived tumor xenografts, the T-cell engager produced early tumor inhibition. But as the experiments progressed, tumor control weakened. The engager was bridging T cells to tumor cells, yet the effect was not sustaining itself. The team examined the tumors and found a consistent pattern: limited accumulation of T cells inside the tumor, increased expression of PD-1, the canonical marker of T-cell exhaustion, sparse populations of CD11c-positive dendritic cells, and weak signaling through CXCL9 and CXCL10, chemokines that normally act as beacons drawing T cells into tissue.

That molecular fingerprint pointed directly at the missing ingredient. Dendritic cells are the professional sentinels and coordinators of the immune system; they present antigen to T cells, secrete inflammatory cytokines such as interleukin-12, and produce the very chemokines—CXCL9 and CXCL10 among them—that recruit effector T cells into tumors. The researchers hypothesized that supplementing the T-cell engager treatment with additional dendritic cells might rebuild the local immune circuit that the cold tumor environment lacked. The experiment confirmed the hypothesis: adding dendritic cells improved TCE-mediated tumor-cell killing in co-culture systems and strengthened tumor control in vivo, converting a waning response into a more durable one.

To understand exactly how the dendritic cells were helping, the team ran mechanistic experiments using blocking antibodies. When they blocked MHC-II, the molecule dendritic cells use to present antigen to helper T cells, the enhancement was only modestly reduced. When they blocked interleukin-12, however, the dendritic-cell-associated boost was consistently diminished. The interpretation is significant for the field: the added dendritic cells were not primarily acting as antigen-presenting machines in this setting, but as cytokine factories, providing the inflammatory support—IL-12 and the downstream CXCL9/CXCL10 chemokine axis—that keeps redirected T cells active, recruited, and functional inside the tumor.

The implications extend well beyond gastric cancer. The study demonstrates that generative protein design tools like RFdiffusion can produce functional, selective therapeutic binders against clinically validated targets, compressing a discovery process that once took years of antibody engineering. Just as importantly, it offers a candid lesson about the limits of that technology: a beautifully designed molecule that bridges T cells to tumor cells cannot, by itself, overcome a tumor microenvironment that lacks the supporting immune infrastructure. The authors’ conclusion—that CLDN18.2-directed T-cell engagers should be tested together with strategies that rebuild local dendritic-cell–T-cell circuits—frames a design principle likely to shape the next generation of solid-tumor immunotherapy combinations.

Caveats remain, as they do in any preclinical translational study. The work rests on engineered cell lines, patient-derived cultures, organoids, and mouse models rather than human clinical trials, and the published version is an early-release, peer-reviewed accepted manuscript that may undergo further editorial refinement. The dendritic-cell supplementation strategy, while mechanistically compelling, would require substantial development before clinical application—questions of cell sourcing, manufacturing, timing, and delivery all remain open. Still, the study offers a clear and testable roadmap: pair AI-designed targeting molecules with interventions that restore the chemokine and cytokine signals cold tumors fail to produce. For patients with CLDN18.2-positive gastric cancer, a disease with limited options in advanced stages, that roadmap represents a rational path from computational design to combination therapy, one that the field will be watching closely as these strategies move toward clinical evaluation.

Subject of Research: RFdiffusion-designed CLDN18.2 T-cell engager combined with dendritic-cell supplementation for gastric cancer immunotherapy

Article Title: RFdiffusion-designed CLDN18.2 T-cell engager coupled with dendritic-cell supplementation enhances antitumor immunity in gastric cancer

Article References: Kan, H., Niu, G., Wan, Z., Zou, Y., Wang, J., Xu, Y., Yang, H., Xu, Z., & Li, B. (2026). RFdiffusion-designed CLDN18.2 T-cell engager coupled with dendritic-cell supplementation enhances antitumor immunity in gastric cancer. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-08841-4

Image Credits: AI Generated

DOI: 10.1186/s12967-026-08841-4

Keywords: RFdiffusion, T-cell engager, CLDN18.2, gastric cancer, dendritic cells, IL-12, CXCL9, CXCL10, protein design, immunotherapy, organoids, patient-derived xenografts

Mitochondria are far more than passive energy suppliers. In cancer cells, they participate in metabolic reprogramming, redox balance, cell death decisions, and signaling to the surrounding tumor microenvironment. To keep these organelles functional, cells run a quality control program that includes mitochondrial fission and fusion, mitophagy, the removal of damaged proteins, and the maintenance of mitochondrial DNA integrity. Disruptions in this program have been implicated in tumor progression and treatment resistance, but the question the Chinese team set out to answer was whether the transcriptional activity of mitochondrial quality regulation genes carries meaningful prognostic information in breast cancer, and whether it connects to the immune and stromal features of tumors.

To build their gene panel, the authors curated twenty mitochondrial quality regulation genes from MitoCarta 3.0, a comprehensive inventory of mitochondrial proteins, together with genes identified in the published literature. They then integrated two large public datasets, The Cancer Genome Atlas breast cancer cohort and the GSE20685 dataset, yielding 1,422 samples for molecular subtyping and 1,421 with survival information. Using consensus clustering, a method that groups samples based on the stability of their expression patterns across repeated clustering runs, the team identified two transcriptional states defined by the activity of these mitochondrial genes. The distinction was not merely descriptive. Patients in Cluster A had significantly better overall survival than those in Cluster B, with a P value of 0.006, indicating that the mitochondrial quality control signature captured clinically relevant tumor biology.

Having established that these transcriptional states matter, the researchers moved from subtypes to a quantitative tool. They applied Cox regression and LASSO-Cox analysis, a technique that shrinks regression coefficients to prevent overfitting and select the most informative predictors, to distill the signal into a seven-gene prognostic score. The score stratified survival in the training subset, the internal testing subset, and the pooled cohort, with all comparisons reaching P values below 0.001. Importantly, the team did not stop at unadjusted associations. After accounting for available clinicopathological factors, the standardized score remained significantly associated with overall survival in the source-stratified pooled analysis, with a hazard ratio of 1.775 per standard deviation increase and a 95 percent confidence interval of 1.554 to 2.027. In other words, each standard deviation rise in the score corresponded to roughly a 78 percent increase in the hazard of death.

A recurring weakness of prognostic signatures in oncology is their failure to replicate outside the data in which they were built. To address this, the authors harmonized expression scales across platforms in an outcome-blind manner and tested the score in GSE199633, an independent cohort of 637 breast cancer patients. The score retained its prognostic association in this external validation, with a hazard ratio of 1.378 and a 95 percent confidence interval of 1.109 to 1.712, and a P value of 0.004. The authors are candid about the limits of this result: adding the score to existing models improved model fit, but the incremental gain in discrimination, the ability to correctly rank patients by risk, was modest. This transparency is notable in a field where headline claims often outpace the statistical fine print.

Beyond survival prediction, the study connects mitochondrial quality regulation to the tumor microenvironment, the ecosystem of immune cells, fibroblasts, blood vessels, and signaling molecules that surrounds cancer cells. Using three complementary computational approaches, single-sample gene set enrichment analysis, CIBERSORT for estimating immune cell composition from bulk RNA data, and ESTIMATE for scoring stromal and immune content, the team found that high-risk tumors displayed distinct immune and stromal features derived from bulk expression profiles. High-risk tumors also carried a higher tumor mutational burden, a measure of the number of mutations per megabase of DNA that has become an important biomarker in immunotherapy research. The convergence of mitochondrial transcriptional states, immune infiltration patterns, and mutational load suggests that mitochondrial quality control may help shape the evolutionary and immunological context in which tumors grow.

The most striking tissue-level finding concerns CXCL9, a chemokine known for recruiting T cells to sites of inflammation and tumors. Using reverse transcription quantitative PCR and Western blotting, the researchers confirmed that CXCL9 protein abundance was higher in paired tumor tissue than in adjacent non-tumor tissue, with a P value below 0.001. They then turned to multiplex immunohistochemistry, a technique that stains multiple protein markers simultaneously on a single tissue section, applied to a 120-case tissue microarray. The spatial analysis revealed a positive correlation between CXCL9 signal and the density of iNOS-positive M1-like macrophages, the classically activated, antitumor-polarized members of the macrophage family, with a correlation coefficient of 0.3203 and a P value below 0.001. This links the bulk-expression signature to an actual cellular population visible under the microscope, a step that many computational prognostic studies never take.

The study’s design reflects the current best practice in bioinformatic oncology research: multi-cohort integration, adjustment for confounders, external validation, and experimental confirmation of selected computational findings. The tissue microarray work was approved by the Ethics Committee on Biological Science and Technology of Hunan Aifang Biological Co., Ltd., and conducted under the Declaration of Helsinki with written informed consent. The research was funded by the Interdisciplinary Research Program in Medicine and Engineering and the Clinical Medical Research 4310 Program of the University of South China. The corresponding author is Yuehua Li, and the first three authors, Huaiwen Pu, Tingjing Li and Renji Liang, contributed equally to the work.

What makes this study worth watching is its framing of mitochondria as a bridge between tumor-intrinsic biology and the immune landscape. If mitochondrial quality regulation genes help determine whether a breast tumor recruits antitumor macrophages or evolves a high mutational burden, they could eventually inform decisions about immunotherapy, patient risk stratification, and the design of trials targeting metabolic pathways. The authors themselves are careful to position the seven-gene score as a framework that warrants further independent clinical validation and mechanistic investigation rather than a ready-made clinical test. The modest incremental discrimination in the external cohort is a reminder that even well-validated signatures must earn their place alongside established clinicopathological variables. Still, the convergence of survival associations, microenvironment characterization, and spatial protein-level evidence gives this mitochondrial angle a credibility that many gene-signature papers lack, and it opens a concrete research agenda for laboratories interested in how the quality control of cellular power plants shapes the fate of breast tumors and the patients who carry them.

Subject of Research: Mitochondrial quality regulation gene expression, prognostic modeling, and tumor microenvironment characterization in breast cancer

Article Title: MQRG-associated transcriptional heterogeneity and a seven-gene prognostic signature in breast cancer: multi-cohort evaluation and tumor microenvironment characterization

Article References: Pu, H., Li, T., Liang, R., Fan, Z., Tang, B., Luo, Y., Yi, X., Xie, L., & Li, Y. (2026). MQRG-associated transcriptional heterogeneity and a seven-gene prognostic signature in breast cancer: multi-cohort evaluation and tumor microenvironment characterization. BMC Cancer. https://doi.org/10.1186/s12885-026-16950-y

Image Credits: AI Generated

DOI: 10.1186/s12885-026-16950-y

Keywords: breast cancer, mitochondrial quality control, prognostic signature, tumor microenvironment, CXCL9, transcriptomics, bioinformatics, TCGA, macrophages, tumor mutational burden, gene expression profiling, multiplex immunohistochemistry

Hirschsprung disease arises when enteric neural crest cells fail to migrate and differentiate properly during development, leaving the distal intestine devoid of the myenteric and submucosal nerve plexuses that normally coordinate peristalsis. The result is abdominal distension and refractory constipation, typically treated by surgically removing the aganglionic segment and reconstructing the bowel. Yet a substantial proportion of patients continues to suffer long-term complications, including enterocolitis, anastomotic strictures, fecal soiling, persistent constipation, and the need for reoperation. These persistent problems have intensified the search for the molecular mechanisms underlying the disease, and the new work points squarely at the glial microenvironment.

Enteric glial cells are the most abundant non-neuronal component of the enteric nervous system, and they are far more than passive scaffolding. They maintain epithelial integrity, coordinate neuro-immune signaling, and support neuronal survival. Recent single-cell RNA sequencing studies have revealed that glia exist in multiple transcriptionally distinct states, some of which resemble reactive microglia in the brain. In tumor-associated macrophages, researchers have described a CXCL9-to-SPP1 polarization axis that separates anti-inflammatory from pro-disease functional states. Whether enteric glia undergo a comparable polarization program in Hirschsprung disease had never been established, and that question became the starting point for the investigation.

The team began with an integrated single-cell transcriptomic analysis of colonic tissue from a mouse model of Hirschsprung disease based on loss of the endothelin receptor B gene, Ednrb, alongside wild-type controls. Among the five transcriptionally distinct glial subclusters they identified, one stood out: a subcluster enriched for both TXNRD1 and SPP1 that expanded markedly in disease samples, while a subcluster marked by Apoe shrank. Enrichment analyses of this expanded population highlighted extracellular matrix organization, cell adhesion, focal adhesion, and PI3K-Akt and MAPK signaling programs. Computational modeling of ligand-receptor interactions further showed that this subcluster engaged in the strongest predicted communication with enteric neurons of any glial population, including a predicted VCAM1-NCAM1 signaling axis.

To probe the regulatory logic behind this glial remodeling, the researchers turned to network-based virtual perturbation using the scTenifoldNet framework, which simulates gene knockout or overexpression within an inferred gene regulatory network. Both virtual knockout and virtual overexpression of TXNRD1 converged on regulatory programs involving redox control, ERK/MAPK-associated immediate-early response genes, and stress-associated glial features. Virtual overexpression produced a more focused pathological shift, with SPP1 emerging as the most prominently perturbed gene. These in silico results positioned TXNRD1, a central regulator of intracellular redox homeostasis, as a plausible upstream driver of the disease-associated glial state.

Human tissue evidence followed. Proteomic profiling of paired proximal dilated and stenotic colonic segments from children undergoing pull-through surgery revealed marked TXNRD1 upregulation in the stenotic, aganglionic tissue, together with significant enrichment of the MAPK signaling pathway. Validation in eight paired patient samples using quantitative PCR, western blotting, and immunofluorescence confirmed elevated TXNRD1 alongside increased phosphorylation of Raf, MEK, and ERK, while the stress kinases p38 and JNK showed no significant changes, indicating selective activation of the ERK branch of the MAPK cascade. Immunofluorescence staining also showed that SPP1 was increased and CXCL9 reduced within S100β-positive enteric glia in the stenotic segments, mirroring the single-cell findings in human tissue.

The mechanistic experiments in cultured rat enteric glial cells tied these observations together. When the researchers overexpressed TXNRD1, the cells selectively increased ERK phosphorylation, elevated their production of SPP1, suppressed CXCL9, accumulated reactive oxygen species, migrated less effectively, and secreted less glial cell line-derived neurotrophic factor, or GDNF. Treatment with the ERK inhibitor SCH772984 partially restored CXCL9 expression and blunted SPP1 upregulation, while the antioxidant N-acetyl-L-cysteine reduced ROS accumulation, dampened ERK activation, and reversed the SPP1-high/CXCL9-low polarization. Notably, ERK inhibition did not normalize ROS levels, suggesting that redox dysregulation acts upstream of ERK/MAPK signaling rather than as a downstream consequence.

The consequences for neurons were striking. In co-culture experiments with primary dorsal root ganglion neurons, TXNRD1-overexpressing glia increased neuronal apoptosis, impaired neuronal migration, reduced phosphorylation of the RET receptor and expression of its co-receptor GFRα1, downregulated the synaptic proteins PSD-95 and synaptophysin, and disrupted intracellular calcium homeostasis. Levels of the neurotrophic factors NGF and BDNF also fell in the co-culture system. Crucially, both ERK inhibition and supplementation with exogenous GDNF partially rescued these defects, implicating suppression of the GDNF/GFRα1/RET signaling axis as a key route by which the remodeled glia damage neuronal development and function.

The in vivo experiments provided the most compelling translational signal. The team generated Ednrb conditional knockout mice, which develop Hirschsprung-like phenotypes, and delivered an adeno-associated virus carrying TXNRD1-targeting shRNA by intraperitoneal injection at postnatal day 5. Knockdown of TXNRD1 markedly ameliorated the disease features, reducing abdominal distension, proximal colonic dilatation, distal stenosis, and fecal retention. Treated mice survived significantly longer, with four animals reaching the study endpoint, and fecal water content rose toward control levels, indicating improved intestinal function. At the molecular level, TXNRD1 suppression decreased SPP1, restored CXCL9, increased the neuronal marker TUJ1 and the neural crest marker p75, and upregulated the communication mediators NCAM1 and VCAM1 in the bowel.

The authors are careful to acknowledge the limitations of their work. The proteomic screen relied on only two paired patient samples, the validation cohorts were small, and the intraperitoneal viral delivery was not restricted to enteric glia, complicating cell-type-specific interpretation. Primary dorsal root ganglion neurons served as a surrogate for enteric neurons, and the temporal origins of the SPP1-high glial state remain unresolved, with lineage tracing and spatial transcriptomics needed to clarify its development. Even so, the study establishes a coherent mechanistic framework linking redox dysregulation, ERK/MAPK activation, pathological glial polarization, and neuronal dysfunction in Hirschsprung disease. By showing that enteric glia are active participants in the disease rather than passive bystanders, and that silencing TXNRD1 alleviates HSCR-like phenotypes in mice, the work nominates TXNRD1 and the glial polarization program it controls as promising targets for future therapies aimed at improving outcomes for children with this challenging disorder.

Subject of Research: The role of TXNRD1-driven enteric glial remodeling in neuronal dysfunction in Hirschsprung disease

Article Title: TXNRD1-driven enteric glial remodeling promotes neuronal dysfunction in Hirschsprung disease

Article References: Li, Y., Wang, Y., Jin, Z., Tang, C., Xia, X., Gong, Y., Du, Q., Huang, L., Li, Z., Liao, Y., He, S., Wang, B., Liu, Y., & Zheng, Z. (2026). TXNRD1-driven enteric glial remodeling promotes neuronal dysfunction in Hirschsprung disease. iScience, 29(10), Article 117636. https://doi.org/10.1016/j.isci.2026.117636

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117636

Keywords: Hirschsprung disease, enteric glial cells, TXNRD1, SPP1, CXCL9, ERK/MAPK signaling, redox dysregulation, GDNF, single-cell RNA sequencing, enteric nervous system, Ednrb knockout mice, neurotrophic signaling

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

To capture that transformation in motion, the team did something many earlier studies failed to do. Rather than relying on the human cell lines that dominate the radioresistance literature, lines that are poorly suited to immunocompetent systems and therefore blind researchers to the immune dimension of treatment failure, the investigators generated radioresistant isogenic variants of murine lung adenocarcinoma cell lines. Because these variants are genetically matched to their radiation-sensitive parents, any differences that emerge after repeated irradiation can be attributed to adaptive changes rather than pre-existing genetic background. The team then interrogated these paired models using transcriptomics, three-dimensional co-culture systems, and syngeneic mouse models that retain fully functioning immune systems.

The results paint a picture of tumors that have reorganized themselves at multiple levels simultaneously. The radioresistant clones displayed stable adaptations to DNA damage, including enhanced clonogenic survival after irradiation and constitutive activation of DNA damage response pathways, with the ATR pathway, named for the ataxia telangiectasia and Rad3-related kinase, standing out as particularly active. ATR functions as a master regulator of the cellular response to replication stress and DNA damage, and its persistent activation appears to give cancer cells a cushion against the strand breaks that ionizing radiation inflicts. Critically, the clinical relevance of this observation was not left to speculation: analyzing data from non-small cell lung cancer patients, the researchers found that high ATR levels were significantly associated with immunosuppression, linking a tumor-intrinsic survival mechanism to a hostile immune environment in human disease.

The transcriptomic analysis uncovered a second layer of the adaptation. Radioresistant tumors showed altered DNA repair programs alongside a marked downregulation of type-I interferon signaling and innate immune pathways. This is a crucial detail because radiation therapy, in addition to directly damaging DNA, is supposed to function as an in situ vaccine. When tumor cells die from irradiation, they are expected to undergo immunogenic cell death, releasing damage-associated molecular patterns that activate the cGAS-STING sensing axis and other innate pathways, which in turn recruit and activate cytotoxic T cells capable of finishing off survivors. By dampening these pathways, radioresistant tumors essentially mute the alarm system that radiation depends on to alert the immune system, reducing immunogenic cell death and converting what should be an inflammatory dying process into a quiet one.

The consequences of that muting were visible throughout the tumor microenvironment. In syngeneic mice, radioresistant tumors recruited a distinctly immunosuppressive cellular cast: increased numbers of T regulatory cells, M2-like macrophages, and myeloid-derived suppressor cells, along with reduced recruitment of cytotoxic T lymphocytes. These cell types are well-known accomplices of tumor survival, suppressing antitumor immunity through a variety of inhibitory signals and metabolic depletion. The three-dimensional co-culture experiments added a functional confirmation: when immunosuppressive cells were present alongside cancer cells in the same culture, they protected the malignant cells from radiation-induced cytotoxicity. In other words, the immune cells that radioresistant tumors attract are not bystanders. They are active participants in treatment failure, forming a shield that absorbs or suppresses the immune attack that radiation is meant to provoke.

Perhaps the most striking evidence for the centrality of immune suppression came from a transplant experiment of sorts. When the same radioresistant tumors were grown in immunodeficient mice, animals lacking the adaptive and innate immune machinery that normally shields the tumor, they regained their sensitivity to radiotherapy. Stripped of their immunosuppressive entourage, the tumors that had been so resilient in immunocompetent hosts became vulnerable again. This single observation reframes the entire problem of radioresistance: a substantial portion of it is not written into the cancer cell’s genome at all, but is instead conferred by the ecosystem the tumor has cultivated around itself.

That reframing immediately suggested therapeutic possibilities, and the team pursued them with encouraging results. The first approach involved artificially kick-starting innate immunity using agonists of Toll-like receptors, specifically the combination of poly(I:C), which activates TLR3, and R848, which targets TLR7 and TLR8. When these agonists were delivered to radioresistant tumors, they restored antitumor responses and improved the efficacy of radiotherapy, in part through a mechanism centered on the chemokines CXCL9 and CXCL10. These chemokines act as beacons for immune cells, and their induction re-established the recruitment of cytotoxic immune cells into the tumor, effectively rebuilding the alarm system that the radioresistant clones had dismantled. The findings give the study its title and its conceptual core: CXCL9-driven immune rewiring can restore the conditions under which radiation therapy works as intended.

The second approach attacked the problem from inside the cancer cell. Using ceralasertib, a pharmacological inhibitor of ATR, the researchers sensitized radioresistant tumors to radiation, directly undermining the constitutive DNA damage response activation that had given the cells their survival cushion. Together with the innate immune agonists, this establishes a rationale for combined strategies that hit both halves of the radioresistance program at once: stripping away the tumor’s molecular repair capacity with ATR inhibition while simultaneously reigniting the immune recruitment signals that allow radiation to trigger a lasting antitumor response. Neither approach alone represents a complete solution, but the logic of the combination, grounded in the mechanistic findings rather than empirical trial and error, is what distinguishes this work from earlier attempts to sensitize tumors to radiation.

The broader implications reach beyond lung cancer. The study demonstrates that the way radioresistance research is conducted matters: models lacking intact immune systems systematically obscure the contribution of the microenvironment, and this team’s decision to build isogenic immunocompetent models paid off with insights that would have been invisible in conventional systems. For clinicians treating non-small cell lung cancer, the work suggests that markers such as ATR expression and chemokine signatures could eventually help identify patients at risk of locoregional relapse before it occurs, and that combining radiotherapy with TLR agonists or ATR inhibitors deserves serious clinical evaluation. For now, the findings rest on preclinical evidence, and the usual caveats about translation from mice to humans apply. But they offer something that has been in short supply in this field: a mechanistically coherent account of how tumors learn to shrug off radiation, and a map of the vulnerabilities created along the way. As radiotherapy continues to anchor the treatment of lung cancer worldwide, strategies that restore rather than replace its power may prove to be the key to preventing the relapses that have long defied the beam.

Subject of Research: Mechanisms of radioresistance in non-small cell lung cancer involving ATR signaling and immune microenvironment rewiring

Article Title: CXCL9-driven immune rewiring and tumor-intrinsic ATR signaling inhibition restore radiotherapy response in lung cancer

Article References: Leon, S., Otegui, N., Redrado, M., Marchena-Perea, E. M., Guruceaga, E., Castro, F., Montuenga, L. M., Oliveira, M. J., Aristu, J. J., Serrano, D., & Calvo, A. (2026). CXCL9-driven immune rewiring and tumor-intrinsic ATR signaling inhibition restore radiotherapy response in lung cancer. Journal of Experimental & Clinical Cancer Research. https://doi.org/10.1186/s13046-026-03834-z

Image Credits: AI Generated

DOI: 10.1186/s13046-026-03834-z

Keywords: radiotherapy, non-small cell lung cancer, radioresistance, ATR inhibitor, ceralasertib, CXCL9, tumor microenvironment, immunosuppression, TLR agonists, innate immunity, DNA damage response, cancer immunotherapy

The challenge the researchers set out to address is well known to immunotherapy researchers. Checkpoint-blocking drugs have transformed outcomes for some patients with melanoma, lung cancer and other malignancies, but many solid tumours remain resistant because their surroundings actively suppress immune activity. Tumour-associated macrophages, or TAMs, are abundant white blood cells found inside tumours that, in their tumour-associated state, help the cancer evade destruction. Professor Zhao explained the central problem succinctly: the immune system may be fully capable of attacking a tumour, but the tumour environment can stop those immune cells from doing their job. Her team’s answer was to change that environment from within, delivering treatment precisely to the cells that maintain the immunosuppressive order and persuading them to switch sides.

The delivery vehicle at the heart of the study is the lipid nanoparticle, the same class of fatty droplet that carries mRNA in COVID-19 vaccines. But these particles were given a targeting upgrade: their surface is coated with an antibody that recognises TREM2, a protein expressed at high levels on tumour-associated macrophages. This molecular address label guides the nanoparticles to the very cells responsible for tumour immune suppression. Once inside the macrophages, the particles release two payloads with complementary functions. The first is an mRNA molecule carrying instructions for the macrophage to produce CXCL9, a chemical signalling protein. The second is Resiquimod, a small-molecule compound that pushes macrophages away from their immune-suppressing behaviour and toward a more inflammatory, immune-supportive state.

The two payloads work in concert in a way that illustrates the elegance of the design. CXCL9 acts as a chemical beacon, drawing cytotoxic CD8+ T cells, the immune system’s primary tumour-killing soldiers, into the tumour mass, where they are often excluded or rendered inactive. Meanwhile, Resiquimod shifts the local macrophage population away from suppression, easing the hostile conditions that would normally exhaust or repel those T cells. Rather than depleting the macrophages outright, an approach that can carry inflammatory side effects, the researchers effectively re-educated them, converting a tumour-protective population into an accomplice of the anti-cancer immune response.

In experiments with mouse models, the approach produced measurable shifts in the tumour immune landscape. Treatment reduced the proportion of immune-suppressing macrophages by more than 60 percent, a substantial remodelling of the tumour’s defensive cellular makeup. Levels of CXCL9 within the tumours rose fourfold, confirming that the delivered mRNA was being translated into functional chemical beacon by the targeted cells. The researchers also documented greater numbers and heightened activity of cancer-fighting T cells inside the treated tumours, along with a moderate reduction in tumour growth. While the growth slowdown alone was not dramatic, the immunological changes suggest a tumour environment becoming markedly more permeable and hospitable to immune attack.

The team then tested whether their nanoparticle therapy could amplify the effects of existing immunotherapies. When combined with immune checkpoint-blocking antibodies targeting PD-L1 and CTLA-4, two of the most widely used targets in clinical oncology, the treatment produced further increases in cytotoxic T cells and, notably, the generation of central memory T cells. These long-lived memory cells could give the immune system the ability to recognise and respond to cancer if it returns, a property highly valued in cancer therapy because it hints at durable protection rather than transient tumour shrinkage. Interestingly, however, the combination did not yield additional tumour-growth inhibition in this particular mouse model, a nuance the researchers report candidly and one that will guide future experimental design.

Professor Zhao framed the findings as an important proof of concept that mRNA and nanoparticle technology can reprogramme the immune environment of a tumour. She emphasised that significant work remains before the approach could be considered for patients, but described the results as an encouraging foundation for developing more targeted cancer immunotherapies. That caution reflects the well-known gap between promising mouse studies and clinical reality: nanoparticle manufacturing, dosing, safety profiling and the variability of human tumour environments all present substantial hurdles. Still, the strategy addresses a specific failure mode of current immunotherapy, immune exclusion, and does so with a precision that conventional drugs have struggled to achieve.

The broader significance of the work lies in the expanding repertoire of mRNA medicine. The pandemic demonstrated that lipid nanoparticles can deliver genetic instructions safely and at scale; researchers worldwide are now exploring whether the same platform can carry therapeutic instructions for enzymes, antibodies, cytokines and tumour antigens. This study adds a subtle new use case: not delivering a drug or an antigen, but delivering the blueprint for a signalling molecule that changes the behaviour of the cells receiving it. By targeting TREM2-positive macrophages, the Adelaide team also taps into a growing body of research on macrophage reprogramming, an area increasingly seen as fertile ground for solid tumour therapy where T-cell-focused approaches alone often fall short.

The research was led by Adelaide University scientists in collaboration with SA Pathology and the Royal Adelaide Hospital, and is published under the title ‘Targeting tumor-associated macrophages using mRNA lipid nanoparticles for cytotoxic T lymphocyte–mediated cancer immunotherapy’. For patients whose tumours shut out immunotherapy, the study offers a vision of treatment that does not fight the tumour’s fortress head-on but instead quietly converts its guards, arming them with the instructions to raise a signal flare that guides the immune system’s most lethal cells inside. If subsequent studies can translate those beacon-lit results into durable clinical benefit, mRNA nanoparticles may find a second act in oncology as architects of the tumour microenvironment rather than mere couriers of vaccines.

The biology underlying the study helps explain why macrophages have become such a sought-after target. Macrophages are not inherently tumour-friendly; in healthy tissue they patrol, clear debris and coordinate inflammatory responses. Tumours, however, gradually reshape the macrophages they recruit, coaxing them into a state that suppresses cytotoxic T cells, promotes new blood vessel growth and remodels the fibrous matrix surrounding the cancer. High densities of these suppressive macrophages have been associated in many cancer types with poorer responses to checkpoint inhibitors, which is why strategies that convert rather than eliminate them have attracted growing interest.

The choice of TREM2 as a targeting marker reflects recent advances in understanding macrophage identity within tumours. TREM2 is a receptor expressed on a subset of macrophages that accumulates in tumours and is linked to immunosuppressive function, and blocking or depleting TREM2-positive cells has emerged as an active area of preclinical investigation. Using an antibody against TREM2 as a homing device, rather than as a therapeutic agent itself, is a distinctive feature of the Adelaide approach: the antibody serves as an address label that concentrates the therapeutic payload where it is most needed, potentially limiting off-target effects in healthy tissue.

The signalling components of the formulation also draw on established immunology. Resiquimod is a synthetic agonist of toll-like receptors, pathways that act as alarm bells for the innate immune system and have previously been explored as topical treatments and vaccine adjuvants. Encapsulating it alongside mRNA allows the two stimuli to act within the same cell, pairing a behavioural switch with a genetic instruction. CXCL9, meanwhile, belongs to a family of chemokines long known for their role in recruiting CD8+ T cells to sites of inflammation, and low CXCL9 expression has been linked to immune-excluded tumours in human studies, making it a rational choice for restoring T-cell infiltration.

The appearance of central memory T cells in the combination experiments deserves particular attention. Memory T cells persist long after an initial immune response subsides and can mount faster, stronger reactions upon re-encounter with their target. In cancer, inducing such cells is a goal of therapeutic approaches such as cancer vaccines and intratumoural therapies, because tumour recurrence remains a leading cause of treatment failure even when initial therapy is successful. The fact that this memory phenotype emerged despite the absence of added tumour shrinkage suggests that immunological benefit and immediate tumour control may follow different timelines, a distinction that longer animal studies would need to resolve.

The candid reporting of the combination result also illustrates a broader principle in immunotherapy research: immune activation and tumour regression are not always tightly coupled in early experiments. Factors such as the mouse model used, the timing of treatment relative to tumour establishment, and the dose and schedule of each component can all influence whether enhanced T-cell activity translates into measurable growth control. The authors’ decision to publish these nuances alongside the positive findings provides a transparent baseline for other laboratories seeking to refine the formulation, test it across additional tumour types, and determine which patient populations might ultimately benefit most from a macrophage-reprogramming strategy.

Subject of Research: mRNA lipid nanoparticle reprogramming of tumour-associated macrophages for cancer immunotherapy.

Article Title: New ‘smart’ nanoparticles help the immune system better attack tumours

Article References: New ‘smart’ nanoparticles help the immune system better attack tumours. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: lipid nanoparticles, mRNA, tumour-associated macrophages, cancer immunotherapy, CXCL9, TREM2, CD8+ T cells, immune checkpoint inhibitors, tumour microenvironment, Resiquimod, Science Advances, Adelaide University