Minimal residual disease, or MRD, refers to the small population of cancer cells that persists after curative-intent treatment. These cells are typically undetectable by conventional imaging but can be identified through sensitive molecular methods such as next-generation sequencing, quantitative polymerase chain reaction and circulating tumor DNA analysis. In hematologic malignancies such as acute lymphoblastic leukemia, MRD has long been established as a robust prognostic indicator, and its relevance in solid tumors is increasingly recognized. The clinical data are striking. In one study of 299 patients with stage I to III colorectal cancer, patients whose ctDNA tested positive for any one of six methylation markers one month after surgery were 17.5 times more likely to experience recurrence than those with negative results. In non-small cell lung cancer, ctDNA was detected before treatment in 24 percent of stage I patients, 77 percent of stage II patients and 87 percent of stage III patients, and post-treatment detection of ctDNA was associated with unfavorable clinical outcomes.

Within this reservoir, drug-tolerant persister cells occupy a pivotal and paradoxical position. The concept originates from microbiology, where persister describes bacterial cells that survive antibiotic exposure through non-genetic mechanisms. In 2010, researchers extended this analogy to human cancer when they identified DTP cells in EGFR-mutant lung cancer cells treated with the EGFR inhibitor erlotinib. Despite lethal drug concentrations, a small fraction of cells persisted in a reversible drug-tolerant state characterized by quiescence and epigenetic reprogramming, distinguishing them from genetically resistant mutants. Three core functional assays now define bona fide DTP populations. In a reversibility assay, DTP cells regain full drug sensitivity and proliferative capacity after drug withdrawal for seven to fourteen days, whereas permanently resistant clones maintain stable insensitivity driven by fixed genetic alterations such as EGFR T790M or BRAF V600E secondary mutations. Single-cell sequencing confirms that DTP cells carry no stable driver resistance mutations, relying purely on epigenetic and metabolic adaptive plasticity. Clonal fluctuation testing shows that DTP populations do not produce stable resistant sublines upon continuous passaging without drug pressure.

The review proposes a hierarchical model that clarifies why these cells matter so much. MRD consists of all heterogeneous residual cancer cells that persist after therapy, covering three distinct subpopulations: therapy-sensitive dormant tumor cells, genetically stable permanent resistant clones, and transient drug-tolerant DTP cells. DTP cells exclusively embody the plastic, non-genetic tolerant faction within MRD, surviving initial cytotoxic treatments through reversible epigenetic, metabolic and cell-cycle adaptive mechanisms, and serving as the critical intermediate reservoir en route to stable acquired resistance. This positioning explains why DTP cells, rather than other MRD components, are the primary initiators of tumor recurrence. In acute myeloid leukemia, MRD identified by flow cytometry frequently includes quiescent cells with persister-like characteristics, indicating a functional connection between the two concepts.

The biological toolkit that DTP cells deploy is remarkably diverse. A prominent feature is cell cycle arrest in the G0/G1 phase, rendering them quiescent and less susceptible to therapies targeting proliferating cells. In HER2-positive breast cancer, HER2 inhibitor treatment induces DTP cells with either luminal epithelial-like or mesenchymal-like transcriptional profiles, and tumor cells stochastically and cyclically enter a G0-like pre-DTP state enriched for dormancy and diapause-associated genes. Metabolic reprogramming represents another adaptation. Under MAPK inhibitor treatment, a subset of lung adenocarcinoma cells shifts from glycolysis to mitochondrial respiration, rendering them highly dependent on antioxidant pathways to mitigate elevated reactive oxygen species. DTP cells also bolster anti-apoptotic mechanisms by upregulating proteins such as MCL-1, and during DTP formation, downregulation of MYC relieves transcriptional repression of PINK1, activating mitophagy that eliminates damaged mitochondria and maintains metabolic homeostasis under therapeutic stress.

The tumor microenvironment plays a decisive role in fostering these cells. Hypoxia, a hallmark of solid tumors, promotes DTP survival by stabilizing hypoxia-inducible factor-1 alpha, which upregulates glycolytic enzymes and anti-apoptotic genes. Under hypoxic or nutrient-deprived conditions, melanoma cells acquire a DTP state conferring resistance to BRAF inhibitors and platinum-based chemotherapies, accompanied by extensive chromatin remodeling. Cancer-associated fibroblasts contribute to DTP induction through balanced regulation of TGF-beta and NF-kappaB signaling. During EGFR inhibitor treatment, fibroblasts transition from a myofibroblastic phenotype to a novel state marked by activation of NF-kappaB and IL-6/JAK/STAT3 signaling, further enhancing DTP drug tolerance. Notably, substantial infiltration of Ccr2-positive macrophages occurs when tumor volume reaches its nadir and enters the DTP phase, coinciding with the transition of tumor cells from a reversible drug-tolerant state to an acquired resistant state.

Metabolic and redox adaptations offer some of the most exploitable vulnerabilities. In BRAF V600E-mutant melanoma, DTP cells develop tolerance to BRAF and MEK inhibitors by shifting from glycolysis to peroxisomal fatty acid oxidation, a process transcriptionally regulated by the PPAR alpha-ACOX1 axis. DTP cells also upregulate Lipin1, a fatty acid esterification enzyme, promoting lipid droplet formation to mitigate lipotoxicity. In obesity-associated colorectal cancer, the FABP4/UCP2 axis drives cetuximab resistance when DTP cells are co-cultured with adipocytes. Across multiple treatment contexts, including BRAF, MEK, ALK and androgen receptor inhibitor-treated tumors, DTP cells show reduced glycolysis but increased oxidative phosphorylation and redox regulation. Persister cells that survive treatment tend to adopt a high-mesenchymal state and become reliant on glutathione peroxidase 4 to detoxify lipid peroxides, thereby evading ferroptosis, an iron-dependent form of cell death. In ALK-positive lung cancer, DTP cells surviving ALK inhibitor treatment upregulate GPX4 to evade ROS-induced cytotoxicity.

Epigenetic reprogramming is a frequent feature of DTP biology. In melanoma under hypoxic or nutrient-deprived conditions, DTP cells acquire resistance through chromatin remodeling characterized by the loss and gain of epigenetic marks such as H3K9me3. Upregulation of O-GlcNAc transferase, in synergy with TET1, promotes H3K4me3 deposition at CpG islands via the hexosamine biosynthesis pathway, driving the transition from a dormant DTP state to a proliferative drug-tolerant persister-proliferative state that manifests as acquired resistance. In BRAF-mutant melanoma treated with combined RAF and MEK inhibition, MRD harbors co-existing DTP populations with pronounced transcriptional and spatial heterogeneity, including a key subpopulation displaying a neural crest stem cell-like program driven by RXRG signaling. Cells that survive lethal drug pressure also enhance methylation of histone H3K9 and H3K27 at LINE-1 repetitive elements, suppressing drug-induced activation of repetitive elements and interferon-stimulated genes.

On the therapeutic front, the review highlights several mechanism-based strategies. The Notch signaling pathway has emerged as an actionable target in lung cancer, where Notch1 and its target genes are upregulated in osimertinib-treated DTP cells, and gamma-secretase inhibitors significantly suppress DTP survival. Combining osimertinib with these inhibitors markedly downregulates phosphorylated ERK and enhances DUSP1 expression, and clinical tissue samples corroborate Notch1/HES1 upregulation following EGFR inhibitor treatment in patients. Epigenetic regulation serves as a cornerstone for DTP-targeted therapies. In melanoma, ALDH1A3 cooperates with ACSS2 in the nucleus to convert glucose metabolic flux into acetyl-CoA, promoting histone H3 acetylation and activating genes associated with the neural crest stem cell-like state. Pharmacological inhibition of ALDH1 delays or prevents relapse after BRAF inhibitor therapy. AMPK activators such as aspirin downregulate OGT expression, preventing DTPP formation and delaying acquired resistance. Autophagy offers a third avenue, as blocking autophagy with chloroquine in combination with chemotherapy eliminates DTP survival in colorectal cancer, and inhibiting PINK1 or mitophagy effectively eliminates DTP cells and enhances sensitivity to targeted therapies.

Translating these insights into clinical practice faces real hurdles. The heterogeneity of DTP cells across cancer types and treatment contexts necessitates robust biomarkers to identify and monitor these populations, and the dynamic plasticity of DTP cells complicates therapeutic strategies as they transition between drug-tolerant and drug-sensitive states. Toxicity associated with combination therapies must be carefully managed, and a gap persists between preclinical models and human tumors, where heterogeneity, off-target effects and resistance mechanisms often diminish therapeutic efficacy. Consensus on MRD thresholds and survival correlates remains unresolved, and aggressive DTP-targeting regimens may increase toxicity, raising concerns about quality of life versus survival gains. Yet emerging technologies offer reason for optimism. Single-cell omics, spatial transcriptomics, digital PCR, live-cell imaging and multiplexed ion beam imaging are pushing detection limits toward ultra-rare DTP cells, while artificial intelligence and machine learning are being deployed to identify DTP signatures, predict relapse and optimize therapies. Adaptive therapy approaches that modulate treatment intensity based on tumor response are being tested to prevent the emergence of DTP populations, and liquid biopsy technologies are being developed to detect circulating DTP cells and DTP-derived ctDNA for real-time MRD monitoring. If these threads converge, the review concludes, targeting DTP cells within MRD holds transformative potential to dismantle the persister niche, prevent relapse and reshape long-term cancer outcomes.

Subject of Research: Cancer drug-tolerant persister cells and minimal residual disease

Article Title: Erasing the invisible: targeting persister cells to cure minimal residual disease

Article References: Wang, Y., Shi, Z., Xu, F., Zhu, Y., Liu, C., Wang, B., Ji, N., Wang, Z., Wang, S., Zhou, X., & Ren, Y. (2026). Erasing the invisible: targeting persister cells to cure minimal residual disease. Holistic Integrative Oncology, 5(1), 81. https://doi.org/10.1007/s44178-026-00307-3

Image Credits: AI Generated

DOI: 10.1007/s44178-026-00307-3

Keywords: minimal residual disease, drug-tolerant persister cells, cancer relapse, epigenetic reprogramming, metabolic reprogramming, tumor microenvironment, ferroptosis, autophagy, liquid biopsy, single-cell sequencing, targeted therapy, drug resistance

The study, published in iScience, followed four AML patients through two critical stages: before transplantation, when their disease had failed to reach remission, and after transplantation, when their leukemia had relapsed with donor-derived immune cells now patrolling the bone marrow. The researchers profiled 95,426 individual cells from these patients alongside four healthy donors, capturing the complete transcriptomic landscape of the bone marrow microenvironment at both time points. What they found was a bone marrow ecosystem fundamentally reshaped by malignancy, with T cell frequencies dramatically lower than in healthy donors and leukemia-like cells dominating the cellular terrain.

Using an integrated computational strategy combining reference-based annotation, copy number variation analysis, and differentiation-state prediction, the team classified the leukemia-like cells into eight distinct subtypes, including hematopoietic stem cell-like blasts, progenitor-like blasts, granulocyte-monocyte progenitor-like blasts, and actively cycling blasts. Pseudotime analysis revealed a developmental trajectory flowing from stem cell-like blasts toward more differentiated subtypes, confirming that the leukemia maintained a cellular hierarchy mirroring normal hematopoiesis. Critically, the stem cell-like blasts carried the highest leukemic stem cell 17-gene score, a validated measure of stemness and treatment resistance, and their proportion increased after transplantation.

These stem cell-like blasts were not merely more primitive — they were more immunosuppressive. Compared with their non-stem counterparts, they upregulated the expression of macrophage migration inhibitory factor (MIF), transforming growth factor beta 1 (TGFB1), galectin-9 (LGALS9), and HLA-E, while simultaneously downregulating HLA class II molecules. This molecular cocktail is a known inducer of T cell and natural killer cell dysfunction. The researchers also noted that antigen-presentation changes after transplantation were molecule-specific rather than uniformly suppressed, with HLA-DR upregulated while HLA-DP and HLA-DQ were downregulated, suggesting a nuanced immune escape strategy rather than blanket silencing.

Turning to the T cell compartment, the team identified three clusters characterized by high expression of inhibitory receptors such as PD-1, LAG-3, and TIGIT, which they defined as IR-high T cells. Clonal expansion analysis showed that highly expanded T cell subsets concentrated mainly in the CD8-positive effector and IR-high clusters in AML patients, while limited expansion was focused in naive and memory populations in healthy donors. A pairwise transition index revealed that CD8-positive IR-high T cells had the highest transition likelihood from effector cells, confirming that these exhausted populations emerged directly from the cytotoxic effector pool. Compared with effector cells, IR-high T cells expressed lower levels of the cytotoxic molecules granzyme B and perforin but higher levels of inhibitory receptors and the transcription factors TOX and NR4A2.

The most striking discovery came when the researchers compared the transcriptional programs of self-derived exhausted T cells before transplantation with donor-derived exhausted T cells after relapse. Self-derived exhausted T cells were characterized by enrichment of NFATC1 and NFATC3 transcriptional programs together with increased TOX and EOMES expression, whereas donor-derived exhausted T cells exhibited enrichment of NFATC2-associated signatures accompanied by elevated NR4A2 expression. Gene regulatory network analysis showed that in self-derived exhausted cells, NFATC1 primarily regulated TOX, EOMES, and multiple inhibitory receptors, while in donor-derived cells, NFATC2 primarily regulated NR4A2 and various inhibitory receptors. This fundamental divergence in regulatory architecture means that the two forms of T cell exhaustion, though phenotypically similar, run on entirely different molecular engines.

Further subclustering revealed four major exhausted T cell populations: a progenitor population, an effector population, a terminally exhausted population, and an actively dividing population. Slingshot trajectory inference showed transitions flowing from the progenitor exhausted subtype toward the other three states. Each exhausted subtype carried higher TOX and EOMES expression before transplantation but higher NR4A2 after relapse. In the progenitor exhausted cells, NFATC3 mainly regulated HAVCR2, EOMES, SELL, and TCF7; in the effector exhausted cells, NFATC2 regulated effector-related genes such as TBX21, ZEB2, and GZMB; and in the terminally exhausted cells, NFATC1 regulated TOX, EOMES, CTLA4, HAVCR2, and PDCD1. These findings suggest that the progenitor exhausted population can differentiate into other exhausted subtypes through activation of different transcription factors whose downstream effects differ dramatically.

The bone marrow microenvironment itself was profoundly immunosuppressive at both disease stages. Monocytes expressed high levels of PD-L1 and interleukin-10 but low levels of the costimulatory ligand ICOSLG compared with healthy donors. Dendritic cells similarly upregulated PD-L1, PD-L2, interleukin-10, and HLA-E while downregulating HLA class II molecules. Regulatory T cells expressed elevated levels of PD-1, CTLA-4, LAG-3, TIM-3, TIGIT, TGFB1, granzyme B, perforin, and the key transcription factors FOXP3, STAT5, and NFKB1, with most of these molecules more prominently expressed after transplantation. Hematopoietic stem cell-like blasts upregulated CD274, CD48, FGL1, interleukin-10, CD47, TGFB1, LGALS9, and HLA class I molecules while downregulating the costimulatory ligand TNFSF4. Together, these cellular components constituted an immunosuppressive bone marrow niche that persisted across both pre- and post-transplantation settings.

Cell-cell communication analysis using CellChat and scMLnet revealed that leukemia-like cells interacted most strongly with CD8-positive effector and exhausted T cells among all T cell subtypes. In the pre-transplant setting, ligand-receptor pairs including HLA-E:CD8A/B, MIF:(CD74+CD44), LGALS9:PTPRC, and TGFB1:TGFBR1/2 were enriched in these interactions, and computational modeling predicted that these signaling pathways further regulated the activation of NFAT family transcription factors in the exhausted T cells. In the post-transplant setting, similar interactions were detected except that the TGFB1:TGFBR1/2 pathway was absent. Leukemia cells also communicated indirectly with exhausted T cells through intermediary populations — regulatory T cells, monocytes, and dendritic cells — modulating their immunosuppressive transcriptional programs and downstream target genes such as LCK, CD86, TNFSF13, and GRN, which in turn participated in signaling interactions associated with NFAT-related transcriptional activity in the exhausted T cells.

Among all leukemia cell subtypes, hematopoietic stem cell-like blasts had the most comprehensive ligand-receptor pairs interacting with exhausted T cells, regulatory T cells, monocytes, and dendritic cells, and their communication probabilities were relatively high at both disease stages. Survival analysis using The Cancer Genome Atlas AML dataset revealed that a gene set consisting of only three ligands — MIF, LGALS9, and TGFB1 — had the best prognostic value for overall survival, outperforming TGFB1 alone and gene sets including HLA class I molecules. The researchers conclude that targeting TOX in self-derived exhausted T cells, NR4A2 in donor-derived exhausted T cells, and the signaling molecules MIF, galectin-9, and TGF-β1 could represent a strategy for reversing T cell exhaustion and improving outcomes in AML patients, though they acknowledge that the findings are primarily computational and will require experimental validation in larger cohorts.

Subject of Research: Single-cell dissection of T cell exhaustion niches in acute myeloid leukemia

Article Title: Single-cell dissection of self-derived versus donor-derived T cell exhaustion niches in patients with acute myeloid leukemia

Article References: Yu, C.-Z., Chen, Y.-L., Zhang, X.-H., Xu, L.-P., Wang, Y., Chen, H., Chen, Y.-H., Wang, F.-R., Sun, Y.-Q., Zhang, Y.-Y., Lv, M., Mo, X.-D., Cheng, Y.-F., Liu, K.-Y., Zhao, X.-S., Sun, X.-Q., Zhao, X.-Y., Huang, X.-J., & Chang, Y.-J. (2026). Single-cell dissection of self-derived versus donor-derived T cell exhaustion niches in patients with acute myeloid leukemia. iScience, 29(11), Article 117918. https://doi.org/10.1016/j.isci.2026.117918

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117918

Keywords: acute myeloid leukemia, T cell exhaustion, single-cell sequencing, stem cell transplantation, immunotherapy, tumor microenvironment, Single-cell, dissection, self-derived, versus, donor-derived, cell

The method, called scDAEC, was developed by Yue Zhu, Chen Chen, Yue Liu, Yanlei Kang and Zhong Li at Huzhou Normal University, together with Hancan Zhu of Shaoxing University. It is an unsupervised clustering framework built on a dual-autoencoder architecture, meaning it trains two different autoencoders in parallel and fuses what each one learns. According to the authors, existing deep clustering models tend to emphasize either gene-expression reconstruction or cell–cell structural learning, which limits their ability to jointly handle expression sparsity while preserving the topology of how cells relate to one another. scDAEC was designed to do both at once.

The first autoencoder is a denoising autoencoder based on an adaptive zero-inflated negative binomial model, abbreviated ZINB. This statistical choice is deliberate. scRNA-seq count data contain many zeros, some of which reflect true biological silence and others of which are artifacts of incomplete sequencing. The zero-inflated negative binomial distribution explicitly models this uncertainty, combining a count component that captures overdispersed gene expression with a zero-inflation component that accounts for excess zeros. By training the denoising autoencoder with an adaptive ZINB loss, the network learns to reconstruct clean gene-expression profiles from corrupted inputs, producing denoised embeddings that retain biologically meaningful variation while suppressing technical noise.

The second autoencoder is a graph autoencoder, or GAE, which approaches the data from a completely different angle. Rather than treating each cell as an isolated vector of gene counts, the graph autoencoder learns from the structure of relationships among cells. A cell–cell graph encodes which cells are similar to which others, and the GAE learns topology-aware representations that preserve this cellular neighborhood structure. This matters because cells of the same type occupy connected regions in the high-dimensional expression space, and capturing that connectivity helps the model keep genuinely related cells together even when individual expression measurements are unreliable.

The innovation that ties the two branches together is a module the authors call dynamic combined multi-head attention, or DCMHA. Multi-head attention, the mechanism popularized by transformer architectures, allows a model to weigh different parts of its input with different importance when making decisions. In scDAEC, the DCMHA module integrates the denoised gene-expression embeddings from the ZINB-based branch with the graph-based embeddings from the GAE branch, dynamically combining the two views for clustering optimization. The dynamic aspect means the fusion weights are not fixed; the model adjusts how much it trusts expression information versus structural information as training proceeds, allowing the two representations to reinforce each other rather than compete.

Once the fused representation is learned, scDAEC optimizes clusters end to end, jointly refining the embeddings and the cluster assignments in a single training loop. This joint optimization is what distinguishes deep clustering from the classical pipeline of dimensionality reduction followed by a separate clustering algorithm such as k-means. In the classical approach, errors introduced during dimensionality reduction propagate unchecked into the clustering step. In scDAEC, the clustering objective feeds back into the representation learning, so the embeddings are shaped specifically to make cell groups separable.

The authors evaluated scDAEC on 13 real scRNA-seq datasets and compared it with representative competing methods. Performance was measured with two standard clustering metrics. The Adjusted Rand Index, or ARI, quantifies how well a clustering agrees with a reference grouping while correcting for chance agreement, and the Normalized Mutual Information, or NMI, measures how much information one clustering shares with another. Based on dataset-level mean results, scDAEC achieved overall mean ARI and NMI values of 0.864 and 0.859 respectively. Compared with the strongest competing methods, the framework improved average ARI by 10.79 percent and average NMI by 10.96 percent, margins the authors describe as reflecting improved clustering accuracy and more stable performance across datasets.

Stability is a point worth emphasizing. Deep clustering methods can be sensitive to initialization and hyperparameter choices, and a method that excels on one dataset but fails on another offers limited practical value. The consistent gains across 13 heterogeneous datasets suggest that the dual-view design, which grounds the model in both expression statistics and cellular topology, generalizes better than single-view alternatives. The adaptive ZINB loss also means the denoising branch does not rely on a fixed noise model; it adapts to the statistical character of each dataset it encounters.

The implications reach beyond benchmark scores. Accurate unsupervised clustering is the foundation for downstream single-cell analyses, including identifying rare cell populations, tracing developmental trajectories, and comparing healthy and diseased tissue. When clustering misassigns cells, rare populations can vanish into larger groups and disease signatures can be diluted. A framework that recovers cell groups more faithfully could sharpen every analysis built on top of it, from atlas construction to drug-response studies.

The work was supported by the National Natural Science Foundation of China under grants U24A20249 and 12171434, and by the Science and Technology Plan Project of Huzhou City under grant 2022GZ51. The study was published open access, and the source code is publicly available on GitHub, allowing other researchers to apply scDAEC to their own data or extend the architecture. As single-cell sequencing continues to scale to millions of cells, methods that can jointly denoise expression data and learn cellular structure are likely to become essential tools for turning raw transcriptomic noise into biological insight.

Subject of Research: Deep learning method scDAEC improves unsupervised clustering of single-cell RNA sequencing data

Article Title: scDAEC: deep clustering of single-cell data based on dual autoencoder

Article References: Zhu, Y., Chen, C., Liu, Y., Kang, Y., Zhu, H., & Li, Z. (2026). scDAEC: deep clustering of single-cell data based on dual autoencoder. BMC Genomics. https://doi.org/10.1186/s12864-026-13261-1

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13261-1

Keywords: single-cell sequencing, deep learning, clustering, autoencoder, genomics, scRNA-seq, scDAEC, deep, single-cell, data, based, dual

The study under discussion approached this question by combining several complementary analytical strategies, each of which addresses a different limitation of conventional observational research. Observational studies of psoriasis and depression are notoriously vulnerable to confounding: shared risk factors such as obesity, smoking, alcohol use, sleep disturbance, and socioeconomic adversity influence both conditions independently, making it difficult to determine whether the association reflects causation or shared predisposition. Mendelian randomization offers a way around this obstacle by exploiting the random allocation of genetic variants at conception. Because alleles are inherited independently of environmental exposures and lifestyle factors, genetic variants that influence gene expression can serve as proxies—so-called instrumental variables—for testing whether altered expression of a particular gene causally affects disease risk. When the researchers used expression quantitative trait loci, or eQTLs, from the eQTLGen consortium as instruments, they identified 340 gene-expression signals associated with psoriasis and 307 associated with depression, of which only nine genes showed evidence of causal involvement in both directions. This small intersection is itself informative: it suggests that the molecular overlap between the two disorders is concentrated in a limited set of genes rather than being distributed diffusely across the genome, narrowing the search space for therapeutic targets considerably.

The nine intersecting genes were then subjected to least absolute shrinkage and selection operator regression, a machine learning technique designed to identify the most predictive features from a larger candidate pool while guarding against overfitting. LASSO works by applying a penalty that shrinks the coefficients of less informative variables toward zero, effectively forcing the model to rely on only the most robust predictors. From this procedure, four genes emerged as central hubs: MAP3K20, WARS2, TBXAS1, and ABHD15. Each of these genes carries a distinct biological profile that helps explain how a skin-limited inflammatory disease and a neuropsychiatric condition could share common molecular ground.

MAP3K20, better known in the literature as ASK1, encodes a mitogen-activated protein kinase kinase kinase that sits at a critical junction in cellular stress signaling. It activates the p38 and JNK pathways in response to oxidative stress, endoplasmic reticulum stress, and inflammatory cytokines such as tumor necrosis factor-alpha. In the context of psoriasis, MAP3K20 signaling in keratinocytes contributes to the proliferative and pro-inflammatory responses that characterize plaque formation. In neural tissue, the same pathway has been implicated in stress-induced neuronal dysfunction and in the neuroinflammatory responses associated with depressive-like behavior in animal models. The dual role of MAP3K20 in both epithelial inflammation and stress-responsive neural signaling makes it a plausible molecular bridge between the two conditions.

WARS2 encodes tryptophanyl-tRNA synthetase, the mitochondrial enzyme responsible for charging transfer RNA molecules with tryptophan during protein synthesis. This gene occupies an intriguing position at the intersection of several pathways relevant to the comorbidity. Tryptophan is not only a building block for proteins but also the precursor for serotonin synthesis, and disruptions in tryptophan availability or metabolism have long been linked to mood disorders. Moreover, tryptophan is diverted along the kynurenine pathway during inflammation, generating neuroactive metabolites that can be either neuroprotective or neurotoxic depending on the enzymatic context. Chronic inflammatory conditions like psoriasis shift tryptophan metabolism toward kynurenine, potentially reducing serotonin availability in the brain while simultaneously producing metabolites that influence glutamate signaling and neuronal survival. A gene that governs tryptophan handling at the cellular level therefore connects inflammatory metabolism directly to monoamine neurotransmission.

TBXAS1 encodes thromboxane A synthase 1, the enzyme that converts prostaglandin H2 into thromboxane A2, a potent vasoconstrictor and platelet aggregator. Thromboxane signaling is well established in cardiovascular biology, and elevated thromboxane production is a recognized feature of the systemic inflammatory state that accompanies psoriasis and contributes to the elevated cardiovascular risk observed in these patients. Less widely appreciated is the role of thromboxane signaling in the central nervous system, where it influences microglial activation, neurovascular function, and inflammatory signaling within the brain. The presence of TBXAS1 among the shared genes suggests that prostanoid metabolism may represent a conduit through which peripheral inflammatory signals generated in psoriatic skin and vasculature are translated into neuroinflammatory changes relevant to depression.

ABHD15, the least characterized of the four, encodes a protein that functions as a regulator of lipolysis and has been implicated in metabolic signaling pathways. Its identification in this analysis points toward the metabolic dimension of the comorbidity. Psoriasis is strongly associated with metabolic syndrome, insulin resistance, and dyslipidemia, and depression likewise shows consistent associations with altered lipid metabolism and insulin signaling. The enrichment results reported in the study highlighted cholesterol metabolism as one of the pathways in which the four hub genes participate, reinforcing the idea that lipid biology constitutes a shared mechanistic layer between the two disorders.

Pathway enrichment analysis of these four genes revealed convergence on the interleukin-17 signaling cascade, nuclear factor kappa B activation, and FoxO transcription factor regulation. The IL-17 pathway is the signature immune axis of psoriasis, and biologic drugs targeting IL-17 or its receptor have transformed the treatment of moderate to severe disease. NF-kappa B serves as a master transcriptional regulator of inflammatory gene expression across virtually all cell types, integrating signals from cytokine receptors, pattern recognition receptors, and stress pathways into coordinated transcriptional responses. FoxO transcription factors, meanwhile, govern cellular responses to oxidative stress, regulate autophagy, and influence neuronal survival. The convergence of all four hub genes on these pathways suggests that the comorbidity operates through a self-reinforcing circuit in which IL-17-driven inflammation activates NF-kappa B in both peripheral immune cells and central nervous system cells, while FoxO-mediated stress responses determine whether cells succumb to or survive the inflammatory assault.

Perhaps the most novel finding to emerge from the single-cell sequencing component of the analysis was the connection to ferroptosis, a form of regulated cell death discovered relatively recently and defined by its dependence on iron-catalyzed lipid peroxidation. Unlike apoptosis or necroptosis, ferroptosis is driven by oxidative damage to lipid membranes and is suppressed by the glutathione peroxidase 4 enzyme and by system xc-mediated cystine import. Ferroptosis has been implicated in a growing list of inflammatory and neurodegenerative conditions, and its involvement in both psoriatic keratinocytes and depressive neurocircuitry represents an emerging area of investigation. In the skin, ferroptotic death of keratinocytes can release damage-associated molecular patterns that amplify inflammatory signaling, potentially perpetuating the psoriatic cycle. In the brain, ferroptosis of neurons and oligodendrocytes has been linked to the hippocampal atrophy and synaptic dysfunction observed in chronic stress models and in patients with major depression. The correlation between the four hub genes and ferroptosis markers identified in this study suggests that iron-dependent lipid peroxidation may serve as a downstream execution mechanism through which chronic inflammation translates into cellular damage in both organ systems.

The single-cell transcriptomic data added a layer of cellular resolution that bulk RNA sequencing cannot provide. By examining individual cells from psoriatic skin lesions and from brain-related tissue in depression cohorts, the researchers were able to determine which cell populations express the four hub genes and how that expression changes in disease states. This cell-type-specific localization matters because the same gene can perform entirely different functions depending on whether it is expressed in keratinocytes, endothelial cells, infiltrating neutrophils, microglia, or neurons. The finding that the hub genes show enrichment in specific immune cell populations and correlate with patterns of immune infiltration provides a spatial framework for understanding how molecular signals propagate across tissue boundaries.

Turning from mechanism to therapeutics, the researchers queried the Comparative Toxicogenomics Database, a publicly available resource that catalogs interactions between chemicals, genes, and diseases, to identify compounds that might modulate the activity of the four hub genes. Among the candidates that emerged, folic acid stood out for its ability to target three of the four genes. Folic acid, the synthetic form of folate or vitamin B9, is an essential cofactor in one-carbon metabolism, supporting the synthesis of purines, thymidine, and the methyl groups required for epigenetic regulation and neurotransmitter synthesis. Folate deficiency has long been associated with depression, and low folate levels predict poorer response to selective serotonin reuptake inhibitors, which is why L-methylfolate is sometimes used as an adjunctive treatment in psychiatry. In the context of psoriasis, folate metabolism intersects with the rapid keratinocyte proliferation that characterizes the disease, and methotrexate—the most widely used conventional systemic agent for psoriasis—works in part by inhibiting folate-dependent enzymes, which is why folate supplementation is routinely co-prescribed to mitigate toxicity.

The identification of folic acid as a potential modulator of the shared molecular network therefore carries both mechanistic and clinical significance. It suggests that the same nutrient pathway that supports neurotransmitter synthesis in the brain also influences the proliferative and inflammatory programs active in psoriatic skin, providing a biological rationale for why nutritional status might modulate both conditions simultaneously. It also raises the possibility that folate-related interventions, whether through dietary optimization, supplementation, or pharmacological modulation of one-carbon metabolism, could be explored as adjunctive therapies aimed at the comorbidity rather than at either condition in isolation.

It is worth emphasizing what this study does and does not establish. The findings rest on computational integration of publicly available datasets rather than on new laboratory experiments or clinical trials, and the causal inferences drawn from Mendelian randomization depend on assumptions—that the genetic instruments are valid proxies, that they influence the outcomes only through the exposure of interest, and that confounding through population stratification has been adequately controlled. The eQTL data were derived from peripheral blood, which may not fully capture gene expression in skin or brain tissue, and the GWAS summary statistics came from populations of European ancestry, potentially limiting generalizability to other populations. The sample sizes for the psoriasis transcriptomic datasets were modest, and single-cell datasets, while informative, reflect specific tissue contexts that may not represent the full cellular diversity of either disease. These limitations do not invalidate the findings but rather frame them as hypothesis-generating results that require experimental validation in cellular models, animal studies, and ultimately prospective clinical cohorts.

Nevertheless, the study contributes a structured framework for thinking about psoriasis and depression as a biologically integrated comorbidity. By triangulating genetic causality, transcriptomic signatures, machine learning feature selection, and single-cell resolution, it narrows a vast molecular landscape down to four genes and a handful of interconnected pathways. The convergence on IL-17 signaling, NF-kappa B activation, FoxO regulation, cholesterol metabolism, synaptic vesicle cycling, and ferroptosis paints a coherent picture in which chronic inflammatory signaling, altered lipid handling, and oxidative stress-induced cell death form a mutually reinforcing network spanning skin, vasculature, and brain. For clinicians, the findings reinforce the importance of routine depression screening in psoriasis care and suggest that monitoring inflammatory markers, metabolic parameters, and nutritional status may provide actionable windows into comorbidity risk. For researchers, the four hub genes and the pathways they anchor offer concrete starting points for experimental validation and for the development of interventions that target the shared biology underlying both conditions, moving the field closer to the goal of precision therapy for patients living with the burden of this dual diagnosis.

Subject of Research: Molecular mechanisms underlying the comorbidity of psoriasis and depression

Article Title: Molecular mechanisms underlying psoriasis and depression: an integrated analysis using mendelian randomization, transcriptomics, and single-cell sequencing

Article References: De, B., Bao, W., Hea, N., & Fang, J. (2026). Molecular mechanisms underlying psoriasis and depression: an integrated analysis using mendelian randomization, transcriptomics, and single-cell sequencing. Frontiers in Molecular Medicine, 6, Article 1770665. https://doi.org/10.3389/fmmed.2026.1770665

Image Credits: AI Generated

DOI: 10.3389/fmmed.2026.1770665

Keywords: psoriasis, depression, Mendelian randomization, single-cell sequencing, transcriptomics, ferroptosis, folic acid, MAP3K20, WARS2, TBXAS1, ABHD15, immune infiltration

The review’s most striking finding is not biological but bibliographic. When the authors systematically searched PubMed for studies directly addressing astrocyte–microglia communication, they found a dramatic imbalance: only 28 records addressed crosstalk during development, 94 during adult homeostasis, and a staggering 1,554 in pathological contexts. This skew, the authors argue, reflects both technical and conceptual biases. Studying two cell types talking to each other in a healthy brain is extraordinarily difficult because most experimental interventions—cell depletion, inflammatory stimulation, genetic manipulation—themselves perturb the system and trigger the very reactive states researchers are trying to avoid. Pathology, by contrast, makes the conversation loud enough to hear.

The story begins before birth. Microglia are the first glial cells to arrive in the developing brain, colonizing it from hematopoietic stem cells in the yolk sac around the fourth gestational week in humans and before embryonic day 8 in mice. Astrocytes, generated later from radial glial neural progenitors after the peak of neurogenesis, enter a brain already populated by these early immune sentinels. That timing matters. In vitro work with neural progenitor cells has shown that microglia-derived factors such as interleukin 6 (IL-6) and leukemia inhibitory factor (LIF) promote astrocyte differentiation, and microglial phagocytosis of newly born cells may help regulate final neuron and astrocyte numbers. In other words, the immune cells that arrive first may help shape the identity and abundance of the glial cells that follow.

As the brain matures, the direction of influence reverses. During spinal cord development, astrocytes secrete IL-33, which acts on microglia to promote synapse engulfment—the process by which excess connections are pruned to sculpt functional neural circuits. In the postnatal cortex, microglia instruct astrocytes through Wnt signaling to disengage synapses before microglia consume them. And in the developing thalamus, astrocytes signal to microglia through interactions between the protein Hevin and Toll-like receptor 4, boosting microglial phagolysosomal activity and synapse removal. These findings reveal a genuinely bidirectional dialogue: microglia first shepherd astrocyte differentiation, then mature astrocytes reciprocate by directing microglial pruning. Disruption of this choreography during critical developmental windows has been implicated in schizophrenia and autism spectrum disorders.

In the healthy adult brain, both cell types are indispensable guardians of homeostasis. Astrocytes metabolically couple to neurons through the lactate shuttle, wrap around synapses to regulate glutamate via transporters such as EAAT2 and EAAT1, buffer potassium to link neuronal firing with blood flow, form gap-junction-coupled syncytia through connexins that integrate activity across broad brain regions, and maintain the blood–brain barrier and water balance through aquaporin 4 channels. Microglia, meanwhile, continuously survey the parenchyma with motile processes, regulate synaptic function, and maintain myelin integrity after developmental ensheathment. Yet the review emphasizes that how these two cells regulate each other under physiological conditions remains poorly understood—many of the inferred interactions come from studies that used proinflammatory stimuli, which likely mimic disease rather than health.

What evidence does exist for homeostatic crosstalk is intriguing. Microglia release trophic and immunomodulatory mediators including transforming growth factor β (TGF-β), IL-10, and insulin-like growth factor 1, which can restrain astrocytic inflammatory responses. Astrocytes and microglia coordinate phagocytosis of neuronal corpses through MERTK-dependent pathways, and astrocytes can even clear microglial debris via complement component 4b–mediated phagocytosis—a process important for microglial turnover in vivo. Pharmacological depletion of microglia causes astrocyte syncytium uncoupling through downregulation of connexins 30 and 43 in the mouse hippocampus, suggesting that microglia actively maintain the astrocyte network itself. These observations hint that the healthy brain’s stability depends on a continuous glial dialogue that science has barely begun to map.

When injury or disease strikes, that dialogue turns dark. Microglia, the first responders to central nervous system damage, release tumor necrosis factor α, IL-1α, and complement component C1q, which synergistically push astrocytes into inflammatory reactive states characterized by impaired homeostatic functions and secretion of neurotoxic molecules. Reactive astrocytes then amplify the cycle by releasing IL-6, GM-CSF, chemokines such as CCL2 and CXCL10, complement proteins, ATP, and damage-associated molecular patterns, sustaining leukocyte recruitment and chronic neuroinflammation. The review’s authors, who have studied this axis experimentally, showed that NF-κB activation in astrocytes is enhanced in cultures with high microglial density and is microglia-dependent in vivo models of inflammation. STAT3 activation in reactive astrocytes at injury borders also depends on microglia, linking the immune cells directly to scar formation dynamics.

Perhaps the most conceptually provocative idea in the review is that pathology reactivates developmental programs. The complement cascade—C1q, C3, and their receptors—orchestrates synaptic pruning during development but drives aberrant synapse elimination in neurodegeneration. In stroke models, astrocytes at injury borders appear to express C3, potentially recruiting C3aR-expressing microglia; in Alzheimer’s disease models, the C3–C3aR pathway mediates both amyloid-beta clearance and neurotoxic inflammation. Similarly, IL-33–ST2 signaling, TGF-β signaling, CSF-1/IL-34–CSF1R, CX3CL1–CX3CR1, purinergic ATP signaling through P2Y12 and P2X7, and the TREM2–ApoE axis all appear redeployed in disease, where their sustained activation promotes maladaptive crosstalk, perpetuating neuroinflammation and progressive neuronal damage. Pathological crosstalk, the authors conclude, is not mediated by a single pathway but by the coordinated reactivation of multiple developmental communication programs.

Studying this partnership has always been methodologically hard, and the review surveys both traditional and emerging tools. Classic approaches—histological correlation, cell-type-specific genetic manipulation with GFAP or ALDH1L1 promoters, pharmacological microglial depletion via CSF1R inhibitors, astrocyte ablation with L-alpha-aminoadipate, and static co-cultures—establish correlations but struggle with causality and physiological relevance. The new generation is more surgical. SPEAC-seq, or systematic perturbation of encapsulated associated cells followed by sequencing, places single NF-κB-reporter astrocytes and CRISPR-edited microglia into oil droplets, lets them communicate, then sorts activated pairs and sequences the guide RNAs to discover which microglial genes drive astrocyte inflammation. RABID-seq uses a barcoded, astrocyte-targeted rabies virus to map the in vivo astrocyte interaction network at single-cell resolution, and has already revealed immunosuppressive astrocyte subtypes interacting with glioblastoma. Human-relevant platforms are advancing too: 3D bioprinted neural tissues, vascularized brain organoids containing human microglia implanted into mouse brains, and microfluidic co-culture devices that model neuroinflammatory interactions using human induced pluripotent stem cell derivatives.

The bigger picture emerging from this synthesis is that astrocytes and microglia are not independent responders but interconnected cellular partners whose molecular dialogue shapes the central nervous system from embryogenesis to aging. Some glial responses traditionally interpreted as cell-autonomous may instead emerge from coordinated interactions between the two cell types, and inflammatory signaling between them can even promote epigenetic remodeling of astrocytes, altering their metabolic support functions. The authors argue that future research must move beyond studying each cell in isolation and instead define how their bidirectional interactions are regionally organized, temporally regulated, and dynamically modified across physiological and pathological states. Only then, they suggest, can therapies be designed to modulate neuroinflammation without compromising the homeostatic functions of glial networks—a balance that may prove decisive for treating Alzheimer’s disease, Parkinson’s disease, multiple sclerosis, epilepsy, and traumatic brain injury.

Subject of Research: Astrocyte–microglia crosstalk in central nervous system development, homeostasis, injury, and disease

Article Title: “Stand by me”: astrocyte–microglia crosstalk in central nervous system development, homeostasis, injury, and disease

Article References: Ramos, A. J., & Villarreal, A. (2026). “Stand by me”: astrocyte–microglia crosstalk in central nervous system development, homeostasis, injury, and disease. Frontiers in Network Physiology, 6, Article 1895622. https://doi.org/10.3389/fnetp.2026.1895622

Image Credits: AI Generated

DOI: 10.3389/fnetp.2026.1895622

Keywords: astrocytes, microglia, glial crosstalk, neuroinflammation, synaptic pruning, complement cascade, neurodegeneration, blood-brain barrier, reactive gliosis, brain development, NF-kB signaling, single-cell sequencing

Genomic imprinting is orchestrated by epigenetic modifications, most notably DNA methylation, which are established during gametogenesis by the de novo methyltransferase DNMT3A and its cofactor DNMT3L, then maintained throughout life by DNMT1. Imprinted genes typically cluster within genomic loci governed by imprinting control regions, or ICRs, which display roughly fifty percent DNA methylation, reflecting parent-of-origin dependent regulation. Zinc finger proteins such as ZFP57 and ZNF445 tightly regulate the establishment of these methylation marks, and during early embryogenesis, when the epigenome is largely reset, ICRs are protected from demethylation by ZFP57 and the PGC7/STELLA complex. Histone modifications, including H3K9me3 and H3K27me3 on the silenced allele and H3K4me3 and H3K27ac on the active allele, reinforce this imbalanced expression pattern. Non-coding RNAs, such as the lncRNA AIRN, which silences the IGF2 receptor gene in mice, and insulator binding by CTCF, which coordinates expression across larger imprinted regions like the IGF2/H19 locus on chromosome 11p15.5, add further layers of regulation.

The most widely accepted evolutionary explanation for imprinting is the parental conflict hypothesis. Paternally expressed genes promote offspring growth during early development, while maternally expressed genes counterbalance this growth stimulation to preserve maternal health. This framework makes imprinting disruption particularly relevant to cancer, as many imprinted genes regulate cell growth, differentiation, and apoptosis. When imprinting is lost, either through reactivation of a normally silenced allele or silencing of a normally active allele, growth-promoting genes can become biallelically overexpressed while growth-limiting genes become completely silenced, directly fuelling tumorigenesis.

The clinical consequences of imprinting dysregulation extend well beyond cancer. Congenital syndromes such as Beckwith-Wiedemann, Prader-Willi, Angelman, Silver-Russell, Temple, and Kagami-Ogata all arise from disrupted imprinting at specific loci, often through DNA methylation defects, uniparental disomy, point mutations, or chromosomal aberrations. Beckwith-Wiedemann syndrome, characterised by excessive foetal growth and elevated childhood cancer risk, stems from dysregulated imprinting at the IGF2/H19 locus. Beyond congenital disorders, altered imprinting has been linked to cardiovascular conditions, including hypertension and atherosclerosis, where the DLK1-MEG3 domain is extensively hypomethylated in plaques, as well as neuropsychiatric disorders, with parent-of-origin effects implicated in autism spectrum disorder and schizophrenia.

In cancer, LOI at the IGF2/H19 locus is the most extensively documented imprinting alteration. Re-expression of the normally maternally silenced IGF2 allele and silencing of the maternally expressed H19 non-coding RNA are considered growth-promoting events. LOI at this locus has been detected in nearly thirty percent of colorectal cancer patients’ normal colonic mucosa, and its presence in blood greatly increases the odds of both colorectal adenoma and carcinoma. IGF2/H19 LOI has also been described in breast, liver, lung, and Wilms tumour, where it recapitulates foetal-program proliferation patterns and leads to poorer treatment response and adverse clinical outcomes, making it a putative predictive and prognostic biomarker. Other imprinted loci, including DLK1 in gliomas and neuroblastomas, PEG3 in gliomas, CDKN1C in several cancer types, HM13 in breast, kidney, and lung tumours, and MEST in breast, ovarian, and lung carcinomas, also undergo LOI or aberrant methylation in cancer.

Studying imprinting in humans presents formidable technical challenges. In model organisms like mice, imprinting can be characterised cost-efficiently by mating parents from different inbred lines and profiling the largely heterozygous progeny. However, imprinting is often species-specific, making human-specific experiments necessary for clinical studies. The retinoblastoma gene RB1, for instance, is imprinted in humans but not in mice, and attempts to generate an imprinted mouse model were unsuccessful. Similarly, the synteny of WT1 and IGF2 on chromosome 11p in humans, which allows a single complex genetic event to produce both aberrations in Wilms tumour, does not exist in mice, making murine models less representative of the human disease.

The gold standard for human imprinting research combines genomics and transcriptomics in a parent-offspring trio setting. By genotyping the offspring, heterozygous loci are identified, upon which RNA sequencing of the tissue under study reveals whether only a single allele is expressed. Comparison with parental genomics data then determines from which parent the expressed allele is inherited. However, this strategy is limited by the need for heterozygosity to discriminate between alleles, requiring sufficiently large numbers of trios and expensive designs combining both genomics and transcriptomics data. Case-control studies that first screen for candidate imprinted loci in controls, followed by analysis in cases, abolish the requirement for trio data but often lead to false positives, particularly due to allele-specific expression effects independent of parental origin, such as random monoallelic expression of human leukocyte antigen genes.

Epigenetics-based strategies, focusing on DNA methylation rather than expression, offer an alternative approach. Whole-genome bisulfite sequencing provides base-resolution DNA methylation profiling by converting unmethylated cytosines to uracil, allowing evaluation of whether methylation at ICRs deviates from the anticipated fifty percent. However, bisulfite treatment induces DNA degradation, and the requirement for large, high-quality, high-purity tumour samples poses significant challenges in cancer research. Enzymatic methylation sequencing (EM-seq) has emerged as a gentler alternative, while reduced representation bisulfite sequencing and targeted capture strategies reduce cost. DNA methylation arrays, particularly the Illumina Infinium MethylationEPIC platform, offer a cost-effective genome-scale option, though they lack allele-specific read-out, making it unclear whether methylation alterations reflect dysregulated imprinting or altered cell-type composition. Methylation-specific multiplex ligation-dependent probe amplification (MS-MLPA), which combines copy number and methylation assessment, remains the most widely used methodology for diagnosing imprinting syndromes.

Three major methodological challenges complicate the interpretation of imprinting analyses in cancer. First, most studies rely on bulk tumour samples that may contain substantial fractions of non-tumour cells, such as tumour-associated fibroblasts or infiltrating immune cells, reaching up to fifty percent in renal cancer. If admixing cell types feature different imprinting patterns, incorrect LOI calls may result. Second, the need for heterozygosity to discriminate paternal from maternal alleles means that the power to detect LOI varies across the genome as a function of genotype frequency, a major bottleneck for clinical assays. Third, transcript-specific imprinting, observed for loci such as MEST and IGF2, can lead to inconsistent LOI calls across a gene’s SNPs, while widespread copy number alterations in cancer can cause both false-positive LOI calls and genuine LOI events through deletion or gain of an originally silenced allele.

Emerging technologies promise to address these challenges. Single-cell RNA sequencing resolves tumour heterogeneity by profiling individual tumour cells, though many platforms miss the SNPs required to separate alleles, and single-cell data sparsity limits genome-wide character. Pseudobulk strategies, combining data from tumour cells or subpopulations, mitigate this resolution problem. Spatial transcriptomics and targeted methods like Quantitative Chromogenic Imprinted Gene In Situ Hybridization (QCIGISH) enable allele-specific visualisation even in homozygous backgrounds, addressing problems with tumour purity, copy number alterations, and heterozygosity. QCIGISH has been validated in cancer patient cohorts, achieving near-perfect sensitivity and specificity for early-stage lung cancer detection and enabling cervical cancer risk stratification. Long-read sequencing technologies offer the possibility to resolve transcript-level imprinting patterns and provide direct DNA methylation read-out during genome sequencing, avoiding bisulfite-related problems. The authors argue that integrating single-cell, spatial, and long-read strategies into multi-omics approaches will largely resolve remaining issues with non-cancer cells in bulk tissues and the need for heterozygosity and transcript-level resolution, ultimately guiding the development of optimised clinical tests and even epigenome-editing therapies for cancer.

Subject of Research: Genomic imprinting and its dysregulation in cancer

Article Title: Challenges and emerging strategies for genome-wide evaluation of loss of imprinting in cancer

Article References: Amin, M. T., Coussement, L., & De Meyer, T. (2026). Challenges and emerging strategies for genome-wide evaluation of loss of imprinting in cancer. British Journal of Biomedical Science, 83, Article 16709. https://doi.org/10.3389/bjbs.2026.16709

Image Credits: AI Generated

DOI: 10.3389/bjbs.2026.16709

Keywords: genomic imprinting, loss of imprinting, cancer, epigenetics, DNA methylation, biomarkers, single-cell sequencing, long-read sequencing, Challenges, emerging, strategies, genome-wide

The clinical stakes are considerable. Head and neck squamous cell carcinoma is broadly divided into two etiological camps: tumors driven by human papillomavirus, which tend to respond relatively well to treatment, and HPV-unrelated tumors, which are overwhelmingly linked to tobacco exposure and carry markedly worse outcomes. Understanding why smoking-associated tumors behave so aggressively has been a central unsolved question in the field. The Emory team, led by corresponding author Yong Teng, with Fan Yang, Fanghui Chen and Xinyu Cui as co-first authors, approached the problem by mining transcriptomic data across multiple patient cohorts, including The Cancer Genome Atlas, to identify genes whose activity tracks with tobacco exposure in HPV-unrelated HNSCC. Among the smoking-associated targets that emerged, MUC5B stood out as an immunomodulatory mucin whose elevated expression correlated with tumor progression.

To understand what MUC5B was actually doing inside tumors, the researchers turned to single-cell RNA sequencing, a technique that reads out gene activity in individual cells rather than in bulk tissue. This cell-type-resolved view revealed that high MUC5B levels were associated with a profound metabolic shift: tumor cells appeared to pivot toward enhanced fatty acid utilization, ramping up the machinery for fatty acid oxidation as a fuel source. Metabolic reprogramming of this kind is a hallmark of aggressive cancers, because rewired energy production can support rapid proliferation, survival under stress, and resistance to therapy. The team confirmed these metabolic signatures using gene set enrichment and gene set variation analyses, standard bioinformatic methods that test whether defined biological pathways are systematically activated or suppressed across samples.

The metabolic story was only half the picture. When the researchers characterized the tumor immune microenvironment, using both conventional analyses and a platform called GigaTIME to generate virtual multiplex immunofluorescence images, they found that MUC5B-high tumors harbored a distinctly immunosuppressive landscape. Macrophages, the immune cells that normally engulf threats and coordinate inflammatory responses, had been polarized toward an M2-like state, a wound-healing, tumor-tolerant phenotype that actively supports cancer growth. At the same time, infiltrating CD8-positive T cells, the cytotoxic soldiers responsible for killing tumor cells, were diminished. In effect, elevated MUC5B coincided with a double blow to anti-tumor immunity: the cancer’s metabolic engine was refitted for fatty acids, and the immune system’s assault force was simultaneously depleted and redirected.

Having established the correlation, the team went hunting for the upstream regulator that connects tobacco smoke to MUC5B in the first place. The answer lay with GATA4, a transcription factor that the researchers found binds directly to the MUC5B promoter, the DNA region that controls the gene’s activity, and represses its transcription. Under normal conditions, GATA4 acts as a molecular brake, keeping MUC5B production in check. But when the researchers exposed HNSCC cells to 4-nitroquinoline 1-oxide, or 4NQO, a carcinogen widely used to model tobacco-induced oral carcinogenesis, this regulatory axis broke down. 4NQO exposure reduced GATA4 expression and, critically, reduced GATA4’s occupancy of the MUC5B promoter, as demonstrated by chromatin immunoprecipitation experiments. With the brake released, MUC5B expression surged.

To test whether MUC5B was merely a bystander or an active orchestrator of these changes, the researchers silenced the gene in laboratory models and observed what happened. MUC5B silencing reversed the metabolic reprogramming, dialing back the fatty acid oxidation program, and restored a more favorable immune profile in cell-based assays. The decisive test came in animals: in immunocompetent mouse models of orthotopic tongue tumors exposed to 4NQO, silencing MUC5B reduced tumor growth. The use of immunocompetent mice is an important methodological detail, because it means the immune system was intact and functional during the experiments, allowing the team to observe genuine immune-mediated effects rather than artifacts of immunodeficient models. Together, the in vitro and in vivo data support a causal chain running from tobacco carcinogen, through GATA4 loss, through MUC5B induction, to metabolic rewiring and immune suppression.

The study’s architecture reflects a broader trend in modern cancer research, in which no single technique is sufficient to untangle a tumor’s complexity. The Emory group layered population-level transcriptomics on top of single-cell resolution, then added spatial context through AI-assisted immune profiling, and finally validated every computational observation with molecular assays, metabolic measurements, and functional immune cell experiments. The GigaTIME-based virtual multiplex immunofluorescence deserves particular note: by computationally reconstructing the spatial arrangement of immune and tumor cells within tissue, it allowed the team to quantify how MUC5B-high tumors organize their microenvironment, information that conventional single-marker staining cannot capture. This integration of artificial intelligence with traditional pathology is increasingly seen as essential for dissecting the tumor microenvironment, the dense ecosystem of cancer cells, immune cells, and supporting tissue that determines whether immunotherapy succeeds or fails.

Therapeutically, the implications are twofold. First, MUC5B itself becomes a candidate drug target, and the demonstration that its silencing reverses metabolic and immune alterations suggests that blocking this mucin, or restoring the GATA4-mediated repression that tobacco carcinogens erode, could resensitize smoking-associated tumors to immune attack. This matters because HPV-unrelated HNSCC patients currently face limited options and poor prognoses, and immunotherapies that work in other cancers often deliver disappointing results in this population, plausibly because the immunosuppressive microenvironment described in this study stands in the way. Second, the 4NQO/GATA4-MUC5B axis provides a mechanistic explanation for a long-standing clinical observation, namely that tobacco-exposed tumors are both more aggressive and more immune-evasive, and it does so through a defined molecular pathway rather than a vague appeal to mutational burden.

Caveats remain, as they always do in translational research. The work relied heavily on the 4NQO carcinogen as a proxy for tobacco exposure, a well-established but imperfect model of the far more complex chemical mixture in cigarette smoke. The mouse experiments, while conducted in immunocompetent animals and in an orthotopic setting that recapitulates the anatomical location of human disease, still require follow-up before MUC5B-targeting strategies reach the clinic. The authors also note that the published version was shared early as accepted peer-reviewed research, subject to final editorial production. Nonetheless, the study was supported by multiple NIH awards including R01DE033433 and R01DE033691, and the authors declare no competing interests. What the work delivers, with unusual clarity, is a complete mechanistic arc: a carcinogen, a transcriptional brake, a mucin, a metabolic switch, and an immune collapse, each link verified experimentally. For a disease in which tobacco’s fingerprints have been visible for decades without a matching molecular explanation, that arc is a genuine advance, and it reframes a humble mucus protein as a central node in one of oncology’s most stubborn problems.

Subject of Research: Tobacco-driven MUC5B regulation of metabolic reprogramming and immunosuppression in head and neck squamous cell carcinoma

Article Title: MUC5B orchestrates tobacco-driven metabolic rewiring and immunosuppression in head and neck squamous cell carcinoma

Article References: Yang, F., Chen, F., Cui, X., Yang, J., Steward-Tharp, S. M., Saba, N. F., & Teng, Y. (2026). MUC5B orchestrates tobacco-driven metabolic rewiring and immunosuppression in head and neck squamous cell carcinoma. Journal of Experimental & Clinical Cancer Research. https://doi.org/10.1186/s13046-026-03851-y

Image Credits: AI Generated

DOI: 10.1186/s13046-026-03851-y

Keywords: MUC5B, head and neck squamous cell carcinoma, tobacco, GATA4, 4NQO, metabolic reprogramming, fatty acid oxidation, immunosuppression, M2 macrophages, CD8 T cells, tumor microenvironment, single-cell sequencing

The neurons in question produce acetylcholine, a chemical messenger that is critical for a range of functions including sleep, cognition and movement. Their gradual loss is linked to some of Parkinson’s most disabling and treatment-resistant symptoms: problems with gait and balance, disturbed sleep, cognitive decline and, in some people, psychosis. While much of the public and scientific attention on Parkinson’s has centred on dopamine-producing neurons in the midbrain, which underlie the condition’s characteristic tremor, rigidity and slowness of movement, this study shifts focus to a brainstem population whose decline shapes the broader, often harder-to-treat burden of the illness.

To interrogate these cells, the team used advanced single-cell mitochondrial DNA sequencing combined with computational analysis, examining individual neurons from brain tissue generously donated by people who had lived with Parkinson’s disease, alongside tissue from healthy controls. Rather than averaging signals across thousands of cells, the single-cell approach allowed the researchers to read out the mitochondrial genome of each neuron separately, revealing the extent and pattern of damage cell by cell. Dr Ilse Pienaar of the University of Birmingham, a senior author of the study, noted that this is the first time single-cell mitochondrial genetic sequencing has been applied to these specific brainstem neurons in people with Parkinson’s disease, and that studying individual neurons has allowed the team to see, in unprecedented detail, the mitochondrial changes associated with their vulnerability.

What the sequencing revealed was striking. The affected neurons carried extensive large-scale deletions in their mitochondrial DNA, and these deletions were particularly concentrated within a stretch of the mitochondrial genome known as the major arc. This region is densely populated with genes that are essential for producing energy in cells, so removing chunks of it strikes directly at the machinery that keeps a neuron powered. Mitochondria are often described as the powerhouses of the cell, and damage to their DNA can interfere with their ability to generate the energy that neurons need, potentially leaving these cells less able to cope with the relentless electrical and metabolic demands placed on them across a human lifetime.

The biological logic of this vulnerability is compelling. Neurons are among the most energy-hungry cells in the body, and brainstem cholinergic neurons must sustain long projections and continuous signalling to regulate sleep-wake cycles, attention and motor control. If their mitochondrial genomes accumulate deletions in the very region that encodes core components of the energy-production system, the resulting power deficit could compound over time, gradually eroding the cell’s capacity to survive. The new findings therefore provide a mechanistic account of why these acetylcholine-producing neurons may be particularly susceptible to degeneration in Parkinson’s disease, complementing decades of work on dopaminergic cell loss.

Yet the study is not simply a catalogue of damage. Despite the extensive mitochondrial DNA lesions, the researchers found evidence that the neurons may be mounting an active protective response. The affected cells showed increased levels of PINK1, a gene central to mitochondrial quality control. PINK1 helps cells identify mitochondria that have become damaged and target them for removal, a process that clears dysfunctional power units from the network and helps maintain a population of healthy mitochondria capable of sustaining energy production. In other words, even as the mitochondrial genomes of these neurons crumbled, the cells appeared to be ramping up the molecular machinery that weeds out the worst-affected organelles.

Perhaps the most intriguing observation concerns individual variation. Higher levels of PINK1 were particularly evident in neurons from people who had survived longer after their Parkinson’s diagnosis. This pattern raises the possibility that a stronger mitochondrial quality-control response helps some neurons withstand damage for longer, effectively buying time and preserving function in the face of accumulating genetic injury. If confirmed, the finding would suggest that the trajectory of Parkinson’s disease is shaped not only by the forces that damage neurons but also by the variable vigour of the brain’s own defences against that damage.

Dr Joanna Elson of Newcastle University, a senior author of the study, described the results as revealing an extraordinary resilience within these vulnerable neurons. Although the team observed extensive mitochondrial DNA damage, the cells appear to mount a protective response through increased PINK1 expression, likely as an attempt to maintain healthy mitochondria and preserve energy production. She added that understanding why this response appears to be more effective in some individuals than others could open new avenues for therapies aimed at supporting the brain’s own protective mechanisms. Dr Pienaar similarly emphasised that the findings point towards biological pathways that could potentially be harnessed therapeutically, and that strengthening the mechanisms which maintain mitochondrial health could ultimately help slow disease progression and improve quality of life for people living with Parkinson’s.

The therapeutic implications are considerable. PINK1 has long been known to the Parkinson’s field because mutations in the gene cause early-onset, inherited forms of the disease, placing it at the heart of the cellular pathway that surveys mitochondrial health and dispatches damaged organelles. The new data suggest that this same pathway is active in the vulnerable brainstem neurons of people with sporadic Parkinson’s, and that its activity correlates with survival after diagnosis. That correlation makes mitochondrial quality control an attractive target for drug development: if researchers can identify why the response is more effective in some patients and find ways to amplify it in others, treatments might be developed that enhance the brain’s natural protective mechanisms rather than simply replacing lost neurotransmitters.

The researchers hope the findings will catalyse further studies of mitochondrial quality-control pathways as potential therapeutic targets in Parkinson’s disease, and that understanding how some neurons withstand damage could ultimately inform treatments that bolster these defences across the brain. For a condition that affects millions worldwide and for which current therapies address symptoms rather than halt progression, the demonstration that single-cell sequencing can expose both the vulnerabilities and the resilience of individual human neurons marks a methodological and conceptual advance. It reframes Parkinson’s disease, in part, as a struggle between mitochondrial damage and cellular self-repair, one in which the outcome may depend on how well each neuron can keep its power supply intact.

Subject of Research: Single-cell mitochondrial DNA sequencing of vulnerable brainstem cholinergic neurons in Parkinson's disease

Article Title: First study to sequence mitochondrial DNA from individual vulnerable brainstem neurons in Parkinson’s disease

Article References: First study to sequence mitochondrial DNA from individual vulnerable brainstem neurons in Parkinson’s disease. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: Parkinson's disease, mitochondrial DNA, single-cell sequencing, brainstem neurons, PINK1, cholinergic neurons, mitochondrial quality control, neurodegeneration, major arc deletions, Brain journal, University of Birmingham, Newcastle University

The technical case for intratumoral microbes rests on several converging methodologies. Bacterial 16S ribosomal RNA gene sequencing has detected microbial signatures in resected tumors across cancer types, including pancreatic ductal adenocarcinoma, non-small cell lung cancer, triple-negative breast cancer, hepatocellular carcinoma, and squamous cell carcinomas of the head, neck, and oral cavity. More decisive have been spatially resolved techniques: fluorescence in situ hybridization, which places bacterial signals inside tumor cells and immune cells under the microscope, and single-cell RNA sequencing analyses that computationally recover microbial reads from individual cells. Studies of pancreatic cancer, for example, reported intracellular bacteria within cancer-associated fibroblasts and immune populations, while work on breast tumors suggested that bacteria can even exploit host cell trafficking pathways, such as the cytoskeletal machinery that normally drives cell migration, to spread within the tumor.

What makes the new review distinctive is its insistence that microbial identity alone explains little. The same bacterial taxon can be associated with tumor suppression in one context and tumor promotion in another. The authors argue that biological function is determined by a combination of microbial traits—such as whether an organism is immunogenic or immunosuppressive, metabolically active or dormant—its spatial localization within the tumor, the characteristics of the host tumor, and the therapeutic pressure applied to the system. This context-dependence reframes the field: instead of asking simply which microbes are present, researchers must ask where they sit, what they are doing, and what the tumor and the treatment are doing to them at that moment.

That framing leads to the review’s central organizing concept, a tripartite framework the authors call microbiota–immunity–therapy. Under this model, the tumor is treated as an ecosystem with three interacting layers. Microbes alter the immune landscape; immunity determines whether cancer cells are eliminated or tolerated; and therapy—particularly immunotherapy—feeds back on both. The framework is designed to replace the older, linear view in which microbes were either passenger bystanders or simple biomarkers. It positions intratumoral microbiota as a dynamic variable that can be measured, monitored, and potentially manipulated to improve treatment outcomes.

The immune reprogramming arm of the framework is where the biology becomes most striking. Intratumoral microorganisms carry pathogen-associated molecular patterns that engage pattern-recognition receptors on innate immune cells, setting off cascades that can either amplify antitumor inflammation or entrench immunosuppression. Tumor-associated macrophages, myeloid-derived suppressor cells, tumor-associated neutrophils, dendritic cells, and natural killer cells all shift their behavior in response to local microbial signals. Bacterial metabolites, including short-chain fatty acids, can diffuse through the microenvironment and reprogram cell states. Outer membrane vesicles shed by bacteria deliver cargo—DNA, proteins, lipids—directly into host cells, adding another channel of communication. The net effect is a microenvironment whose immune tone is partly written by its microbial residents.

Adaptive immunity is equally entangled. Microbial antigens presented within the tumor can recruit and activate T cells, and in some settings bacterial signals enhance dendritic cell maturation and T cell priming against tumor antigens. In others, microbial stimulation drives exhaustion phenotypes in T cells, expands regulatory populations, or skews the balance toward immune escape. The review emphasizes that these effects are not fixed properties of a microbe but outcomes of its ecological position: a bacterium sequestered inside cancer cells may behave very differently from the same species sitting in the extracellular matrix adjacent to a high endothelial venule or a tertiary lymphoid structure.

Nowhere is this context-dependence more consequential than in immunotherapy. Immune checkpoint inhibitors targeting PD-1 and PD-L1 have transformed outcomes in melanoma, lung cancer, and many other malignancies, yet most patients do not respond durably. A growing body of work links intratumoral microbial composition to checkpoint inhibitor efficacy, and the review catalogues examples in which specific organisms or microbial signatures correlate with response, resistance, or immune-related toxicity. Crucially, the authors highlight that microbial effects on immunotherapy can appear paradoxical: the same microorganism may sensitize a tumor to PD-1 blockade under one set of ecological and therapeutic conditions and undermine it under another. Tumor bacterial burden—the overall load of bacteria within the tumor—emerges as a variable that may stratify patients in ways that conventional genomic biomarkers do not capture.

These insights are already generating microbiota-targeted therapeutic strategies. Fecal microbiota transplantation, which reprograms the gut ecosystem that indirectly shapes systemic immunity, has shown early promise in rescuing responses to checkpoint inhibitors in small clinical studies. More directly, researchers are exploring intratumoral approaches: engineering bacteria to deliver therapeutic payloads, deploying bacterial outer membrane vesicles as adjuvant-like immunostimulants, combining antibiotics with immunotherapy to deplete immunosuppressive microbes, and designing probiotic or oncolytic microbial constructs that reprogram the tumor microenvironment from within. Each strategy inherits the same caveat that runs through the entire review—manipulating an ecosystem whose rules are only partially understood carries the risk of unintended immune consequences.

The authors are candid about the methodological fragility of the field. Intratumoral microbial ecosystems are low-biomass environments, which makes them exquisitely vulnerable to contamination from reagents, surgical equipment, and sequencing pipelines. Distinguishing genuine tumor-resident microbes from environmental noise requires rigorous negative controls, careful sample handling, orthogonal validation by imaging and culture, and standardized reporting. Spatial heterogeneity adds another layer of difficulty: a single biopsy may miss microbial niches concentrated in hypoxic or necrotic regions. The review argues that emerging technologies—spatial transcriptomics, single-cell multi-omics, and improved culture-independent methods—will be essential to move the field from cataloguing microbes to understanding their functions in situ.

Even so, the trajectory is clear. The discovery that tumors host their own microbiomes has added a previously unrecognized layer of complexity to cancer biology, and the West China Hospital team’s synthesis argues that this layer belongs at the center of immunotherapy research rather than at its margins. If the microbiota–immunity–therapy framework holds up under experimental scrutiny, the next generation of cancer immunotherapy trials may routinely profile intratumoral microbes alongside tumor mutational burden and PD-L1 expression, and microbial manipulation may become a legitimate lever for converting non-responders into responders. For a field that long treated tumors as sterile, the message is blunt: the ecosystem inside a cancer is crowded, active, and consequential—and learning to manage its microbial residents could reshape how the immune system is aimed at cancer.

Subject of Research: Intratumoral microbiota and their role in tumor immune reprogramming and immunotherapy response

Article Title: Intratumoral microbiota: spatiotemporal dynamics, microenvironmental immune reprogramming, and immunotherapeutic implications

Article References: Xiang, Z., Cheng, W., Wu, Y., Niu, T., Zhao, A., & Kang, K. (2026). Intratumoral microbiota: spatiotemporal dynamics, microenvironmental immune reprogramming, and immunotherapeutic implications. Molecular Cancer. https://doi.org/10.1186/s12943-026-02811-6

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02811-6

Keywords: intratumoral microbiota, tumor microenvironment, cancer immunotherapy, immune checkpoint inhibitors, PD-1/PD-L1, tumor-associated macrophages, bacterial metabolites, fecal microbiota transplantation, tumor bacterial burden, microbial ecology, single-cell sequencing, Molecular Cancer

The research focused on non-small cell lung cancer, or NSCLC, the umbrella term for the most common forms of the disease, including lung squamous cell carcinoma and lung adenocarcinoma. Rather than studying established tumours, the team turned its attention to premalignant lesions — abnormal patches of tissue in the central airways that have not yet become cancerous. These lesions are routinely detected in high-risk smokers through a technique called autofluorescence bronchoscopy, in which special light reveals suspicious areas of the bronchial lining. The clinical stakes are considerable: while low-grade lesions rarely progress, between 50 and 80 per cent of high-grade lesions invade to become frank cancer within two to three years, making these patients an ideal population in which to test strategies for interception.

Using single-cell RNA sequencing on bronchial biopsies from patients under surveillance at University College London Hospitals, the researchers profiled thousands of individual T cells — the immune system’s principal assassins and regulators. Among the ten distinct T cell populations they identified, one stood out with unusual clarity: a subset of effector regulatory T cells, known as Treg cells, marked by high expression of a transcription factor called BATF. These BATF-positive Treg cells expressed the highest levels of FOXP3, CD25, CTLA-4 and a suite of highly suppressive molecules including OX40, GITR, 4-1BB and CCR8 — a molecular profile previously associated with aggressive tumours and poor responses to immunotherapy. Crucially, these cells were selectively and significantly enriched in high-grade premalignant lesions compared with normal tissue, while no other CD4 or CD8 T cell population showed a comparable increase. The immune system, in other words, was being actively disarmed before cancer had even invaded.

The team then asked where these suppressive cells came from. By sequencing T cell receptors — the unique molecular barcodes that identify individual T cell clones — in both lesions and matched blood samples, they uncovered a remarkable clonal connection. Nearly 18 per cent of all BATF-positive Treg cell clonotypes found in the lesions were also present in the peripheral blood, and the dominant blood population sharing these clonotypes was a circulating effector Treg cell population expressing CD39. Statistical modelling confirmed that this sharing was far greater than chance, with an odds ratio of nearly 39. Pseudotime analysis, which reconstructs cellular differentiation trajectories, suggested a blood-to-lung journey: circulating effector Treg cells appeared to migrate into the airways, where antigen-presenting cells displaying MHC class II molecules and engaging CD86–CTLA-4 and TNFSF9–TNFRSF9 interactions drove their activation, expansion and tissue adaptation.

This clonal link opened the door to something clinically transformative: a non-invasive blood test. In a cohort of 129 blood samples from 47 patients in the bronchoscopy surveillance programme, the researchers found that patients carrying high-grade lesions had significantly elevated frequencies of circulating CD39-positive effector Treg cells compared with those with only low-grade disease — a result that held even after adjusting for age, smoking history and previous cancer diagnosis. The signal extended to patients with established, screen-detected early-stage lung cancer as well. In the ASCENT study of patients diagnosed through CT screening, 90 per cent of whom had stage I disease, circulating effector Treg cells were significantly increased compared with healthy individuals. An independent validation cohort comparing 77 early-stage lung adenocarcinoma cases with 67 CT-screened individuals showing no signs of disease reproduced the finding with striking significance.

Perhaps most compelling was the longitudinal evidence. For a small group of patients who developed lung squamous carcinoma during CT surveillance, blood samples had been collected at three- to twelve-month intervals from enrolment to diagnosis. Deep sequencing of T cell receptors in these serial samples revealed a marked and progressive increase in effector Treg cell clonotypes from baseline to the moment of cancer diagnosis, while naive-like and central-memory Treg clones remained stable or changed only modestly. The rising immune signature was not a static correlate but a dynamic process tracking tumour development in real time. When the researchers combined the blood effector Treg frequency with the intratumoural BATF-positive Treg signature in 89 patients with clinical follow-up, they found that patients with high values on both measures faced a nearly fourfold increase in the hazard of recurrence or death — a risk that remained significant after adjustment for age, stage, sex, smoking status and technical batch effects.

To establish causality rather than mere correlation, the team turned to a mouse model in which the carcinogen NTCU induces lung squamous tumorigenesis that faithfully recapitulates the human disease, progressing from low-grade to high-grade dysplasia and finally to invasive carcinoma. The same cross-tissue effector Treg axis emerged: BATF-positive Treg cells accumulated selectively around the bronchial tree, shared clonotypes with circulating effector Treg cells, and were seeded partly from the blood. Upstream, the tumour-draining mediastinal lymph nodes showed elevated PD-L1 expression on migratory type 1 conventional dendritic cells — a change absent from non-draining lymph nodes — alongside an expansion of Treg cells. Blocking lymph node egress with the drug FTY720 sharply reduced both lung Treg cells and circulating effector Treg cells, confirming that the lymph node acts as a supply depot feeding the suppressive network. The conservation of this axis across species underscored its biological centrality to early lung carcinogenesis.

The therapeutic implications were tested directly. Because pathway analysis revealed heightened PI3K signalling in BATF-positive Treg cells, the researchers administered PI-3065, an inhibitor of the PI3K-delta enzyme that is known to preferentially impair Treg cell maintenance, starting at week 15 when preinvasive disease first appears. The results were dramatic: tumour incidence fell from 83 per cent to 42 per cent, and the tumours that did emerge were significantly smaller, with the largest 30 per cent eliminated entirely. A control inhibitor targeting the unrelated PI3K-alpha enzyme had no such effect. Spatial transcriptomics using the Xenium platform revealed the mechanism: carcinogenesis had spawned peribronchial immature tertiary lymphoid structures — dense immune aggregates rich in Treg cells, CXCL13-associated macrophages, germinal-centre-like B cells and specialized vasculature. PI3K-delta inhibition dismantled these structures, selectively depleting Treg cells while increasing conventional type 1 dendritic cells, effectively converting suppressive hubs into potential sites of immune surveillance. A complementary experiment with an anti-CD25 depleting antibody confirmed that Treg cells actively restrain CD8 T cell differentiation during high-grade dysplasia.

The authors propose a ‘theranostic’ framework in which circulating effector Treg cells serve multiple roles: as a non-invasive diagnostic for high-grade premalignant lesions, as an adjunct for risk-stratifying CT-detected nodules, as a pharmacodynamic readout during treatment, and as a tool for selecting patients for Treg-directed immunotherapy. Because immune interventions in people without invasive cancer demand an exceptionally high safety margin, the researchers suggest that next-generation PI3K-delta inhibitors with better toxicity profiles, intermittent dosing schedules, or delivery directly to the airway by inhalation or robotic bronchoscopy could minimize risk. They also acknowledge that specificity remains a challenge — chronic inflammation, infection and injury can all perturb Treg cell homeostasis — and propose pairing the immune signature with routine infection markers such as C-reactive protein, or using longitudinal monitoring to detect a rising rather than a merely elevated signal.

If validated at scale in prospective interception trials, this work could mark a turning point in lung cancer prevention. As CT screening programmes expand worldwide and identify ever more premalignant lesions, clinicians currently have limited tools to decide which patients need aggressive follow-up and which can be safely monitored. A simple blood test that reads the immune system’s own account of tumour development — and a drug strategy that dismantles the suppressive network before invasion begins — would give medicine something it has rarely possessed in oncology: the ability to stop cancer not by treating it, but by intercepting it on the way to becoming dangerous.

Subject of Research: A preinvasive regulatory T cell axis that tracks and can be targeted to intercept early lung squamous cancer development

Article Title: A preinvasive regulatory T cell axis for lung cancer interception

Article References: Gamble, S., Whiteman, Z. E., Peinador-Marín, C., Lebrusant-Fernandez, M., Shurr, A. Y. L., Enica, A., Sahwangarrom, T., Ung, S. K. A., Rogers, A., Fessas, P., Khaw, C. R., Kalinke, L., Ahlmann-Eltze, C., Alhendi, A. S. N., Otter, K., Hu, X., Thakkar, K., Gration, B., Newsham, I., … Reading, J. L. (2026). A preinvasive regulatory T cell axis for lung cancer interception. Nature. https://doi.org/10.1038/s41586-026-11066-6

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11066-6

Keywords: lung cancer, regulatory T cells, BATF, cancer interception, premalignant lesions, PI3K-delta inhibitor, immunotherapy, CT screening, tertiary lymphoid structures, single-cell sequencing, tumour immunology, early detection